Split (1)

train · 5.09k rows

pmcid stringlengths 6 6	title stringlengths 9 374	abstract stringlengths 2 4.62k ⌀	fulltext stringlengths 167 106k	file_path stringlengths 64 64
521734	Correction: Nicotine's Defensive Function in Nature	July correction	In PLoS Biology , volume 2, issue 8 Nicotine's Defensive Function in Nature Anke Steppuhn, Klaus Gase, Bernd Krock, Rayko Halitschke, Ian T. Baldwin DOI: 10.1371/journal.pbio.0020217 The Academic Editor was erroneously listed as Michael Levine. The Academic Editor for this paper is Joy Bergelson, Chicago University. This correction note may be found online at DOI: 10.1371/journal.pbio.0020382.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC521734.xml
529426	A National Health Insurance Program for the United States	The US will spend $1.79 trillion on health care in 2004, yet 44 million Americans remain uninsured. What the country needs, argues McCanne, is publicly funded universal health coverage	The total projected spending on health care in the United States for 2004 is $1.79 trillion—15.5% of its gross domestic product [ 1 ]. That amounts to $6,167 per person, almost twice what most nations with comprehensive systems spend on care. Most policy analysts agree that this level of spending should be more than enough to provide all Americans with high quality, comprehensive health care. Yet the United States falls far short of these goals. What are the flaws in the United States health system that prevent Americans from receiving value from this huge health care investment? And what are the options for improvement? Physicians for a National Health Program First, I should reveal my personal bias. Physicians should be well represented in the forefront of reform. As we look back on the past half century of failed health policy decisions, we see that the dominant physicians' organization in the United States, the American Medical Association (AMA), has opposed most reform measures that would result in an equitable, affordable system for everyone. Instead, the AMA has supported an agenda that promotes physicians' freedom to maximize their personal financial reward, even though those policies may deprive tens of millions of Americans access to affordable care. The AMA agenda has contributed significantly to the current high costs of American health care and to our failure to adequately address the mediocrity that characterizes health care in the United States. Many American physicians—including myself—believe that the funding infrastructure should be redesigned to maximize heath care resource allocation for the primary benefit of patients. Because of the failure of organized medicine to advocate on behalf of our patients, we decided that a new organization was needed. We established Physicians for a National Health Program ( www.pnhp.org ) [ 2 , 3 ]. The Uninsured and the Poorly Insured There are 45 million Americans with no health care coverage, and not surprisingly, lack of insurance is associated with worse health outcomes [ 4 ]. About 18,000 young adults die each year because they lack health insurance [ 4 ]. The uninsured are less likely than the insured to receive the professionally recommended standard of care for their chronic diseases, such as diabetes ( Figure 1 ) [ 5 ]. And if you have a serious health crisis while you are uninsured, you risk major debt or bankruptcy. Figure 1 Diabetes Management among Insured and Uninsured Adults, Ages 18–64 The figure is based on data from a United States national survey of 105,764 adults in 1997 and 117,364 in 1998 [ 5 ]. The proportions have been adjusted to the demographic characteristics of the study cohort, controlling for age, sex, race/ethnicity, region, employment status, education, and income. (Reprinted, with permission, from [ 4 ], courtesy of the National Academies Press, Washington, DC, United States.) Even the insured are inadequately covered. Employers and individuals who purchase coverage are rebelling at the high price of insurance premiums. To maintain competitive premiums, insurers are designing products that reduce the benefits they pay out by increasing the out-of-pocket portion that patients are required to pay for services received. Insured patients may have to pay cash for care until a designated amount is reached (the deductible)—which could be thousands of dollars. In addition, patients are often required to pay a dollar amount (co-payment) or a percentage of the charges (coinsurance) each time services are received. Insurers may also exclude specified services from coverage, such as maternity benefits or mental health services. Most insurance plans now use lists of contracted physicians and hospitals, and impose severe financial penalties for using health care providers that are not contracted. All of these measures reduce the value of insurance by shifting costs from the insurers to the patients who actually need care. Inadequate insurance coverage is making average-income Americans poorer. A recent study found that for 29% of individuals who had average or greater-than-average incomes and were continually insured, medical bills had caused significant financial problems [ 6 ]. For those who were not continuously insured, the percentages were even higher. These financial barriers are impairing access to beneficial services. The United States insurance market is now dominated by insurance plans that provide neither adequate health security nor financial security. Does Higher Health Spending Mean Better Quality? There is a widespread belief that the high spending in the United States means that high quality care is being delivered to the majority, who can afford both comprehensive coverage and the attendant out-of-pocket expenses. But international comparisons of industrialized nations have shown that the United States is in the bottom quartile of population health indicators such as life expectancy and infant mortality [ 7 ]. And in regional comparisons within the United States, increased levels of spending have not produced a commensurate improvement in health care outcomes. In fact, a recent study found that in a state-by-state comparison, there is an inverse relationship between spending and quality outcomes—the more expenditure, the worse the quality [ 8 ]. In 2000, the World Health Organization rated the United States first in its health expenditures per capita, but 37th in its overall health system performance, below most industrialized nations [ 9 ]. The United States is clearly not receiving adequate value for its health care investment. Some contend that the poor performance of the United States system is due to the funding of health care in the private sector, and that all would be well if the government would just take over funding. But it is not quite that simple. The greater part of health care in the United States—59%—is already funded by the tax system. On a per capita basis, the public, taxpayer-funded health care expenditures alone total more than the health care spending of every other nation's public and private funding combined (with the exception of Switzerland, in which total spending per capita equals our public spending alone) [ 10 ]. Flaws in Funding and Allocation How can the United States spend as much as it does and end up with such mediocre health care? Of the many reasons that exist, two are particularly important. The United States has a highly flawed system of funding health care and a flawed system of allocating its health care resources. In the United States, a multitude of private health plans cover the lucrative sector of society—low cost, healthy workers and their healthy families. But public programs must cover the higher costs of the elderly, individuals with permanent disabilities, and some low-income individuals. Since the uninsured are frequently unable to pay for the care they receive, the costs for their care are shifted to government programs or private plans, or to the charity of providers, even if unintended. The costly administrative excesses of private health plans, especially when contrasted to government programs, have been well documented [ 11 ]. This fragmented system of funding care places an even greater administrative and financial burden on the providers of health care. Although the exact amount is disputed, most policy analysts agree that replacing this fragmented system of funding care with a single, universal, publicly administered insurance program could recover 200 billion dollars or more, which are currently being wasted on useless and sometimes detrimental administrative services [ 11 ]. And what is wrong with the way that the United States allocates its resources? Many studies have confirmed that supporting a strong primary care base provides better outcomes at a lower cost [ 12 ]. But in the United States, specialized, high-technology care is heavily marketed, and providers of that high-tech care are rewarded more generously than primary care professionals. Yet studies show that these greater expenditures result in no additional benefit—and sometimes even in worse outcomes [ 8 , 13 ]. Excessive resources are allocated to inappropriate expansion of high-tech facilities and to training an excessive number of specialists to provide high-tech services [ 8 , 13 ]. Health Care Reform What has been the response to these deficiencies in the United States health care system? In the 1990s, the Clinton administration attempted to introduce a comprehensive system of funding universal health care. The system would have used marketplace principles in a program of managed competition, but their complicated idea pleased no one, and it was never even brought to a vote. Because of this miserable political failure, policymakers decided that any comprehensive approach should be avoided, and that reform must take place in incremental steps. To date, with the notable exception of the State Children's Health Insurance Program ( http://www.cms.hhs.gov/schip ), the accomplishments of these incremental health reform measures have been unimpressive. Over the past decade, those interested in reform have been preoccupied with managed care measures and, more recently, with consumer-directed measures that increase costs to patients by requiring greater out-of-pocket spending. But these measures are designed more to control costs than to increase coverage and access. In the debate on universal coverage, three general concepts have been put forward: (1) the expansion of our current system of public and private programs, (2) the establishment of a national health service with government ownership of the system, or (3) the replacement of all current funding with a single, publicly administered, publicly funded program of social insurance that does not alter the existing ownership status of the delivery system. The greatest political support today is for incremental expansions of our current programs, which, theoretically, would eventually result in universal coverage. There are innumerable variations of this approach. Most would increase the affordability of insurance premiums for private group and individual plans by providing financial assistance through tax policies and by modifying the benefits and coverage of the plans. Some policy analysts recommend that employers be mandated to offer coverage to their employees. Others recommend that individuals be required to purchase their own coverage. Since some individuals would be left without coverage, a public program, such as the existing Medicaid program for low-income individuals, would be used to cover everyone else. Many simulation studies have shown that these approaches could be effective in covering almost everyone, but they are the most expensive models of reform since they leave in place the administrative excesses of the fragmented system of funding care [ 14 ]. Also, to keep premiums affordable, these approaches may fall short on comprehensiveness of coverage and on the affordability of the out-of-pocket component, especially for those individuals with greater health care needs. In contrast, simulation of both the national health service and public social insurance models of reform have shown that they would provide truly comprehensive benefits for everyone, and that they are the least expensive models [ 14 ]. By integrating funding with the health care delivery system, both models are well suited for the introduction of an integrated information technology system. Such a system would provide invaluable data to assist with decisions on resource allocation, enabling incentives to be established that would strengthen the primary care base. It would also improve capacity planning for high-tech and specialized services, thereby ensuring appropriate access without excessive queues [ 15 ]. Even with insurance, a serious health crisis can lead to major debt or bankruptcy (Illustration: Rusty Howson, sososo design) The political threshold for adopting a government-owned health service model in the United States is very high, since most citizens fear the specter of “socialized medicine.” In contrast, the Medicare program, an insurance program for the retired and for those with long-term disabilities, is very popular. There is an increasing public perception that we may need to accept a greater government role in health insurance if we are to adequately address the deteriorating status of our health care system. Correcting the flaws in Medicare and then using the program to cover everyone may be a concept that can gain political traction in the United States. Conclusion Our political process is currently dominated by those who are enticed by the siren song of the market theorists and turn a deaf ear to the health policy scientists who plead for health care justice. The debate needs to focus on defining the best role for government in ensuring that people receive the best health care value. That debate needs to be guided by a thorough understanding and diligent application of sound health policy science. The continuing deterioration of affordability, coverage, and quality in health care makes it imperative that United States policymakers broaden their reform efforts beyond the ineffectual tinkering of incrementalism. A universal, single-payer, publicly funded and publicly administered program of social insurance would ensure access to affordable, comprehensive, high-quality health care for all. It should be the standard by which any other proposals are judged. If a better proposal can be crafted, now is the time to do it. People are dying while we delay.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC529426.xml
314463	Dorsoventral Patterning of the Mouse Coat by Tbx15	Many members of the animal kingdom display coat or skin color differences along their dorsoventral axis. To determine the mechanisms that control regional differences in pigmentation, we have studied how a classical mouse mutation, droopy ear ( de H ), affects dorsoventral skin characteristics, especially those under control of the Agouti gene. Mice carrying the Agouti allele black-and-tan ( a t ) normally have a sharp boundary between dorsal black hair and yellow ventral hair; the de H mutation raises the pigmentation boundary, producing an apparent dorsal-to-ventral transformation. We identify a 216 kb deletion in de H that removes all but the first exon of the Tbx15 gene, whose embryonic expression in developing mesenchyme correlates with pigmentary and skeletal malformations observed in de H / de H animals. Construction of a targeted allele of Tbx15 confirmed that the de H phenotype was caused by Tbx15 loss of function. Early embryonic expression of Tbx15 in dorsal mesenchyme is complementary to Agouti expression in ventral mesenchyme; in the absence of Tbx15 , expression of Agouti in both embryos and postnatal animals is displaced dorsally. Transplantation experiments demonstrate that positional identity of the skin with regard to dorsoventral pigmentation differences is acquired by E12.5, which is shortly after early embryonic expression of Tbx15 . Fate-mapping studies show that the dorsoventral pigmentation boundary is not in register with a previously identified dermal cell lineage boundary, but rather with the limb dorsoventral boundary. Embryonic expression of Tbx15 in dorsolateral mesenchyme provides an instructional cue required to establish the future positional identity of dorsal dermis. These findings represent a novel role for T-box gene action in embryonic development, identify a previously unappreciated aspect of dorsoventral patterning that is widely represented in furred mammals, and provide insight into the mechanisms that underlie region-specific differences in body morphology.	Introduction A fundamental question in developmental biology is how adjacent regions of the vertebrate body acquire differences in their appearance or morphology. Mechanisms that establish the general body plan make use of a relatively small number of signaling pathways shared among all animals (reviewed in Pires-daSilva and Sommer 2003 ), but the extent to which these pathways control finer differences between body regions is not clear. Among vertebrates, differences in the shape or number of skeletal elements, altered morphology of epidermal appendages, and variation in pigment distribution combine to produce the majority of what distinguishes one animal from another. Among these, pigment patterns are an excellent system to investigate how morphological differences arise, both for different regions of the body within a species and for different animals from closely related species. In natural environments, color variation is a nearly universal mechanism for recognition, camouflage, or both; consequently, a large number of pigment patterns have been characterized from an evolutionary and ecological perspective ( Boughman 2001 ; Jiggins et al. 2001 ). In the laboratory, color variation has been the subject of vertebrate genetics for more than a century ( Searle 1968 ; Silvers 1979 ), and many pigmentary components have been identified whose actions are understood in a cellular or organ-based context (reviewed in Bennett and Lamoreux 2003 ). Several mechanisms may contribute to regional differences in vertebrate pigmentation. In the embryo, alterations in the determination or migration of melanoblasts from the neural crest affect the number or distribution of pigment cells in the skin (reviewed in Reedy et al. 1998 ). Within hair follicles, paracrine signals control the type of pigment made in specific regions of the body or at specific times during the hair cycle (reviewed in Furumura et al. 1996 ; Barsh et al. 2000 ). Finally, movement of pigment granules within melanocytes or from melanocytes to keratinocytes makes use of cellular machinery that is shared by a variety of cell types, but that can vary in different regions of the body (reviewed in Marks and Seabra 2001 ). However, for all of these mechanisms—white spotting, pigment-type switching, and melanosome biogenesis—more is known about the identity of the molecular components than their spatial and temporal control. One of the most obvious aspects of regional color variation in vertebrates is a dark dorsal surface juxtaposed to a light ventral surface, apparent in the color of skin, scales, feathers, or hair, in which the boundary between dorsal and ventral compartments is often sharp and lies in register with the limbs. In rodents and probably other mammals, this dorsoventral difference in hair color is brought about by differences in pigment type as determined by allelic variation of the Agouti gene ( Bultman et al. 1992 ; Miller et al. 1993 ). Secreted by dermal papilla cells within each hair follicle ( Millar et al. 1995 ), Agouti protein causes melanocytes in that follicle to switch from the production of brown/black eumelanin to red/yellow pheomelanin. Agouti protein has a short radius of action ( Silvers and Russel 1955 ) and can be switched on and off during a single hair cycle ( Bultman et al. 1992 , 1994 ; Miller et al. 1993 ; Vrieling et al. 1994 ); thus, its regulated expression is thought to be responsible for the cream-colored or yellow ventral surface of mice carrying the black-and-tan ( a t ) allele and for the yellow markings around the feet, ears, or head, i.e., tan points or head spots, of certain dog breeds. In laboratory mice, previous studies from our group and others identified two predominant Agouti mRNA isoforms that differ by virtue of their transcriptional initiation site and 5′ untranslated exons. A “hair cycle-specific” transcript is expressed in both dorsal and ventral skin for 2–3 days during early hair growth, while a “ventral-specific” transcript is expressed throughout the entire period of active hair growth, but only in ventral skin ( Bultman et al. 1994 ; Vrieling et al. 1994 ). Animals carrying the a t allele express only the ventral-specific Agouti transcript ( Bultman et al. 1994 ; Vrieling et al. 1994 ) and have black dorsal hairs and cream-colored to yellow ventral hairs, with a sharp boundary at the level of the limb–body wall articulations and in the middle of the whisker pad. Ventral-specific Agouti isoforms are also expressed in developing skin from embryonic day 10.5 (E10.5) and beyond and may play a role in pigment cell differentiation ( Millar et al. 1995 ). Thus, regulatory elements for ventral-specific Agouti isoforms are responsive to dorsoventral positional cues established in the embryo and whose effects persist after birth. The boundary between dorsal and ventral color compartments in a t / a t mice bears superficial resemblance to dorsoventral boundaries apparent for many other mammals, but morphogenetic differences between dorsal and ventral skin seem likely to include more elements than the type of pigment made by hair follicle melanocytes. In particular, dermis of the flank has at least two distinct origins: dermatomal derivatives of somites and loose mesenchyme derived from the lateral plate mesoderm ( Mauger 1972 ; Christ et al. 1983 ; Olivera-Martinez et al. 2000 ; Nowicki et al. 2003 ); these lineages are established early in development and could, in principle, set up compartments whose identity contributes to dorsoventral differences in adult skin. To better understand the mechanisms that give rise to differences between dorsal and ventral skin and to the boundary between them, we have determined how several morphologic characteristics vary along the dorsoventral axis of the mouse and how these characteristics correspond to ventral-specific Agouti expression and the lineage boundary that distinguishes somite from lateral plate derivatives. Our results indicate that the apparent uniformity of the dorsoventral boundary represents the sum of independent mechanisms that affect melanocyte density and/or differentiation, pigment-type synthesis, and hair length; surprisingly, none of these coincide with the somite–lateral plate lineage boundary. We also make use of a classical mouse mutation, droopy ear ( Curry 1959 ), that produces a dorsal-to-ventral transformation of flank coat color by allowing expansion of the ventral-specific Agouti transcript. By positional cloning and gene targeting, we identify an allele of droopy ear , de H , as a loss of function for Tbx15 , which encodes a T-box transcription factor expressed in a dynamic and spatially restricted manner in the developing skin and musculoskeletal system. Embryonic expression and transplantation studies suggest that Tbx15 is required to establish certain characteristics of dorsal patterning in mesenchymal cells of the developing flank. These results identify a previously unappreciated aspect of dorsoventral patterning that is widely represented in furred mammals and provide insight into the mechanisms that underlie region-specific differences in body morphology. Results Morphological Components of Dorsoventral Skin Differences Besides the obvious change in hair color that frequently distinguishes dorsal from ventral skin, casual observation suggests there are additional differences in hair length, distribution of hair type, and skin thickness. Furthermore, dorsoventral differences in pigmentation can represent differences in the number and/or differentiated state of pigment cells, as well as the type of pigment synthesized in response to expression of Agouti . In particular, ventral hair of a t / a t animals can vary from cream-colored to reddish-yellow depending on age, strain background, and position along the dorsoventral axis. To evaluate the relationship among these components, we compared their features among mice of different Agouti genotypes. Semiquantitative measurements of hair length plotted as a function of dorsoventral position reveal that the apparent sharp boundary between dorsal and ventral pigment compartments in a t / a t mice coincides with a more gradual change in both hair color and hair length ( Figure 1 A– 1 D). Within the region of transition from dorsum to ventrum ( Figure 1 B), flank hairs from a t / a t mice become progressively shorter and exhibit increasing amounts of pheomelanin deposition progressing from the tip to the base of the hair. However, the region of transition for hair length is considerably broader than that for pigmentation and independent of Agouti genotype. Although hair-cycle timing varies along the rostrocaudal axis, measurements of absolute hair length for mice matched for age and rostrocaudal level are remarkably similar ( Figure 1 D). Furthermore, measurements of relative hair length for animals of different age, size, and Agouti genotype also are very similar when normalized to body circumference ( Figure 1 C). Taken together, these observations indicate that variation of hair length along the dorsoventral axis is stereotyped and maintained through multiple hair cycles, with a transition in hair length that is gradual and encompasses the pigment-type transition in a t / a t mice. Figure 1 Dorsoventral Skin Characteristics (A) Skin slices from animals of different age and genotype demonstrate similar patterns of hair-length variation along the dorsoventral axis (scale bar = 1 cm). (B) Enlarged area from (A), demonstrating the transition in hair length and color in a t / a t mice (scale bar = 0.375 cm). (C) Proportional hair length for (A) plotted as a function of relative position along the dorsoventral axis. (D) Hair length plotted as a function of absolute position along the dorsoventral axis for 8-wk-old BA strain mice. (E) Proportion of zigzag hairs (± SEM) differs slightly between dorsum and ventrum of inbred mice ( p < 0.0001, χ 2 test, n = 1,958, 1,477, 1,579, 1,502). (F) Differences in dorsal and ventral skin development at P4.5 (scale bar = 1 mm, upper; 200 μm, lower). (G) Differences in hair melanin content and DOPA staining for dorsum (d), flank (f), and ventrum (v) in a e / a e and a t / a t mice. The upper panel also demonstrates a cream-colored appearance of the a t / a t ventrum. The middle panel shows representative awls (scale bar = 100 μm). The lower panel shows DOPA-stained dermis (scale bar = 200 μm). Dorsal and ventral skin develop at different rates. Transverse sections of skin at postnatal day 4.5 (P4.5) exhibit dorsal hair follicles that are noticeably more developed than ventral hair follicles, along with a gradual dorsoventral decrease in dermal thickness ( Figure 1 F). However, differences in skin thickness disappear by 3–4 wk of age ( Forsthoefel et al. 1966 ), and, overall, the proportion of different hair types is also similar in dorsa and ventra of adult mice. In age-matched inbred mice, we observed a small decrease in the ratio of undercoat hairs (zigzags) to overcoat hairs (auchenes, awls, and guard hairs) in dorsum compared to ventrum ( Figure 1 E), but there was no consistent difference in hair-type distribution for outbred mice (data not shown). Differences between dorsal and ventral pigmentation of a t / a t mice are usually attributed to pigment-type differences caused by ventral-specific expression of Agouti , but animals homozygous for a null allele of Agouti , extreme nonagouti ( a e ), have ventral hairs that contain less melanin than dorsal hairs, giving a slightly paler appearance to the ventral coat ( Figure 1 G). Using DOPA staining as an indicator of tyrosinase activity, we observed a gradual dorsoventral transition in isolated dermis preparations from P4.5 a e / a e mice ( Figure 1 G). By contrast, skin from a t / a t mice reveal an abrupt dorsoventral transition of DOPA staining, which probably reflects the additive effects of reduced melanin content (as in a e / a e mice) and downregulation of tyrosinase activity induced by Agouti . Melanin content of individual hairs is likely to be influenced both by the number of pigment cells and their follicular environment. Regardless, dorsoventral differences in hair pigment content of a e / a e mice persist throughout multiple hair cycles into adulthood, similar to hair length (but unlike skin thickness). Thus, at least three characteristics distinguish dorsal from ventral skin: differences in pigment-type synthesis (depending on Agouti genotype), differences in hair length, and differences in melanin content. Ventralization of Skin Morphology by the droopy ear Mutation Named after its effects on craniofacial morphology, droopy ear is a recessive mutation on mouse Chromosome 3; the original allele described more than 40 years ago by Curry (1959) is extinct, but a spontaneous remutation that occurred in Harwell, de H , is available through The Jackson Laboratory (Bar Harbor, Maine, United States). External craniofacial malformations are the most obvious characteristic of de H / de H animals, including widely spaced eyes, small palpebral fissures, a broad nasal area, and a shortened skull held in an elevated position, which presumably causes or contributes to the abnormal position of the ears. We became interested in droopy ear because the original allele was described to affect pigment pattern in a way that suggests a possible dorsal to ventral transformation: “On a genetic background ( a t and A W ) which causes the belly hair to be lighter than the back hair, the belly hair comes up farther round the sides of the body and face” ( Curry 1959 ). An abnormal dorsoventral pigment pattern is readily apparent in a t / a t ; de H / de H mice, but comparison to nonmutant animals is more accurately described in terms of ventral, lateral, and dorsal regions (Figures 1 G and 2 A). The ventral region has short hairs with a gray base and cream-colored tip whose boundary coincides with the limb–body wall junction; both the appearance of this region and position of the boundary are approximately similar in a t / a t compared to a t / a t ; de H / de H mice. The lateral region contains yellow hairs of progressively increasing length; in a t / a t mice, the lateral region appears as a thin yellow stripe along the flank, but in a t / a t ; de H / de H mice, the lateral region is considerably expanded with a diffuse boundary along the dorsal flank, and a dorsal eumelanic region whose size is correspondingly reduced ( Figure 2 A and 2 B). Total body size is smaller in mutant compared to nonmutant animals, but the proportion of body circumference occupied by the lateral region in mutant animals is increased about 2-fold, from 11.9% to 22.2% ( Figure 2 C). The proportion of the ventral cream-colored region is also expanded a small amount, 47.9% in mutant compared to 37.8% in nonmutant animals, but expansion of the lateral region, which occurs at all levels of the body, including the limbs and the cranium (but not the whisker pad), is the major feature responsible for the ventralized appearance caused by de H . Figure 2 The de H Pigmentation Phenotype (A) 10-wk-old de H /de H and nonmutant animals on a a t background. A thin stripe of yellow hair normally separates the dorsal black hairs from the ventral cream hairs. In de H , the yellow stripe is extended dorsally, and the boundary between the yellow and the black hairs is fuzzier. (B) Skin slices taken from 1.5-mo-old de H /de H and nonmutant littermates (scale bar = 0.5 cm). (C) Proportion of total skin area as determined by observation of pelts taken from the interlimb region. The proportion occupied by the yellow lateral compartment (± SEM) differs between mutant and nonmutant littermate flanks ( p < 0.0005, paired t -test, n = 6 pairs). There is also (data not shown) a small increase in the proportion of total skin area occupied by the ventral cream-colored compartment, 47.9 % in mutant compared to 37.8% in nonmutant ( p < 0.005, paired t -test, n = 6 pairs). (D) On an a e /a e background, the extent of dorsal skin pigmentation is reduced in de H /de H neonates (P3.5). (E) Hair length in a representative pair of 1.5-mo-old de H /de H and nonmutant littermates, averaged over three skin slices at different rostrocaudal levels, and plotted as a function of the absolute distance from middorsum or the percentage of total slice length. To investigate whether de H affects other dorsoventral skin characteristics besides pigment-type switching, we examined its effects on hair length and pigmentation in an a e / a e background. Overall, de H causes a small but consistent reduction in hair length in both dorsum and ventrum; when mutant and nonmutant animals are normalized for body circumference, reduced hair length is most apparent in the lateral region ( Figure 2 E). Adult a e / a e ; de H / de H animals exhibit body-size reduction and skeletal abnormalities, but display no coat-color phenotype (data not shown). However, a e / a e and a e / a e ; de H / de H neonates are clearly distinguishable in the first few days after birth, when a dorsoventral gradient of melanogenic activity is apparent under the skin ( Figure 2 D). At this stage, melanoblast migration from the neural crest is mostly complete, but there is a dorsoventral gradient in melanocyte differentiation and pigment synthesis. The skin of a e / a e neonates appears uniformly dark over the entire dorsum, but in a e / a e ; de H / de H neonates, the area of dark skin is more restricted, particularly above the limbs, and resembles the pattern of dorsal eumelanin in a t / a t ; de H / de H adult animals. Taken together, these observations suggest that de H interferes with the establishment of dorsoventral patterning during skin development by causing dorsal expansion of a lateral region that is normally 3–5 mm in width. This same region may serve as a boundary between dorsal and ventral skin by inhibiting melanocyte differentiation, by promoting pheomelanin synthesis, and by supporting a progressive increase in hair growth from ventrum to dorsum. As described below, the gene defective in de H , Tbx15 , is normally expressed in the dorsal region and therefore is likely to play a role in establishing the size and dorsal extent of the lateral region. Positional Cloning of de H As a visible marker, early linkage studies with the original droopy ear allele or the de H allele identified a map position in the middle of Chromosome 3, distal to matted and proximal to Varitint-waddler ( Carter and Falconer 1951 ; Curry 1959 ; Lane and Eicher 1979 ; Holmes et al. 1981 ). We used an F 2 intercross with CAST/Ei mice to localize de H to a 0.1 cM interval between D3Mit213 and 16.MMHAP32FLF1, which was refined by development of a bacterial artificial chromosome (BAC) contig and additional markers to a 1.4 Mb region that contained eight genes, including Tbx15 ( Figure 3 A). We considered Tbx15 as an excellent candidate for the skeletal abnormalities caused by de H , based on studies by Agulnik et al. (1998) , who described its embryonic expression in the craniofacial region and developing limbs. Figure 3 Molecular Genetics of de H and Tbx15 (A) Genetic and physical map, as described in the text. Markers M1 to M3 are SSCP markers generated from a BAC contig of the region; marker M4 is STS 16.MMHAP32FLF1 and was also used as an SSCP marker. M2 and M3, which flank the Tbx15 and M6pr-ps on the UCSC genome browser map and lie 634 kb apart, were nonrecombinant with de H in 2340 meioses. (B) The de H mutation is a deletion that starts in Tbx15 intron 1 and ends in the M6pr-ps . (C) Sequence of deletion breakpoints. (D) Diagram of Tbx15 LacZ allele constructed by gene targeting. As described in the text, this allele is predicted to give rise to a protein truncated after approximately 154 codons and is lacking critical residues of the T box. Heterozygotes for the targeted allele exhibit normal size, morphology, and hair-color patterns, but homozygotes and Tbx15 LacZ / de H compound heterozygotes are identical to de H homozygotes. Using sequence information from Agulnik et al. (1998) and the partially completed mouse genome sequence, we found that portions of several Tbx15 exons could not be amplified from de H / de H genomic DNA. The same gene was initially referred to as Tbx8 ( Wattler et al. 1998 ) and then later renamed Tbx14 , but is currently referred to in several vertebrate genomes as Tbx15 ( Agulnik et al. 1998 ; Begemann et al. 2002 ). By comparing the sequence of a 1.3 kb junction fragment amplified from de H / de H genomic DNA to publicly available mouse genome sequence, we identified a 216 kb deletion that extends from Tbx15 intron 1 to 148 kb downstream of the polyadenylation sequence in a region annotated as a mannose-6-phosphate receptor pseudogene, M6pr-ps ( Figure 3 B and 3 C). ( Ludwig et al. 1992 ). By Northern blot analysis, we identified a fusion transcript produced from the de H chromosome (data not shown). However, the deletion removes 534 of the 602 amino acids encoded by Tbx15 (including the T-box DNA-binding domain), de H /+ animals are grossly normal, and the phenotype of de H / de H animals is identical to that described for the original allele. In addition, other than M6pr-ps , no other genes or transcripts have been annotated to the 216 kb deletion. While the positional cloning work was underway, one of us (A. Russ) generated an independent mutation of Tbx15 by gene targeting in embryonic stem cells. The targeted allele, Tbx15 LacZ , carries an IRES-LacZ-neo cDNA cassette that disrupts the open reading frame at codon 154 early in the T-box domain ( Figure 3 D). Animals heterozygous for the targeted allele are completely normal with regard to size, skeletal morphology, and hair-color distribution, but Tbx15 LacZ / Tbx15 LacZ homozygotes were noted to exhibit reduced body size and an abnormal craniofacial appearance identical to that caused by de H . We generated Tbx15 LacZ / de H compound heterozygotes; on an A w / a t background, these animals exhibited the same abnormal restriction of dorsal pigmentation at P3.5 and expanded yellow flank area as described above for de H / de H animals (see Figure 2 ). These observations demonstrate that the pigmentary and craniofacial characteristics of de H are caused by loss of function for Tbx15 . Expression of Tbx15 and Agouti Previous studies by Agulnik et al. (1998) using whole-mount in situ hybridization described expression of Tbx15 as first detectable at E9.5 in the limb buds, progressing to the branchial arches, flanks, and craniofacial regions through E12.5. To investigate this pattern in more detail, we hybridized a Tbx15 mRNA probe to a series of transverse sections at E12.5 and observed expression in multiple mesenchymal tissues of the head, trunk, and developing limbs ( Figure 4 A), much of which is consistent with the skull, cervical vertebrae, and limb malformations reported for mice carrying the original droopy ear allele. Figure 4 Developmental Expression of Tbx15 (A) At E12.5, transverse sections at different levels show expression in head mesenchyme (a and b); myotome, occipital, and periocular mesenchyme (b); palatal shelf, cervical sclerotome, and nasal cartilage (c); maxillary and mandibular processes (d); limbs (e); and myotome and lateral mesenchyme (e and f) (scale bars = 500 μm). (B) Transverse sections through the flank at different times show expression in lateral mesenchyme (E11.5), expanding dorsally at E12.5, and both ventrally and dorsally at E13.5, detectable in loose mesenchyme underlying the dermis and the abdominal and subcutaneous muscles (scale bar = 500 μm). At P3.5, Tbx15 is expressed in the entire dermis and is most strongly expressed in dermal sheaths (scale bar = 200 μm). We were particularly interested in determining the exact nature of the embryonic flank expression relative to the ventralized phenotype of adult de H / de H mice. Transverse abdominal sections from different times during development reveal a dorsolateral band of expression in the superficial mesenchyme at E11.5 that broadens both dorsally and ventrally over the next several days ( Figure 4 B). By E13.5, the developing dermis has become separated from the loose mesenchyme by a subcutaneous muscle layer; Tbx15 is expressed in all of these layers as well as the underlying abdominal muscles. In P3.5 skin, Tbx15 is expressed in both dorsal and ventral skin, most strongly in the condensed upper dermis and developing dermal sheaths of hair follicles; faint expression can also be detected in rare dermal papillae cells ( Figure 4 B). Although the effects of Agouti on pigment-type switching occur during postnatal hair growth, the ventral-specific isoform of Agouti is expressed in developing skin beginning at E11.5. We compared adjacent sections hybridized with probes for Tbx15 and Agouti and observed complementary patterns at E12.5, with expression of Agouti in ventral skin and expression of Tbx15 in dorsal skin ( Figure 5 A and 5 B). The junction between expression domains is indistinct, and by E14.5, Tbx15 expression extends ventrally and overlaps extensively with Agouti expression ( Figure 5 C and 5 D). Figure 5 Embryonic Expression of Tbx15 Compared to Agouti in a t / a t Mice (A and C) Tbx15 . (B and D) Agouti . At E12.5, expression of Tbx15 in dorsal skin is approximately complementary to that of Agouti in ventral skin. At E14.5, the levels of expression for both genes are lower, but Tbx15 expression has expanded ventrally and overlaps extensively with that of Agouti . In all four panels, arrows mark the approximate ventral limit of Tbx15 and the approximate dorsal limit of Agouti (scale bars = 500 μm). We also examined the effect of de H on expression of Agouti and found no difference between mutant and nonmutant at E12.5 or E13.5 (data not shown). However, at E14.5, the normal ventral-to-dorsal gradient of Agouti expression appeared to extend more dorsally in de H / de H embryos ( Figure 6 A). In P4.5 skin, expression of Agouti is also extended dorsally in de H / de H animals and is most apparent in the midflank region within the upper dermis and dermal papillae cells ( Figure 6 B). Thus, while the pigmentation phenotype of de H / de H mice can be explained, not surprisingly, by dorsal extension of Agouti expression after birth, patterned expression of Tbx15 and Agouti are apparent some 10 days earlier, between E12.5 and E13.5, and the effects of Tbx15 deficiency on expression of Agouti can be detected by E14.5. Figure 6 Effect of de H on Agouti Expression Comparable sections from a t /a t ; de H / de H and a t /a t ; +/+ littermates. (A) At E14.5, de H / de H embryos have a smaller body cavity and loose skin within which Agouti expression appears to be shifted dorsally, as marked by arrows (scale bars = 500 μm). (B) At P4.5, Agouti expression in both dorsal and ventral skin is similar in de H / de H compared to nonmutant, but in the midflank region, there is increased Agouti expression in de H / de H , especially in the upper dermis (scale bars = 200 μm). Sections shown are representative of two mutant and two nonmutant samples examined at each time. Relationship of Embryonic Tbx15 Expression to Dorsal and Ventral Pigmentation Domains The observations described above are consistent with a model in which transient expression of Tbx15 in the embryonic dorsal flank is required to establish positional identity of the future dermis, at least with respect to pigment-type synthesis caused by the ventral-specific Agouti isoform. To further investigate this hypothesis, we carried out transplantation experiments in which pieces of embryonic skin were isolated from different dorsoventral positions. We evaluated the embryonic skin fragments for their potential to give rise to different hair colors and for their expression of Tbx15 and Agouti . Previous studies by Silvers and colleagues ( Poole and Silvers 1976 ) showed that dorsal and ventral skin isolated from a t / a t embryos gives rise to black and yellow hair, respectively, when transplanted into testis and allowed to develop for several weeks. Furthermore, dermal–epidermal recombination experiments carried out at E14.5 demonstrated that positional identity is carried by the embryonic dermis. In a variation on this experiment, we divided embryonic skin from a t / a embryos into dorsal, flank, and ventral pieces and analyzed the different pieces for their ability to give rise to black or yellow hair after testis transplantation, and, in parallel, for gene expression using in situ hybridization. For the purposes of a reproducible morphologic boundary, we divided flank from ventral skin based on a change in skin thickness and divided dorsal from flank skin at the level of an ectodermal notch that lies at the same level as the ventral extent of the myotome ( Figure 7 ) ( Huang and Christ 2000 ; Olivera-Martinez et al. 2000 ; Sudo et al. 2001 ; Burke and Nowicki 2003 ; Nowicki et al. 2003 ). Figure 7 Embryonic Establishment of Dorsoventral Skin Patterning Pieces of skin from dorsal, flank, and ventral regions of a t /a E12.5 embryos were transplanted into the testes of congenic animals as described in the text. Hair color of the grafts was examined 3 wk later. Grafts of ventral embryonic skin ( n = 3) produced yellow hairs, dorsal embryonic skin ( n = 4) produced black hairs, and flank embryonic skin produced mostly (13 out of 15) black and yellow hairs in distinct regions as shown. In parallel, in situ hybridization studies revealed that the embryonic flank contains the boundary of expression between Agouti and Tbx15 (scale bars = 1 mm for hairs and 200 μm for in situ hybridization results). We found that E12.5 is the earliest time at which embryonic ventral skin is able to produce hair when transplanted to the testis. Of the grafts that produced hair, ventral skin gave rise to yellow hair ( n = 3), and dorsal skin gave rise to black hair ( n = 4). Transplantation of flank skin gave rise to a patch of yellow hair juxtaposed against a patch of black hair in 85% of the successful grafts ( n = 13); the remaining two flank grafts produced solely black or yellow hair. In no case did we observe intermingling of black and yellow hairs. As predicted from the experiments using tissue sections (see Figures 5 and 6 ), dorsal pieces expressed Tbx15 but not Agouti , while flank pieces expressed both genes (see Figure 7 ). Thus, dorsoventral identity for adult pigmentation is established by the time when patterned expression becomes apparent for Tbx15 and Agouti (E11.5–E12.5); furthermore, positional identity is maintained throughout later stages of skin development, even though expression of Tbx15 broadens to include ventral as well as dorsal skin. Relationship of the Dorsoventral Pigment Boundary to Lineage Compartments and the Lateral Somitic Frontier The ectodermal notch that we used to mark the boundary between embryonic dorsum and embryonic flank is a characteristic feature in vertebrate embryos. In cell lineage studies carried out in the chick system, the notch serves as a landmark for the boundary between dermis derived from somitic mesoderm and dermis derived from lateral plate mesoderm and has been termed the “lateral somitic frontier” ( Olivera-Martinez et al. 2000 ; Sudo et al. 2001 ; Burke and Nowicki 2003 ; Nowicki et al. 2003 ). Although fate-mapping studies have not been carried out in mammalian embryos, somite- and lateral plate-derived mesoderm could give rise to precursors for dermis dorsal and ventral to the limb–body wall junction, respectively. However, this notion conflicts with our observation that the future pigmentation boundary lies ventral to the ectodermal notch (see Figure 7 ). To examine directly the relationship between the pigmentation boundary and dermis derived from lateral plate mesoderm, we made use of a Cre transgene driven by the Hoxb6 promoter that was developed by Kuehn and colleagues ( Lowe et al. 2000 ). As described by Lowe et al. (2000) , midgestation embryos carrying both the Hoxb6-Cre transgene and the R26R lacZ reporter gene ( Soriano 1999 ) exhibit X-Gal staining in lateral plate mesoderm but not somite-derived mesoderm of the trunk. In whole-mount skin preparations from P1.5 or P4.5 neonatal animals, we observed a ventral band of dark X-Gal staining corresponding to lateral plate-derived dermis, which represents 63% of the total circumference ( Figure 8 A). However, in parallel preparations from a t / a t mice, the ventral pheomelanin domain represents 47% of the total skin circumference; therefore, the proportions of total skin circumference occupied by dorsal eumelanin and somite-derived dermis are 53% and 37%, respectively ( Figure 8 B). These results indicate that the pigmentation boundary is clearly distinct from, and more ventral to, the boundary between lateral plate- and somite-derived dermis. Figure 8 Comparison of the Dorsoventral a t /a t Pigmentation Boundary to the Lateral Somitic Frontier (A) Dorsoventral slices of skin from at the midtrunk region prepared such that the dorsal midline lies in the center of the slice. Sections were taken at P1.5 (a) or P4.5 (b–e) from a t /a t or R26R /+; Tg.Hoxb6-Cre /+ mice (the latter were stained with X-Gal), as described in Materials and Methods. For purposes of comparison, images were proportionally scaled. The boundary of X-Gal staining marks dermis derived from lateral plate versus dermis derived from mesoderm (the lateral somitic frontier) and lies more dorsal than the a t /a t pigmentation boundary. (B) Quantitation of mean (± SEM) dorsal pigmentation area ( n = 5) and somite-derived dermis area ( n = 3) shows a significant difference ( p < 0.005, t -test). (C) RNA in situ hybridization showing that Tbx15 expression at E11.5 is complementary to En1 expression on the flank (scale bars = 200 μm). The arrow indicates the boundary between the expression domains of the two genes. Because the pigmentation boundary lies in register with the limb–body wall junction (see Figure 2 ), we wondered whether mechanisms used for dorsoventral limb patterning might be related to those used to establish the pigmentation boundary. In the developing limb, Engrailed1 ( En1 ), Wnt7a , and Lmx1b are part of a network whose restricted domains of expression help to establish dorsoventral identity (reviewed in Niswander 2003 ). En1 is transiently expressed in the developing flank; at E11.5, transverse abdominal sections reveal domains in the neural tube, somite-derived mesenchyme, and the ventral body wall ( Figure 8 C). An adjacent section hybridized with Tbx15 reveals a complementary pattern in the flank, which provides additional evidence for developmental mechanisms that establish a pigmentation boundary entirely within lateral plate mesoderm and independent of lineage restrictions imposed by the lateral somitic frontier. Discussion Several mutations and genes have been identified that affect the pattern of hair follicle development, but Tbx15 is the only gene of which we are aware that affects the pattern of hair pigmentation in different body regions. Ventral areas that normally produce yellow hair in the trunk, limbs, and craniofacial regions are expanded in de H / de H mice and, in the trunk at least, represent inappropriate dorsal expression of an Agouti mRNA isoform that is normally restricted to ventral skin. The de H allele is caused by a large deletion that removes most of the Tbx15 coding sequence, but the pleiotropic phenotype is caused by a simple loss of function for Tbx15 rather than a dominant-negative or contiguous gene effect. In particular, there is no heterozygous phenotype, no other genes lie within or close to the deletion breakpoints, and the expression pattern of Tbx15 is consistent with the spectrum of phenotypic abnormalities in both the original de allele and the de H allele. Finally, a Tbx15 targeted allele has the same phenotype as de H . Our results suggest that patterned expression of Tbx15 provides an instructional cue required to establish the future identity of dorsal dermis with regard to pigmentary and hair length patterning. The ventral edge of Tbx15 expression in the developing flank does not correspond to a known lineage compartment, but, like limb development, occurs within lateral plate mesoderm. These findings represent a novel role for T-box gene action in embryonic development and provide evidence for a previously unappreciated complexity to acquisition of dorsoventral positional identity in mammalian skin. Distinct Morphologic Regions Represent the Sum of Different Gradients The visual boundary between dorsal and ventral skin in a t / a t mice is reminiscent of other systems in which adjacent compartments enforce a binary choice between alternative patterns of gene expression and cell fate (reviewed in Dahmann and Basler 1999 ). However, Agouti mRNA in both embryonic and postnatal skin is distributed along a gradient whose dorsal boundary is indistinct and overlaps with two additional gradients recognized by their effects on hair length and histochemical staining for melanocytes. The three gradients are close but not congruent, and it is their proximity that gives rise to the superficial distinction between dorsal and ventral skin of a t / a t mice. Indeed, slight differences between the regions of transition for pigment-type switching and pigment content give rise to a subtle yellow stripe along the flank (see Figures 1 , 2 , and 9 A). Levels of Agouti mRNA remain high throughout the entire ventrum, but hair pigment content is reduced, giving rise to a cream-colored region in the ventrum that, depending on age and genetic backgrounds, may appear more or less distinct from the yellow flank stripe. Figure 9 Model for Acquisition of Dorsoventral Patterning in the Trunk and the Role of Tbx15 (A) A tricolor pigmentation pattern is generated by the combination of distinct mechanisms that affect distribution of Agouti mRNA and histochemical staining for melanocytes; effects of the latter mechanism by itself are evident in a e / a e mice (see Figure 1 ). In a t /a t mice, reduced hair melanocyte activity and high levels of Agouti mRNA in the ventrum lead to a cream color; as melanocyte activity gradually increases towards the dorsum, a lateral stripe is apparent on the flank. The distributions of Agouti mRNA and histochemical staining for melanocytes are both affected by Tbx15 and are externally evident by a widening of the lateral stripe and an increased proportion of total skin occupied by the cream-colored area. (B) The lateral yellow stripe in a t /a t mice lies at the same level as the limb dorsoventral boundary. As described in the text, we propose that distinct dorsoventral compartments in ectoderm of the trunk provide an instructional cue to the mesoderm, leading to expression of Tbx15 in dorsal trunk mesenchyme and acquisition of dorsal dermis character. In the absence of Tbx15 , dorsal mesenchyme assumes ventral characteristics instead. Loss of Tbx15 affects dorsoventral transitions of hair length, pigment content, and expression of the ventral-specific Agouti isoform; however, the former two effects are subtle and contribute little, if at all, to the abnormal pigmentation of adult de H / de H mice. Thus, despite the abnormal pattern of dark skin in neonatal de H / de H mice (e.g., Figure 2 D), the most obvious feature in adults is dorsal displacement of the “boundary” between black and yellow hair ( Figure 9 A). Genetics of Tbx15 Named for the presence of a DNA-binding domain first identified in the mouse Brachyury gene (haploinsufficiency causes a short tail), T box–containing genes have been identified as developmental regulators in a wide spectrum of tissues and multicellular organisms (reviewed in Papaioannou 2001 ). The Tbx15 subfamily, which also includes Tbx18 and Tbx22 , is likely to have arisen during early chordate evolution since there is a single gene in amphioxus but no obvious homolog in the fly genome ( Ruvinsky et al. 2000 ). Consistent with this relationship, the three genes are expressed in partially overlapping patterns that include anterior somites ( Tbx18 and Tbx22 ), limb mesenchyme ( Tbx15 and Tbx18 ), and craniofacial mesenchyme (all three genes, Tbx15 more broadly than Tbx18 or Tbx22 ) ( Agulnik et al. 1998 ; Kraus et al. 2001 ; Braybrook et al. 2002 ; Bush et al. 2002 ; Herr et al. 2003 ). These observations suggest that an ancestral gene for Tbx15 , Tbx18 , and Tbx22 may have been important for craniofacial development in cephalochordates, with acquisition of additional expression patterns and developmental functions in the limb and the trunk during early vertebrate evolution. Expression of Tbx18 and Tbx22 has not been reported in embryonic flank mesenchyme, which suggests that Tbx15 is the only family member involved in establishing the dorsoventral identity of the trunk. However, it would not be surprising to find some degree of functional redundancy in animals mutated for two or three of the subfamily members in other body regions, particularly the limbs and the head. For example, mutations in Tbx22 cause the human syndrome X-linked cleft palate and ankyloglossia ( Braybrook et al. 2001 ). Despite high levels of Tbx22 expression in periocular embryonic mesenchyme ( Braybrook et al. 2002 ; Bush et al. 2002 ; Herr et al. 2003 ), the condition does not affect the eye, perhaps because residual activity is provided by Tbx15 in the same region. In an initial description of the expression and map location of mouse Tbx15 , Agulnik et al. (1998) suggested human Tbx15 that lies on Chromosome 1p11.1 as a candidate for acromegaloid facial appearance (AFA) syndrome, for which there is a weak positive LOD score to Chromosome 1p ( Hughes et al. 1985 ). Originally described as a rare autosomal-dominant syndrome with progressive facial coarsening, overgrowth of the intraoral mucosa, and large, doughy hands, more recent case reports describe macrosomia, macrocephaly, or both and generalized hypertrichosis with progressive coarsening ( Dallapiccola et al. 1992 ; Irvine et al. 1996 ; da Silva et al. 1998 ; Zelante et al. 2000 ). The de H phenotype exhibits little overlap with these features; instead, we suggest a more likely candidate for mutations of human TBX15 would be frontofacionasal syndrome, an unmapped autosomal recessive condition characterized by brachycephaly, blepharophimosis, and midface hypoplasia ( Reardon et al. 1994 ). Two of us (S. Kuijper and F. Meijlink) became interested in the de H mutation because of its effects on skeletal development ( Curry 1959 ) and the possibility that the aristaless -related gene Alx3 might be allelic with droopy ear ( ten Berge et al. 1998 ). In spite of similarities between skeletal phenotypes of de H and Alx3 or Alx4 mutants, subsequent experiments (unpublished data) excluded allelism of Alx3 and de H , and a full description of the Tbx15 skeletal phenotype will be published elsewhere. Developmental Mechanism of Tbx15 Expression and Action in the Skin Our attention to the role of Tbx15 in pigment patterning was motivated by the effects of Agouti in postnatal animals. However, Agouti is also expressed in the embryo, where it provides a convenient marker of ventral dermis identity. Because an expanded domain of embryonic Agouti expression in de H / de H animals is detectable by E14.5, the effects of Tbx15 on dorsoventral patterning must occur prior to this time. Among other T-box genes whose developmental actions are at least partially understood, two general themes have emerged, one focused on the ability to specify alternative fates for an undifferentiated group of precursor cells and another focused on the ability to support proliferative expansion of a cell population whose fate is already determined (reviewed in Tada and Smith 2001 ). Either mechanism may apply to the apparent dorsal-to-ventral transformation in de H / de H mice. For example, while the expanded domain of Agouti expression in postnatal de H / de H animals can be traced to events that occur between E11.5 and E13.5, the underlying cause may be that embryonic cells in dorsolateral mesenchyme acquire a ventral rather than dorsal identity or that those cells fail to proliferate normally, followed by compensatory expansion of ventral cells. Cell lineage studies should provide a definitive answer, but we favor the latter hypothesis, because measurements of dorsoventral regions according to hair color in de H / de H mice revealed a small increase of the cream-colored ventral region in addition to the approximate doubling of the yellow flank region (see Figure 2 ). In embryonic mesenchyme, expression of Tbx15 and Agouti are complementary, and it is possible that Tbx15 acts directly to inhibit Agouti transcription in dorsolateral mesenchyme. However, the ability of Tbx15 to suppress expression of the ventral-specific Agouti isoform in postnatal mice is likely to be indirect, since postnatal expression of Tbx15 occurs broadly along the dorsoventral axis and overlaps extensively with that of Agouti . In either case, the targets of Tbx15 action in the skin include genes in addition to Agouti , since hair length and melanocyte distribution exhibit a demonstrable, albeit subtle, alteration in animals that carry a null Agouti allele. One potential target is Alx4 , which, like Agouti , is expressed in ventral embryonic mesenchyme, and, when mutated, affects hair-follicle as well as limb and craniofacial development ( Qu et al. 1997 , 1998 ; Wu et al. 2000 ; Wuyts et al. 2000 ; Mavrogiannis et al. 2001 ). However, expression of ventral markers such as Alx4 , as well as Alx3 and Msx2 , appears to be unaffected at E11.5 in de H / de H embryos (data not shown). Differences and Similarities to Dorsoventral Limb Patterning Loss of Tbx15 also affects regional distribution of hair color in the limbs, with areas that would normally produce black hair giving rise to yellow hair instead. However, neither normal patterns of pigment-type synthesis in the limb nor their disruption in de H / de H mice correspond to obvious developmental compartments. Furthermore, losses of function for En1 or Wnt7a , which cause a partial transformation of the distal limb from dorsum to ventrum ( Loomis et al. 1996 ) or ventrum to dorsum ( Parr and McMahon 1995 ), respectively, have no effect on regional patterns of Agouti expression or distribution of hair-color regions (Y. Chen, unpublished data). (Ectopic pigmentation of the ventral footpads that develops in En1 mutant mice is unrelated to pigment-type synthesis and instead likely reflects a requirement for En1 , independent of Wnt7a , to repress migration or proliferation (or both) of pigment cells in ventral epidermis [ Cygan et al. 1997 ; Loomis et al. 1998 ].) These considerations notwithstanding, control of dorsoventral trunk pattern by Tbx15 shares certain features with control of dorsoventral limb patterning by Lmx1b , a LIM domain transcription factor that acts downstream of Wnt7a and En1 ( Riddle et al. 1995 ; Vogel et al. 1995 ; Cygan et al. 1997 ; Logan et al. 1997 ; Loomis et al. 1998 ; Chen and Johnson 2002 ). Both Tbx15 and Lmx1b act autonomously in mesenchymal cells to promote a dorsal identity, yet have expression domains that do not correspond to cell lineage compartments in the flank ( Tbx15 ) or the limb ( Lmx1b ) ( Altabef et al. 1997 ; Michaud et al. 1997 ). In the case of Lmx1b , its expression in the distal limb depends on Wnt7a produced in the overlying dorsal ectoderm ( Riddle et al. 1995 ; Cygan et al. 1997 ; Loomis et al. 1998 ). Wnt7a , in turn, is restricted to dorsal ectoderm by En1 in the ventral ectoderm ( Loomis et al. 1996 ; Cygan et al. 1997 ; Logan et al. 1997 ), whose expression marks a lineage boundary coincident with the dorsoventral midline of the apical ectodermal ridge ( Altabef et al. 1997 ; Michaud et al. 1997 ; Kimmel et al. 2000 ). As described above, En1 or Wnt7a mutations have not been reported to affect patterns of hair-color distribution (C. Loomis, personal communication; Parr and McMahon 1995 ; Loomis et al. 1996 ). However, the essential theme that ectodermal lineage compartments control the fate of underlying mesenchyme in developing limbs may apply to the trunk as well as the limb. The mammary glands also develop at a stereotyped dorsoventral position and depend on epithelial–mesenchymal interactions. However, the number and apparent position of the mammary glands are normal in de H / de H animals, indicating the existence of additional mechanisms that control dorsoventral patterning in the trunk as well as in the limbs. These ideas are summarized in the model shown in Figure 9 B. We speculate that a diffusible signal from dorsal trunk ectoderm, at or prior to E11.5, promotes expression of Tbx15 in dorsal trunk mesenchyme, which then establishes dorsal positional identity of those cells as manifested by differences in Agouti expression, pigment-cell development, and hair growth. Because the ventral limit of Tbx15 expression corresponds to the dorsal limit of En1 expression and because the normal position of the pigmentation boundary lies approximately in register with the limb-bud outgrowths, we depict the position of a putative dorsoventral boundary in trunk ectoderm as coincident with the limb dorsoventral boundary. This model is consistent with studies in the chick, where distinct dorsal and ventral lineage compartments exist for ectoderm in both the limb ( Altabef et al. 1997 , 2000 ; Michaud et al. 1997 ; Kimmel et al. 2000 ) and interlimb regions ( Altabef et al. 1997 , 2000 ), but not for limb mesoderm ( Altabef et al. 1997 ; Michaud et al. 1997 ). In fact, the same mechanism that determines dorsoventral position of the limbs and the apical ectodermal ridge may also act on expression of Tbx15 in the trunk, since ectopic limbs induced in the interlimb region by application of FGF beads develop along a single line that is coincident with normal limb buds (and the future pigmentation boundary) ( Cohn et al. 1995 ; Crossley et al. 1996 ; Vogel et al. 1996 ; Altabef et al. 1997 , 2000 ). Our model predicts that ectopic expression of Tbx15 in ventral mesenchyme should give rise to a dorsalized pigmentation phenotype and could be tested with gain-of-function approaches. However, Tbx15 expression is very dynamic and is restricted to dorsal mesoderm only from E11.5 to E13.5. It is possible that Tbx15 influences skin patterning in a very narrow window of development; alternatively, establishment of dorsal identity by Tbx15 may require another as-yet-unidentified factor that is only present in the mesenchyme underlying dorsal ectoderm. Pigmentation Patterns and Tbx15 in Other Mammals The lateral somitic frontier, defined as the lineage boundary between somite-derived versus lateral plate-derived mesoderm, is established during somitogenesis early in development ( Mauger 1972 ; Christ et al. 1983 ; Olivera-Martinez et al. 2000 ; Nowicki et al. 2003 ), but remains distinct in postnatal animals despite the potential for extensive cell mixing (see Figure 8 ). However, our transplantation and fate-mapping studies demonstrate that the lateral somitic frontier lies dorsal to the pigmentation boundary and does not obviously correlate with a difference in skin morphology. An additional dorsoventral domain that is not externally apparent has emerged from studies of Msx1 , whose expression marks a subgroup of somite-derived mesenchymal cells that contribute to dermis in a narrow stripe along the paraspinal region ( Houzelstein et al. 2000 ). Thus, there exist at least three distinct boundaries in postnatal mammalian skin that are parallel to the sagittal plane, marked by differences in pigment-type synthesis, differences in cell lineage, and differences in expression of Msx1 . In rodents, only the pigmentation boundary is evident externally, but many mammals have more complicated patterns of hair type, length, and/or color that vary along the dorsoventral axis. Raccoons, squirrels, skunks, and many different ungulates exhibit lateral stripes whose developmental origins have not been investigated, but may correspond to the lateral somitic frontier, the paraspinal Msx1 compartment, or an interaction between these domains. The effect of Tbx15 on pigmentation in laboratory mice is reminiscent of coat-color patterns in both selected and natural populations of other mammals. Saddle markings are common in some dog breeds, such as German shepherds, and in certain populations of Peromyscus polionotus , in which a dorsal extension of ventral depigmentation provides an adaptive advantage to subspecies that live on white sand reefs ( Blair 1951 ; Kaufman 1974 ; Belk and Smith 1996 ). Neither German shepherds nor deer mice have craniofacial characteristics similar to the de H mutation, but the pigmentation patterns in these animals could represent alterations in the regulation or action of Tbx15 activity. From the opposite perspective, the effects of Tbx15 on coat color are only apparent in certain genetic backgrounds and may not be evident at all in mammals that lack dorsoventral pigmentation patterns. Studying the sequence and expression of Tbx15 in other vertebrates may provide additional insight into patterns that affect the skeleton as well as the pigmentary system. Materials and Methods Mice All mice were obtained originally from The Jackson Laboratory (Bar Harbor, Maine, United States), except the BA strain (Stanford Veterinary Services Center, Stanford, California, United States), Hoxb6-Cre transgenic mice (kindly provided by M. Kuehn of the National Institutes of Health, Bethesda, Maryland, United States), mice carrying the R26R lacZ reporter allele (kindly provided by P. Soriano, Fred Hutchinson Cancer Research Center, Seattle, Washington, United States), and C57BL/6J (B6) a e /a e mice (kindly provided by L. Siracusa, Jefferson Medical College, Philadelphia, Pennsylvania, United States). The de H mutation arose in the 1960s in Harwell, probably on the BN strain background (C. Beechey, personal communication). We obtained de H on a B6/EiC3H background, introduced the a t allele from the BTBR strain, and have maintained the line as a mixed de H /+ × de H /+ intercross stock with periodic outcrossing to BTBR or B6. For timed matings, the morning of the plug was considered E0.5. Postnatally, the day of birth was considered to be P0.5. Phenotypic analysis For measurements of hair length and color, the entire interlimb region of skin was first dissected with a single incision at the dorsal midline and preserved with powdered sodium bicarbonate. Slices 2–2.5 mm in width were then prepared parallel to the dorsoventral axis, hair length boundaries determined from electronic images with Adobe Photoshop (San Jose, California, United States), and measurements obtained using ImageJ (National Institutes of Health). This approach samples awls and auchenes, because they are much thicker and therefore visually more predominant than zigzag underhairs. To assess dorsoventral variation in hair-type distribution, several hundred hairs were plucked from the middorsum or midventrum of 8-wk-old male BA strain animals, then sorted and categorized using a dissection microscope. No attempt was made to distinguish between awls and auchenes. For skin histology, 12 μm sections from paraffin-embedded tissue were stained with hematoxylin and eosin. For DOPA staining, the dermis and epidermis were split after 3 h of incubation in 2 M sodium bromide at 37°C (this preparation causes most hair follicles to remain with the dermis), individually fixed for 1 h, then rinsed and stained with 0.1% L-DOPA (Sigma, St. Louis, Missouri, United States), 0.1 M sodium phosphate buffer (pH 6.8) for 5 h at 37°C in the dark, changing the staining solution after 1 h. The samples were then fixed overnight, dehydrated, and mounted. This method is sufficient to stain interfollicular melanocytes without creating a high background. The fixative used was always 4% paraformaldehyde. Positional cloning A high-resolution map for de H was generated from an intersubspecific intercross between de H / de H and CAST/Ei mice. We initially localized de H to a 1 cM interval between D3Mit233 and D3Mit11. F 2 animals carrying recombinant chromosomes between these markers whose genotype at de was indeterminate ( de H /+ or +/+) were progeny-tested by crossing to de H / de H animals. Further genetic mapping established a minimal region of 0.1 cM between D3Mit213 and 16.MMHAP32FLF1; these markers were used to initiate construction of a physical map with BAC genomic clones (Research Genetics, Huntsville, Alabama, United States, and Genome Systems, St. Louis, Missouri, United States). End sequence from those BACs was used to develop SSCP markers M1 to M3, as depicted in Figure 3 , and to establish a minimal physical interval of 1.4 Mb. Primer pairs used were TTCCCTCCAATAAGTTCTGGGTACC and AAGCTTGCTGCTCTGGATTCCATTTGTAG for M1, CCTTCATTTTTTTTTCAAGTAAAA and AAGCTTGGCTTAGTCCCAGTGGC for M2, CCTCCAGGAAGATCTACTAGGCAC and ATGGAAAAAAAAAAGTAAGATTGAAAG for M3, and TGGTTATCGATCTGTGGACCATTC and AAGTGAGAGAGCAGGATGGACCAC for M4 (the M4 marker represents STS 16.MMHAP32FLF1). Genomic sequence and annotations were obtained from the UCSC Genome Browser February 2003 assembly version mm3 ( http://genome.ucsc.edu ); the 1.4 Mb interval between M1 and M4 contains eight genes: four hydroxysteroid dehydrogenase isomerases, Hsd3b3, Hsd3b2 , Hsd3b6 , and Hsd3b1 ; an hydroacid oxidase, Hao3 ; a tryptophanyl-tRNA synthetase, Wars2 ; a T-box gene, Tbx15 ; and a novel gene, 4931427F14Rik . In the genome sequence, M1 primers correspond to AGGCCTCCAATAAGTTCTGGGTACC and AAGCTTGCTCTCTGGATTCCATTTGTAG, the M2 reverse primer corresponds to AAGCTTGGCTTTAGTCCCAGTGGGC, and the M3 primers correspond to CCTCCAGGAAGAATCTACTAGGCAC and AATGAAAAAAAAAAAAGTAAGATTGAAAG. Minor differences among the sequences of the primers we obtained from the BAC ends and the public genome sequence may represent strain differences or sequencing errors on the BAC DNA. A multiplex genotyping assay was developed to genotype for the de H deletion using primers GGAGCAGATCCAATTGCTTT, TCCATAGCCCATCTTCACAA, and CATGTCCACTTCTGCTTCCA. This PCR assay produces a 392 bp product from the de H chromosome and a 595 bp product from the nonmutant chromosome. Gene targeting A targeted allele of Tbx15 was constructed using the same approach described in Russ et al. (2000) . In brief, an IRES-LacZ-neo cassette with 5′ and 3′ homology arms of 3.5 kb and 1.8 kb was inserted into a unique BamHI site that lies 479 nucleotides downstream of the transcriptional initiation site (relative to the mRNA sequence) in exon 3. Positive ES clones were injected into B6 blastocysts, and chimeric founders crossed to either B6 mice or to de H /+ animals. In situ hybridization In situ hybridization was carried out on 12-μm paraffin sections using digoxigenin-labeled RNA probes (Roche Diagnostics, Indianapolis, Indiana, United States) according to standard protocols ( Wilkinson and Nieto 1993 ). Embryos and postnatal skin samples were obtained from intercrosses of de H / + mice. Embryos E13.5 or younger were fixed for 24 h; those older than E13.5 and postnatal skin were fixed for 36–48 h prior to embedding. The Tbx15 probe was generated by RT–PCR using primers GGCGGCTAAAATGAGTGAAC and TGCCTGCTTTGGTGATGAT (corresponds to exons 1 and 2), and the En1 probe was generated by PCR from genomic DNA using primers ACGCACCAGGAAGCTAAAGA and AGCAACGAAAACGAAACTGG (located in the last exon). The Agouti probe corresponds to the protein-coding sequence. Embryonic skin transplantation (BTBR- a t / a t × B6- a / a )F 1 embryos at E12.5 were dissected in sterile Tyrode's solution, and embryonic skin was divided into dorsal, flank, and ventral pieces, each 1–2 mm 2 in size, as shown in Figure 7 . Skin fragments were grafted to the testes of congenic animals as follows. After anesthetization with 2.5% Avertin, a 1.5-cm incision in the skin and body wall was made at a point level with the top of the limbs. The fat pads were pulled out and laid on the outside of the body, exposing the testes. Forceps were used to introduce a small hole in the testis capsule through which a piece of dissected embryonic skin was inserted, the testes were then replaced into the abdominal cavity, and the wound was closed in both the body wall and the skin. After 21 days, mice that received grafts were sacrificed and the resulting hair was dissected from the testes and examined. Fate-mapping the lateral somitic frontier The Hoxb6-Cre transgene described by Kuehn and colleagues ( Lowe et al. 2000 ) is expressed in the lateral plate but not the somitic mesoderm of the trunk, beginning at E9.5. Animals doubly heterozygous for this transgene and the R26R reporter gene were used as a source of whole skin at P1.5 or P4.5. Skin sections parallel to the dorsoventral axis were prepared with a single incision along the ventral midline and stained for β-galactosidase activity using standard protocols at room temperature. The P1.5 sample was stained overnight and the P4.5 samples were stained for 5.5 h. Similar nonstained skin sections were prepared from animals carrying the a t allele. Images of the different skin fragments were aligned and scaled, and the relative position of the somite–lateral plate and the pigmentation boundaries were measured using ImageJ. Supporting Information The GenBank ( http://www.ncbi.nlm.nih.gov/Genbank/index.html ) accession numbers discussed in this paper are for 4931427F14Rik (AK016477), Agouti gene (L06451), Alx3 (U96109), Alx4 (AF001465), En1 (L12703), M6pr-ps (X64069), Tbx14 (AF013282), Tbx15 (AF041822), Tbx18 (AF306666), and Tbx22 (NM_145224). The OMIM ( http://www.ncbi.nlm.nih.gov/omim/ ) accession numbers discussed in this paper are for acromegaloid facial appearance (MIM 102150), frontofacionasal syndrome (MIM 229400) and human syndrome X-linked cleft palate and ankyloglossia (MIM 303400).	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC314463.xml
544190	Method for determination of (-102C>T) single nucleotide polymorphism in the human manganese superoxide dismutase promoter	Background Manganese superoxide dismutase (MnSOD) plays a critical role in the detoxification of mitochondrial reactive oxygen species constituting a major cellular defense mechanism against agents that induce oxidative stress. The MnSOD promoter contains an activator protein-2 (AP-2) binding site that modifies transcription of MnSOD . Mutations have been identified in the proximal region of the promoter in human tumor cell lines. One of these mutations (-102C>T) has been shown to change the binding pattern of AP-2 leading to a reduction in transcriptional activity. The aim of our study was to develop a method to identify and determine the frequency of this (-102C>T) polymorphism in human tissues. Results A new TaqMan allelic discrimination genotype method was successfully applied to genomic DNA samples derived from blood, buccal swabs, snap frozen tissue and paraffin blocks. The polymorphism was shown to be in Hardy-Weinberg Equilibrium in an evaluation of 130 Caucasians from Warsaw, Poland: 44 (33.8%) were heterozygous and 6 (4.6%) were homozygous for -102T. Conclusion This report represents the first description of the MnSOD -102C>T polymorphism in human subjects by a novel Taqman allelic discrimination assay. This method should enable molecular epidemiological studies to evaluate possible associations of this polymorphism with malignancies and other diseases related to reactive oxygen species.	Background Antioxidant enzymes such as superoxide dismutase (SOD) protect cells from oxidative stress. Generation of reactive oxygen species (ROS) has been implicated in the etiology of a diversity of human diseases, including cancer[ 1 ], aging[ 2 ], atherosclerosis[ 3 ] and neurodegenerative diseases[ 4 , 5 ]. Superoxide dismutase catalyzes the dismutation of superoxide radical (O 2 - ) to H 2 O 2 and O 2 . Three distinct types of SODs have been identified in human cells: 1) a homodimeric cytosolic CuZnSOD [ 6 ], 2) an extracellular homotetrameric glycosylated SOD [ 7 ], and 3) a mitochondrial matrix homotetrameric manganese superoxide dismutase (MnSOD) [ 8 ]. Numerous reports indicate a relative deficiency of superoxide dismutase catalytic activity, including mitochondrial MnSOD, in many types of solid tumors [ 9 , 10 ]. Interest in this relative deficiency of SOD activity has been greatly increased by observations that over-expression of SOD in tumor cells will suppress cell division in culture and tumor growth in vivo[ 11 ]. In addition recent reports have suggested a possible association between decreased SOD activity and malignant phenotype[ 12 ]. While the precise reasons for this relationship between tumor cell growth rate and intracellular SOD activity are not known, these findings support the general idea that decreased expression of SOD may promote tumor growth. In fact, as a result of these and other observations, MnSOD is considered a tumor suppressor gene[ 1 ]. Further evaluation of MnSOD suggests that it is critically important in maintenance of mitochondrial function. Mice with deficiency of this enzyme exhibit progressive cardiomyopathy, neurodegeneration and perinatal death[ 13 ]. These studies went on to confirm that transgenic mice that express human MnSOD in the mitochondria are protected from environmental oxygen-induced lung injury [ 14 ] and adriamycin-induced cardiac toxicity[ 15 ]. In contrast, disruption of the other two SODs yielded viable mice which were normal in non-stressful conditions [ 16 ]. Thus the mitochondrial MnSOD represents a major cellular defense against oxidative stress. Genetic polymorphism in the MnSOD mitochondrial targeting sequence has been associated with risks to various diseases including breast cancer[ 17 , 18 ], lung cancer[ 19 ], cardiomyopathy[ 20 ] and Parkinson's disease[ 21 ]. A reduction of MnSOD activity has been shown to exist in many types of human cancer cells when compared to normal cells [ 22 ]. A recent report has also demonstrated the possible association between decreased SOD activity and malignant phenotype[ 12 ]. A recent report demonstrated a new mutation L60F, in exon 3 of the mature protein in the Jurkat human T-cell leukemia-derived cell line that reduced MnSOD [ 12 ]. Thus, it appears that reduced levels of MnSOD activity in human cancer cells can be associated with coding region mutations that alter protein sequence as well as promoter region mutations that alter gene expression [ 23 ]. The human MnSOD gene is localized to chromosome 6 (6q25). The MnSOD promoter region is characterized by a lack of TATA or CAAT boxes but the presence of a GC rich region containing multiple SP-1 and AP-2 binding sites [ 24 ]. Further work identified one cause for the reduced expression of MnSOD in some human tumor lines; the occurrence of three heterozygous mutations in the upstream promoter region of this gene [ 25 ]. One of these mutations in the MnSOD promoter sequence (MnSOD -102C>T) has been shown to change the binding pattern of AP-2 leading to a reduction in transcriptional activity. However the presence of this polymorphism has not been reported in human tissue. In this study we developed a TaqMan allelic discrimination assay to reliably genotype DNA from many tissues (i.e. blood, buccal swabs, paraffin blocks, and snap frozen tissue) for the -102C>T polymorphism in the MnSOD promoter. Results We confirmed the presence of the -102C>T single nucleotide polymorphism in human subjects and submitted the sequence variant to Genbank[ 26 ]. The genotyping success rate with this technique in the Polish Caucasian population was 85%. An evaluation of 130 DNA samples successfully genotyped from Polish Caucasians not known to have cancer, demonstrated 80 (61.5%) were homozygous (-102C), 44 (33.8%) were heterozygous (-102CT) and 6 (4.6%) were homozygous (-102TT). This distribution is consistent with the Hardy-Weinberg Law. The success rate with this technique in an additional American control population was blood (95%), buccal swabs (90%), snap frozen tissue (80%) and paraffin-embedded samples (75%). The success rates were influenced by DNA quality, DNA extraction technique, and the ability to acquire enough DNA from the buccal swab. Discussion Reactive oxygen species in the form of superoxide radicals, hydrogen peroxide, and hydroxy-radicals are formed during incomplete reduction of molecular oxygen during normal respirations. The production of reactive oxygen species remains relatively stable during normal physiologic respirations. A significant increase in the production of reactive oxygen species such as superoxide radicals can be greatly increased as a result of metabolic disorders or more commonly from exposure to toxins such as cigarette smoke, well-cooked meat, urban residency, and excessive alcohol consumption. Under normal physiologic conditions, superoxide radicals are detoxified by superoxide dismutase. Among the three SODs, MnSOD has been demonstrated to be the only form that has been essential for survival of aerobic life [ 27 ]. Inactivation of the MnSOD gene in E. coli significantly increased mutation frequency and cell death when bacteria were grown under aerobic conditions [ 28 ]. This has been further demonstrated in the evaluation in mammals in which the inactivation of MnSOD gene has led to detrimental effects. Polymorphisms of the human MnSOD gene have been found in the promoter region, the sequence coding for mature protein, and the mitochondrial targeting sequence. Initial evaluation of the five prime flanking regions from human tumor cell lines indicated that there were no major additions or deletions in the five prime flanking regions of the human MnSOD gene [ 29 ]. However, there were three mutations that were identified in these tumor cell lines: a C to a T at the – 102 position; a C to a G at the – 38 and an insertion of an A in 11 straight Gs at the – 93 position in relation to the transcription initiation site. The significance of these mutations was felt to be important because this region includes multiple binding sites for SP-1 proteins as well as AP-2 binding sites. Further evaluation of these mutations identified that the C to T change at the – 102 position effected the overall transcription of the MnSOD gene [ 30 ]. This change in transcription may result from an effect on the AP-2 binding site. Although the -102 C to T mutation was reported in human tumor cell lines [ 25 ], no evaluation has been assessed in human subjects. Evaluation of the -102C>T polymorphism is complicated by difficulty in PCR because of the excessive GC rich region in which this polymorphism exists. This location, upstream from the transcription start site was extremely difficult to identify through multiple PCR-restriction fragment length polymorphism (RFLP) assays, which failed to adequately digest at this polymorphism site, and led to multiple false negative results. We found only the highest quality DNA (i.e. blood) was able to be evaluated using a PCR RFLP assay with only 50% genotyping success. This failure to accurately reproduce the PCR-RFLP assay [ 31 ], led us to the development of this TaqMan allelic discrimination assay. The TaqMan allelic discrimination assay provided results that were confirmed by automated DNA sequencing and blind repeat genotyping. Although we did not test it use on DNA from multiple tissues from the same individual, it was successful for DNA samples derived from buccal swabs and paraffin blocks. It has significant advantages over RFLP analysis, allele-specific amplification, allele-specific hybridization, and oligo-nucleotide ligation assay techniques. The reasons for this advantage come from the reduction in labor intensive work up, the lack of need for special handling of radioactive probes, and the ability to modify this technique to evaluate multiple polymorphisms in this gene. In addition as more significant polymorphisms within the MnSOD gene are discovered, this technique will facilitate detection within the MnSOD gene. The limitations of this technique ultimately come from the quality of DNA that is available and the significant initial expense that is required for a TaqMan assay instrumentation. Conclusions This report represents the first description of the MnSOD -102C>T polymorphism in human subjects by a novel Taqman allelic discrimination assay. This method should enable molecular epidemiological studies to evaluate possible associations of this polymorphism with malignancies and other diseases related to reactive oxygen species. Methods DNA sources Most DNA samples (130) were isolated from buffy coats of Caucasian controls derived from a population-based case-control study of stomach cancer carried out in Warsaw, Poland as previously described [ 32 ]. To test the utility of the method to genotype DNA from various tissue sources, peripheral blood (20 samples), buccal swabs (40 samples), paraffin blocks (40 samples), and snap frozen tissues (15 samples) were collected from research subjects in the USA (Louisville, Kentucky). DNA extraction from paraffin sections was performed after tissue sections (10 sections, 10 μm thick) were cut from paraffin blocks. Samples were removed from paraffin through a sequential extraction with histaclear, 100% ethanol and acetone, and dried under vacuum. The pellet was incubated overnight with proteinase K (200 μg/ml in 50 mM Tris-HCL pH 8.5, 1 mM EDTA and 0.5% Tween-20) at 55°C. After heating at 100°C for 10 min, digestion was sequentially extracted with phenol, phenol/chloroform and chloroform. DNA was precipitated with the addition of 3X volume 95% ethanol. Primer design SNP-specific polymerase chain reaction (PCR) primers and fluorogenic probes (Table 1 ) were designed using Primer Express (Version 1.5; Applied Biosystems, Foster City, CA). This technique has been utilized extensively in genotyping other candidate genes with multiple single nucleotide polymorphisms[ 33 , 34 ]. The fluorogenic probes were labeled with a reporter dye (either FAM or VIC) and are specific for one of the two possible bases (-102 C or T) in the MnSOD promoter region. A MGB quencher probe was utilized on the 3' end by a linker arm. TaqMan Universal PCR Master Mix (Applied Biosystems) was used to prepare the PCR. The 2X mix was optimized for TaqMan reactions and contained AmpliTaq-Gold DNA polymerase, AmpErase, UNG, dNTPs with UTP and a Passive Reference. Primers, probes and genomic DNA were added to final concentrations of 300 nM, 100 nM, and 0.5–2.5 ng/μl respectively. Controls (no DNA template) were run to ensure there was no amplification of contaminating DNA. Reference control DNA was also utilized to verify the polymorphisms identified. The amplification reactions were carried out in an ABI Prism 7700 Sequence Detection System (Applied Biosystems) with two initial hold steps (50°C for 2 min, followed by 95°C for 10 min) and 50 cycles of a two step PCR (95°C for 15 sec, 60°C for 1 min). The fluorescence intensity of each sample was measured at each temperature change to monitor amplification of the 278 base pair MnSOD promoter region. The -102 nucleotide was determined by the fluorescence ratio of the two SNP-specific fluorogenic probes. The fluorescence signal increases when the probe with the exact sequence match binds to the single stranded template DNA and is digested by the 5'-3' exonuclease activity of AmpliTaq-Gold DNA polymerase (Applied Biosystems). Digestion of the probe releases the fluorescent reporter dye (either FAM or VIC) from the quencher dye. As shown in figure 1 , the method readily distinguishes between at C or T at -102 in the MnSOD promoter region. Table 1 Primers and Fluorgenic Probes for -102C>T MnSOD Allelic Discrimination Primers -102-Forward Primer (-252 to -234) 5'-gcagacaggcagcgaggtt-3' -102-Reverse Primer (35 to 19) [287 bp] 5'-ctgaagccgctgccgaa-3' Probes -102C-Taqman Probe (-97 to -107) fam-ccgcgggcccc -102T-Taqman Probe (-97 to -107) vic-ccgcgagcccc Figure 1 Fluorescence ratios of FAM-labeled/VIC-labeled fluorogenic probes specific for -102C>T polymorphism in MnSOD . Each bar represents mean standard error for determinations in DNA from 3 human subjects. Open bar represents DNA samples homozygous for -102C. Solid bars represent DNA samples homozygous for -102T. Crossed bars represent DNA samples heterozygous for the SNP. The fluorescence ratios differed significantly (p < 0.05) among homozygous and heterozygous genotypes. Twenty samples with genotypes C/T (4 samples), T/T (3 samples), and C/C (13 samples), some of which were derived from paraffin-embedded tissues, were all confirmed by automated DNA sequencing. These sequence-confirmed samples served as reference standards for the remaining samples. In addition, 10% of the samples were genotyped blind a second time with identical results obtained. Author contributions RM: Participated in design of study and manuscript preparation KH: Participated in genotyping samples MD: Participated in design of methods of assay QL: Participated in statistical analysis BM: Participated in design of methods of assay JL: Participated in sample collection NR: Contributed to the study design and the analysis and interpretation of the data DH: Participated in design of study and manuscript preparation	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544190.xml
521085	RNA-binding proteins to assess gene expression states of co-cultivated cells in response to tumor cells	Background Tumors and complex tissues consist of mixtures of communicating cells that differ significantly in their gene expression status. In order to understand how different cell types influence one another's gene expression, it will be necessary to monitor the mRNA profiles of each cell type independently and to dissect the mechanisms that regulate their gene expression outcomes. Results In order to approach these questions, we have used RNA-binding proteins such as ELAV/Hu, poly (A) binding protein (PABP) and cap-binding protein (eIF-4E) as reporters of gene expression. Here we demonstrate that the epitope-tagged RNA binding protein, PABP, expressed separately in tumor cells and endothelial cells can be used to discriminate their respective mRNA targets from mixtures of these cells without significant mRNA reassortment or exchange. Moreover, using this approach we identify a set of endothelial genes that respond to the presence of co-cultured breast tumor cells. Conclusion RNA-binding proteins can be used as reporters to elucidate components of operational mRNA networks and operons involved in regulating cell-type specific gene expression in tissues and tumors.	Background Many recent studies have described the use of microarrays to identify genes expressed in different types of cancers (reviewed in [ 1 , 2 ]. Most of these transcriptomic studies monitor the steady state levels of expressed mRNAs in order to derive the "molecular signatures" of tumors [ 2 ]. However, the gene expression profile of a whole tumor corresponds to the combined profiles of the different cell types contained within it (e.g. endothelial cells, T-cells, cancer cells, stromal cells, etc.). Moreover, the multiple cell types present in a tumor or organ are interdependent and exchange biochemical signals as a means of cell-cell communication [ 3 ]. An important example of cell-cell communication is evident in angiogenesis, the mechanism by which new blood vessels vascularize tumors and other organs (reviewed in [ 4 ]). Monitoring the dynamics of gene expression in each cell type of a tumor during angiogenesis will advance understanding of tumorigenesis as well as organogenesis, in general. Methods have been devised to generate mRNA samples from specific types of tumor cells. These include microdissection, laser capture (reviewed in [ 4 - 6 ], and cell sorting based on specific membrane markers [ 7 ]. Here we demonstrate that RNA-binding proteins can be used to isolate mRNA populations representing total cell mRNA from specific types of cells, as well as discrete mRNA subpopulations that represent post-transcriptionally regulated subsets of mRNAs that encode functionally related proteins. We propose that these represent genes whose regulation is important for tumor growth and maintenance. RNA binding proteins play a key role in post-transcriptional regulation, participating in splicing, mRNA transport and localization, mRNA stability and translation (for overview see ref. [ 8 ]). Our lab has devised biochemical and immunological approaches to gene expression profiling by using RNA-binding proteins as reporters of discrete mRNA subsets in metazoan cells [ 8 - 10 ]. For example, we identified subpopulations of mRNAs that are associated with ELAV/Hu RNA-binding proteins that are expressed in specific cell types [ 10 ]. While we and other labs have demonstrated the isolation of mRNA subsets that are potentially co-regulated using RNA binding proteins as reporters of gene expression, methods have not been described that provide information about coordinated posttranscriptional regulation within specific types of cells during tumorigenesis and development. Moreover, because many different mRNA-binding proteins in specific cell types are known to interact with unique subpopulations of mRNAs encoding functionally related proteins [ 9 - 15 ] they can be informative of the dynamic effects of cells on one another. Therefore, it will be necessary to assess changes in gene expression that occur when cells such as tumor cells and endothelial cells interact in order to understand growth control and critical processes such as angiogenesis. In this study, we define a model system for using poly (A) binding protein (PABP) to recover mRNAs from specific cell-types in mixed cell cultures. Using this approach, we were able to determine how the gene expression profiles of endothelial cells change in response to the presence of breast cancer cells. Among the advantages of this approach are: a) no manipulations or treatments are required prior to the preparation of cell extracts, b) the recovered mRNA population can be identified directly using genomic methods, and c) RNA binding proteins can be engineered for expression in different cell types using various molecular tags in order to discriminate cell-specific mRNA populations. These studies provide a methodological basis for creating mouse models in which different types of cells within a tumor express RNA binding proteins to reveal unique populations of posttranscriptionally regulated mRNAs. Results and Discussion The goals of these experiments are to validate procedures for the isolation and characterization of discrete mRNA sub-populations associated with RNA binding proteins expressed in specific cell types within a tumor or organ in order to assess the responses of cells to their surroundings. Earlier studies have shown that mRNA subpopulations in single cell types reflect the functions of the RNA binding proteins with which they associate and can provide key information about post-transcriptional regulatory mechanisms of gene expression [ 8 - 11 , 13 , 15 - 18 ]. In model organisms, such information can be obtained by expressing epitope-tagged RNA binding proteins using tissue-specific promoters [ 19 ] or by using virus-specific receptors (M.D.B., L.O.F.P. and J.D.K. unpublished). In this study we demonstrate the feasibility of this approach by using two different cell types in culture that each express specific RNA binding proteins as reporters of gene expression profiles. Comparison of total mRNA of PY4.1 endothelial cells with their PABP-associated mRNA patterns Microarray analysis was used to compare the gene expression profiles obtained using total RNA and PABP-associated RNA of PY 4.1 endothelial cells (Figure 1 and supplementary data). These patterns were highly reproducible and consistent. Very few if any qualitative differences were observed when comparing these mRNA patterns indicating that the same set of expressed genes was detected in both preparations. However, quantitative differences were observed between some of the PABP-associated mRNA levels and those of the total mRNA population (Figure 1 and supplementary data). Approximately 19% of the expressed mRNAs varied more than two fold. In the case of the total mRNA profile (transcriptome), the signal intensity reflects the steady-state level of each mRNA, while the signal intensity of the PABP associated messages likely correlates with their translational activity [ 20 ]. PABP is an essential RNA binding protein that is highly conserved among eukaryotic organisms. It mediates interactions between polyadenylated mRNA sequences at the 3' ends of mRNAs and the eIF-4G protein [ 21 ]. Interactions of eIF-4G with the cap-binding protein, eIF-4E, are believed to circularize the mRNA and to prepare it for association with ribosomes. Many studies have shown that PABP is involved in activating the stability and translation of mRNAs to which it is bound (reviewed in [ 22 , 23 ], by protecting the poly (A) tail from exonuclease attack [ 24 ], preventing mRNA decapping [ 25 ], by promoting mRNA maturation [ 26 ] and by stimulating the initiation of translation [ 20 ]. Most studies are in general agreement that PABP functions as a translational activator by facilitating the assembly of mRNAs and ribosomes. On the average, a single PABP is expected to recognize approximately fifteen adenylate residues, suggesting that approximately ten molecules of PABP are bound to the average mRNA that has a poly (A) stretch of 150–200 in length [ 27 ]. Figure 1 Comparison of the gene expression profiles of total RNA and PABP-associated mRNA populations. Total RNA from PY4.1 murine endothelial cells and mRNA immunoprecipitated from cell extracts using anti-PABP serum were radiolabeled and hybridized on 1.2 mouse Atlas arrays (CLONTECH). The image overlay comparing total RNA with PABP-associated gene expression profiles was derived using the Atlas software with a global normalization showing quantitative, but not qualitative differences. Several reports indicate that substantial differences can be found when comparing the steady state levels of mRNAs (transcriptome) with proteins (proteome) in the same cell population [ 28 , 29 ]. The accumulated levels of some proteins and their corresponding mRNAs can vary by as much as 30-fold [ 28 - 30 ]. The differential between steady state levels of mRNA and protein are expected to be more dramatic under conditions in which post-transcriptional regulation plays a major role. For example, following T cell-activation or during neuronal differentiation, translational control is thought to affect a significant proportion of the proteomic outcome [ 31 , 32 ]. It is possible that gene expression profiles obtained by immunoprecipitating mRNA-PABP complexes may reflect the functional state of protein production from these mRNAs [ 10 ]. For the purposes of this study, PABP is used as a functionally relevant RNA-binding protein with which to compare changes in bound mRNAs across gene expression profiles. Expression of tagged PABP does not interfere with cell growth In order to compare mRNA profiles from mixed cell populations, we prepared two different cell lines stably expressing different epitope tagged PABPs. Figure 2 outlines the experimental approach. T98G human glioma cells and PY4.1 mouse endothelial cells were co-cultured, cell extracts were prepared and antibodies against PABP, Flag-PABP and G10-PABP were used to immunoprecipitate the mRNP complexes in order to generate mRNA populations for gene expression analysis. Concerns that the epitopes represented in the tags might affect the results were addressed using PY4.1 cells stably expressing either Flag-tagged PABP or G10-tagged PABP. Cell extracts from both of the stable cell lines grown separately were prepared and immunoprecipitated with the respective antibodies. The two mRNA populations generated by this procedure were compared using an RNAse protection assay (RPA) and a microarray analysis. The results showed no significant qualitative or quantitative differences between these profiles in that 96% of the genes detected by microarray were within 1.5-fold of one another (data not shown). Figure 2 A) Experimental design for sorting cell type specific mRNA populations using RNA binding proteins in mixed cell cultures . Two cell lines from two different species (murine endothelial PY4.1 and human glioma tumor T98G) are engineered to express G10-tagged and FLAG-tagged PABP, respectively. Cell-type specific gene expression profiles are obtained from co-cultured cells or mixed cell extracts after immunoprecipitation with specific antibodies against the different tags. The RNA samples are phenol extracted, precipitated and subsequently analyzed by RPAs or microarrays. B) Western blots of cell lines expressing tagged-PABP. Immunoblots of extracts from T98G cells expressing Flag-PABP were probed with anti-Flag antibody, while extracts from PY4.1 and PY4.1 cells expressing G10-PABP were probed with anti-G10 antibody. Control blots of extracts of T98G and PY4.1 cells with anti-sera against PABP. C) Comparison between the overall levels of PABP of T98G cells and T98G cells expressing Flag-PABP . Immunoblots of extracts from T98G cells and T98G cells expressing Flag-PABP were probed with anti-sera against PABP and anti-α tubulin antibody, as a loading control. D) Comparison of the cell cycle status of T98G and T98G expressing Flag-PABP cells – Cells were arrested at G0/G1 by serum deprivation and stimulated to re-enter the cell cycle by addition of serum. At the indicated times, aliquots of cells were processed for FACS analysis to determine the population distribution in G1, S, and G2 stages of the cell cycle. No differences between of the cell cycle of T98G cells and T98G cells expressing Flag-PABP were observed. A potential complication for this type of analysis is that expression of a tagged-RNA binding protein, in this case PABP, could affect cell growth. While these cell lines appeared unaffected morphologically, the levels of PABP in cell lines expressing tagged-PABP and respective control cell lines were evaluated and compared by Western blotting. No substantial change in overall PABP expression was observed when expressing exogenous tagged PABP (Figure 2C ). This result was expected, since PABP has been shown to inhibit the translation of its own mRNA by binding to poly (A) sequences found in the 5' UTR. This fortuitous auto-regulatory mechanism is believed to keep the level of PABP constant in the cell, thereby avoiding excessive overexpression [ 33 ]. No changes in cell growth or mortality of the cell lines used in this study or in other cells lines expressing tagged-PABP were observed. Moreover, the cell cycle kinetics of T98G cells expressing Flag-PABP and cells that were subsequently stimulated by serum addition were compared to those of T98G cells containing the empty vector, pCMVneo and no substantial differences were observed using fluorescent cell sorting (Figure 2D ). We conclude that expression of neither the authentic PABP, nor the tagged-PABP has untoward effects on the growth and homeostasis of these cells. PABP does not exchange between mRNAs in cell extracts Early studies of PABP binding to mRNA indicated that a dynamic exchange or hopping of PABP from mRNA to mRNA might be an important aspect of its function [ 27 , 34 ]. For the expression profiling methods described above to be precise, it is critical to avoid post-lysis exchange (or adventitious reassortment) of PABP with mRNAs. In other words, does free mRNA in a cell extract displace the mRNA originally bound to PABP during the incubation period; or instead, does free PABP in an extract exchange by binding to available mRNAs? PABP is a good test model in this case because it has been suggested to "hop" based on in vitro studies, and it is a highly abundant RNA-binding protein. To examine these possibilities, we added increasing amounts of competitor poly (A) RNA with an average length of 550 nucleotides to lysates of mouse endothelial PY4.1 and human glioma T98G cells. After immunoprecipitation with anti-PABP serum, mRNAs were isolated from the pellets and analyzed using a highly sensitive multi-probe RNase Protection Assay (RPA) as described previously [ 10 ]. Figure 3 shows that the mRNAs originally bound to PABP were not displaced by the competing poly (A) RNA even at concentrations as high as 1000 fold excess. The inability of free poly (A) RNA to compete bound PABP off of endogenous mRNA reflects a stable interaction between PABP and endogenous mRNA. These data suggest that exchange of mRNA into PABP RNPs is not likely to distort gene expression profiles obtained by immunoprecipitating PABP, and in addition, this observation is not compatible with a previous "hopping model" for PABP [ 27 ]. Figure 3 Reassortment of mRNA and PABP was not detected in cell lysates. The potential for displacement of PABP was tested by adding increasing amounts (0–1500 μg) of pure competitor poly (A) to 400 μl of cell lysates of PY4.1 and T98G cells prior to incubation with antibody-coated (anti-PABP) beads. Following immunoprecipitation, mRNAs were isolated from the pellets and analyzed using the RNase Protection Assay (RPA) mouse angiogenesis (mAngio) and human tumor suppressor (hTS1) multi-probe sets. Unprotected RNA probes were used to identify the nature of the different sized protected fragments. The experiment shows that the interaction between PABP and the endogenous mRNA targets cannot be disrupted by competing poly (A) RNA. We have also addressed the potential problem of having a pool of free PABP in an extract that could be available to bind mRNA during incubation. Sucrose gradient analysis indicated that this is very unlikely since the majority of PABP was found in heavy polysomes and associated with mRNA, while only a small percentage was found in the upper portion of the sucrose gradients (H.S. and J.D.K., unpublished data). In total, these results demonstrate that reassortment of PABP in these cell extracts was not a significant limitation to using PABP RNPs for gene expression profiling of bound mRNAs. As noted above, it has been suggested that yeast PABP uses a "hopping" mechanism in vivo by moving from RNA to RNA [ 27 , 34 ]. While this experiment is not a direct test of that hypothesis, these data are not consistent with a hopping or exchange of PABP among the mRNAs in our cell extracts, but suggest instead that PABP forms a stable RNP complex with polyadenylated transcripts. Detection of cell-specific mRNAs using epitope-tagged PABP The question of whether reassortment of PABP occurs among mRNAs in cell extracts was also examined using lysates from mixed mouse and human cells. We used two different cell lines, murine endothelial PY4.1 and human glioblastoma T98G, that express G10-PABP and Flag-PABP, respectively (Figure 2B ). T98G cells have a volume approximately 3 to 4 times larger than PY4.1 cells. These cell lines were co-cultured in an approximate 1:1 cell ratio and subsequently lysed, or in separate experiments, lysates of each were mixed at equivalent amounts of total protein prior to immunoprecipitation. Both approaches gave the same results. Using the G10 antibody to precipitate only PY4.1 mRNAs and Flag antibody to precipitate T98G mRNAs we were able to examine the separate populations using a multiprobe RPA (Figure 4 ). Thus, by mixing a human (T98G) and a murine (PY4.1) cell line we could take advantage of the species-specific RPA probe sets. Given our interest in tumor angiogenesis, we used mouse angiogenesis (mAngio) and human tumor suppressor (hTS1) RPA probe sets after they were tested for cross species hybridization. Both probe sets showed good specificity of discrimination with the exception of the L32 and GAPDH control genes as expected (Figure 4A and 4B ). In Figure 4C and 4D , it is apparent that the same expressed genes were detected whether using total RNA, or mRNA obtained by immunoprecipitation with anti-PABP, anti-Flag or anti-G10 antibodies. Extracts from mixed cell lines were immunoprecipitated and analyzed using both human and mouse probe sets (Figure 4E,4F and 4G ). Both mouse and human mRNAs were detected when anti-PABP rabbit serum was used to precipitate both endogenous and exogenous PABP, while immunoprecipitations with anti-G10 and anti-Flag antibodies enriched the mRNA population for each species with only minor background from the other species. While the discrimination obtained in these experiments was excellent, a low degree of background due to non-specific binding of mRNA to the agarose beads was consistently observed even in the absence of antibody. Figure 4 RNase Protection Assay (RPA) of mRNAs from mixed mouse and human cell lines . Species specificity of multiprobe RPAs using: A) the human tumor suppressor probe set (hTS1), and B) the mouse angiogenesis RPA probe set (mAngio), was verified using total RNA extracted from murine PY4.1 cells and human T98G cells. Unprotected RNA probes were used to identify the nature of the different sized protected fragments. C) RPA gene expression profile of PY4.1 cells expressing G10-PABP obtained with total RNA and with mRNA derived from immunoprecipitations with anti-PABP serum or anti-G10 antibody D) RPA gene expression profile of T98G cells expressing Flag-PABP obtained from total RNA and from mRNA derived from immunoprecipitations with anti-PABP serum or anti-Flag antibody. E, F and G) Mixed extracts from cells expressing T98G Flag-PABP and PY4.1 G10-PABP were immunoprecipitated with anti-PABP serum or anti-Flag or anti-G10 antibodies. The mRNA populations generated by immunoprecipitations were analyzed with RPA of both the mouse angiogenesis (mAngio) and the human tumor suppressor probe (hTS1) sets. GAPDH and L32 are controls in both probe sets and show cross species hybridization. The asterisk in B and G also indicate a band resulting from cross species hybridization. The experiments indicate that species-specific mRNA populations can be isolated and quantified by the use of distinct tagged-PABPs. Analysis of PABP-associated mRNA populations using microarrays In order to evaluate the degree of mRNA enrichment over background using a genome–wide methodology, we analyzed immunoprecipitated mRNAs from mixed cell populations on CLONTECH Atlas arrays. We first tested these arrays for cross species hybridization using total RNA and it was minor (data not shown). The mouse RNA on human Atlas arrays did not show any detectable cross hybridization signal, while the human RNA on mouse Atlas arrays showed a small percent (2–3%) of cross hybridizing mRNAs (Figure 5B , blue squares). Figure 5 Discrimination of the gene expression profiles of mixed human and mouse cell lines using microarrays. Cell extracts from T98G Flag-PABP and PY4.1 G10-PABP cells were prepared, mixed and immunoprecipitated with both anti-Flag and anti-G10 antibodies. The mRNA populations generated by both immunoprecipitations were analyzed on human and mouse 1.2 CLONTECH arrays. When anti-Flag antibodies were used, the T98G mRNA population was enriched in relation to the PY4.1 mRNA population. When anti-G10 antibodies were used, the PY4.1 mRNA population was enriched in relation to the T98G mRNA population. To identify the PABP-associated mRNAs in the T98G Flag-PABP and PY4.1 G10-PABP cells, extracts were prepared as described above, followed by immunoprecipitation with either anti-Flag or anti-G10 antibodies. The mRNA populations generated by both immunoprecipitations were analyzed on human and mouse 1.2 Atlas arrays. Cross-species hybridization was monitored and genes showing cross-species reactivity were eliminated from consideration. A comparison of Flag versus G10 PABP-associated mRNAs was performed to assess the degree of enrichment. In an average experiment for the mouse genes, 91 % (184 out of 202 detected genes) were enriched at least 4 fold in the G10 PABP population when compared to the Flag PABP population. For the human genes, 82.4 % (122 out if 148) were enriched at least 4 fold in the Flag PABP population in relation to the G10 PABP population (Figure 5 and supplementary data). Changes in gene expression induced by co-cultivation of PY4.1 endothelial cells with 4T1 breast cancer cells Having demonstrated that the approach we described using PABP can be used to efficiently recover cell type specific mRNAs from mixed cell types, we addressed the consequences of cell-cell communication and changes in gene expression that were induced in the endothelial cells by co-cultivation with the tumor cell. The goal of these experiments is to gain insight into how endothelial cells respond to the presence of cancer cells in cell culture as a first approximation of changes in gene expression that may be involved in early stages of angiogenesis. We used two murine cell lines, the PY4.1 line described above and a 4T1 breast tumor cell line that can produce tumors and spread by metastasis in nude mice. Figure 6 represents a schematic view of two branches of our experimental strategy. In the first line of inquiry, PY4.1 cells stably expressing Flag-PABP and 4T1 cells were co-cultured. The number of plated cells of each type was calculated based upon measurements of their different growth rates. After 48 hours the co-cultured cells reached confluence. Approximately 50% of the petri dish surface was covered with PY4.1 cells and 50% was covered with 4T1 cells. Cells were harvested and extracts were prepared as described above. In the second line of inquiry, the two cell types were plated separately. Cells were harvested after 48 hours of incubation, the extracts were prepared, and mixed proportionally to those used in the first experimental line of investigation. Subsequently, extracts from the co-cultured and the mixed cells were immunoprecipitated with anti-Flag antibodies and mRNAs analyzed on microarrays. The comparison between the two mRNA populations (co-culture versus mix) was used to detect changes in the gene expression profile of PY4.1 cells as a response to the presence of 4T1 cells. In a separate experiment, a comparison between PY4.1 total RNA labeled with Cy3 and Cy5 was performed to rule out dye bias (not shown). Figure 6 Effects of co-cultivated mouse tumor cells on gene expression profile of mouse endothelial cells. PY4.1 cells expressing Flag-PABP and 4T1 cells were grown either separately or together in a co-culture. Cell extracts from the co-culture and from a mixture of the monocultures were prepared. Immunoprecipitation of extracts with anti-Flag antibodies generated two distinct PY4.1 cells mRNA populations that were compared by microarray. The comparison revealed PY4.1 cellular genes that changed their gene expression profile in response to presence of 4T1 breast cancer cells. As expected, the great majority of the genes expressed in PY4.1 cells were not altered in their levels of expression while in co-culture. Of interest, a small number of genes were consistently upregulated in four independent experiments. To assess the consistency of the fold change, we plotted the p-value (from modified t-test) against the average fold change (Figure 7 ). A 'volcano plot' summarizes both the magnitude of change and the corresponding statistical significance for all genes. We sorted the candidate genes according to their p-values (Table 1 ), and the top-20 genes were identified and are listed along with comments concerning their biological function(s). [See Supplementary Data for the complete microarray analysis – ]. Several of the genes present in this population are gene expression regulators that fall into two major categories: RNA binding proteins and DNA binding proteins/transcription factors. We expected to find gene expression regulators as part of an early response to cell surface interactions or secreted factors from the other cell line. As with any biochemical cascade event, changes in the expression of global regulators as well as structural genes (such as those encoding membrane or cytoskeletal proteins) often precede downstream alterations in the expression of other important genes. Among the RNA binding proteins identified in our screen were stem-loop binding protein, a highly conserved RNA binding protein that binds a stem loop structure in the 3'UTR of histone mRNAs and is required for both processing and translation of histone messages [ 35 ], Brul4, the mouse homologue of Drosophila bruno , a translation repressor which functions at the early steps of embryogenesis [ 36 ], and quaking, described as a key gene involved in the myelination of the central nervous system and other regulatory functions [ 37 ]. It should be noted that we did not identify PY4.1 genes whose expression decreased as a response to the presence of 4T1 cells. Figure 7 "Volcano" plot of p-value versus fold change in expression level. Dashed line indicates the cutoff of the top 20 enriched genes shown in Table 1. Table 1 List of the top 20 PY4.1 genes that were upregulated in response to the presence of 4T1 tumor cells. Genes are classified according to their biological function. Gene expression regulators (GR). Genes involved in metabolism (M). Genes related to cell cycle or cell division (C). Genes encoding structural proteins (S). Other genes (O). Name, classification gbID fold P value biological function 1 -Heterogeneous nuclear ribonucleoproteinH1, (G R) NM_021510 2.6 6.40E-06 RNA binding, RNA processing and modification 2 -High mobility group box 1, (G R) NM_010439 2.7 9.78E-06 DNA binding, nitric oxide biosynthesis, inflammation mediator, cell differentiation 3 -Prothymosinalpha, (C) NM_008972 3.7 1.38E-05 cell proliferation, cell division 4 -RIKEN cDNA2610016F04 gene, (G R) AK009120 2.6 4.59E-05 putative DNA binding, transcritionfactor 5 -ATPase, H+ transporting, (M) NM_024173 1.8 4.92E-05 hydrogen-exporting, ATPaseactivity, phosphorylativemechanism 6 -RIKEN cDNA2510010F10 gene, (O) AF215660 1.8 7.45E-05 described as a carnitinedeficiency-associated gene 7 -Stem-loop binding protein, (G R) NM_009193 2.1 7.88E-05 RNA binding, histonemRNA processing 8 -Quaking, (G R) NM_021881 1.9 7.79E-05 RNA binding, participates in myelination 9 -RIKEN cDNA2410004I17 gene, (O) AK010391 2 8.26E-05 unknown 10 -Purinerich element binding protein A, (G R ) NM_008989 2.8 8.95E-05 DNA and RNA binding, association with rough endoplasmic reticulum, postnatal brain development 11 -Similar to isopentenyl-diphosphatedelta isomerase, (M) BC004801 3 9.80E-05 cholesterol biosynthesis, steroid biosynthesis 12 -P53 apoptosis effectorrelated to Pmp22, (O) NM_022032 2.4 1.11E-04 induction of apoptosis 13 -Tumor differentially expressed 1, like, (S) NM_019760 2 1.17E-04 plasma membrane 14 -RIKEN cDNA5830409B12 gene, (S) AK017914 7.6 9.73E-05 putative cytoskeleton associated protein 15 -G7e protein, (S) NM_033075 2 1.42E-04 resembles viral envelope genes 16 -Procollagenlysine, 2-oxoglutarate 5-dioxygenase 2, (M) NM_011961 2.8 1.43E-04 protein metabolism 17 -Receptor-like tyrosine kinase, (M) L21707 1.7 2.08E-04 ATP binding, kinaseactivity 18 -RIKEN cDNA4930506D01 gene, (G R) BC006745 2.8 1.77E-04 putative transcription factor 19 -MusmusculusBRUL4 (Brul4) mRNA, (G R) AF314173 5.8 1.91E-04 RNA binding, translation regulator 20 -CyclinI, (C) NM_017367 2.3 1.67E-04 cell cycle, cyclin-dependent protein kinaseregulator activity relevant references 1 -none 2 -Gastroenterology. 2002. 123:790–802; J LeukocBiol. 2002. 72:1084–1091 3 -Peptides. 2000. 21:1433–1446; IntJ Biochem Cell Biol. 1999. 31:1243–1248. 4 -none 5 -J BiolChem. 2002. 277:36296–36303; Gene. 2003. 302: 147–153. 6 -Biochim BiophysActa. 2002. 1577:437–444. 7 -J. Cell Sci. 2002. 115: 4577–4586. 8 -Neuron. 2002. 36:815–829; Nucleic Acids Res. 2003.31:4616–4624 9 -none 10 -J BiolChem. 2002.277:37804–37810; Mol Cell Biol. 2003. 23:6857–6875 11 -J Mol Evol. 2003. 57: 282–291. 12 -Genes Dev. 2000. 14: 704–718. 13 -J Exp Biol. 2000. 203:447–457. 14 -none 15 -Genomics. 1996. 15:5–12 16 -J BiolChem. 2003. 278:40967–40972. 17 -none 18 -J Immunol. 1995. 154:1157–1166; Proc NatlAcadSciU S A. 1992. 89:11818–11822. 19 -CytogenetGenome Res. 2002.97:254–260. 20 -Gene. 2000. 256: 59–67. The presence of several RNA binding proteins among the top-20 genes affected by co-cultivation may result in downstream effects on gene expression, and we plan to examine the target mRNAs of these RNA binding proteins in endothelial cells. This should help elucidate additional post-transcriptional pathways and networks regulating cell growth mechanisms and tumorigenesis [ 9 ]. Conclusion This study describes changes in the gene expression profile of an endothelial cell when co-cultivated with a tumor cell by isolating ribonucleoprotein complexes and identifying their associated mRNAs using genomic arrays. Moreover, it presents a model system that can be used to elucidate post-transcriptional operons in specific types of cells by using various RNA binding proteins from mixed cell cultures as a novel approach to understanding how cell-cell communication affects gene expression during tumorigenesis and organogenesis. Methods Cell lines and media Murine endothelial PY4.1 cells were kindly provided by Dr. Christopher Kontos, Duke University Medical Center. Human gliobastoma T98G and murine breast cancer 4T1 cells were obtained from American Type Culture Collection. All cell lines were maintained in DMEM Medium (Gibco) supplemented with 10% Fetal Bovine serum. Constructs and stable cell lines The ORF of human PABP I containing the Flag tag (GACTACAAGGACGACGATGACAAG) or the G10 tag (CCACCATGGCT AGCATGACTGGTGGACAGCAAATGGGT) at the 5' end was cloned into the pCMV-Neo retroviral vector [ 38 ]. Stable lines expressing the Flag-PABP (T98G and PY 4.1 cells) and the G10-PABP (PY 4.1 cells) were obtained according the protocol described in the Pantropic Retroviral Expression System (Clontech). Antibodies Monoclonal anti-G10 antibodies were obtained as previously described [ 39 ]. Antibodies against Flag and α-tubulin were obtained from SIGMA. A PABP carboxy-terminal (last 172 amino acids) was prepared by cloning a PCR product into the pGEXCS expression vector. The protein was purified by their affinity to glutathione beads (Amersham Biosciences). The purified proteins were dialyzed against 1 × PBS, 20% glycerol and sent to COVANCE Inc., where a rabbit was immunized. Protein preparation and Western analysis Protein extracts were prepared from T98G and PY4.1 cell lines by homogenization in polysomal lysis buffer [ 10 ]. 50 μg of extract were fractionated by electrophoresis in 10% polyacrylamide-SDS Laemmli gels. Proteins were transferred to nitrocellulose membranes using a transfer cell (Bio-Rad). After blocking with 5% nonfat milk in PBS-Tween 20 buffer, the membranes were incubated with anti-PABP rabbit serum (1:10,000 dilution), anti-Flag antibody (1:1,500 dilution) or anti-G10 antibody (1:10,000 dilution). Anti-rabbit or anti-mouse HPC IgGs (Amersham Biosciences) were used as secondary antibodies at a 1:3000 dilution. Blots were developed using an ECL detection kit (Amersham Biosciences) and exposed to film. Cell cycle experiments Analysis of the cell cycle of T98G cells was performed as described [ 40 ]. Immunoprecipitation of mRNP complexes from cell lysates Cell lysates and immunoprecipitation of mRNP complexes were essentially performed as described [ 41 ]. Polyadenylated RNA (free poly-A) used in competition experiments was obtained from Amersham Biosciences. RNase Protection Assay Total or immunoprecipitated RNAs were assayed by RNase protection by using the PharMingen Riboquant assay according to the manufacturer's recommendations (45014K). mAngio-1 (mouse angiogenesis) and hTS1 (human tumor suppressor) template sets were used (551418 and 556161, respectively). Protected riboprobe fragments were visualized on a phosphorimaging screen (Molecular Dynamics). Phosphorimages were scanned by using the Molecular Dynamics STORM 860SYSTEM at 100 μm resolution and analyzed by using Molecular Dynamics IMAGE QUANT software (version 5.0). Clontech microarrays, probing and analysis cDNA array analysis was performed by using Atlas Mouse and Human 1.2 Arrays (CLONTECH). Probing of cDNA arrays was performed as described in the CLONTECH Atlas cDNA Expression Arrays User Manual (PT3140-1). Reverse-transcribed probes were radiolabeled with 32 P α-dATP (Amersham Biosciences). After hybridization, the array membrane was washed and the results were visualized on a phosphorimaging screen (Molecular Dynamics). Phosphorimages were scanned by using the Molecular Dynamics STORM 860SYSTEM at 100 μm resolution and stored as .gel files. Images were analyzed by using ATLASIMAGE 2.01 software (CLONTECH). Global normalization was used when arrays being compared had approximately the same number of positive hits. Printed oligo arrays, probing and analysis Printed oligo arrays using the Operon Mouse Oligo set version 2.0 (16,423 genes) were produced by the Duke Microarray Core Facility. Protocols used for preparation of slides, labeling, amplification, hybridization and scanning are described in . GenePix data were normalized with pin-tip specific lowess normalization [ 42 ]. Differentially expressed genes were identified with a moderated t-test, which shrinks the estimated sample variances towards a pooled estimate [ 43 ]. This moderated t-test is more robust when the number of arrays is small. The candidate gene list is sorted by the p-value. All calculations were conducted using the bioconductor package [ 44 ]. List of abreviations ELAV – embryonic lethal abnormal vision RPA – RNase Protection Assay PABP – Poly A binding protein Author's contribution LOFP was responsible for the experimental design and performed Western blots, immunoprecipitations, RPAs and microarray experiments. MDB helped perform RPAs and microarray experiments. SML performed statistic analysis of microarray data and prepared the webpage with supplementary data. HS generated the stable cell lines used in this study and performed the cell cycle experiment. JDK conceived the project and assisted in experimental design. Supplementary material	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC521085.xml
503393	Bereavement care interventions: a systematic review	Background Despite abundant bereavement care options, consensus is lacking regarding optimal care for bereaved persons. Methods We conducted a systematic review, searching MEDLINE, PsychINFO, CINAHL, EBMR, and other databases using the terms (bereaved or bereavement) and (grief) combined with (intervention or support or counselling or therapy) and (controlled or trial or design). We also searched citations in published reports for additional pertinent studies. Eligible studies had to evaluate whether the treatment of bereaved individuals reduced bereavement-related symptoms. Data from the studies was abstracted independently by two reviewers. Results 74 eligible studies evaluated diverse treatments designed to ameliorate a variety of outcomes associated with bereavement. Among studies utilizing a structured therapeutic relationship, eight featured pharmacotherapy (4 included an untreated control group), 39 featured support groups or counselling (23 included a control group), and 25 studies featured cognitive-behavioural, psychodynamic, psychoanalytical, or interpersonal therapies (17 included a control group). Seven studies employed systems-oriented interventions (all had control groups). Other than efficacy for pharmacological treatment of bereavement-related depression, we could identify no consistent pattern of treatment benefit among the other forms of interventions. Conclusions Due to a paucity of reports on controlled clinical trails, no rigorous evidence-based recommendation regarding the treatment of bereaved persons is currently possible except for the pharmacologic treatment of depression. We postulate the following five factors as impeding scientific progress regarding bereavement care interventions: 1) excessive theoretical heterogeneity, 2) stultifying between-study variation, 3) inadequate reporting of intervention procedures, 4) few published replication studies, and 5) methodological flaws of study design.	Background Give sorrow words; the grief that does not speak Whispers the o'er fraught heart and bids it break. Shakespeare, Macbeth IV, iii, 209 Grieving the death of a loved one has an ancient history: from time immemorial, cultures have provided the bereaved with advice and rituals to address – and express – the experience of grief [ 1 ]. Over the past several decades, efforts to aid the bereaved have increasingly focused on the physical and psychological morbidity, and the spiritual suffering and social isolation associated with bereavement. The resulting plethora of intervention options, ranging from mutual-help support groups to prescribed pharmacotherapy and professionally led psychotherapy, is striking, as is the panoply of settings in which bereavement care can be found: hospitals, hospices, churches, palliative care units, community-based services, and bereavement-specific foundations all provide an array of bereavement care interventions. This welter of activity testifies to the broadly valued goal of decreasing the severity of bereavement-related symptoms. Given the abundance of care options, what is the best way to care for a bereaved person? Numerous studies measuring the impact of bereavement interventions have been published in diverse journals, yet no consensus has emerged in the medical, mental health, or social work communities regarding whether one form of treatment is preferable to another [ 2 - 5 ]. We therefore have conducted a systematic review of bereavement care interventions. Our goal is to present a comprehensive yet coherent synthesis of the current literature that will promote the advancement in the quality of care and research on behalf of bereaved individuals. Methods Data sources To identify studies in the traditional medical literature as well as the complementary and alternative medicine literature, we searched the following databases: MEDLINE; PsychINFO; Cumulative Index to Nursing and Allied Health (CINAHL); BIOSIS Previews; ISI Science Citation Index Expanded and Social Sciences Index; Evidence Based Medicine Reviews (EBMR), including the Cochrane Database of Systematic Reviews (DSR), the Cochrane Controlled Trial Registry (CCTR), Database of Abstracts of Reviews of Effectiveness (DARE), and the American College of Physicians' (ACP) Journal Club Review; Sociological Abstracts; Alt HealthWatch; and Wilson Web from 1966 to 2003. We identified all relevant articles on bereavement care interventions by using the primary search terms of " bereaved or bereavement " and " grief ", combined with secondary descriptors of " intervention or support or counselling or therapy " and " controlled or trial or design ". Study selection Our inclusion criteria specified that each study: 1) addressed the treatment of bereaved individuals, and 2) included an evaluation of a selected method of therapy aimed at reducing the grief reaction due to bereavement. We considered only articles written in the English language. We then reviewed the titles and abstracts of all articles we retrieved through our initial database search, and obtained the full texts of all applicable studies. We also reviewed the references in all applicable studies for additional pertinent studies. Data extraction The full articles of all studies that met inclusion criteria and passed subsequent title and abstract reviews were retrieved and examined independently by two of the authors. Each article was reviewed for measured outcomes, patient and decedent characteristics, and intervention characteristics. These measures included sample size, type of intervention, length of intervention, patient's relationship to the deceased, time since the bereaved death, and patient demographics. Data was extracted and any disagreements were resolved through discussion, clarification, and consensus within the research team. Characteristics of reviewed studies The initial literature search generated 737 citations. Elimination of duplicate citations yielded 340 references. 2 studies, written in Chinese and Spanish, were excluded. Reviewing the titles culled the sample to 243 citations, and a review of the abstracts found 87 of these to be potentially relevant. Of these, 9 were dissertations, 2 were irretrievable, 2 were duplicate publications of the same study, and 15 were ineligible because they did not meet our inclusion criteria. The resulting set of 74 articles was subject to review for data extraction. A list of all citations found, including those excluded from this analysis, is available [see Additional file 1 ]. Of the 74 studies that met inclusion criteria, almost 6,000 participants within these studies experienced a multitude of losses – of parents, spouses, children, and other loved ones who had died from a wide range of causes, both sudden and protracted. The therapies utilized and outcomes evaluated varied widely. Heterogeneity among both the outcomes and the measures used to assess similar outcomes precluded an effort to summarize data across studies, even in the form of generic effect-size measures. Furthermore, for a significant portion of the studies, concerns regarding the internal or external validity of the reported results cautioned against making quantitative summary statements regarding treatment effects. Results The 74 studies selected for detailed review evaluated diverse types of interventions designed to ameliorate the adverse physical and psychological outcomes associated with bereavement. These interventions can be classified according to various schemes, including their underlying theoretical framework (ranging from Freudian psychoanalysis to neurotransmitter imbalances), the format of the intervention (individual, group, family, marital), the timing of the intervention (acute, intermittent crisis, chronic), the tasks assigned to the bereaved (ranging from verbalizing feelings to taking medication), or the population targeted for the intervention (children, adults, seniors). We chose to organize this review on the basis of the social framework used to implement the intervention (that is, either personalized structured therapeutic relationships or less personal systems-level interventions), as this attribute of the interventions emerged as the most verifiable and salient measure. Structured therapeutic relationship Eight studies feature pharmacotherapy , but only four compared active therapy to non-pharmacotherapy controls, and only one study clearly reported their random allocation method (Table 1 ) [ 6 - 13 ]. These studies targeted adults and seniors, ranged in sample size from 10–80 subjects, and used a variety of drugs, including tricyclic antidepressants (TCA), selective serotonin reuptake inhibitors (SSRI), buprioion, and benzodiazepines. Overall, these studies demonstrated a statistically significant beneficial effect of pharmacotherapy on ameliorating symptoms of depression and improving subjective sleep quality [ 6 - 11 , 13 ]. These benefits persisted only as long as the subjects continued to receive pharmacotherapy. Pharmacotherapy was found, however, to have a mixed effect on bereavement intensity as measured by symptoms of grief (i.e., Texas Revised Inventory of Grief, Inventory of Complicated Grief). For example, Warner and colleagues (2001) did not find evidence of an effect of benzodiazepines (diazepam) on bereavement-related grief intensity[ 12 ]. One study combined pharmacotherapy with psychotherapy in a 16-week double-blinded factorial design trial of nortriptyline (NT) and interpersonal psychotherapy [ 6 ]. The 80 patients were randomly assigned to one of four treatment conditions: NT plus interpersonal psychotherapy, NT plus medication clinic (i.e., no interpersonal psychotherapy), placebo pill plus interpersonal psychotherapy, and placebo pill plus medication clinic (i.e., no interpersonal psychotherapy conditions). Details of the psychotherapy were not described. While the results displayed a statistically significant benefit of nortriptyline over placebo regarding remission of depression, none of the treatment conditions were associated with diminishment of grief. Table 1 Pharmacotherapy Interventions Medication Pop CG RA Num * TSL (days) Dose DT (days) Key Outcome Measures Article Nortriptyline Senior Y Y-NE 80/66 216–279 Steady-state plasma level: 50–120 ng/mL 112 Depression (HAM-D); Grief (TRIG) Reynolds, Miller, et al, 1999** Senior Y Y-NE 27/27 210 (mean) Steady-state plasma level: 79.9+/-28.3 ng/mL Daily dose: 70.8+/-22.2 mg <180 Sleep (PSQI); Depression (HAM-D, BDI) Taylor, Reynolds, et al, 1999 Senior Y NR 30/24 276 Steady-state plasma level: 72.7 ng/mL Daily dose: 53.0 mg 112 Sleep (PSQI) Pasternak, Reynolds, et al, 1994 Senior N NA 13/13 150–750 Daily dose: 49.2 mg 9–184 Depression (HAM-D, BDI, BSI); Grief (TRIG, JGI); Sleep (PSQI) Pasternak, Reynolds, et al, 1991 Nortriptyline and Paroxetine Adult N NA 21/15 183–4158 PT Daily dose: 20–50 mg NT Daily dose: 50–160 mg 120 Depression (HAM-D); Grief (ICG); Sleep (PSQI) Zygmont, Prigerson, et al, 1998 Desipramine Adult N NA 10/9 NR Daily dose: 75–150 mg 28 Depression (HDRS, CGI, Raskin DS); Grief (Separation Distress) Jacobs, Nelson, et al, 1987 Bupropion Adult N NA 22/14 42–56 Daily dose: 150–300 mg 56 Grief (TRIG, ICG); Depression (HAM-D) Zisook, Schuchter, et al, 2001 Diazepam Senior Y Y 35/30 <14 2 mg/pill, self-administered <42 Bereavement (BPQ) Warner, Metcalfe, et al, 2001 Notes: * All Ns are reported as (starting population of bereaved individuals/bereaved population completing all follow-ups), unless only study included only one assessment. Study also included psychotherapy condition. Legend: Pop, Target Population; CG, Control Group; RA, Random Assignment; Num, Number of subjects; TSL, Time Since Loss; DT, Duration of Trial; NA, Not Applicable; NR, Not Reported; UC, Unclear; Y, Yes; N, No; Y-NE, Randomization mentioned, but allocation method not explicitly stated; RS, Randomization Subverted. Support groups or counselling constituted the intervention in 39 studies, of which 23 had control groups and 15 claimed random allocation, yet only three of these included clearly described allocation methods (Table 2 ) [ 14 - 52 ]. Ten of these were mutual/self-help, with the majority taking the form of informal group therapy. The remaining 29 studies were professionally led support groups targeting select subgroups including parentally bereaved children, college students, and seniors, as well as many specific adult populations. Program implementation across studies varied even further. This variation was found in terms of number of sessions (one to 25), whether the sessions proceeded with full-fledged patient-driven discussion or highly structured protocols, whether attendance was mandatory or individually motivated, as well as in the nature of the group leadership and the format (individual, group, or marital). Perhaps due to these or other differences in the interventions, some studies documented study treatment effects [ 22 , 26 , 29 - 31 , 33 , 34 , 52 ] while other studies showed no effect [ 15 , 17 , 27 , 37 , 46 , 51 ]. Table 2 Support/Counselling Interventions Type Format Pop CG RA Num TSL (days) DT Key Outcome Measures Article Mutual/Self-help Individual Adult Y Y-NE 162/62 ~30 NR Psychiatric Functioning (GHQ); Social Support/psychological and psychophysiological variables (author-created) Vachon, Lyall, et al, 1980 Mutual/Self-help (included professionally-lead groups) Group Senior Y RS 339/295 30–60 56, 365 days Self-Esteem (Rosenberg's Self-Esteem Scale); Life Satisfaction (LSI-A); Depression (GDS); Grief (TRIG) Caserta & Lund, 1983 Mutual/Self-help Group Senior Y N 23 34–474 21 days; 7 sessions Domain Specific State Locus of Control (Zeigler-Reid State Locus of Control Measure); Trait Locus of Control (I-E); Distress (BSI, GSI) McKibbin, Guarnaccia, et al, 1997 Mutual/Self-Help Group Adult Y Y 113/67 90–365 63 days; 9 sessions Depression (GHQ, BDI); Anxiety (STAI); Social Functioning (SAS); Social Support (SSQ) Tudiver, Hilditch, et al, 1992 Mutual/Self-help Group Adult Y Y-NE 113/112 90–365 63 days Healthcare visit rates (Family Physician, Specialist, Psychiatrist) Tudiver, Permaul-Woods, et al, 1995 Mutual/Self-help Group Adult Y N 38/21 90–750 70 days; 10 sessions Treatment Expectancy (Expectancy Scale); Depression (BDI); Avoidance, Anxiety (Social Anxiety and Distress Scale); Enjoyability (Pleasant Events Scale); Life Satisfaction (Life Satisfaction Scale) Walls & Meyers, 1985 Mutual/Self-help Group Adult Y N 721/502 ~1290 365 days; >3 sessions Depression, Anxiety, Somatization (Hopkins Symptom Checklist); Self Esteem, Well-being, Mastery (Not reported) Lieberman & Videka-Sherman, 1986 Mutual/Self-help Group Adult Y N 667/391 365–1095 365 days Depression, Anxiety, Somatization (Not reported); Self Esteem (Rosenberg 1965); Life Satisfaction, Mastery, Medication (Not reported); Social Functioning Parental Functioning Attitudes (BPQ) Videka-Sherman & Lieberman, 1985 Mutual/Self-help Group Adult N Y-NE 61/55 120–1095 84 days; 12 sessions Avoidance/Intrusion (IES); Stress Symptoms (SRRS); Depression (BDI); Mental Distress (BPRS, SCL-90); Social Functioning (SAS-SR); Overall Functioning (GAS) Marmar, Horowitz, et al, 1988 Mutual/Self-help Group Adult N NA 53/33 <730 8 sessions, optional 4 Psychosomatic Symptoms (SCL-90 subscales: somatization, obsessive-compulsive, interpersonal sensitivity, depression, anxiety, hostility, phobic anxiety, paranoid ideation, psychoticism, GSI) Rogers, Sheldon, et al, 1982 Professionally Lead Individual Adult Y NR 493/225 120 1 day; 1 session Grief (HGRC) Kaunonen, Tarkka, et al, 2000 Professionally Lead Family Adult Y Y-NE 50/30 1–2 1–120 days; up to 8 sessions General Health Questionnaire (self-rated); Anxiety, Depression (Leeds Scale) Forrest, Standish, & Baum, 1982 Professionally Lead Family Adult Y UC 334/161* <1–180 7–70 days; up to 10 sessions Medical Illness (CMI, MMPI); Psychiatric Illness (Boston Bereavement, Mood Inventory); Family Functioning (Ferriera-Winter, Bodin Drawing); Crisis Coping (Intrapersonal, Family, Job/Financial, Social); Social Cost (Gross Income, Living Expenses, Absenteeism, Economic Loss) Williams, Lee, & Polak, 1976 Polak, Egan, et al, 1975 Professionally Lead Family Adult Y UC 176/86* <1–180 7–70 days; up to 10 sessions Neurotic Symptoms Scale; Bodin Family Closeness; Crisis Coping Scale; Religious helping of others; Authoritarian Family Functioning; Depression; Monthly Income; Monthly Expenses; Social Costs; Bereavement Adjustment Williams & Polak, 1979 Professionally Lead Family Adult N NA 77/37* <1 >360 days Personal and social phenomena of death (structured interview) Oliver, Sturtevant, et al, 2001 Professionally Lead Family Child Y Y-NE 72/55 <730 15 sessions Depression (CDI, CBCL, PERI Demoralization Scale); Parental Warmth (CRPBI); Family Cohesion (Family Environment Scale); Parent perception of support (author-created scale); Family Coping (F-COPES) Sandler, West, et al, 1992 Professionally Lead Group Senior N NR 28/11 90–7300 <140 days; up to 20 sessions Social Support (ASSIS); Affect/Mood (PANAS); Emotional/Social Loneliness (ESLI) Stewart, Craig, et al, 2001 Professionally Lead Group Adult Y Y 197/166 <180 70 days; 10 sessions Grief (TRIG, GRI); Distress (POMS); Depression and Anxiety (SIGH-AD) Goodkin, Blaney, et al, 1999 Professionally Lead Group Adult Y Y-NE 242/185 0–35 77 days; 3 sessions Distress (POMS-TMD, Anxiety-tension, Depression-dejection, Anger-hostility, Confusion-bewilderment, Overall emotional disturbance); Self-Esteem (Rosenberg 1965 scale) Swanson, 1999 Professionally Lead Group Adult Y Y-NE 150/120 30–240 270 days Depression (CES-D, BDI); Anxiety (A-Sta); Somatic Symptoms (SOM); Emotional Symptoms (EMOT); Life Satisfaction (Lsat, SelfAnch) Kay, Guernsey de Zapien, et al, 1993 Professionally Lead Group Adult Y Y-NE 119/119 <180 70 days; up to 10 sessions Immunological measures (CD3+CD4+ cell count, CD3+CD8+ cell count, CD4/CD8 ratio, CD3+ cell count, CD4 cell count, Lymphocyte count, T-lymphocyte count); Neuroendocrine measure (Plasma cortisol level) Goodkin, Feaster, et al, 1998 Professionally Lead Group Adult Y Y-NE 110/80 ~730 28 days; 8 sessions Coping and Adaptation (TAT) Balk, Lampe, et al, 1998 Professionally Lead Group Adult Y Y-NE 36/36 <180 70 days; up to 10 sessions Plasma Viral Load (HIV-1 RNA copy number) Goodkin, Baldewicz, et al, 2001 Professionally Lead Group Adult Y N 159/127 42–140 Up to 25 sessions Social Support (SSES); Group Involvement (Liberman & Videka-Sherman, 1986); Depression (CES-D, POMS-D); Anger (POMS-A); Anxiety (POMS-T); Stress (IES) Levy, Derby, et al, 1993 Professionally Lead Group Adult Y N 121 30–4745 30–365 days Grief (HGRC subscales: Despair, Panic behavior, Personal growth, Blame and Anger, Detachment, Disorganization) DiMarco, Menke, & McNamara, 2001 Professionally Lead Group Adult N Y 139/107 90–17155 84 days; 12 sessions Avoidance/Intrusion (IES); Grief (TRIG); Interpersonal Distress (IIP); Social Functioning (SAS-SR); Depression (BDI); Anxiety (STAI); Mental Distress (BSI, GSI); Self-Esteem (SES); Physical Functioning (SF-36); Symptomatic Distress (SCL-90) Piper, McCallum, et al, 2001** Professionally Lead Group Adult N N 83/70 <30–8030 49 days; 7 sessions Physical, Emotional, and Social Functioning (author created measures); Self Esteem (Rosenberg, 1962); Locus of Control (I-E); Life satisfaction (Neugarten, Havighurdt, & Tobin, 1961); Attitude Toward Women (Spence & Helmreich, 1972, Gump 1972) Barrett, 1978 Professionally Lead Group Adult N NA 392/77 NR 56 days Distress (BSI); Group Process and Satisfaction (author created questionnaire) Glajchen & Magen, 1995 Professionally Lead Group Adult N NA 174/138 780 730 days Motives for joining (Lieberman 1979); Interpersonal relations (Porat 1987); Group leadership style (Porat 1987); Perceived contribution of treatment on recovery Geron, Ginsberg, & Solomon, 2003 Professionally Lead Group Adult N NA 21/21 NR 70 days; up to 10 sessions Perceived Social Support (PRQ); Perceived Stress (PSS) Davis, Hoshiko, et al, 1992 Professionally Lead Group Adult N NA 21/21 365–3650 52 days; up to 8 sessions Depression (CDI); Anxiety (HSC-25); Knowledge of Death and Bereavement (KDBQ) Stoddart, Burke, & Temple, 2002 Professionally Lead Group Adult N NA 20 90–1095 <1095 days; unlimited sessions Grief (TRIG); Social Network (SNM, SNG) Forte, Barrett, & Campbell, 1996 Professionally Lead Group Adult N NA 12/5 60–780 28 days; 8 sessions Emotional Distress (EPI); Family Adjustment (FACES-III); Social Adjustment (SAS-SR) Heiney, Ruffin, & Goon-Johnson, 1995 Professionally Lead Group Child Y Y-NE 17/17 >730 42 days; 6 sessions Self-Esteem (PH); Depression (CDI); Behavior (CBCL-TRF, CBCL-YSR) Huss & Ritchie, 1999 Professionally Lead Group Child N NA 38/29 <900 300 days; 12 sessions Depression (BID); Attitude/ Conception of Death (ATCD) Shilling, Koh, et al, 1992 Professionally Lead Group Child N NA 18/18 <730 52 days; 8 sessions Bereavement Survey (author created); Loss Resolution (LRS-Modified); Distress and Somatic Complaints (ALAC) Opie, Goodwin, Finke, et al, 1992 Professionally Lead Group Child N NA 6/6 240–1020 42 days; 6 sessions Psychological measures (Lewis Counselling Inventory, IPAT) Quarmby, 1993 Professionally Lead Group Child N NA 4/4 <90 77 days; 11 sessions Self-Esteem (Piers-Harris Self-Concept Scale); Descriptive (Risk Impact, Negative Chain Events, Opening Up Opportunities) Zambelli & DeRosa, 1992 Professionally Lead Couple/ Marital Adult Y Y-NE 57/31 NR Mean of 6 sessions Grief (TRIG); Irritability, Depression, Anger (IDA) Lilford, Stratton, et al, 1994 Notes: * Families, not individuals. Study also included psychotherapy condition. Legend: Type, Type of Intervention; Format, Format of Intervention; Pop, Target Population; CG, Control Group; RA, Random Assignment; Num, Number of subjects; TSL, Time Since Loss; DT, Duration of Trial; NA, Not Applicable; NR, Not Reported; UC, Unclear; Y, Yes; N, No; Y-NE, Randomization mentioned, but allocation method not explicitly stated; RS, Randomization Subverted. Several studies documented substantial spontaneous improvements in bereavement symptomology in the control groups. Kay and others (1993) report a bereavement intervention for Mexican-American widows [ 33 ]. They found that all widows improved on all depression scales, state anxiety, life satisfaction, and emotional and somatic symptom scales over the course of two years. However, those widows in the experimental support group exhibit significantly improved changes in these scores. Tudiver and colleagues (1992) conducted a mutual-help support group for recently bereaved widowers [ 17 ] that can be compared to Vachon and colleagues' (1980) and Barrett's (1978) widow studies [ 14 , 39 ]. Tudiver and others found significant improvement over time (baseline to eight months) for all widowers, but found no significant differences between those who received treatment and a comparison group of windowers who were on the wait list to receive treatment but had not. Psychotherapy-based treatments , another form of psychological interventions, can be done in different formats (family, group, or individual), and via different approaches. Of the 25 studies that use psychotherapy as an intervention, approaches included cognitive-behavioral, psychodynamic, psychoanalytical, and interpersonal approaches, as well as combinations of these and modality and social support (Table 3 )[ 6 , 19 , 22 , 35 , 38 , 53 - 72 ]. Seventeen of these studies utilized control groups, only 13 claimed randomization, and only five of these clearly stated their method of allocation. Table 3 Psychotherapy Interventions Type Format Pop CG RA Num TSL (days) DT Key Outcome Measures Article Cognitive-behavioral Individual Senior Y N 58/NR 120–180 70 days; 4 sessions Mastery (Personal Mastery Scale); Well-being (MHI subscales, ABS Subscale, PERI self-esteem); Distress (PERI Demoralization Scales, MHI subscales) Reich & Zautra, 1989 Individual Senior N NA 4/4 540–730 98 days; 14–18 sessions Distress (SUDS); Grief (ICG); Depression (BDI); Anxiety (BAI) Harkness, Shear, et al, 2002 Individual Adult Y Y 30/25 >90 35 days; 10 sessions Avoidance/Intrusion (IES); Anxiety (SCL-90); Depression (SCL-90); Mood (POMS) Lange, van de Ven, et al, 2001 Individual Adult Y Y-NE 26/14 180–7300 70 days; 6 sessions Depression (Wakefield, BDI); Physical Symptoms (Mawson et al, 1981); Fear (FQ); Grief (TRIG); Avoidance (Bereavement Avoidance Tasks) Sireling, Cohen, & Marks, 1988 Group Adult Y Y-NE 261/147 46–229 84 days; 8 sessions Mental Distress (BSI, GSI); PTS Symptoms (TES); Grief (GES); Physical Health (HHB); Marital Strain (DAS) Murphy, Johnson, et al, 1998 Murphy, 1997 Group Adult Y Y-NE 110/80 ~730 28 days; 8 sessions Coping and Adaptation (TAT) Balk, Lampe, et al, 1998 Group Adult Y N 38/21 90–750 70 days; 10 sessions Treatment Expectancy (Expectancy Scale); Depression (BDI); Avoidance, Anxiety (Social Anxiety and Distress Scale); Enjoyability (Pleasant Events Scale); Life Satisfaction (Life Satisfaction Scale) Walls & Meyers, 1985 Group Adult N NA 8/8 >30 56 days; 8 sessions Avoidance/Intrusion (IES); Depression (BDI, SCL-90-R); Anxiety (SCL-90-R, STAI); Grief (GRI); Distress (PERI Demoralization) Sikkema, Kalichman, et al, 1995** Group Child Y UC 19/18 <730 NR Behavior (BRIC-S, BRIC-H); Depression (DSRS); Grief (BP) Hilliard, 2001 Psycho-dynamic Individual Senior Y Y 228 ~60 <180 days; Unlimited sessions Number of Office Visits, Types of Illnesses Gerber, Wiener, Battin, et al, 1975 Individual Senior Y Y-NE 33/30 90–1170 14 days; 4 sessions Mental Distress (BSI); Depression (GDS); Hopelessness (GHS); Avoidance/Intrusion (IES); Mood (PANAS) Segal, Bogaards, et al, 1999 Individual Adult Y Y 66/56 <49 90 days; up to 9 sessions General Health(general health questionnaire) Raphael, 1977 Individual Adult Y N 72/63 60–462 12–20 sessions Avoidance/Intrusion (IES-A, IES-I); Depression (SCL-90); Anxiety (SCL-90); Total Pathology (SCL-90); Stress-Intrusion (SRRS); Neurotic Symptoms (BPRS) Horowitz, Weiss, et al, 1984 Individual Adult N Y-NE 12/6 365–3650 196 days Depression (Wakefield); Grief (TRIG) Phobic Avoidance (FQ); Hostility/Anger/Guilt (HAG); Attitude to self and deceased (author-created scales); Avoidance (Bereavement Avoidance Tasks); Physical Symptoms (Maddison & Viola, 1968); Compulsive Behavior (Compulsive Activity Checklist); Social Adjustment (Watson & Marks, 1971) Mawson, Marks, et al, 1981 Individual Adult N NA 1/1 <180 112 days; 10 sessions Grief (Grief Scale); Coping (CRI) Orton, 1994 Group Senior Y Y 150/117 <365–7300 540 days; 6 sessions Depression (BDI); Socialization (RSAS) Constantino, 1988*** Psycho-dynamic Group Adult Y Y-NE 56/53 120–330 8 sessions Depression, Anxiety, Somatization (Hopkins Symptom Checklist); Grief Intensity, Preoccupation, Guilt, Anger (Lieberman & Videka-Sherman, 1986); Psychological Distress (Bradburn Affect Balance Scale); Locus of Control (Pearlman et al, 1981); Self-Esteem (Rosenberg scale, 1965); Social Adjustment (Pearlman et al, 1981, Lieberman & Videka-Sherman, 1986) Lieberman & Yalom, 1992 Group Adult Y N 50/50 NR 90 days; 14 sessions Grief (TRIG) Sabatini, 1988–89 Group Adult N Y 139/107 90–17155 84 days; 12 sessions Avoidance/Intrusion (IES); Grief (TRIG); Interpersonal Distress (IIP); Social Functioning (SAS-SR); Depression (BDI); Anxiety (STAI); Mental Distress (BSI, GSI); Self-Esteem (SES); Physical Functioning (SF-36); Symptomatic Distress (SCL-90) Piper, McCallum, et al, 2001* Group Adult N Y-NE 61/55 120–1095 84 days; 12 sessions Avoidance/Intrusion (IES); Stress Symptoms (SRRS); Depression (BDI); Mental Distress (BPRS, SCL-90); Social Functioning (SAS-SR); Overall Functioning (GAS) Marmar, Horowitz, et al, 1988 Group Child Y N 16/16 30–365 56 days Depression (CDI); Behavior (AP, AT); Grief (BP); Family alliance (TP); Grief, family relationship (TC) Tonkins & Lambert, 1996 Group Child N NA 45/37 30–3650 70 days Trauma (CPTSRI) Salloum, Avery, & McClain, 2001 Psycho-analytic Group Adult N N 154/59 NR 84 days; 12 sessions Affect (author created); Psychodynamic Work (PWORS); Severity of objectives (author created) McCallum, Piper, & Morin, 1993 Inter-personal Individual Senior Y Y-NE 80/66 216–279 112 days; up to 16 sessions Grief (TRIG) Reynolds, Miller, et al, 1999* Behavioral and Psycho-dynamic Individual Adult Y N 83/83 <1825 15–20 sessions Anger (State Trait Anger Inventory); Anxiety (STAI); Avoidance/Intrusion (IES); Somatic/psychoneurotic symptoms (SCL-90); (Locus of Control Scale) Kleber & Brom, 1987 Notes: * Study also included support/counselling condition. Treatment also included aspects of interpersonal psychotherapy. * Study also included pharmacotherapy condition. ** Treatment also included aspects of social support. *** Study also included social activities condition. Legend: Type, Type of Intervention; Format, Format of Intervention; Pop, Target Population; CG, Control Group; RA, Random Assignment; Num, Number of subjects; TSL, Time Since Loss; DT, Duration of Trial; NA, Not Applicable; NR, Not Reported; UC, Unclear; Y, Yes; N, No; Y-NE Randomization mentioned, but allocation method not explicitly stated. Cognitive-behavioral therapy was employed in nine trials, four of which used individual sessions while five studies used group sessions. Murphy and colleagues (1998) studied an intervention for parents bereaved by the violent death of their children [ 57 ]. The results show no treatment effect between intervention and control groups over the five main tested outcome variables. The authors then proceeded with a post-hoc subgroup analysis, which identified mothers with high Global Severity Index scores and grief at baseline as potentially benefiting from intervention during the period, while fathers who received the intervention appeared to have more posttraumatic stress disorder (PTSD) symptoms at six-month follow-up. Kleber and Brom (1987) conducted a comparative outcome study of three forms of short-term psychotherapy [ 69 ]. They compared the results of 83 patients suffering from a major loss who had been randomized into hypnotherapy (behavioral), trauma desensitization (behavioral), psychodynamic therapy, and a delayed-treatment control group. They found all three therapies successful in improving patients' conditions, but did not find any particular therapy to be significantly more effective than another. While the control group showed slight recovery, over time the three therapies were more effective in reducing symptoms of the bereavement response. Studies of psychodynamic therapy, which strives for the patient to understand and cope better with feelings by re-experiencing them and talking them through with the aid of the therapist, was found to be quite prevalent in the bereavement care literature. Overall, the results are mixed, with more support found in the group format of psychodynamic therapy than in individual therapy. Of the studies we evaluated as psychodynamic therapy, six were individual in format, seven had a group format, and eight employed control groups; five of these claimed random allocation (one additional study randomly assigned subjects to two experimental conditions but lacked a control group). Psychoanalysis , as exemplified by Freud, proceeds with an inward investigation of unconscious mental processes and childhood experiences as the principal therapeutic procedure. Problematic measurement methodology beset the one study that utilized a group format to provide a psychoanalytic-based intervention (with no details regarding the tasks assigned to the patients)[ 68 ]. This study focused primarily on the relationship between the patient's personal affect (measured by an unvalidated affect assessment scale) and a favorable treatment outcome (measured again by an ad-hoc unvalidated measure). Behavioral therapy uses learning principles (such as behavior modification, systematic desensitization, and aversion) to eliminate or reduce unwanted reactions to either external situations, one's thoughts and feelings, and bodily sensations or functions. Behavioral therapy was used in only one study, which compared traumatic desensitization to hypnotherapy and psychodynamic therapies [ 69 ]. As described in the section on cognitive-behavioral therapies above, all three therapies resulted in significant improvements from pre- to post-treatment as compared to controls, and no one therapy was found to be more effective than the others in treating bereavement-related symptoms. Interpersonal therapy aims to improve communication skills and increase self-esteem during a short time period by focusing on a patient's behaviors and social interactions with family and friends, directly teaching how better to relate to others. Only one study used interpersonal therapy as a bereavement care intervention, and this study found no effect on grief as the only measured outcome [ 6 ]. Systems-oriented interventions Seven studies featured interventions that altered the manner in which the healthcare system interacted with patients, family, and friends prior to death, guided by an underlying (yet not fully explicated) notion that interactions experienced by loved ones prior to death can influence the subsequent bereavement process (Table 4 ) [ 73 - 79 ]. Six of the seven interventions provided enhanced or augmented care, in the form of palliative care, hospice care, or care coordination. One intervention gave family members the option of witnessing resuscitation efforts [ 79 ]. Overall, the studies that reported systems-oriented interventions produced mixed results of efficacy, with only three of the seven studies showing any treatment effect, mostly in long-term follow-up ranging from 60–365 days post-death. In fact, no study found significant treatment effects when measured during the intervention. Table 4 Systems-Oriented Interventions Intervention Pop CG RA Num Time of Evaluation Key Outcome Measures Article Care Coordination Relative of cancer death Y Y 94 365 days pre-death 56 days post-death Anxiety (HADA, Leeds Depression and Anxiety Scale); Depression (HADD, Leeds Depression and Anxiety Scale); Social Support (Family Apgar Scale) Addington, MacDonald, et al, 1992 Emergency Room Relative of Emergency Room Death Y N 100/66 180–365 days post-death Changes in satisfaction of care, information received (author-created questionnaire) Adamowski, Dickinson, et al, 1993 Hospice Care Relative of cancer death Y Y-NE 96 42 days post-death 540 days post-death Depression (CES-D); Anxiety (Rand Health Insurance Study); General Health (Rand Health Insurance Study); Social Functioning Kane, Klein, et al, 1986 Palliative Care Relative of cancer death Y Y 183 60–270 days pre-death 390 days post-death Grief (TRIG2) Ringdal, Jordhoy, et al, 2001 Relative of cancer death Y NR 119/49 0–60 days post-death Anxiety, Depression, Mental Exhaustion ("observations and ratings") Haggmark & Theorell, 1988 Relative of cancer death Y N 49/37 365 days post-death Health, Anger, Mental State, Depression (Holland & Segroi's instrument) Haggmark, Bachner, & Theorell, 1991 Witnessed Resuscitation Relative of unsuccessful resuscitation Y Y 18 30 days post-death 90 days post-death Grief (TRIG1, TRIG2); Avoidance/Intrusion (IESA, IESI); Depression (BDI, HADD); Anxiety (HADA, BAI) Robinson, Makenzie-Ross, et al, 1998 Legend: Pop, Target Population; CG, Control Group; RA, Random Assignment; Num, Number of subjects; NR, Not Reported; Y, Yes; N, No; Y-NE Randomization mentioned, but allocation method not explicitly stated. Ringdal and colleagues (2001) found no significant differences between those family members whose relative received palliative care and those who received traditional care [ 76 ]. This intervention, however, was not directed to the bereaved relatives, but rather to their terminally ill relatives. The bereaved relatives did show an overall significant decline in TRIG grief scores over one year post-bereavement for both palliative and traditional care groups. Robinson (1998) examined the psychological effect of witnessing resuscitation efforts of patients in the emergency room on bereaved relatives [ 79 ]. They found no psychological differences between the control group who did not witness the resuscitation attempt and the experimental group who had the option of viewing the resuscitation effort. In fact, at the three- and nine-month follow up, the experimental group exhibited median scores lower (that is, better) than the controls on five of the eight measured scales. At nine months, the authors found the difference in TRIG2 scores approaching the 5% significance level with a reported p = 0.08. These findings provide no evidence to support the popular belief that relatives should be excluded from the resuscitation room, and provide only weak evidence of possible psychological benefit of witnessed resuscitation; they do not, however, suggest that having witnessed an unsuccessful resuscitation attempt alleviates the grief reaction of the bereaved. Discussion When reviewed systematically, the current bereavement intervention literature – notwithstanding the existence of many intriguing reports – yields few reliable conclusions to guide treatment. Good evidence supports the pharmacological treatment of depression occurring in the context of bereavement. For all other forms of intervention, however, and for all attempts to diminish grief per se , no consistent pattern of treatment benefit has been established across well-designed experimental studies. Why – despite prevalence of bereavement, the intense dedication on the part of the bereavement research community, and the multitude of peer-reviewed published bereavement studies – does the field of bereavement care lack a formidable evidence base? In order to improve the effectiveness and quality of bereavement care, this question begs to be addressed. On the basis of our systematic review of the literature, we postulate the following five factors as hindering methodical scientific progress regarding bereavement care interventions. Excessive theoretical heterogeneity As the history of science and medicine suggests, successful scientific inquiry into a topic is typically a cumulative process undertaken by a community of investigators working within a shared scientific paradigm [ 80 , 81 ]. The field of bereavement care intervention studies does not appear to be organized in such a manner, but instead consists of distinct groups of investigators working within disparate theoretical frameworks: pharmacologic, psychodynamic, psychoanalytic, behavioral, cognitive-behavioral, interpersonal, and social supportive theories each vie for attention. Indeed, although specification of an underlying treatment-theory conceptual model may improve causal inference [ 82 ], the bereavement care literature may be too invested in and reliant on theoretical justifications of treatments. Consequently, the compiled published reports demonstrate a cumulative 'Tower of Babel' phenomenon, with the different theory-dominated perspectives failing to engage each other meaningfully: the sum is no greater than the parts, and perhaps less. Stultifying between-study variation Treatments featured in published studies vary almost as much as the authors who tested them. One can observe substantial variation across studies regarding the type of intervention generally or regarding the specific implementation of a specific type of intervention (such as different doses of pharmaceuticals); regarding characteristics of targeted patient populations; regarding outcome measurements and study design methodology. Scrutinizing the key outcome measures listed in the accompanying tables illustrates this remarkable heterogeneity. Although these differences have been due in part to diverse treatment-theory paradigms, even studies conducted within the same theoretical paradigm often differed markedly in terms of what potential benefit was being tested, and how it was being measured. Such substantial variation between studies stymies comparison or confirmation of treatment effects. Inadequate reporting of intervention procedures and implementation Aside from the pharmacological studies, which reported the dosing of the intervention medication, very few reported studies describe the intervention procedures in sufficient detail for readers to envision clearly what tasks or activities intervention subjects were asked to perform. This under-specification prevents sensible analysis, within a class of treatments (such as cognitive-behavioral therapy), of observed differences in treatment effects (since the implementation of cognitive-behavioral therapy, for instance, may have been quite different in seemingly similar intervention studies). Few published "replication" studies Inadequate specification of intervention procedures, combined with other factors at work within the community of bereavement care investigators, may have resulted in the dearth of published replication studies. This lack of replication prevents the accumulation of a body of evidence that would confirm, refute, or refine prior estimates of treatment effects. Methodological study-design and data-analysis flaws A final factor inhibiting research progress in the realm of bereavement care interventions encompasses a number of recurring methodological flaws that greatly limit inferences regarding treatment effects. First and foremost is the omission of control groups. Control groups are essential for the valid evaluation of a bereavement intervention, particularly because of the typically self-limited course of grief: even absent any treatment, most bereaved people show "diminished pathological symptoms and fewer signs of disturbance within two years of the loss"[ 65 ]. Purported beneficial treatment effects observed in an intervention group without a suitable control group therefore may in fact be simply the natural grief remission process. A second common study design feature is the non-random assignment of study subjects into treatment and control groups, which again limits the strength of inference regarding observed 'treatment' effects, as these differences between treatment and control groups may be due to selection or assignment bias. Third, many studies measured subject outcomes using untried assessment tools that had been created on an ad hoc basis, and which may therefore have compromised measurement accuracy and inference validity. Lastly, studies that failed to demonstrate a statistically-significant difference for the main outcome measure often performed numerous post hoc subgroup analyses, a practice that negates the rigor of statistical hypothesis testing. If these five factors are indeed hampering progress towards improving bereavement care interventions and quality of care for bereaved individuals, then concrete actions could facilitate progress within the field of bereavement care, specifically: 1) Convening a consensus-building conference among key stakeholders and investigators to define a specific research agenda that would draw on a limited number of theoretical paradigms and delineate key elements of treatment theory [ 82 ]; 2) Focusing on interventions to improve key outcomes that are valued by bereaved individuals; 3) Targeting well-defined patient populations at well-defined phases of bereavement; 4) Conducting high-quality randomized controlled trial research designs, employing rigorous tests of hypotheses defined prior to the conduct of the study, and eschewing unplanned subgroup analyses; 5) Weighing the ethical arguments for and against the use of randomized control subjects in such research; 6) Increasing incentive to conduct and publish highly-comparable replication studies; and 7) Enforcing the adoption of uniform standards regarding clinical trial study reporting (such as outlined in the CONSORT statement [ 83 ]) by journal editors and the bereavement research community. Competing interests None declared. Abbreviations CG Control group NT Nortriptyline SSRI Selective serotonin reuptake inhibitors TCA Tricyclic antidepressants TRIG Texas Revised Inventory of Grief Authors' contributions AF assisted in the design of the review, conducted data collection, data abstraction, and drafted the manuscript. MH assisted in the design of the review and critically reviewed several drafts of the manuscript. RP assisted in data collection and abstraction. CF conceived of and designed the review, assisted in drafting and revision of the manuscript, and provided support and mentorship through the process. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 All citations. This file contains citations to all of the studies identified by our literature search and screened for inclusion in this review, as well as other scholarly works consulted during the conduct of this review. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC503393.xml
555937	The Use of Edge-Betweenness Clustering to Investigate Biological Function in Protein Interaction Networks	Background This paper describes an automated method for finding clusters of interconnected proteins in protein interaction networks and retrieving protein annotations associated with these clusters. Results Protein interaction graphs were separated into subgraphs of interconnected proteins, using the JUNG implementation of Girvan and Newman's Edge-Betweenness algorithm. Functions were sought for these subgraphs by detecting significant correlations with the distribution of Gene Ontology terms which had been used to annotate the proteins within each cluster. The method was implemented using freely available software (JUNG and the R statistical package). Protein clusters with significant correlations to functional annotations could be identified and included groups of proteins know to cooperate in cell metabolism. The method appears to be resilient against the presence of false positive interactions. Conclusion This method provides a useful tool for rapid screening of small to medium size protein interaction datasets.	Background Protein interaction datasets are typically presented as graphs (or networks), in which the nodes are proteins and the edges represent the interactions between the proteins. These graphs can be used to investigate the functions of unannotated proteins through their interactions with neighbouring annotated proteins. Protein interaction datasets frequently contain many false positives and false negatives, (Bader et al [ 1 ], von Mering et al [ 2 ]) but studies have shown that true positives are frequently associated with areas where there are many interactions between neighbours (clusters). For example Giot et al [ 3 ] used independent datasets to remove false positives from a large-scale protein interaction dataset and as a result were able to demonstrate that true positives had a strong positive correlation with the clusters. Spirin and Mirney [ 4 ] found that clusters of highly interconnected proteins are significant features of protein interaction networks. These could not have occurred by chance and are therefore likely to represent groups of proteins that have co-evolved to serve a common biological function. Identification of clusters is therefore likely to capture the biologically meaningful interactions in large scale datasets. Edge-Betweenness clustering [ 5 ], the method used here, has been exploited in the social and ecological sciences to study communities [ 6 ] and in the study of biochemical pathways [ 7 ]. It has proved to be a useful and adaptable method. As discussed by Holme et al [ 7 ] edge-betweenness uses properties calculated from the whole graph, allowing information from non-local features to be used in the clustering. Many other clustering methods, which have proved useful for clustering protein interaction graphs, are based on calculation of local quantities such as node degree (number of attached edges) [ 8 , 9 ]. These 'local' methods will exclude nodes with a low degree e.g. the many prey nodes attached to their bait by a single edge, which are common in yeast two-hybrid (Y2H) datasets. Methods using whole graph properties will automatically include these poorly connected nodes in clusters [ 5 ], whilst a 'local' method would need to restore such nodes in a post-processing step [ 9 ]. Clusters created using edge-betweenness clustering are therefore useful when the information associated with these nodes is required. Other methods based on whole graph properties will also have this advantage, for example Markov Clustering [ 10 ]. A discussion of different clustering methods can be found in [ 11 ] We applied the edge-betweenness method to a set of human protein interactions from our laboratory [ 12 , 13 ]. In these experiments interactions were identified using the Y2H method. For comparison, two datasets of yeast protein interactions [ 14 , 15 ] were also analysed. One yeast dataset also used the Y2H method [ 14 ] whereas the other was prepared using affinity purification [ 15 ]. The functions identified for clusters by the automatic method were compared with the expert biologists' interpretations presented in these papers. Results Allocation of GO terms Differences in clustering between the datasets The three datasets used differ in content, purpose, size, structure and species. A more detailed description of each dataset is given in the 'Methods' section and in Table 1 , but briefly, the Gavin and Uetz datasets were large scale screens of the yeast proteome, not focused on particular metabolic pathways, whereas the Lehner dataset is focused on a few metabolic areas/complexes related to the human MHC class III region. While Lehner and Uetz both used the Y2H method to detect protein-protein interactions, Gavin used a combination of affinity purification and mass-spectroscopy. The two yeast datasets (Gavin and Uetz) have approximately 5× more nodes than the Lehner dataset. Whilst the Gavin and Uetz datasets have roughly the same number of nodes, the Gavin (affinity purification) dataset has twice as many edges (3145 vs 1498) as the Uetz (Y2H) dataset. The affinity purification method (Gavin) retrieves fairly stable complexes of proteins whereas the Y2H method detects direct protein-protein interactions which may be weak or transient. From Tables 2 and 3 it can be seen that the affinity purification dataset gives much bigger clusters with the removal of a similar proportion of edges, when compared to the Y2H datasets. When 15% of edges were removed from the Gavin dataset, the clusters (with more than one member) had an average of 23 nodes whilst for Uetz the average was just over 7 nodes. The Lehner dataset fell between these values. Diagrams showing the Lehner dataset before and after clustering are presented in Additional files 11 and 12 . The choice of the number of edges removed needs to be guided by the dataset and problem under consideration. A number of criterion could be used. (i) Range of cluster sizes: To decide what a sensible distribution of cluster sizes would be, the range of sizes of clusters found by affinity purification was used as a guide. Gavin [ 15 ] reported the distribution of cluster sizes as follows:-51% had 1–5 nodes, 18% 6–10 nodes, 15% 11–20 nodes, 6% 21–30 nodes, 4% 31–40 nodes, and 6% > 40 nodes. In order to emulate this type of distribution with the automatic clustering (see Table 2 ) it is necessary to remove more than 13% of edges from the Uetz and Lehner datasets and more than 25% from the Gavin dataset. Therefore it is necessary to remove a much higher proportion of edges from the affinity purification dataset. Other results from Tables 2 , 3 and 4 that could also be used to try and determine the appropriate number of edges to remove are (ii) increasing the significant number of GO terms per protein (iii) aiming for an average size of cluster of 5–20 proteins (iv) reducing the size of the biggest cluster to < 20% of the dataset, a useful metric to indicate reasonable decomposition of the dataset (but which could be varied according to the total number of nodes in the dataset) (v) reducing the number of nodes not associated with any other nodes to < 30%. The proportion of edges that need to be removed in order to attain each of these criteria would be:- (i)distribution cluster size Gavin 25% Uetz 13% Lehner 14% edges (ii) significant GO terms For all datasets, the more edges that are removed the more terms become significant down to the smallest cluster sizes investigated (iii)average cluster 5–20 Gavin 25% Uetz 2–13% Lehner 7–25% edges (iv)biggest cluster < 20% Gavin 25% Uetz 13% Lehner 14% edges (v)single nodes < 10 % Gavin 25% Uetz 27% Lehner 25% edges The data above shows that most of these criteria give similar results and suggest that the method used to produce the data (Y2H or affinity purification) will be a major determinant of the proportion of edges to remove. To summarise, for Y2H, useful results are obtained by removing 10%–15% of edges whereas for affinity purification, removing 25% edges gives better results. Newman and Girvan [ 16 ] have developed methods for assessing the 'modularity' of the clusters produced by edge-betweenness clustering. It would also be possible to use methods of this type, as a more objective way of deciding how many edges to remove in different datasets. Size of cluster is important, because the quantity of significant annotation information i.e. the average number of significant GO terms per protein, (Table 4 ) increased, for all datasets, as cluster size decreased. However the detail of the information, measured as average depth of GO per node, did not change with cluster size. It is noticeable that human proteins in the Lehner dataset [ 12 , 13 ] had been annotated to a greater level of detail (average depth of nearly 6 in the GO hierarchy) than the yeast proteins (average dept of approx 4.7, see Table 4 ) and whereas virtually all of the clusters in the Lehner dataset had a correlation with at least one GO term there were many clusters in the yeast dataset which had no significant GO terms (the majority in the case of the Uetz dataset). This could be a peculiarity of the metabolic areas chosen for the Lehner study. Scaling The utility of this approach is currently restricted by the size of the dataset being analysed, especially when a large number of edges are being removed. For the Gavin dataset, when 57 edges were removed the total time to cluster was 1 h 25 min but when removing 1500 edges it took 10 h 10 min. According to the software documentation [ 17 ] and as discussed by Newman [ 6 ] the running time for sparse graphs (such as these) is proportional to both the number of edges removed and the total number of nodes. The Ito dataset (see below) took >>24 h when > 500 edges were removed. This method is therefore of greater utility for small to medium datasets, having less than 2000 nodes or edges. Significance of GO terms After performing the Chi Squared tests and checking them against a random reallocation of GO terms across the network, all the significant GO cluster correlations remained significant. In no case were more than 5% of the lowest p values of the randomly reallocated GO terms lower than the lowest p value in the original dataset. In almost every case the significant annotations were informative about a potential function for the clusters (see Table 5 ), providing distinctive groupings of annotations which distinguished different functions for the different clusters (the aim of the method). It was often a very small proportion of the proteins which provided the annotations which were used to characterise the cluster, (Table 5 and Additional Files 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 , 10 , which provide complete sets of clustering results and details of the proteins which contributed the significant annotation). Correlation with biological function One test of this method was to determine whether the clusters generated and the associated GO terms corresponded to clusters previously identified by expert biologists. With respect to the Lehner et al dataset [ 12 ], the authors identified groups of interacting proteins which appeared to be involved in distinct biological processes including transcription regulation, protein-ubiquination, cell cycle regulation and mRNA processing When edge-betweenness clustering was used to remove 57 edges, 21 clusters (with size greater than one) were created (Table 3 ). From Table 5 (and from the more detailed information in Additional file 7 ), it can be seen that these clusters differ in the significant GO terms associated with them i.e. the method does separate groups of proteins with different metabolic functions. Significantly, clusters were generated with functions corresponding to all of the metabolic areas identified by informed biological interpretation. These were transcription (cluster 9), ubiquination (cluster 15), cell cycle reg (cluster 14) and mRNA processing (cluster 21, cluster 3, cluster 6) Only one cluster, cluster 10, had a description ("biological process") which was too general to give useful information about function. However when this cluster was broken down further (in the test with 100 edges removed) more informative terms ("response to abiotic stimulus", "eukaryotic translation elongation factor 1 complex") were associated with the new, smaller parts of the cluster. Interestingly cluster 10 contained very few proteins with GO terms assigned to them and therefore may represent an under-investigated module in the human proteome. This highlights the dependence of this method on the quality (depth) and quantity of the GO annotations available. This was good for the H.sapiens proteins but less good for the yeast proteins. One important question is whether the functions identified for these protein clusters are confirmed by biological experimentation. The Lsm complex is mentioned by the authors of all 3 papers [ 13 - 15 ], It has been extensively studied in both yeast and human [ 13 ]. The Lsm complex has been shown to have a number of functions related to RNA processing, including the splicing of nuclear pre-mRNA and the decapping of cytoplasmic mRNA prior to degradation. Clusters in the Lehner dataset In the Lehner dataset two GO terms, GO:6371 "mRNA splicing" and GO:8380 "RNA splicing", were always associated with only one cluster in the dataset. This was a good candidate for the Lsm complex. Of the 8 Lsm proteins examined in [ 13 ], all eight were found in the cluster associated with these two GO terms for the tests when 10, 30 and 57 edges were removed. A diagram showing the cluster containing these proteins, in the dataset with 57 edges removed, can be seen in Additional file 13 . When 100 edges were removed, the cluster labeled as RNA splicing contained 5/8 of the Lsm proteins. The three clusters containing the other 3 proteins had the following significant descriptions (the number in parenthesis shows; the number of proteins with this annotation / the total number of proteins in the cluster). GO:15980 energy derivation by oxidation of organic compounds (2/19) GO:5837 26S proteosome (2/16) GO:6350 transcription (2/3) For the Lehner data, when 15, 30 and 57 edges were removed, the clusters labeled as being associated with RNA splicing are large containing 190, 143 and 60 proteins respectively (see below). The cluster with 5/8 Lsm proteins (100 edges removed) had only 17 proteins. In addition to the Lsm proteins the large clusters contained other proteins known (i.e. having GO labels) to be involved in RNA splicing. The proportions are shown in Table 13 . This data clearly shows that as the cluster size gets smaller, the cluster is more focused round the RNA splicing function. Larger clusters must have sub-clusters related to other functions. The last column in the table above shows that many of the RNA splicing proteins grouped in these clusters were the prey of the Lsm proteins in the original experiments [ 13 ], which is what we hoped this method would achieve. Therefore for the Lehner data, the cluster identified by Edge-Betweenness clustering as the "RNA splicing" cluster, did contain the proteins expected to be associated with this process. However this is a small dataset focused around a specific biological process. A more stringent test of this method is provided by the yeast proteome datasets where screening was not functionally focused. Clusters in the yeast datasets Gavin et al [ 15 ] and Uetz et al [ 14 ] both describe the Lsm complex. One complication in both of these datasets, is that the yeast proteins are not annotated to the same level of detail as the human proteins. For example there is no annotation for "RNA splicing" but only the higher level GO term GO:16070 "RNA metabolism", which covers a much broader range of cellular processes. In Gavin et al [ 15 ], the Lsm proteins are found in the complex described as TAP-C128. This contained 36 proteins. The distribution of the TAP-C128 proteins between the clusters are shown in Table 6 . It can be seen that a minimum of 6/7 Lsm proteins and proteins associated with RNA metabolism are clustered together, at all numbers of edges removed. Therefore in a dataset not focused round RNA metabolism, the edge-betweenness algorithm successfully clustered the Lsm proteins with a number of other proteins that were co-purified in the TAP-C128 complex and a cluster produced using the graph topology was shown to correspond to a cluster of known function. In Uetz et al 2000 [ 14 ], the Lsm complex is described as a set of 16 interacting proteins. The one cluster containing all of these proteins does not correlate with the GO term for "RNA metabolism" in the datasets with 30 or 57 edges removed. This correlation only emerged once 100 edges had been removed. With 400 edges removed 11/16 are still in the same "RNA metabolism" cluster (the other 5 are spread between 5 different clusters). Therefore in the Uetz dataset although all the Lsm proteins clustered together, it was only once more than 10% of edges had been removed that it was possible to get a significant association with the relevant GO term. Finding the correct number of edges to remove is obviously essential to extracting the required information. Overall it can be seen that the method is capable of finding clusters of proteins with known biological function and of correctly assigning a relevant annotation to a particular group. Stable and transient clusters In Gavin et al [ 15 ] the authors discuss two clusters which are described as "stable and "transient". TAP-C162 is an example of a "stable" complex which was always isolated with the same members. It is part of the poly-adenylation machinery. In contrast, TAP-C151, the "transient" complex was frequently isolated with different components. It is a signaling complex formed around protein phosphatase 2a. The distribution of these two complexes between the clusters generated by edge-betweenness clustering, was compared at different levels of clustering, (see Tables 7 and 8 ). While TAP-C162 remains mainly associated with one cluster at all numbers of edges removed, TAP-C151 becomes distributed much more evenly between a greater number of clusters. Therefore it seems likely that the method described here favours the detection of more stable clusters, as the number of edges removed increases. False positive interactions Clustering the Lehner dataset with added false positive edges (see "Methods" section and Table 9 ) gave no obvious difference in cluster size (Tables 10 and 11 ) or quality or quantity of GO annotation (Table 12 ). The dataset with false positives is slightly larger than the original dataset, but this did not change the number of clusters. The slight increase in average cluster size led to a commensurately small fall in annotation quality (GO per node), but there were no dramatic differences in cluster size distribution or any of the other measurements. Fourteen out of twenty-one of the clusters in the original dataset remained completely intact, and even when this was not the case a minimum of 70% of the original proteins in the other clusters could still be found together in one of the new clusters. Therefore adding the false positives did not render any of the original clusters unrecognisable. When the dataset with the false positive edges removed was compared to the dataset with the same number of edges removed at random, the differences were more marked. The dataset where edges were removed at random had smaller clusters (Tables 10 and 11 ) and more single nodes (Table 11 last column). The identity of the clusters was perturbed to a greater extent. Further analysis showed that when the false positives were removed 12/21 clusters still remained completely intact. With removal of random edges only 4/21 clusters were completely intact. However even in this dataset 14/21 clusters had 80% of proteins from the original clusters co-occurring i.e. 3/4 of clusters were still recognisable. Randomly removed edges can be considered to be false negatives and so the method is also showing good tolerance to false negatives, and can still preserve a good level of cluster identity. Overall, even though the false negatives reduce the average sizes of the clusters and splits off many single nodes (as would be expected because nodes with single edges are much more abundant than nodes with multiple edges, in Y2H datasets) the same clusters are still being found 75% of the time. In other words the presence of false positives and false negatives in the dataset does not seem to distort the composition of the clusters created by the Edge-Betweenness method in a way that obliterates cluster identity. But false negatives do appear to have a slightly more detrimental effect than false positives. Looking at the edges which were removed during clustering, when 57 edges were removed (from the dataset containing false positive edges) 3/57 (5%) had false positive nodes at one or both ends. When clustering was done by removing 100 edges 15/100(15%) were attached to false positive nodes. This compares with 68/465(14.6%) edges attached to false positive nodes in the whole dataset. There is no obvious bias in the presence of false positive edges between or within clusters. Overall it appears that the clustering is fairly robust to the presence of false positives and also to the random removal of edges i.e. false negatives. With the Ito et al [ 18 ] dataset it was hard to say whether there was much effect from the removal of false positives or addition of false negatives, as the proportion of nodes and edges affected was so small, but again there were no obvious differences. Discussion Edge-Betweenness clustering can be used to separate protein interaction networks into clusters which have correlations with annotated gene functions. This can be done in an automated fashion and thus can provide a means of rapidly screening the results of protein interaction experiments. Clusters produced by this method contain groups of proteins which are known to cooperate to perform common functions, described by the correlating annotations. Therefore the clusters detected by this method correspond to active protein complexes found in the cell. Moreover the method worked for different types of dataset (Y2H and affinity purification) different organisms (yeast and human) and for datasets with a 5× difference in the number of edges. The smaller the clusters generated by this method, the higher the average number of significant annotations. The preliminary results presented here suggest that, in general, useful information was obtained once approximately 10% of edges were removed from Y2H datasets and a slightly higher proportion (25%) from affinity purification data. This method is particularly good at detecting "stable" clusters. The method is also flexible and can be adjusted according to the nature of the dataset and to the function being studied. Currently scaling to very large datasets when large numbers of edges need to be removed is problematic, but this may soon be alleviated by new developments of the algorithm [ 6 ]. The level of detail and amount of available annotation will have a significant effect on the utility of this method although it is possible to tune the amount of annotation found by the method, by altering the number of edges removed. The amount of available annotation will increase as proteome annotation progresses. Spirin and Mirny [ 4 ] have demonstrated the robustness to false positives and negatives of various clustering methods (not including the Edge-Betweenness method used here). They found that 80% of clusters could still be detected if up to 20% of links were added or removed. Our results suggest that Edge-Betweenness clustering is similarly robust. This robustness is undoubtedly for the reason identified in [ 4 ] which is "the use of multiple interactions to identify a cluster", in other words the interconnectedness of a pair of proteins is reconfirmed by the interconnectedness of their neighbours. The biological significance of these interconnected sets of proteins was shown by the high correlation between true positive interactions and clusters in Drosophila protein interaction networks, found by Giot et al [ 3 ]. Giot et al [ 3 ] also found that prey (but not bait) with a large number of neighbours had a significant negative correlation with the reliability of the interactions. These highly connected prey correspond to the promiscuous prey which we identified as false positives and which although highly connected do not have neighbours which are themselves highly interconnected. As this method appears robust to the presence of such proteins it is not necessary to "clean up" the datasets before using them. The hierarchical nature of the Gene Ontology made this a very useful system of annotation to exploit in this method. It allows proteins to be grouped according to the most detailed shared level of annotation but also enables higher level (less informative) annotation to be used when this is all that is available. The very high level terms which apply to almost all proteins are usually ignored as they are not concentrated in a particular cluster, although these terms occasionally appear as significant, in clusters with higher than average levels of annotation. Conclusion Edge-Betweenness clustering provides a quick way of picking out functionally interesting areas of protein interaction datasets. It also appears to be robust against false positives and negatives. As such this approach can be applied to any quality of data. It also deals effectively with poorly connected nodes, such as the many prey with single connections found in Y2H graphs. Because the Edge-Betweenness algorithm does not scale well to larger graphs, this method is currently most appropriate for studies focused on specific areas of the proteome. However, modifications of the algorithm are being developed and these should allow it to be applied to larger datasets in the future [ 6 ]. The implementation described here is particularly effective where good quality GO annotation is available, which is especially true for many human proteins. It will be a useful method for detecting functions for unannotated proteins based on the knowledge of the functions of their neighbours and for exploring functional modules within the proteome. Methods Datasets The datasets used for analysis are described in Table 1 . Briefly the Lehner dataset comes from our work on the function of the MHC class III region [ 12 , 13 ] and is a small, highly focused dataset of H. sapiens protein interactions, detected using the Y2H method [ 12 ]. The other datasets, Gavin [ 15 ] and Uetz [ 14 ], are larger datasets resulting from mass screens of the yeast proteome, using either Y2H (Uetz) or affinity purification (Gavin). The method presented here was developed for the Lehner dataset. In order to test the method, it was applied to the larger, less selective yeast datasets. The Ito dataset [ 18 ], an even larger yeast dataset, was included in order to test the effect of false positive proteins. This dataset contained 16 proteins identified by Gavin et al [ 15 ] as false positives. However it was not used for other aspects of the investigation as clustering takes a long time when large number of edges are removed. Thus the Ito dataset represents the upper limit of the size of datasets suitable for use with the method described here. Protein function The Gene Ontology (GO) [ 19 ] was used as the source of functional annotations. It was chosen because it provides hierarchically structured, controlled vocabularies. Genes or gene products may be labeled with terms from any level in any of the three hierarchies (ontologies). By searching up through the hierarchy, it was possible to find terms shared by proteins which had been initially labeled with different descriptions. The search through the hierarchy is easy to automate, which makes it possible to group together proteins participating in the same general functions, even when they were originally annotated for different, more specific functions. Steps of the analysis The steps of our method to cluster the graph and assign functions to the clusters, were as follows:- 1. Transform the protein interaction data to GraphML (an XML format for graphs [ 20 ]), removing any parallel edges, to make the data ready for import into JUNG. 2. Use the JUNG graph analysis framework [ 21 ] to cluster the data using the "Edge-Betweenness" [ 5 ] algorithm. 3. Find GO terms and the parents of those GO terms for each GO annotated protein in every cluster. 4. Test the association between each GO term and each cluster, from a 2 by 2 contingency table. 5. Correct the association tests for multiple comparisons, using a permutation test with random re-allocation of GO terms to proteins. 6. Generate reports on cluster size and significant GO terms. Perl scripts were used to perform most of these steps, the other software used is described below. Details of the steps listed above are as follows: Clustering JUNG version 1.3 [ 21 ] was used to cluster the graph by the Edge-Betweenness clustering method [ 5 ]. This algorithm removed those edges which lay on routes between interconnected clusters. "Betweenness" is calculated by finding the shortest path(s) between a pair of vertexes and scoring each of the edges on this/these path(s) with the inverse value of the number of shortest paths. (So if there was only one path of the shortest length, each edge on it would score 1 and if there were 10 paths of that length, each edge would score 1/10.) This is done for every pair of vertexes. In this way each edge accumulates a "betweenness" score for the whole network. The network is separated into clusters by removing the edge with the highest "betweenness", then recalculating betweenness and repeating until the desired number of edges have been removed. The method is fully described in [ 5 ]. The number of edges to remove was supplied as a parameter. Removing a larger number of edges reduced the size of the clusters produced. The number of edges removed was varied to see whether (a), clusters of certain sizes gave better correlations with GO terms and (b), whether datasets of different types cluster in different ways (likely, as the affinity purification dataset has approximately 3× as many edges as the Y2H dataset with a similar number of nodes). Source of GO annotations GO terms available for each of the proteins in the graph were retrieved. In the case of the Lehner dataset these were taken from the RefSeq records [ 22 ], for the Uetz and Gavin data these were provided by BIND [ 23 ]. Processing GO annotations The Gene Ontology "termdb" release from December 2003 was used as the source of the parent GO terms [ 24 ]. Tables to hold these GO data were set up using the PostgreSQL relational database management system [ 25 ] (version 7.3.4-RH). The parents of each GO term were found by using an adaptation of the sample query provided on the GO web site [ 26 ]. This query was called from either perl scripts or Java programs, which allocated the terms to the clusters. Detecting GO terms with significant associations to clusters The 'R' statistical package [ 27 ] (version R 1.8.1 (2003-11-21)) was used to perform the statistical analysis on the data retrieved. The association between each cluster and each GO term was tested using a 2 by 2 contingency table by Fisher's exact test. Re-testing significant GO associations The GO terms (significant and non-significant) were redistributed across the clustered network at random. The p value was recalculated for each GO/cluster combination. This randomisation was repeated 1000 times. The overall significance was calculated as the proportion of randomisations in which the smallest p value for a GO-cluster association was less than or equal to the smallest p value in the original data. We considered the GO numbers to be significantly associated with the clusters if the overall significance was less than 5% (i.e. fewer than 50 of the 1000 randomisations' lowest p values were smaller than the smallest p value from the observed data). Reports on significant GO/cluster associations In order to compare the informativeness of the GO/cluster associations, the following ratios were calculated (a), the average number of GO terms per node in the clusters and (b), the average depth of the GO terms per node per cluster. These provided an indication of the 'quantity' and 'quality' of the GO information. A GO at a greater depth in the GO hierarchy provides more detailed information than one higher in the GO hierarchy. False positives In our original experiments [ 12 , 13 ] there were a number of prey that interacted with many different bait. Prey found by more than three different bait were defined as false positives (of the 'promiscuous' type). There were 14 of these (approximately 4% of the dataset nodes). 10 of these 14 had been excluded from the original data. To investigate their effect on clustering, these nodes and all associated edges were added back to the data. This contributed 59 new edges to the dataset (13% of dataset edges). This dataset was clustered and the clusters compared to those found in the original experiment. If these nodes were disconnected this removed 68 edges, so nine of the edges connected to false positives were part of the original data. In a control experiment 68 edges were removed at random (from the dataset with false positives added), this dataset was clustered. This was repeated 100 times and the results were compared to the clusters obtained from the dataset which had false positive edges removed. Gavin 2002 Supplementary Information Table S2 [ 15 ] provided a list of false positive proteins, which were excluded from their yeast dataset. They were excluded because they either appeared in more than 20 of the purifications or were isolated in mock transformations. The data describing the edges created by these proteins was not provided, therefore it was not possible to add them back to the Gavin data. The Uetz data contained only 2 of the false positive proteins, however the Ito dataset contained 16(0.5% of dataset). The Ito dataset is large and 16 out of 3271 nodes is a very small proportion, so any effect will not be large. Disconnecting these nodes removed 26 edges from the dataset (0.6% of edges). A control dataset had 26 edges removed at random before clustering All false positive datasets (see Table 9 ) and controls were clustered by removing 57 edges (a number chosen originally because it gave a tractable number of clusters of a reasonable size in the Lehner dataset). Authors' contributions RD: analysis, interpretation and writing. FD: statistical analysis. CS: data acquisition and supervision. Supplementary Material Additional File 11 Additional file 11 shows the Lehner dataset before it was clustered. The Lsm proteins are highlighted. Click here for file Additional File 12 Additional file 12 shows all the clusters produced when the Lehner dataset was clustered by removing 57 edges. The whole cluster containing the Lsm proteins is highlighted. Click here for file Additional File 1 These files provide information on all the clusters (of size > 1) formed when the Gavin, Uetz and Lehner datasets were clustered. Each file name begins with the name of the dataset used in clustering, followed by the number of edges removed to perform the clustering. The files with a name ending 'clusters.txt' list all the members of each cluster and all the GO terms which were found to be significant for that cluster. Files ending with 'proteins-per-GO.txt', give further detail, showing the transcripts in the cluster which had the significant GO annotations. The examples provided show each dataset with the 'best' number of edges removed (see Results section: Difference in clustering between the datasets). For the smallest (Lehner) dataset, examples with greater and fewer numbers of edges removed are also provided. Click here for file Additional File 2 These files provide information on all the clusters (of size > 1) formed when the Gavin, Uetz and Lehner datasets were clustered. Each file name begins with the name of the dataset used in clustering, followed by the number of edges removed to perform the clustering. The files with a name ending 'clusters.txt' list all the members of each cluster and all the GO terms which were found to be significant for that cluster. Files ending with 'proteins-per-GO.txt', give further detail, showing the transcripts in the cluster which had the significant GO annotations. The examples provided show each dataset with the 'best' number of edges removed (see Results section: Difference in clustering between the datasets). For the smallest (Lehner) dataset, examples with greater and fewer numbers of edges removed are also provided. Click here for file Additional File 3 These files provide information on all the clusters (of size > 1) formed when the Gavin, Uetz and Lehner datasets were clustered. Each file name begins with the name of the dataset used in clustering, followed by the number of edges removed to perform the clustering. The files with a name ending 'clusters.txt' list all the members of each cluster and all the GO terms which were found to be significant for that cluster. Files ending with 'proteins-per-GO.txt', give further detail, showing the transcripts in the cluster which had the significant GO annotations. The examples provided show each dataset with the 'best' number of edges removed (see Results section: Difference in clustering between the datasets). For the smallest (Lehner) dataset, examples with greater and fewer numbers of edges removed are also provided. Click here for file Additional File 4 These files provide information on all the clusters (of size > 1) formed when the Gavin, Uetz and Lehner datasets were clustered. Each file name begins with the name of the dataset used in clustering, followed by the number of edges removed to perform the clustering. The files with a name ending 'clusters.txt' list all the members of each cluster and all the GO terms which were found to be significant for that cluster. Files ending with 'proteins-per-GO.txt', give further detail, showing the transcripts in the cluster which had the significant GO annotations. The examples provided show each dataset with the 'best' number of edges removed (see Results section: Difference in clustering between the datasets). For the smallest (Lehner) dataset, examples with greater and fewer numbers of edges removed are also provided. Click here for file Additional File 5 These files provide information on all the clusters (of size > 1) formed when the Gavin, Uetz and Lehner datasets were clustered. Each file name begins with the name of the dataset used in clustering, followed by the number of edges removed to perform the clustering. The files with a name ending 'clusters.txt' list all the members of each cluster and all the GO terms which were found to be significant for that cluster. Files ending with 'proteins-per-GO.txt', give further detail, showing the transcripts in the cluster which had the significant GO annotations. The examples provided show each dataset with the 'best' number of edges removed (see Results section: Difference in clustering between the datasets). For the smallest (Lehner) dataset, examples with greater and fewer numbers of edges removed are also provided. Click here for file Additional File 6 These files provide information on all the clusters (of size > 1) formed when the Gavin, Uetz and Lehner datasets were clustered. Each file name begins with the name of the dataset used in clustering, followed by the number of edges removed to perform the clustering. The files with a name ending 'clusters.txt' list all the members of each cluster and all the GO terms which were found to be significant for that cluster. Files ending with 'proteins-per-GO.txt', give further detail, showing the transcripts in the cluster which had the significant GO annotations. The examples provided show each dataset with the 'best' number of edges removed (see Results section: Difference in clustering between the datasets). For the smallest (Lehner) dataset, examples with greater and fewer numbers of edges removed are also provided. Click here for file Additional File 7 These files provide information on all the clusters (of size > 1) formed when the Gavin, Uetz and Lehner datasets were clustered. Each file name begins with the name of the dataset used in clustering, followed by the number of edges removed to perform the clustering. The files with a name ending 'clusters.txt' list all the members of each cluster and all the GO terms which were found to be significant for that cluster. Files ending with 'proteins-per-GO.txt', give further detail, showing the transcripts in the cluster which had the significant GO annotations. The examples provided show each dataset with the 'best' number of edges removed (see Results section: Difference in clustering between the datasets). For the smallest (Lehner) dataset, examples with greater and fewer numbers of edges removed are also provided. Click here for file Additional File 8 These files provide information on all the clusters (of size > 1) formed when the Gavin, Uetz and Lehner datasets were clustered. Each file name begins with the name of the dataset used in clustering, followed by the number of edges removed to perform the clustering. The files with a name ending 'clusters.txt' list all the members of each cluster and all the GO terms which were found to be significant for that cluster. Files ending with 'proteins-per-GO.txt', give further detail, showing the transcripts in the cluster which had the significant GO annotations. The examples provided show each dataset with the 'best' number of edges removed (see Results section: Difference in clustering between the datasets). For the smallest (Lehner) dataset, examples with greater and fewer numbers of edges removed are also provided. Click here for file Additional File 9 These files provide information on all the clusters (of size > 1) formed when the Gavin, Uetz and Lehner datasets were clustered. Each file name begins with the name of the dataset used in clustering, followed by the number of edges removed to perform the clustering. The files with a name ending 'clusters.txt' list all the members of each cluster and all the GO terms which were found to be significant for that cluster. Files ending with 'proteins-per-GO.txt', give further detail, showing the transcripts in the cluster which had the significant GO annotations. The examples provided show each dataset with the 'best' number of edges removed (see Results section: Difference in clustering between the datasets). For the smallest (Lehner) dataset, examples with greater and fewer numbers of edges removed are also provided. Click here for file Additional File 10 These files provide information on all the clusters (of size > 1) formed when the Gavin, Uetz and Lehner datasets were clustered. Each file name begins with the name of the dataset used in clustering, followed by the number of edges removed to perform the clustering. The files with a name ending 'clusters.txt' list all the members of each cluster and all the GO terms which were found to be significant for that cluster. Files ending with 'proteins-per-GO.txt', give further detail, showing the transcripts in the cluster which had the significant GO annotations. The examples provided show each dataset with the 'best' number of edges removed (see Results section: Difference in clustering between the datasets). For the smallest (Lehner) dataset, examples with greater and fewer numbers of edges removed are also provided. Click here for file Additional File 13 Additional file 13 shows more detail for this cluster, including the transcript ID for each node. The images were produced using the BioLayout [ 32 ] graph visualisation tool Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC555937.xml
538265	Reduced bio-efficacy of permethrin EC impregnated bednets against an Anopheles gambiae strain with oxidase-based pyrethroid tolerance	Background Insecticide-treated nets (ITNs) are an integral component of malaria control programmes in Africa. How much pyrethroid resistance in malaria vectors will impact on the efficacy of ITNs is controversial. The purpose of this study was to evaluate knockdown and killing effects of ITNs on a metabolic-based resistant or tolerant malaria vector strain. Methods Bio-efficacy of 500 mg/m 2 permethrin EC treated bednets was assessed on the OCEAC laboratory (OC-Lab) strain of Anopheles gambiae s.s.. This strain is resistant to DDT and tolerant to pyrethroids, with elevated mixed function oxidases. The Kisumu reference susceptible strain of A. gambiae s.s . was used as control. Nets were impregnated in February 1998 and used by households of the Ebogo village. Then they were collected monthly over six months for Bio-assays (WHO cone test). Knockdown and mortality rates were compared between the OC-Lab and the Kisumu strains, by means of the Mantel-Haenszel chi-square test. Results During the whole trial, permethrin EC knockdown rates were impressive (mostly higher than 97%). No significant difference was observed between the two strains. However, the mortality rates were significantly decreased in the OC-Lab strain (40–80%) compared with that of the Kisumu strain (75–100%). The decrease of killing effect on the OC-Lab strain was attributed to permethrin EC tolerance, due to the high oxidase metabolic activity. Conclusion These data suggested an impact of pyrethroid tolerance on the residual activity of ITNs. More attention should be given to early detection of resistance using biochemical or molecular assays for better resistance management.	Background Malaria is the most important vector-borne disease in Africa. It is estimated that 80 to 90% of the 300 million annual cases and one million deaths occur on this continent [ 1 ]. The sharp rise of its incidence in the past decades resulted in dramatic economic consequences for African countries [ 2 ]. The global strategy adopted by the World Health Organization (WHO) in 1992 recommended an integrated management of the disease, including selective vector control [ 3 ]. Selective vector control is defined as: application of site-specific targeted use of different and cost-effective vector control methods alone or in combination to reduce human-vector contact. Insecticide-treated nets (ITNs) are one of the main vector control tools against malaria. They are as effective as indoor residual spraying (IRS) [ 4 ] and strongly advocated for malaria prevention [ 5 , 6 ]. Implementation does not systematically require vector control services that no longer exist in many countries. At this time, insecticides belonging to the pyrethroid family are the only compounds available for the impregnation of materials. They strike mosquitoes with knockdown and killing effects at dosages far below the threshold of mammalian toxicity [ 7 ]. However, the emergence of pyrethroid resistance in the Anopheles gambiae complex and the Anopheles funestus group, the most important malaria vectors in Africa, is a threat to the effectiveness of ITNs [ 8 - 11 ]. This resistance is based on several mechanisms that could segregate according to their operational impact on vector biology and control. Some modifications of insecticide effects associated with reduced sensitivity of the sodium ion channel along nerve axons due to kdr mutation have been reported in A. gambiae s.s. from West and East Africa [ 12 , 13 ]. In addition, there is strong evidence for metabolic-based resistance mechanisms in African malaria vectors [ 14 , 15 ]. Three major enzyme families (esterases, glutathione S-transferases and cytochrome P 450 oxidases) are involved in insect detoxification. Elevation of their activity usually results in resistance to insecticides such as pyrethroids. In Cameroon, elevated esterase, oxidase or glutathione S-transferase activities were reported as the main resistance mechanisms in many populations of the A. gambiae complex [ 16 ]. Malaria vector resistance to pyrethroids has been clearly demonstrated in Africa. However, its operational implication in terms of reducing efficacy of ITNs, especially in the case of metabolic-based resistance is not well documented. The aim of this study was to assess the knockdown and killing effects of ITNs on a metabolic-based pyrethroid resistance or tolerance strain of malaria vector. This study reports on the decrease of ITN's killing effect against a laboratory strain of A. gambiae s.s with a likely oxidase-based pyrethroid tolerance. Methods The study was undertaken in the entomology laboratory of the Organisation de Coordination pour la lutte contre les Endémies en Afrique Centrale (OCEAC), in Yaoundé (Cameroon). Bednets impregnation and sampling Bednets were made of white multifilament polyester fabric (75 denier; 156 meshes, 12 × 13 holes/inch 2 ) manufactured by SiamDutch Mosquito Netting Co. Ltd. (Bangkok, Thailand). Two sizes of bednets were used : X-family (16.3 m 2 ) and Family (13.13 m 2 ). Both were strengthened on the lower part by a 20 cm sheeting border (made of more polyester filaments) to prevent tearing while being tucked in. They were impregnated with the target dosage of 500 mg/m 2 permethrin EC and hung in households of the Ebogo village, for use during the period of March-September, 1998. Ebogo-village (3°20 N, 11°20 E) is about 65 km far from Yaoundé (the capital city of Cameroon), in the equatorial forest. Anopheles moucheti is the main malaria vector there, with 307 infected bites/man/year [ 17 ]. This village was chosen for the implementation of ITNs because the people there were used to bednets, since a deltamethrin SC trial was conducted there in 1994. A total of 50 permethrin EC impregnated bednets were distributed in the village, in addition to about 30 old nets that were retreated by the study team. All the new nets were identified by a code number. Immediately after impregnation, two nets were randomly chosen and brought to the laboratory. Then, two others were collected each month from the Ebogo households and systematically replaced by unused ones. Replacement nets and old ones were properly identified so that they could not later be collected from the field and used for bio-assays. People were asked not to wash their bednets during the trial. Laboratory procedure Netting section In the laboratory, netting portions were isolated from the lower part of bednets collected from the field and wrapped in aluminum sheets. Each sample was identified by a code number and kept at 4°C until Bio-assays were performed (less than one month). Mosquito strains The bio-efficacy of treated nets was assessed on the OC-Lab strain of A. gambiae s.s., originated from Yaoundé and laboratory-reared for about 15 years without insecticide selection. The Kisumu susceptible reference strain of A. gambiae s.s., originated from Kenya and provided by LIN/IRD Montpellier, was used as a control. The OC-Lab. strain is known to be strongly resistant to DDT and tolerant to pyrethroids, response to WHO susceptibility test [ 18 ] performed in 1997 is given in table 1 . We registered 26 per cent mortality rate to 4 per cent DDT, 78–95 per cent mortality rates to 0.25 percent permethrin (former diagnostic concentration), 0.025 per cent deltamethrin (former diagnostic concentration) and 0.2 per cent cyfluthrin. With these diagnostic concentrations, the time of knockdown for 50 per cent mosquitoes during exposure to insecticide-impregnated papers was 2–5 fold increased compared with that of the Kisumu strain. However, mortality rates to 1.0 per cent permethrin, 0.05 per cent deltamethrin (revised current diagnostic concentrations) was higher than 98 per cent, with knockdown time ratio less than 2 fold. Table 1 Kisumu susceptible and OCEAC Laboratory strains of Anopheles gambiae s.s. response to WHO susceptibility test. Strains Insecticides No TKd 50 (CI) TKd 95 (CI) Tkd 50 R (CI) Mt ST. Kis. 4% DDT 100 18.8 (17.6–20.0) 28.7 (25.8–33.7) -- 100 S 1.0% permethrin 99 9.2 (8.6–9.7) 14.3 (13.2–16.0) -- 100 S 0.25% permethrin 100 12.4 (11.2–13.7) 28.8 (24.8–35.4) -- 94.1 T 0.05% deltamethrin 89 9.4 (8.4–10.2) 17.2 (15.6–20.0) -- 100 S 0.025% deltamethrin 100 8.9 (8.1–9.7) 19.7 (17.5–23.0) -- 100 S 0.2% cyfluthrin 120 8.6 (8.0–9.1) 15.5 (14.1–17.7) -- 100 S OC-Lab. 4% DDT 100 9.9 (86.1–119.6) 268.9 (195.0–465.5) 5.2 (4.0–6.7) 26 R 1.0% permethrin 101 12.2 (11.5–12.8) 17.5 (16.3–19.5) 1.3 (0.9–1.7) 98.7 S 0.25% permethrin 125 45.7 (13.6–153.3) 109.2 (3.3–3593.0) 3.7 (0.6–22.9) 78.7 R 0.05% deltamethrin 100 16.8 (13.2–21.3) 36.2 (23.7–55.5) 1.8 (1.1–2.9) 100 S 0.025% deltamethrin 125 24.9 (23.6–26.5) 41.2 (37.5–46.9) 2.8 (2.3–3.4) 94.8 T 0.2% cyfluthrin 108 17.8 (12.7–24.8) 34.1 (32.5–37.3) 2.1 (1.3–2.9) 94.4 T Kis.: Kisumu strain, OC-Lab.: OCEAC Laboratory strain, No: Number of tested mosquitoes, Tkd 50 : knockdown time in minutes for 50% tested mosquitoes, Tkd 95 : knockdown time in minutes for 95% tested mosquitoes, CI: confidence interval at 95%, Tkd 50 R: Tkd 50 OC-Lab strain / Tkd 50 Kisumu strain, Mt.: Mortality rate 24 h post exposure, ST: status, S: indicates susceptibility, T: suspects resistance to be confirmed (or tolerance), R: suggests resistance. Biochemical analysis of esterase, mixed function oxidase and glatathione S-transferase enzyme systems using microtitre plates and spectrophotometer as described by Penilla et al. [ 19 ] and Brogdon et al. [ 20 ] revealed elevation of mixed function oxidases activity in the OC-Lab strain (Figure 1 ). For this enzyme system, the activities (mean ± standard deviation) of the OC-Lab and Kisumu strains were 0.049 ± 0.018 and 0.027 ± 0.011 nmol cytochrome unity equivalent/mg protein (rank-sum normal statistic with correction Z = -3.924, p < 0.001). Esterase and glutathione S-transferase levels were lower in the OC-Lab strain than in the Kisumu strain. For esterases, two substrates were used, α-naphtyl acetate and paranitrophenyl acetate. With α-naphtyl acetate, the activities of the OC-Lab and Kisumu strains were 0.057 ± 0.011 and 0.117 ± 0.083 μmol α-naphtol produced/min/mg protein (rank-sum normal statistic with correction Z = 6.302, p < 0.001). Figure 1 Oxidase levels in Kisumu and OCEAC laboratory strains of Anopheles gambiae s.s. through biochemical assays. KISUMU: Pattern of cytochrome P 450 UE/mg protein in individuals of the Kisumu susceptible laboratory strain, OC-Lab: Pattern of cytochrome P 450 UE/mg protein in individuals of the OCEAC laboratory strain. With paranitrophenyl acetate, the activities of the OC-Lab and Kisumu strains were 0.002 ± 0.008 and 0.053 ± 0.132 μmol p -nitrophenol produced/min/mg protein (rank-sum normal statistic with correction Z = 7.028, p < 0.001). For glutathione S-transferases, the activities of the OC-Lab and Kisumu strains were 0.013 ± 0.024 and 0.087 ± 0.104 μmol GSH conjugated/min/mg protein (rank-sum normal statistic with correction Z = 3.172, p = 0.001). PCR analysis [ 21 ] showed that individuals of the OC-Lab strain belonged to the M molecular form of A. gambiae s.s. and those of the Kisumu strain to the S molecular form. Individuals of the OC-Lab. strain which survived to WHO susceptibility test were screened by PCR [ 22 ], all of them appeared free of kdr Leu-Phe mutation. Bio-assays After treatment, ten batches of five unfed females (2–5 days old) of the OC.Lab strain and those of the Kisumu strain were exposed under WHO's plastic cones to netting from newly treated nets for three minutes. Ten other batches of each mosquitoe strain were exposed to netting from untreated nets as control. Mosquitoes were then transferred in white cups and the knockdown rates were recorded at 60 minutes post-exposure. They were then supplied with a 15% glucose solution and held under laboratory conditions, at 80 per cent relative humidity and 27°C (± 2°C) temperature. The mortality rates were recorded after 24 hours. Bio-assays were also performed on used nets. Between March and September 1998, 100 females of A. gambiae s. s. from the OC-Lab strain and 100 specimens from the Kisumu strain were tested each month (from M 0 to M 6 ). For the control, 50 specimens from the OC-Lab and 50 others from the Kisumu strain were exposed to untreated netting. Each month knockdown and mortality rates of the OC-Lab strain and the Kisumu strain were then compared by means of the Mantel-Haenszel chi-square test. Results Efficacy of freshly treated nets Table 2 indicates knockdown and mortality rates in mosquitoes after exposure to nettings from freshly treated and untreated nets. Table 2 Kisumu and OCEAC Laboratory strains of Anopheles gambiae s.s. response to permethrin EC freshly treated nets. Variables Nets Kisumu strain OC-Lab strain X 2 p No % No % Kd rates Untreated 50 0 50 0 Permethrin EC (500 mg/m 2 ) 100 100 100 97 3.04 0.08 Mt rates Untreated 50 0 50 2 Permethrin EC (500 mg/m 2 ) 100 89 100 68 13.06 <0.001 Kd: Knockdown rates 60 minutes post-exposure, Mt: Mortality rates 24 hours post-exposure, No: number of tested mosquitoes, p : Probability at 5%. Knockdown Rates No knockdown effect was observed in mosquitoes exposed to netting from untreated nets (control), either in the Kisumu strain or in the OC-Lab strain. With netting from permethrin EC freshly treated nets (M 0 ), more than 95 per cent of mosquitoes from both strains were knocked down 60 minutes post-exposure. The difference between the two strains was not significant at the five per cent level ( p = 0.08, df = 1). Mortality rates With netting from untreated nets, mortality rate in each strain did not exceed 2 per cent. Using netting from permethrin EC freshly treated nets, the mortality rate in the OC-Lab strain did not exceed 70 per cent, while about 90 per cent mosquitoes of the Kisumu strain were killed, difference between the two strains was highly significant ( p < 0.001, df = 1). Efficacy of treated bednets during domestic utilization Knockdown rates During the whole trial, no knockdown effect was observed in mosquitoes exposed to netting from untreated nets, either in the Kisumu strain or in the OC-Lab strain, while most of the mosquitoes exposed to netting from treated nets were knocked down during the 60 minutes post-exposure. The profile of knockdown rate variations during the six month evaluation is given in Figure 2 . No significant difference was observed between the Kisumu strain and the OC-Lab strain ( p > 0.05, df = 1). For both strains, the knockdown rate was mostly higher than 90 per cent, except in the fifth month during which about 70 per cent were registered. Figure 2 Knockdown rates in Kisumu and OCEAC laboratory strains of Anopheles gambiae s.s. to permethrin EC used nets. OC-Lab: Permethrin EC treated net knockdown rates on the OCEAC laboratory strain, KISUMU: Permethrin EC treated net knockdown rates on the Kisumu susceptible laboratory strain. Mortality rates The mortality rates in the control netting were constantly lower than five per cent for both strains. Conversely, numerous mosquitoes exposed to netting from treated nets were killed during the 24 hours post-exposure. The graph of the mortality rate variations during the six month evaluation is given in Figure 3 . During the first five months, the killing effect was higher in the Kisumu strain than in the OC-Lab strain. The decrease of net efficacy on the OC-Lab strain was significant during the first three months ( p < 0.001, df = 1). From the fourth to the sixth month, there was no longer a significant difference between the two strains (0.13 < p < 0.57, df = 1). Figure 3 Mortality rates in Kisumu and OCEAC laboratory strains of Anopheles gambiae s.s. to permethrin EC used nets. OC-Lab: Permethrin EC treated net mortality rates on the OCEAC laboratory strain, KISUMU: Permethrin EC treated net mortality rates on the Kisumu susceptible laboratory strain, * stars indicate months during which the mortality rates were significantly lower in the OCEAC strain than in the kisumu strain. Discussion DDT resistance in the OC-Lab. strain of A. gambiae s.s. which originated from Yaoundé city was not selected in the laboratory. The selective pressure was performed in the field several years prior to the collection of the strain. DDT was used in Yaoundé for residual indoor spraying during the 1950's [ 23 ]. Furthermore, Desfontaines et al . [ 24 ] reported the intensive use of household insecticides containing mixture of compounds such as pyrethrins and pyrethroids (coils, mats, etc ...) in this city for protection against mosquito bites. The first susceptibility tests on the OC-Lab strain were performed in 1997 using WHO's protocol. Samples were tested for 4.00 per cent DDT, 0.25 per cent and 1.00 per cent permethrin, 0.025 per cent and 0.05 per cent deltamethrin, then 0.20 per cent cyfluthrin. Susceptibility tests were carried out with these ranges of pyrethroid dosage because the OC-Lab strain had to be used for the evaluation of cyfluthrin bio-efficacy in phase III of the World Health Organization Pesticide Scheme (WHOPES). This strain was found resistant to 4.00 per cent DDT and 0.25 per cent permethrin, tolerant to 0.025 per cent deltamethrin and 0.20 per cent cyfluthrin, but susceptible to 1.00 per cent permethrin and 0.05% deltamethrin. It was also shown that the kdr Leu-Phe mutation was not involved in this case of DDT resistance or pyrethroid tolerance. Bio-assays using WHO cone test with a cyfluthrin EW 50 mg active ingredient per m 2 of netting resulted in 35 per cent mortality rate versus 95 per cent rate for the Kisumu strain. The difference in knockdown rates was not significant (95–100 per cent for both strains). It was seen that the strain was not suitable for that trial. In fact, the cyfluthrin bio-efficacy was assessed with the Kisumu susceptible reference strain [ 25 ]. Subsequently, biochemical analysis revealed over-production of mixed function oxidases in the OC-Lab strain and the same metabolic-based resistance was reported in wild populations of A. gambiae s.l. from cotton and rice fields in northern Cameroon [ 16 ], which is a threat for the efficacy of treated nets in this area. It was, therefore, essential to investigate pyrethroid-treated material effectiveness against a metabolic-based resistant malaria vector population. With these rationales, the OC-Lab strain was found suitable for a laboratory trial compared with a susceptible reference strain of A. gambiae s. s., such as the Kisumu strain. The study was not carried out with field mosquitoes because the cotton fields are actually located 1,000 km from the laboratory, it would be difficult to collect sufficient field samples for bio-assays. From current data, the insecticide activity of treated nets on the Kisumu reference strain was clearly demonstrated, despite some breakdowns observed after the third month. The decrease of knockdown and mortality rates at this period may be related to bad conditions of net utilization. Previous reports have underlined the impact of external factors such as dirt and fume on the bio-efficacy of treated nets [ 26 - 28 ]. However, the activity of permethrin in this study was similar to that usually reported in field trials [ 29 , 30 ]. Conversely, nets were less effective against the OC-Lab strain, especially in term of mortality rate. These data are consistent with those previously obtained with cyfluthrin. The decrease of knockdown rate prior to that of mortality rate is known as one of the major modifications of pyrethroid effects associated with kdr mutation [ 12 , 31 ]. The contrast between knockdown and mortality rates in this trial is relevant to the involvement of metabolic detoxification in insecticide resistance which does not systematically induce the decrease of knockdown effect. In Cameroon, ITNs were found effective in reducing malaria transmission and morbidity during the early 1990s [ 32 , 33 ] and, until now, they have been strongly advocated by the national malaria control programme. Therefore, the emergence of pyrethroid resistance in the A. gambiae complex [ 34 ] is of a particular concern for the efficacy of interventions. Generally, insecticide resistance has a major impact in reducing efficacy of IRS programmes. Detoxification through mixed function oxidases was reported to delay the deltamethrin IRS programme against A. funestus populations from northern Kwazulu/Natal [ 10 ]. By the same token, high activities in glutathione S-tranferases, esterases and mixed function oxidases resulted in the failure of the IRS programme against A. albimanus in southern Mexico [ 19 ]. ITNs tested in laboratory and experimental huts in West Africa were found partially effective against DDT or pyrethroid resistant populations of A. gambiae s.s. with kdr gene frequency higher than 70% [ 35 ]. Nevertheless, the epidemiological impact at community level was similar to that observed in areas with susceptible vectors (Henry, personal communication). The lessening of pyrethroid exito-repellent and irritancy effects against knockdown resistant mosquitoes allowed their contact with treated nets and resulted in killing many of them. ITNs could, therefore, work positively against pyrethroid-resistant malaria vectors with kdr gene. Considering the genetic diversity of pyrethoid resistance mechanisms, the efficacy of ITNs against knockdown resistant populations could not be extrapolated to vector populations with elevated monoxygenases activity. From the current study, it has been seen that the knockdown rates were not decreased in the permethrin-tolerant strain; likewise pyrethroid properties (excito-repellent and irritancy) may not be impeded against metabolic-based resistance mosquitoes. There is a converging suggestion that the impact of insecticide resistance on the efficacy of ITNs used as personal protection tools might not be limited when resistance is due to high metabolic detoxification. Conversely, the decrease of mortality rates puts forward a potential limited impact of ITNs when used at community level as vector control intervention aiming at mass reduction of vector density. Conclusions Current data call attention to early detection of resistance as one of the key guidelines for insecticide resistance management. Susceptibility tests are the entry point for insecticide resistance studies. However, whether resistance is detected or not, it is necessary to go through biochemical or molecular assays for detection of resistance genes which may not have a great impact on the level of resistance when they stand at very low frequency. In order to preserve efficacy of relevant tools such as ITNs against malaria vectors, it would be easy to control insecticide resistance when the occurrence of the involved gene is at low rather than at high frequency. Moreover, the impact of insecticide resistance on vector control interventions is a complex phenomenon that depends not only on the resistance itself (mechanisms, gene frequency, etc...), but also includes vector behaviour, environment and insecticide properties such as the excito-repellent effect. Drawing a general conclusion on the efficacy of ITNs in areas with metabolic-based resistant vector populations needs further investigation in experimental huts to study their behaviour and at community level to assess epidemiological impact. Authors' contributions JE participated in the conception of the study, carried out field and laboratory procedures, analysed and interpreted data, drafted and revised the manuscript. FC carried out biochemical assays, participated in data analysis and interpretation. PG participated in the design of the study and revising it critically. LM conceived the study, participated in its design, coordinated, revised and gave final approval of the version to be published.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC538265.xml
423137	Integrative Analysis of the Mitochondrial Proteome in Yeast	In this study yeast mitochondria were used as a model system to apply, evaluate, and integrate different genomic approaches to define the proteins of an organelle. Liquid chromatography mass spectrometry applied to purified mitochondria identified 546 proteins. By expression analysis and comparison to other proteome studies, we demonstrate that the proteomic approach identifies primarily highly abundant proteins. By expanding our evaluation to other types of genomic approaches, including systematic deletion phenotype screening, expression profiling, subcellular localization studies, protein interaction analyses, and computational predictions, we show that an integration of approaches moves beyond the limitations of any single approach. We report the success of each approach by benchmarking it against a reference set of known mitochondrial proteins, and predict approximately 700 proteins associated with the mitochondrial organelle from the integration of 22 datasets. We show that a combination of complementary approaches like deletion phenotype screening and mass spectrometry can identify over 75% of the known mitochondrial proteome. These findings have implications for choosing optimal genome-wide approaches for the study of other cellular systems, including organelles and pathways in various species. Furthermore, our systematic identification of genes involved in mitochondrial function and biogenesis in yeast expands the candidate genes available for mapping Mendelian and complex mitochondrial disorders in humans.	Introduction About half of the expected mitochondrial proteins in humans are known to date, and already a fifth of these known proteins are associated with human Mendelian disorders (Online Mendelian Inheritance in Man [ http://www.ncbi.nlm.nih.gov/Omim/ ]; DiMauro and Schon 1998 ; Andreoli et al. 2004 ). Mitochondrial core functions such as oxidative phosphorylation, amino acid metabolism, fatty acid oxidation, and iron-sulfur cluster assembly have been highly conserved during evolution, suggesting that a systematic identification of mitochondrial proteins in model organisms will accelerate the search for new human mitochondrial disease genes ( Steinmetz et al. 2002 ). In yeast, 477 proteins (469 encoded by the nuclear genome) show conclusive evidence of mitochondrial localization (this study and those listed in the Mitochondrial Proteome 2 [MitoP2] database [ http://ihg.gsf.de/mitop ]). About 30% of these proteins have evidence of orthologs in humans (MitoP2 database). Identification of the yeast mitochondrial proteome is far from complete. Thirty to forty percent of the predicted complement of proteins that make up the organelle are still considered unknown although many genome-wide and functional systematic studies have been applied ( Westermann and Neupert 2003 ). These include systematic identification of mitochondrial proteins by mRNA expression analysis under various conditions ( DeRisi et al. 1997 ; Lascaris et al. 2003 ), DNA microarray analysis of mRNA populations associated with mitochondrion-bound polysomes ( Marc et al. 2002 ), deletion phenotype screening ( Dimmer et al. 2002 ; Steinmetz et al. 2002 ), large-scale localization studies ( Kumar et al. 2002 ), protein–protein interaction studies ( Uetz et al. 2000 ; Ito et al. 2001 ; Gavin et al. 2002 ; Ho et al. 2002 ), mass spectrometry (MS) of mitochondria ( Pflieger et al. 2002 ; Ohlmeier et al. 2003 ), and various computational predictions of mitochondrial proteins ( Nakai and Horton 1999 ; Drawid and Gerstein 2000 ; Small et al. 2004 ). In addition, two recent studies reduced the gap of missing mitochondrial localized proteins: a comprehensive proteomic study of mitochondria claimed to reduce the gap to 10% and identified 749 proteins ( Sickmann et al. 2003 ), and a protein localization study identified 527 mitochondrial localized proteins by green fluorescent protein (GFP) tagging ( Huh et al. 2003 ). Here we generated a component list of the mitochondrial organelle by first identifying mitochondrial proteins using MS and then integrating 22 datasets relevant to the study of mitochondria, including our proteomic data. The integration generated a comprehensive definition of the proteins involved in mitochondrial function and biogenesis and allowed for a comparison of genomic approaches, with implications beyond mitochondria. Results/Discussion Proteomics We identified mitochondrial proteins by combining different methods for purification of whole mitochondrial organelles from yeast cell cultures and directly measured the proteins present in these fractions using MS. Mitochondria from yeast cells grown under four different conditions, including fermentable (glucose) and nonfermentable (lactate) substrates for both natural and synthetic culture media, were purified by either density gradient or free-flow electrophoresis. Preparations were separated into mitochondrial membrane and matrix fractions and analyzed separately for protein content. In total, 20 fractions were digested with trypsin and analyzed by reversed phase high resolution liquid chromatography/tandem MS (LC/MS/MS) ( Ferguson and Smith 2003 ; Washburn et al. 2003 ). In addition, eight of the fractions were further analyzed by liquid chromatography/Fourier transform-ion cyclotron resonance MS (LC/FTICR) ( Lipton et al. 2002 ; Smith et al. 2002 ). Altogether, 28 experimental datasets were generated ( Table S1 ), which in combination identified 546 proteins ( Table S2 ); listed also in the Yeast Deletion and Proteomics of Mitochondria [YDPM] database and the MitoP2 database). The performance of our proteomic and other systematic approaches in identifying mitochondrial proteins was evaluated against a reference set of 477 proteins classified as mitochondrial localized based on single gene studies. Of the 546 proteins identified by our proteomic approach, 47% were known mitochondrial, covering 54% of the reference set (256/477). Sorting the 546 candidates by the number of experiments in which they were found demonstrated that the probability of identifying a mitochondrial protein correlated with its detection frequency and with the confidence associated with its identification based on the number of peptide tags identified ( Figure 1 A). A separate analysis of membrane and matrix preparations showed that membrane and matrix proteins were more likely to be identified in membrane and matrix preparations, respectively. In addition, similar proportions of known mitochondrial proteins were identified from both fractions, indicating no significant bias towards the identification of either primarily soluble or primarily membrane-associated proteins ( Figure 1 B). Figure 1 Enrichment for Mitochondrial Proteins by MS (A) shows the 546 proteins (in rows) identified from 28 datasets (columns). The proteins are sorted in decreasing order down rows by the number of experiments in which peptide tags were identified by MS and binned into three classes of detection frequency. The number at the bottom of each class indicates the total number of proteins in the class. Proteins that are part of the reference set, and thus are previously known mitochondrial proteins (M), are marked to the left. The experiments are divided according to fermentable (F) and nonfermentable (NF) mitochondrial preparations. (B) Proportions of proteins identified in membrane and matrix fractions. Whether a protein was detected predominantly in either the membrane or matrix fraction, or equal in both fractions, was determined based on where it was detected with an average higher tag number. Shown are the proportions for all 546 proteins, for known matrix proteins (i.e., matrix and intermembrane space, n = 109), for known membrane proteins (i.e., inner and outer membrane, n = 101), and for detected proteins not previously known to be mitochondrial ( n = 290). (C) Distribution of proteins identified under fermentable and nonfermentable conditions by proteomics, and overlap with previously known mitochondrial proteins. Total numbers are given in parentheses. (D) Breakdown by localization of the 546 proteins identified. For mitochondrial localization the reference set was chosen; for localization outside mitochondria the GFP fusion protein data were used ( Huh et al. 2003 ). The inner circle represents the distribution for all proteins in yeast. (E) Distribution of median mRNA expression under fermentable and nonfermentable conditions, protein abundance under fermentable conditions ( Ghaemmaghami et al. 2003 ), and protein length across bins of confidence of identification (maximum number of tags identified in any of the 28 datasets). The bars indicate fold differences from the median for the known mitochondrial proteins that were not detected by MS (“M not det.”). A comparison between fermentable and nonfermentable growth conditions revealed that more proteins were detected under respiration (448) than fermentation (378) conditions ( Figure 1 C), consistent with the known activation of oxidative phosphorylation during aerobic growth. Notably, of the 477 known mitochondrial proteins, 183 were identified under both growth conditions, suggesting that at least 38% of the mitochondrial machinery is present at moderate to high abundance even under fermentable growth conditions. This finding indicates the presence of a core mitochondrial protein set that exists under multiple growth conditions, which is consistent with previous observations ( Ohlmeier et al. 2003 ). Of the 546 proteins identified by proteomics, 182 proteins are known to localize outside mitochondria, mainly to the cytoplasm, nucleus, endoplasmic reticulum, and plasma membrane ( Figure 1 D). In addition to contaminants copurified with the fractions, identification of these proteins lends further support to the physical interaction of mitochondria with other cellular compartments and the existence of proteins with multiple localizations ( Achleitner et al. 1999 ). In the analysis of complex protein mixtures by MS, low abundance of proteins can preclude their identification ( Patterson and Aebersold 2003 ). This might explain why 46% of the mitochondrial reference set escaped detection (221 proteins). To assess the correlation between protein detection and expression level systematically, we performed genome-wide mRNA expression analysis by means of high-density oligonucleotide arrays under the same fermentable and nonfermentable growth conditions. This analysis showed that absolute mRNA expression levels increased with the known index for establishing confidence of protein identification (tag number; Figure 1 E): while genes identified by proteomics had median expression levels 1.2- to 7.1-fold higher than their unidentified mitochondrial counterparts, they did not differ in protein length, supporting a bias of current proteomic approaches primarily towards the detection of more abundant proteins. We also extended our comparison to the analysis of protein abundance, which was recently determined for about two-thirds of the yeast proteome under fermentation ( Ghaemmaghami et al. 2003 ). To visualize the distribution of identified proteins by their copy number per cell, we divided the 3558 proteins from that study into ten abundance classes, each consisting of an equal number of proteins. We then analyzed the distribution of known mitochondrial proteins across the classes. Figure 2 A shows that we were able to detect known mitochondrial proteins over the whole range of expression levels, from 195 to 519,000 copies per cell. However, there is a clear bias towards the detection of more abundant proteins (i.e., in the highest abundancy class, 82% of the reference-set proteins were identified). A recently published study using multidimensional chromatography, Sickmann et al. (2003) , achieved a higher coverage of known mitochondrial proteins, but the distribution of their identified proteins is also characterized by a bias against proteins of very low abundance ( Figure 2 A). Interestingly, even among the most abundant mitochondrial reference proteins, several remained undetected by either proteomic approach. Some of these proteins have a dual localization for which only a minor amount localizes to mitochondria (i.e., tRNA nucleotidyltransferase or synthases), further supporting the failure of proteomics to detect rare proteins in the samples. Figure 2 Evaluation of Proteomic Data for Protein Abundance and Mitochondrial Localization (A) Coverage of known mitochondrial proteins (Mref) by two MS proteome studies (this study and Sickmann et al. [2003] ). We evaluated the 340 proteins of the mitochondrial reference set for which protein abundance data existed ( Ghaemmaghami et al. 2003 ). The x-axis represents the median protein abundance of ten consecutive, equally sized bins of proteins. (B) Distribution and overlap of proteins identified by the two MS studies and known mitochondrial proteins. The total number of entries for each dataset is indicated in parentheses outside each circle. The number inside each circle indicates the number of proteins in each of the categories. In addition, the percentage of proteins that were localized to mitochondria by GFP tagging ( Huh et al. 2003 ) is given in parentheses for each category. Analysis of the overlap between both proteomic datasets ( Figure 2 B) shows that 337 proteins, corresponding to 62% of our study, were identified by both proteomic approaches, while 209 and 412 proteins, respectively, were present in only one or the other dataset. The majority of the proteins identified by both approaches were already known mitochondrial proteins (71%) or were localized to mitochondria by GFP-fusion proteins (an additional 13%; Huh et al. 2003 ). This high coverage stands in contrast to the much lower number found for proteins detected by only one dataset. Only 23% of the proteins identified by only one method were known mitochondrial proteins. In addition, while 52% of the new candidates (not previously known mitochondrial) identified by both proteomic studies were confirmed by GFP localization to mitochondria ( Huh et al. 2003 ), only 8% and 15% of the candidates identified by only one or the other study were confirmed by the GFP-localization dataset. This analysis suggests that most of the proteins not found by both studies may be nonmitochondrial contaminants. Further indicative of this conclusion is the observation that proteins identified from localization categories outside mitochondria (see Figure 1 D) also were among the high-abundancy proteins in those classes (data not shown). Since mitochondria were purified with different methods in the two proteomic studies, these observations suggest the importance of an integration of approaches. Integration Are there classes of proteins that were not captured by a proteomic analysis, whether integrative or not, but that could be found using different approaches, and vice versa? To address this question, we performed a comparative analysis of functional categories identified by our proteomic dataset in comparison to functional approaches of gene expression analysis and quantitative deletion phenotype screening—datasets which were generated in this study and by Steinmetz et al. (2002) , respectively. In the proteomic dataset, proteins annotated as localized outside mitochondria were not significantly enriched for any of the known functional classes ( Mewes et al. 2002 ; Huh et al. 2003 ). In contrast, known mitochondrial proteins were primarily enriched for known mitochondrial functions such as energy production, transport and sensing, protein fate, and amino acid metabolism ( Figure 3 ). Deletion phenotype screening enriched mainly for proteins involved in genome maintenance, transcription, and translation. Very low enrichment of mitochondrial proteins was achieved by mRNA expression, which predominantly detects proteins involved in energy production, the majority of which seem to localize outside the mitochondrial organelle. Figure 3 Functional Categories and Cellular Localization of Our Proteomic, Deletion, and Expression Datasets Each field shows the proportion of proteins found by the experiment out of the total number of proteins known with a given combination. Fields are color coded by the level of coverage gained by the experiment (color scale upper right). Localization outside mitochondria was based on the GFP fusion protein data ( Huh et al. 2003 ). Fields with less than three identified proteins were not evaluated and left blank. In the upper left corner is the distribution and overlap of proteins identified by each experiment. In parentheses are the known mitochondrial proteins based on the reference set. A comparison of the distribution of protein enrichments shows that different functional categories are targeted by different approaches. Overall, the proteomic approach and the deletion approach identified about equal numbers of previously known mitochondrial proteins; however, they overlapped for less than 30% of the proteins. These observations suggest that combining complementary approaches and an integrative data analysis could be advantageous for predicting new mitochondrial proteins. To improve our comparison of methods and to generate a high confidence list of mitochondrial proteins, we expanded our comparison to a total of 22 datasets relevant to the study of the mitochondrial proteome that had been collected to date. These included the experimental and computational approaches listed in the introduction, including a very recent proteome analysis ( Sickmann et al. 2003 ) and GFP-tag localization study ( Huh et al. 2003 ). For most approaches the mitochondrial candidate genes were taken directly from the publication. From the mRNA expression analyses three datasets were generated. Genes were considered predictive of mitochondrial function if they were differentially expressed between fermentable and nonfermentable growth conditions (our study), differentially regulated in response to the diauxic shift ( DeRisi et al. 1997 ), or differentially expressed in response to Hap4p overexpression (Hap4p is a transcription factor of mitochondrial proteins; Lascaris et al. 2003 ). The protein interaction datasets were screened for genes that interacted with known mitochondrial proteins. Most of the computational predictions searched for signal peptides indicative of mitochondrial targeting sequences. The homology studies searched for proteins similar to, for example, Rickettsia prowazekii, believed to be closest to a common ancestor with mitochondria. The details for each dataset are given in the MitoP2 database (a flatfile with the datasets is also available as Dataset S1 ). We first assessed the performance of each method. For this purpose, the sensitivity and specificity of the different approaches were calculated by comparing each dataset with the reference set of known mitochondrial proteins ( Figure 4 ). This comparison showed that the multidimensional proteomic data ( Sickmann et al. 2003 ) covered 76% of the reference set (sensitivity of 76%), followed by the GFP fusion protein data (69%; Huh et al. 2003 ) and our proteomic dataset (54%). Among the experimental approaches that yielded sensitivity and specificity values of 45% or more were the same three datasets as above, in addition to the deletion phenotype screen ( Steinmetz et al. 2002 ) and another localization study ( Kumar et al. 2002 ). Fifty-three proteins were detected by all five methods, all of which were known mitochondrial proteins. In addition, only 51 proteins of the 477 mitochondrial reference-set proteins were not detected by any of these five methods. In comparison, a comprehensive dataset (union) of all 22 approaches covered 6,324 annotated open reading frames (ORFs) in which all 477 known mitochondrial proteins were included. Figure 4 Specificity and Sensitivity of Systematic Approaches with Regard to Mitochondria Various datasets were benchmarked against the mitochondrial reference set. Each dot in the graph represents an entire dataset: PSORT (Nakai and Harton 1999), hap4 expression ( Lascaris et al. 2003 ), deletion phenotype screen ( Steinmetz et al. 2002 ), tag localization ( Kumar et al. 2002 ), GFP localization ( Huh et al. 2003 ), MitoProt greater than 90 ( Scharfe et al. 2000 ), Bayesian prediction ( Drawid and Gerstein 2000 ), pet phenotypes ( Dimmer et al. 2002 ), three MS proteome studies ( Pflieger et al. 2002 ; Sickmann et al. 2003 ; Ohlmeier et al. 2003 ), mitochondria localized ribosomes ( Marc et al. 2002 ), Predotar ( Small et al. 2004 ), and yeast proteins with known human mitochondrial orthologs (MitoP2 database). High-throughput protein–protein interaction datasets (PPI) were combined and divided into confidence classes ( von Mering et al. 2002 ). Medium and high confidence PPI datasets were defined by interactions with known mitochondrial proteins (MitoP2 database). The predictive score for a mitochondrial protein (MitoP2) was based on the integration of 22 datasets, most of which are shown, and was calculated for different thresholds. Specificity and sensitivity are current best estimates owing to the incompleteness of the reference set. We next set out to determine whether the information supplied by each one of the different methods could be combined to achieve a predictive power that exceeded that of any single approach. We assessed the overlap among different combinations of the 22 datasets and defined a metric for attaching a numerical value to the likelihood of a protein being mitochondrial. A predictive score (MitoP2 score) was estimated based on the specificity of the best combination of approaches: we calculated for each approach as well as for all possible combinations of approaches, the percentage (R) of observed proteins present in the mitochondrial reference set relative to the total number of proteins detected. Most proteins belonged to more than one combination, and for these proteins multiple R values were calculated. The MitoP2 value was chosen to represent the highest R value calculated for a protein, representing the specificity of the best combination of methods. Figure 4 shows that the list of proteins selected with the MitoP2 score yields a sensitivity and specificity higher than those achieved by any single approach. Among 435 proteins with a MitoP2 value greater than 96, 353 proteins were known mitochondrial. Using a MitoP2 value of 90 as a threshold, 691 yeast proteins were found of which 399 were known mitochondrial localized and 292 were new candidates. These data indicate that the power of defining mitochondrial proteins through combining various genome-wide datasets is significantly greater than that of any single method alone, including proteomics and GFP fusion protein localization. Three lines of evidence further support the success of this integrative analysis for defining the yeast mitochondrial proteome. First, the enrichment level for known mitochondrial proteins correlated with the level of the MitoP2 score and the number of experiments in which candidates were identified by proteomics: for the high, medium, and low classes (see Figure 1 A) the median MitoP2 scores were 98, 94, and 82, respectively. Second, MitoP2 prediction was confirmed by import experiments. Ten out of 15 tested candidates with MitoP2 scores greater than 90 were imported into isolated yeast mitochondria, and seven of these were supported with signal sequence cleavage ( Figure 5 ). This ratio (10/15) predicts that 67% of the 292 new candidates could be imported into mitochondria, indicating that 594 of the 691 proteins (with MitoP2 scores greater than 90) may thus be localized to mitochondria (399 plus 195). Third, an investigation of known subunits in mitochondria revealed that most of the components of known complexes were assigned a high MitoP2 score ( Figure 6 ). Comparison with our proteomic dataset showed that while some of the assembly factors of respiratory chain complexes IV and V and subunits of the TIM22 complex were not detected by proteomics, the integrative analysis defined them correctly as mitochondrial proteins. This observation provides further support of the advantage gained by an integrative approach that combines various datasets. Figure 5 Verification of Proteomic Candidates by Mitochondrial Import Samples were incubated in the presence or absence of a membrane potential (MP) and of proteinase K (PK). Cases where import was accompanied by removal of the signal peptide (SP) are marked as “SP-processing” (+). Su9(1–69)DHFR and AAC serve as positive controls for a processed matrix protein and a nonprocessed inner membrane protein, respectively. The bar graphs indicate if a protein was more likely to be found in either the membrane or the matrix fractions of our proteomic data. The height of the bar corresponds to the number of samples in which a protein was identified with higher tag number—in the mitochondrial membrane or mitochondrial matrix fractions, respectively. Figure 6 Verification of Prediction in Selected Mitochondrial Protein Complexes The assignment of complexes to mitochondrial compartments is based on known localizations of the protein subunits. Complexes are shown as clusters of circles, where each circle represents one protein. Red denotes a protein that was detected under fermentable and green under nonfermentable growth conditions by our proteomic dataset; white indicates proteins that were not detected. The numbers indicate the MitoP2 predictive score. For proteins without a number, no predictive score was assigned by the integrative analysis. Ac, acetyl; CoA, coenzyme A; α-KG, α-ketoglutarate; GDC, glycine decarboxylase; NDH, NADH-oxidoreductase; OAA, oxaloacetate; PDH, pyruvate dehydrogenase; RCC, respiratory chain complex; TIM, transport across inner membrane; TOM, transport across outer membrane; MOM and MIM, mitochondrial outer and inner membrane, respectively. A list of the genes for the plotted complexes is available in Table S4 . Implications Our use of mitochondria as a model system for an integrative analysis of a subcellular proteome was aided by the large set of reference proteins known and previous experiments performed. All individual systematic approaches were biased to some extent and incomplete. An integration of data sources is therefore essential to go beyond the limitations of any single method and to achieve a more comprehensive view of the mitochondrial organelle. In similar approaches for other organelles and pathways, the use of reference sets to integrate functional genomic approaches and to define parts lists may prove useful. Most of the mitochondrial reference proteins (399 of 477; 84%) had MitoP2 scores greater than 90, and since we have no evidence for a bias in the current reference set, the mitochondrial proteome as defined by the integration of 22 datasets is nearing saturation. In fact, our integration can be used to obtain an estimate of the number of mitochondrial proteins in yeast. Since outer membrane proteins are often not protease protected, the import analysis is conservative and allows us to estimate a lower boundary for the number of mitochondrial localized proteins. Considering that 84% of the reference proteins had MitoP2 scores greater than 90, we can predict a lower bound estimate of approximately 700 mitochondrial localized proteins in yeast (594 predicted true positives/0.84). This number is at the lower level of previous estimates and indicates that the mitochondrial organelle may consist of fewer proteins than the 800 anticipated ( Westermann and Neupert 2003 ). In order to make a prediction as to which combination of methods may be best applied to study a new system where no prior datasets exist, we performed an analysis of all pairwise combinations of methods. Among the comparisons, the union of proteomics ( Sickmann et al. 2003 ) and subcellular localizations via GFP fusion proteins ( Huh et al. 2003 ) achieved the highest coverage of previously known mitochondrial proteins (sensitivity 87%; specificity 45%). Higher specificity can be achieved by considering the overlap between the two datasets; however, coverage is then severely reduced due to a drastic reduction in gene number (sensitivity 58%; specificity 78%). Union of the two most complementary studies, our proteomics and deletion phenotype datasets—even though they are significantly less exhaustive—also achieved high values (sensitivity 76%; specificity 42%). If we concentrate on datasets that can be generated without massive genetic manipulations, as is required for gene tagging and deletion phenotype approaches, we can achieve a similar sensitivity of 78% with a specificity of 35% through combining in silico predictions (Predotar analysis; Small et al. 2004 ), expression profiling of a transcription factor mutant ( Lascaris et al. 2003 ), and our proteomic data. These data argue that a combination of even a few complementary datasets may identify the majority of expected proteins. A better balance between sensitivity and specificity, however, can be achieved by an integrative analysis of as many complementary approaches as possible. The advantage of integrative analysis combining structural and functional approaches is the high coverage of various mitochondrial components and functions. With this approach we were able to detect with high confidence proteins that had dual localization. For example, Met7p, which was assigned a MitoP2 score of 96, has a cytoplasmic and mitochondrial dual localization ( DeSouza et al. 2000 ). Met7p was not detected as localized to mitochondria in any structural approach, but was identified by the deletion phenotype screen ( Steinmetz et al. 2002 ). Altogether, 40 known mitochondrial reference proteins were not detected by proteomics or by subcellular localization studies. Through the inclusion of functional datasets in the calculations and the use of a localization list as a reference, our candidate list is strongly enriched for mitochondrial localized proteins, but is not limited to those. Consequently, because the MitoP2 calculation is based on both structural and functional datasets, the score not only predicts mitochondrial localized proteins but also reflects proteins that may localize outside mitochondria but affect mitochondrial function and biogenesis from there. It is clear that the current list of mitochondrial proteins is not complete. The addition of further datasets will improve the prediction, as evidenced by the fact that less than 8% of known mitochondrial proteins have a MitoP2 score less than 70. These proteins thus remain rather undefined by the current integration, and further experimentation is needed to capture this class of mitochondrial proteins, consisting in part of three carrier proteins, 12 dual localized proteins, a few small proteins, and 11 mtDNA-encoded proteins (MitoP2 database). Our method of integration serves as one example; other ways of analyzing and integrating the datasets are possible and may reveal more proteins involved in other aspects of the mitochondrial system. Finally, our study has implications for human diseases ( Foury 1997 ). To date, 129 mitochondrial proteins have been implicated in human disorders (MitoP2 database; DiMauro and Schon 1998 ; Wallace 1999 ). The integration in yeast identified 143 new human orthologs of the 292 new yeast mitochondrial candidates defined by a MitoP2 score greater than 90 ( Table S3 and MitoP2 database). This set of 143 proteins provides new candidates for putative human mitochondrial disorders where intervals have been mapped but no responsible gene has been identified to date ( Steinmetz et al. 2002 ). Materials and Methods Purification of mitochondria Saccharomyces cerevisiae strains were grown aerobically at 30 °C in SC or YP medium, and cells were harvested in logarithmic growth phase (OD600 < 1.3). Mitochondria were isolated by one of two different methods. One method involved differential centrifugation followed by a Nycodenz density gradient ( Glick and Pon 1995 ), where the progress of mitochondrial purification was controlled by Western blot analysis using organelle-specific marker protein antibodies. In the other method, isolated mitochondria were purified by zone electrophoresis using a ProTeam FFE Free-Flow Electrophoresis apparatus (Tecan, Grödig, Austria) ( Zischka et al. 2003 ). The anodic and cathodic circuit electrolytes consisted of 100 mM acetic acid and 100 mM triethanolamine acetate (pH 7.4). The electrolyte stabilizer was 280 mM sucrose, 100 mM acetic acid, and 100 mM triethanolamine (pH 7.4). The separation medium was 280 mM sucrose, 10 mM acetic acid, and 10 mM triethanolamine (pH 7.4). The counterflow medium was 280 mM sucrose. Table S1 lists the strains, growth conditions, and purification methods used for each dataset. Prior to FFE fractionation, the mitochondria sample was equilibrated with separation medium and adjusted to a final protein concentration of 1–2 mg/mL. Electrophoresis was performed in horizontal mode at 5 °C with a total flow rate of 280 mL/h within the separation chamber at a voltage of 750 V. The samples were applied to the separation chamber with a flow rate of 1–2 mL/h via the cathodic inlet. Fractions were collected in 96-well plates, and the distribution of separated particles was monitored at a wavelength of 260 nm with a SynergyHT reader (Bio-Tek, Winooski, Vermont, United States). The peak fraction was isolated, shock-frozen in liquid nitrogen, and used for electron microscopy. To assess purity, the preparations were analyzed by electron microscopy. The mitochondrial preparations were fixed with 4% formaldehyde, 2% glutaraldehyde, 4% sucrose, 2 mM calcium acetate, and 50 mM sodium cacodylate (pH 7.2) at 4 °C. The fixed samples were dissected with a scalpel, washed for 1 h in cacodylate buffer with 1% osmium tetroxide, and dehydrated with alcohol in increasing concentrations. After embedding in Araldite, the preparations were cut into 50-nm slices by means of an ultramicrotome (LKB-Produkter, Bromma, Sweden) and then analyzed on a Zeiss (Oberkochen, Germany) EM 10 electron microscope. Fractionation of matrix and membrane proteins Reagents used for the preparation of peptide samples were purchased from the indicated suppliers. Ammonium bicarbonate and methanol were from Fisher Scientific (Fair Lawn, New Jersey, United States). Sodium carbonate, urea, dithiothreitol, and calcium chloride were obtained from Sigma-Aldrich (St. Louis, Missouri, United States). Thiourea, trifluoroacetic acid, and acetonitrile were from Aldrich Chemical Company (Milwaukee, Wisconsin, United States). Sequencing-grade, modified porcine trypsin was obtained from Promega (Madison, Wisconsin, United States). Ammonium formate was obtained from Fluka (St. Louis, Missouri, United States). CHAPS and bicinchoninic acid (BCA) assay reagents and standards were from Pierce (Rockford, Illinois, United States). Purified water was generated using a Barnstead Nanopure Infinity water purification system (Dubuque, Iowa, United States). Purified mitochondrial samples were disrupted using a Mini Beadbeater-8 (Biospec Products, Bartlesville, Oklahoma, United States) for 3 min at 4,500 rpm with 0.1 mm zirconia/silica beads (Biospec Products) in a 0.5-mL, sterile siliconized microcentrifuge tube. The lysed mitochondria, containing membrane and matrix proteins, were removed from the beads through a puncture at the bottom of the microcentrifuge tube, by centrifugation at 16,000 xg for 2 min at 4 °C, and the flow-through was collected in a second microcentrifuge tube. The collected lysate was then centrifuged at 356,000 xg for 10 min at 4 °C to pellet the mitochondrial membranes. The soluble supernatant was used for the study of mitochondrial matrix proteins, and the pellet was retained for identifying mitochondrial membrane proteins. Mitochondrial membrane protein preparation Using a sonication bath (Branson 1510, Danbury, Connecticut, United States), the membrane pellet was resuspended in 50 mM ammonium bicarbonate (pH 7.8) in an ice bath. The resuspended sample was diluted with ice-cold 100 mM sodium carbonate (pH 11.0) and incubated on ice for 10 min. The membranes were then pelleted by ultracentrifugation at 356,000 xg for 10 min at 4 °C. The pelleted membranes were washed using two aliquots of ice-cold water and pelleted again by centrifugation. The BCA protein assay was performed to determine protein concentration. The membrane pellet was resuspended in 7 M urea, 2 M thiourea, 1% CHAPS in 50 mM ammonium bicarbonate (pH 7.8), using vortexing and sonication in an ice bath. Dithiothreitol was added to a final concentration of 9.7 mM in the resuspended sample, and the proteins were then treated with thermal denaturation for 45 min at 60 °C. The denatured and reduced protein sample was then diluted 10-fold with 50 mM ammonium bicarbonate (pH 7.8), and calcium chloride was added to a final sample concentration of 1 mM. Tryptic digestion was performed for 5 h at 37 °C using a 1:50 (w/w) trypsin-to-protein ratio. Snap-freezing the sample in liquid nitrogen quenched the digestion. The tryptic peptides were cleaned using a 1-mL strong cation exchange column (Discovery DSC-SCX , Supelco, Bellefonte, Pennsylvania, United States) per the manufacturer's instructions. The eluted peptide sample was concentrated by lyophilization and a BCA assay was performed to determine final peptide concentration. The peptide sample was stored at −80 °C until time for LC/MS/MS analysis. Mitochondrial matrix protein preparation The BCA protein assay was performed on the soluble matrix supernatant. The proteins were thermally denatured and reduced using 7 M urea, 2 M thiourea, and 5 mM dithiothreitol and incubating at 60°C for 30 min. The denatured and reduced protein sample was diluted 10-fold with 50 mM ammonium bicarbonate (pH 7.8), and the concentration of calcium chloride was adjusted to a final concentration of 1 mM. The tryptic digestion of the protein sample was performed in the same manner as described above for the membrane protein sample. The tryptic peptides were cleaned using a 1-mL LC-18 SPE column (Reversed Phase Supelclean LC-18 SPE, Supelco) per the manufacturer's instructions. The eluted peptide sample was concentrated by lyophilization, a BCA protein assay was performed, and the sample was stored at −80 °C until time for LC/MS/MS analysis. Identification of potential mass and time tags by LC/MS/MS The LC/MS/MS analysis of the tryptically digested peptides was performed as previously reported ( Shen et al. 2001 ). In brief, the high-resolution reversed phase capillary liquid chromatography (LC) system was composed of a column assembled in-house using a 150-μm id × 360-μm od × 65-cm capillary (Polymicro Technologies, Phoenix, Arizona, United States) fixed with a 2-μm retaining mesh and packed with 3-μm Jupiter C18 stationary phase (Phenomenex, Torrence, California, United States). The column was equilibrated with 100% mobile phase A (0.05% trifluoroacetic acid in water) at 5,000 psi. Ten minutes after injecting a 10-μL sample (∼0.5 μg/μL), the exponential gradient began mixing mobile phase A with mobile phase B (0.1% trifluoroacetic acid:90% acetonitrile:9.9% water [vol/vol/vol]) while maintaining constant pressure. Using an in-house-manufactured electrospray ionization source, the capillary LC was interfaced with an LCQ ion trap mass spectrometer (ThermoFinnigan, San Jose, California, United States) with settings of 2.2 kV and 200 o C for the ESI voltage and heated capillary, respectively. The data-dependent tandem MS analysis was conducted using a series of segmented mass/charge ( m/z ) ranges. A collision energy setting of 45% was employed for the collision-induced dissociation of the three most abundant ions detected in each MS scan. Dynamic exclusion was used to discriminate against previously analyzed ions. Peptides were identified by searching the tandem MS spectra against the complete annotated S. cerevisiae genome database (available at http://www.yeastgenome.org/ ) using SEQUEST (ThermoFinnigan) ( Eng et al. 1994 ). “MudPIT” filtering rules were adopted as the acceptance criteria for peptides generated from the SEQUEST results ( Washburn et al. 2001 ). Fully tryptic peptides with a 1+ charge state that had a cross-correlation (Xcorr) factor of 1.9 or greater were accepted. Fully or partially tryptic peptides with a 2+ charge state that had an Xcorr of 2.2 or greater were accepted as well. Peptides with a 2+ charge state that had an Xcorr of 3.0 or greater were accepted. Finally, fully or partially tryptic peptides with a 3+ charge state were accepted if an Xcorr of 3.75 or greater was obtained. Identification of accurate mass and time tags by LC/FTICR Some of the samples analyzed by LC/MS/MS were further analyzed by LC/FTICR. In LC/FTICR, tryptic peptides are analyzed using the same high-resolution reversed phase capillary LC described in the previous section, coupled to an electrospray ionization interface with a Fourier transform-ion cyclotron resonance mass spectrometer ( Smith et al. 2002 ). We used both a custom-made 11.5 Tesla FTICR instrument, designed and constructed in house at Pacific Northwest National Laboratory, and a commercial 9.4 Tesla Bruker Apex III FTICR instrument (Bruker Daltonics, Billerica, Massachusetts, United States). The acquired FTICR spectra (10 5 resolution) were processed and deconvoluted using ICR-2LS (software written in-house at Pacific Northwest National Laboratory) to obtain peak lists containing the monoisotopic mass, observed charge, and intensity of the major ions in each spectrum. The masses were calibrated using the masses of internal calibrant peaks infused at the beginning and end of each LC/FTICR analysis. The peak lists for each analysis were then matched against the potential mass and time (PMT) tags defined previously (see above; by LC/MS/MS analyses among any of the previous samples) using VIPER (software written in-house at Pacific Northwest National Laboratory). The matching involved finding the groups of ions in the data, computing a median monoisotopic mass for each group, and then comparing the mass and elution time of the group with the mass and normalized elution time of each peptide in the PMT tag database (match tolerance of ± 8 ppm and ± 0.05 normalized elution time), resulting in the generation of an accurate mass and time (AMT) tag. Because the PMT tag database consisted only of the peptide tags produced via the previous LC/MS/MS analyses (a PMT tag database for the whole genome does not exist to date), the LC/FTICR analysis could identify only AMT tags which corresponded to previously identified PMT tags from one of the LC/MS/MS runs. Identification of proteins For the purpose of deriving a final list of proteins identified by MS, we included only proteins that had been detected by at least two tags in any single experimental dataset. As such we adapted the rules that are standard for minimizing false positives from MS and defining the detected proteins ( Wu et al. 2003 ). Gene expression profiling Each sample was done in duplicate. Log phase cultures were grown overnight to an O.D. of 1 in 100 mL of YPD, YPL, SCD, or SCL medium. Total RNA was isolated using a hot phenol glass beads protocol. PolyA+ mRNA was purified using Qiagen's Oligotex kit (Qiagen, Valencia, California, United States). Then 4.5 μg of polyA+ mRNA were reverse transcribed to generate single stranded cDNA. Product was fragmented to approximately 50 bp using DNase digestion, biotin end labeled, and hybridized to Affymetrix S98 arrays as described in the Affymetrix user handbook (Affymetrix, Santa Clara, California, United States). Hybridizations were normalized and duplicate samples integrated to arrive at an estimate of absolute transcript abundance using the dChip computational package (Wong Lab, Harvard University). For genes with multiple probe sets on the array, only the probe set with the highest signal was used. For every gene, we calculated the fold difference between fermentable and nonfermentable growth conditions and considered significant only genes with a 1.2-fold or greater difference (either increased or decreased expression). In the final list we included only genes that showed a consistent direction of expression difference (increase or decrease) in both rich and synthetic media conditions. Comparative genomic analysis between yeast and other organisms All-against-all comparison of genes belonging to human, yeast, R. prowazekii , and Encephalitozoon cuniculi genomes has been conducted using the PSI-BLAST algorithm ( Altschul et al. 1997 ). For each PSI-BLAST match, the following information has been stored in the MitoP2 database: the identification numbers of two matching proteins, the BLAST E-value of the match, the coverage of the BLAST alignment (defined as the fraction of amino acids of the shorter protein covered by the alignment), and whether the match is a bidirectional best hit (ortholog). A compendium of the yeast–human bidirectional blast hits for all yeast proteins with a MitoP2 score greater than 90 is given in Table S3 . Prediction of mitochondrial targeting sequences Psort was downloaded locally as a perl5 script (from E-mail: nakai@imcb.osaka-u.ac.jp) . MitoProt was run in the same way as in Scharfe et al. (2000) . Predotar analysis was performed as described by Small et al. (2004) . The protein lists are available in the MitoP2 database. Integration of published datasets and calculation of MitoP2 score To calculate the MitoP2 score, the percentage R of known mitochondrial proteins (reference set of 477 proteins) identified in each single genome-wide experiment (specificity) or in the overlap of all possible combinations of datasets (specificity of the combination of several methods) was calculated. Most proteins belonged to more than one combination, and for those proteins multiple R values were calculated. For example, proteins identified by two approaches received three R values: the specificity of the first approach alone, the specificity of the second approach alone, and the specificity of the overlap of both approaches. The MitoP2 value represented the highest R value calculated for a protein. The relevancy was checked according to the binomial law. The value gives a lower limit of the specificity of a defined combination because the mitochondrial reference dataset is not complete. For more detailed description, please see the MitoP2 database. Protein import into isolated mitochondria For T7 polymerase–driven synthesis of preproteins in vitro, the ORFs were amplified from ATG to STOP-codon by PCR, including the T7 RNA polymerase promoter and transcription initiation site within the 5′ primer. Using reticulocyte lysate (Promega), the resulting PCR products were utilized for coupled in vitro transcription/translation reactions to synthesize preproteins in the presence of 35S-radiolabeled methionine. Mitochondria were isolated by differential centrifugation from yeast strain W334 grown on lactate medium and resuspended at 25 °C in import buffer (0.3 mg/mL fatty-acid-free BSA, 0.6 M sorbitol, 80 mM KCl, 10 mM magnesium acetate, 2 mM KH2PO4, 2.5 mM EDTA, 2.5 mM MnCl2, 2 mM ATP, 5 mM NADH, and 50 mM HEPES/KOH [pH 7.2]). Import was initiated by adding 1% to 4% (vol/vol) of reticulocyte lysate containing radiolabelled preprotein. After 15 min, samples were placed on ice and subsequently treated with proteinase K (50 μg/mL) or not for 15 min to remove nonimported proteins. Protease was inhibited by the addition of 2 mM PMSF. Mitochondria were reisolated and analyzed by SDS-PAGE and autoradiography. Control experiments were performed in the absence of membrane potential in the presence of 1 μM valinomycin and 20 μM oligomycin. Supporting Information Dataset S1 Flatfile with the Integrated Datasets (358 KB TXT). Click here for additional data file. Table S1 Sample Details for Each Proteomic Experiment (34 KB DOC). Click here for additional data file. Table S2 Proteins Identified by MS (544 KB DOC). Click here for additional data file. Table S3 Human Orthologs of Yeast Mitochondria-Related Proteins (828 KB DOC). Click here for additional data file. Table S4 Members of Selected Mitochondrial Protein Complexes (224 KB DOC). Click here for additional data file. URLs The YDPM database is the supporting online database for the proteomic, expression, and deletion datasets discussed in this paper, providing access to data analysis files, candidate lists, and a search function for individual ORFs. Available at http://www-deletion.stanford.edu/YDPM/YDPM_index.html . The MitoP2 database is a mitochondrial proteome database for yeast and human that integrates published datasets and is available at http://ihg.gsf.de/mitop . The database provides annotated ORF information and the MitoP2 scores for the predicted mitochondrial proteins.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC423137.xml
538259	The development of a new measure of quality of life for young people with diabetes mellitus: the ADDQoL-Teen	Background This study evaluated the psychometric properties of the ADDQoL-Teen, an innovative individualised, patient-centred questionnaire measuring perceived impact of diabetes mellitus on quality of life (QoL) of teenagers. Respondents rate all 30 life domains for frequency, and personally applicable domains for 'bother'. Two overview items measure present QoL and diabetes-dependent QoL. ADDQoL-Teen design was based on the ADDQoL (for adults with diabetes). Methods Interviews and discussion groups were conducted with 23 teenagers aged 13–16 years, during work to design the ADDQoL-Teen. The new questionnaire was then completed by 152 young people, (mean age 16.4 ± 2.4 years), attending diabetes clinics at six UK centres. Results Five domains detracted from the measure's reliability and factor structure, four of which were analysed separately and one deleted. The 25-domain ADDQoL-Teen had high internal consistency reliability [Cronbach's alpha = 0.91, (N = 133)] and could be summed into an overall Average Weighted Impact score. There were two subscales: a 10-item Impact-Self subscale (measuring impact of diabetes and its treatment on the individual) and a 15-item Impact-Other subscale (measuring impact on interactions with others and the external world). Both subscales had good internal consistency reliability, [Cronbach's alpha coefficients of 0.82 (N = 142) and 0.88 (N = 138) respectively]. Domains reported as most severely (and negatively) impacted by diabetes were (mean weighted impact ± SD): lie in bed (-3.68 ± 3.41), interrupting activities (-3.5 ± 3.23), worry about the future (-3.45 ± 3.28), career (-3.43 ± 3.15) and sweets (-3.24 ± 3.24), (maximum range -9 to +3). Analysis of the overview items showed that although 72.5% considered that their present QoL was good or brilliant , 61.8% felt that having diabetes had a negative impact on QoL, but 35.6% reported no impact and 2.6% reported a positive impact on QoL. Conclusions The ADDQoL-Teen is a new measure of perceived impact of diabetes and its treatment on QoL of teenagers. It will help healthcare professionals and parents consider QoL issues as well as medical outcomes when caring for young people with diabetes. It may be used in clinical trials and for routine clinical monitoring in a context of continuing evaluation.	Background Increasing numbers of children are being diagnosed with diabetes mellitus [ 1 ] and, once diagnosed, these children and their families face major changes to their lives. However, the emphasis from health professionals is often on control of blood glucose levels and far less consideration is given to the impact of diabetes and the complex daily treatment regimen on each child's quality of life (QoL) and the child's perceptions of the disorder and its management. QoL is an outcome of diabetes management that is important in its own right and the significance of interacting biopsychosocial factors in the management of chronic disorders is recognised [ 2 ]. Thus both psychological and physiological effects of diabetes need to be measured. Measures of the impact of diabetes on the QoL of children are needed, to provide healthcare professionals with information to help protect the QoL of their patients. Such information can be used not only in research, to measure the impact of educational interventions, care provision and treatment regimens on QoL, but also in consultations where completed questionnaires can form the basis of structured discussions between the child, their parents and healthcare professionals. Professionals can be encouraged to be more patient-centred, and help children to overcome the negative impact of diabetes and its treatment on their QoL. Adults may not be able to assess a child's point of view adequately, so children themselves should rate their own QoL wherever possible [ 3 - 5 ], and a more child-centred approach helps clinicians to treat patients successfully [ 4 ] and has produced data which are both valid and informative [ 6 , 7 ]. The questions asked in diabetes-specific adult measures such as the ADDQoL [ 8 , 9 ] are usually too abstract for younger children and/or inappropriate. Existing paediatric diabetes-specific QoL measures do not allow each child to say which aspects of diabetes matter to them personally: they are not sufficiently child-centred or individualised. For example the Diabetes Quality of Life Measure for adults [ 10 ] was simplified and modified to be suitable for adolescents [ 11 ], but children were not involved in the generation of items. The PedsQL [ 12 ], whilst completed by the children themselves, does not use an individualised approach, i.e. it is not possible for the individual to indicate the relevance or importance of a specific aspect of life to his or her QoL. This paper describes the design and subsequent psychometric validation of a new teenager-centred, individualised measure of the impact of diabetes on the QoL of teenagers, the ADDQoL-Teen. The ADDQoL-Teen follows the philosophy underpinning the individualised ADDQoL measure for adults, but ideas in the teenager version are more specific and concrete than the broader, more abstract concepts of the adult version. Methods I. Design of the ADDQoL-Teen questionnaire Four hospitals in the Greater London area participated in the research to design the questionnaire, following Ethical Committee approval. To help identify QoL issues for inclusion in the ADDQoL-Teen measure, clinic sessions were observed, health professionals consulted, and a literature review undertaken. Semi-structured interviews using open-ended questions were then conducted with 10 teenagers with diabetes, and discussions took place with 13 teenagers in small groups of 2–4 teenagers each. The views of 23 young people, aged 13–16 years, were obtained in all. The groups were single-sex as teenagers might be inhibited from talking freely about sensitive issues with members of the opposite sex present. This qualitative research identified important QoL issues that formed the content of 30 items in the new ADDQoL-Teen questionnaire. The items were designed to reflect the teenagers' own perceptions of life with diabetes and measure their individual feelings about the importance of the issues in their everyday lives, rather than being based on researchers' or professionals' opinions. The questionnaire items, response choices and format were based on comments from the teenagers to ensure that the items were child-centred and had face validity for the teenagers themselves. The design of a questionnaire for teenagers was part of a wider study to design child-centred questionnaires for children with diabetes in three age ranges including 5–8 years (ADDQoL-Junior) and 9–12 years (ADDQoL-Junior Plus) [ 13 ]. Description of the ADDQoL-Teen questionnaire In order to produce an individualised questionnaire, the ADDQoL measure of the impact of diabetes on QoL of adults [ 8 ] measures the impact of diabetes on each aspect of life and the importance of that aspect for the QoL of the individual. Design of the ADDQoL was, in turn, influenced by the generic individualised interview measure, the SEIQoL (the Schedule for the Evaluation of Individual Quality of Life) [ 14 ]. In the ADDQoL, adults' impact ratings for each applicable aspect of life (domain) are multiplied by importance ratings to provide a weighted impact score for each domain. In the new ADDQoL-Teen, however, teenagers are asked about the frequency ('a' stem) with which diabetes impacts on each aspect of life, and then how much that particular domain bothers them ('b' stem). The majority of stem 'a'/frequency items are in the format: Do you ever ..... because of your diabetes? and stem 'b'/bother items in the format: Does it bother you when ..... because of your diabetes? (See example in Fig. 1 ). Multiplying frequency and bother ratings for each domain gives a domain weighted impact score. Each item provides an assessment in stem 'a' of whether the aspect of life described is relevant to the teenager, and thus contains a 'no' response option as well as multiple 'yes' response options. Stem 'a'/frequency scoring is 3, 2, 1, 0 (from Yes – a lot ..... No – I do not ). Stem 'b'/bother has response options scoring from -3, -2, -1, 0 ( Yes – it bothers me very much ..... No – it does not bother me , it's OK ) and a positive response option ( No – it does not bother me , I like it ) scoring 1. Figure 1 Example of an ADDQoL-Teen domain item. The ADDQoL-Teen questionnaire has 30 items dealing with specific life domains, in which the wording of item stems and response choices is appropriate to teenagers. Table 1 contains a full description of the wording of each 'a'/frequency stem as well as the item abbreviations that will be used throughout this article. The majority of items have a negative sense but there are three items (7: extra things , 13: out of fix , and 30: holidays ) that have a positive sense. The 'b'/bother stem is scored differently for these three positive items. For example, the responses to item 7b ( How do you feel about having extra things because of your diabetes? ) are scored 3, 2, 1, 0, -1 (from I like having them very much ...... I don't like having them ). Table 1 ADDQoL-Teen item wording and abbreviations No: Abbreviation Overview item A present QoL In general, I feel my quality of life is..... B diabetes-dependent QoL Does diabetes usually make your quality of life worse or better? Full item as in the 'a'/frequency stem 1 others fuss Do you ever feel people fuss or worry about you because of your diabetes? 2 sweets Do you ever feel you want to eat sweets but don't because of your diabetes? 3 drink Do you ever want to drink something but you don't drink it because of your diabetes? 4 eat Do you ever want to eat something but you don't eat it because of your diabetes? 5 insulin Do you take insulin? 6 bleed Do you ever bleed or have any bruises or lumpy bits where you take your insulin? 7 extra things Do you ever have extra things, like snacks, money, treats or days out because of your diabetes? 8 interrupt do Do you ever find diabetes interrupts what you are doing, like watching TV, working at home or school, playing computer games or any other activities? 9 finger tests Do you have finger prick blood tests? 10 control Do you ever feel you want to take more control of diabetes on your own, with less help from other people? 11 moody Do changes in your blood sugars ever make you feel moody? 12 unwell Do you ever feel unwell because of your diabetes, like having a headache or pain, or feeling tired, sick or dizzy? 13 out of fix Do you ever find that having diabetes gets you out of a fix, or gets you out of doing something you don't want to do? 14 sleep away Do you ever get asked to sleep away from home or at a friend's house, but you don't because of your diabetes? 15 wake nights Do you ever wake up in the night feeling hypo with low blood sugar? 16 lie in bed Do you ever want to have a lie in bed, but you don't because of your diabetes? 17 miss events Do you ever miss a party, a school trip, going out or any other event because of your diabetes? 18 low BG Do you ever feel your blood sugar is too low? 19 high BG Do you ever feel your blood sugar is too high? 20 worry future Do you ever worry about the future, like getting married, having children or your future health because of your diabetes? 21 career Do you ever feel that having diabetes will make a difference to your future job or career? 22 different Do you ever feel 'different' because of your diabetes? 23 not allowed Are you ever told that things are 'not allowed' because of your diabetes? 24 family life Do you ever feel that diabetes makes a difference to life with your family or the people you live with? 25 responsibility Do you ever find you are expected to take more responsibility than you would like because of your diabetes? 26 play sport Do you ever find that having diabetes makes any difference to playing sport? 27 go toilet Do you ever find that you need to go to the toilet too often because of your diabetes? 28 social life Do you ever find you need to fit diabetes into your social life, like carrying equipment, planning when to eat, or where to take insulin when away from home? 29 clinic visits Do you go to a diabetes clinic? 30 * holidays Have you ever been to B.D.A holidays or weekends away, or made new friends because of your diabetes? positive item. B.D.A: British Diabetic Association (now known as Diabetes UK). Finally there are two overview/global items: QA: present QoL and QB: diabetes-dependent QoL . QA ( In general, I feel my quality of life is --- brilliant --- good --- OK --- not OK --- bad ) is scored 3, 2, 1, -1, -2 respectively. QB ( Does diabetes usually make your quality of life worse or better? --- a lot worse --- a fair bit worse --- a bit worse --- neither worse nor better --- better ) is scored -3, -2, -1, 0, 1. There is a free comments section at the end of the questionnaire where respondents are asked if there is anything else they would like to say about their life with diabetes. Weighting the items and summation to an Average Weighted Impact score Negative items: The 'a'/frequency ratings in categories scoring 1, 2, and 3 are multiplied by the corresponding 'b'/bother ratings to give a weighted score from -9 to +3 (maximum negative to maximum positive impact of diabetes on a domain). Zero scores on the 'a'/frequency rating are ignored as these items are not applicable to the individual and no 'b'/bother rating is made. The overall ADDQoL-Teen Average Weighted Impact score (ADDQoL-Teen AWI) can be calculated by summing weighted impact scores for all applicable domains before dividing by the number of domains applicable to the individual teenager. ADDQoL Teen AWI varies from -9 to +3, the maximum negative to maximum positive weighted impact of diabetes on overall QoL. Positive items: items 7, 13, and 30 have weighted scores from -3 to +9 (maximum negative to maximum positive impact of diabetes on that domain). The weighting procedure for positive items is similar to that for negative items. Overview items: QA and QB are not included in the calculation of AWI, but analysed individually. II. Study to determine the psychometric properties of the ADDQoL-Teen Patient recruitment In order to determine the psychometric properties of the 30-item ADDQoL-Teen, at least 150 completed questionnaires were needed, as factor analyses ideally require five or more respondents per item [ 15 ]. Young people with Type 1 diabetes mellitus (N = 78) were recruited to an interview study conducted by the National Children's Bureau [ 16 ], and also completed the questionnaire. Another 74 young people were recruited to complete the questionnaire only. Six UK centres were involved, (Centres A to F), of broad geographical spread, and serving diverse communities. Recruitment was undertaken by diabetes specialist nurses. The criteria for inclusion were: the patient was expected to move from paediatric to adult care in the following year or the patient had moved from paediatric to adult care in the previous year. Moving from paediatric to adult care was defined as moving out of the care of the paediatric team. Depending on the size of the caseload in each research area, the nurse either included all patients or a random sample that fitted the criteria in the sampling frame. Ethical Committee approval was obtained for the study to be conducted at all the centres. Statistical analyses The 'No – I do not' response option and loss of data None of the data from any respondent who selected a No – I do not response option (i.e. not applicable, hereafter referred to as "N/A") would normally be included in factor and reliability analyses, as the SPSS statistical package treats N/A responses as missing. Furthermore, if the SPSS default of listwise deletion of missing data is used, all cases that have any missing values across all 30 items are lost to analysis. Results of reliability and factor analyses are therefore reported below with SPSS set to pairwise deletion of missing data, and N/A responses to read as zero, to avoid considerable loss of data. This procedure has been fully described for the original development of the ADDQoL for adults [ 8 ]. Homogeneity of the patient sample There was a risk of systematic differences in responses from the six UK centres creating artefactual correlations within a data set combined to provide sufficient numbers for the psychometric analyses. To check that the final sample was sufficiently homogenous ADDQoL-Teen weighted item scores were converted to standardised z scores for each subgroup, and then recombined. All questionnaire items were forced onto one factor in a Principal Components Analysis of (1) raw weighted scores and (2) recombined z scores, and results compared (a procedure used previously in the original development of the ADDQoL [ 8 ]). Normality issues Normality of distributions was investigated through histograms, box plots and standardised z(skew) values, whereby acceptable z(skew) values between ±2.58 indicate normality [ 17 ]. The ADDQoL-Teen is not a questionnaire where a normal spread of scores and normal distributions would be expected. Respondents were expected to report predominantly negative effects of diabetes with few indicating that diabetes had some positive effects on their lives. The ADDQoL-Teen identifies individuals with extreme responses – the ones most affected by their health condition. Although normality of data is desirable for factor analyses, finding transformations for skewed variables, where N/A was set to zero, that did not adversely affect normal distributions of other items in the questionnaire, proved difficult. The assumption was made that if reliability were high, the factor analysis robust, and the number of respondents sufficiently high, then a degree of non-normality was tolerable. Factor analyses were conducted on data with reflect and inverse transformations, but reliability of non-transformed variables is reported. Internal consistency reliability Cronbach's alpha coefficient [ 18 ] was determined in reliability analyses. Nunnally [ 19 ] regarded an alpha of 0.9 as the minimum acceptable for making decisions about individuals, but 0.8 adequate for comparing groups. Others consider that an acceptable minimum alpha can be 0.7 – 0.8, or even lower for short subscales [ 20 ]. In the present analyses a minimum alpha of 0.9 was regarded as ideal, but alphas above 0.8 were considered very acceptable. Acceptable corrected item-total correlations were those ≥0.2 [ 21 ]. Factor structure Factor structure was explored with Principal Components Analysis, using Varimax rotation. A forced one-factor solution was obtained to confirm the validity of calculating the ADDQoL-Teen AWI score, and unforced analysis to investigate the existence of any subscales. Salient loadings were taken as ≥0.4, higher than the recommended minimum 0.3 [ 22 ], erring on the side of caution in an effort to reduce the risk of spurious loadings that owed their origin to non-normality of item distributions, and also to avoid multiple loadings. Bonferroni correction In exploratory investigations of correlations and subgroup differences in responses, the Bonferroni correction for familywise error was adopted (i.e. alpha was set initially to 0.05/ n where n was the number of variables within a "family") and then the Holm's sequential Bonferroni procedure for multiple tests was applied [ 23 ]. Assessing tolerance of missing data To assess the effects of respondent missing data on the measure's reliability, reliability analyses were run sequentially deleting the strongest item each time, (i.e. deleting the item having the lowest "alpha if item deleted" and therefore contributing most to the internal consistency reliability of the scale, as described elsewhere [ 24 ]. Analysis was conducted using SPSS for Windows (Release 9). Results Patient sample Diabetes services have introduced age-appropriate clinics for teenagers with diabetes to help their transition from paediatric to adult care [ 16 ]. Some services offer adolescent clinics for 14–16 year olds, followed by transition clinics run jointly by paediatric and adult services. Other services run young person's clinics for young people up to 25 or 30 years. The ADDQoL-Teen was originally designed for age range 13–16 years, with the intention that it could be completed by those aged 17–18 years, many of whom would still be attending school. However, 28 of the 152 young people who completed the questionnaire in this study fell outside the 13–18 year age range, of whom 21 were older and 7 younger, and 31 individuals were aged 17 or 18 years. Table 2 shows the ages of the sample, broken down by clinic. There were 72 males (47% of the sample), [mean age 16.79 ± 2.64; range: 12.8 to 24.0 years] and 80 females (53%), [mean age 15.96 ± 2.17, range: 10.4 to 22.7 years]. Over three-quarters (78%) of the sample were attending school or sixth form college at the time of questionnaire completion, 7% were at university, 12% working, and the remainder unemployed. In order to have sufficient data for the psychometric analyses and to evaluate the questionnaire for the wider age range, it was decided to use the data for all 152 respondents, even those who fell outside the 13–18 age range. Key analyses were re-run on the subset of 13–18 year olds to check that results were similar to those from the full data set, although sample size (124) in this age range was less than optimal. Table 2 The sample of 152 young people who completed the ADDQoL-Teen Hospital Male Female Paediatric clinic Adult clinic Mean age Total A 16 13 21 8 16.58 ± 1.28 29 B 21 25 38 8 15.80 ± 2.12 46 C 11 13 17 7 18.27 ± 3.50 24 D 8 14 16 6 14.92 ± 1.27 22 E 10 11 11 10 15.11 ± 1.29 21 F 6 4 10 - 19.44 ± 1.32 10 Total 72 80 113 39 16.36 ± 2.43 152 Questionnaire completion rates Completion rates were very high, providing an indication of the acceptability of the questionnaire to respondents: 'a'/frequency stem items (99.6%); 'b'/bother stem items (99.4%); overview items (98.0%). Homogeneity of the patient sample Initial analyses demonstrated that the six subgroups (recruited from six centres) could be treated as one for the purposes of reliability and factor analyses where a larger N is desirable. Percentage variance accounted for by the loadings on the forced 1-factor analyses was very similar: raw weighted scores (27.52%); standardised z scores for each hospital subgroup recombined (25.62%). Regression analysis found no significant difference between the two sets of loadings: the correlation of 0.987 was close to a perfect 1 (p < 0.001), the constant (0.027) close to zero [t (28) = 1.83, p > 0.05] and the slope (0.916) also close to 1, [t (28) = 32.8, p < 0.0001]. Descriptive statistics Frequency analyses of domains indicated that from 1 to 82% of respondents used the No – I do not (N/A) response option (Table 3 ). The items with the highest frequency of N/A responses were 17: miss events (82%), 14: sleep away (80%) and 30: holidays (72%). Thus the great majority of respondents did not consider that they missed any events, or sleeping over at a friend's house as a result of their diabetes, nor had their diabetes resulted in going away on British Diabetic Association (B.D.A, now known as Diabetes UK) holidays or weekends or making new friends. The areas of greatest importance/frequency of feeling (Response: Yes – a lot in the 'a'/frequency stem) were: 5: insulin (69%), 9: finger tests (49%), 28: social life (26%), and 2: sweets (25%). The areas where the highest percentage of young people considered they were very much bothered were: 12: unwell (25%), 6: bleed , 8: interrupt do and 19: high BG (all 21%), (Table 3 ). The positive response option ( No – it does not bother me , I like it ) was used by up to 19% of respondents: 19% liked going to the diabetes clinic, 7% liked finger prick blood tests, 5% liked taking insulin, and 3% liked other people fussing or worrying about them because of their diabetes (items 29, 9, 5, and 1 respectively). Table 3 Descriptive statistics of ADDQoL-Teen domain items No: Abbreviation N % N/A % 'a'/frequency: Yes, a lot ‡ % 'b'/bother: very much § Weighted impact: Mean ± SD* Median [range] 1 others fuss 148 2.6 21.1 9.9 -2.16 ± 2.52 -2 [-9 to 3] 2 sweets 112 25.3 24.7 16.0 -3.24 ± 3.24 -2 [-9 to 0] 3 drink 80 46.4 6.0 5.3 -2.11 ± 2.31 -1 [-9 to 0] 4 eat 117 21.9 16.6 14.7 -2.62 ± 3.09 -1 [-9 to 1] 5 insulin 151 0.7 69.1 14.5 -2.21 ± 3.27 0 [-9 to 3] 6 bleed 142 6.0 15.9 21.2 -2.39 ± 2.70 -1.5 [-9 to 1] 7 extra things 87 42.8 7.9 - 1.78 ± 2.80 0 [-2 to 9] 8 interrupt do 90 40.0 14.7 20.7 -3.50 ± 3.23 -2 [-9 to 0] 9 finger tests 149 0.7 49.0 19.3 -2.16 ± 3.21 -1 [-9 to 3] 10 control 87 41.3 16.0 5.4 -1.85 ± 2.90 -1 [-9 to 2] 11 moody 125 16.7 19.3 18.4 -2.89 ± 3.04 -2 [-9 to 3] 12 unwell 131 13.2 13.8 24.5 -2.93 ± 2.87 -2 [-9 to 3] 13 out of fix 85 44.1 11.2 - 1.47 ± 2.78 1 [-3 to 9] 14 sleep away 30 80.3 2.0 4.6 -2.23 ± 2.75 -1 [-9 to 3] 15 wake nights 111 27.0 4.6 18.4 -2.08 ± 2.22 -1 [-9 to 2] 16 lie in bed 77 48.3 16.6 14.7 -3.68 ± 3.41 -2 [-9 to 3] 17 miss events 27 82.2 2.6 5.9 -2.67 ± 3.10 -1 [-9 to 1] 18 low BG 120 20.5 6.6 11.3 -1.71 ± 2.18 -1 [-9 to 3] 19 high BG 130 13.2 9.9 20.7 -2.77 ± 2.77 -2 [-9 to 1] 20 worry future 93 38.2 17.1 13.9 -3.45 ± 3.28 -2 [-9 to 3] 21 career 100 34.9 15.8 20.4 -3.43 ± 3.15 -2 [-9 to 2] 22 different 81 46.1 9.9 12.6 -2.72 ± 2.93 -1 [-9 to 0] 23 not allowed 105 30.9 10.5 20.4 -3.16 ± 2.99 -2 [-9 to 2] 24 family life 89 40.8 9.2 13.2 -2.60 ± 3.11 -1 [-9 to 3] 25 responsibility 93 37.7 11.9 6.7 -1.83 ± 2.85 -1 [-9 to 3] 26 play sport 96 36.8 11.2 8.6 -2.17 ± 2.86 -1 [-9 to 3] 27 go toilet 97 36.2 8.6 11.8 -2.14 ± 2.63 -1 [-9 to 3] 28 social life 135 10.6 25.8 17.9 -2.84 ± 3.28 -1 [-9 to 0] 29 clinic visits 148 2.0 23.0 2.6 -0.30 ± 2.08 0 [-9 to 3] 30 * holidays 43 71.5 2.6 - 2.40 ± 2.83 2 [-1 to 9] positive item. max possible range negative items [-9 to 3] and positive items [-3 to 9]. ‡ valid % of Yes – a lot in response to the 'a'/frequency stem. § valid % of Yes – it/they bother/s me very much in response to the 'b'/bother stem. As expected, all negative items showed negative weighted impact of diabetes on the domains, whereas positive items indicated positive impact of diabetes on domains. The most severe negative impact of diabetes was felt (in descending order of impact, means in brackets) for 16: lie in bed (-3.68), 8: interrupt do (-3.5), 20: worry future (-3.45), 21: career (-3.43) and 2: sweets (-3.24) (Fig. 2 ). The least severe negative impact of diabetes was felt for 29: clinic visits (-0.3), 18: low BG (-1.71), 25: responsibility (-1.83) and 10: control (-1.85). Diabetes had the most positive impact on 30: holidays (2.4) (noting that this item was only applicable to 28% of respondents) and 7: extra things (1.78). Overview items found that although the majority (72.5%) considered that their present QoL was good or brilliant (mean 1.79), 61.8% felt that having diabetes had a negative impact on QoL (mean -0.83), but 35.6% considered it had no impact on QoL, and 2.6% that the disorder had a positive impact on QoL (Table 4 ). Figure 2 Mean weighted impact scores of the domains of the 25-item ADDQoL-Teen for the whole sample and 13–18 year age group. Table 4 Descriptive statistics of ADDQoL-Teen overview items No: Abbreviation N Mean ± SD Median [range] A present QoL 149 1.79 ± 0.97 2 [-2 to 3] B diabetes-dependent QoL 149 -0.83 ± 0.88** -1 [-3 to 1] max possible range [-2 to 3]; *max possible range [-3 to 1]. Preliminary factor and reliability analyses of the 30-item ADDQoL-Teen Preliminary factor and reliability analyses were conducted to determine the number of items in the scale that could be summed into the overall ADDQoL-Teen AWI score. Full results are not provided, but these analyses resulted in the decision not to include five items (items 7, 13, 14, 29, 30) in the summation of an overall scale AWI, for the following reasons. The three positive items (7: extra things , 13: out of fix , 30: holidays ) had unsatisfactory loadings, (<0.4), in a forced 1-factor analysis of the 30-item scale. It was decided to omit them from summation of AWI and, as further reliability and factor analyses did not indicate that they formed a subscale, to analyse each of them as separate items. Item 29: clinic visits had a relatively low corrected item-total correlation (0.218), reduced the reliability of the whole scale, and had an unsatisfactory forced 1-factor loading (<0.2). Indeed a high percentage reported that they were not bothered by attending clinic (57%) or that they liked it (19%). Item 29 can also be analysed separately. However, item 14: sleep away had an unsatisfactory forced 1-factor loading (<0.4), and although it contributed to the overall scale reliability, a very high percentage (80%) regarded the domain as N/A and, of those for whom it was applicable, only 8% found that it impacted a lot or a fair bit on their QoL. As the domain was covered by 17: miss events , it was decided to delete item 14 from the scale. All further analyses below were conducted on the 25-item scale. The 25-item ADDQoL-Teen Internal consistency reliability Cronbach's alpha was close to the ideal level of 0.9 (0.913, N = 133). All corrected item-total correlations were >0.37, i.e. well above the acceptable minimum. None of the 25 items would increase the alpha coefficient if deleted from the scale (Table 5 ). Table 5 Reliability analysis of 25-item ADDQoL-Teen (whole sample) Item Scale mean if item deleted Scale variance if item deleted Corrected item-total correlation Alpha if item deleted 1: others fuss -44.67 1412.16 0.4881 0.9102 2: sweets -44.33 1357.63 0.6016 0.9079 3: drink -45.60 1429.77 0.5125 0.9103 4: eat -44.56 1363.22 0.5800 0.9084 5: insulin -44.50 1334.19 0.6447 0.9070 6: bleed -44.37 1391.05 0.5024 0.9099 8: interrupt do -44.56 1364.32 0.5632 0.9088 9: finger tests -44.59 1385.38 0.4451 0.9114 10: control -45.51 1426.39 0.3727 0.9121 11: moody -44.35 1376.36 0.5240 0.9096 12: unwell -44.19 1391.50 0.4842 0.9103 15: wake nights -45.15 1423.99 0.5022 0.9102 16: lie in bed -44.75 1404.51 0.3802 0.9127 17: miss events -46.22 1442.99 0.4281 0.9114 18: low BG -45.28 1430.58 0.4483 0.9109 19: high BG -44.33 1392.77 0.5171 0.9096 20: worry future -44.64 1377.43 0.5054 0.9100 21: career -44.42 1364.17 0.5803 0.9084 22: different -45.20 1362.36 0.7117 0.9062 23: not allowed -44.55 1361.46 0.6364 0.9073 24: family life -45.16 1372.68 0.6134 0.9078 25: responsibility -45.50 1415.25 0.4332 0.9111 26: play sport -45.26 1396.84 0.5138 0.9097 27: go toilet -45.38 1422.21 0.5002 0.9102 28: social life -44.09 1349.43 0.5856 0.9083 Alpha = 0.9129, standardised item alpha = 0.9144 (N = 133). Factor structure A forced 1-factor Principal Components Analysis indicated that all but one item (16: lie in bed ) loaded at ≥0.4 (Table 6 ). However, whilst item 16 loaded slightly low (0.389) it contributed to overall scale reliability and there did not seem sufficient reason to remove it from the scale, especially as descriptive analysis showed this domain to be the most severely impacted by diabetes (mean weighted impact -3.68 ± 3.41). Thus both reliability and factor analyses of the 25-item scale gave support for the calculation of an AWI score by summing the weighted scores of applicable items. Table 6 Forced 1-factor analysis of 25-item ADDQoL-Teen (whole sample) Item Loading 1: others fuss 0.455 2: sweets 0.631 3: drink 0.543 4: eat 0.639 5: insulin 0.620 6: bleed 0.581 8: interrupt do 0.622 9: finger tests 0.475 10: control 0.403 11: moody 0.555 12: unwell 0.498 15: wake nights 0.554 16: lie in bed 0.389 17: miss events 0.447 18: low BG 0.455 19: high BG 0.534 20: worry future 0.500 21: career 0.590 22: different 0.745 23: not allowed 0.650 24: family life 0.646 25: responsibility 0.415 26: play sport 0.546 27: go toilet 0.486 28: social life 0.630 % variance 30.4% Subscales An unforced Principal Components Analysis with Varimax rotation found seven factors (not shown). Items referring to other people/the external world loaded on Factor 1 (e.g. 1: others fuss , 21: career , 22: different , 26: play sport , 28: social life ). Factor 2 was concerned with consumption of food and drink (2: sweets , 3: drink , 4: eat ). Items concerning the effects of diabetes and its treatment on the individual loaded on the remaining five factors in a pattern that was difficult to interpret and with some items double loading. The scree plot indicated two factors. A forced 2-factor analysis gave the clearest factor structure (Table 7 ). Factor 1 contained items that related to the way diabetes and its treatment affected interactions with others and the "external world". It included 1: others fuss , items 2, 3, 4 (consumption of food and drink), and 28: social life . Item 27: go toilet double loaded, loading slightly higher (0.358) on this factor than on Factor 2 (0.329). Factor 2 contained items connected with diabetes, its treatment and effects on the individual, e.g. 5: insulin , 6: bleed , 9: finger tests , and 11: moody . Item 25: responsibility double loaded, slightly higher on Factor 2 (0.295) than on Factor 1 (0.293). The factors accounted for 20.7% and 17.0% of the variance respectively. Table 7 Forced 2-factor analyses of the whole sample compared with the 13–18 age group Whole sample 13–18 years Impact-Other Impact-Self Impact-Other Impact-Self 1: others fuss .571 .545 2: sweets .593 .277 .559 .285 3: drink .530 .518 4: eat .611 .268 .645 .274 5: insulin .354 .540 .341 .542 6: bleed .285 .560 .575 8: interrupt do .511 .359 .451 .499 9: finger tests .480 .471 10: control .482 .389 11: moody .598 .302 .519 12: unwell .757 .714 15: wake nights .270 .538 .273 .534 16: lie in bed .452 .501 17: miss events .516 .578 18: low BG .602 .664 19: high BG .721 .697 20: worry future .570 .527 21: career .495 .329 .547 .272 22: different .664 .371 .637 .382 23: not allowed .670 .654 24: family life .691 .713 25: responsibility .293 .295 .325 26: play sport .463 .298 .496 .365 27: go toilet .358 .329 .412 28: social life .648 .610 Loadings >0.25 are shown, with all loadings >0.4 in bold typeface. The best solution seemed to be that the 25-item scale had two subscales, one relating to the effects of diabetes and its treatment on interactions with others and the external world (the "Impact-Other" subscale) and the second to effects on the individual (the "Impact-Self" subscale). Item 25: responsibility double loaded slightly higher on Factor 2 than on Factor 1, and it was decided to retain this item in the Impact-Self subscale because taking responsibility for diabetes and its treatment will rest increasingly on the individual child as he/she grows older. Domains of others fuss, miss events, career, different, not allowed, family life, play sport and social life , on the Impact-Other subscale, clearly relate to interactions with the others and the external world. The consumption of food and drink very often occurs in a social context. Frequent visits to the toilet (27: go toilet ) or having to stop an activity to inject insulin (8: interrupt do ) may cause embarrassment socially as well as being annoying for the individual. The association of the other items on this scale with the external world is also explicable: item 10: control refers to the individual taking control of diabetes, with less help from other people; and having to get up early in the morning to test/inject may be a major issue, particularly for teenagers, again making the young person with diabetes feel different from others (16: lie in bed ). Eight of the ten items of the Impact-Self subscale clearly relate to the effects of diabetes and its treatment on the individual (domains of insulin, bleed, finger tests, moody, unwell, wake nights, high BG and low BG . As pointed out above, taking responsibility for treatment may have greater impact on the individual child with increasing age (item 25), at the same time the child with diabetes may worry about his/her own future (item 20). Internal consistency reliability of the 15-item Impact-Other subscale was very satisfactory (Cronbach's alpha = 0.883, N = 138), but falling short of the optimal alpha of 0.9. All corrected item-total correlations were satisfactory (>0.38) and only one item (16: lie in bed ) would increase alpha if deleted, and then only by 0.001. Similarly the 10-item Impact-Self subscale also had very satisfactory reliability (alpha = 0.818, N = 142). All corrected item-total correlations were satisfactory (>0.38) and no item would increase alpha if deleted. These analyses confirmed the reliability of the subscales, and gave support for summing the subscale items into their respective subscale total scores. Dealing with missing data The whole 25-item scale was found to be reliable at alpha ≥ 0.9 with maximum one item of missing data and reliable at alpha ≥ 0.8 with up to 10 items of missing data. We recommend that AWI is calculated as the mean of the completed domains with no more than one item of missing data, if the desired alpha level is 0.9, or up to 10 missing values, if the desired alpha level is set at 0.8, which is very acceptable for most research purposes involving group comparisons. The scale is reliable at >0.7 with up to 15 items missing data but we do not advise calculating AWI with this number of missing items, as questionnaire content may well be distorted. The 15-item Impact-Other subscale was reliable at alpha ≥ 0.8 with maximum four items of missing data, but the 10-item Impact-Self subscale was reliable at alpha ≥ 0.8 with no item of missing data. Higher levels of reliability (alpha ≥ 0.9) are required of measures that are being used to compare an individual's scores across time [ 19 ] and for such purposes the full scale score would be needed with no more than one applicable item missing (excluding N/A items). ADDQoL-Teen AWI and subscale scores Analysis of the data for the whole sample found that mean overall ADDQoL-Teen AWI was -2.39 ± 1.68, mean Impact-Other was -2.44 ± 1.86 and mean Impact-Self was -2.31 ± 1.86, (maximum possible range -9 to 3) implying that young people perceived that diabetes had a negative impact on their QoL, on interactions with others and the external world, and on themselves. Sex differences There were no significant sex differences in ADDQoL-Teen AWI and subscale scores after a Bonferroni correction requiring significance of p = 0.017 or less for that family of variables. However, the sex difference in Impact-Self approached significance (p = 0.028) on a Mann-Whitney test, with female respondents tending to show greater perceived negative impact of diabetes on self-related factors (-2.6 ± 1.85) than did male respondents (-1.99 ± 1.84). Considering the 25 ADDQoL-Teen items as another group (with Bonferroni correction requiring minimum significance of p = 0.002), sex differences in 6: bleed reached significance. Female respondents showed significantly greater perceived negative impact of having bleeding or bruising at site of insulin injection (-3.01 ± 2.92) than did males (-1.65 ± 2.21) [U = 1764.5, p = 0.002, 2-tailed]. Sex differences also approached significance for 20: worry future (p = 0.011), and 22: different (p = 0.043) and, considering the three positive items as another family of variables, for 7: extra things (p = 0.026). Compared with males, females showed a tendency towards greater perceived negative impact of diabetes on feeling different from peers, worries about the future, but greater positive impact on getting extra things because of their diabetes. Correlations with age Small but significant positive correlations with age were found for AWI and the two subscales (Table 8 ) indicating lessening impact of diabetes on overall QoL as measured by the ADDQoL-Teen, lessening impact of diabetes on relationships with others and external world (Impact-Other) and on self-related factors (Impact-Self) with increasing age. There were also significant positive correlations with age for the two overview items, indicating improving present QoL with increasing age, and lessening impact of diabetes on QoL. Moderate correlations were found between ADDQoL-Teen AWI and the overview item QB: diabetes-dependent QoL (rho = 0.49), and a smaller correlation, as expected, with overview item QA: present QoL (rho = 0.34). Table 8 Correlations between ADDQoL-Teen AWI, subscales, overview items and age at completion of questionnaire (whole sample) Age AWI ADDQoL-Teen AWI 0.21 (p = 0.01) Impact-Other subscale 0.22 (p = 0.006) 0.90 (p < 0.001) Impact-Self subscale 0.16 (p = 0.043) 0.85 (p < 0.001) QA: present QoL 0.19 (p = 0.02) 0.34 (p < 0.001) QB: diabetes-dependent QoL 0.26 (p = 0.002) 0.49 (p < 0.001) N (range) 149 – 152 149 – 152 All correlations are 2-tailed non-parametric Spearman's rho, and significant after Bonferroni corrections applied. The 13–18 year age group The mean age of those in the 13–18 year age group was 15.82 ± 1.47, a little less than that of the whole sample (16.36). Mean weighted impact scores of the younger group were very similar to those of the full sample (Fig. 2 ). The most negatively impacted domains, in descending order (mean ± SD) were: 20: worry future (-3.53 ± 3.36), 16: lie in bed (-3.51 ± 3.32), 8: interrupt do (-3.48 ± 3.33), and 23: not allowed (-3.4 ± 3.12). A forced 1-factor analysis of the scores of the 124 teenagers in the 13–18 age range on all 30 items, found support for excluding the same five items from the scale as described above for the whole sample (i.e. the three positive items, and items 14 and 29). All 25 ADDQoL-Teen items loaded >0.4 on a forced 1-factor analysis except 10: control and 25: responsibility (loading at 0.356 and 0.394 respectively, full results not shown). Regression analysis found no significant difference between the forced 1-factor loadings for the subset of 13–18 year olds and those for whole sample (N = 152). The correlation of 0.954 was close to 1, the constant (0.027) was close to zero [t (23) = -0.71, p > 0.05] and the slope (1.04) was also close to 1, [t (23) = 15.21, p < 0.001]. This high correlation indicated that data from the whole sample could substitute for that from the narrower age range. Table 7 compares loadings obtained from the forced 2-factor analyses of the 13–18 year age group with those of the whole sample. The loadings are very similar, except that the double loading of 8: interrupt do is higher on Impact-Self with the 13–18 year group, perhaps implying that the younger age group may have less responsibility for deciding on whether to interrupt an activity because of their diabetes, and this is seen as impacting more on the self than on others; and 10: control loads less than optimally (0.389) on Impact-Other in the 13–18 age group. 27: go toilet loads >0.4 on Impact-Other in the 13–18 age group, but double loads with the wider age range. Cronbach's alpha of the whole 25-item scale was 0.9132, (N = 106) and only 10: control would marginally increase alpha if deleted (0.9133). All corrected item-total correlations were satisfactory. The scale was found to be reliable at 0.9 with up to two items missing and reliable at 0.8 with up to 10 items missing. The 15-item Impact-Other subscale had good internal consistency reliability (Cronbach's alpha = 0.887, N = 111) and was reliable at 0.8 with up to four items missing. The 10-item Impact-Self subscale was reliable (alpha = 0.805, N = 114) if no items were missing. All corrected subscale item-total correlations were satisfactory. Note: Cronbach's alpha for the sample in the 13–18 age range was only marginally lower (by 0.005) than that for the narrow 13–16 age range (0.918, N = 76), again indicating that the addition of respondents aged 17–18 years is not harmful to the questionnaire's reliability. Free comments section The free comments section at the end of the ADDQoL-Teen was used by 49 young people in all. The majority of respondents' comments emphasised a response that they had already made to a questionnaire item. The following areas were mentioned by at least four individuals and are not directly covered in the questionnaire. Consideration will be given to adding further items to cover these new areas in the future: • The effect of diabetes on patient's lives, and having to organise/plan life around diabetes and its treatment (nine respondents). • Other people, including healthcare professionals, not understanding diabetes and its effects on the young person's life (five respondents). • Concerns about weight, and difficulty losing weight (four respondents). Although 17–18 year olds were not included in the focus groups at the questionnaire design stage, analysis of the free comments showed that only five of the 49 respondents offering comments fell outside the narrower age range of 13–16 years: four young people were aged 17, and one was within a few days of their 13 th birthday. However, each of these four 17 year olds commented on a different aspect of life with diabetes not already covered by the questionnaire (i.e. there was no salient aspect missing from the questionnaire on which all four commented). If the questionnaire was not suitable for those aged 17–18, and was missing important domains for these older respondents, it is very likely that a greater number of older respondents would have taken the opportunity to comment at this point. We can be reassured therefore that the questionnaire is suitable for the older age group (17–18 years), even though the measure was not specifically piloted with them. Discussion The ADDQoL-Teen is a new child-centred, individualised questionnaire measuring the impact of diabetes and its treatment on the QoL of teenagers. The items not only reflect the concerns of teenagers with this condition, as expressed in interviews and focus groups, but also use teenagers' wording where possible. Twenty-five of the life domains form a scale with excellent internal consistency reliability. Summation of the weighted impact scores from the applicable items into a single score, the ADDQoL-Teen AWI, gives a measure of the Average Weighted Impact of diabetes on the QoL of the individual. There are two subscales: the 15-item Impact-Other subscale, measuring the impact of diabetes and its treatment on interactions with others and the external world, and the 10-item Impact-Self subscale, measuring the impact of diabetes and its treatment on the individual. Both subscales have good internal consistency reliability. The two overview items (QA and QB) provide global measures of an individual's present QoL, and the perceived impact of diabetes and its treatment on their QoL respectively and, as expected, QB has a higher correlation with AWI than QA, as both QB and AWI measure impact of diabetes on QoL. Of the original 30 items, one item, concerning sleeping away from home, was deleted from the scale as it detracted from scale reliability and factor structure, and was not applicable to the great majority of respondents. Four items, three of which concerned potential positive aspects of diabetes such as getting extra things like snacks or treats, either did not load well with the 25 items in the single scale, or detracted from reliability, but can be analysed individually. Despite the majority describing their present QoL as good or brilliant , young people perceived overall negative impact of diabetes on QoL (AWI), including negative impact on interactions with others and the external world (Impact-Other), and on themselves (Impact-Self). However, interesting information can also be gleaned by analysing frequencies of individual domains. Domains reported as most severely (and negatively) impacted by diabetes were lie in bed , interrupt do , worry future , career and sweets . These show the particular concerns of young people about not being able to stay in bed in the morning like many of their contemporaries, owing to the demands of the diabetes treatment regimen, and the way that this treatment regimen interrupts their normal day-to-day activities. Respondents were also looking to the future and were concerned about their career prospects, getting married, having children, and their longer-term health. The impact of diabetes on consumption of carbohydrates was most notable in relation to eating sweets. The usefulness of the questionnaire's bi-polar scale was indicated by the numbers of individuals who chose a positive response: almost a fifth of respondents liked attending their diabetes clinic, and perhaps a surprising number liked taking insulin or doing finger prick blood tests (5% and 7% respectively). It was also interesting to note that concerns about having a low blood glucose level had the least negative impact on QoL of any of the domains, although this aspect of diabetes is of major concern to healthcare professionals. Some sex differences were found. Girls and young women showed significantly greater perceived impact of experiencing bleeding or bruising at the site of insulin injection, and there was a non-significant tendency for females to show greater perceived negative impact than males with respect to feeling different from peers, and worries about the future, but greater positive impact on getting extra things because of their diabetes. With increasing age, correlations indicated reduced perceived negative impact of diabetes on overall QoL (AWI), on relationships with others and the external world (Impact-Other) and on self-related factors (Impact-Self). Present QoL also improved with increasing age. The moderate correlation between ADDQoL-Teen AWI and the overview item QB: diabetes-dependent QoL was too low (rho = 0.49) to allow the single overview item to replace the 25-item scale for most purposes. Content validity was also good: relatively few respondents mentioned new domains in the free comments section at the end of the questionnaire. However, consideration will be given in the future to adding further items to cover new areas: organising life around diabetes, other people's understanding of the condition, and concerns about excess weight. The teenagers involved in interviews and focus groups during work to design the questionnaire lived in and around London. However, the respondents in the questionnaire study were from six areas in Britain, and there were clear indications of acceptability to all in terms of very high completion rates, and that neutral, non-regional vocabulary had been chosen. In order to have sufficient data for the psychometric analyses it was necessary to use the data for all 152 respondents, even those who fell outside the age range for which the questionnaire was originally designed (13–16 years). Nevertheless, completion rates indicated the acceptability of the questionnaire to a much wider age range than that for which it was originally intended. This is a valuable outcome, as there is considerable variability in age range at paediatric, adolescent, transition and adult diabetes clinics between different diabetes services in the UK. Indeed the mean age of respondents from one of the centres in the present study, a paediatric clinic, was 19.4 years. The questionnaire can also be recommended for 13–18 year olds, as analyses performed on the subset of data for this age group found results very similar to those for the full data set. Although the sample size (124) in the 13–18 year age range was less than optimal, the factor structure was clear and very similar to that of the wider age range, and the full 25-item scale and two subscales also had very good internal consistency reliability. There appeared to be some slight differences in mean weighted impact scores between the two groups (Fig. 2 ). The negative impact of not being allowed to do things because of diabetes was higher in the 13–18 year age group (who had a lower mean age), as was the negative impact on diabetes on eating, and of having to take more responsibility than they would like. As might be expected, those in the 13–18 age group also perceived greater negative impact of not being allowed to do things because of their diabetes, and also for high blood glucose levels (glycaemic control often deteriorates in adolescence [ 25 ]). Not being able to lie in bed was the most negatively impacted of all domains for the whole sample, and the second most extreme response for the younger age group. Both groups were concerned about the future and the effects of diabetes on their careers. Moreover, there was no evidence from analysis of free comments that the measure was unsuitable for 17–18 year olds, as only four representatives of this age group took the opportunity to comment here, and no aspect was mentioned by more than one of these older respondents. We would not recommend that the measure is used above the age of 18, unless the cognitive development of the young person seemed to indicate that the equivalent adult measure, the ADDQoL, were unsuitable. However, if a hospital has young people over 18 years in its adolescent clinic, clinicians might welcome a measure that has been found in practice to be suitable for young people above this age cut-off when conducting studies on their patients. Although physiological measures are used to monitor the treatment of children and young people with diabetes, there are no child-centred, individualised psychological instruments currently in use in paediatric clinics that measure the impact of diabetes on children's everyday lives and on their QoL. The ultimate aim of QoL measurement is to improve patients' QoL wherever possible, by taking into account the impact of the treatment regimen and the effects of diabetes on their experience of daily living. Use of the ADDQoL-Teen would facilitate understanding of these issues and would provide healthcare professionals with valuable information about the psychosocial effects of diabetes on teenagers' everyday lives, which will help them consider psychological issues as well as medical outcomes when caring for teenagers with diabetes. Children (and parents) are faced with the day-to-day responsibility for the management of diabetes, and any improvements in QoL, whilst welcome in themselves, may also mean that these young patients will be more likely to follow the planned treatment regimen which will, in turn, help improve control of blood glucose levels and contribute to a reduction of long-term complications of diabetes. Conclusions The internal consistency reliability and some aspects of the validity of the new child-centred, individualised ADDQoL-Teen have been established for young people with diabetes, and the measure may be recommended for use with individual patients. The new questionnaire should help health professionals to consider psychological issues as well as medical outcomes when caring for young people with diabetes. The instrument is also expected to be useful in evaluating new treatments and educational interventions for diabetes in clinical trials. Authors' contributions CB and RJH designed the ADDQoL-Teen, with RJH conducting the interviews and focus groups with teenagers that informed the design of the measure. JD and NJHM conceived and designed the interview study and JD conducted the interviews and collated the questionnaires. CVM carried out the psychometric and statistical analyses of questionnaire data and drafted the manuscript. CB contributed to the interpretation of psychometric analyses, decision-making regarding item selection, and manuscript preparation. All authors read and approved the final manuscript. ADDQoL-Teen copyright For access to and a licence to use the ADDQoL-Teen, contact the copyright holder, Clare Bradley PhD, Professor of Health Psychology, Health Psychology Research, Royal Holloway, University of London, Egham, Surrey, TW20 0EX. Email: c.bradley@rhul.ac.uk	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC538259.xml
538271	Validation of a prefractionation method followed by two-dimensional electrophoresis – Applied to cerebrospinal fluid proteins from frontotemporal dementia patients	Background The aim of this study was firstly, to improve and validate a cerebrospinal fluid (CSF) prefractionation method followed by two-dimensional electrophoresis (2-DE) and secondly, using this strategy to investigate differences between the CSF proteome of frontotemporal dementia (FTD) patients and controls. From each subject three ml of CSF was prefractionated using liquid phase isoelectric focusing prior to 2-DE. Results With respect to protein recovery and purification potential, ethanol precipitation of the prefractionated CSF sample was found superior, after testing several sample preparation methods. The reproducibility of prefractionated CSF analyzed on 2-D gels was comparable to direct 2-DE analysis of CSF. The protein spots on the prefractionated 2-D gels had an increased intensity, indicating a higher protein concentration, compared to direct 2-D gels. Prefractionated 2-DE analysis of FTD and control CSF showed that 26 protein spots were changed at least two fold. Using mass spectrometry, 13 of these protein spots were identified, including retinol-binding protein, Zn-α-2-glycoprotein, proapolipoproteinA1, β-2-microglobulin, transthyretin, albumin and alloalbumin. Conclusion The results suggest that the prefractionated 2-DE method can be useful for enrichment of CSF proteins and may provide a new tool to investigate the pathology of neurodegenerative diseases. This study confirmed reduced levels of retinol-binding protein and revealed some new biomarker candidates for FTD.	Background Frontotemporal dementia (FTD) accounts for up to 20% of presenil dementia cases [ 1 ] and is, after Alzheimer's disease (AD), the second most common form of early onset dementia (at age < 65 years) [ 2 ]. The clinical picture in FTD is characterized mainly by changes in personality and social behavior, signs of disinhibition, lack of insight and changes in eating preferences [ 1 ]. Memory disturbances, which prevail in AD, may also be found in FTD but not usually to the same extent [ 3 ]. Post-mortem pathological examination reveals bilateral atrophy of the frontal and anterior temporal lobes in FTD and the ventricular system is sometimes widened frontally [ 4 ]. The histological findings provide a basis for the division of FTD into various subtypes. Neurofibrillary tangles, a prominent neuronal accumulation of hyperphosphorylated and filamentous forms of the microtubule associated protein tau, are found in FTD with Parkinsonism linked to chromosome 17, the hereditary variant of FTD, caused by mutations in the tau gene [ 5 ]. Another FTD variant, Pick's disease, is characterized by the presence of neuronal inclusion bodies called Pick bodies, containing filamentous tau and ubiquitin aggregates. The most common type of FTD is frontal lobe degeneration of non-Alzheimer type, which is clinically indistinguishable from Pick's disease, and histologically characterized by neuron loss and gliosis in the absence of distinctive histopathology, such as neurofibrillary tangles or other intracellular inclusions [ 4 ]. The diagnosis of FTD is often difficult and would be greatly enhanced by the use of disease specific neurochemical markers [ 6 ]. Several neuro-specific proteins in the cerebrospinal fluid (CSF) of FTD have been investigated [ 7 , 8 ] and elevation of cytoskeleton markers such as neurofilament light protein and tau have been found [ 7 - 10 ]. In order to expand the search for diagnostic biomarkers, which would also lead to a better understanding of the pathophysiological mechanisms of neurodegeneration, two-dimensional electrophoresis (2-DE) investigations of the CSF have previously been performed [ 11 - 15 ]. 2-DE can effectively separate several proteins and their isoforms simultaneously and is a useful tool for identifying quantitative and qualitative protein differences between the diseased and normal state. Previous proteomic studies by our group have shown for the first time that several proteins involved in FTD pathology are not effected in the CSF of AD patients and vise versa, thus establishing a likely difference in the pathophysiological mechanism between FTD and AD [ 11 , 12 ]. Some abundant proteins, for example albumin and immunoglobulins, limit the total amount of CSF proteins that can be loaded on the 2-D gel, resulting in difficulties detecting low abundant proteins of CSF. By using liquid phase isoelectric focusing (LP-IEF) as a prefractionation step prior to 2-DE we have previously shown that less abundant CSF proteins can be enriched, thus making them more easily detected and identified by mass spectrometry (MS) [ 16 ]. The advantage of this method is that a larger volume of CSF (3 ml) can be used as starting material and that proteins outside the selected pH interval of the 2-D gel can be excluded. For several years alternative prefractionation methods prior to 2-DE have been reported [ 17 - 22 ], each with its advantages and disadvantages. The aim of the present study was to improve the prefractionation procedure for individual CSF samples and to determine its reproducibility. Moreover, the second aim of this study was to apply the method and further explore disease-influenced proteins in CSF from FTD patients compared to controls. Results Preparation of prefractionated CSF samples prior to 2-DE Several strategies to reduce impurities, i.e. salt and ampholytes, of the prefractionated CSF samples were performed including, trichloroacetic acid (TCA)-acetone precipitation, ethanol precipitation, chloroform/methanol/water precipitation and, micro Bio-Spin desalting column. The protein pattern on the 2-D gels and protein recovery from the different sample preparation methods were compared to that of acetone precipitation, previously used prior to prefractionated 2-DE of CSF [ 16 ]. We found that ethanol precipitation gave the best result and allowed us to reduce the focusing time from 25 000 Vh to 20 000 Vh, as recommended by the Protean cell manufacturer. Evaluation of gels with ethanol precipitated sample showed that the same protein spots were present, or even a few more, with similar intensities as in the gels with the acetone treated CSF-samples (figure 1a and 1b ). Furthermore, the optimal ethanol concentration was tested showing that a concentration of at least 70% ethanol was sufficient to generate a total protein recovery, determined by the RC DC protein assay (figure 1c ). An ethanol concentration of 71.25% (3/4 95% ethanol and 1/4 sample) was used in the subsequent studies. Figure 1 Comparison of different protein precipitation methods . a) 75% acetone precipitation and b) 70% ethanol precipitation of prefractionated SYPRO Ruby stained CSF proteins with pH 4.5–6.0, separated on IPG-strips, pH 4–7. The molecular weights (M W ) are in kDa. c) The protein recovery as % of control (untreated CSF) after precipitation with indicated concentrations of ethanol and 75% acetone. The protein recovery was measured using the RC DC protein assay (Bio-Rad). Both TCA-acetone and chloroform/methanol/water precipitation substantially reduced the number of spots and resulted in streaky 2-D gels (data not shown). The use of biospin columns gave well-focused 2-D gels but several protein spots were reduced or lost (data not shown). Reproducibility of prefractionated and unfractionated CSF on 2-D gels Four identical CSF samples were individually prefractionated by LP-IEF. Fractions 6–9 (having a pH of 4.5–6.0) were pooled and analyzed on pH 4–7 strips in four replicates. Coefficients of variation (CVs) were calculated for the protein spots quantities, determined by the PD-Quest software, and a selection of 20 spots had CVs ranging from 1–33% (15.4% ± 8.3%, mean ± SD) (Figure 2a ). The spots were selected to represent a broad range of proteins present in each replicate, i.e. proteins of different molecular weights and isoelectric points (pIs), low, medium and high abundant proteins as well as different isoforms of the same protein. The reproducibility of direct 2-DE of CSF, using pH 4–7 gels, in four replicates was also determined and the CVs of a similar selection of 20 spots ranged from 1–35% (14.6% ± 7.9%, mean ± SD) (Figure 2b ). Figure 2 Reproducibility study of direct and prefractionated 2-DE of CSF on SYPRO Ruby stained 2-D gels . a) Represents a standard 2-D gel image of prefractionated (using LP-IEF) CSF fractions with pH 4.5–6.0, separated on a pH 4–7 IPG-strip. Numbers represents the CVs of encircled protein spots from four individually prefractionated, identical CSF samples separated by 2-DE. The mean CV of the 20 marked spots is 14.6% ± 7.9% (mean ± SD). b) Represents a standard 2-D gel image of directly analysed (after acetone precipitation) CSF proteins, separated on a pH 4–7 IPG-strip. Numbers represents the CVs of encircled protein spots from four replicate gels. The mean CV of the 20 marked spots is 15.4% ± 8.3% (mean ± SD). Molecular weights (M W ) are in kDa. Comparison of direct and prefractionated 2-DE Comparison of direct and prefractionated 2-DE showed that the prefractionated gels contained more spots with a higher protein concentration (Figure 3 ). The spots in the pH 3–6 and pH 5–8 gels in particular were increased in both number and density after prefractionation. Figure 3 Comparison of direct and prefractionated CSF on SYPRO Ruby stained 2-D gels . The upper figures represents standard images of direct 2-DE, and the lower figures represents prefractionated 2-DE. The pH interval of the IPG strips is denoted in the upper left corner of the gels and the pH range of the prefractionated CSF samples at the bottom of the gel images. Molecular weights (M W ) are in kDa. A proteomic study comparing prefractionated CSF from FTD patients and control subjects The prefractionated 2-DE method was used to screen for changes in the CSF proteome of five FTD patients compared to that of five non-dementia controls in the pH intervals 3–6, 4–7, and 5–8 (Figure 4a,4b,4c ). Figure 4 Prefractionated CSF from FTD patients compared to non-demented controls . Protein densities increased (squares) or decreased (circles) at least two times in prefractionated FTD CSF, analyzed using SYPRO-Ruby stained 2-DE gels. The five FTD patients were 70.6 ± 5.6 (mean ± SD) year-of-age and the five non-demented controls were 59.2 ± 11.9 (mean ± SD) year-of-age. The numbers on the 2-D gel pattern correlate each identified protein to the data given in table 2. Molecular weights (M W ) are in kDa. a) Fraction 2–5 with pH 1.5–4.5 was analysed using a pH 3–6 linear IPG-strip. b) Fraction 6–9 with pH 4.5–6.0 was analysed using pH 4–7 linear IPG-strip. c) Fraction 10–14 with pH 6.0–7.5 was analysed using pH 5–8 linear IPG-strip in the first dimension. Comparing the protein densities of the gels, 10 spots were up regulated and 16 spots down regulated, at least two fold, in FTD patients compared to non-dementia controls. Increased proteins are marked with a square and decreased with a circle (Figure 4a,4b,4c ). Thirteen of the protein spots, corresponding to 7 different proteins, were identified by mass spectrometry (MS) Table 1 ). In the FTD group the following protein spots were increased: One isoform of Zn-α-2-glycoprotein (ZAG), proapolipoproteinA1 (ProapoA1), β-2-microglobulin (β-2-m) and two isoforms of transthyretin (TTR), while a reduction was seen in four isoforms of serum albumin, two isoforms of alloalbumin and retinol binding protein (RBP), compared to controls (Table 1 ). Table 1 CSF proteins increased or decreased, at least two fold in FTD vs. control Spot no. Protein identity NCBI Acc. no. Theor. Mw (kDa) Theor.pI No. peptides matched Seq. cov. (%) Levels in FTD vs. control FTD spot norm. density (mean ± SD) Control spot norm. density (mean ± SD) 1 Zn-α-2-glycoprotein 141596 31.6 5.70 4 22 ↑ 42256 ± 7227 4189 ± 979 2 proapolipo- protein A1 178775 28.9 5.45 14 43 ↑ 11572 ± 10432 4575 ± 2309 3 retinol- binding protein 20141667 20.9 5.27 4 35 ↓ 3680 ± 1675 8131 ± 4857 4 serum albumin 113576 52.0 5.69 11 22 ↓ 1819 ± 1657 7272 ± 1812 5 serum albumin 113576 52.0 5.69 11 22 ↓ 2415 ± 964 7839 ± 1906 6 serum albumin 113576 66.0 5.69 12 20 ↓ 1003 ± 309 3098 ± 1428 7 serum albumin 113576 66.0 5.69 9 15 ↓ 1528 ± 587 4896 ± 2551 8 alloalbumin 178345 69.2 5.99 12 19 ↓ 1361 ± 530 7454 ± 1748 9 alloalbumin 178345 69.2 5.99 12 19 ↓ 2463 ± 806 8395 ± 1475 10 retinol-binding protein 20141667 20.9 5.27 4 35 ↓ 1744 ± 551 5871 ± 2040 11 transthyretin 339685 13.8 5.3 8 81 ↑ 45198 ± 26202 9149 ± 2685 12 transthyretin 339685 13.8 5.3 8 81 ↑ 303002 ± 72750 147732 ± 30928 13 β-2-microglobulin 4757826 12.9 5.77 5 46 ↑ 41938 ± 43510 12287 ± 3161 The proteins were identified by MALDI-MS. The spot numbers refer to those given in figure 4. Discussion In this study we present an improved method for increased detection of CSF proteins by a combination of LP-IEF and 2-DE, followed by SYPRO Ruby protein staining and protein identification by mass spectrometry, for investigation of protein differences in CSF of FTD patients compared to controls. To our knowledge no other prefractionation method combined with 2-DE has so far been developed and evaluated for CSF proteins. The study showed that the reproducibility of prefractionated 2-D gels could be compared to that of direct 2-D gels, indicating that the extra prefractionation step did not introduce additional variation and could be reproduced from sample to sample. The protein detection and quantitative reproducibility of Coomassie Brilliant Blue [ 23 ], silver, [ 24 , 25 ] and SYPRO Ruby [ 26 , 27 ] staining of direct 2-DE gels has previously been described. In one SYPRO Ruby study [ 26 ] the reproducibility of the quantification of 20 proteins, selected to represent well matched proteins of different molecular weight and intensity, from four replicate gels, had CVs ranging from 3 to 33%. This is in agreement with our findings using a similar selection of 20 proteins, where the prefractionated 2-DE CVs ranged from 1–33% (mean 14.5%) and the direct 2-DE ranged from 1–35% (mean 15.4%). Mainly very faint spots have CVs in the higher range (Figure 2 ). In addition to the high salt concentration of CSF (> 150 mmol/L), ampholytes are also introduced into the sample in the prefractionation step, LP-IEF. We previously reported that the focusing time in the first dimension of prefractionated 2-DE had to be increased [ 16 ] probably due to insufficient "clean up" of the sample by acetone precipitation. Therefore, different "clean up" procedures were tested. Precipitation using ethanol was found to be most effective, keeping the number and intensity of the protein spots constant and allowing us to reduce the focusing time in the first dimension. We found that TCA-acetone precipitation reduced the protein content of the sample in agreement with a study of directly analyzed CSF samples [ 13 ]. In contrast to the results of Yuan et al. [ 13 ] a substantial loss of protein spots using the Bio-Spin column was found. The reason might be that proteins are retained in the spin column to a higher degree in the presence of ampholytes (Servalytes), which are small charged peptides. Ethanol precipitation of plasma samples has previously been performed[ 28 ], showing that a concentration of 66.6% ethanol was sufficient to precipitate 99% of the proteins [ 28 ]. This is in agreement with our findings, that 100% of CSF proteins are precipitated at ethanol concentrations above 70%. When comparing direct and prefractionated 2-DE it is evident that the prefractionated gels contain more spots, with higher protein quantities. Thus, the CSF proteins are enriched in the prefractionation step, simplifying their identification by MS, as shown in our previous study [ 16 ]. CSF analysis on the pH 3–6 and pH 5–8 gels in particular is improved by the prefractionation step, probably because the amount of CSF proteins in these pH ranges, without prefractionation, is rather low. In order to widen our search for protein differences in the CSF of FTD patients the improved prefractionated 2-DE procedure was applied to CSF from five FTD patients and five control subjects. 26 protein spots were changed at least two fold and 13 of these protein spots, representing seven different proteins, were identified as ZAG, ProapoA1, β-2-m, TTR, RBP, serum albumin and alloalbumin. Our previous direct 2-DE study [ 12 ] of the FTD proteome showed that 7 proteins were significantly altered compared to controls, including granin like neuroendocrine precursor, apolipoprotein E, pigment epithelium derived factor, RBP, haptoglobulin and albumin. A reduced level of RBP was consistent between our two studies, and in this case RBP was found reduced in both the pH 4–7 and the pH 5–8 gels. In contrast, CSF analysis of AD showed increased levels of one isoform of RBP [ 11 ], indicating a different role of RBP in the pathology of AD and FTD. RBP is synthesized by hepatic parenchymal cells, after binding to its ligand retinol, the complex is secreted into the circulation [ 29 ], where it further complexes with the plasma protein TTR. CSF RBP concentration has been shown to correlate to those of serum [ 30 ]. Serum RBP and retinol have been found to be reduced during acute infection and the decrease is proportional to the extent of the infection [ 31 ], suggesting that reduced RBP levels may result from an inflammation in the FTD brain. Brain TTR is exclusively produced, secreted and regulated by the choroid plexus [ 32 - 34 ]. TTR makes up 25% of the total CSF protein content [ 32 ] and even higher concentrations exist during prenatal and early postnatal life, indicating an importance of the protein in CNS development [ 35 ]. In this study, the levels of two isoforms of TTR were increased in the CSF from FTD patients. To our knowledge, the TTR levels of FTD CSF have not previously been studied, but in AD the CSF levels were decreased in an immunological study, not differentiating between TTR isoforms [ 36 ]. In contrast, the direct 2-DE study of the AD proteome [ 11 ] showed an increased level of TTR, but of a more acidic isoform, compared to this study. This highlights the capacity of 2-DE to quantify specific isoforms. One isoform of β-2-m was found increased in this study. Other studies have shown that CSF β-2-m is elevated in patients with various neurological diseases including AD [ 11 ], infectious meningoencephalitis [ 37 ], neurosarcoidosis [ 38 ] and AIDS dementia complex [ 39 ]. β-2-m constitutes the non-covalently bound light chain of major histocompatibility complex class I molecule (MHCI) [ 40 ]. The MHCI complex is expressed on the surface of all nucleated cells and the association of β-2-m to the MHCI transmembranal chain is an absolute requirement for the antigenic presentation function of the complex [ 40 ]. It has been proposed that conformational changes of the MHCI complex, associated with cell injury, can be responsible for increased shedding of β-2-m from the cell membranes with consequent expansion of the circulating β-2-m pool [ 41 ]. The function of ZAG is unknown but studies have shown that it is present in several body fluids, including CSF, sweat, saliva, seminal fluid, plasma, milk, amniotic fluid and urine, suggesting a fairly widespread exocrine function of the protein [ 42 ]. In this study increased levels of ZAG were found in FTD CSF and to our knowledge, ZAG has not previously been associated with dementia. The level of ProapoA1 was also increased in this study. Our previous study of the FTD proteome [ 12 ] did not show any increase in ProapoA1. Nevertheless, our study of the AD proteome [ 11 ] detected reduced levels of 3 isoforms of ProapoA1. The reason that several protein changes were inconsistent between this and our previous study of the FTD proteome may be explained by the fact that a smaller sample size and a different population of FTD patients, which is a rather heterogeneous disease, were used in this study. Due to the small sample size it must also be emphasized that the protein changes found in this study are preliminary. Moreover, direct and prefractionated 2-DE are still two different proteomic approaches and a somewhat different analytical window was not unexpected. Indeed, direct and prefractionated 2-D gels show different protein patterns, for example, the apolipoprotein E and apolipoprotein J isoforms seem to be missing in the prefractionated 2-D gels, which may be explained by the fact that lipoproteins tend to adhere to plastics [ 43 ] and could be lost during LP-IEF or additional sample transfer steps in the prefractionation procedure. However, the lipoprotein, ProapoA1 could still be detected in the prefractionated 2-D gels. The proteins most likely to be favored by a prefractionation step are low abundant hydrophilic proteins, which most likely are present in CSF. Nevertheless, this investigation of the FTD proteome failed to detect any very low abundant brain specific proteins. As shown in the present study, the levels of specific isoforms are altered and these are unlikely to be detected using methods measuring the total concentration of a protein. Therefore, determination of posttranslational modifications is of importance for understanding the neuropathology in FTD, and 2-DE is a useful method for sensitive detection of different protein isoforms. Conclusions We have shown that the prefractionated 2-DE method is reproducible to the same extent as traditional 2-DE and can enrich CSF proteins in the gel. This approach may offer new perspectives on the pathology of neurodegenerative diseases. Prefractionated 2-DE analysis of FTD CSF proteins confirmed some of the proteins previously detected by direct 2-DE and revealed some new biomarker candidates. The protein changes should be further validated on a larger patient material, preferably also with complementary methods, in order to assess any of the proteins potential as biomarkers for FTD. Methods CSF samples CSF samples were obtained from the Clinical Neurochemical Laboratory (Sahlgrenska University Hospital/ Mölndal). All CSF samples had a normal white-cell count, normal blood-brain barrier function and absence of intrathecal IgG and IgM production. The CSF samples for the proteomic study, described in table 2 , were obtained from 5 FTD patients aged 70.6 ± 5.6 (mean ± SD) years and 5 non-dementia controls aged 59.2 ± 11.9 years (mean ± SD). FTD was diagnosed according to the Lund Manchester criteria [ 4 ]. The severity of dementia was evaluated using the Mini Mental State Examination (MMSE) [ 44 ]. The control group, "non-demented controls" consisted of subjects with minor psychiatric complaints or subjective memory complaints that could not be verified by clinical examination, CSF analysis or neuropsychological testing. All control individuals had MMSE scores of 29–30. Lumbar puncture was performed in the L4–L5 vertebral interspace. The first 12 mL of CSF was collected and gently mixed to avoid possible gradient effects. The CSF samples were then centrifuged at 2,000 g for 10 min to eliminate cells and other insoluble material, and stored at -80°C. Table 2 CSF samples included in the prefractionated 2-DE study Subject Age a) Sex Albumin ratio b) Tau (ng/L) Aβ (ng/L) MMSE FTD 1 64 F 6.3 633 779 20 FTD 2 72 F 6.7 364 668 22 FTD 3 73 M 4.1 409 245 15 FTD 4 78 F 8.0 283 833 10 FTD 5 66 F 6.9 302 642 19 Control 1 63 M 6.7 368 599 29 Control 2 63 F 8.7 274 1070 30 Control 3 74 M 6.4 270 412 29 Control 4 54 F 5.5 210 1160 30 Control 5 42 M 2.8 318 1620 29 a) The age of the FTD group is 70.6 ± 5.6 years (mean ± SD) and the age of the control group is 59.2 ± 11.9 years (mean ± SD) b) [CSF albumin(mg/L)] / [serum albumin(g/L)] The study was approved by the Ethical Committee of Göteborg University. All participants or their relatives gave their informed consent to participation in the study, which was performed in accordance with the Declaration of Helsinki. Purification and precipitation methods To find a method for effective reduction of impurities and maximal protein recovery, 300 μL of prefractionated pooled CSF samples (fraction 6–9, pH 4.5–6.0) were purified using each of the following methods: 1. Ice cold acetone precipitation ; acetone: sample (4:1, v/v) precipitated at -20°C for 2 hours. 2. Ice cold acetone-TCA precipitation ; 2a) acetone: TCA: sample (4:10%:1, v/w/v) at -20°C for 45 min. 2b) acetone: TCA: sample (4:20%:1, v/w/v) at -20°C for 45 min. The protein pellet was washed 2 times with acetone after centrifugation. 3. Chloroform/methanol/water precipitation , chloroform: methanol: sample (4:8:3, v/v/v) at room temperature for 2 hours. 4. Ice cold ethanol precipitation ; final concentrations of 60%, 70% and 80% ethanol was added to the sample and precipitated for 2 hours at -20°C. 5. Purification using micro Bio-Spin column (Bio-Rad, Hercules, CA, USA) with a M W cut off of 6 kDa. The purification procedure w two-dimensional electrophoresis (2-DE) as performed according to the manufacturer's instructions. After precipitation all samples were centrifuged and the protein pellet analyzed on 2-D gels, described below. The protein recovery of acetone and ethanol treated samples was measured using the RC DC protein assay (Bio-Rad) according to the manufacturer's instructions. Prefractionation, sample preparation and 2-DE procedure The CSF samples from individual patients were prefractionated using LP-IEF in the Rotofor cell (Bio-Rad). Three mL CSF sample was mixed with 9 mL millipore water, 1% ampholytes (Servalyte pH range 3–10, Serva Electrophoresis, GmbH, Germany), 20 mM dithiothreitol (DTT) and 1 × Complete antiprotease solution (Roche Diagnostics, Mannheim, Germany). The focusing was performed at 4°C and at 12 W constant power for 2.5 hours. Then the 20 Rotofor fractions were harvested and fraction 2–5 corresponding to pH 1.5–4.5, fraction 6–9 corresponding to pH 4.5–6.0 and fraction 10–14 corresponding to pH 6.0–7.5 were pooled and concentrated to 300 μL in a vacuum centrifuge prior to 2-DE. In the FTD-study, the prefractionated pooled protein fractions were precipitated using 900 μL 95 % ice-cold ethanol (71.25% final conc. ethanol) for two hours at -20°C. The mixture was centrifuged at 10,000 × g for 10 min at 4°C. The protein pellets were air-dried and then resolved in a buffer containing 9 M urea, 35 mM tris, 42 mM DTT, 2% 3-((3-cholamidopropyl) dimethylammonio)-1-propanesulfonate (CHAPS), 0.66% sodium dodecyl sulfate (SDS), 2% IPG buffer and bromophenol blue. The first dimension was carried out using immobilized pH gradient (IPG) strips (Bio-Rad), 7 cm, pH 3–6 for Rotofor fractions 2–5, pH 4–7 for Rotofor fractions 6–9 or pH 5–8 for Rotofor fractions 10–14. The IPG-strips were actively rehydrated in the CSF-protein sample for 12 h at 50 V followed by protein focusing for 20,000 Vh using the Protean IEF Cell (Bio-Rad). The IPG strips were placed in 5 ml equilibration solution (50 mM Tris-HCl pH 8.8, 6 M urea, 30% glycerol, 2% SDS, bromophenol blue) containing 1% DTT, and 2.5% iodoacetamide in the second equilibrium step for 2 × 15 min. The second dimension was performed using the Nu-PAGE gel system (NOVEX, San Diego, CA, USA) with (2-(N-morpholino) ethane sulfonic acid (MES) buffer: 50 mM MES, 50 mM tris, 3.5 mM SDS, 1 mM EDTA), for 35 min at 200 V. In the direct 2-DE procedure, 300 μL CSF proteins were precipitated using 900 μL ice-cold acetone and stored for two hours at -20°C. The mixture was then centrifuged at 10,000 × g for 10 min at 4°C. The 2-DE procedure was performed as described above for prefractionated 2-DE. Visualisation and evaluation The gels were stained using SYPRO Ruby Protein Stain (Molecular-Probes, Eugene, Oregon, USA) according to the manufacturer's instructions. Image acquisition and analysis were performed on a Fluor-S MultiImager (Bio-Rad). The protein spots were detected, quantified and matched with the PD-Quest 2-D gel analysis software, v.7.0 (Bio-Rad). The gels were normalized according to the total quantity in valid spots (the raw quantity of each spot in a member gel is divided by the total quantity of all the spots in that gel that have been included in the Master gel). Protein levels increased or decreased two fold were taken into account. In-gel tryptic digestion and sample preparation The protein digestion method has been previously described in detail [ 11 ]. Briefly, the gel pieces were digested with porcine trypsin (Promega Corporation, Madison, USA) and the peptides were extracted with formic acid (FA) and acetonitrile (ACN). The digested protein sample was dried under vacuum and then dissolved in 10 μL 0.2% triflouroacetic acid (TFA) (v/v). The samples were applied to the MS probe using the AnchorChip™ technology (Bruker daltonics, Bremen, Germany) as previously described [ 45 ]. Briefly α-cyano-4-hydroxy-cinnamic acid (CHCA) solution (100 g/L in 90% acetone, 0.005 % TFA) was spread out evenly on the sample plate surface creating the CHCA matrix layer. Then 2μL of the protein sample solution was applied to each anchor spot. After 2 min, the remaining liquid was removed by absorption using a paper tissue. Mass spectrometry and database searching Matrix assisted laser desorption/ionization time-of-flight (MALDI-TOF) MS analysis was performed using an upgraded Reflex II MALDI-TOF MS (Bruker-Franzen Analytik GmbH, Bremen, Germany) equipped with a two-stage electrostatic reflectron, a delayed extraction (time-lag-focusing) ion source, a high resolution reflector detector and a 2 GHz digitizer. The spectra were acquired in the reflection mode at an accelerating voltage of 20 kV. The mass spectra, acquired and analyzed using Bruker software, were initially calibrated by external calibration using a mixture of known peptides and later recalibrated using two auto digestion products of porcine trypsin as internal calibrants. The protein database search tool "MASCOT Peptide Mass Fingerprint" on the Matrix Science web site [ 46 ] was used to compare the monoisotopic m/z values of the tryptic fragments to those of known proteins in the NCBI database. A mass deviation of 100 ppm was tolerated and Homo sapiens was specified. Statistical analysis Coefficient of variation (CV) was calculated (standard deviation (SD)/ Mean × 100) of the normalized protein spot densities from four replicate 2-D gels. In the proteomic study a 2-fold increase or decrease of normalized protein quantities was taken into account. Competing interests The authors declare that they have no competing interests. Authors' contributions SFH carried out all the 2D gel experiments, mass spectrometry analysis, participated in results evaluation and drafted the manuscript. MP participated in the 2D gel experiments, results evaluation and participated to the manuscript writing. KB contributed with material and critical reading of the manuscript. MJ contributed with material and participated in the design of the study. PD conceived the study, participated in its design and coordination, results analysis and supervised the manuscript writing.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC538271.xml
533867	Improvement of alignment accuracy utilizing sequentially conserved motifs	Background Multiple sequence alignment algorithms are very important tools in molecular biology today. Accurate alignment of proteins is central to several areas such as homology modelling, docking studies, understanding evolutionary trends and study of structure-function relationships. In recent times, improvement of existing progressing programs and implementation of new iterative algorithms have made a significant change in this field. Results We report an alignment algorithm that combines progressive dynamic algorithm, local substructure alignment and iterative refinement to achieve an improved, user-interactive tool. Large-scale benchmarking studies show that this FMALIGN server produces alignments that, aside from preservation of functional and structural conservation, have accuracy comparable to other popular multiple alignment programs. Conclusions The FMALIGN server allows the user to fix conserved regions in equivalent position in the alignment thereby reducing the chance of global misalignment to a great extent. FMALIGN is available at	Background The advent of large genome projects has led to an explosion of sequence data in public databases. Analysis of protein families, understanding their evolutionary trends and detection of remote homologues are now the primary objectives. Genome annotation and analysis tools like fold prediction, homology modelling, protein-ligand docking and clustering algorithms rely heavily on accurate multiple alignments to provide a genome-wide perspective. The most popular approach for multiple sequence alignment has been the progressive alignment method [ 1 ]. A multiple alignment is built up gradually by aligning the closest sequences first and successively adding in more distant relatives. A number of alignment programs imply this algorithm, for example MULTALIGN [ 2 ], MULTAL [ 3 ] and CLUSTALX [ 4 ]. They employ a global alignment algorithm to construct an alignment over the entire length of the sequences and differ mainly in the procedure employed to determine the order of alignment of the sequences. The most common usage is the sequential branching method to identify two closest sequences first and subsequently align the next closest sequence to those already aligned. MULTALIGN [ 2 ] constructs a guided tree using UPGMA [ 5 ] method. A consensus method is then used to align larger and larger groups of sequences according to the branching order of the tree. CLUSTALX uses the alternative neighbour-joining algorithm [ 6 ] to construct the tree. In contrast to the above global method, PIMA [ 7 ] uses a local dynamic programming algorithm to align only the most conserved motifs. In addition, numerous new alignment algorithms have recently been developed which offer fresh approaches to the multiple alignment problem. A common point of interest has been the application of iterative strategies to refine and improve the initial alignment. A local alignment approach is implemented in the DIALIGN program [ 8 , 9 ] to construct multiple alignments based on segment-to-segment comparison rather than residue-to-residue comparison using an iterative strategy to improve alignment accuracy. Alignment programs like MATCH-BOX [ 10 ] utilize statistical similarity measures to delineate sequentially conserved regions and the final alignments are derived by those of the conserved "box" regions. The regions outside the "boxes" are not aligned. METAMEME [ 11 ] is a motif based search engine that aligns motif regions found in the target sequences. DbClustal [ 12 ] combines the advantages of both local and global alignment algorithm in a traditional tree-based progressive alignment. Starting from ClustalW [ 22 ], which is a global alignment program, local alignment data or anchor points are incorporated in the dataset. The global alignment is then weighted towards, but not constrained, to the locally conserved segments and the alignment is not subject to iterative refinements. MACAW [ 13 ] provides a user interactive interface to select conserved segments from the alignment but these segments are not utilized further to refine the resulting alignment. In this paper, we present an alignment algorithm that combines the properties of both progressive alignment methods and iterative refinement algorithms. This algorithm offers the user the selective advantage of guiding the course of alignment by simultaneous inputs of multiple conserved motif regions that in turn guarantees retention of structure/function in the final alignment. FMALIGN is an alignment server that provides the user to obtain a control over the alignment by providing important conserved regions as input to the alignment program to achieve a more structurally relevant and functionally useful alignment of protein sequences. It employs the sequential branching method to identify the closest pair of sequences and subsequently includes the next closest sequences to generate a guided tree using UPGMA [ 5 ], which in turn dictates the sequential order of the alignment. FMALIGN also considers the local similarity of the sequences in the conserved motif regions; as the name implies, it allows local conserved regions of the sequences to be fixed and aligns the rest based on normal progressive alignment. The chances of global misalignment are thereby reduced and the possibility of obtaining overall better alignment is increased. The FMALIGN server also offers an iterative refinement option where a routine (FINDMOTIF) identifies more conserved regions in the derived alignment and allows the user to provide fresh 'equivalences' to obtain an overall better alignment. Benchmarking studies on difficult alignments, examined in BAliBASE [ 14 ], show promising prospect for the FMALIGN server to be an useful alignment algorithm. Implementation: methodology and description of FMALIGN server The algorithm of FMALIGN (Fixed Motif ALIGNment) program combines three main criteria: (a) Progressive global alignment method, (b) substructure (local) conserved region fixation and (c) iterative refinement of alignment with identification of more conserved regions. The procedure involves three steps: (i) identification and fixing of the specified sequential conserved regions. This alignment method requires specific regions of the sequences to be aligned as anchor and these anchors are generally meant for sequentially conserved parts which do not undergo many changes. (ii) derivation of the progressive multiple sequence alignment guided by the tree. During this step, excluding the fixed anchored regions, the rest of the sequences are divided into several sub-segments that are aligned employing a dynamic alignment algorithm in a sequential order from N to C-terminus. The phylogenetic guide tree has been derived using Unweighted Pair Group Method with Arithmetic Mean (UPGMA) [ 5 ] which is the simplest method of tree construction. The UPGMA method was adopted since the whole length of the protein is divided into several segments considering them as ultrametric After all the sub-segment alignments are performed, the aligned parts as well as the selected, fixed motif regions are combined to produce the full-length alignment for a group of proteins under the hierarchical cluster. This process continues until all the protein sequences are aligned multiply. (iii) The iterative refinement of alignment subsequent to the identification of more motifs. In this step, more conserved regions are identified by observing the amino acid exchanges in the resultant alignment derived from the previous iteration. These conserved regions are then used as anchors together with previously identified motifs and the whole process is repeated until an optimal alignment having maximum number of conserved regions and alignment score is obtained. The algorithm thus combines the progressive dynamic algorithm for global multiple alignment and selected conserved regions or local alignments. Once the primary alignment is derived, the second step can be repeated by including more conserved regions from this alignment as motifs to derive a better alignment. Figure 1 shows a cartoon representation of the methods in a flowchart diagram. FMALIGN server can accept amino acid patterns for multiple motifs provided by the user. It also provides option to the user to search motifs within their sequences using FINDMOTIF routine or to obtain them for the alignment by providing a link to SMoS [ 15 ], a structural motif database for aligned protein superfamilies. The FINDMOTIF routine in the server provides sequential conserved regions for a set of proteins on the basis of sequence similarity and a 20 × 20 substitution matrix by consulting large number of structure-based sequence alignments of homologous families [ 16 ]. Amino acid exchange scores at every alignment position are assigned the same as the element in this matrix for all possible pairs of proteins and averaged over the number of pairs. Contiguous alignment positions with an average amino acid exchange score over 50 (for homologues) or 40 (for superfamilies) are recognised as motifs. FMALIGN also offers an option for the user to refine the derived alignment by generating more sequential conserved regions through FINDMOTIF option. The inter-motif regions are aligned by normal progressive alignment using standard substitution matrices like BLOSUM62. The gap penalties used in this version of FMALIGN are all maintained according to standard multiple alignment parameters. Alignment scores To assess the performance of FMALIGN in comparison to other programs, Sum-of-Pair-score (SPS) and Column-Score (CS) alignment scoring scheme [ 14 ] are applied on FMALIGN derived alignments to assess the quality of alignments compared to BAliBASE reference alignments. SPS and CS are calculated such that the score increases with the number of sequences aligned accurately and is used to determine the extent to which the programs succeed in aligning some, if not all, of the sequences in an alignment correctly [ 14 ]. The scores used to measure the performance of the various alignment programs may not be appropriate for all the datasets. Therefore, for each reference test the most suitable scoring function have been selected according to the nature of the benchmarking. Results and discussion Benchmarking at family level FMALIGN derived alignments retain high degree of conservation in secondary structures. Six members from the globin family were selected comprising a wide range of sequence identity between them (7% to 61%). Conserved regions for six aligned globin sequences were identified by the FINDMOTIF routine starting from CLUSTALX [ 4 ] alignment with default parameters and aligned by the FMALIGN server. The resultant alignment shows high degree of secondary structural conservation despite the difficulty of aligning a set of sequences having very wide range of sequence identity (Figure 2 ). In order to evaluate the FMALIGN server, the results have been compared with 10 other alignment programs and objective criteria were employed to assess the quality of an alignment. We selected the BAliBASE benchmark alignment database [ 14 ] to compare the performance of the FMALIGN alignment server. The BAliBASE benchmark alignment database contains 142 reference alignments, divided into five reference sets each containing at least 12 representative alignments. Performance of FMALIGN is checked on all five reference sets provided in BAliBASE datasets. Reference 1 alignments consist of a small number of equidistant sequences of similar length, i.e. the percent residue identity (% ID) between any two sequences is within a specified range and no large extensions or insertions have been introduced. Reference 2 contains alignments of a family of closely related sequences with >25% ID, plus up to three 'orphan' sequences (distant members of the family with <20% ID, sharing a common fold). It is designed to evaluate program accuracy according to two criteria: (i) the stability of the family alignment when orphans are introduced into the sequence set and (ii)the quality of the alignment of the orphan sequences. Reference 3 checks the ability of the programs to correctly align equidistant divergent families into a single alignment. The reference alignments consist of up to four families, with <25% ID between any two sequences from different families. Reference 4 and 5 contain alignments of upto 20 sequences including N/C-terminal extensions (upto 400 residues) and insertions (upto 100 residues), respectively. Reference 1: a small number of approximately equidistant sequences This dataset is designed to study the effect of sequence length and percentage identity on the performance of the alignment program and provides a basis for the remaining tests. The overall performance of FMALIGN server for this dataset is comparable to the two best performing alignment programs, like PRRP [ 17 ] and CLUSTALX [ 4 ], in all three categories (VI, V2, and V3) based on sequence identity and alignment lengths (short, medium and long) as shown in Figure 3 , 4 , 5 . Reference 2: a related family with divergent, orphan sequences It is possible to assess the performance of the methods to align divergent 'orphan' sequences (10–20% ID with the family and between orphans) with a family of highly related (>25% ID) sequences using this data set. It is also interesting to observe the disruption of the family alignment due to the introduction of orphans. Figure 6 shows SPS for the alignment of a single orphan against a closely related family. The global alignment programs again perform better than the local ones in this test. However, CLUSTALX and SAGA [ 18 ] now rank above PRRP. The performance of FMALIGN server is significantly better than other programs for all the three length categories. Reference 3: families of related sequences This allows the assessment of the programs to correctly align approximately equidistant divergent families (<20% ID) composed of highly related sequences (>25% ID) into a single multiple alignment. Figure 7 shows the scores for the programs in the order. The iterative strategies of PRRP [ 17 ] and SAGA [ 18 ] perform better in this test than the traditional progressive alignment methods. However, FMALIGN performs better than the other progressive methods, with the global methods generally ranking higher than the local methods. Reference 4: N/C-terminal extensions This dataset includes large N/C-terminal extensions to investigate whether the programs are capable of aligning the core blocks flanking the extensions. No large internal insertions are introduced at this stage. Mostly local alignment strategies out-perform the global methods. PILEUP (Wisconsin Package v.8; Genetics Computer Group, Madison, WI) is the only program based on a global alignment method which does reasonably well compared to other global methods. Performance of FMALIGN is comparable to the best three methods (DIALIGN [ 8 ], SB_PIMA [ 19 ] and PILEUP) as shown in Figure 8 . Reference 5: internal insertions In contrast to reference 4, in this dataset the insertions are internal to the homologous domains and not at the N/C-terminus as overhangs. FMALIGN also performs well in this category and results are comparable to the other better performing global alignment methods like PRRP and CLUSTALX as shown in Figure 8 . Benchmarking at superfamily level Utilization of structural motifs to improve alignment accuracy The performance of FMALIGN has been tested on a dataset of representative superfamilies of proteins belonging to different structural classes in PASS2 [ 20 ] and SMoS [ 15 ] databases. The structural motifs from the SMoS database have been utilized as conserved regions for the member of the superfamily (average sequence identity less than 30%). All the proteins of a superfamily have been aligned. Each alignment is assigned a quality score by averaging the amino acid exchange score of each column over the length of the alignment (see method for details). The alignments derived from FMALIGN have been compared against the CLUSTALX-derived sequence alignment as well as the sequence-structure alignment derived from COMPARER [ 21 ]. Figure 9 shows an equivalent or better accuracy of the FMALIGN server compared to CLUSTALX and COMPARER-derived alignment. This indicates that FMALIGN is particularly efficient for specific sets of sequences for which the degree of conservation is known. The initial results also indicate that FMALIGN can perform very well and provide an alignment which is very similar or better than a structurally derived alignment. Utilization of sequence motifs to improve alignment accuracy Representative superfamily alignments belonging to different structural classes from the PASS2 [ 20 ] superfamily alignment database are taken for a benchmarking test (Table 1 ). These superfamily alignments are of different lengths (short, medium, and long) and possess an average sequence identity that ranges from 10% to 26%. Sequential conserved regions or motifs are identified for each superfamily alignment and utilized in the FMALIGN server to realign the superfamily members. Similarly, the superfamily members are also aligned by the two best performing multiple alignment methods, CLUSTALW [ 22 ] and T-Coffee [ 23 ]. These alignments are then compared against the structure-based COMPARER alignments provided by PASS2 database using the same alignment scoring scheme (Sum-of-Pair score) of the BAliBASE benchmarking database. FMALIGN-derived alignments performed better for most of the cases compared to the other two methods as shown in Figure 10 . Conclusions FMALIGN server provides a web interface for pairwise and multiple sequence alignments of proteins. FMALIGN provides an alignment by combining progressive dynamic algorithm, local substructure alignment and iterative refinement presenting an improved, user interactive alignment procedure. FMALIGN server allows the user to fix conserved regions in equivalent positions amongst the sequences to be aligned leading to alignments that are reliable and biologically more meaningful. Additional options for the users to choose substitution matrices and gap penalty values may be incorporated in future. All the alignments provided in BAliBASE are realigned by FMALIGN and the resulting scores (both SPS and CS) are calculated using a standard program kindly provided by the authors of BAliBASE. The inter-motif regions are aligned in FMALIGN by normal progressive alignment using standard substitution matrices like BLOSUM62. The gap penalties used in this version of FMALIGN are all maintained according to standard multiple alignment parameters. Benchmarking at the superfamily level has also been done utilizing both structural and sequential conserved regions. The dataset is wide enough to include proteins from different structural classes and of different length. The average sequence identity for this dataset was less than 30% since the proteins are related at the superfamily level. The performance of FMALIGN has been tested against one of the best performing multiple sequence alignment (T-Coffee) as well as structure based alignment tool (COMPARER). Studies on large and different datasets revealed an overall better performance of FMALIGN server for all categories in BAliBASE benchmarking database. It works especially well at lower sequence identity range, such as superfamily level, where no two proteins are more than 25% identical to each other in comparison to other popular methods like CLUSTALX and T-Coffee. It is well-known that automatic multiple sequence alignments at poor sequence identity are often subject to careful manual validations and improvements to avoid offsets of critical functionally important residues. FMALIGN can sensitively address this issue to avoid manual intervention subsequent to final alignment. FMALIGN-derived alignments also show a high conservation of secondary structural elements and provide better alignments for comparative modelling. The implementation of this alignment algorithm can be used to include new members into an existing protein superfamily with the help of motif regions that provide a reliable approach to connect protein sequences with their structural homologues within a particular superfamily. FMALIGN is available via the following URL Authors' contributions S.C. had carried out the benchmarking, was involved in the development of the server and has written the first draft of the manuscript. N.B. had written the initial part of the code and P.A. has written the latter part of the code and developed a web server. R.S. had initiated the idea, was involved in discussions and in the critical reading of the manuscript. List of abbreviations FMALIGN F ixed M otif ALIGN ment SPS Sum-of-Pairs Score	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC533867.xml
514572	Attenuation of dengue virus infection by adeno-associated virus-mediated siRNA delivery	Background The need for safe and effective treatment of dengue virus (DEN), a class A agent that causes dengue hemorrhagic fever/dengue shock syndrome, has been a critical global priority. An effective vaccine for DEN is not yet available. In this study the possibility of attenuating DEN infection using adeno-associated virus (AAV)-encoded short interfering RNAs (siRNA) was examined in Vero cells and human dendritic cells (DCs). Methods A cassette encoding siRNA targeted to a 3' untranslated sequence common to all DEN serotypes was designed and tested for its ability to attenuate DEN infection by use of AAV delivery. Results Vero cells or DCs infected with AAV-siRNA showed a significant, dose-dependent reduction in DEN infection. Treatment of DCs with AAV-siRNA also decreased the DEN-induced apoptosis of DCs and did not induce significant inflammation. Conclusion These results demonstrate that AAV-mediated siRNA delivery is capable of reducing DEN infection in cells and may be useful in decreasing DEN replication in humans.	Introduction The need for a safe and effective prophylaxis or treatment for dengue virus (DEN) infection, a category A mosquito-borne human pathogen, is a critical global priority. DEN causes dengue hemorrhagic fever/dengue shock syndrome (DHF/DSS), which is associated with heterologous secondary DEN infection and affects thousands of people worldwide. The incidence of DHF/DSS is increasing in the western hemisphere and, although many different approaches are being tried to develop prophylactic DEN vaccines, none have been licensed for public health and there are no specific antiviral treatments available. DEN belongs to the family of Flaviviruses and is an enveloped single plus-stranded RNA virus with four distinct serotypes. The DEN genome of approximately 11,000 nucleotides encodes a polyprotein (C-prM-E-NS1-NS2a-NS3-NS4a-NS4b-NS5) consisting of three structural proteins (C, prM and E) and seven nonstructural proteins. The open reading frame is flanked by a 100 nucleotide-long noncoding region (NCR) at the 5' end and a 400 to 600 nucleotide-long NCR at the 3'end [ 1 ]. Although the mechanism of DEN pathogenesis is unclear, DEN typically appears to replicate locally in skin or blood dendritic cells (DCs) and may also involve monocytes and macrophages. Secondary infection is usually more serious because of antibody-dependent enhancement (ADE). We reasoned that an effective antiviral approach aimed at attenuating the DEN virus burden might protect infected subjects from DHF/DSS and, therefore, we examined the potential of an in vivo gene-silencing approach using short interfering RNA (siRNA) to decrease DEN replication. RNA interference is triggered by dsRNA that is cleaved by an RNAse-III-like enzyme, Dicer , into 21–25 nucleotide fragments with characteristic 5' and 3' termini [ 2 ]. These siRNAs act as guides for a multi-protein complex, including a PAZ/PIWI domain containing the protein argonaute2 , that cleaves the target mRNA [ 3 ]. These gene-silencing mechanisms are highly specific and can potentially inhibit the gene expression of different viruses [ 4 , 5 ]. This approach was found to be effective in blocking DEN replication in insect cells [ 6 , 7 ]. Plasmid DNAs or adenoviruses encoding appropriate DNA sequences allow transient siRNA expression in cells and in vivo leading to specific gene silencing. However, plasmids transfect mammalian cells poorly, and adenoviruses produce an acute inflammatory response and an immune response to viral vector-encoded antigens [ 7 , 8 ]. We therefore developed an adeno-associated virus (AAV) system capable of expressing siRNA cassettes, and tested this vector with a siRNA cassette composed of a nucleotide sequence from the 3' NCR of the DEN genome (siDEN3UT), which is common to all four serotypes. The results obtained in Vero cells and human DCs infected with AAV-siDEN3UT show significant decreases in DEN infection and DEN-induced apoptosis. Methods Plasmid constructs The pCMV-MCS plasmid (Stratagene) was digested with Not I and the larger fragment was ligated to the synthetic adapter containing in order, Not I- Kpn I- Apa I- Xho I- Hind III- EcoR I- Bam HI- Sac II- Sac I- Cla I- Sal I- Bgl II- Not I. The U6 promoter was obtained by PCR amplification, using specific primers with the desired restriction sites from the template pSilencer 1.0-U6 (Ambion), and inserted into the adaptor at the Kpn I and Apa I sites to get a novel plasmid pCMV-U6. Pairs of oligos were synthesized to develop siRNA constructs. The nucleotide sequence for each siRNA is as follows: siEGFP: 5'-GGC GAT GCC ACC TAC GGC AAG CTT CTC GAT TCG AAG CTT GCC GTA GGT GGC ATC GCC CTT TTT G-3' [ 10 ]; siRSVNS1:5'-GGC AGC AAT TCA TTG AGT ATG CTT CTC GAA ATA AGC ATA CTC AAT GAA TTG CTG CCT TTT TG-3'; siDENPrM: 5'-GGA AGA CAT AGA TTG TT G GTG CAC TCG AGT CAA CGT GCA CCA ACA ATC TAT GTC T TC CCT TTT TG-3'; siDEN3UT:5'-GGA AAA ACA GCA TAT TGA CGC TGC TCG AGT CAA CGC AGC GTC AAT ATG CTG TTT TTC CCT TTT TG-3'. Each pair of oligos was annealed and then inserted into pCMV-U6 digested with Apa I/ Xho I and Xho I/ EcoR I respectively. The modified pCMV-U6 plasmid was then redigested with Not I and the smaller fragment was ligated to the 2.9 kb fragment of pAAV-MCS (Stratagene) obtained following its Not I digestion to generate the corresponding si-vector for EGFP, RSV and DEN. HEK-293 cells were cotransfected with the helper plasmid and the si- plasmid to generate recombinant AAV. Cell culture and viral packaging HEK293 cells were cultured with DMEM (Cellgro) plus 10% FBS (Cellgro) and cotransfected with pSMWZ-siDEN, pHelper and pAAV-RC (Stratagene) by standard calcium phosphate transfection. Cells were harvested 48 hr post-transfection and the cell pellets were stored at -80°C. Purification of recombinant adenoassociated viruses Cells were lysed by 5 cycles of freezing and thawing to release the virus. Crude viral lysate were collected by centrifugation at 27,000 × g for 30 min, and the supernatants were harvested and put onto a CsCl gradient (density 1.20/1.50) in fresh tubes and centrifuged for 16 h at 100,000 × g. Opalescent bands were collected after ultracentrifugation. Titers of purified AAVs were measured using an AAV titration ELISA kit (Progen Biotechnik, Germany). Isolation and culture of dendritic cells from human peripheral blood Conditions were similar to those described previously [ 11 ]. Buffy coats were diluted with one volume of DMEM (Cellgro) and PBMCs were isolated by density-gradient centrifugation using Histopaque-1077 (SIGMA) according to the instructions. The PBMC layer was harvested, washed twice with DMEM, resuspended in DMEM supplemented with 10% FBS, and then seeded into six-well culture plates. After 2 h at 37°C/5%CO 2 the nonadherent cells were removed and the adherent cells were cultured with fresh DMEM supplemented with 10% FBS (Cellgro), 200 ng/ml IL-4 (BD-Pharmingen) and 50 ng/ml GM-CSF (BD-Pharmingen) for 7 days prior to infection with DEN. Blocking dengue virus infection in vitro 1 × 10 5 Vero cells or DCs were seeded into six-well tissue culture plates and infected with different numbers of recombinant AAV carrying the DEN-siRNA silencing cassette. After 2 days the cells were infected with DEN-2 virus (strain16803) at an MOI of 0.1. Flow cytometry Cells were harvested and centrifuged for 10 min at 150 × g. The cell pellets were washed with PBS and resuspended with antibody to DEN-2 virus envelope protein, (Microbix Biosystems Inc, Clone No 3H5) on ice for 40 min. The cells were centrifuged and pellets were washed with PBS and then resuspended with secondary antibody conjugated with FITC (Sigma) for an additional 30 min. The number of infected cells was measured by flow cytometry 5 days post-infection. DCs were also stained with CD11c antibody conjugated with PE (BD-Pharmingen). Plaque assay The supernatants from DEN-2-infected DCs were collected at day 5 post-infection and 10-fold serial dilutions were allowed to adsorb to monolayers of Vero cells in six-well culture plates for 2 h. The medium was then removed and replaced by an agarose-containing overlay (2X DMEM, 10%FBS, non-essential amino acids (Gibco BRL), 1% low melting-point agarose (Gibco BRL) and the plates were incubated at 37°C/5% CO 2 for 5 days. Afterwards, 1% neutral red (Sigma) was added to each well and the plaques were counted 48 h later. Apoptosis assay Infected-DCs were harvested on day 5 of infection with DEN-2 virus. Aliquots of DCs were put onto slides using a Cytospin and fixed with 4% paraformaldehyde. Apoptotic DCs were determined using the terminal dUTP nick end-labeling assay (TUNEL, Promega, Madison WI) and the annexin V apoptosis detection kit (BD Biosciences, CA). Cytometric bead array and ELISA analysis Supernatants from infected-DCs were harvested at 24 h, 48 h, 72 h and 96 h post-infection. Cytokine concentrations were measured by cytometric bead array (CBA) and ELISA (BD-Pharmingen) following instructions in the manuals. Statistical analysis Data were expressed as arithmetic mean ± SEM. Levels of significance of the differences between groups were determined by the student t test. Values of p < 0.05 were considered statistically significant. Results Development of an AAV-siRNA system for gene silencing In order to develop an AAV-siRNA system, a plasmid pSMWZ-1 was engineered that comprised a mouse U6 promoter linked to a siRNA cassette (Fig. 1A ). To test whether this plasmid was functional and capable of suppressing gene expression, HEK293 cells were cotransfected with pEGFP, a plasmid expressing green fluorescent protein, and pSMWZ-siEGFP. The percentage of cells expressing EGFP was determined and the results showed that there was a dose-dependent silencing of EGFP expression (Fig. 1B ). In contrast, control cells cotransfected with siRSVNS1 (targets the NS1 gene of human respiratory syncytial virus) in place of siEGFP did not show any reduction in EGFP expression. To test various siDEN candidates, Vero cells were transfected with either pSMWZ-siDENpreM (siDENpreM) or pSMWZ-siDEN3UT(siDEN3UT), then two days post-transfection infected with DEN-2 (strain 16803) at an MOI of 0.1. At five days post-infection, the numbers of DEN-2 virus-infected cells were quantified by fluorescence microscopy using antibody to DEN-2 envelope protein and FITC-conjugated secondary antibody. The results showed that siDEN3UT was better than siDENpreM in suppression of DEN-2 infection (Fig. 1C ). The AAV-siRNA system was similarly tested using HEK293 cells which were infected with AAV-siEGFP and then transfected with pEGFP. The decrease in the percentage of cells expressing EGFP showed that there was a silencing of EGFP expression in a dose-dependent and sequence-specific manner (Fig. 1D ). Figure 1 Construction and characterization of the siDEN suppressor. (A) Diagram of the construction of the plasmid vector, pSMWZ-1, capable of expressing a DEN infection suppressor cassette. Abbreviations: N, Not I; K, Kpn I; A, Apa I; E, EcoR I; SUP-1, Suppressor cassette; (B) Co-transfection with pSMWZ-siEGFP and pEGFP inhibits the expression of EGFP in cultured cells. HEK293 cells were transfected with different concentrations of plasmid DNA and three days later, EGFP-positive cells were counted by fluorescence microscopy. Results are expressed as mean ± SEM. p < 0.05 compared to control (siRSVNS1, an unrelated siRNA construct against respiratory syncytial virus). (C) pSMWZ-siDEN suppression of DEN-2 virus replication. Vero cells were transfected with pSMWZ-siDEN3UT or pSMWZ-siDENpreM plasmid. Two days later, the cells were infected with DEN-2 virus (MOI of 0.1) and 5 days later, the numbers of DEN-2 virus infected cells were counted by fluorescence microscopy. Data are mean ± SEM. * p < 0.05 compared to control DEN-2. (D) AAV-siEGFP inhibits the expression of EGFP in cultured cells. HEK293 cells were infected with different concentrations of AAV-siEGFP, and three days later the cells were transfected with pEGFP. EGFP-positive cells were counted by fluorescence microscopy. Statistically significant differences, ** p < 0.01, when compared to pEGFP plasmid control, AAV-siRSV (10 7 ) and AAV-siEGFP (10 8 ) group, respectively. To test whether AAV-siDEN-2 expression decreases DEN-2 virus infection in cultured Vero cells, cells were infected with recombinant AAV carrying siDEN-2 (MOI 10) or siEGFP (MOI ~1000) silencing cassettes. After 2 days the cells were infected with DEN-2 virus at an MOI of 0.1. Five days later, the numbers of DEN-2 virus infected cells were quantified by flow cytometry using anti-DEN-envelope protein. Cells pre-infected with AAV-siDEN3UT, but not AAV-siEGFP, showed a significant reduction in DEN infection, and the reduction was dose dependent (Fig. 2 ). Figure 2 AAVsiDEN expression decreases DEN-2 virus infection in cultured Vero cells. Cells were infected with different amounts (PFU/ml) of AAV carrying the siDEN3UT silencing cassette and after 2 days the cells were infected with DEN-2 virus (MOI of 0.1). Five days later, the numbers of DEN-2 virus infected cells were measured by flow cytometry. siRNA suppresses DEN infection in human dendritic cells DEN is transmitted through Aedes aegypti mosquito bites, and resident skin DCs are regarded as the targets of DEN infection [ 12 ]. DCs are thought to be 10-fold more permissive for DEN infection than monocytes or macrophages [ 13 ]. We therefore tested the ability of AAV-siDEN3UT to attenuate DEN infection in human DCs. DCs were isolated from human blood and cultured in the presence of IL-4 and GM-CSF for 5 days to generate immature DCs (iDCs). These DCs were then infected with 10 9 PFU/ml of recombinant AAV carrying siDEN3UT or siEGFP (control) silencing cassette. After 2 days the cells were infected with DEN-2 at an MOI of 0.1. Five days later, the numbers of DEN-2 infected cells were quantified by flow cytometry using DEN-2 antibody. Cells preinfected with AAVsiDEN3UT showed a 50% reduction in the number of infected cells (Fig. 3A ). To test whether the reduction in the number of infected DCs involved a reduction in DEN titer, the culture supernatants were examined using a Vero cell-based plaque assay. AAV-mediated siDEN3UT expression significantly decreased DEN-2 virus titer compared to control (Fig. 3B ). These results indicate that AAV-siDEN3UT can significantly decrease DEN titers in human DCs. Figure 3 Suppression of DEN-2 replication in DCs by siDEN. (A) DCs were isolated from human peripheral blood and cultured in DMEM medium supplemented with FBS, IL-4 and GM-CSF. Non-adherent DCs were harvested on day 7, infected with AAVsiDEN, and two days later the cells were infected with DEN-2 at 0.1 MOI. DCs were harvested 5 days after DEN-2 infection and DEN-2 titers were measured by flow cytometry. (B) Supernatants from DEN-infected DCs were collected and added to culture plates containing confluent Vero cell monolayers. After virus adsorption, the Vero cells were overlaid with agarose and stained with 1% neutral red. Viral plaques were counted 48 h after neutral red overlay. Data are the averages of two independent experiments. * p < 0.05 compared to control. Silencing DEN-2 genes inhibits apoptosis in dendritic cells It has been reported that DEN infection induces apoptosis of DCs [ 11 ] leading to an immunosuppressed condition. To examine the effect of AAV-mediated siRNA delivery in DCs, apoptosis was investigated in infected DCs using the TUNEL assay. The results showed that a small percentage of DCs undergo apoptosis naturally during culture, but DEN infection causes much more apoptosis. The AAV-siRNA-treated DEN-infected DCs showed significantly fewer apoptotic cells compared to DEN-infected cells without AAV-siRNA (Fig. 4 ). Figure 4 AAVsiDEN reduces apoptosis in human DCs infected with DEN-2. DCs were isolated from human peripheral blood and infected with AAVsiDEN followed by DEN-2. Five days after infection, DCs were put onto slides and apoptosis was determined using the terminal dUTP nick end-labeling assay (TUNEL). Nuclei were stained with diamidinophenylindole (DAPI). Representative fields were visualized by fluorescence microscopy. Differential expression of cytokines by infected dendritic cells The supernatants of infected DCs were collected at different time points and cytokines were measured using cytokine bead array (CBA) and ELISA assays. As showed in Figure 5 , cultured DCs spontaneously produced increased IL-1b. A variety of cytokines including IFN-γ, TNF-α, IL-8, IL-6, IL-12 were measured (data not shown). In the presence of AAV-siRNA infection, the production of IFN-γ, TNF-α, IL-8, IL-6, and IL-12 did not change significantly compared with cultured DCs. IL-1b secretion at 72 h post-infection was increased, however. These results indicate that in our system, AAV-siRNA delivery does not induce acute inflammation in DCs, in vitro, Figure 5 Differential expression of cytokines in the supernatants of infected DCs. Supernatants from DEN-2-infected DCs with or without siDEN treatment were harvested at the indicated time points and analyzed by CBA and ELISA to measure the concentrations of cytokines. Data are the averages of two independent experiments. ** p < 0.01 in comparison with the value of DCs within individual group. Discussion The significant findings of this report include the development of an adeno-associated virus-based siRNA approach for downregulating gene expression. A recombinant AAV-siDEN3UT was utilized to induce significant decreases in DEN infection compared to control in both Vero cells and human DCs. The results indicate that siRNAs may be used to attenuate DEN infection in human DCs and may have therapeutic value. Interference of gene expression by siRNAs is a novel strategy to knock down specific genes in cells or tissues, and the specific silencing of pathogen genes using siRNA is a very attractive approach for the clinical treatment of infectious diseases. Long dsRNAs (of >30 nt in length) activate a dsRNA-dependent protein kinase and 2', 5'-oligoadenylate synthetase in mammalian cells, which leads to a non-specific reduction in levels of mRNAs [ 14 ]. The endogenous expression of siRNAs from introduced DNA templates is thought to overcome some limitations of exogenous siRNA delivery, in particular its transient effects on silencing specific genes and loss of phenotype [ 15 ]. AAV vectors have been proven to be safe and efficacious in Phase I clinical trials for gene therapy of cystic fibrosis and hemophilia B and are regarded as a potential alternative to retroviral and adenoviral vectors for gene therapy in humans. The AAV vectors have a number of advantages over other vectors. They are not pathogenic and do not induce production of neutralizing antibodies that could reduce transgene function. They possess a broad-range of tissue tropism and the capability of inducing long-term transgene expression [ 16 ]. In this study, we utilized a novel AAV system to deliver DEN siRNA into mammalian cells and estimated its anti-DEN effect in vitro. In this AAV system, we incorporated the mouse U6 promoter, which is important for transcription and folding of the suppressor RNA, into a plasmid pCMV-U6. The choice of appropriate target genes is necessary for the success of the siRNA strategy, and two siRNAs derived from either the pre-M or the 3' NCR region of DEN-2 were used in our study. An internal deletion of 3' NCR nucleotide sequences was found to be lethal for DEN virus replication in an in vivo study [ 17 ]. The 3' NCR of the flavivirus genome, which presumably functions as a promoter for minus-strand RNA synthesis, is predicted to form a stem-and-loop secondary structure. Computational analyses have revealed that there is conserved sequence in all flaviviruses within the 3' end [ 18 , 19 ]. Thus, two siRNA cassettes were tested in this study that included the 3'NCR sequence common to all four DEN serotypes. The other siRNA cassette is from the gene encoding the preM protein which is important for maturation of the virus into an infectious form. Our test of anti-DEN efficiency showed that siDEN3UT attenuated DEN Infection better than siDENpreM. Knocking down viral genes at the earlier stage of the viral multiplication cycle rather than in the structural protein synthesis phase may provide better antiviral protection, although the limited plasmid transfection ratio appeared to influence the suppression efficiency of siDEN to DEN-2 infection in Vero cells in the present study (Fig. 1C ). The other DEN serotypes will be investigated with our 3'NCR cassette. DEN is transmitted through Aedes aegypti mosquito bite, and resident skin DCs are an early target of DEN infection [ 12 ]. Immature DCs are the most permissive for DEN infection and serve as a source of DEN replication and production [ 20 ]. Replication in the early target cells may be essential for dengue pathogenesis in the human host. In this study, we also utilized peripheral blood iDCs as a cell model to test our AAV system. Similar to results in Vero cells, AAV-mediated siDEN3UT delivery down-regulated DEN-2 protein expression in iDCs. However, the magnitude of suppression in iDCs at the same infectious titer of AAV-siDEN was less compared to that found in Vero cells. Previous data showed that variations in the efficiency of transduction among DCs derived from different normal blood donors is between 2% and 50% [ 21 ], and we found that the infectious ratio for AAVEGFP is about 45%~50% in Vero cells. That may be due to limited expression of the AAV receptor or differential activation of the mouse U6 promoter in Vero cells compared to DCs [ 22 ]. Increasing the AAV infection titer or utilizing a more effective promoter within the AAV vector backbone might elevate the suppression for DEN replication in iDCs. Nevertheless, DCs treated with recombinant AAV showed a significant reduction in DEN virus titer compared to control. This is important as viral titer is the gold standard for measuring antiviral activity. DCs are one of the most powerful of APCs. After infection with virus in the periphery, iDCs process viral antigens, then differentiate into mature DCs and migrate from peripheral tissues to lymph nodes where they prime naïve CD4 and CD8 T lymphocytes to maintain protective antiviral cytotoxic T cell memory [ 23 , 24 ]. Thus, DCs play an important role in the initiation of antiviral immunity and provide a crucial step in the development of adaptive antiviral immunity. Previous data showed that DEN infection induces apoptosis of DCs [ 11 ], which leads to a state of temporary immune-suppression during DEN fever. An important observation in our study is that AAV-siDEN treatment resulted in a significant decrease in apoptotic iDCs. The attenuation of apoptosis in iDCs following AAV-mediated siRNA delivery suggests that AAV-siRNA may be immunologically protective. After the primary DEN infection, most patients appear viremic in the early febrile phase, but the viruses are quickly cleared from the blood system after defervescence [ 25 ]. The activation of both a humoral and cellular immune response is considered to be involved in DEN clearance. The most severe outcome in DEN infection is development of DHF/DSS, which is associated with secondary infections by heterotypic DEN serotypes. It is postulated that the preexisting, cross-reactive, adaptive immune response leads to excessive cytokine production, complement activation, and the release of other inflammatory factors that produce DHF/DSS [ 20 ]. Therefore, it should be imperative for prophylaxis of DHF/DSS to eliminate DEN infection by different serotypes in the early target cells. Attenuation of DEN infection in DCs and protection of infected DCs from apoptosis would be a benefit for the elimination of the early DEN infection and the development and maintenance of antiviral innate/adaptive immune response in vivo . One of the important features of AAV vectors is the lack of inflammation following infection. We failed to detect significant IFNγ or IL-12 production in the supernatants of AAV-siDEN-infected DCs. This is in accordance with previous data [ 26 - 28 ], which demonstrated our AAV delivery system did not induce significant acute inflammatory responses and, therefore, is useful in gene therapy for DEN infection in humans. In conclusion, we developed a novel AAV-mediated siRNA delivery system. Our results demonstrate significant downregulation of DEN protein expression in Vero cells and human DCs, which strongly suggest that our AAV vector can be useful for siRNA delivery and that this AAV system may be applied in clinical settings to attenuate DEN infection. List of abbreviations AAV, adeno-associated virus; DCs, dendritic cells; DEN, dengue virus; DHF/DSS, dengue hemorrhagic fever/dengue shock syndrome; MOI, multiplicity of infection; pEGFP, enhanced green fluorescent protein; siRNA, small interfering RNA; TUNEL, terminal deoxynucleotidyl transferase-mediated dUTP nick end-labeling. Competing interests None declared. Authors' contributions WDZ constructed the si-plasmids, performed cell culture, isolated and administered the virus and isolated and cultured dendritic cells. RS assisted with virus handling and cell culture, and performed flow cytometry. GH did TUNEL assays. XK did cytometric bead array assays and ELISAs. HSJ measured virus titer by plaque assay. RFL, SW, KP and SSM designed and implemented the experiments, performed troubleshooting, and did the analysis and interpretation of the data.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514572.xml
509245	A randomised controlled trial and cost effectiveness study of systematic screening (targeted and total population screening) versus routine practice for the detection of atrial fibrillation in the over 65s: (SAFE) [ISRCTN19633732]	Background Atrial fibrillation (AF) has been recognised as an important independent risk factor for thromboembolic disease, particularly stroke for which it provides a five-fold increase in risk. This study aimed to determine the baseline prevalence and the incidence of AF based on a variety of screening strategies and in doing so to evaluate the incremental cost-effectiveness of different screening strategies, including targeted or whole population screening, compared with routine clinical practice, for detection of AF in people aged 65 and over. The value of clinical assessment and echocardiography as additional methods of risk stratification for thromboembolic disease in patients with AF were also evaluated. Methods The study design was a multi-centre randomised controlled trial with a study population of patients aged 65 and over from 50 General Practices in the West Midlands. These purposefully selected general practices were randomly allocated to 25 intervention practices and 25 control practices. GPs and practice nurses within the intervention practices received education on the importance of AF detection and ECG interpretation. Patients in the intervention practices were randomly allocated to systematic (n = 5000) or opportunistic screening (n = 5000). Prospective identification of pre-existing risk factors for AF within the screened population enabled comparison between high risk targeted screening and total population screening. AF detection rates in systematically screened and opportunistically screened populations in the intervention practices were compared to AF detection rate in 5,000 patients in the control practices.	Background Atrial fibrillation (AF) has been recognised as an important independent risk factor for thromboembolic disease, particularly stroke with which it is associated with a five fold increase in risk [ 1 ]. There are few data on the prevalence of AF in the United Kingdom. Local data derived from the Echocardiographic Heart of England Screening (ECHOES) study suggested a prevalence of AF in people over the age of 65 of 3.8% (95% CI: 2.5–5.1) [ 2 ]. A review of four large community based studies of AF suggested that the overall community prevalence in the United States is 0.89% [ 3 ]. In these studies, the prevalence increased sharply with age: 2.3% of people aged 40 or over; 5.9% of people aged over 65 (higher than the local estimate), and 10% of those over 80. The vast majority (84%) of people with AF are over the age of 65. AF is a particularly important risk factor for stroke in the elderly – while 15% of all strokes are associated with the arrhythmia, it is associated with 36% of strokes in people over the age of 80. The incidence of new cases of AF in people over the age of 65 is of the order of 1% per annum [ 4 ]. Screening for AF in the elderly fulfils many of the Wilson-Jungner criteria for a screening programme [ 5 ]. It is a common and important condition which can be diagnosed by means of a simple test, and the risk of serious sequelae such as stroke can be dramatically reduced by treatment. One UK study has compared systematic nurse-led screening with prompted opportunistic case finding for AF in primary care [ 6 ]. This small scale study (four practices, n = 3001) demonstrated that systematic nurse-led screening detected more cases than opportunistic case finding, however most of those cases detected were already diagnosed. Two further single practice based studies have investigated the role of practice nurses in the screening process [ 7 ], and whole population screening [ 8 ]. 5% of total NHS expenditure can be attributed to stroke, and there would be expected to be about 1,000 new cases of stroke per annum in a typical health authority of a half million population. Therefore, any programme that might lead to an important reduction in stroke incidence needs serious consideration, both because of the potential for health gain, and the potential for reduced overall NHS expenditure. Screening for AF might be one such programme since, in population terms, AF is an important risk factor for stroke and anticoagulation provides a highly effective treatment to reduce this risk. A meta-analysis of randomised controlled trials has shown a 68% relative risk reduction in patients' with AF receiving oral anticoagulation [ 9 ]. It has been estimated that optimal treatment of AF in the population might reduce the overall incidence of stroke by 10%. However, before implementing screening programmes, unresolved questions over how the screening should be conducted must be answered. The appropriate screening strategy to be employed Opportunistic screening The simplest strategy was opportunistic case finding , where a health care professional took the opportunity to feel a patient's pulse during a consultation. If the pulse is irregular, they might make a clinical diagnosis of AF, or request/perform an electrocardiogram (ECG) as a confirmatory test. However, opportunistic case finding is likely to miss a significant proportion of people who would otherwise have benefited from treatment. For example, detection of hypertension in general practice was traditionally detected in an opportunistic way until the introduction of health checks with the 1990 GP contract. The Health Survey for England shows that in 1991, 42% of the population over the age of 75 had hypertension for which they were not taking any medication [ 10 ]. This figure had fallen to 31% by 1994, after the GP contract had taken effect. Targeted screening One possible approach was to screen patients who are at higher risk of AF – a targeted screening programme . Cardiac failure, hypertension and rheumatic heart disease are important precursors of AF [ 7 ]. AF is more common in people with a history of myocardial infarction, angina, diabetes mellitus, hyperthyroidism, stroke or transient ischaemic attack (than in people without these conditions) [ 11 ]. Most general practices were computerised, and some have disease registers. A targeted screening programme could exploit these to identify such high risk patients, either through disease registers, or through prescribing information on the computerised records. Whole population screening Another approach was to screen everyone 65 and over (65+) for AF – a whole population screening programme . A modelling exercise using decision analysis to inform on the methodology for this study indicated that there were not sufficient primary data available to recommend which of these (targeted or whole population) would be the optimum policy The most appropriate screening test for AF 12-lead ECG is recognised as the gold standard test, but this test is time consuming (taking at least 15 minutes to perform in an outpatient setting). Therefore, it is important to consider simpler tests. This study assessed simpler methods compared to the gold standard, both in terms of accuracy, time taken and patient acceptability. These include taking the pulse, and simpler ECGs. Interpreting the ECG Cardiologists offer the most accurate readings of ECGs, but can satisfactory interpretations be obtained by the GP, the practice nurse, or computerised diagnostic software? This study assessed the accuracy of these different approaches to interpreting the ECG. The value of echocardiography The main treatment options to reduce risk of stroke in patients' with AF are currently warfarin or aspirin. Aspirin is much less effective than warfarin – it achieves a barely significant 21% reduction in stroke risk [ 9 ]. However, it is safer to use, since it confers a lower risk of serious haemorrhage. Therefore, in practice, the clinical decision as to which treatment to use depends upon the balance of risks and benefits for the individual patient. Thromboembolic risk is currently determined primarily on clinical criteria. Data from the SPAF study [ 12 ] suggested that echocardiography (Echo) may inform on risk stratification, assisting in therapeutic decision making. The role of routine Echo for patients with AF identified in the community remains to be proven. Data also needs to be quantified regarding the cost effectiveness of Echo versus clinical impression alone. Studies have suggested that the clinical utility in people aged over 74 is poor [ 13 , 14 ]. Therefore this study focused on patients aged 65–74. Once somebody has been identified as having AF, should they also receive an echocardiogram to assess their risk of stroke, or is clinical assessment of risk adequate? Optimum strategy This study, by providing answers to these questions, allowed the optimum strategy for introducing a screening programme for AF in the over 65s to be determined. However, before a decision is made as to whether to institute a screening programme, not only must the question of the best strategy be considered, but also, the question of whether any screening programme at all should be introduced. This study provided data to assist in answering this fundamental question by providing: i) An accurate estimate of the community prevalence and incidence of AF in over 65s; ii) An assessment of the health economic implications of screening for AF; iii) An assessment of the service provision implications of implementing such a programme; iv) An assessment of the impact on patient quality of life and anxiety after various screening methodologies. Health economics of screening Although the cost effectiveness of different approaches to screening is often put in terms of the average cost per case detected, such an approach ignores the sensitivity and specificity of the screening test. This is because average cost per case detected focuses entirely on true positives, paying no attention to false positives, false negatives and true negatives. False positives and false negatives impose costs on patients and health services which would be neglected if the focus was confined to true positives [ 15 ]. An undue emphasis on the average cost per case detected could justify opportunistic screening of a small number of high risk patients who present, with no consideration of the number of cases missed. This study compared the incremental cost per case detected for different methods of AF screening. This refers not to the average cost but rather approximates the incremental cost per case detected in moving from one of the screening options to another. Use of incremental cost per case detected by option shows how the cost per additional case detected is likely to increase as the intensity of screening increases. This method has been used to deal with similar uncertainties about the cost effectiveness of screening for other diseases, including breast and colorectal cancer and has been recommended by the US guidelines [ 16 ]. Objectives Primary objective • To determine baseline prevalence and the incidence of AF based on a variety of screening strategies and in doing so to evaluate the incremental cost-effectiveness, in terms of cost per case identified, of the different screening strategies (targeted or whole population screening) compared with routine clinical practice for detection of AF in people aged 65 and over. Secondary objectives • To evaluate the relative cost-effectiveness of screening methods for AF diagnosis, comparing 12 lead ECG (gold standard) with pulse taking, lead II rhythm strip from standard ECG limb leads alone and single lead thoracic placement ECG. • To evaluate the most cost-effective method of test interpretation, comparing cardiologist (gold standard), with GP, practice nurse, or computerised diagnostic software. • To assess the differing combinations of screening strategies and procedures in terms of patient acceptability and impact on patient quality of life, including any psychological effects of screening. • To determine the community prevalence of AF in people 65+. • To evaluate the value of clinical assessment and echocardiography as additional methods of risk stratification for thromboembolic disease in patients with AF. • To evaluate the service provision implications should screening for AF become a national programme, and identify the optimum screening algorithm for identification of patients with AF. Outcome measures Primary outcome • The incidence of AF according to a variety of screening strategies • The associated costs providing an incremental cost per case detected. The cost data was collected from an NHS and patient perspective. It has focused on resources required to establish screening, time taken to complete screening and the cost of the equipment. Secondary outcomes • Cost effectiveness of 4 different methods of screening for AF. The cost data focused on the difference in the cost of the equipment and the time taken for each of the different methods of screening to be completed. This was from both an NHS and patient perspective. • Cost effectiveness of 4 different methods of ECG interpretation. The cost data focused on the difference in the cost of the grade of staff interpreting the ECG and the accuracy of their interpretation. • Overall community prevalence and incidence of AF • Patient acceptability to AF screening was measured using an adapted version of the screening specific questionnaire used in the Colorectal Screening Programme [ 18 ]. Patient uptake of screening was also monitored. The impact on quality of life was assessed using EQ-5D [ 24 , 25 ]. Patient anxiety was measured using the Spielberger 6 item Anxiety Questionnaire [ 17 ]. • Modelling techniques were used to identify the implications of AF screening on health service provision nationally. This included the effect on echocardiography and anti-coagulation clinic provision. Methods This was a multi-centre randomised controlled trial. The study schema is shown in figure 1 . Figure 1 Study schema for the multi-centre randomised controlled trial. 50 computerised general practices within the West Midlands were recruited through MidReC (Midlands Research Practices Consortium). This was undertaken by writing to all practices in the West Midlands and surrounding counties explaining the study and asking whether they were interested in participating. Practices showing an interest were given further information about the study and invited to attend an investigators meeting. Following the investigator meetings sixty practices interested in participating in the project were randomised (stratified based on Townsend score and practice list size): 25 as intervention, 25 as control practices with 10 reserve practices. A computerised list of all patients aged 65+ was obtained from each practice, and from this a random sample of 10,000 patients from the intervention practices (representing approximately 1 / 3 rd of the total population of patients 65+ in this group), and 5000 from the control practices (representing approx. 1 / 6 th of the total population of patients 65+ in this group) were identified. Patients from intervention practices were randomised, by patient, to opportunistic or systematic groups. All patients within the systematic screening arm, including those with a history of AF, were invited by post to attend a screening clinic. For patients in the opportunistic arm, their notes were flagged within the practice to encourage practice staff to undertake pulse recording. Patients with an irregular pulse were invited to attend a screening clinic. Once this process had been undertaken, the flag was removed from the notes and returned to the research team. The screening clinic was run by practice nurses. Patients gave informed consent. Data collected was baseline information, past medical history (including any previous diagnosis of AF), radial pulse and a 12 lead ECG. The 12 lead ECG was performed using an electronic ECG machine which allowed print-out of single lead thoracic placement ECG and a rhythm strip of lead 2 using limb leads from standard ECG. All 12 lead ECGs were sent to two cardiologists for reporting (GL, MD). Where there was disagreement over the diagnosis a third cardiologist was used to decide. The cardiologists were asked to state whether the ECG showed AF or not, and to state whether there were any other significant abnormalities. Patients were informed of the result within two weeks. Patients with normal ECGs were informed of this, patients with any abnormality were asked to make an appointment with their GP. At the GP appointment patients with AF aged 65 – 74 were offered echocardiography. GPs were asked to make a clinical decision as to thromboprophylaxis both before and after the echocardiogram. Patients with other ECG abnormalities were managed as clinically indicated. At the end of the screening process, GPs and Practice Nurses from both intervention practices (who had received education on ECG interpretation) and control practices (who had received no education) were sent ECGs to interpret for the presence or absence of AF. All ECGs recorded within the study were printed off as either 12 lead, single lead thoracic placement or limb lead recordings. Allocation to ECG type was random and resulted in three equal ECG groups. In order for each interpreter to read all three types of ECG, batches of ECGs were collated with the same numbers of each type of ECG. Allocation to a batch was also random. In total, there were 25 batches of ECGs to match the number of practices in each arm. The GP and Practice Nurse from the same practice read the same batch of ECGs and each batch was read by one control practice and one intervention practice. Therefore each ECG was read by two GPs and two Practice Nurses. All ECGs were anonymised, and practices did not receive any ECGs from their own practice. The interpreters were given a sheet to fill in to indicate for each ECG the presence or absence of AF. All ECGs (as 12 lead) were also analysed by the specific software package accompanying the electronic ECG and results recorded. Patient acceptability and quality of life for different screening strategies were established using EuroQol (EQ-5D) combined together with the Speilberger 6-item Anxiety Questionnaire. EQ-5D allowed the measurement of broad aspects of quality of life. The shortened Speilberger anxiety questionnaire also has proven validity and is more specific to anxiety than is the SF-12 [ 17 ]. An adapted version of the screening-specific tool used in the Colorectal Screening Programme [ 18 ] was used to assess the acceptability of the screening process. A random sample of 750 patients (375 screened patients and 375 opportunistically screened patients) were sent postal versions of the psychological instruments (EQ-5D and Spielberger) on entry to the study (i.e. before the intervention group has received their invitation to attend for screening). One reminder was sent a month later to non-responders. The same questionnaires were sent to the same groups plus those patients who had screened postive at the end of the screening period, approximately 17 months later. This allowed a non-randomised comparison between the effects on quality of life and anxiety in screen positive and screen negative patients. In addition, all patients who were screened were asked to complete the acceptability and Spielberger questionnaire immediately after screening. The patient acceptability questionnaire was also administered to all patients who proceeded for echo. The value of clinical assessment and echocardiography in risk stratification were determined in patients aged 65–74. This compared GP assessment based on the Birmingham guidelines for thromboprophylaxis in AF with any changes in recommendations for treatment once echocardiography results were available to the GP. Sample size and power calculations The assumptions for the power calculations were that patients aged 65 and over represent 17% of the total population; that 40% of study population will be in the high risk group. Also assuming that: 1. Minimum worthwhile change in detection rate was 1% for targeted screening versus routine practice. It is estimated that this change would equate to £10,000 per life year gained. This is based on the following assumptions: a) 60% of new cases of identified AF would be suitable candidates for warfarin b) Annual risk of stroke in this population was 5%, reduced by 60% to 2% if treated c) Costs: £25 to screen a patient; £100 to treat with warfarin pa; £6,000 NHS costs to treat a stroke 2. 50% of patients with AF will be already known to their general practitioner (estimates range from 30% [ 19 ] to 76% [ 20 ]) 3. Community prevalence of AF in this population was 6% [ 2 ] See figure 1 . It was assumed that the baseline prevalence of AF known to the practice (A1) would be 3% (i.e. half of real prevalence of 6%) and that the prevalence of known AF in the control practices would remain constant over the screening period. Thus, the change in the prevalence of known AF in the control practices between baseline to follow up (C2-C1) should be approximately 0%. The change in the GP educated arm (B2-A1) should be marginally higher and is assumed to be between 0 and 1%. The change in the systematic screening arm should, on average, be between 0 and 3% and was assumed to be approximately 2% for the total screening arm (A3-A1) and in the high risk arm (A2-A1) was approximately 3%. All sample size calculations were for 90% power and 5% significance levels unless otherwise stated. a) To detect a 1% difference in detection rate between intervention (GP educated) and control practices (B2-A1) vs (C2-C1). This requires 1,236 patients. However, since this is a difference based at the practice level of randomisation, it needed to be inflated by the design factor. Based on AF prevalence data from the EcHoES (Echocardiographic Heart of England Screening) Study [ 2 ], the between practice variance is 3.7 and the within practice variance is 246. This gave an intra-cluster correlation coefficient of 0.015. The most efficient design in this circumstance would be a cluster size of 200, which gives a design factor of 4. Therefore, 5,000 patients would be needed in 25 practices in both intervention (GP educated) and control groups. b) To detect a 1% difference in detection rate between intervention (Systematic screening total arm) and control practices (A3-A1) vs (C2-C1) . This requires 1,236 patients but when scaled by the design factor of 4 required 5,000 patients. c) To detect a 1.8% difference in detection rate between intervention (Systematic screening high risk arm) and control practices (A2-A1) vs (C2-C1) . This requires 684 patients. However, since this is a difference based at the practice level of randomisation, it also needed to be inflated by the design factor. This meant that 2,736 patients would be needed in each arm. Since the ratio of patients in the two arms is 2:5 this means that 1,916 patients would be needed in the high risk arm and 4,789 in the control arm. With the 2,000 patients expected to be at high risk in this arm – resulting from the 5,000 needed for the previous comparison there were more than enough patients to detect the required difference. Although comparison b) required fewer patients to detect the expected difference (2%) stated in the assumptions, it would be possible to detect differences as low as 1%, should the detection rate not be as high as expected. The a), b) and c) comparisons are all at practice level randomisation. d) To detect a 1% difference in detection rate between high risk screening strategy and routine practice prompted by education (opportunistic arm) (A2-A1) vs (B2-A1). This requires 1,236 patients in both the high risk systematic screening and the GP educated (opportunistic) screening arms of the intervention practices assuming the high risk screening detects a 1% increase and opportunistic screening detects 0% increase. Should the increased detection rates be higher in each arm (1.7% in the high risk arm and 0.7% in the opportunistic arm) then this could require 2,686 patients in each arm. However, since there is a ratio of 2:5 patients in these arms there will be sufficient patients as only 1,880 are needed in the high risk arm and 4,700 in the opportunistic arm to be able to detect this 1% difference . e) To detect a 1% difference in detection rate between total screening strategy and routine practice prompted by education (opportunistic arm) (A3-A1) vs (B2-A1).This requires 3,300 patients in both the total screening and the GP educated (opportunistic) screening arms of the intervention practices. f) To detect a relative risk (RR) of 2 (1% detection rate difference) between total population and high risk screening (A3 vs A2). It was assumed that 40% of the study population fall into the high risk group, and the prevalence of undetected AF is 3%. This meant that 1,434 patients would be needed in each of the two risk groups to detect a two fold difference in risk (i.e. RR of AF in high risk as compared to low risk group is 2). This RR of 2 equates to an increase in AF detection rate from 3% in the total population arm to 4% in the high risk arm. Since there was a 40:60 split in the two risk groups unequal sample size calculations only require a minimum of 1,200 patients in the high risk group and 1,800 in the moderate/low risk group. This was achievable with a screening arm of 5,000 patients, as there would actually be 1,320 in the high risk group and 1,980 in the moderate risk group if a 66% screening acceptance rate was assumed. Sample size for quality of life assessment Although some of the variances are from North American populations we have no reason to suspect that the variation will be different in a British population since data from the ECHOES study on the SF36 gives variations very similar to the North American norms. Spielberger The shortened (6-item) version of the Spielberger state anxiety questionnaire has been validated and used in populations different from that under consideration in SAFE, namely it tends to have been used in young and mostly female populations [ 17 , 21 , 22 ]. The variance obtained from these papers appears to be approximately 144 for Marteau [ 17 ] but higher for the Ubhi [ 22 ] paper. However, the women in the latter paper were being informed of major illness outcomes (either benign or malignant breast cancer). A full Spielberger on people undergoing physiological tests also gave a variance of the order of 144 [ 23 ]. The full version of the Spielberger state anxiety when used with an elderly population also seems to give a variance that is not too far from the previously mentioned papers being 188.8 [ 23 ]. Taking this latter value as being the nearest to our population we can detect a 4 point difference in the mean values obtained with 249 patients in each arm. EQ5D The VAS scale The VAS variance as reported for an elderly population aged 75 and over was 365 [ 24 ] but for a group of recovered stroke patients (ages not given) it was approximately 100 [ 25 ]. Taking the former value as a worst case this means it will be possible to detect a 6% difference between groups on the VAS with 213 patients within each group. The Utility index This was reported in different ways in the Johnson [ 24 ] and Dorman [ 25 ] papers. Using the utility values from the Dorman et al paper the variance is approximately 0.066 and this allows us to detect a 0.1 difference with 139. Using the Johnson paper the variance is approximately 576 and using this means that we can detect a mean change of 7 with 247 patients in each arm. Statistical analysis Intention to treat analysis will be used. Any previously known Atrial Fibrillation cases will be subtracted from the totals obtained at the end of the study to ensure there is no double counting in the incidence figures. Chi-squared, independent t-tests and log-linear models will be used to describe demographic data. If there are differences between the groups this may need to be adjusted for in later analyses. Primary objective: to determine baseline prevalence and incidence of AF on a variety of screening strategies Proportions and rates will be used as the measures of prevalence and incidence. The independent t-test and ANOVA with random effects (as appropriate) will be used to examine the detection rate differences between the intervention and control screening strategies. Should the data be strongly non-normal a non-parametric equivalent will be used. Secondary objectives a) to assess patient acceptability and impact on QoL of different screening strategies The independent and related t-test and ANOVA will be used to examine the differences between the intervention screening strategies on the Spielberger and EuroQol EQ5D. Should the data be strongly non-normal a non-parametric equivalent will be used. Chi squared tests will be used on the screening tool. b) to assess the value of echocardiography in risk stratification for thromboembolic disease in patients with AF McNemar's test will be used to see whether there is any significant change in the doctor opinion on risk of CVA and treatment decision before and after echo screening. c) to evaluate the most cost effective method of test interpretation The cost effectiveness will be covered in the economic section. However, the use of sensitivity, specificity, Cohen's κ and conditional logistic modelling will allow for comparison of the various methods for detecting AF between the GPs, nurses and consultants. d) to evaluate the most cost effective method of screening This will be covered in the economic analysis section. Multivariate and logistic modelling analyses will be undertaken in order to determine which markers might be the best predictors of the presence of AF. This will act as confirmatory analysis for the risk factors used in the screening strategy to define a high risk patient. Economic analysis Framework for the economic analysis This trial evaluated a large number of alternative screening scenarios for identifying atrial fibrillation (2 screening strategies i.e. target v population; 6 screening methods i.e. pulse plus 3 types of ECG if pulse abnormal, and 3 types of ECG regardless of pulse; and 4 screening test interpretations, making 264 = 48 plus control and opportunistic screening = 50). The study has been powered to detect a difference in the targeted versus population arms, and in systematic screening versus routine clinical practice. However, the economic analysis will compare the cost-effectiveness of all alternative screening approaches using a modelling framework, whereby data will be drawn from the trial (where appropriate and available) and from external sources. The use of a modelling approach allows the timescale for the economic analysis to be extended beyond the follow-up period allowed for in the trial. The use of such extrapolation will enable estimation of the incremental cost-effectiveness ratios (ICERs) for each approach: Incremental cost per case detected Incremental cost per life year gained Incremental cost per quality-adjusted life year (QALY) gained Data on consequences The use of EQ-5D allows the measurement of broad aspects of quality of life. EQ-5D allows changes in health status to be measured but also valued, using the University of York Measurement & Valuation of Health general population survey tariff [ 27 ]. Cost data The cost analysis adopted a broad perspective to include costs incurred within the health sector and by patients and carers. Data collection was undertaken on all trial patients in order to allow a stochastic cost analysis to be conducted. The focus of the data collection will be upon the key cost drivers which will include: a) the resources required to establish screening (invitation to patient, follow ups, communication of results), b) the time taken to carry out the various tests, and c) the cost of the equipment (expressed as cost per test). a), b) and c) will be based on data collected in the study. The analysis adopted an incremental approach such that data collection concentrated on resource use differences between alternative screening scenarios. The process of collecting data on resource use was undertaken separately from data collection on unit costs. Resource use data on the screening process was principally collected within the trial. Unit costs were collected from published sources and a representative sample of NHS providers in order to increase generalisability. The methods used in collecting data will include patient questionnaires (see above) and review of patient records (both GP and hospital). Data on private costs were collected from a survey of a sub-cohort of the trial population. Cost effectiveness analysis The plan for the analysis is: 1. Report a cost consequence analysis, which will involve providing a full description of all important results relating to costs and consequences. 2. Conduct both a cost-effectiveness analysis and a cost utility analysis (using data on true positive cases detected as the measure of effect and data on EQ-5D to estimate QALYs). An incremental approach will be used in order to compare the large number of alternative screening strategies. We are interested in comparing the mean costs per patient since our concern is with predicting overall programme costs. However, the data on costs are likely to have a skewed distribution. Therefore, the plan for the analysis of costs is: 1. To explore the nature of the distribution of costs 2. If required, to use non-parametric comparison of means (e.g. bootstrapping) 3. If the distribution of the data is approximately normal, parametric methods will be used. This approach is in line with recent recommendations [ 26 ]. If missing data are a problem at the economic analysis stage, then imputation techniques will be employed. Longer term costs and consequences will be explored by extrapolating beyond the end of the trial using a modelling framework using data from a range of trial and non-trial sources. The precise form of modelling is yet to be determined, but is likely to be either Markov or Discrete Event Simulation, depending upon the extent to which the Markov assumptions are justified. An advantage of using such an approach is that it will allow the additional costs of increasing survival to be explicitly incorporated into the analysis. In particular, modelling will provide estimates of the optimal frequency of screening for AF, based on the estimates of incidence and prevalence from the trial. Identifying untreated patients with AF will have implications for service provision. These will depend upon the prevalence of atrial fibrillation, the screening mechanism employed, the use made of echocardiography, and the additional requirements for anticoagulation monitoring. The separate parallel study, BAFTA, will provide empirical data that will allow such implications for service provision to be assessed. The modelling exercise, which will draw on both SAFE and later BAFTA results, will combine best estimates of both screening and anticoagulation options. The robustness of the results of the economic analysis will be explored using sensitivity analysis [ 27 ]. This will explore uncertainties in the trial based data itself, the methods employed to analyse the data and the generalisability of the results to other settings. Uncertainty in the confidence to be placed on the results of the economic analysis will be explored by estimating cost-effectiveness acceptability curves. These plot the probability that the intervention is cost-effective against threshold values for cost-effectiveness. Inclusion and exclusion criteria Inclusion criteria Patients aged 65 years or over (65+). Exclusion criteria Patients who were terminally ill. Randomisation Randomisation of practices and patients was performed by statisticians from the Department of Primary Care and General Practice at The University of Birmingham. Cluster randomisation of practices to intervention or control was stratified by Townsend quartiles and practice size. Computer searches were carried out to identify cases of known AF, within the sample of patients identified above, using a published strategy [ 15 ]. The randomisation of patients within the intervention practices ensured that the study patients in each practice were divided equally between systematic and opportunistic screening arms and also that there was an even distribution of patients with known AF between the two arms. Patients within the systematic screening arm were identified by computerised record searching as either being at high risk (target population) or moderate risk (non-target population) of AF by recognised criteria [ 28 , 11 ]. Cleaning of lists post randomisation Following initial sampling of the total population the list of patients from each practice were returned to the practices who were asked to remove any patients who had died, moved or were terminally ill. Patients removed following this process were replaced with patients from a reserve list, which had been randomised at the same time as those on the initial lists. Patient information and consent Patients aged 65+ who were selected to the systematic screening arm received an information sheet with an invitation letter to attend the ECG clinic. Entry to the trial was discussed with the practice nurse at the clinic. The practice nurse then obtained written consent from those patients who were willing to participate. Study patients found opportunistically to have an irregular pulse were given an information sheet and invited to attend the screening clinic. Practice staff education and ECG training GPs and other members of the primary health care team in the intervention practices attended investigator days at which they were given educational materials informing them of the importance of detection of AF, and the treatment options that are available. The materials encouraged them to consider opportunistic screening of patients. Members of the primary health care team in control practices received no educational input from the research staff. Practice nurses attended an ECG training day prior to starting the ECG screening clinics. Training included how to perform an ECG (using the Biolog) to ensure a standardised high quality tracing and basic ECG interpretation (specifically how to identify AF). Computerised and note searches of GP records Prevalence and incidence data Computer searches were carried out to identify cases of probable AF in the 15,000 study patients using a published strategy [ 15 ]. Searches were tailored towards the information that is held on computer in each practice. If practices hold AF registers, or use READ diagnosis coding, then these were used. In addition, a search was carried out to identify prescriptions of digoxin, a beta-blocker, a class 1,3 or 4 anti-arrhythmic agent, aspirin or warfarin. This information was recorded into computerised case report forms. Case notes of patients identified as 'known' or 'probable' AF in any of these computer searches were reviewed for mention of a diagnosis of AF. AF diagnosis were drawn from hospital letters stating the existence of the condition or ECG recordings from the last 5 years. An additional 5% random sample of case notes of patients not identified as 'known' or 'probable' AF by computer searching were reviewed (750 in all) to estimate how many other patients who are known to have AF were not identified by the computer searches. If this had revealed a significant number of extra cases of known AF, then the sample size for manual searching would have been increased to allow a precise estimate of the baseline rate of known AF. Unidentified extra AF cases were not found to be significant so no additional note search was required. The same computer searches on both intervention and control practice patients notes were performed prior to, and 12 months after, commencement of screening. Screening clinics All patients in the systematic screening arm and those found to have an irregular pulse in the opportunistic screening arm of the study were invited to attend an ECG screening clinic. At the clinic the practice nurse explained the aims of the study and answered any questions about the study. Written informed consent to participation in the study was obtained from the patient. The nurse then recorded baseline information on age, sex, present smoking and alcohol status and past medical history, including previous diagnosis of AF, and any treatment the patient may be receiving for AF. Radial pulse rate, and whether regular or irregular, was noted. A 12 lead ECG, the gold standard by which other traces were compared, was then recorded using the Biolog machine, which was also able to produce a trace corresponding to the single lead thoracic placement and a rhythm strip of lead II. Finally, the patient was asked to complete an acceptability questionnaire. Discussion This study will identify the most cost-effective strategy for identifying atrial fibrillation in patients aged 65 and over. The policy implications will be dependent on the findings and one of the strengths of the current study is the utilisation of modelling techniques to investigate the implications of different screening strategies and frequency of screening within different health care environments. The initial draft of the report has been submitted to the Department of Health and publication of the results should be expected later this year (2004). Competing interests None declared. Authors' contributions RH and DF were principal investigators. EM and SJ were project managers. DS collected data and prepared the first draft of the paper. JM, EM, JR, SB, MD and GL all contributed to the project design. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC509245.xml
544351	Allele frequencies of hemojuvelin gene (HJV) I222N and G320V missense mutations in white and African American subjects from the general Alabama population	Background Homozygosity or compound heterozygosity for coding region mutations of the hemojuvelin gene ( HJV ) in whites is a cause of early age-of-onset iron overload (juvenile hemochromatosis), and of hemochromatosis phenotypes in some young or middle-aged adults. HJV coding region mutations have also been identified recently in African American primary iron overload and control subjects. Primary iron overload unexplained by typical hemochromatosis-associated HFE genotypes is common in white and black adults in Alabama, and HJV I222N and G320V were detected in a white Alabama juvenile hemochromatosis index patient. Thus, we estimated the frequency of the HJV missense mutations I222N and G320V in adult whites and African Americans from Alabama general population convenience samples. Methods We evaluated the genomic DNA of 241 Alabama white and 124 African American adults who reported no history of hemochromatosis or iron overload to detect HJV missense mutations I222N and G320V using polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) technique. Analysis for HJV I222N was performed in 240 whites and 124 African Americans. Analysis for HJV G320V was performed in 241 whites and 118 African Americans. Results One of 240 white control subjects was heterozygous for HJV I222N; she was also heterozygous for HFE C282Y, but had normal serum iron measures and bone marrow iron stores. HJV I222N was not detected in 124 African American subjects. HJV G320V was not detected in 241 white or 118 African American subjects. Conclusions HJV I222N and G320V are probably uncommon causes or modifiers of primary iron overload in adult whites and African Americans in Alabama. Double heterozygosity for HJV I222N and HFE C282Y may not promote increased iron absorption.	Background Homozygosity or compound heterozygosity for coding region mutations of the hemojuvelin gene ( HJV ) in whites is a cause of early age-of-onset iron overload (juvenile hemochromatosis), and of hemochromatosis phenotypes in some young or middle-aged adults [ 1 - 4 ]. HJV coding region mutations have also been identified recently in African American primary iron overload and control subjects [ 5 ]. We estimated the frequency of the HJV missense mutations I222N and G320V in Alabama white and African American adults who reported no history of hemochromatosis or iron overload, because primary iron overload unexplained by typical hemochromatosis-associated HFE genotypes is common in whites and blacks in Alabama [ 6 - 8 ], and HJV I222N and G320V were detected in an Alabama juvenile hemochromatosis index patient [ 3 ]. In addition, HJV I222N and G320V have been described in white adults with hemochromatosis phenotypes in other geographic areas [ 1 , 2 , 9 ]. Methods General criteria for selection of study subjects Performance of this study was approved by the Institutional Review Boards of the University of Alabama at Birmingham and Brookwood Medical Center. Unrelated white subjects were spouses of patients who attended a hematology and medical oncology outpatient clinic; hospital employees; and controls who participated in a sleep study. Unrelated control black subjects were spouses of patients who attended a hematology and medical oncology outpatient clinic; hospital employees; controls who participated in a sleep study; and controls who participated in a study of gestational diabetes mellitus. All subjects were adults (≥ 18 years old). No subject reported a previous diagnosis of hemochromatosis or iron overload. Serum transferrin saturation, serum ferritin concentration, or other indicators of iron metabolism were not measured in most study subjects. Polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) analyses Genomic DNA was isolated from blood buffy coat as described previously [ 6 ]. HJV I222N and G320V were genotyped by standard PCR-RFLP and agarose electrophoresis techniques. Exon 4 of the HJV locus was amplified using the primers HJEx4AF and HJEx4BR previously described [ 3 ]; this allowed for the amplification of the region containing both the I222N and G320V mutations of the HJV locus. Reaction conditions were as follows: 10 mM KCl, 10 mM (NH 4 ) 2 SO 4 , 20 mM Tris-HCl (pH 8.8), 0.1% Triton X-100, 3.2 mM MgSO 4 , 200 μM dATP, 200 μM dCTP, 200 μM dGTP, 200 μM dTTP, 1 μM of each primer, 0.2 U/μL of Taq Polymerase. Amplification conditions on a Biometra UNO II thermocycler (LabRepCo/LabRepNet, Horsham, PA) consisted of an initial denaturing at 94°C for five minutes followed by thirty cycles of 94°C for 30 seconds, 60°C for 30 seconds and 72°C for 30 seconds. Delineation of c.665T→A of the HJV exon 4 nucleotide sequence defining I222N was made utilizing the restriction endonuclease Bcc I (New England Biolabs, Beverly, MA), which recognizes the nucleotide c.665T of I222 as a restriction site, and not the nucleotide c.665A of N222. Similarly, the nucleotide polymorphism site c.959G→T defining the G320V polymorphism of the HJV exon 4 locus was determined by using the restriction endonuclease Ban I (New England Biolabs, Beverly, MA). Ban I recognizes the nucleotide c.959G of G320 as part of the restriction site sequence, but not the nucleotide c.959T of V320. DNA specimens from the parents of a white juvenile hemochromatosis index case were used as positive controls for HJV I222N and G320V [ 3 ]. Statistical considerations The data set consisted of observations on 241 Alabama white and 124 African American adults. Analysis for HJV I222N was performed in 240 whites and 124 African Americans. Analysis for HJV G320V was performed in 241 whites and 118 African Americans. Results are expressed as allele frequencies and 95% confidence intervals (CI). To estimate the CI for frequencies of HJV mutations that were not detected in the present subjects, we computed the allele frequency as the quotient of single minimal hypothetical value (= 0.01) and the number of chromosomes corresponding to the respective groups of subjects. Computations were performed with GB-STAT (v10.0; Dynamic Microsystems, Inc., Silver Spring, MD). Results Allele frequencies of HJV I222N and G320V These results are displayed in Table 1 . One of 240 white subjects was heterozygous for HJV I222N, but this mutation was not detected in African American subjects. HJV G320V was not detected in either white or African American subjects. Table 1 HJV missense mutations in Alabama adult subjects from the general population. 1 HJV mutation I222N G320V Whites tested, n 240 241 Whites with mutation, n 1 (heterozygote) 0 Allele frequency (95% CI) 0.0021 (0, 0.0062) 0 (0, ~0.00006) African Americans tested, n 124 118 African Americans with mutation, n 0 0 Allele frequency (95% CI) 0 (0, ~0.00012) 0 (0, ~0.00013) 1 HJV mutations were detected using PCR-RFLP technique. DNA specimens from the parents of a white juvenile hemochromatosis index case were used as positive controls for HJV I222N and G320V [3]. CI = confidence interval. Characteristics of a HJV I222N heterozygote Medical records of the white subject heterozygous for HJV I222N were available for review. This 29 year-old woman was also heterozygous for HFE C282Y. She reported that she had had two normal pregnancies and one spontaneous abortion. She has always experienced heavy menstrual flow, but she reported no other significant blood loss. She reported no blood donation. She reported that she eats a variety of meats, vegetables, and other foods; there was no history of supplemental iron ingestion. Her hemoglobin level was 12.8 g/dL, platelets 113,000/mm 3 , transferrin saturation 21%, and serum ferritin 23 ng/mL; marrow examination performed to evaluate thrombocytopenia revealed normal cellular morphology and normal iron stores. Discussion A variety of HJV coding region mutations occur in homozygous or compound heterozygous configuration in persons with juvenile or adult-onset hemochromatosis phenotypes [ 1 - 5 , 9 - 11 ]. However, these types of HJV -associated hemochromatosis are collectively uncommon [ 1 - 3 , 11 , 12 ]. HJV mutations were also uncommon in control whites studied with dHPLC (denaturing high-performance liquid chromatography) [ 9 ]. In 200 Greek volunteer blood donors, none carried the HJV G320V mutation detectable by PCR-RFLP analysis, suggesting that the frequency of the G320V allele in the Greek population is lower than 0.004 [ 11 ]. In French control subjects, for example, only the HJV missense mutations L101P (two heterozygotes) and E302K (one heterozygote) were identified in 333 control subjects, representing an aggregate HJV mutation frequency of 0.0045 (95% CI: 0, 0.0096) [ 9 ]. The present results in Alabama white control subjects are consistent with these previous reports. Taken together, these observations confirm and extend previous observations that the individual or aggregate frequency of HJV coding region mutations is low in general populations of whites [ 9 ]. Heterozygosity for a HJV missense mutation was associated with increased severity of iron overload in nine of 310 French patients with hemochromatosis and HFE C282Y homozygosity who were studied with dHPLC [ 9 ]. In 48 white U.S. hemochromatosis patients with HFE C282Y homozygosity studied with dHPLC, no HJV coding region mutation was detected [ 5 ]. Thus, aggregate HJV coding region mutation frequencies in C282Y homozygotes with a hemochromatosis phenotype in France and the U.S. are similar (0.0148 vs. 0, respectively; p = 0.2716, Fisher exact test). The aggregate frequency of HJV coding region mutations was similar in French hemochromatosis patients with HFE C282Y homozygosity and in French control subjects (0.0145 vs. 0.0045, respectively; p = 0.0564, Fisher exact test), although the hemochromatosis patients with HJV coding region mutations had more severe iron overload, on the average, than did hemochromatosis patients without HJV coding region mutations [ 9 ]. Double heterozygosity for HJV I222N and HFE C282Y was not associated with evidence of increased iron absorption in the present case. Some persons with JH have common HFE genotypes, including C282Y heterozygosity, H63D heterozygosity or homozygosity, and S65C heterozygosity [ 13 - 18 ]. However, these HFE genotypes are infrequently associated with a severe hemochromatosis phenotype [ 6 , 7 , 19 - 25 ]. Further, sequencing HFE introns and exons in English and French Canadian JH cases did not reveal novel HFE mutations that could likely explain the development of iron overload [ 14 , 15 ]. In 310 C282Y homozygotes in France, HJV mutations were relatively common and were associated with greater severity of iron overload [ 9 ]. In 48 C282Y homozygotes in the U.S. who had severe iron overload phenotypes, no HJV coding region mutation was detected [ 5 ]. The frequency of H63D is significantly greater in persons with cardiomyopathy than in normal control subjects [ 26 ]. The brother of a Spanish JH proband had evidence of iron overload associated with H63D homozygosity and heterozygosity for a putative JH-associated Ch1q haplotype [ 17 ]. Although these reports are consistent with observations in another JH patient with cardiomyopathy who had HFE H63D [ 3 , 12 ], other persons with JH and cardiomyopathy did not have H63D [ 15 ]. Altogether, these reports indicate that further exploration of the potential role of variant HJV alleles in the clinical expression of iron overload with particular reference to HFE hemochromatosis is warranted [ 2 ], and that the frequency and biological significance of HJV alleles may vary among racial/ethnic groups. Primary iron overload in African Americans is phenotypically and genetically heterogeneous [ 8 ]. Some patients have common HFE mutations, a common allele of the ferroportin gene ( FPN1 Q248H), or heritable types of anemia [ 8 , 27 , 28 ], but the genetic basis of most cases remains unknown. HJV I222N and G320V were not detected in the present African American subjects. However, HJV coding region mutations were recently identified in two of 51 African American subjects with iron overload: exon 3, nt 205–207 ins GGA (ins G69) (frequency 0.0098); and exon 4, nt 929 C→G (A310G) (allele frequency 0.0196) [ 5 ]. The HJV alleles ins G69 and A310G were also identified in African American control subjects (allele frequencies 0.0038 and 0.0720, respectively) [ 5 ]. Taken together, these observations suggest that HJV coding region mutations are uncommon in African Americans, but may account for some cases of primary iron overload [ 5 ]. Conclusions HJV I222N and G320V are uncommon causes or modifiers of primary iron overload in adult whites and African Americans in Alabama. Double heterozygosity for HJV I222N and HFE C282Y may not promote increased iron absorption. Competing interests The author(s) declare that they have no competing interests. Authors' contributions JCB participated in conceiving the study, provided DNA specimens, performed statistical evaluation, and wrote part of the manuscript. CAR devised the PCR-RFLP assay, provided DNA specimens, participated in data collection, and wrote part of the manuscript. SN and SB performed testing on DNA specimens and participated in data collection. RTA participated in conceiving the study, provided DNA specimens, and wrote part of the manuscript. All authors approved the final version of the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544351.xml
546234	Human embryonal epithelial cells of the developing small intestinal crypts can express the Hodgkin-cell associated antigen Ki-1 (CD30) in spontaneous abortions during the first trimester of gestation	Background Ki-1 (CD30) antigen expression is not found on peripheral blood cells but its expression can be induced in vitro on T and B lymphocytes by viruses and lectins. Expression of CD30 in normal tissues is very limited, being restricted mainly to a subpopulation of large lymphoid cells; in particular, cells of the recently described anaplastic large cell lymphoma (ALCL), the Reed-Sternberg (RS) cells of Hodgkin's lymphoma and scattered large parafollicular cells in normal lymphoid tissues. More recent reports have described CD30 expression in non-hematopoietic and malignant cells such as cultured human macrophages, human decidual cells, histiocytic neoplastic cells, mesothelioma cells, embryonal carcinoma and seminoma cells. Results We investigated the immunohistochemical expression of CD30 antigen in 15 paraffin-embedded tissue samples representing small intestines from fetuses after spontaneous abortion in the 8th, 10th and 12th weeks using the monoclonal antibody Ki-1. Hormones had been administered to all our pregnant women to support gestation. In addition, a panel of monoclonal antibodies was used to identify leukocytes (CD45/LCA), B-lymphocytes (CD20/L-26) and T-lymphocytes (CD3). Our findings were correlated with those obtained simultaneously from intestinal tissue samples obtained from 15 fetuses after therapeutic or voluntary abortions. Conclusions The results showed that: (1) epithelial cells in the developing intestinal crypts express the CD30 (Ki-1) antigen; (2) CD30 expression in these epithelial cells is higher in cases of hormonal administration than in normal gestation. In the former cases (hormonal support of gestation) a mild mononuclear intraepithelial infiltrate composed of CD3 (T-marker)-positive cells accompanies the CD30-positive cells.	Introduction CD30 antigen, a member of the tumor necrosis factor (TNF) receptor superfamily [ 1 - 3 ], was originally identified as a cell surface antigen on primary and cultured Hodgkin's and Reed-Sternberg cells by use of the monoclonal antibody Ki-1 [ 4 , 5 ]. CD30 antigen is normally expressed by a subset (15–20%) of CD3+ T cells after activation by various stimuli [ 6 ]. Its expression is stimulated by interleukin (IL)-4 during lineage commitment of naïve human T cells [ 7 , 8 ] and is augmented by the presence of CD28 co-stimulatory signals [ 9 ]. CD30 also is expressed at variable levels in different non-Hodgkin's lymphomas (NHL) as well as in several virally transformed T and B cell lines [ 5 , 10 ]. In particular, CD30 is a specific marker of a subset of peripheral T cell NHLs known as anaplastic large cell lymphomas (ALCL) [ 5 ]. More recently, preferential CD30 expression has been detected on a subset of tissue and circulating CD4+ and CD8+ T cells producing mainly Th2 cytokines in immunoreactive conditions [ 11 - 14 ]. CD30 appears to have an important immunoregulatory role in normal T cell development. Within the thymus, CD30L is highly expressed on medullary thymic epithelial cells and on Hassal's corpuscles [ 15 ]. Pallesen and Hamilton-Dutoir [ 16 ] were the first to report CD30 expression outside lymphoid tissue in 12 out of 14 cases of primary or metastatic embryonal carcinoma (EC) of the testis, using immunostaining with the monoclonal antibodies (MAbs) Ber-H2 and Ki-1. Subsequently, several investigators have confirmed their results and have detected CD30 in these carcinomas at the protein [ 17 - 20 ] and the mRNA [ 10 ] level. Two reports demonstrated CD30 expression in 4/21 and 4/63 cases of testicular and mediastinal seminoma [ 21 ] and in the seminomatous components of 7/14 cases of mixed germ cell tumours of the testis [ 22 ]. Suster et al. detected the CD30 antigen in 6/25 yolk sac tumours of the testis and mediastinum [ 22 ]. CD30 expression has also been reported in other non-lymphoid tissues and cells such as soft tissue tumours [ 23 ], decidual cells [ 24 , 25 ], lipoblasts [ 26 ], myoepithelial cells [ 27 ], reactive and neoplastic vascular lesions [ 28 ], mesotheliomas [ 29 ], cultivated macrophages, and two histiocytic malignancies [ 30 ]. Primitive crypts (epithelial downgrowths into the mesenchyme between the small intestinal villi), appear in the postpharyngeal foregut between the 9 th and 12 th weeks of embryo development. Goblet cells are present in small numbers after 8 weeks, Paneth cells differentiate at the base of the crypts in weeks 11 and 12, and enteroendocrine cells appear between weeks 9 and 11. The fact that the CD30 molecule can mediate signals for cell proliferation or apoptosis [ 2 ] prompted us to perform a systematic investigation of CD30 antigen expression in non-hematopoietic embryonal tissues during the proliferation and differentiation stages, beginning with the epithelial cells of the developing intestinal crypts. Materials and methods Samples representing 15 small intestines from fetuses after spontaneous (involuntary) abortion occurring in pregnant women treated with progesterone (300–600 mg per day until the 12th gestational week), and 15 small intestines from fetuses after therapeutic or voluntary abortion, were obtained in the 8th, 10th and 12th weeks of gestation. The Regional Ethics Committees approved the study. Written informed consent was obtained from all individuals and the procedures followed accorded with institutional guidelines. Small intestines were cut in 3 mm slices and fixed in 10% neutral buffered formaldehyde at 4°C for 24 h, then processed for routine paraffin embedding. Paraffin blocks were available in all cases, and 3 μm thick tissue sections were stained routinely with hematoxylin-eosin, PAS and Giemsa, and subsequently by immunohistochemistry. Immunoperoxidase labeling was performed as follows: sections were deparaffinized in 70% alcohol and endogenous peroxidase was blocked with 3% H 2 O 2 in methanol. The sections were preincubated in 20% serum of the species from which the secondary antibody was raised, and the primary antibody was applied. After overnight incubation at room temperature, the secondary biotinylated antibody was applied for 30 min. Staining was visualized with a Vector Elite System (Vector Laboratories, Burlingame, CA) using diaminobenzidine as the chromogen. The sections were counterstained with dilute hematoxylin. The primary antibodies used were as follows: (CD30/Ki-1) activated lymphoid cells, mouse monoclonal antibody (Novocastra); (CD45/LCA) leukocyte common antigen, mouse monoclonal antibody (Dako); (CD20/L-26) B-lymphocytes, mouse monoclonal antibody (Dako); and (CD3) T-lymphocytes, mouse monoclonal antibody (Dako). We used the high temperature antigen unmasking technique for immunohistochemical demonstration of CD30/Ki-1 on paraffin sections (Novocastra). Control slides were incubated with nonimmunized rabbit serum. An anaplastic lymphoma case-slide (positive control) was run in parallel with the assay. Analysis of CD30/Ki-1 positive cryptae cells For each sample, the CD30/Ki-1 positive population was assessed by enumeration of labeled cells in each tissue compartment for a minimum of five random fields per section viewed at 40-fold magnification through a grid. Cell numbers were calculated per mm 2 of tissue section. The counted areas were selected from random tissue sections, taking into account that the ratio of the area of the intestinal stroma to the area of surface epithelium covering the crypts was representative of the entire field. Areas with obvious necrosis or haemorrhages were excluded. Statistical analysis was performed using the ANOVA test. Results Five microscopic fields of the small intestines were evaluated in each case without knowledge of the clinical data (TABLE 1 ). Two observers examined the sections independently, and positive cellular staining for each antibody was manifested as fine brown cytoplasmic granularity and/or surface membrane expression. Table 1 Expresion of CD30 antigen in fetal intestinal cryptae cells during the first trimester of gestation. Spontaneous abortions Voluntary abortions 8th week 10th week statistics 8th week 10th week statistics CD30(+)cells/mm 2 3.61+/0.16 5.27+/-0.19 p < 0.0001 3.42+/-0.17 3.43+/-0.18 p = 0.95 8th week 12th week statistics 8th week 12th week statistics CD30(+)cells/mm 2 3.61+/-0.16 5.34+/-0.23 p < 0.0001 3.42+/-0.17 3.41+/-0.17 p = 0.95 8th week of gestation In cases of spontaneous (involuntary) abortion, immunohistochemistry revealed small clusters or scattered, large-sized CD30/Ki-1 positive cryptae cells within the intestine in all settings examined (Fig. 1 ), with percentages varying from 3.2 to 3.9 (mean ± sd = 3.61 ± 0.16). In the neighbouring intestinal stroma a slight cellular infiltration was observed, consisting of rounded mononuclear cells approximately 10 μm in diameter with eccentric kidney-shaped nuclei and expressing a CD45/LCA and CD3 phenotype. In cases of voluntary or therapeutic abortion, immunohistochemistry showed a smaller number of large-sized CD30/Ki-1 positive cryptae cells in all settings examined (Fig. 2 ), with percentages varying from 3.1 to 3.7 (mean ± sd = 3.42 ± 0.17). No inflammatory infiltrates or necrosis were noted in the neighbouring intestinal stroma. Figure 1 8 th week of gestation (involuntary abortions) . Ki-1 (CD30) antigen is expressed by a small number of epithelial cryptae cells. Immunohistochemical stain X 400. Figure 2 8 th week of gestation (voluntary abortions) . Weak to moderate expression of Ki-1 (CD30) antigen in the developing crypts. Immunohistochemical stain X 400. 10th week of gestation In cases of spontaneous abortion, immunohistochemistry showed a higher number of positive CD30/Ki-1 cryptae cells than at the 8th week of gestation (Fig. 3 ), with percentages varying from 4.9 to 5.6 (mean ± sd = 5.27 ± 0.19). There were very few inflammatory infiltrates in the intestinal stroma expressing the phenotype CD45/LCA and CD3. In cases of voluntary or therapeutic abortion, the frequency of CD30/Ki-1 positive cryptae cells was similar to that at the 8th week of gestation, with percentages varying from 3.2 to 3.9 (mean ± sd = 3.43 ± 0.18). No inflammatory infiltrates or necrosis were noted in the neighbouring intestinal stroma. Figure 3 10 th week of gestation (involuntary abortions) . Strong expression of Ki-1 (CD30) antigen in the developing crypts. Immunohistochemical stain X 400 12th week of gestation In spontaneous abortion cases the number of CD30/Ki-1 positive cryptae cells was even higher than at 10th week, with percentages varying from 4.8 to 5.7 (mean ± sd = 5.34 ± 0.23). The number in cases of voluntary or therapeutic abortions was more or less the same as at 8th and 10th weeks, with percentages varying from 3.2 to 3.7 (mean ± sd = 3.41 ± 0.17). No differences in immune reaction were noted in the neighbouring intestinal stroma in cases of either spontaneous or voluntary/therapeutic abortion in comparison to the 8th and 10th gestational weeks. The differences among the numbers of CD30/Ki-1 positive cells at the 8th, 10th and 12th gestational week after spontaneous abortion were statistically significant (p < 0.0001). No significant differences were observed in the numbers of these cells after voluntary or therapeutic abortions (p = 0.95). Discussion The value of the CD30 antigen as a diagnostic marker for Hodgkin's lymphoma and anaplastic large cell lymphoma is well documented [ 4 , 5 , 31 ]. However, the function of this cytokine receptor in Hodgkin's lymphoma and other CD30-positive diseases is still not clear. CD30 appears to have an important immunoregulatory role in normal T cell development. In normal cells, this transmembrane glycoprotein can be induced on B and T lymphocytes by mitogen stimulation or viral transformation [ 32 - 34 ]. cDNA cloning has revealed that the CD30 protein is a cytokine receptor of the tumor necrosis factor receptor superfamily [ 1 , 35 ], the ligand of which belongs to the tumor necrosis factor family [ 22 , 23 ]. Recent in vitro data indicate that the CD30 receptor-ligand complex can mediate signals for cell proliferation, apoptosis and cytotoxicity in lymphoid cells [ 20 , 36 , 37 ]. Our results give the first indication that the CD30 antigen is expressed in the epithelial cells of developing intestinal crypts. This observation has a number of important implications. First, our findings are of significance with regard to the accepted origin of R-S cells. Care must be taken when drawing histogenetic conclusions based on the identification of a single marker in different cell types. Shared expression of CD30 antigen does not necessarily relate Hodgkin and R-S cells to activated lymphocytes. The identification of this antigen in cells as apparently disparate as activated lymphocytes, R-S cells and now human epithelial cells of the developing fetal intestinal crypts suggests that previous views about the nature of the Ki-1 antigen must be re-examined. The Hodgkin and Reed-Sternberg cells are indeed lymphocytes as they harbor rearranged immunoglobulin (in more than 90% of cases) and T cell receptors [ 38 ]. Although the expression of CD30 antigen may indicate a relationship between these cell types, it is likely to be less straightforward than was previously supposed. Identification of the normal physiological role of CD30 antigen is thus made even more imperative if these relationships are to be understood. Second, these findings indicate that outside the lymphatic system, CD30 antigen expression in the epithelial cells of developing intestinal crypts can mediate signals for cell proliferation and differentiation in a region where other cell types (stem, goblet, Paneth and enteroendocrine) grow throughout life. Third, CD30 expression in the epithelial cells of the developing intestinal crypts is induced by progesterone. This is a novel mechanism of CD30 induction, distinct from neoplastic transformation and viral infection of lymphocytes. The demonstration of large R-S like cells in the developing crypts within a lymphoplasmacytic infiltrate, in the same way that similar R-S like cells are observed in reactive lymph nodes, especially within the parafollicular areas, is evidence that such cells might represent the physiological counterparts of R-S cells. The possibility that CD30 is an oncofetal antigen is supported by our positive findings in fetal intestinal cryptae cells. We have so far been able to investigate only a single tissue from a small number of fetuses of early gestational age. Pallesen and Hamilton-Dutoit [ 16 ] examined CD30 expression in normal adult, neonatal and fetal (week 28) testes, as well as other tissues (brain, spinal cord, lung, gut, kidney, erythropoietic tissue, muscle, bone and connective tissue) from fetuses of 11 and 12 weeks gestational age, with negative results. This is the first demonstration of CD30 in epithelial cells in fetal tissue. Although our results require confirmation from frozen sections, they – together with a reported positive staining in placenta [ 24 , 25 ] – suggest that the antigen is expressed by proliferating and differentiating epithelial cells of other than lymphoid origin. Clearly the extent of expression of CD30 antigen in embryonal tissues warrants further investigation.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546234.xml
514566	In vivo transcriptional profiling of Plasmodium falciparum	Background Both host and pathogen factors contribute to disease outcome in Plasmodium falciparum infection. The feasibility of studying the P. falciparum in vivo transcriptome to understand parasite transcriptional response while it resides in the human host is presented. Methods A custom made oligonucleotide array with probes based on the P. falciparum 3D7 laboratory strain chromosome 2 sequence was used to detect in vivo P. falciparum transcripts. This study analyzed transcripts from total RNA derived from small blood samples of P. falciparum infected patients and compared the in vivo expression profile to the in vitro cultivated 3D7 strain transcriptome. Results The data demonstrated that in vivo transcription can be studied from a small blood sample, despite the abundance of human RNA. The in vivo transcriptome is similar to the 3D7 ring stage transcriptome, but there are significant differences in genes encoding a sexual stage antigen and surface proteins. Conclusions Whole genome transcription analysis of P. falciparum can be carried out successfully and further studies in selected patient cohorts may provide insight into parasite in vivo biology and defense against host immunity.	Background Plasmodium falciparum infection remains a major health problem worldwide. Its complex life cycle has hampered standard methods for the study of pathogenesis. New approaches to elucidate parasite biology using whole genomic methods have provided insight into gene function, transcriptional regulation and stage specific biology [ 1 - 4 ]. Characterization of the in vivo biology of this pathogen, through adaptation of a whole genome approach, would provide insight into the host-parasite relationship, parasite virulence factors and inform new strategies for intervention. Genomic scale transcriptional profiling of P. falciparum during a natural infection is presented. Small amounts of parasite RNA, isolated from a few milliliters of a blood sample are found to be sufficient for whole genome transcriptional analysis. This data show that several genes are differentially expressed in vivo , indicating differences between the transcriptional program of 3D7 laboratory strain parasites growing in culture and naturally occurring infections in the human host. Whole genome expression has been used in studies of bacterial pathogenesis to identify genes that are specifically transcribed under in vivo conditions [ 5 - 7 ]. For example, genes involved in amino acid transport and metabolism are upregulated in Pasteurella multocida in vivo as compared to in vitro conditions [ 8 ]. Similarly, analysis of P. falciparum gene expression patterns, particularly the subset of genes that are specifically expressed in the in vivo state may identify unique parasite biology when it resides in the host environment. Processes involving parasite metabolism, immune evasion and transmission may be altered in the highly specialized environment of the human host as compared to in vitro conditions. In addition, approximately 12% of P. falciparum's predicted genes have not been found to be expressed in any of the life cycle stages previously studied [ 9 ]. Whole genomic analysis of the parasite in vivo may reveal the unique expression of such genes in vivo , providing additional targets for intervention. Methods Parasite isolates This study was conducted as part of an ongoing P. falciparum chloroquine resistance study in Senegal [ 10 ]. Patients with mild P. falciparum malaria gave consent for the study and were enrolled at an outpatient health clinic. Patients underwent venipuncture using K 3 EDTA coated Vacutainers (Beckton Dickinson) and from this sample, 1.6–2.5 ml of whole blood was collected and passed through a white cell depletion filter using a 20 ml syringe. The filtered sample was centrifuged for 5 minutes at 3,200 rpm in a clinical centrifuge and placed in Tri-Reagent BD (Molecular Research Center). The samples were vortexed and stored at minus 70°C. Samples were thawed in a room temperature bath one month later and isolation of RNA was completed per Molecular Research Centers protocol. Three samples obtained in Senegal that had the highest parasitemia and largest blood volume are presented. A 14 ml blood sample from a P. falciparum infected traveler from Nigeria was similarly processed in Boston. Oligonucleotide array analysis Labeling and hybridization of total RNA was performed as described [ 2 ]. Expression levels were calculated using the Match Only Integral Distribution Algorithm (MOID) [ 11 ]. The presence or absence of gene expression was determined using methods previously described [ 9 ]. The design of the probes to human ESTs was based on UniGene version 116. Real time PCR To confirm array data, a subset of genes (PFB0120, PFB0100, PFB0270 PFB0355, and PFB0065) that vary from high to low level abundance by array were quantified using real time PCR from cDNA. PFB0120: forward primer 5'-CAG CCC TCT TAG CTC TCA ACT TC-3', reverse primer 5'-AGC AAC AGC AGA GGC TAT AGA ACT-3', PFB0100: forward primer 5'-CAC CAA ATG GCT ATG CTT ATG GA-3', reverse primer 5'-TTC CAG GAG CAC CAT TAA ATC CT-3', PFB0270: forward primer 5'-ACA CTT ACT GGT ATT TCG GAA TTT-3', reverse primer 5'-TAA TTG TCC ATA TTC TTC AAT ATA T-3', PFB0355: forward primer 5'-ATT GTA AGA AAT AGT TGG GGT-3', reverse primer 5'-TAT ATC ATG CTC CTT CTT ATC A-3', PFB0065: forward primer 5'-CGT TGG TAG TGC GTT CCT TAC AA-3', reverse primer 5'-GTT CCT GCT ATA TCA GGA GCA CCA-3'. Sequence analysis confirmed the identity of the amplification products. 3D7 strain parasites were cultivated under standard conditions and synchronized with 5% sorbitol to obtain ring stage parasites for extraction of total RNA [ 12 , 13 ]. cDNA was synthesized from total RNA from the Nigerian in vivo sample and 3D7 ring stage total RNA using Super Script 1 st Strand synthesis system (Invitrogen). Duplicate reactions using real time PCR were performed with 1 μl cDNA with gene specific primers in 50 μl reaction volume using fluorescent dye SYBR Green (SYBR Green PCR Master Mix, Applied Biosystems). The reactions were carried out on an ABI PRISM model 7700-sequence detector and all PCR reactions amplified a single product as determined by dissociation curve analysis (Dissociation Curve Software, Applied Biosystems). Statistical tests Variation between samples was assessed using Kruskal-Wallis method (non-parametric ANOVA) to test the null hypothesis. To normalize samples, the mean gene expression level was calculated for all Plasmodium genes between the 10 th and 90 th percentile with at least six probes. Analysis based on rank was a second method used; in each experiment the probe intensity was ranked and this resulted in equivalent quantile distribution for all probes between two experiments. This method is more conservative and will define relative rank changes between experiments and is independent of potential normalization artifacts. Human subjects Patient blood samples were collected after informed consent was obtained. The study was approved by the institutional review boards at Harvard School of Public Health, Brigham and Women's Hospital and Cheikh Anta Diop University. Results To evaluate the integrity of the RNA transcripts from the in vivo isolated samples a denaturing RNA gel was carried out (Figure 1 ). The ribosomal bands are sharp with minimal RNA degradation. Despite buffy coat depletion there are human ribosomal bands present in addition to P. falciparum ribosomal bands. Human ribosomal bands are not seen on a denaturing RNA gel from in vitro cultivated 3D7 (data not shown). The most abundant transcript of human origin in the in vivo was haemoglobin RNA (Table 1 ). Human transcripts are also detected in 3D7 in vitro samples, but at a lower level of abundance. Figure 1 Denaturing gel of total RNA isolated from P. falciparum infected patient blood reveals both human and parasite ribosomal RNA. Total RNA was electrophoresed on a 1.3% formaldehyde agarose gel and stained with ethidium bromide. Marker (M) marks RNA size ladder. Three patient samples were run in lane 1–3. Closed arrows represent human ribosomal RNA 28s: 4700 bp, 18s:1900 bp; Open arrows mark P. falciparum ribosomal RNA 28s:4104 bp, 18s:1384 bp. Table 1 Human transcripts detected in P. falciparum infected patient blood samples and in vitro cultivated 3D7 samples Comparison of expression level of the most abundant human transcripts detected in in vivo blood and in 3D7 in vitro cultivated samples. Average expression level of in vivo isolated RNA derived from the average expression level of three samples from Senegal and one from Nigeria compared with the average expression level from three 3D7 ring stage in vitro isolated RNA samples. Transcripts are listed in order of highest abundance in vivo as detected by array. Expression levels are reported as expression units (EU), using the MOID algorithm. Expression Units Gene Description in vivo in vitro 3D7 Hs.155376_at haemoglobin, beta 757266 101307 Hs.36977_at haemoglobin, delta EST, similar to tctp human translationally controlled tumor protein H. sapiens 395849 12265 Hs.203820_at 241506 6932 Hs.21295_at EST, Weakly similar to KIAA0902 protein H. sapiens 190454 9321 Hs.247921_at haemoglobin, theta 1 188961 4528 Hs.87246_at Bcl-2 binding component 3 187575 5688 Hs.256047_at ESTs, similar to B24264 proline-rich protein MP3 - M. musculus 160290 5681 Hs.14587_at ESTs, similar to AF1 51 859_1 CGI-101 protein H. sapiens 155768 6389 Hs.117848_at haemoglobin, epsilon 1 150921 5903 Hs.168073_at DKFZP727M231 protein 147492 6615 Hs.24545_at hypothetical protein FLJ11137 146324 6292 Hs.6318_at peroxisomal short-chain alcohol dehydrogenase 145846 5136 Hs.155833_at ESTs, similar to spliceosomal protein SAP 155 H. sapiens Homo sapiens mRNA; cDNA DKFZp434H0820 (from clone DKFZp434H0820); partial cds 140563 5508 Hs.109857_at 138856 7397 Hs.248677_at ESTs, similar to A48018 mucin 7 precursor, salivary – H. sapiens 136390 5253 Hs.4205_at hypothetical protein FLJ20124 ATPase, H+ transporting, lysosomal (vacuolar proton pump), member D 132244 4050 Hs.106876_at 127582 3930 Hs.172914_at retinol dehydrogenase 5 (11-cisand 9-cis) 125310 4560 Hs.23898 at paraneoplastic antigen 124648 5631 The corresponding peripheral blood smears for the four in vivo samples contained only ring forms. Notably, only ring stages are present in the peripheral blood of P. falciparum infected patients; later stages are sequestered in the microvasculature. For this reason, the in vivo whole genome transcription data was compared to the in vitro chromosome 2 ring stage transcriptome. Three samples with parasitemias that were less than 0.3% and total volumes of up to 2.5 ml from Senegal were studied: this resulted in the detection of fewer transcripts than the sample obtained from a Nigerian patient who had parasitemia of 0.4% and underwent a larger blood draw. However, 50% of the top twenty five expressed transcripts in all four samples were shared (data not shown). Further analysis was performed on the Nigerian sample. Only one parasite line was detected in this sample through DNA genotyping of the K1, MAD20, RO33 alleles of msp1 and FC27 and IC1 alleles of msp2 using primers and methods previously reported [ 14 ]. After total RNA was isolated, aliquots of 8 μg of total RNA were labeled using a modified Eberwine procedure [ 2 ]. To maximize parasite transcript detection, 15 μg to 120 μg of cRNA were hybridized to the array and a quantitative expression level was calculated using the MOID algorithm for the Plasmodium genes on the array [ 2 ]. Correlation coefficients comparing P. falciparum chromosome 2 expression levels utilizing 15 μg to 30 μg or 15 μg to 60 μg cRNA were 0.95 and 0.92, respectively. However, as cRNA concentration was increased to 120 μg, the background to noise ratio increased significantly, resulting in a decreased correlation coefficient (R = 0.72) (Figure 2a,2b and 2c ). Figure 2 Scatter plot of expression level variance between samples to define the highest signal to background ratio of calculated expression units for P. falciparum probes. Increasing concentrations of cRNA from the Nigerian sample were hybridized to the array and expression levels (Expression Units) for each transcript were derived using the MOID algorithm. 15 μg cRNA data is presented on the Y axis ( a ) 15 μg v. 30 μg of starting cRNA. ( b ) 15 μg v. 60 μg starting cRNA. ( c ) and 15 μg v. 120 μg starting cRNA. (R = correlation coefficient). The most abundant transcripts detected from the Nigerian in vivo sample are listed in Table 2 . Genes in bold are uniquely expressed in the in vivo sample compared to 3D7 ring stage previously reported using the Kruskal-Wallis method [ 2 ]. Notably, a number of genes encoding surface proteins such as rifins and SERA antigens appear overexpressed in vivo . Table 2 P. falciparum genes expressed in vivo encoded by chromosome 2. In vivo transcripts from the Nigerian sample were defined as present as compared to uninfected blood control hybridization. Asterisk () denotes transcripts that were also detected in a Senegal derived blood sample. Genes in bold are uniquely expressed in vivo and were not found to be expressed in the previously reported 3D7 ring stage transcriptome. Gene locus is from PlasmoDB 4.1 . gene locus description gene locus description membrane proteins cellular function PFB0120w etramps 2 PFB0175c* prt of the MAK16 family PFB0405w* transmission blocking target antigen PFB0205c prt with 5'-3' exonucl. domain PFB0015c rifin PFB0210c monosaccharide transporter PFB0025c rifin PFB0265c* RAD2 endonucl. PFB0035c rifin PFB0295w* adenylosuccinate lyase (OO) PFB1010w* rifin PFB0380c phosphatase (acid phosphatase family) PFB1020w* rifin PFB0390w* ribosome releasing factor (OO, TP) PFB1035w* rifin PFB0410c* phospholipase A2-like a/b fold hydrolase PFB1050w* rifin PFB0445c elF-4A-like DEAD family RNA helicase PFB0330c SERA antigen/papain-like protease PFB0510w GAF domain prt PFB0340c SERA antigen/papain-like protease PFB0525w asparaginyl-tRNA synthetase (OO, TP) PFB0345c* SERA antigen/papain-like protease PFB0585w Leu/Phe-tRNA prt transferase PFB0355c* SERA antigen/papain-like protease PFB0595w* prt with DnaJ domain, DNJ1/SIS1 family PFB0360c SERA antigen/papain-like protease PFB0760w Mtn3/RAG1IP-like prt PFB0095c membrane protein, PfEMP3 PFB0795w ATP synthase alpha chain PFB0975c PfEMP1 fragment PFB0815w* calcium-dept. prt kinase PFB1045w Pf EMP1 fragment PFB0875c chromatin-binding prt (SKI/SNW family) PFB1055c PfEMP1 (var gene) hypothetical proteins PFB0085c* prt with DnaJ domain (RESA-like) PFB0135c hypothetical protein PFB0100c* knob-associated His-rich prt PFB0490c* hypothetical protein PFB0300c merozoite surface antigen MSP-2 PFB0575c hypothetical protein PFB0475c predicted multiple-TM membrane prt PFB0580w hypothetical protein PFB0770c* predicted multiple-TM membrane prt PFB0630c hypothetical protein PFB0125c predicted membrane associated prt PFB0705w hypothetical protein PFB0735c predicted integral membrane protein PFB0745w hypothetical protein PFB0275w* membrane transporter PFB0870w hypothetical protein PFB0465c membrane transporter other PFB0675w predicted secreted protein PFB0990c predicted secreted protein To confirm the accuracy of the results from the oligonucleotide array and to confirm the in vivo overexpression of a SERA antigen (PFB0355), the relative expression of five genes that had varying transcript abundance by array was carried out using real time PCR of cDNA generated from total RNA isolated from the Nigerian in vivo sample and a 3D7 in vitro ring stage sample. There is good correlation between the array results and those obtained by real time PCR (Figure 3 ). To compare abundance of PFB0355 cDNA between the in vivo and in vitro samples, the data is normalized to cDNA of PFB0120 to account for differences in starting parasite cDNA, secondary to human cDNA. The in vivo sample contained 0.15 ng cDNA of PFB0355c and 3D7 ring stage cDNA had 0.09 ng by real time PCR. When PFB0355c is normalized to PFB0120c, it was found to be ten fold overexpressed in vivo as compared to in vitro , consistent with the array results. Figure 3 Correlation of expression levels derived from the oligonucleotide array with ng cDNA determined by real time PCR. Log ng of cDNA from the Nigerian sample and the in vitro 3D7 ring stage sample was determined for five genes that vary from high to low level abundance by array: PFB0120, PFB0100, PFB0270, PFB0355, and PFB0065. cDNA concentration was determined with real time PCR using a standard curve based on 3D7 genomic DNA. These results were correlated to the expression units (EU) found by the array. (R 2 = correlation coefficient) Discussion This data confirms that in vivo whole genomic expression can be performed despite the potential technical challenges of scarce RNA contained in a small blood volume sample and presence of abundant human RNA. Tri-Reagent BD was used to stabilize RNA and the samples were stored at -80°C before transport to the US. The denaturing gel suggests that the RNA remains intact using this reagent. Notably Kyes et al. reported that minus 80°C rather than 4°C or minus 20°C is the optimal temperature to store field sample RNA for detection of long transcripts such as var [ 15 ]. Surprisingly, abundant human RNA was detected in the denaturing gel despite buffy coat depletion of the samples. In addition, the degree of hybridization to human probes on the oligonucleotide array used here was not seen in the previously studied in vitro sample [ 2 ]. Haemoglobin is found to be the most abundantly expressed human transcript (Table 1 ) with considerably higher levels noted in the in vivo samples as compared to the 3D7 in vitro sample. Although human red cells are used for culturing in the in vitro system, reticulocytes which contribute to the haemoglobin detected are not abundant in in vitro samples. This is most likely due to the observation that reticulocyte levels display a 75% loss at 48 hours, when placed at 37°C, which is the condition of in vitro culture [ 16 ]. In addition, other human RNA such as ribosomal RNA is more abundant in the in vivo samples and may be secondary to white cell contamination. Cross hybridization of human RNA to parasite probes may occur. However, Zhou et al have shown that human genomic DNA does not highly cross hybridize to the parasite probes on this custom array [ 17 ]. The high specificity of this array is due to the nature of the highly AT rich parasite genome as compared to the human genome and the careful selection of parasite unique 25 mers probes [ 2 , 18 , 19 ]. Due to this high specificity, it is likely that buffy coat depletion is not necessary for analysis of in vivo parasite transcripts when using these probes. The amount of blood volume necessary for comprehensive detection of transcription depends on level of parasitemia and method of microarray analysis. This analysis utilized the Affymetrix system which requires very little starting RNA as compared to other methods [ 20 ]. Due to the presence of human RNA in the in vivo samples it was not possible to determine how much parasite RNA is required for whole genome analysis. The samples from Senegal were of low parasitemia and small volumes, whereas the 14 ml blood sample with 0.4% parasitemia was sufficient to detect a greater number of chromosome 2 transcripts. Four to five mls of packed blood in a patient with a greater than 2% parasitemia should provide sufficient material for whole genome analysis using these methods. The data demonstrates the reproducibility of the method by independent hybridizations and that maximal senstivity can be achieved with up to 60 μg of cRNA, using this array. PFB0120w is the most abundant in vivo transcript in all samples encoded on chromosome 2. This gene is a member of a recently described gene family, etramps , expressed at early ring stage encoding a protein thought to be involved in erythrocyte remodeling; this was also the most highly expressed transcript in in vitro ring stage cultures [ 2 , 21 , 22 ]. The in vivo expressed genes from the Nigerian sample was compared to that of the in vitro 3D7 ring stage chromosome 2 transcriptome [ 2 ]. As expected, there was a good correlation between the in vivo ring stage transcriptome and the 3D7 ring stage transcriptome. Most of the genes that are expressed in vivo are also expressed in vitro particularly those involved in cellular function and genes encoding hypothetical proteins (Table 2 ) [ 2 ]. A number of differentially expressed genes involved in transmission and antigenic variation were identified. There is high transcription level of transmission blocking target antigen (PFB0405w) in the in vivo samples. Previously this gene has been demonstrated to be expressed and transcribed only in the sexual transmission stage [ 3 , 9 ]. Since no transmissible forms were identified by microscopy this suggests that the in vivo samples may have more parasites that are undergoing or are committed to sexual development than is detectable by microscopy. Several genes encoding membrane proteins, including SERA and rifin genes, were also found to be differentially expressed. These genes are members of multigene family that encode surface proteins which are thought to be involved in immune evasion [ 23 , 24 ]. The observed increase in transcription for these genes in vivo could be due either to genetic or transcriptional modulation of the parasite's defense repertoire or geographic variation. Differences between the in vivo and in vitro expression of SERA antigens may be due to differences in the in vivo stage of development as this gene family is transcribed at later stages in the in vitro life cycle [ 2 , 4 , 25 ]. There was overall higher hybridization intensity of the in vitro samples due to higher parasite counts and a subset of genes were found to be overexpressed in vitro after normalization. This analysis focussed on genes overexpressed in vivo as these results would not be influenced by normalization algorithms. Overall, this data demonstrates that P. falciparum transcripts can be detected from in vivo samples, and that there are potentially important differences between trancription of in vivo samples and that of the 3D7 in vitro trancription profile. In summary, this study provides evidence that whole genome gene expression in P. falciparum can be studied in vivo from a small blood sample of an infected patient. The in vivo sample however contains human RNA, whose quantity may vary from sample to sample and therefore differenes in parasite transcript level between samples must be reported relative to a reference transcript. Despite the abundance of human RNA the genomes are sufficiently different with resultant probes specificity. Predictably, there was a high correlation of in vivo expression with the in vitro ring stage 3D7 transcriptome [ 2 ]. Importantly these data also suggest differences between in vivo and in vitro expression levels in genes typically found in transmissable forms and encoding variant surface proteins. Evaluation of trancription of genes specific for gametocyte development in specific patient populations may uncover the in vivo conditions that favor development of transmissable forms. Similarily, a whole genome analysis can comprehensively characterize expression of multigene families that encode variant surface proteins under in vivo conditions. Further exploration of the in vivo biology of P. falciparum using specific probes to all annotated genes will be undertaken to confirm and explore other important biological differences. This new approach will further the understanding of the host-pathogen interaction and may result in the development of new strategies to combat this disease.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514566.xml
546220	Geographical clustering of prostate cancer grade and stage at diagnosis, before and after adjustment for risk factors	Background Spatial variation in patterns of disease outcomes is often explored with techniques such as cluster detection analysis. In other types of investigations, geographically varying individual or community level characteristics are often used as independent predictors in statistical models which also attempt to explain variation in disease outcomes. However, there is a lack of research which combines geographically referenced exploratory analysis with multilevel models. We used a spatial scan statistic approach, in combination with predicted block group-level disease patterns from multilevel models, to examine geographic variation in prostate cancer grade and stage at diagnosis. Results We examined data from 20928 Maryland men with incident prostate cancer reported to the Maryland Cancer Registry during 1992–1997. Initial cluster detection analyses, prior to adjustment, indicated that there were four statistically significant clusters of high and low rates of each outcome (later stage at diagnosis and higher histologic grade of tumor) for prostate cancer cases in Maryland during 1992–1997. After adjustment for individual case attributes, including age, race, year of diagnosis, patterns of clusters changed for both outcomes. Additional adjustment for Census block group and county-level socioeconomic measures changed the cluster patterns further. Conclusions These findings provide evidence that, in locations where adjustment changed patterns of clusters, the adjustment factors may be contributing causes of the original clusters. In addition, clusters identified after adjusting for individual and area-level predictors indicate area of unexplained variation, and merit further small-area investigations.	Background Ideally, contextual analysis allows for consideration of both attributes that are generalizable across multiple settings, and geographically referenced relationships – influences that occur in context with each other. However, a tension exists between geographic variation analysis, which identifies the location and nature of the variation, and non-spatial analysis, which may identify characteristics of environments or individuals associated with variation, but does so without spatially specific models. Spatial variation in disease characteristics occurs, and multiple statistical methods have been developed to determine whether patterns of variation occur by chance alone, or whether variation is unlikely to have happened at random[ 1 ]. One type of variation analysis is cluster detection analysis, which specifically examines geographic clustering – spatial groups of outcomes that are statistically unlikely to occur by chance alone, given the overall distribution of the outcome of interest across the entire space being examined. Examples might be the occurrence of the disease itself [ 2 , 3 ], or distributions of factors of interest, such as characteristics of the disease, intermediate events such as extent of the disease at time of diagnosis [ 4 ] and receipt of certain treatments [ 5 ], or outcomes such as mortality related to the disease [ 6 ]. However, if clustering of an outcome is identified and determined to occur non-randomly, there is still little information on which to act, because the reasons for these clusters remain hidden. Conversely, conventional non-spatial analysis methods may be used to identify important influences on individual or area-level disease variation. For example, hierarchical or multilevel regression can be used to simultaneously examine individual and area-level characteristics which are associated with variation in disease incidence, characteristics, or outcome [ 7 ]. However, these methods usually consider areas as discrete, even when they are contiguous, without examining relationships between the largest units of analysis. If areas in analyses are geographically related, after building multilevel models, it is still necessary to examine the data for spatial dependence, and to determine whether the model fully accounts for geographic patterns, or whether there is remaining unexplained variation that is spatially dependent – including, but not limited to, geographic clustering. The study of disease patterns in prostate cancer, for example, can be informed by geographic analyses. Prostate cancer is a disease with strong geographic variation, both internationally and also within individual countries or regions [ 8 ]. Like most cancers, the development of prostate cancer typically occurs over a long period of time. Both age of onset and disease course vary enormously, but it has been demonstrated through autopsy study that most men will develop some degree of prostate cell abnormality in older age. It is likely that many factors contribute to its development; from inherited genetic risk, to lifestyle patterns in diet, use of substances such as tobacco and alcohol, exercise and body size and composition, to environmental exposures to a range of protective and detrimental agents [ 9 ]. Furthermore, although much is still unknown about prostate cancer etiology and development, there is sufficient information to argue that prostate cancer is most likely caused by a complex combination of factors, rather than a single explanatory risk. Beyond simple incidence, outcomes such as stage at diagnosis, tumor biology and histologic grade, receipt of standard-of-care treatment, and high quality survivorship are also geographically patterned. When considering the utility of a geographic approach to prostate cancer influences, it may be useful to think of three broad categories of factors. There are factors which may be, at first consideration, purely non-geographic in influence. An example of this might be the influence of the biological characteristics of the cancer on the disease course, such as the relationship between histologic grade of tumor on the stage or extent of disease at diagnosis [ 10 ]. This relationship is considered important and tumor characteristics such as grade are almost always included when modeling outcomes. Yet we can consider this influence to be relatively non-geographic, because we might speculate that this relationship does not change under local geographic influences. Other factors, such as age, might be considered to be pseudo-geographic in influence. The age distribution of the male population would vary across almost any geographic area under consideration, and there is also a strong age-disease relationship in prostate cancer, with the risk of the disease increasing with age. However, the age-disease relationship is not likely to be primarily driven by geography. Adjusting for the distribution of age within a population of interest is often desirable, in order to remove the confounding caused by age, and simulate the geographic variation we would expect to see if we had populations with identical age distributions. A third and more complex category of influences are those for which geographic context is critical to their causal pathway, and thus these variables may be only partially understood outside of their geography. Examples might be individual social or behavioral characteristics such as ethnicity or race, income, insurance or education, occupation, diet or body size. For example, the consistently greater risk for prostate cancer among men of African ancestry compared to all other ethnic groups in the world suggests fundamental biologic causes that supersede geographic influences. However, substantial geographic variation within the US African-American population, as well as international variation between African, Afro-Caribbean, and US men of African ancestry suggests complex multigenerational social and geographic influences [ 11 ]. Even influences that we may confidently classify as so fundamental as to be geographically immutable, such as the relationship between tumor biology and disease progression, could be influenced by geographic variation in access to care or medical practices, dietary, occupational, or environmental agents, or individual variation in behaviors such as tobacco use, exercise, or body size. Therefore, the extent to which any factor's influence on a cancer outcome varies by context or location offers tremendous insight into the mechanisms of influence. The purpose of this research was to combine cluster detection analysis techniques with multilevel modeling of area-level influences on disease patterns, in order to examine the relationship between social-environmental influences and spatial patterning. We used data from the Maryland Cancer Registry on incident cases of prostate cancer occurring in Maryland from 1992 to 1997, and examined variation in two disease characteristics which contribute significantly to overall disease burden: histologic grade of tumor, and stage of disease at time of diagnosis. The use of geographic analysis of prostate cancer outcomes of interest, in combination with modeling of known risk factors, may prove useful in understanding how much of the strong geographic patterns in prostate cancer can be explained by individual and area-level influences, and how much remains, as of yet, unexplained. For each of our two outcomes of interest, higher tumor grade and later stage of disease at diagnosis, we first modeled the "crude" or unadjusted variation in these outcomes across the entire State. This was done by calculating a block group-specific expected rate of each outcome, based simply on the number of cases within the block group and the overall rate of the outcome across the State, and comparing the ratio of observed to expected cases with the given outcome at the block group level. We then used estimates from multivariate models to refine our estimates of the expected number of higher grade or later stage cases, and recalculated, at the blockgroup level, the ratio of observed to expected cases with the outcome of interest. Throughout each set of three analyses, the observed number of cases remained the same, and the expected number (the denominator) varied with each adjustment. Therefore, if an independent variable in a regression model was positively associated with excess risk for the outcome of interest, it increased the regression-estimated expected number of such cases, and thus decreased the observed-to-expected ratio in areas where it was observed. Factors which were negatively associated with risk for the outcome, when adjusted for, reduced the number of such cases expected, and, in turn, increased the observed-to-expected ratio. The methods used are explained in greater detail in the methods section. Results Table 1 describes the overall population of prostate cancer incident cases reported to the Maryland Cancer Registry during 1992–1997, as well as the population used for each analysis. Cases ranged in age from 16 to 106, with a median age of 69. Among cases retained for analysis, 26% were African-American. Overall, in Maryland during the time period 1992–1997, 23% of cases whose record contained histologic grade information had a tumor grade of 3 (poorly differentiated) or 4 (non-differentiated), and 21% staged cases had their disease detected after it had spread outside the prostate gland (stage 2 through 7). Table 1 Characteristics of prostate cancer cases in Maryland, 1992–1997 Registry Population N = 23993 Stage Analysis N = 19223 Grade Analysis N = 18947 Age Group n % n % n % 16–49 403 2 352 2 325 2 50–69 11777 49 10228 53 9868 52 70–79 8739 36 6833 36 6853 36 80–106 3002 13 1810 9 1901 10 Missing 72 1 0 0 0 0 Race/Ethnicity White 16565 69 14255 74 14114 74 Black 5779 24 4968 26 4833 26 Asian 11 1 0 0 0 0 Native American 12 1 0 0 0 0 Other, Not Specified 343 1 0 0 0 0 Missing 1283 5 0 0 0 0 SEER Summary Stage at Diagnosis 0 80 1 0 0 0 0 1 15679 65 15233 79 13798 73 2 2250 9 2190 11 2000 10 3 263 1 255 1 220 1 4 170 1 165 1 152 1 5 150 1 145 1 127 1 7 1274 5 1235 7 945 5 Missing 4127 17 0 0 1705 9 Grade at Diagnosis 1 2505 10 2042 10 2289 12 2 13112 55 11301 59 12335 65 3 4425 18 3786 20 4199 22 4 128 1 113 1 124 1 Missing 3823 16 1981 10 0 0 Figure 1 and table 2 provide information on the four-item county-level social resource index used in the multilevel analysis. The six suburban counties surrounding Washington, D.C. have the highest scores on this index, with Baltimore City and the rural areas of western Maryland and the Eastern Shore of the Chesapeake Bay region having the lowest scores. Both low and high scoring counties contribute substantial numbers of African-American cases to the analysis. Figure 1 Maryland counties, ranked by county resource index score . Maryland's 24 counties ranked from lowest (Baltimore City) to highest (Howard County), based on combined score on four 1990 US Census population characteristics (table 2). Table 2 County resource index score and subcomponents, Maryland 1990 Census County County Resource Index Score 1 Index Subcomponents: 1990 Census # Cases % Cases Who Are Black % High School Graduate 2 % Employed 3 % Moved in last 5 years 4 Median Household Income ($1000) 5 1. Balto. City -1.58 61 91 42 24 3645 61 2. Garrett -1.57 68 93 34 23 109 1 3. Somerset -1.54 61 92 42 23 100 33 4. Allegany -1.48 71 92 36 22 480 1 5. Dorchester -1.41 65 94 38 25 202 33 6. Caroline -1.01 67 96 42 28 161 24 7. Washington -.72 69 96 45 30 517 3 8. Kent -.68 71 97 43 30 104 32 9. Worcester -.57 71 95 50 28 247 21 10. Wicomico -.47 72 95 51 29 300 25 11. Cecil -.37 72 95 45 36 256 7 12. Talbot -.34 77 98 44 32 286 15 13. Queen Anne's -.07 77 96 45 39 199 18 14. Balto. Co. -.06 78 96 43 39 3890 11 15. St. Mary's .11 77 96 51 37 202 21 16. Carroll .11 79 97 43 42 605 4 17. Frederick .36 80 97 49 41 497 7 18. Harford .38 82 97 49 42 760 9 19. Calvert .45 79 97 47 48 197 20 20. Anne Arundel .52 81 97 49 45 1671 13 21. Prince Geo .55 83 96 51 43 2457 52 22. Charles .55 81 97 49 46 350 33 23. Montgomery 1.41 91 97 53 54 3077 11 24. Howard 1.59 91 98 57 54 618 19 1. County resource index scores were calculated by summing the raw score of four measures (percent high school graduates, percent employed, percent moved in last 5 years, and median household income), subtracting the mean of the raw composite scores, and dividing by the standard deviation of the raw composite scores. 2. Percent of persons 25 years or older who have received a high school diploma or its equivalent (e.g. GED) or higher (e.g. college or professional school). 3. Percent of persons 16 years old and over in the labor force who are currently employed. 4. Percent of residents age 5 and older who were not living in the same dwelling five years ago.5. Summed incomes (in thousands of dollars) of household members 15 years of age and older. Cluster detection results – higher grade of tumor Figures 2 , 3 , and 4 , and the related table 3 , show the block group-level patterns of tumor histology across the State. Of the 3670 Maryland 1990 Census block groups, 3313 (90%) contained cases used in this analysis; the number of cases per block group ranged from 1 to 99 with a median of 4. Figure 2 Observed vs. expected block group rates of high grade tumors, and significant clusters . Proportion of prostate cancer cases with histologic grade of 3 or 4, compared to proportion expected based on overall Maryland rate, Maryland Cancer Registry, 1992–1997, N = 18949. A spatial scan statistic was used to identify non-overlapping clusters of statistically significant high or low rates (table 3). Figure 3 Observed vs. expected block group rates of high grade tumors, adjusted for case characteristics, and significant clusters . Proportion of prostate cancer cases with histologic grade of 3 or 4, compared to proportion expected based on case characteristics of age, race and year of diagnosis. Figure 4 Observed vs. expected block group rates of high grade tumors, adjusted for case and area-level characteristics, and significant clusters . Proportion of prostate cancer cases with histologic grade of 3 or 4, compared to proportion expected based on case characteristics of age, race and year of diagnosis, and area-level Census characteristics. Table 3 Cluster Analysis of Higher Grade* Prostate Cancer Cases – Maryland Cancer Registry, 1992–1997 Radius (km) # Block groups in Cluster # Higher Grade Cases Expected # Higher Grade Cases Observed Relative Risk P Value Map 2. Unadjusted Analysis Cluster 1 5.99 550 522.5 669 1.28 .001 Cluster 2 44.93 201 305.3 210 0.69 .001 Cluster 3 10.34 173 253.9 176 0.69 .004 Cluster 4 5.93 38 93.8 49 0.52 .017 Map 3. Adjusted Analysis Cluster 1 14.81 292 487.3 362 0.74 .001 Cluster 2 54.65 162 247.6 164 0.66 .003 Cluster 3 8.27 80 99.6 156 1.56 .013 Map 4. Adjusted Analysis * Cluster 1 6.02 554 444.0 643 1.45 .001 Cluster 2 48.62 1181 1825.4 1587 0.87 .001 Cluster 3 30.88 99 155.8 94 0.60 .004 * Among cases with a histologic grade, those cases graded as 3 or 4 vs. 1 or 2. Expected Rate Adjusted for Race, Age, Year of Diagnosis. * Expected Rate Adjusted for Age, Race, Year of Diagnosis, and Area-Level Census Characteristics. Figure 2 shows that most block groups vary from the expected proportion of high grade cases (23%); block groups with lower proportions of high grade cases are displayed in blue, and those with greater than expected rates of high grade cases are shown in red. (Block groups contributing no cases to the analysis are identified in white on the maps.) Much of this variation is random, and not statistically different than we would expect by chance. Furthermore, because a block group's spatial size is inversely proportional to population density, large individual block group areas of deep color, although striking to the eye, are unlikely to include a substantial proportion of the case population, and therefore would not constitute a statistically significant area of variation on their own. However, figure 2 identifies four non-overlapping clusters with statistically significant (p < .05) higher or lower rates of aggressive grade. The most likely cluster is a geographically small densely populated area in Baltimore City, with a relative risk (RR) of 1.28 (p = .001). The second most likely non-overlapping cluster is a large area in the center of the Eastern Shore of the Chesapeake Bay, with a significantly lower rate of high grade tumors in men with prostate cancer (RR = 0.69, p = .001). Two small areas of lower rates in the suburban areas outside of Baltimore City were identified, one to the north of the City (RR = 0.69, p = .004) and one to the southwest (RR = 0.52, p = .017). Figure 3 shows that adjustment for individual case characteristics (older age, black race, and earlier year of diagnosis) changes the number and location of statistically significant clusters of high and low rates of aggressive grade. The most likely cluster is an area of lower risk for aggressive grade located between Baltimore City and Washington DC (RR = 0.74, p = .001); this area overlaps with the area contained in cluster 3 in figure 2 . Similarly, a large area of the Eastern Shore is again identified as the second most likely cluster with a lower relative risk for higher grade tumors (RR = 0.66, p = .003). There are no longer significant non-overlapping clusters in Baltimore City or northwestern Baltimore County. However, a previously non-significant area of excess risk in Anne Arundel County is now identified, based on the number of cases expected from individual case risk characteristics, as having statistically significant excess risk (RR = 1.56, p = .013). This cluster was identified but not reported due to borderline statistical significance (p = .09) in the unadjusted analysis (figure 2 ). Figure 4 shows results of a cluster detection analysis comparing the observed and expected numbers of cases of high grade tumor in each block group, based on individual case characteristics, and also block group and county-level population characteristics (block group median household income, as well as the composite index of county-level high school attainment, employment, income, and residential mobility). Adjusting for these area-level social influences changes both the number and location of block groups found to have higher or lower rates than expected by chance. The most likely cluster in this analysis is an area of higher than expected rates of aggressive tumor among cases, located to the west of Baltimore City (RR = 1.45, p = .001). This small area was previously identified as the most likely cluster in figure 2 , but with a lower relative risk, and was not identified as having higher rates than expected in the analysis adjusting for individual characteristics (figure 3 ). The second most likely cluster in this analysis is a large area of lower than expected rates (RR = 0.87, p = .001), located in several counties to the north and west of Washington DC. This area includes small clusters 3 from figure 2 and cluster 1 from figure 3 , but the majority of block groups in this cluster were not previously included in the clusters found in the previous analyses. The third most likely cluster in this analysis is located on the Eastern Shore, and although it includes areas identified in the two previous clusters detected on the Eastern Shore, it is both smaller in area and has lower estimate of relative risk for aggressive disease among cases in this area (RR = 0.60, p = .004). Cluster detection results – stage at diagnosis Figures 5 , 6 , and 7 , and the related table 4 , display results of the cluster detection analysis for later stage diagnosis. Cases were located in 90% (3313/3670) of Maryland's 1990 Census block groups; cases per block group ranged from 1 to 90 with a median of 4. Figure 5 Observed vs. expected block group rates of later stage at diagnosis, and significant clusters . Proportion of prostate cancer cases with stage of disease at diagnosis of 2 to 7, compared to proportion expected based on overall Maryland rate, Maryland Cancer Registry, 1992–1997, N = 19223. A spatial scan statistic was used to identify non-overlapping clusters of statistically significant high or low rates (table 4). Figure 6 Observed vs. expected block group rates of later stage at diagnosis, adjusted for case characteristics, and significant clusters . Proportion of prostate cancer cases with stage of disease at diagnosis of 2 to 7, compared to proportion expected based on case characteristics of age, race, tumor grade, and year of diagnosis. Figure 7 Observed vs. expected block group rates of later stage at diagnosis, adjusted for case and area-level characteristics, and significant clusters . Proportion of prostate cancer cases with stage of disease at diagnosis of 2 to 7, compared to proportion expected based case characteristics of age, race, tumor grade, and year of diagnosis, and area-level Census characteristics. Table 4 Cluster Analysis of Later Stage* Prostate Cancer Cases – Maryland Cancer Registry, 1992–1997 Radius (km) # Block groups in Cluster # Later Stage Cases Expected # Later Stage Cases Observed Relative Risk P Value Map 5. Unadjusted Analysis Cluster 1 85.32 1436 1481.0 1743 1.18 .001 Cluster 2 20.72 88 93.8 182 1.94 .001 Cluster 3 41.92 291 512.7 366 0.71 .001 Cluster 4 10.61 316 455.8 325 0.71 .001 Map 6. Adjusted Analysis Cluster 1 24.71 96 98.3 191 1.94 .001 Cluster 2 69.96 326 372.5 533 1.43 .001 Cluster 3 39.82 1208 1633.4 1398 0.86 .001 Cluster 4 4.12 286 248.4 329 1.32 .029 Map 7. Adjusted Analysis * Cluster 1 20.65 95 104.9 188 1.79 .001 Cluster 2 69.96 326 394.8 533 1.35 .001 Cluster 3 47.90 676 1014.0 831 0.82 .001 * Among cases receiving staging, those cases diagnosed at Stages 2–7 vs. Stage 1. Expected Rate Adjusted for Race, Age, Grade, and Year of Diagnosis. * Expected Rate Adjusted for Age, Race, Grade, Year of Diagnosis, and Area-Level Census Characteristics. Statistically significant clusters of high or low rates were identified in four geographic areas in the unadjusted analysis (figure 5 ). As described in detail in table 4 , the most likely cluster is the largest, covering most of the Eastern Shore and some of the adjacent Western Shore of the Chesapeake Bay region of Maryland, with cases in this area have a modestly elevated relative risk of later stage diagnosis (RR) = 1.12, p = .001). A smaller geographic area in the western area of the State was identified as the second most likely cluster, with a relative risk of 1.94 (p = .001). Two relatively affluent areas of the State were identified has having lower probability of later stage diagnosis: Montgomery County, a suburb of Washington D.C. (RR = 0.71, p = .001), and the suburban and rural areas to the north and west of Baltimore City (RR = 0.71, p = .001). Figure 6 shows that, after adjusting for individual case attributes associated with late stage (black race, younger age, aggressive or missing tumor grade, and earlier year of diagnosis), the relationship between the observed number of later stage cases and the expected number changes in many block groups across the State. Although the visual pattern remains similar, the location and size of statistically significant clusters, as well as the relative risk of late stage diagnosis within those clusters, changes. The most likely cluster is now in western Maryland, with a relative risk which is essentially unchanged by adjustment for individual case characteristics (RR = 1.94, p = .001). The largest cluster has now been reduced in size and includes primarily the lower Eastern Shore, but the estimate of relative risk for later stage diagnosis in this area has increased (RR = 1.43, p = .001). The area of lower risk for cases in suburban Washington DC has grown to include much of the suburban area between Washington and Baltimore (RR = 0.86, p = .001), and a new area, centered in Baltimore City, has been identified as having greater risk for later stage diagnosis (RR = 1.32, p = .029). Figure 7 displays results of a cluster detection analysis for later stage diagnosis, comparing actual counts to those expected when considering both individual men's age, race, year of diagnosis, and tumor biology, as well as their immediate neighborhood and county level of social resources – including occupation, education, employment, poverty and residential mobility. These additional adjustments change both the visual patterning of higher and lower rates, as well the location and estimates of relative risk for the statistically significant clusters identified. Two clusters of higher than expected rates of later stage diagnosis remain, covering essentially the same areas as in figure 6 . The relative risk for later stage diagnosis in western Maryland has been reduced only slightly, from 1.94 to 1.79 (p = .001), and the relative risk for the secondary cluster on the Eastern Shore has been reduced from 1.43 to 1.35 (p = .001). Both the Baltimore City cluster and the suburban Washington DC clusters seen in the first two maps are no longer identified as statistically significant. However, a large area in the north central part of the State has been identified as having lower than expected rates of later stage diagnosis, with a relative risk of 0.82 (p = .001). Discussion These geographic analyses provide information on both biological influences on cancer, as well as those more closely influenced by patterns of medical care. For tumor biology, the results of the unadjusted analysis suggest that one primarily rural area of the State, as well as two affluent suburban areas, appear to offer protection from high grade tumor histology. On the other hand, the urban Baltimore area has higher than expected rates of high grade tumors among men diagnosed in the time period 1992–1997. Individual case characteristics change this picture dramatically, but do not "explain away" all variation in this important disease characteristic. For example, black race is an important risk factor for aggressive tumor biology; therefore, it is reasonable to speculate that area differences in the proportion of African-American men in the case population may have accounted for some of the clustering in figure 2 , with clusters in primarily white northern Baltimore County and primarily black Baltimore City no longer statistically significant with race adjustment. Figure 3 shows that, despite individual case differences accounted for with adjustment, there are still three areas of the State with unusually high or low rates of aggressive disease. Figure 4 shows some impact of further adjustment for social resource composition within small areas (block groups) and larger areas (counties). The interpretation of this adjustment is more speculative than confirmatory, but suggests some avenues for further research. Large areas of the rural Eastern Shore of Maryland are no longer identified as being contained inside non-overlapping areas of statistically significant lower risk for aggressive tumor biology, with the protected area being narrowed from a radius of 54.65 kilometers to 30.88 kilometers. Conversely, the small protective area in affluent Howard County between Washington and Baltimore has now grown from a radius of 14.81 kilometers in figure 3 to 48.62 kilometers in figure 4 . Anne Arundel County is no longer at excess risk but Baltimore City is. The influence of area level social resources on high grade of tumor was complex: the men with the lowest risk for aggressive tumors were white men living in small areas of greater income, nested within counties of overall low social resources. Therefore, clusters remaining in figure 4 are those whose rates are either higher or lower than expected given their social characteristics. The cluster in Baltimore City reflects the fact that Baltimore City does not fit the overall model of low resource counties as protective. Baltimore City is the single urban county in the lowest range of the index; the rest of the lower resource counties are predominantly rural. Therefore, moving from figure 3 (only individual adjustment) to figure 4 (area-level adjustment) identified that Baltimore City's low social resources are not protective, to same effect as in rural counties. This difference may be caused by any number of lifestyle differences between urban and rural low income communities. Although individual case race is not likely to be driving this difference, it may be that area-level racial composition is another piece of this puzzle, given that Baltimore City's racial composition differs so dramatically from the other low resource counties. Conversely, the protective clusters are found in counties with high social resource index scores, centered in Montgomery County in the Washington, D. C. suburbs, and in an area with slightly low scores, Talbot and Queen Anne's counties on the Eastern Shore. For the D.C. suburbs, their rate in figure 2 is neither high nor low; however, their high social resource index score would predict high rates; therefore they create a lower- than-expected cluster. For the Eastern Shore, the lower rate of aggressive disease has been consistent across all three cluster analyses. For low social resource counties such as Dorchester, the adjusted predicted rate in figure 4 is now consistent with expected low rates, and therefore this area is no longer part of a cluster. However, the rate is lower than expected in counties with slightly higher resources, and therefore the most affluent Eastern Shore counties (Talbot and Queen Anne's) continue to be identified as lower than expected. Finally, Anne Arundel County, which had higher than expected rates in the individually adjusted analysis, is now no different than expected, arguing that the relatively high social resource index score for this county led to a closer approximation of expected proportion of cases with aggressive disease. When considering the geographic patterning of later stage at diagnosis for men with prostate cancer in Maryland during the time period 1992 to 1997, it appears from the unadjusted analysis that men in certain rural areas were much more likely to come into treatment with more advanced disease than those in the suburban, more affluent areas of the State. Individual characteristics of the patients appear in some ways to have masked these geographic differences, in that the clusters generally remain or become more important once the case population mix of characteristics such as age, race, tumor biology, and year of diagnosis is taken into account (figure 5 versus figure 6 ). Additionally, an area of Baltimore City, which has a greater proportion of young, African American men than the rest of the State, became significantly more likely to have later stage cases, after adjustment for age and race. This suggests that men in Baltimore are specifically disadvantaged in terms of early detection of disease, beyond what would be predicted by their age, race, or tumor biology. From figure 7 , we see evidence that area-level socioeconomic resources may contribute to these patterns. The relative risk for late stage diagnosis in the two rural clusters has been reduced somewhat, and the cluster of lower risk in the north-central part of the State is less different than in the unadjusted map. Baltimore City and the Washington suburbs no longer differ significantly from the rest of the State, supporting a socioeconomic influence on the previous clusters. Prostate-specific antigen (PSA) testing was widely available in Maryland during the entire time period of this study (1992–1997). This suggests that more global barriers to health care, rather than differential access to this specific diagnostic tool, were more important in creating these patterns of late stage diagnosis. Conclusions Although there is no statistical test to evaluate the proportion of variation explained in a multilevel model, because of the inclusion of random effects, it is reasonable to state that, overall, our adjustment methods did account for substantial variation in rates of aggressive disease and late stage diagnosis, by considering important influences – characteristics of the men themselves, as well as characteristics of their environments. In spite of this, variation in these cancer characteristics remained substantial across the State. In fact, whether the measure used is the number of cases, number of block groups, or geographic area, figures 4 and 7 identify as much, if not more, deviation from the predicted pattern than the unadjusted figures 2 and 5 respectively. Many areas are included in clusters across all three analyses of a specific outcome. Given that the exact boundaries of clusters are always approximate, and would be expected to vary from analysis to analysis, it is important to note similarities as well as differences within each outcome-specific set of maps. This suggests that there are underlying causal influences that remain, despite the important relationships with the measures used for adjustment. An additional caveat in the interpretation of these cluster detection analyses is related to the choice of criteria used for reporting clusters. In consideration of the large amount of data being examined, both in terms of geographic area and number of cases, we chose to report clusters only if they had a statistical significance of p < .05, and contained no geographic overlap with a more significant cluster. These restrictions meant that a given geographic area could possibly be described as part of a cluster of excess or reduced risk in one analysis, and not in other, based on small changes in expected number of cases, or based on the identification of a more significant cluster nearby. These findings have implications on both a practical cancer control level, as well as for further research in prostate cancer. For state and local health agencies, trends in area-level patterns of cancer outcomes over time can be used to monitor change, whether to evaluate the effectiveness of geographically distributed interventions such as screening or treatment programs, or identify population changes which may increase need for services. Unadjusted cluster analyses provide valuable information for cancer control planners who need to address areas of greatest need, regardless of the cause. However, adjusted analyses identify geographically unique situations, such as the persistently elevated rates of later stage diagnosis in two rural areas of Maryland. For researchers, analytic techniques which identify both explained and unexplained geographic variation may provide information about the multilevel synergistic factors influencing cancer patterns, or, at a minimum, identify areas and populations meriting further study. Methods Data and data sources More detailed information on these data and methods has been reported previously [ 10 ]. With IRB approval from the Johns Hopkins School of Public Health and the Maryland Department of Health and Mental Hygiene, and a data agreement between the two institutions, we obtained all incident cases of prostate cancer reported to the Maryland Cancer Registry during the years 1992–1997 (n = 24,189). Based on case residence address, we geocoded cases to latitude and longitude coordinates. For cases unable to be geocoded, we assigned cases to a coordinate location within their zipcode using a weighted imputation algorithm, based on 1990 US Census race-, age-, and gender-specific population distributions within their zipcode [ 12 ]. We thus assigned each nongeocoded case to a Census block centroid, based on the best-known distribution of men like himself within his zipcode. Of 24,189 cases, 23,993 had verifiable Maryland addresses. Ninety-one percent (21,904) were successfully geocoded, and nine percent (2,089) were assigned to an imputed location within their zipcode by algorithm. An additional 3063 cases were not used, due to missing demographic or clinical data, or because their race was neither African-American or white, leaving a final analysis population of 20,928. We used individual case characteristics from the Registry record, including age at diagnosis, race, year of diagnosis, tumor stage, and tumor histologic grade. Based on case residence, we added to each case record selected 1990 US Census characteristics of three nested geographic units surrounding the case location – the Census block group, Census tract, and county. Our record for each case therefore contained individual demographic and clinical characteristics based on the Cancer Registry data, point location of residence, and Census measures for the case's block group, tract and county of residence. We created seventeen possible area-level social indicators for each block group, tract, and county from the 1990 US Census STF-3 file [ 13 ]: three measures of housing resources (median sale price, percent of owner occupied units, and housing value percentile rank, based on both rental and sale values, weighted by the proportions of rental and owner-occupied housing in an area), three income measures (median household income, median family income, and median per capita income), four population composition measures (percent white, percent black, percent born outside the US, and percent age 5 or older not living in the same residence for at least five years), two social class measures (percent high school graduates among persons age 25 and older, and percent employed in white collar jobs, defined as Census job classifications of managerial, professional, technical, sales, and administrative support) and five material deprivation measures (percent households without a car, percent households without a telephone, percent persons 16 and older who were unemployed, percent persons living in "crowded" residences, defined as more than one person per room, and percent of persons living in poverty). For continuous variables (case age and census measures), we compute standardized measures to reduce collinearity, by centering each case value at the population mean, and dividing by the standard deviation. For year of diagnosis, we centered the values at 1994, a midvalue in our six year time window. We chose two outcomes of interest which are associated with differences in prostate cancer disease severity and longterm survival for patients – histologic grade, or degree of cell differentiation, of the tumor, and stage, or extent, of disease at time of diagnosis. For each outcome, we dichotomized the data and examined the likelihood of the more negative outcome. For tumor grade, we compared tumors staged as 3 or 4 (poorly differentiated or undifferentiated) to those graded as 1 or 2 (well or moderately well differentiated). For stage at diagnosis, we used the Surveillance, Epidemiology and End Stage (SEER) summary stage [ 14 ], and compared cases diagnosed at stages 2 through 7 (regional to distant metastases) to those diagnosed at stage 1 (localized disease). Cases only missing staging information were retained for analyses of grade, and cases only missing grade were used in the analysis of stage at diagnosis. Because tumor histology is an important predictor of rapidly spreading disease, tumor grade was used as an independent predictor in modeling late stage disease. Cluster detection methods In order to explore the geographic patterns of our two outcomes of interest, we used the spatial scan statistic [ 15 ] to detect and evaluate the statistical significance of any geographic clusters of each outcome. This method imposes a very large number of overlapping circles of different location and size on the map, each of which is a potential cluster, and adjusts for the multiple testing inherent in the many circles considered. Our cluster detection method identified clusters of both high and low rates, with a maximum scanning window size to include up to 50% of the population at risk. Secondary clusters were reported if they had no geographic overlap with more likely clusters. P-values were derived from 999 simulated Monte Carlo replications under the null hypothesis of spatial randomness of outcomes of interest. We conducted three separate cluster detection analyses for each of the two prostate cancer outcomes: higher histologic grade of tumor, and later stage at diagnosis. In the unadjusted analysis, under the null hypothesis, the expected number of more aggressive grade or late stage cases in a block group was calculated by multiplying the total case population of the block group by the statewide rate of the outcome of interest. Thus, in the unadjusted analysis, a block group would be expected to have the same rate or proportion of late stage or high grade cases in its case population as the State. In the two adjusted analyses, the expected number of aggressive grade or later stage cases was calculated from a regression model containing individual case characteristics, or from a regression model with both individual and area-level covariates. Based on the expected counts, the number of aggressive grade and later stage cases in each block group was modeled as a Poisson distribution. For the unadjusted analyses, we also used a Bernoulli model to compare the distribution of so-called "cases" (those with aggressive grade or late stage) to "controls" (less aggressive grade or early stage) based on point location of each residential address, rather than rates within block groups. This was useful to compare the sensitivity of the Poisson model assumption for aggregated data to that of the unaggregated Bernoulli method. No major differences in results were found, and to allow proper comparison between the adjusted and unadjusted analyses, the Poisson model results are presented for all three types of analyses. For each cluster identified, we list the radius, number of block groups in the cluster, the observed versus expected number of late stage or aggressive grade cases, the relative risk and the p value. The relative risk is the risk of the respective outcome within the cluster, compared to the population's risk. We report clusters with statistical significance p < .05 that do not overlap with another reported cluster with a lower p-value. Calculations were done using the freely available SaTScan v4.0 software . Sources of expected population counts We used the results of two multivariate modeling methods to calculate the expected count of aggressive grade and late stage cases in each block group. In prior work [ 10 ] we built multivariate hierarchical logistic regression models [ 16 , 17 ] to identify individual and area-level factors significantly associated with aggressive grade and late stage among cases, and these findings, summarized below, served as the basis for calculating our expected population in each block group. In logistic regression models including only individual level predictors, our final model included the following statistically significant associations with higher histologic grade of tumor: older age (Odds Ratio (O.R.) 1.17, 95% Confidence Interval (C.I.) 1.13, 1.21), black race (O.R. 1.46, 95% C.I. 1.35, 1.57), more recent year of diagnosis (O.R. 0.92, 95% C.I. 0.90, 0.94), and an interaction between age and year of diagnosis (O.R. 1.06, 95% C.I. 1.04, 1.08). To build the multilevel models, we tested each of the 17 area-level indicators at each level, starting with block group, and also tested for interactions at each level and between levels. To avoid unstable models, when we found multiple significant Census predictors, we computed and tested simple indices by summing relevant Census measures. We also tested for random effects, to account for additional variability. In a multilevel logistic regression model of aggressive tumor grade, each of the above individual level variables remained significant. In addition, two area-level indicators were significant in the final model: block group median household income (O.R. 0.92, 95% C.I. 0.87, 0.96), with an interaction between black race and income (O.R. 1.12, 95% C.I. 1.02, 1.223), and a standardized county resource index, composed of four summed county-level measures: percent high school graduates, percent employed, percent moved within the past five years, and median household income (in $1000 units) (O.R. 1.23, 95% C.I. 1.16, 1.31). Random intercept terms were found to be significant at the block group and county level. In logistic regression models including only individual level predictors, our final model included the following statistically significant associations with late stage at diagnosis: older age (O.R. 0.85, 95% C.I. 0.82, 0.90), black race (O.R. 2.97, 95% C.I. 1.35, 1.59), higher tumor grade (O.R. 2.97, 95% C.I. 1.35, 1.59), missing tumor grade (O.R. 5.56, 95% C.I. 2.77, 3.17), more recent year of diagnosis (O.R. 0.83, 95% C.I. 0.78, 0.88) and interactions between age and black race (O.R. 1.18, 95% C.I. 1.09, 1.27), grade and year of diagnosis (O.R. 1.06, 95% C.I. 1.02, 1.10), and missing grade and year of diagnosis (O.R. 1.19, 95% C.I. 1.10, 1.30). In the multilevel logistic regression model of late stage at diagnosis, each of the above individual level variables remained significant. In addition, two area-level indicators were significant in the final model: block group percentage of white collar workers among the employed population (O.R. 0.93, 95% C.I. 0.89, 0.98), and the standardized county resource index (O.R. 0.94, 95% C.I. 0.89, 0.98). A statistically significant interaction existed between county resource score and older age (O.R. 0.95, C.I. 0.92, 0.99), and random intercept terms were found to be significant at the block group and county level. Calculation of block group-specific predicted populations The models described above were used to calculate an expected count of aggressive grade and late stage cases, respectively, for each block group. This was accomplished by taking the inverse logit transform of the expected linear predictor in each logistic regression model, yielding a set of estimated probabilities for each outcome. These probabilities were then aggregated to the block group level, providing expected block group-specific counts of later stage and aggressive grade cases. By definition the expected linear predictors includes only estimates from fixed effects in each multilevel logistic regression model. The random effects at the block group and county level, although influential on parameter estimation, were not included in these calculations. Compared to the unadjusted results, geographic patterns shown in the adjusted analyses could be interpreted as those existing after controlling for individual and area-level factors, respectively. In essence, this approach explores residual geographic variation. For this reason, information from the random effects is not included in determining expected outcome counts. Authors' contributions AK obtained funding and data for this research, and was PI on the project. She conceptualized and conducted the analysis, and drafted the manuscript. MK directed the development and interpretation of the spatial analysis. FC worked with AK to develop and interpret the multilevel models used, as well as the methodology for imputation. All authors participated in the preparation and approval of the final version of the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546220.xml
545067	Tetanus immunity in nursing home residents of Bolu, Turkey	Background Tetanus is a serious but vaccine-preventable disease and fatality rate of the disease is high in the neonates and the elderly. The aim of this study was to detect the tetanus antibody prevalence in the over sixty-year age residents of the nursing homes in Bolu. Methods A voluntary-based study was done in the residents of two nursing homes in Bolu, Turkey. Blood samples were taken from 71 volunteers residing in there nursing homes. Tetanus IgG antibodies were measured by a commercial ELISA kit. Results Among overall subjects, only 11 (15.7 %) had the protective tetanus antibody titers at the time of the study. Totally, 10 subjects were examined in emergency rooms due to trauma or accidents within the last ten years and, four (40%) of them had protective antibody levels. Of the remaining 61 subjects only 7 (11%) had protective antibody levels (p < 0.05) [Relative Risk = 3.49, 95% Confidence Interval 1.24–9.77]. Conclusions Tetanus antibody level is below the protective level in the majority of the over-sixty-year-age subjects residing in the nursing homes. Each over sixty-year age person in our country should be vaccinated. Until this is accomplished, at least, nursing home residents should be vaccinated during registration.	Background Tetanus is an acute disease caused by the tetanus toxin, released by the bacterium Clostridium tetani . Tetanus spores are present in soil and manure and may be introduced into the body through a puncture wound, burn or scratch. Tetanus can never be eradicated because the spores are widely spread in the environment, and found in the intestinal flora of domestic animals, horses, chickens, and humans. Tetanus is not spread from person to person [ 1 ]. Tetanus is preventable by proper immunization. However, the disease is still prevalent even in the most developed countries and it often emerges in the elderly [ 2 ]. The prevalence of neonatal tetanus is considered a criterion for developmental level of the country. Every year, the cases of neonatal tetanus are reported in four to six European countries; Turkey and Albania have the highest occurrence rates [ 3 ]. The mortality rate of tetanus is high in the neonates and the elderly patients [ 4 ]. Age-related mortality rates were found to be 79.4 % in the neonatal period [ 5 ], 11 % in adults younger than 50 years old and 54 % in adults older than 50 years old [ 4 , 6 ]. Totally, 280 cases were observed in Turkey between 1994 and 1995. The mortality rate was 29.8% for these cases [ 7 ]. Risk is greater in people over 60, particularly in the developed countries [ 8 ]. In the USA, the disease occurs at a six-times higher rate in this age group [ 8 , 9 ]. Of 99 tetanus patients whose complete information was reported to the Centers for Disease Control and Prevention (CDC) during 1987 and 1988, 68% were over 50, while only six were younger than 20. No cases of neonatal tetanus were reported. The disease continues to occur almost exclusively among people who are unvaccinated or inadequately vaccinated or whose vaccination histories are unknown or uncertain [ 10 ]. Bolu is a city in the Western Black Sea Region of Turkey and has a population of approximately 283.000. There are two nursing homes (Izzet Baysal and Neziha Baysal Nursing Homes) in the each with a boarding capacity of 60 people. Tetanus antibody level in such a specific group has not been researched in the country yet. The purpose of this study was to detect tetanus antibody prevalence in the voluntary residents of the nursing homes over 60 years old. Methods This study was performed May 1–30, 2004. Subjects were selected from Izzet Baysal and Neziha Baysal Nursing home residents who volunteered for antibody measurement. A standard form was filled including subjects' clinical information, demographic traits along with vaccination and injury stories. 10 cc venous blood was taken from each subject and kept at -20°C until the day of study. Tetanus Ig G antibodies were measured by a commercial ELISA kit [Novatec Dietzenbach, Germany] at a 450/620 nm wavelength. The results were evaluated in the way previously defined by Schoroder et al. [ 11 ]. Briefly, antitoxin levels below 0.1 IU/L were defined as "below protective level" and antitoxin levels above 0.1 IU/L were defined as "at protective level". Students't-test was used to compare quantitative variables while chi-square and Fischer's exact tests (two-tailed) were used for qualitative data, p < 0.05 was considered to be significant. Statistical analyses were performed with Epi-info 6.0 (Center for Disease Control, Atlanta, USA). Results Subjects We aimed to include all residents of the nursing homes. In the study period, 78 subjects were staying in the two nursing homes (40 residents in Izzet Baysal Nursing Home and 38 residents in Neziha Baysal Nursing Home). However, 7 subjects did not want to participate in the study. Therefore, these seven subjects could not be included. Antibody results and properties of subjects A total of 71 voluntary subjects [54 (76%) male and 17 (24%) female] were included in the study. The mean age was 71. Only 11 (15.4%) of the 71 subjects had a protective antibody level against tetanus. 15 subjects were aged between 60–65 (3 of them had protective antibodies), 22 subjects were aged between 66–75 (2 of them had protective antibodies) and 34 of 71 subjects were aged ≥ 76 years (6 of them had protective antibodies). Protective antibody level decreased with every age group (60–65, 66–75, and ≥ 76) above 60 years old, but there were no statistically significant difference between the five-year age periods above 60 (p > 0.05) (Fig. 1 ). Figure 1 Antibody levels of subjects Totally, 10 subjects were examined in emergency rooms due to trauma or accidents within the last ten years. While four (40%) of them had protective antibody levels, only 7 (11%) of 61 subjects who didn't get examined in the emergency room had protective antibody levels (p < 0.05) [Relative Risk = 3.49, 95% Confidence Interval 1.24–9.77]. Only 10 (14%) of the 71 subjects had been vaccinated in the last ten years. Five of them had been vaccinated as a result of emergency room visits. Antibody level and demographic properties of subjects were presented in the Table 1 . Table 1 Demographic characteristics and antibody levels of subjects according to age groups Characteristics TOTAL n = 71 (%) Age (mean ± SD) 71.1 ± 8.6 Female/Male 17/54 Diabetes mellitus 7 (9.8) Emergency room visits due to injury within the last 10 years 10 (14) Tetanus vaccine within last 10 years Emergency room visit/No emergency room visit 5 (7)/5 (7) Occupation with soil / garden-work 20 (28.1) Average tetanus Ig G level 0.080 Tetanus Ig G >0.1 IU/ml 11 (15.4) Discussion According to the records of the Ministry of Health, 2039 tetanus cases were detected in Turkey between 1980–2002, 462 (23%) of which died [ 12 ]. In our country, a vaccination campaign against tetanus was initiated in the mid 1960s. However, it was carried out irregularly until 1985. After 1985, 3 doses of tetanus toxoid were given to all neonates after birth and a booster dose was applied in the 16th month followed by vaccination of primary school children at aged 7 and 12. While 67 neonatal tetanus cases were detected in 1990 in a population of 57.582.244, 32 neonatal tetanus cases were detected in 2002 in a population of 70.415.244 [ 12 ]. The neonatal tetanus rate has decreased with an increased rate vaccination in the country. In Turkey, females are vaccinated during pregnancy and males are vaccinated during military service but there is no vaccination program for the elderly. Unfortunately, a falling could not be achieved in the prevalence of tetanus in the elderly. Similar results were obtained in the developed countries, too. For instance, it was established that, rate of occurrence of tetanus in the people older than 60 was ten times higher than the young in Italy [ 9 ]. Routine vaccination programs focus on the childhood period in the developing countries. However, according to researches, the majority of adult population is sensitive to tetanus [ 13 ]. In our study group, only 11 (15.4%) of 71 subjects had protective antibody levels while the rest were under the risk of tetanus. This resultsuggested that lack of protective immunity among nursing home residents was a significant public health concern. In industrialized countries tetanus has become a rare disease and an infrequent cause of death, mainly due to the implementation of comprehensive immunization programs. But, tetanus is still an important problem for the developing countries, due to poor immunization standards and inadequate hygiene [ 14 ]. Deaths due to tetanus in Turkey mostly occur in the elderly. The 27.4 % of the total deaths caused by tetanus in 1989 occurred in the patients over 40. But today mortality rate in this age group ranges between 48.8–60.6% [ 15 ]. Those who are not vaccinated and the elderly are at risk [ 16 ]. Individuals over 60 are usually retired people and spending most of their time in garden or land-work. Tetanus spores may be introduced into the body through a puncture wound, scratch during these works. So they face risk of getting tetanus. Also, 28 % of our subjects still were busy with soil and garden work. According to our results, antibody levels in individuals who had been examined in an emergency room within the last ten years due to injury are significantly higher than who didn't have such a history (p < 0.05). This situation suggests that vaccination programs are not well established in countries like Turkey, thus injuries and accidents expose people over 60 to a great risk of getting the tetanus prophylaxis [ 17 ]. Once the existence of tetanus is suspected, intensive, and effective management is essential. The patient should receive intensive care aimed at prevention of muscle spasms, prevention of respiratory tract and metabolic complications, and neutralization of circulating toxins [ 18 ]. Assuming that one tetanus case stays in the intensive care unit for at least fifteen days, it costs $9.000 per case approximately. In Turkey, 30.000 people can be vaccinated with a booster dose of tetanus vaccine for this amount of money [ 19 , 20 ]. Moreover, vaccination will not only ensure economic benefits but also protect thousands of people against tetanus. Apart from being more profitable, such policy will improve the health statistics of the country [ 20 ]. It is imperative that every person over 60 should be vaccinated against tetanus. Conclusions The lack of protective immunity against tetanus among the nursing home residents is a significant public health concern. A vaccination program including every individual over 60 should be charted immediately. After the primary series of 3 doses, protection against tetanus should be sustained by scheduling booster doses routinely in every 10 years [ 21 ]. Until this campaign is accomplished, at least, nursing home residents should be vaccinated during registration. Competing interests The author(s) declare that they have no competing interests. Authors' contributions All authors read and approved the final manuscript. OK designed the study and drafted the manuscript. AT, KA, analysed the data. OK oversaw the microbiological research. OK and FO interpreted the results of the analysis and critically reviewed the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545067.xml
449891	On the Brink: How Biology and Humans Affect Extinction Risk	null	Close to a quarter of the world's mammals are at high risk of extinction. Save for the periodic “great extinctions,” mammalian extinction has been a relatively rare event in geological terms, with one species disappearing from the fossil record every 1,000 years or so. Over the past 400 years, species have been disappearing 50 times faster than this “background” rate, with one vanishing every sixteen years. Human population growth and all its consequences—habitat destruction, propagation of invasive species, poaching—are largely to blame. Top predators often suffer heavily from encounters with humans, especially when those predators are perceived as economic threats. Thirty-four Mexican gray wolves have been reintroduced in Arizona since 1998, and five have been shot, reportedly by ranchers. The IUCN lists lions as vulnerable. Photo, with permission, by Nicky Jenner, Institute of Zoology, London Species in the most densely populated areas are expected to face the greatest risk, yet some survive while others perish, suggesting biological factors play a role in their fate. If, for example, the same external force drastically reduces populations of species with different biological profiles, then a species with a relatively short gestation period may stand a better chance of recovering than a long-gestating species. Effective conservation strategies depend on understanding which factors are likely to increase extinction risk, but it's unclear how important intrinsic biological traits are relative to external pressures from humans and whether biology's influence on survival depends on the intensity of the threat. Ecologists often use human population density as a proxy for anthropogenic threats such as habitat destruction and hunting. To tease out the relative importance of all these factors, Marcel Cardillo et al. analyzed the impact of various biological traits and human population density on extinction risk in the mammal order Carnivora, which includes the red panda, lion, and members of the photogenic weasel-like viverrid family. By identifying the most salient factors in predicting extinction, the authors have created a model to identify those species at greatest risk. The biology of a species combined with human population density, the researchers found, is a stronger predictor of risk than exposure to humans alone; those biological traits that increase risk vary depending on a species' exposure to human populations. Carnivores with low exposure to humans, for example, are likely to be at greater risk if their population density is low and they have small ranges, possibly because this makes them more vulnerable to loss of habitat. Species living near densely populated human areas must often contend with hunting and other direct threats on top of habitat loss and are more at risk if they also have long gestation periods—they can't repopulate fast enough to offset the additional pressures. Based on projected human population growth, this model predicts the addition of a number of species—mostly from Africa, where population growth rates largely exceed the global average—to the endangered list by the year 2030. Most of these species—including African viverrids such as the common genet, which not only lives in areas where human populations are rapidly expanding but is also biologically predisposed to decline—are currently considered a low conservation priority. While it's possible that the direct effects of human population density are past—that is, species most sensitive to human incursions are already gone—human population density likely modulates biology. That might explain why gestation length didn't predict risk for species living in sparsely populated areas—all else being equal, their numbers remained relatively stable. A species with a small population forebodes a high extinction risk regardless of human density, though species with long gestation periods, again, face greater danger in the company of humans. Altogether, these results suggest that as human population pressures increase, the importance of species-specific biology in predicting extinction risk also increases, with biology affecting which species are most vulnerable to external threats. With most conservation efforts focused on damage control, these findings make the case for interceding before a species reaches the brink of extinction. “There is no room for complacency about the security of species,” the authors warn, “simply because they are not currently considered threatened.”	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC449891.xml
400249	Noise Minimization in Eukaryotic Gene Expression	All organisms have elaborate mechanisms to control rates of protein production. However, protein production is also subject to stochastic fluctuations, or “noise.” Several recent studies in Saccharomyces cerevisiae and Escherichia coli have investigated the relationship between transcription and translation rates and stochastic fluctuations in protein levels, or more generally, how such randomness is a function of intrinsic and extrinsic factors. However, the fundamental question of whether stochasticity in protein expression is generally biologically relevant has not been addressed, and it remains unknown whether random noise in the protein production rate of most genes significantly affects the fitness of any organism. We propose that organisms should be particularly sensitive to variation in the protein levels of two classes of genes: genes whose deletion is lethal to the organism and genes that encode subunits of multiprotein complexes. Using an experimentally verified model of stochastic gene expression in S. cerevisiae , we estimate the noise in protein production for nearly every yeast gene, and confirm our prediction that the production of essential and complex-forming proteins involves lower levels of noise than does the production of most other genes. Our results support the hypothesis that noise in gene expression is a biologically important variable, is generally detrimental to organismal fitness, and is subject to natural selection.	Introduction Stochasticity is a ubiquitous characteristic of life. Such apparent randomness, or “noise,” can be observed in a wide range of organisms, resulting in phenomena ranging from progressive loss of cell-cycle synchronization in an initially synchronized population of microbes to the pattern of hair coloration in female calico cats. An important source of stochasticity in biological systems is the random noise of transcription and translation, which can result in very different rates of synthesis of a specific protein in genetically identical cells in essentially identical environments ( Elowitz et al. 2002 ; Ozbudak et al. 2002 ; Blake et al. 2003 ). Understanding how stochasticity contributes to cellular phenotypes is important to developing a more complete picture of how cells work. Accordingly, noise in gene expression and other cellular processes has been a major focus of research for more than a decade. While several cases have been described where stochasticity is advantageous (e.g., phase variation in bacteria [ Hallet 2001 ] and the lysis/lysogeny decision in phage lambda [ Arkin et al. 1998 ]), it is expected that noise is not advantageous in most cellular processes, as precisely controlled levels of gene expression are presumably optimal (c.f. Barkai and Leibler 2000 ). However, whether noise in expression is of consequence to organismal fitness has not previously been investigated, despite the centrality of this question to our understanding of the role of noise in biological systems. In this study, we investigate whether the differences in noise levels among genes are consistent with the hypothesis that noise in gene expression has been subject to natural selection to reduce its deleterious effects. We propose that random fluctuations in the expression levels of two groups of genes in yeast, essential genes and genes encoding protein complex subunits, should be particularly consequential for organismal fitness. If noise in gene expression is not an important factor to yeast—i.e., if the level of stochasticity experienced by yeast in gene expression is below that which would have negative consequences—then we would expect to see no difference in the randomness of expression in genes for which noisy expression is predicted to be relatively more or less deleterious. However, if stochasticity is an important variable on which natural selection has acted, we would expect to see the strongest signature of such selection in the expression of genes for which yeast are the most sensitive to randomness. Results If deletion of a gene has only a small deleterious effect on the fitness of yeast, then random fluctuations in the amount of protein produced from that gene are likely to have a similarly small, or even smaller, impact. In contrast, the same fluctuations in the level of a protein essential for viability may have a profound effect on fitness; in the extreme, fluctuation to levels below that required for normal cellular function could compromise viability. Considering this predicted difference in the sensitivity of yeast to randomness in expression of essential versus relatively dispensable genes, we reasoned that if noise in gene expression is a biologically important variable, selection for reduction of stochasticity in expression levels would likely be stronger for essential genes than for nonessential ones. A recent study linking noise in protein levels to transcription and translation rates in yeast ( Blake et al. 2003 ) allows us to test this prediction. In the study, noise in the expression of a green fluorescent protein (GFP) reporter gene was measured by flow cytometry; stochasticity was measured as the amount of variation in GFP levels per cell in a population. Thus if all cells in a population had very similar levels of GFP, there was little noise in the production of the GFP. The effect of transcription and translation on noise levels was studied by independently varying these two parameters and measuring the resulting noise levels for a population of cells. This experimental approach, as well as a mathematical model of protein production ( Blake et al. 2003 ), indicates that noise in protein production is maximized at intermediate levels of transcription (at approximately one-third of the maximal transcription rate of a gene, regardless of what that maximum is; see Materials and Methods ), as well as at maximal levels of translation per mRNA molecule. To produce a given amount of any particular protein, yeast could adopt one of three qualitatively different strategies ( Thattai and van Oudenaarden 2000 ) ( Figure 1 ): (1) maximize transcription and minimize translation per mRNA, (2) maximize translation per mRNA and minimize transcription, or (3) employ intermediate levels of both transcription and translation per mRNA. Importantly, strategy 1 should result in less stochasticity than strategy 2 or 3. Strategy 2 is noisy due to the high translation, and strategy 3 is noisy due to both intermediate transcription and translation (the data currently available do not allow us to predict whether expression strategy 2 is more or less noisy than strategy 3). In contrast, noise is minimized at both transcription and translation steps for genes that exhibit strategy 1. Thus we predicted that if noise in protein production is an important factor in yeast, then genes that are essential for viability would be biased towards having high transcription rates and a low number of translations per mRNA. Figure 1 Strategies for Expression Three different strategies for achieving a given rate of protein production (four proteins will be produced in each case) and the amount of noise in expected to result from each strategy. Curved lines represent mRNA molecules, with ribosomes translating them; a larger number of mRNA molecules represents higher transcription, and a larger number of ribosomes per mRNA represents higher translation per mRNA. To test this prediction, we estimated protein production rates (proteins/s; see Materials and Methods ) for all yeast genes and asked whether essential genes tended to adopt strategy 1 more often than nonessential genes with similar protein production rates. It was critical to control for overall rates of protein production, as there is an overall correlation between a gene's dispensability (defined as the growth defect of a yeast strain missing that gene in rich glucose medium, i.e., an essential gene is indispensable) and its rate of protein production ( Figure S1 ). This correlation between dispensability and the rate of protein synthesis may have nothing to do with stochasticity; most essential proteins may simply be needed in somewhat greater quantity than most nonessential proteins, so their genes must be more highly transcribed and/or translated. Since such a relationship could lead to an association between gene importance and the likelihood of adopting expression strategy 1, we employed two statistical methods to control for this possibility. In the first of these two methods, we binned yeast genes by their protein production rate, so that all genes in each of 15 bins had approximately equal levels of protein production (see Table S1 for details). The genes in each bin could have achieved their similar protein production levels by any of the three strategies listed above; our prediction was that if noise in gene expression is relevant to yeast, then essential genes would be biased towards having the highest transcription and lowest translation per mRNA (strategy 1) in each bin. Indeed, this was confirmed by the data: when the genes within each bin were separated into thirds by their number of translations per mRNA, a larger number of essential genes were in the third with the lowest number of translations (low noise) than in the third with the highest number of translations (high noise) for all but one of 15 bins ( Figure 2 A). A Fisher's exact test ( Sokal and Rohlf 1994 ) demonstrated that for all of the 14 bins with more essential genes in the low noise third than the high noise third, this difference was significant ( p ≤ 0.02). Similar results were found when using different numbers of bins, when using halves or quartiles instead of thirds, or when separating bins by transcription rate instead of by number of translations per mRNA (data not shown). This result cannot be explained by the overall positive correlation between dispensability and rate of protein synthesis. (In the binning analysis, the third of each bin with the lowest translation rate had, on average, a slightly lower overall protein synthesis rate than the third with the highest translation rate [data not shown]; this bias is the opposite of what would be expected from the positive correlation between protein synthesis rate and fitness effect or protein complex membership, and thus it acts against our observed bias to make the results of this analysis conservative estimates of the true bias.) Figure 2 Essential Genes and Protein Complex Subunits Minimize Noise in Expression Binning analysis of (A) essential genes and (B) protein complex subunits. All genes for which transcription and translation rate data were available were separated into 15 bins by their protein production rate. Each bin was then separated into thirds by number of translations per mRNA. The two-thirds in each bin with the most extreme transcription and translation are shown: black bars are the number of each type of gene (essential or complex subunit) in the third of each bin with the lowest number of translations per mRNA and the highest transcription rate, and thus low noise; gray bars are the number of each type of gene in the third with the highest number of translations per mRNA and the lowest transcripton rate, and thus high noise. Bins are ordered by their rate of protein synthesis. The number of asterisks indicates the Fisher's exact test probability of observing the values for each bin under the null model of independence. , p ≤ 0.02; , p < 0.005; *, p < 0.0005. Because binning genes still allows for a small amount of variability in protein production within each bin (see Table S1 ), we sought to control for protein production rate in another fashion as well. We employed partial correlation, a method that allows one to examine the relationship between two variables when other, possibly confounding, variables are statistically held constant (see Materials and Methods ). The stochastic model of gene expression ( Blake et al. 2003 ) led us to the prediction that, when protein production rate is controlled for, fitness effect ( f , where f = 0 indicates no effect on growth when a gene is deleted, f = 1 indicates that a gene is essential, and 0 < f < 1 indicates a quantitative growth defect [ Hirsh and Fraser 2001 ]) would correlate positively with transcription rate and negatively with translation rate per mRNA. Indeed, this is what we observed ( f versus transcription [txn] rate \| protein production rate, Spearman partial r = 0.282, n = 4,746, p = 10 −87 ; f versus translations [tlns] per mRNA \| protein production rate, Spearman partial r = −0.258, n = 4,746, p = 10 −75 ). We also expected that the relationship between gene importance and implementation of the expression strategy that minimizes noise could additionally be seen by considering transcription rate and translation rate per mRNA together, as a ratio; a large ratio of transcription rate to translations per mRNA would indicate that transcripts are produced quickly but are translated slowly, corresponding to our expression strategy 1. Confirming this, the correlation between fitness effect and the ratio of transcription rate to translations per mRNA (controlling for protein production rate) is highly significant ( f versus txn rate/tlns per mRNA \| protein production rate, Spearman partial r = 0.275, n = 4,746, p = 10 −86 ). Partial correlation analysis is thus in accordance with the trend illustrated in Figure 2 A: essential genes preferentially use expression strategy 1, which minimizes stochasticity. In addition to essential genes, genes whose protein products participate in stable protein complexes (“complex subunits”) would also be expected to exhibit sensitivity to randomness in expression: producing too little or too much of a single protein complex subunit can compromise the proper assembly of the entire complex and waste the energy invested in the production of the other complex subunits. In support of this, it has been found that both under- and overexpression of complex subunits is more likely to result in a reduced growth rate or inviability of yeast than is misexpression of other genes, and also that complex subunits tend to be more precisely coexpressed with other genes than noncomplex subunits ( Papp et al. 2003 ). Using data from two high-throughput studies that identified proteins involved in stable complexes ( Gavin et al. 2002 ; Ho et al. 2002 ), we assigned genes to two groups: those whose protein products were members of a stable complex found in either study and those whose protein products were not. (Since the protein complex data do not include all protein complexes, we expect that many protein complex subunits will not be classified as such in our list; this, as well as any false positives in the data, makes our results a conservative estimate the true strength of the effect.) We then performed the same binning analysis as described above, substituting our list of complex subunits for our list of essential genes. Again the prediction was confirmed: in all 15 bins, the third of the bin with the least translation per mRNA (and thus the lowest noise level) contained more complex subunits than the third with the most translation per mRNA ( Figure 2 B). The association between low translation per mRNA and protein complex membership was significant (Fisher's exact test, p ≤ 0.02) for all but one bin. As in Figure 2 A, this result is robust with respect to the number of bins and the size of the divisions within bins (data not shown). Also as in Figure 2 A, the bias is the opposite of that expected from the positive correlation between fitness effect and protein production rate; it is also the opposite of what would be the result of highly expressed genes being more likely to appear in the list of protein complex subunits than are poorly expressed genes. (It has been found that highly expressed genes are overrepresented in protein complex data [whether this is an experimental artifact or a true relationship is unclear; von Mering et al. 2002 ]; this would also act against our observed bias of complex subunits being overrepresented in the third with the lowest overall protein synthesis rate in each bin, thus making our results conservative.) When we repeated the partial correlation analysis for complex subunits (genes were assigned a value of one if they were a complex subunit, zero if not), we found similar results. When total protein synthesis was controlled for with the partial correlation, complex subunits were more likely to have a high transcription rate (complex subunit versus txn rate \| protein production rate, Spearman partial r = 0.203, n = 4,900, p = 10 −46 ) and a low number of translations per mRNA (complex subunit versus tlns per mRNA \| protein production rate, Spearman partial r = −0.200, n = 4,900, p = 10 −46 ). Using the ratio of transcription rate to translations per mRNA also yielded similar results (complex subunit versus txn rate/tlns per mRNA \| protein production rate, Spearman partial r = 0.220, n = 4,900, p = 10 −56 ). Thus, partial correlations confirm the finding illustrated in Figure 2 B. Since proteins that participate in many protein–protein interactions are more likely to be essential ( Jeong et al. 2001 ; Fraser et al. 2002 ), it was not immediately clear whether protein fitness effect and membership in a multiprotein complex are independently associated with the expression strategy that minimizes stochastic fluctuations. To address this question, we calculated the partial correlation between fitness effect and the ratio of transcription rate to translations per mRNA, while controlling for both protein production rate and protein complex membership. Likewise, we calculated the correlation between protein complex membership and the ratio of transcription rate to translation rate per mRNA while controlling for both protein production rate and fitness effect. The two partial correlations were both quite significant ( f versus txn rate/tlns per mRNA \| protein production rate, complex membership: Spearman partial r = 0.227, n = 4,746, p = 10 −57 ; complex membership versus txn rate/tlns per mRNA \| protein production rate, f: Spearman partial r = 0.147, n = 4,746, p = 10 −24 ), suggesting that fitness effect and protein complex membership are independently associated with the expression strategy that minimizes stochastic fluctuation. (However, the relative strengths of the partial correlations cannot be interpreted as their true relative contributions because of the differing quality of fitness effect and protein complex membership data.) Repeating the partial correlations above with either transcription rate or translations per mRNA in place of their ratio gave significant partial correlations with both fitness effect and protein complex membership as well (data not shown). The hypothesis that genes of large fitness effect are under stronger selection to reduce stochastic fluctuation in expression provides an explanation for a previously observed, but as yet unexplained, correlate of protein evolutionary rate. Pal et al. (2001 ) noted a weak but significant negative correlation ( r = −0.11, p = 10 −9 ) between an mRNA's rate of decay and the evolutionary rate of the protein it encodes. This correlation was surprising, as it is precisely the opposite of what one would expect if the relationship between the rates of mRNA decay and protein evolution were mediated by the level of expression: slow decay would result in increased expression, which is known to be associated with slow evolution ( Pal et al. 2001 ). Thus, we would expect a positive correlation between rates of mRNA decay and protein evolution, not the negative one that is observed. However, under the present hypothesis that relatively important genes are under stronger selection to reduce noise, the relationship between mRNA decay and protein evolutionary rate is interpretable. Both genes of large fitness effect and genes that encode protein complex subunits are known to evolve slowly ( Hirsh and Fraser 2001 ; Fraser et al. 2002 ; Jordan et al. 2002 ). (While the reason why genes of large fitness effect evolve slowly has been debated [ Hirsh and Fraser 2003 ; Pal et al. 2003 ], the presence of the correlation has not been disputed, and it can be seen to be much stronger than previously reported when using more accurate fitness effect and evolutionary rate data [data not shown]). Here we have shown that they are also associated with a strategy of expression that maximizes the rate of transcription and minimizes the number of translations per mRNA. Given a desired rate of protein production, one way to maximize transcription rate while minimizing the number of translations per mRNA is to maximize the mRNA decay rate. Thus, we would expect rapid mRNA decay among essential genes and protein complex subunits, both of which evolve slowly, yielding the observed negative correlation between the rates of mRNA decay and protein evolution. In support of this prediction, both essential genes and protein complex subunits have substantially shorter mRNA half-lives than the rest of the genome (e.g., mRNA half-lives of nonessential genes are 32% longer than those of essential genes, and the bias remains when controlling for protein production rate; p i= 10 −36 by the Wilcoxon test [ Sokal and Rohlf 1994 ]). Discussion We found that noise in protein production is minimized in genes for which it is likely to be most harmful, specifically essential genes and genes encoding protein complex subunits. This finding supports the hypothesis that noise in gene expression is generally deleterious to yeast. Yeast appear to control the noise in their gene expression at both transcriptional and translational levels preferentially for some genes; however, this noise minimization is not without a cost, as the high transcription and high mRNA decay rates that are needed to minimize noise are energetically expensive and are thus expected to be advantageous only when the benefit of reducing noise in a particular gene's expression outweighs this cost ( Thattai and van Oudenaarden 2000 ). Protein degradation rates may also play a role in controlling noise, but this cannot be tested until genome-wide protein degradation rates have been measured. As is the case with many genome-wide studies, it is possible that a hidden variable could bias our results. For example, it is possible that essential genes and genes encoding protein complex subunits tend to have high transcription and low translation for reasons unrelated to noise minimization. However, until such a reason is identified, the most parsimonious interpretation of our results is that yeast adaptively minimize noise in the expression of certain genes. As genome-wide transcription and translation rate data become available for other organisms, it will be interesting to see if the apparent tendency to minimize noise in the expression of important genes extends to organisms other than yeast. Considering that several anecdotal examples of indispensable genes with unusually low translation rates, and thus low noise in expression, have already been noted in Escherichia coli ( Ozbudak et al. 2002 ), this could well be the case. Materials and Methods Functional genomic data sources Transcription rates were calculated from mRNA abundances and decay rates in log-phase yeast growing in rich glucose medium ( Wang et al. 2002 ) according to the steady-state equation R = −ln(0.5) A / T , where R is transcription rate, A is mRNA abundance, and T is mRNA half-life. Translation rates per mRNA in rich glucose medium were calculated from ribosome occupancy data by Arava et al. (2003 ); specifically, ribosome density per mRNA present in the polysome fraction was multiplied by the fraction of each mRNA that was found in the polysome fraction to estimate the average ribosome density for all copies of each mRNA in a cell. This density is equivalent to a relative translation rate, assuming that the speed at which ribosomes produce proteins is constant over different mRNAs. An estimate of the actual translation rate was found by multiplying the relative translation rates by a constant: the speed of translation, which is approximately ten amino acids/s ( Arava et al. 2003 ). Protein production rate (proteins/s) was then calculated by multiplying translation rate per mRNA with mRNA abundance. Note that the protein production rate can also be represented as the product of transcription rate and number of translations per mRNA. It is this latter variable that was used to separate each bin into thirds in Figure 2 , since it is thought to be more directly relevant to noise in protein production than related quantities such as translation rate per mRNA ( Berg 1978 ); the variable was calculated by dividing protein production rate by transcription rate for each gene. However, bins could also be separated into thirds by transcription rate, transcript abundance, or translation rate per mRNA, all yielding similar results (data not shown). Fitness effect ranks were calculated from 12 replicate growth experiments for all viable homozygous yeast deletion strains in rich glucose medium; growth experiments were conducted using the method described in Giaever et al. (2004 ). The logarithms of deletion strain tag fluorescence intensities on high-density oligonucelotide microarrays for each growth time course were fitted to a linear model that accounted for time-course-specific effects and variable initial strain concentrations. The slope of the linear regression was then used as the relative growth rate for each strain. Estimates of percent induction levels Blake et al. (2003 ) showed for two different promoters in yeast, as well as in their mathematical model, that noise due to transcription peaked at approximately one-third of maximal transcriptional induction. Importantly, one of their promoters (P ADH1* ) was 10-fold weaker than the other two at full induction, but all three showed very similar relationships between noise strength and percent transcriptional induction. Since we do not have genome-wide data for the percent induction for genes in rich glucose medium (or any other environment), in our analysis we make the assumption that the promoters of more highly transcribed genes tend to be at higher percent induction levels. While this certainly does not hold for all genes, we believe that it is a reasonable approximation for most genes. Partial correlations Partial correlations were calculated as described by Sokal and Rohlf (1994 ). Briefly, let r XY be the correlation coefficient between variables X and Y. To control for a third variable Z, To assess the significance of the partial correlation, it is transformed to be distributed according to a Student's t distribution, by the equation The two-sided p -value can then be calculated according to where the t -value falls with respect to its expected distribution. Supporting Information Figure S1 The Relationship between Fitness Effect and Protein Production Rate Fitness effect ranks are shown on the y-axis (the large number of points at 519.5 are the essential genes, with fitness effect = 1). Protein production rate (proteins/s) is shown on the x-axis. The Spearman rank correlation coefficient is r = –0.202 (p = 10 –49 ). (316 KB PPT). Click here for additional data file. Table S1 Details of the Protein Production Rates (Protein/s) within Each Bin from Figure 2 (37 KB DOC). Click here for additional data file.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC400249.xml
509251	The usefulness of the Korean version of modified Mini-Mental State Examination (K-mMMSE) for dementia screening in community dwelling elderly people	Background We assessed whether the Korean version of modified Mini-Mental State Examination (K-mMMSE) has improved performance as a screening test for cognitive impairment or dementia in a general population compared with the Korean Mini-Mental State Examination (K-MMSE). Methods Screening interviews were conducted with people aged 65 and over in Noam-dong, Namwon-city, Jeonbuk province. There were 522 community participants, of whom 235 underwent clinical and neuropsychological examination for diagnosis of dementia and Cognitive Impairment No Dementia (CIND). Sensitivity, specificity and areas under the receiver operating characteristic (ROC) curves for the K-mMMSE and the K-MMSE were the main outcome measures. Results Cronbach's alpha for the K-mMMSE was 0.91, compared with 0.84 for the K-MMSE. The areas under the ROC curves in identifying all levels of CIND or dementia were 0.91 for the K-mMMSE and 0.89 for the K-MMSE ( P < 0.05). For the K-mMMSE, the optimal cut-off score for a diagnosis of CIND was 69/70, which had a sensitivity of 0.86 and a specificity of 0.79, while, for a diagnosis of dementia, the optimal cut-off score of 59/60 had a sensitivity of 0.91 and a specificity of 0.78. The K-mMMSE also had a high test-retest reliability ( r = 0.89). Conclusion Our findings indicate that the K-mMMSE is more reliable and valid than the K-MMSE as a cognitive screen in a population based study of dementia. Considering the test characteristics, the K-MMSE and modified version are expected to be optimally used in clinical and epidemiologic fields.	Background The Mini-Mental State Examination (MMSE) is a brief screening test that quantitatively assesses the cognitive status of elderly people [ 1 , 2 ]. It is easy to administer and has shown good reliability. Although its validity as a screening test is acceptable for clinical samples, it has been shown to have difficulty in discriminating between demented and non-demented individuals in community-based samples [ 2 ]. The MMSE has been found to be influenced largely by pre-morbid ability and is less sensitive to focal brain dysfunction [ 3 ] or mild dementia [ 4 ]. These limitations led to the development of the Modified Mini-Mental State Examination (3MS) in 1987 [ 5 ], which expanded the MMSE from 30 to 100 points to provide finer discrimination. In addition, the 3MS added four items: personal information, including date and place of birth; verbal fluency; abstract reasoning; and a second delayed recall trial. The 3MS also graded temporal orientation and broadened the delayed recall measures, which included cued and recognition formats. While retaining the brevity and ease of administration of the MMSE, the 3MS improved the validity and reliability of identifying individuals with dementia [ 6 , 7 ], and in predicting functional outcomes in patients with stroke [ 8 ]. In 2002, a group from the Cache County Study modified the 3MS for use as a cognitive screen in an epidemiologic study of dementia [ 9 ]. These modifications substituted the recall of personal demographic information with the recall of current and past prominent politicians. The main reason for this modification was the difficulty the researchers had in verifying personal demographic information. In addition, the scaling of the items in the time orientation and writing parts of the test was changed, and the time allotted for animal naming was shortened. This revised form of the 3MS (3MS-R) demonstrated good sensitivity in detecting dementia in a general population and providing age- and education-specific normative data and cut-off values at the 7 th and 10 th percentiles [ 9 ]. In Korea, the MMSE was translated into two versions, the Korean version of the Mini-Mental State Examination (MMSE-K) and the Korean Mini-Mental State Examination (K-MMSE) [ 10 , 11 ]. Both Korean versions of the MMSE were tested for validity and efficacy in clinical settings [ 11 , 12 ] and partly in epidemiologic research [ 13 ]. Both were somewhat modified to adjust better to the cultural background in Korea, but both shared all the limitations of the original MMSE. Recently the Korean Modified Mini-Mental State Examination (K-3MS) was introduced and validated in a clinical setting [ 14 ]. Although the K-3MS was found to be a reliable cognitive screening measure, there was no significant difference between the K-3MS and extracted MMSE for detecting individuals with dementia. In addition, components of the K-3MS could not be compared with items extracted from the K-MMSE and MMSE-K. We have therefore introduced the Korean version of modified Mini-Mental State Examination (K-mMMSE), and we have determined whether its validity is superior to that of the K-MMSE as a screen for cognitive impairment or dementia in a community setting. Methods Subjects Potential participants for this study were recruited from all inhabitants of Noam-dong, Namwon-city, Jeonbuk-province, South Korea, aged 65 and over in 2003, as recorded in national residents registration lists. The area surveyed covered 8.93 km 2 and had an estimated population of 6,883, of whom about 7% were aged 65 or over. All participants gave informed consent, and the study was conducted in accordance with the guidelines in The Declaration of Helsinki and approved by the appropriate research ethics committee. Among the 522 eligible subjects aged 65 and over identified from registration lists, 235 (45%) completed clinical examinations after the interview and formed the study sample for principal analysis. Of the remaining subjects, we were unable to establish contact with 162 (31%), 75 (14%) refused to participate, 18 (3%) did not complete the survey, 7 (1%) had severe pre-morbid illness including blindness and deafness, 4 (0.8%) had changed address, and 3 (0.6%) had died before the visit. The principal apparent reason for the difficulties in establishing contact was that the person was in a regular daily activity or away from home, visiting family members living elsewhere. The 18 individuals (3%) who did not complete the survey questionnaire or were not examined clinically had a mean age of 74.9 ± 10.3 years; 13 (72%) were females, and 4 (22%) were educated. Of all the eligible subjects, participants and non-participants did not differ in age (73.5 ± 6.8 y vs. 74.6 ± 7.8 y, respectively) or gender (66% and 62%, all P values > 0.1). Assessment and measurements Interviewers received a seven-day training session on administering the screening instruments and were supervised throughout by the project neurologist. Cognitive status was classified in two stages. In the first stage, interviewers carried out home-based interviews for data on cognitive function, past medical history, and demographic characteristics. All participants were contacted for cognitive screening using a formulated battery, from which the K-mMMSE and K-MMSE were extracted. And they were rated by a knowledgeable informant using the Short form of Samsung Dementia Questionnaire (S-SDQ) and Korean Instrumental Activities of Daily Living (K-IADL) [ 15 , 16 ]. The S-SDQ is a Korean version of the Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE), with 15 items and scores ranging from 0 to 30 [ 15 ]. The K-IADL is composed of 11 items that grade functional abilities, with scores calculated as the sum of points over the number of applicable questions and ranging from 0.0 to 3.0 [ 16 ]. High scores on the S-SDQ and K-IADL indicate poor performance. Both tests were found to be uncontaminated by pre-morbid ability, including education or age [ 15 , 16 ]. At the second interview, physicians who were blinded to the cognitive scores performed a clinical examination and neuropsychiatric inventory on participants who completed the first survey questionnaire. The clinical examination verified the presence of cognitive impairment. The diagnostic criteria for dementia were based on those of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) [ 17 ] and were subdivided according to the guidelines of the National Institutes of Neurological and Communicative Disorders and Stroke and the Alzheimer's disease and Related Disorders Association (NINCDS-ADRDA) [ 18 ]. Severity of dementia was staged using the Korean version of the Expanded Clinical Dementia Rating (CDR) scale [ 19 ]. The physician and neurologist made independent diagnoses and CDR scoring and subsequently held a case conference to reach a consensus diagnosis, classifying the person as either cognitively normal, cognitively impaired with no dementia (CIND), or having dementia. A diagnosis of CIND represents an attempt to classify people with recognizable cognitive decline who did not meet the criteria for dementia. This group included people who complained of cognitive decline and showed impaired memory function, but did not have any non-cognitive alterations, including intact activities of daily living. At both stages, home visits were repeated on at least two occasions if no contact was made. Instruments We translated the original 3MS-R into Korean according to the guidelines recommended by the modifiers [ 9 ]. The item regarding political figures, which asked the participants to name the current president, vice president, and state governors, was replaced with questions about current and previous presidents, because there is no vice president in the government of South Korea. We assessed temporal orientation according to three methods used to calculate year and time in Korea: the solar, lunar and Tangun era. We replaced the words "shirt," "nickel" and "honesty" in the memory task with the words "airplane," "pine tree" and "sincerity." In Korean, the first two words coincided with the K-MMSE items and ended with vowel sounds ("bee-haeng-gi," and "so-na-mu," respectively); while "seong-sil" in Korean, which means sincerity, was substituted for "jeong-jik," which is equivalent to the English word honesty, inasmuch as a pilot study found that "jeong-jik" was more difficult to hear or perceive than "seong-sil", possibly because the latter ended with a voiced sound and could be heard more comfortably. In explaining the appropriate questions and answers, we presented a simple example prior to asking the first question, specifically, "the eyes and nose are different, but they are similar in being part of our face." In a pilot study, most elderly subjects could not understand the concept of similarity without this example, and they became intolerant to our interview unless an example was provided. After providing the example, however, most subjects were more cooperative and tried to answer properly. We asked subjects to write a spontaneous sentence and scored whether it was legible and correct, with or without prompting. The total possible score was 100, and a K-MMSE score could be generated from it. Statistical analysis For demographic factors, mean or median values and proportions were calculated according to the cognitive impairment strata. Cronbach's alpha was used to quantify internal consistencies of the K-mMMSE and K-MMSE [ 20 ]. We assessed the stability and test-retest reliability with 30 subjects who took the K-mMMSE twice. Receiver operating characteristic (ROC) curves were used to determine the validity of the two screening tests graphically and statistically. The areas under the ROC curves (AUC) were calculated and described with standard errors (SE) using the trapezoidal rule [ 21 ], and comparison between AUCs was made by the algorithm with an estimated covariance matrix [ 22 ]. Cut-off points were chosen to optimize the trade-off between false-negative and false-positive rates. The choice of whether to judge the screening tests by their ability to identify CIND or dementia was addressed practically by assessing both. The validity of the K-mMMSE and K-MMSE were compared, first between those with CIND or dementia and those who were normal, and then between those with dementia and those with CIND or normal. All statistical calculations were performed using Stata 8.2 software (Stata, College Station, TX). Results Descriptive statistics Descriptive statistics are displayed in Table 1 for the total sample and according to the cognitive impairment strata. Of the 235 participants, 46 (19.6%) were classified as having dementia and 54 (22.9%) as having CIND. Overall levels of education were very low, in that 118 participants (50.2%) had no formal education, of whom 83 (70.3%) were illiterate. Severity of cognitive impairment was directly related to mean age and inversely related to number of years of education. Women had higher rates of being cognitively impaired or demented than men. The median scores on the K-mMMSE and K-MMSE decreased with severity of cognitive impairment. Median scores on the informant questionnaires of the S-SDQ and K-IADL increased with poorer cognitive status. The inter-quartile ranges (IQR) also steadily increased, reflecting increasing variability in cognitive and functional status. Four respondents scored perfectly on the K-MMSE, whereas none scored perfectly on the K-mMMSE, suggesting that the K-mMMSE might be less prone to the ceiling effect in this population. Table 1 Subject characteristics All subjects (n = 235) Normal (n = 135) CIND (n = 54) Dementia (n = 46) Demographic characteristics Age, mean ± SD, y 73.5 ± 6.7 71.9 ± 5.3 73.8 ± 7.4 77.8 ± 7.8 Women, % 66.4 57.8 74.1 82.6 Education, median (IQR), y 1 (0–6) 4 (0–6) 0 (0–5) 0 (0–2) Cognitive measures, median (IQR), score K-mMMSE 64 (48–80) 78 (66–85) 54 (44–63) 38 (29–47) K-MMSE 20 (14–25) 24 (20–27) 16 (13–21) 12 (9–15) S-SDQ 9 (5–13) 7 (3–10) 10 (5–14) 14 (10–21) K-IADL 0.22 (0.09–0.50) 0.11 (0.00–0.29) 0.27 (0.11–0.60) 1.10 (0.40–1.67) CIND; Cognitive Impairment No Dementia, SD; standard deviation, IQR; Inter-quartile range, K-mMMSE; Korean version of modified Mini-Mental State Examination, K-MMSE; Korean Mini-Mental State Examination, S-SDQ; Short form of the Samsung Dementia Questionnaire, K-IADL; Korean Instrumental Activities of Daily Living. Reliability The estimated Cronbach's alpha was 0.91 for the K-mMMSE and 0.84 for the K-MMSE. Relative to each cognitive impairment stratum (normal, CIND, and dementia), the alphas were 0.84, 0.81 and 0.81, respectively, on the K-mMMSE and 0.74, 0.72, and 0.63, respectively, on the K-MMSE. Neither age nor gender had any substantial impact on internal consistency. Stratum-specific alphas of the K-mMMSE for the different subgroups ranged from 0.86 to 0.91 in men and 0.89 to 0.90 in women. The retest of the K-mMMSE was assessed in 30 subjects (mean interval, 26 days; range, 19–32 days). The correlation coefficients for the total scores were 0.89 on the K-mMMSE and 0.85 on the K-MMSE. The coefficients of the 15 items of the K-mMMSE were all significant, ranging from 0.37 for similarities to 0.83 for time orientation. The re-tested subjects were representative of the entire study population, in that the sociodemographic characteristics of the 30 retested subjects were similar to the other participants in mean age (74.0 ± 7.4 y vs. 73.5 ± 6.7 y), educational years (3.5 ± 3.2 y vs. 3.4 ± 3.9 y), K-mMMSE scores (63.8 ± 21.0 vs. 63.1 ± 20.4), and proportion of women (80.0% vs. 66.4%; P = 0.132). Validity of the K-mMMSE and K-MMSE Construct validity Construct validity data between the K-mMMSE and other cognitive or functional measures are shown in Table 2 . K-mMMSE was found to be significantly correlated with all measures, including CDR, Sum of Boxes of CDR (CDR-SB), and informant questionnaires such as the S-SDQ and K-IADL. The correlation coefficient between K-mMMSE and K-MMSE scores was 0.94. According to the CDR scores, the median values of the K-mMMSE and K-MMSE changed significantly (Table 3 ). Table 2 Correlations between K-mMMSE, K-MMSE and cognitive or functional measures (CDR, KIADL, and S-SDQ) K-mMMSE K-MMSE CDR CDR-SB KIADL S-SDQ K-mMMSE 1.000 K-MMSE 0.945* 1.000 CDR -0.755* -0.710* 1.000 CDR-SB -0.750* -0.702* 0.966* 1.000 KIADL -0.648* -0.614* 0.740* 0.793* 1.000 S-SDQ -0.489* -0.454* 0.529* 0.555* 0.617* 1.000 * P < 0.001 by Pearson's correlation analyses. K-mMMSE; Korean version of modified Mini-Mental State Examination, K-MMSE; Korean Mini-Mental State Examination, CDR; Clinical Dementia Rating, CDR-SB; Sum of Boxes of CDR, K-IADL; Korean Instrumental Activities of Daily Living, S-SDQ; Short form of the Samsung Dementia Questionnaire Table 3 K-mMMSE and K-MMSE scores for each CDR group CDR 0 (n = 137) 0.5 (n = 52) 1 (n = 33) 2+ (n = 13) K-mMMSE, Median (IQR)* 78 (66–85) 54 (44–63) 44 (35–50) 29 (12–31) K-MMSE, Median (IQR)* 24 (20–27) 16 (13–21) 12 (9–16) 10 (5–11) * P < 0.001 by Kruskal-Wallis test. K-mMMSE; Korean version of modified Mini-Mental State Examination, K-MMSE; Korean Mini-Mental State Examination, CDR; Clinical Dementia Rating, IQR; Inter-quartile ranges. Identification of combined CIND and dementia The performance of the K-mMMSE (AUC ± SE, 0.91 ± 0.02) was significantly superior to that of the K-MMSE (0.89 ± 0.02; P = 0.041). The ROC curves plotted in the same graph suggested that the performance of the K-mMMSE was superior to that of the K-MMSE at almost all cut-off points (Figure 1 ). At a cut-off of 69/70 for CIND, the K-mMMSE had a sensitivity of 0.86 (95% Confidence Intervals, 0.78–0.92), a specificity of 0.79 (0.71–0.86), a positive likelihood ratio (LR) of 4.15, and a negative LR of 0.18. In comparison, at a cut-off of 20/21 for CIND, the K-MMSE had a sensitivity of 0.82 (0.73–0.89), a specificity of 0.79 (0.71–0.86), a positive LR of 3.95 and a negative LR of 0.23. Figure 1 Receiver operating characteristic (ROC) curves of the K-mMMSE and the K-MMSE for combined CIND and dementia. K-mMMSE (light blue), K-MMSE (brown), and diagonal line. Area under ROC Curves (AUC): K-mMMSE = 0.91, K-MMSE = 0.89. Identification of dementia Both the K-mMMSE and the K-MMSE could properly discriminate demented from normal individuals, but there was no significant difference between them (AUC, 0.92 vs 0.91; P = NS). At the cut-off of 59/60, the K-mMMSE had a sensitivity of 0.91 (0.79–0.98), a specificity of 0.78 (0.72–0.84), a positive LR of 4.21 and a negative LR of 0.11. At a cut-off point of 18/19, the K-MMSE had a sensitivity of 0.91 (0.79–0.98), a specificity of 0.76 (0.69–0.82), a positive LR of 3.82 and a negative LR of 0.11. Discussion We have shown here that the K-mMMSE is a valid, reliable, and stable cognitive screening instrument, as well as being more sensitive to all levels of CIND and dementia, compared to the K-MMSE. The K-mMMSE has been shown to have a broader spectrum of cognitive domains, including political figures, word fluency, similarities, and delayed recall. Furthermore, the expanded 100 point scoring allows finer discrimination of cognitive impairment. Thus, the K-mMMSE represents a summary form of administration and scoring. Internal consistency results of the K-mMMSE and K-MMSE were comparable to those observed in previous community studies. Cronbach's alpha (á) for the 3MS has been reported to be 0.91 in a community study, a value identical to that found here [ 23 ]. Another population study has reported alphas for the 3MS and MMSE of 0.87 and 0.78, respectively, which were slightly lower than our findings of 0.91 and 0.84 [ 6 ]. Cronbach's alpha has been reported to be influenced by educational status or variability of response, in that it was higher in groups having fewer years of education [ 24 ] and in clinical populations having greater variability [ 25 ]. Our population consisted of a high percentage with no formal education (50.2%), and their scores were very variable (inter-quartile ranges for the K-mMMSE and the K-MMSE of 48–80 and 14–25, respectively). Test-retest reliability results of the K-mMMSE were also comparable to those in previous studies. Correlation coefficients of 0.91 to 0.93 have been reported for small samples of community dwelling residents and dementia patients, which are slightly higher than our value of 0.89 [ 26 ]. The Stirling County Study found a coefficient of 0.78, but items requiring less judgment exhibited lower reliability than items requiring more judgment [ 23 ]. In contrast, we found markedly lower reliability in items requiring more judgment, i.e., similarities ( r = 0.37), compared with simple items, i.e., temporal orientation ( r = 0.81). The discrepancy might be due to a difference of time lag, in that the Stirling County Study retest was performed over a 3 year interval, with individual retests ranging from 0.9 to 4.0 years. Furthermore their retested subjects were not representative of all participants. We observed a correlation coefficient of 0.85 for the K-MMSE over all levels of cognitive status, which is in line with generally acceptable findings [ 2 ]. Scores on the K-mMMSE and K-MMSE increased after retest, with differences in mean values of 4.4 and 2.6 points, respectively, presumably due to a practice or studying effect after a short interval [ 1 , 27 , 28 ]. The K-mMMSE was superior to the K-MMSE for diagnosis of all levels of CIND or dementia, as well as being slightly superior at almost all cut-off points. Since dementia is usually preceded by CIND or mild cognitive impairment (MCI), the definition of both requires explication [ 29 , 30 ]. Subjects with CIND or MCI have been found to be at increased risk for developing dementia or, more specifically, Alzheimer's disease and some vascular subtypes of dementia [ 30 , 31 ]. The difference between the K-mMMSE and the K-MMSE in diagnosing this condition might mean that the former was more sensitive to the mild stage or pre-dementia than the latter. In this respect, the K-mMMSE seemed to partially overcome a weakness of the K-MMSE, that is, insensitivity to mild dementia [ 4 ]. The present findings suggested that the K-MMSE was actually a fairly reasonable instrument as well. Given the faster administration of the K-MMSE, it would be a choice of clinicians to use optimally, recognizing that the K-MMSE was slightly inferior in terms of its test characteristics. The two cognitive screening measures did not differ significantly, however, in the detection of dementia. These results are comparable to the findings of McDowell et al. [ 6 ]; however, their validity results differed between the two language groups studied, namely French and English speakers. The 3MS was superior only in the diagnosis of combined CIND or dementia in French, but not English, speaking participants. There were fewer French than English participants (434 vs. 1166), and they had fewer years of education (6.8 vs. 9.2 years). These differences were also observed in our study samples, with the most important being the smaller sample size, inasmuch as statistical significance was directly influenced by the total number of participants [ 22 ]. Even Cache County modifications to the 3MS showed a good sensitivity and specificity, 3MS-R was also dependent on their cultural and social factors which might limit general use in non-US population [ 9 ]. For this reason, cross validation was a very important step for a cultural validation of the instrument. The K-mMMSE was shown to be more significantly correlated with other tests for cognitive status or functional abilities, such as the CDR, S-SDQ, and K-IADL, than was the K-MMSE. The correlation coefficients of the CDR were higher than the informant questionnaires, which might be due to the characteristics of the questions. That is, the K-mMMSE and K-MMSE are cognitive screening measures, and the CDR includes items about the cognitive aspects for scoring, whereas the informant questionnaires (S-SDQ and K-IADL) are comprised only of questions related to functional abilities. To the best of our knowledge, this is the first report showing concurrent validity of the modified MMSE series. There are important limitations to our findings. First, the subjects who participated in this study showed very low levels of educational background, perhaps limiting its general usefulness, especially regarding the cut-off points for a diagnosis of CIND or dementia. The low educational attainment, however, has been one of important characteristics of our elder population, because they were largely deprived of education due to Korean War and Japanese colonial dominion over the country [ 32 ]. And the study design, which showed the validities of and comparison between the two cognitive screening measures, would be appropriate for selected community samples, because all participants have a two-stage interview and a clinical examination, thus reducing verification bias. Second, although the trapezoidal rule provide a more accurate method of estimating the "true" AUC, an AUC derived from the parameters of a straight-line fit to the ROC plot tends to slightly underestimate the AUC of a Gaussian-based ROC. Finally, although we observed no significant differences between participants and non-participants, the rate of participation in our study was somewhat low. The majority of non-participants were those with whom we could not meet on two separate visits, suggesting that individuals who refused to participate may be more intelligent or active than the participants. If this were true, however, our results would not change, and additional statistical power may be added to our analysis. Conclusions We conclude that the K-mMMSE is a valid, stable, and reliable cognitive screen in an epidemiologic study. The K-mMMSE is more sensitive to all levels of CIND and dementia than the K-MMSE. Future investigations with the K-mMMSE will examine age-, sex-, and education-specific reference values to determine how performance patterns on individual items may discriminate between those with or without cognitive impairment and dementia subtypes. Competing interests None declared. Authors' contributions SKJ performed physical measurements, collected data, and drafted the manuscript. KHC participated in data collection and reviewed the manuscript. JMK conceived of the study and participated in its design, and also performed physical measurements. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC509251.xml
520833	Comparison in gene expression of secretory human endometrium using laser microdissection	Background The endometrium prepares for implantation under the control of steroid hormones. It has been suggested that there are complicated interactions between the epithelium and stroma in the endometrium during menstrual cycle. In this study, we demonstrate a difference in gene expression between the epithelial and stromal areas of the secretory human endometrium using microdissection and macroarray technique. Methods The epithelial and stromal areas were microdissected from the human endometrium during the secretory phase. RNA was extracted and amplified by PCR. Macroarray analysis of nearly 1000 human genes was carried out in this study. Some genes identified by macroarray analysis were verified using real-time PCR. Results In this study, changes in expression <2.5-fold in three samples were excluded. A total of 28 genes displayed changes in expression from array data. Fifteen genes were strongly expressed in the epithelial areas, while 13 genes were strongly expressed in the stromal areas. The strongly expressed genes in the epithelial areas with a changes >5-fold were WAP four-disulfide core domain 2 (44.1 fold), matrix metalloproteinase 7 (40.1 fold), homeo box B5 (19.8 fold), msh homeo box homolog (18.8 fold), homeo box B7 (12.7 fold) and protein kinase C, theta (6.4 fold). On the other hand, decorin (55.6 fold), discoidin domain receptor member 2 (17.3 fold), tissue inhibitor of metalloproteinase 1 (9 fold), ribosomal protein S3A (6.3 fold), and tyrosine kinase with immunoglobulin and epidermal growth factor homology domains (5.2 fold) were strongly expressed in the stromal areas. WAP four-disulfide core domain 2 (19.4 fold), matrix metalloproteinase 7 (9.7-fold), decorin (16.3-fold) and tissue inhibitor of metalloproteinase 1 (7.2-fold) were verified by real-time PCR. Conclusions Some of the genes we identified with differential expression are related to the immune system. These results are telling us the new information for understanding the secretory human endometrium.	Introduction Many studies have sought to understand the mechanism of implantation. Recently, the rate of pregnancy in the in vitro fertilization and embryo transfer (IVF-ET) cycle has declined, and this has been attributed to a decrease in the rate of implantation. The recently developed laser microdissection method has gained widespread use throughout the research field. Information about cells can be determined without contamination by using this method. Moreover, with the macroarray technique, which was already widely used for this purpose, the profiling of the gene expression of specific cells types has become possible. Torres et al. have already reported differences in gene expression between cell types or regions within the monkey endometrium using laser microdissection and differential display [ 1 , 2 ]. Identification of cell-specific proteins, which are expressed in the endometrium during the secretory phase, has been performed using a multi-disciplinary approach in the same trial. IGF-II mRNA is expressed in the mid-to-late secretory phase and in early pregnancy [ 3 ]. During decidualization, interstitial collagen in the mouse endometrium increases [ 4 ]. Collagen IV and laminin reactivity increased in the basal lamina and underlying subepithelial stroma as pregnancy proceeds [ 5 ]. The most abundant expression of IL-15 mRNA during the menstrual cycle is observed in the midsecretory phase [ 6 ]. Leukaemia inhibitory factor (LIF) is known as an indispensable factor for implantation and is expressed in the glandular epithelium at the time of implantation in human endometrium [ 7 - 9 ]. In this study, we demonstrate differential in gene expression between the epithelial and stromal areas obtained from secretory human endometrium using laser microdissection and the macroarray method. Confirmation of differential expression of candidate genes was performed by real-time PCR. Materials and Methods Materials Human endometrium was obtained from 8 patients (25–38 years old) with normal menstrual cycles (28~30 days) during the mid secretory phase. These patients had had at least one intrauterine pregnancy in the past (3 patients for Microarray analysis, 5 patients for real-time PCR). Part of the endometrial biopsy was obtained with a curetting technique. The day of the menstrual cycle was determined by the patient's history, plasma progesterone levels (9.8~17.3 ng/ml) and the histological criteria of Noyes et al [ 10 ]. These patients did not receive any hormonal therapy. Informed consent was obtained from all patients who participated in this study. The Institutional Review Boards of Showa University approved the use of human subjects and the procedures. Methods Laser Microdissection and RNA extraction The endometrium was embedded in OCT compound and frozen immediately in isopentane that had been cooled in the liquid nitrogen. This freezing block was sliced by a cryomicrotome at 8 μm thickness. Frozen sections were fixed in 100% methanol for 3 min and stained with 1% toluidine blue. The section was laser-microdissected by the PALM MicroBeam system (PALM Microlaser. Technologies A.G.) for epithelial and stromal areas and collected in a small tube (Fig 1a,1b,1c,1d ). Approximately 30–50 sections were laser-microdissected. Contamination with non-target components was monitored morphologically. Total RNA was extracted from the tissue section using the acid guanidinium-phenol-chloroform method [ 11 ]. Figure 1 Human endometrial epithelial areas (a-b) and stromal areas (c-d) were laser-microdissected. (200×) Macroarray The RNAs obtained were synthesized from cDNA using a modified oligo (dT) primer and the BD SMART™ PCR cDNA Synthesis Kit (BD Biosciences Clontech, Palo Alto, CA). cDNA was PCR amplified for 24–29 cycles according to the user manual. (BD Atlas™ SMART™ Probe amplification Kit (BD Biosciences Clontech, Palo Alto, CA)). 550 ng of cDNA sample was labeled with α- 32 P dCTP (3000 Ci/mmol) using a randam primer. Labeled probes were hybridized to a nylon array (BD Atlas™ Nylon cDNA Expression Arrays, Human 1.2 Array (BD Biosciences Clontech, Palo Alto, CA)) in ExpressHyb solution at 68°C overnight. After hybridization, the nylon membrane was washed with 2 × standard saline citrate (SSC) + 1% sodium dodecyl sulphate (SDS) (WAKO Pure Chemical Ltd, Japan) once, twice with 1.0 × SSC + 0.5% SDS at 68°C [ 12 , 13 ]. The membrane was exposed to a phosphor screen (Fujifilm, Japan) for 24 hours and scanned using a STORM 830 Scanner and IMAGEQUANT 4.1-J (Molecular Dynamics). Hybridization signal intensities for individual genes were a subtracted from the background and normalized to the signals for GAPDH and the beta-actin gene, respectively, using AIS (Analytical Imaging Station) Array™ (IMAGING Research INC.). Each normalized a gene expression signal the epithelial and stromal areas was compared, and was automatically calculated as a ratio [ 14 , 15 ]. In this study, a change in expression <2.5-fold in all three samples was excluded. Real-time PCR RNA was reverse transcribed using oligo (dT) primers by TaKaRa RNA PCR Kit (AMV) Ver 2.1 (TAKARA BIO INC, Shiga, Japan) according to the manufacturer's instructions. PCR was performed using the ABI PRISM 7700 Sequence Detection System. TaqMan Universal PCR MasterMix and Assays-on-Demand Gene Expression probes (Applied Biosystems) were used for the PCR step (Assay ID for MMP7; Hs00159163 m1, WFDC2; Hs00707910 s1, TIMP1; Hs00171558 m1 Decorin; Hs00370385 m1). Primer sequences are not publicly available, although their validity has been established by the manufacturer. The expression values obtained were normalized against those from the control human GAPDH [ 16 ]. Statistical significance was determined by the Wilcoxon test and defined as p < 0.05. Results Microdissection and Microarray Secretory endometrium was collected from 8 patients (3 patients for Microarray, 5 patients for real-time PCR). Each sample was carefully dissected by laser microdissection for epithelial and stromal areas (Fig 1a to 1d ). Total RNA was extracted and subjected to macroarray with nearly 1000 genes on the nylon membrane. Fifteen genes were strongly expressed in the epithelial areas (Table 1 ), while 13 genes were strongly expressed in the stromal areas (Table 2 ). Mean values are shown in Tables 1 and 2 . Genes strongly expressed in the epithelial areas that increased >5-fold in expression included WAP four-disulfide core domain 2 (WFDC2), matrix metalloproteinase 7 (MMP7), homeo box B5, msh homeo box homolog, homeo box B7 and protein kinase C, theta (PKC theta). On the other hand, decorin, discoidin domain receptor member 2 (DDR2), tissue inhibitor of metalloproteinase 1 (TIMP1), ribosomal protein S3A, and tyrosine kinase with immunoglobulin and epidermal growth factor homology domains (Tie1) were strongly expressed in the stromal areas. Table 1 Expressed gene list in epithelial areas. Ratio GENE BANK LOCUS LINK Gene Name Classifications 44.1 X63187 10406 WAP four-disulfide core domain 2 inhibitors of proteases 40.1 X07819 4316 matrix metalloproteinase 7 metalloproteinases 19.8 M92299 3215 homeo box B5 CDK inhibitors 18.8 M97676 4487 msh homeo box homolog 1 (Drosophila) transcription activators and repressors 12.7 M16937 3217 homeo box B7 transcription activators and repressors 6.4 L07032 5588 protein kinase C, theta intracellular kinase network members 4.8 X59798 595 cyclin D1 cyclins 4.5 M97796 3398 inhibitor of DNA binding 2, dominant negative helix-loop-helix protein transcription activators and repressors 3.9 X02920 5265 serine proteinase inhibitor, clade A inhibitors of proteases 3.8 D14520 688 Kruppel-like factor 5 basic transcription factors 2.9 U24166 22919 microtubule-associated protein, RP/EB family, member 1 adaptors and receptor-associated proteins 2.9 X67055 3699 pre-alpha (globulin) inhibitor inhibitors of proteases 2.8 U26710 868 Cas-Br-M (murine) ectropic retroviral transforming sequence b adaptors and receptor-associated proteins 2.7 D45132 7799 PR domain containing 2, with ZNF domain transcription activators and repressors 2.5 AF059244 10047 cystatin 8 inhibitors of proteases Ratio shows average of 3 samples Table 2 Expressed gene list in stromal areas Ratio GENE BANK LOCUS LINK Gene Name Classifications 55.6 M14219 1634 decorin cell surface antigens 17.3 X74764 4921 discoidin domain receptor family, member 2 intracellular transducers 9 X03124 7076 tissue inhibitor of metalloproteinase 1 extracellular secreted proteins 6.3 M77234 6189 ribosomal protein S3A ribosomal proteins 5.2 X60957 7075 tyrosine kinase with immunoglobulin and epidermal growth factor homology domains intracellular transducers 4.9 M57399 5764 pleiotrophin (heparin binding growth factor 8) growth factors 4.8 M62424 2149 coagulation factor II (thrombin) receptor intracellular transducers 3.5 S40706 1649 DNA-damage-inducible transcript 3 other apoptosis-associated proteins 3.3 M81757 6223 ribosomal protein S19 other cell cycle proteins 2.8 J00123 5179 proenkephalin neuropeptides 2.8 M15395 3689 integrin, beta 2 major histocompatibility complex 2.8 U32944 8655 dynein, cytoplasmic, light polypeptide other apoptosis-associated proteins 2.6 D15057 1603 defender against cell death 1 other apoptosis-associated proteins Ratio shows average of 3 samples Real-time PCR Real-time PCR (Taqman analysis) was used to verify the changes in expression of certain candidate genes that are highly expressed in the array. Five samples were used for this study. WFDC2 and MMP7, which are both strongly expressed in the epithelial areas and decorin and TIMP1, which are both strongly expressed in the stromal areas by the cDNA array were chosen for verification. Each value was corrected for differences in loading relative to GAPDH mRNA expression. WFDC2 and MMP7 mRNA expression increased by 9.4- and 9.7-fold, respectively, compared to that of stromal cells. Decorin and TIMP1 mRNA expression increased by 16.3-and 7.2-fold, respectively, in stromal cells compared to that of epithelial cells. Statistically significant changes in expression of these genes were observed (p < 0.05) (Fig 2A,2B,2C and 2D ). Figure 2 A: WFDC2 mRNA expression in the secretory phase of the endometrium was determined by real-time PCR (n = 5). Values were normalized to GAPDH mRNA expression. Epithelial areas WFDC2 mRNA expression is shown relative to that of the stromal areas. The mean change is 19.4-fold. Statistical analysis was carried out using Wilcoxon test. B: MMP7 mRNA expression in the secretory phase of the endometrium was determined by real-time PCR (n = 5). Values are normalized to GAPDH mRNA expression. Epithelial areas MMP7 mRNA expression is shown relative to that of the stromal areas. The mean change is 9.7-fold. Statistical analysis was carried out using Wilcoxon test. C: Decorin mRNA expression in the secretory phase of the endometrium was determined by real-time PCR (n = 5). Values were normalized to GAPDH mRNA expression for each sample. Stromal areas decorin mRNA expression is shown relative to that of the epithelial areas. The mean change is 16.3-fold. Statistical analysis was carried out using Wilcoxon test. D: TIMP1 mRNA expression in the secretory phase of the endometrium was determined by real-time PCR (n = 5). Values are normalized to GAPDH mRNA expression for each sample. Stromal areas TIMP1 mRNA expression is shown relative to that of the epithelial areas. The mean change is 7.2-fold. Statistical analysis was carried out using Wilcoxon test. Discussion To date, various methods have been used for understanding the function of the endometrium. It is a well-known fact that epithelial cells and stromal cells in the endometrium play specific roles and are influenced by steroid hormones. However, it is very difficult to understand the molecular composition of each cell type as a function of time during the menstrual cycle. One of the problems of a cell culture experiment is that separation cultivation makes changes the composition of the cells. This is especially true as these cells are influenced by neighboring cells in vivo . It has recently become possible to acquire the information about the cell by the microdissection method. In this study, laser microdissection was used to isolate epithelial and stromal areas from the human endometrium. RNA was amplified by PCR and global gene expression was demonstrated by cDNA macroarray. Twenty-eight genes were identified in this study. These constitute only 2.8% of the 1000 genes on the array. Although this seems to be a small number, these genes were expressed at least 2.5-fold greater in all three samples and normalized to two house keeping genes. A similar percentage (1.2–5.8%) of genes with differential expression were reported using array analysis [ 17 - 19 ]. However, included genes below the 2.5-fold that we established as a criterion for inclusion in this study should be considered. Recently, some papers focused on endometrial gene expression have been reported. However, lots of them were compared between phases in the menstrual cycle. While Okulicz et al. demonstrated a difference in the gene expression between cell compartments in the monkey endometrium, the genes they identified are not the same as ours [ 1 , 2 ]. One of the reasons for this is because they tried to find new genes using differential display RT-PCR. Fifteen of 1000 genes were strongly expressed in the epithelial areas compared to the stromal areas. Of these, WFDC2 and MMP 7 were strongly expressed in the epithelial areas as confirmed by real-time PCR. WFDC2 was originally described as an epididymis-specific protein is expressed in a number of normal human tissues. A possible role for this gene in sperm maturation is indicated by amino acid similarities to extracellular proteinase inhibitors of genital tract mucous secretions [ 20 ]. Although WFDC2 has been recently reported in the secretory endometrium of monkey, this is the first report of its localization in the epithelium of the human endometrium [ 21 ]. However, the physiological role of this gene in the endometrium is presently unknown. Baboon endometrial epithelia express MMP 7 was reported by Cox et al [ 22 ]. The highest expression of MMP 7 occurred on day 7 of pregnancy in the rat uterus [ 23 , 24 ] and has been reported to have close associations with tumor invasion and metastasis [ 25 , 26 ]. HOXB gene induction is related to the immune system, and is specifically associated with IL-2-induced NK cell proliferation [ 27 , 28 ]. Although Hox-7 is reported to be in human cervical tumor tissue [ 29 ], this is the first report or its localization in human endometrium. Msh genes play a role in the regulation of cell-cell adhesion [ 30 ]. Friedmann et al. reported that regulated expression of homeobox genes Msx-1 and Msx-2 in mouse mammary gland development suggests a role in hormone action and epithelial-stromal interactions [ 31 ]. PKC theta cooperates with Vav1 to induce JNK activity in T-cells [ 32 , 33 ]. Take Catalano et al. report that JNK pathways are altered by RU486 which is an antiprogestins. PKC therefore seems to be important factor for control secretory endometrium [ 17 ]. Of the strongly expressed genes in the stromal aeas, decorin and TIMP1 gene expression were verified by real-time PCR. San Martin et al. reported the expression of decorin which is a leucine-rich proteoglycan in the mouse uterine and suggested it localized in the undifferentiated interimplantation site stroma [ 34 ]. Some reports also demonstrated its presence in the human uterin cervix and myometrium, but not in human endometrium [ 35 , 36 ]. DDR2 is a new type of receptor tyrosine kinases, and is thought to be involved in the metastasis of some tumors. Its ligand is fibrillar collagen, which suggests a role in controlling celluar responses to the extracellular matrix [ 37 ]. The changes in the extracellular matrix may play an important role in implantation, in invasion of trophoblastic cells and in the maintenance of pregnancy [ 38 ]. Decidualized stromal cells stained strongly positive for TIMP-1 [ 39 ]. Matrix metalloproteinases and their endogenous inhibitors, tissue-specific inhibitors of matrix metalloproteinases, play key roles in the cyclic remodeling events that occur in the human endometrium in preparation for pregnancy [ 40 ]. Ribosomal protein S3A is through direct or indirect actions on B and T cells and cytokine secretion, could participate in the immunoregulatory processes that play a role in the balance of the Th1 and Th2 immune response [ 41 ]. It has been recently said that the ratio of Th1 to Th2 influences pregnancy. Ribosomal protein S3A may be an interesting factor for stromal cell research. In this study, these results show the fact that many of the genes, which are related to the immune system, are expressed in the endometrium during the mid secretory phase of the menstrual cycle. This time, however, differences in gene expression between cell compartments of the endometrium were considered. Interactions among cells are key factors in understanding endometrial function.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC520833.xml
533873	Functional characterization in Caenorhabditis elegans of transmembrane worm-human orthologs	Background The complete genome sequences for human and the nematode Caenorhabditis elegans offer an opportunity to learn more about human gene function through functional characterization of orthologs in the worm. Based on a previous genome-wide analysis of worm-human orthologous transmembrane proteins, we selected seventeen genes to explore experimentally in C. elegans . These genes were selected on the basis that they all have high confidence candidate human orthologs and that their function is unknown. We first analyzed their phylogeny, membrane topology and domain organization. Then gene functions were studied experimentally in the worm by using RNA interference and transcriptional gfp reporter gene fusions. Results The experiments gave functional insights for twelve of the genes studied. For example, C36B1.12, the worm ortholog of three presenilin-like genes, was almost exclusively expressed in head neurons, suggesting an ancient conserved role important to neuronal function. We propose a new transmembrane topology for the presenilin-like protein family. sft-4 , the worm ortholog of surfeit locus gene Surf-4, proved to be an essential gene required for development during the larval stages of the worm. R155.1, whose human ortholog is entirely uncharacterized, was implicated in body size control and other developmental processes. Conclusions By combining bioinformatics and C. elegans experiments on orthologs, we provide functional insights on twelve previously uncharacterized human genes.	Background The nematode Caenorhabditis elegans has been used as a simple model for understanding animal biology for nearly four decades. After the sequencing of entire genomes from several metazoans, we are now in an excellent position to take a gene-centric approach to the worm as a model organism. A majority of human genes have homologs in C. elegans . In a comparative proteomics study, 83% of the worm proteome was found to have human homologous genes [ 1 ]. Only 11% or less contains nematode specific genes. This makes the worm a suitable model organism for delineating human gene function [ 2 - 4 ]. In a previous study, all transmembrane protein families in the C. elegans genome were classified and the human orthologs identified [ 5 ]. Predicted proteins with two or more membrane domains were clustered and for each cluster a multiple alignment was created. From the alignments, HMMs (Hidden Markov Models) were built and subsequently used to search for mammalian homologs. The consensus of nine different phylogenetic methods and BLAST were used to assign orthology. This resulted in a total of 174 worm-human orthology assignments with a high confidence. Orthologs are sequences that arose from a common ancestor gene and were separated by a speciation event [ 6 ]. Identification of orthologs is important, since they might share functionality. In closely related species, such as human and mouse, orthologs are normally trivial to find. However, when comparing distantly related species, e.g. human and worm, this is no longer the case because the similarity levels overall are low. Instead, one needs to rely on sophisticated phylogenetic reconstruction techniques to infer whether two genes stem from a node that corresponds to a speciation split or to a duplication event within a lineage. Close orthologs are likely to have the same biological role in the two organisms. Distant orthologs on the other hand, are less likely to have the same phenotypical role, but may have the same role in the corresponding pathway. Consequently, by studying true C. elegans orthologs to human genes experimentally in the worm, one can potentially learn more about the gene function also in humans. Depending on whether duplication(s) have occurred in one or both lineages since the speciation event, orthologs can form one-to-one, one-to-many or many-to-many relationships. Paralogs arise from a duplication event. A common scenario when genes are duplicated is that one of the gene copies is under negative selective pressure and therefore retains the function of the ancestor. The other copy might then be more free to evolve a new function different from the ancestral function. This is the reason why paralogs in different species are less likely to share functionality compared to orthologs. Paralogs can be divided into two subtypes – outparalogs and inparalogs [ 7 ]. Outparalogs are paralogs that evolved by gene duplications that happened before the speciation event and therefore they do not form orthologous relationships. Inparalogs, on the other hand, form co-orthologous relationships, since they are paralogs that evolved by gene duplications that happened after the speciation event. Here we present an initial functional characterization in C. elegans of seventeen genes. The criteria for selecting these genes were that they are high confidence candidate orthologs to human genes [ 5 ] and that their function is unknown. They are all predicted to encode transmembrane proteins, which imply that they could constitute as yet unknown receptors, channels or transporters playing important roles in various biological processes in multicellular organisms. We are particularly interested in studying those genes that might have a neuronal function. The phylogeny, membrane topology and domain organization were analyzed. Gene function was explored experimentally in the worm by means of RNA interference induced knock-down phenotypes and gene expression patterns. Results Membrane topology predictions The consensus of nine different methods was used to predict membrane topologies for the putative C. elegans proteins, and two different methods were used to predict signal peptides (see Methods for details). Each predicting method has some margin of error; therefore the consensus from several different predictors is more likely to give a better estimate of the true topology. Results were viewed using the SFINX tool [ 8 , 9 ], an example of output can be seen in Fig. 1 . The number of transmembrane (TM) regions ranges between six and ten (except for one of the splice variants of R155.1), with a majority of proteins having six or seven TM regions (see Table 1 ). Such proteins are likely to be receptors, channels, or transporters. One case, however, (C36B1.12) is likely to be an intramembrane protease. Figure 1 Output from SFINX for (A) C. elegans protein C36B1.12 and (B) one of its assigned human orthologs Q8TCT8. The overall membrane topology of the two proteins is very similar. The consensus from the different topology predictors is nine transmembrane (TM) regions, a N-terminal signal peptide, and a >150 amino acids non-cytoplasmic N-terminal region. Conserved aspartic acid residues are marked with an asterisk (residues 433 and 516 for C36B1.12; residues 351 and 412 for Q8TCT8). Phobius is the only program used that predicts both TM regions and N-terminal signal peptides. The other programs are designed to only detect TM regions, and therefore, they commonly mistake the signal peptide for a TM region. Numbers on the horizontal axis indicate amino acid positions. Color coding: black = TM region, white = non-cytoplasmic, gray = cytoplasmic, striped = N-terminal signal peptide predicted by Phobius. Table 1 Description of C. elegans – human orthologs. The Swiss-Prot accession numbers are given for the worm sequences and their human orthologs. TM: the number of transmembrane regions predicted using the consensus of nine different methods in the SFINX tool. Bootstrap support (%) is given for the inferred speciation node in the phylogenetic tree constructed using PHYLOWIN with PAM distances. Identity (%): sequence percentage identity from the Blastp output between the C. elegans gene and the nearest human ortholog. Four of the worm genes are predicted to have two splice variants (a and b). However, there are only small differences in the protein sequences between the two variants, except for R155.1. Splice variant R155.1b is predicted to have a truncation of more than one hundred amino acids in the N-terminus compared to R155.1a. The PF03062 domain (MBOAT) is still present in both splice variants. The protein sequence for R155.1a was used in the phylogenetic analysis. C. elegans orthologs TM Pfam-A domains Human orthologs Bootstrap support (%) Identity (%) Nearest human ortholog with putative function C30H6.2 (Q9XVR4) 6 PF02535 (ZIP) Q9H6T8, Q9NXC4, Q96NN4 57 50 Q9H6T8: SLC39A4, involved in intestinal absorption of zinc. T11F9.2a (Q8I4G0), T11F9.2b (Q22395) 7 PF02535 (ZIP) Q15043, Q96SM9, Q9C0K1, Q96BB3 91 30 Q9C0K1: BIGM103, involved in intracellular zinc retention and accumulation. H13N06.5 (Q9XTQ7) 9 PF02535 (ZIP) Q92504 98 52 Transport of zinc out of ER 2 and other intracellular stores. T28F3.3 1 (Q9XUC4) 7* PF02535 (ZIP) Q92504 79 38 Transport of zinc out of ER 2 and other intracellular stores. T01D3.5 (Q9XVJ5) 8 PF02535 (ZIP) Q9NUM3 99 38 Function unknown. F40F9.1a (Q8MQ56), F40F9.1b (Q8MQ55) 7 PF01027 (UPF0005) Q9BWQ8, Q969X1 70 39 Q9BWQ8: Lifeguard protein, protects cells from Fas-mediated cell death. F40F9.2 (Q20241) 7 PF01027 (UPF0005) Q9BWQ8, Q969X1 70 40 Q9BWQ8: Lifeguard protein, protects cells from Fas-mediated cell death. F08F1.7 (O17388) 9* PF02990 (EMP70) Q99805 100 62 Endosomal integral membrane protein. ZK858.6a (Q94422) 9 PF02990 (EMP70) Q92544 100 53 Function unknown. ZK858.6b (Q7YTF9) 9* F14F3.3 (Q19468) 9 PF03062 (MBOAT) Q96N66, Q99908 100 27 Q99908: BB1 protein, malignant cell expression-enhanced gene. R155.1a (O01925) 8 PF03062 (MBOAT) Q92980 100 32 Function unknown. R155.1b (Q86DC4) 4 C36B1.12 (Q93346) 9* PF04258 (Peptidase_A22B) Q8TCT7, Q8TCT8, Q8IUH8 100 30 Q8TCT7: Presenilin-like protein, may act as intra-membrane protease. sft-4 (Q18864) 7 PF02077 (SURF4) O15260 100 55 Surfeit locus protein 4, probable ER 2 integral membrane protein. D2013.10 (O62126) 6 None Q15055 98 53 Function unknown. T04A8.12 (Q22141) 6 None Q9UHJ9 50 35 FRAG1 (FGFR (fibroblast growth factor receptor) activating gene 1). Y6B3B.10 (Q9XWE9) 6 PF03798 (LAG1) P27544 99 37 Function unknown. ZK721.1 (Q9GYF0) 10* None Q9NXL6, Q9Y357 97 34 Function unknown. 1 T28F3.3 was initially included because of strong similarity to Q92504; however, the phylogenetic analysis showed that it is probably an outparalog to the human gene. 2 ER = endoplasmic reticulum. * predicted to have a N-terminal signal peptide. Phylogenetic analysis The results from the phylogenetic analysis are presented in Table 1 . The previous orthology assignments are still valid [ 5 ], although for some C. elegans genes, additional human orthologs have emerged from the sequencing efforts. At present, 29% (5 of 17) of the worm genes have one-to-many ortholog relationship with human genes, which means that there has probably been an expansion in the human lineage. This is the case for C36B1.12 (Q93346) and ZK721.1 (Q9GYF0) (see Fig. 2A and 2D , respectively). 53% (9 of 17) showed a one-to-one relationship with its human ortholog, for example sft-4 (Q18864) (see Fig. 2C ). Two of the worm genes, F40F9.1 (Q8MQ55, Q8MQ56) and F40F9.2 (Q20241), seem to have a many-to-many relationship (see Fig. 2B ). The phylogenetic tree in Fig. 2B show somewhat inconclusive support for where the speciation event could have taken place. It is possible that the human gene Q8IVW7 is also an ortholog to the two worm genes. To investigate the ortholog relationship further, Orthostrapper was used [ 10 ]. Orthostrapper analyzes a set of bootstrap trees instead of the optimal tree for orthologs. The algorithm detects orthologous relations between two (groups of) species. The frequency of orthology assignments in the bootstrap trees can be interpreted as a confidence value for the possible orthology of two proteins. Orthology assignments in the optimal phylogenetic tree that might be incorrect can be identified by their low ortholog bootstrap value. This makes it possible to resolve complicated many-to-many orthologous relationships. When analyzing the multiple alignment for the phylogenetic tree in Fig. 2B using Orthostrapper, the results showed a stronger support for the human genes Q9BWQ8 and Q969X1 to be the orthologs compared to Q8IVW7 (65% vs. 23%). It seems that the orthologous relationship in this particular case is complicated to elucidate. Still, Q9BWQ8 and Q969X1 are the best candidate human orthologs for the C. elegans genes F40F9.1 and F40F9.2. Figure 2 Phylogenetic trees for genes (A) C36B1.12 (Q93346), (B) F40F9.1a (Q8MQ56), F40F9.1b (Q8MQ55) and F40F9.2 (Q20241), (C) sft-4 (Q18864), and (D) ZK721.1 (Q9GYF0). The trees were constructed using PHYLOWIN with neighbor-joining method and PAM distances. 500 bootstrap replicates were run. All gene identifiers are Swiss-Prot accession numbers, except in (D) where XP_148505 is the NCBI accession number. The C. elegans genes studied here are marked with asterisks. F40F9.1 is predicted to have two splice variants; however, the putative proteins have the same length and only differ in the two most C-terminal amino acids. Species abbreviations: Arabidopsis thaliana (AT), Caenorhabditis elegans (CE), Drosophila melanogaster (DM), Fugu rubripes (FR), Homo sapiens (HS), Mus musculus (MM), Oryza sativa (OS), Rattus norvegicus (RN), Saccharomyces cerevisiae (SC), Schizosaccharomyces pombe (SP), Xenopus laevis (XL). H13N06.5 and T28F3.3 both show sequence similarity to the human gene Q92504. However, the phylogenetic tree and results from Orthostrapper (data not shown), suggest that H13N06.5 is the putative ortholog to Q92504, whereas T28F3.3 may be an outparalog. The bootstrap support for the speciation node between T04A8.12 and its human ortholog Q9UHJ9 barely made the cutoff of 50% when PAM distances was used. With observed divergence and Poisson correction as distance methods, the bootstrap support improved to 99% and 67%, respectively. When analyzing the phylogenetic tree using Orthostrappper, there was a very strong support for this orthology assignment (94%). Therefore, we conclude that T04A8.12 is probably the ortholog to Q9UHJ9. C30H6.2 has three potential human orthologs, Q9H6T8, Q9NXC4 and Q96NN4. The bootstrap support for the speciation node with PAM distances was 57%. This improved to 87% and 88% with observed divergence and Poisson correction, respectively. Orthostrapper results showed a strong support for Q9H6T8 and Q9NXC4 as orthologs to C30H6.2 (84%), whereas the support for Q96NN4 was weaker (57%). Considering these results, we believe that all three human genes are orthologs to the worm gene; however, the ortholog relationship seems to be weaker between Q96NN4 and C30H6.2. Putative domain assignments The domain organization of the predicted proteins was analyzed using the Pfam database [ 11 , 12 ] (see Table 1 ). Conclusions about possible functions cannot be drawn from the mere presence of a putative domain, although it can give some indication. Five of the proteins (C30H6.2, T11F9.2, H13N06.5, T28F3.3 and T01D3.5) may have a PF02535 domain, which is annotated as being a ZIP domain. The ZIP family is believed to include zinc and other metal transporters. The ZIP proteins have been classified into four groups based on sequence conservation [ 13 ]; the ZIP subfamily I and II, the gufA subfamily and the LIV-1 subfamily (also called the LZT subfamily). The ZIP I subfamily contains mostly plant and yeast sequences; however, it also includes T01D3.5 and its putative orthologs in Drosophila melanogaster , mouse and human (Q9V4C6, Q8BFU1 and Q9NUM3, respectively). The other four worm genes appear to belong to the LIV-1 subfamily. This subfamily has a unique metalloprotease motif that raises the possibility that they might have protease activity [ 14 ]. Within the LIV-1 subfamily there is a subgroup called the KE4 group, to which H13N06.5 and its human ortholog hKE4 (Q92504) belong. C36B1.12 was predicted to have a PF04258 domain, a probable signal peptide peptidase (SPP) domain. SPP catalyzes intramembrane proteolysis of some signal peptides after they have been cleaved from a preprotein. This processing by SPP is related to protein cleavage by presenilins. Homologs to SPP are divided into five subfamilies based on phylogenetic analysis (subfamily SPP and subfamilies SPPL1-4, for SPP like) [ 15 ]. C36B1.12 and its putative human orthologs Q8TCT7, Q8TCT8 and Q8IUH8 belong to the SPPL2 subfamily. The members of subfamilies SPPL1-4 only show homology to SPP in the C-terminal half of the protein and in the N-terminus there is substantial variation. This suggests that the C-terminal part may constitute the proteolytic subdomain, whereas the N-terminus defines the specific function of the respective proteins. F40F9.1 and F40F9.2 seem to have a PF01027 domain (UPF0005), which is an uncharacterized protein family. Both F08F1.7 and ZK858.6 may belong to the PF02990 domain family (EMP70). Proteins in this family might be located to endosomal membranes [ 16 ]. F14F3.3 and R155.1 were predicted to have a PF03062 domain, which is annotated as a MBOAT (Membrane bound O-acyl transferases) domain. Biochemically characterized proteins of this group encode enzymes that transfer organic acids onto hydroxyl groups of membrane-embedded targets [ 17 ]. SFT-4 most likely has a PF02077 (SURF4) domain. Members of this family are believed to encode integral membrane proteins located to the endoplasmic reticulum [ 18 ]. A PF03798 domain (LAG1) was found in Y6B3B.10. This domain is associated with longevity in yeast (Jiang et al. 1998). Three of the seventeen putative proteins (D2013.10, T04A8.12 and ZK721.1) do not match to any Pfam-A domain. RNA interference studies Out of the seventeen genes studied, sft-4 and R155.1 exhibited phenotypes when both the N2 (wildtype) and the RNAi sensitive rrf-3(pk1426) II [ 19 , 20 ] strains were subjected to RNAi by feeding (see Table 2 ). The phenotypes were enhanced with strain rrf-3 , although the Dpy (dumpy) phenotype seen for R155.1 was still low penetrant and relatively weak. The Lva (larval arrest) observed for sft-4 occurred at larval stages L2–L3 and there was an almost complete penetrance with the sensitive strain. The RNAi phenotypes for both genes were detected at all temperatures, although, for sft-4 they were more severe at higher temperatures. The positive results were verified with RNAi by injection in strain N2. Table 2 RNAi phenotype and major tissues of gene expression for C. elegans orthologs. Abbreviations: RNAi phenotype: clear (Clr), dumpy (Dpy), larval arrest (Lva), ruptured (Rup), sterile (Ste), wildtype (WT). Gene expression: body wall muscle (bwm), commissures (c), excretory system (exc), gonad (g), hypodermis (h), hypodermal seam cells (hs), intestine (i), neuronal (n), pharyngeal muscle (phm), rectal epithelial cells (re), spermatheca (s), vulva (v), ventral nerve cord (vnc). A limitation when extrachromosomal array transgenes are used is that expression in the germ line is not possible to evaluate. No transgenic lines could be generated for T11F9.2, H13N06.5 and T04A8.12. Possible reasons for this could be that the injected DNA concentration was too low or that the sequence was toxic. In either case, the extrachromosomal array formed may not have been sufficiently large to be inheritable [46]. The transgenic lines for ZK858.6 and F14F3.3 showed no expression of gfp . This might be caused by conditional gene expression, germline silencing or absence of the promoter:: gfp fusion from the inheritable extrachromosomal array [46]. C. elegans orthologs RNAi phenotype Major tissues of gene expression C30H6.2 WT h, phm T11F9.2a, T11F9.2b WT No transgenic line H13N06.5 WT No transgenic line T28F3.3 WT h, i, n, v, vnc T01D3.5 WT hs F40F9.1a, F40F9.1b WT bwm, c, h, n, phm, vnc F40F9.2 WT exc, n, phm F08F1.7 WT h, n, phm, re, s, v ZK858.6a, ZK858.6b WT No expression F14F3.3 WT No expression R155.1a, R155.1b Dpy bwm, h, i, phm C36B1.12 WT i, n sft-4 Clr, Lva, Rup, Ste bwm, h, i, n, phm, v D2013.10 WT bwm, h, i, n, s, v T04A8.12 WT No transgenic line Y6B3B.10 WT phm ZK721.1 WT bwm, g, h, i, n, phm, s, v Analysis of gene expression Transcriptional fusions with gfp were established for fourteen genes and the resulting gene expression was analyzed. The results are presented in Table 2 . Because the arrays are extrachromosomal and not integrated; mosaic patterns of expression were observed. Also, germ line expression could not be analyzed, due to germ line silencing. For 18% (3 of 17) of the genes no transgenic lines could be established despite several attempts, and out of the lines established, 14% (2 of 14) showed no expression. This could be due to several reasons (see Discussion). In half of the transgenic lines established, expression was found in more than three different tissues. The most prevalent major tissues of expression were hypodermis (9 of 14 transgenic lines), nervous system and pharyngeal muscle (8 of 14) and intestine (6 of 14). Examples of gene expression patterns observed are presented in Fig. 3 , 4 , 5 , 6 . C36B1.12 shows expression restricted to head neurons and intestine (see Fig. 3 ). The intestinal expression was stronger during larval stages compared to the adult stage, and it was predominantly located to posterior intestinal nuclei. F40F9.1 and F40F9.2 demonstrate some overlapping expression in nervous system and pharyngeal muscle (see Fig. 4A,4G ); however, F40F9.1 appear to be more widely expressed in the nervous system with expression in more neuronal cell bodies and in commissures and ventral nerve cord (see Fig. 4C ). Expression of F40F9.1 is also located to body wall muscle and hypodermal cells in the tail (see Fig. 4E ). F40F9.2 is also expressed in the excretory system, although it is weaker compared to the other tissues (see Fig. 4G ). Widespread expression patterns were observed for sft-4 and ZK721.1 both during larval and adult stages (see Fig. 5 and 6 , respectively). For sft-4 it was highly mosaic with pharyngeal muscle as the most consistent tissue of expression. Figure 3 Major tissues of expression for C36B1.12. (A) Fluorescence micrograph of an L4 larvae hermaphrodite carrying a transcriptional fusion between gfp and a putative promoter of C36B1.12 expressed in neurons in the head. (C) Fluorescence micrograph of a young adult hermaphrodite carrying the same construct expressed in intestine. The observed intestinal expression is mostly located to posterior intestinal nuclei and is more prominent in younger worms. (B and D) Corresponding DIC images. Scale bars, 20 μm. Figure 4 Major tissues of expression for F40F9.1 and F40F9.2. Fluorescence micrographs of an adult hermaphrodite carrying a transcriptional fusion between gfp and a putative promoter of F40F9.1 expressed in (A) neurons and pharyngeal muscle, (C) commissures (c) and the ventral nerve cord (vnc), and (E) body wall muscle (bwm) and hypodermis (h). (G) Fluorescence micrograph of an L4 hermaphrodite carrying a transcriptional fusion between gfp and a putative promoter of F40F9.2 expressed in the excretory system (exc), neurons, and pharyngeal muscle. (B, D, F, and H) Corresponding DIC images. Scale bars, 20 μm. Figure 5 Major tissues of expression for sft-4 . Fluorescence micrographs of an adult hermaphrodite carrying a transcriptional fusion between gfp and a putative promoter of sft-4 expressed in (A) pharyngeal muscle, (C) vulva and (E) intestinal nuclei (i). The intestine shows some unspecific autofluorescence, but there is also specific expression in the intestinal nuclei. Expression in body wall muscle, hypodermis and neurons is not shown. (B, D, and F) Corresponding DIC images. Scale bars, 20 μm. Figure 6 Major tissues of expression for ZK721.1. Fluorescence micrographs of an adult hermaphrodite carrying a transcriptional fusion between gfp and a putative promoter of ZK721.1 expressed in (A) body wall muscle, (C) hypodermis and (E) gonad. Gene expression in hypodermis is weaker compared to expression in other tissues. Expression in intestine, neurons, pharyngeal muscle, spermatheca, and vulva is not shown. (B, D and F) Corresponding DIC images. Scale bars, 20 μm. Putative function assignments C36B1.12 The three putative human orthologs (Q8TCT7, Q8TCT8 and Q8IUH8) to the C. elegans gene C36B1.12 are thought to be presenilin-like (PSL) proteins (also called PSH proteins for presenilin homologs). Presenilins are an important group of proteases acting in the nervous system. Abnormal proteolytic cleavage may result in accumulation of pathogenic insoluble proteins, implied in e.g. Alzheimer's disease. We have shown that C36B1.12 is probably expressed in head neurons and intestine; however, the intestinal expression might be ectopic (see Discussion for details), which would imply that the gene is exclusively expressed in neurons. This suggests that the three human orthologs may also encode neuronal functions. The membrane topology of human proteins belonging to the presenilin-like family has been analyzed previously. Q8IUH8 was predicted to have seven transmembrane (TM) segments and a cytoplasmic C-terminus [ 21 ], and the same was predicted for HM13_HUMAN (Swissprot: Q8TCT9) [ 15 ]. However, it should be noted that although the number of TM segments of these predictions is the same, the topologies are in fact different. The fourth segment in the Q8IUH8 prediction is missing from the HM13_HUMAN prediction, and the C-terminal segment in the HM13_HUMAN prediction is missing from the Q8IUH8 prediction. This means that the four C-terminal TM segments are not in register between the predictions, and consequently the loops are on opposite sides. This includes the loop between the putatively catalytic aspartic acid residues also present in presenilins that was predicted cytoplasmic by Ponting et al., and non-cytoplasmic by Weihofen et al. Because of the TM segment disagreement, these aspartic acid residues were predicted to be located in TM5 and TM6 in the Ponting et al. prediction, but in TM4 and TM5 in the Weihofen et al. prediction. Merging these two proposed topologies by accepting all TM segments predicted by one or the other study would yield a topology with nine TM segments. Our own analysis of the proteins in question using the SFINX tool [ 8 , 9 ] provides strong support for this topology, with the C-terminus in the cytoplasm (data not shown). We therefore propose that both previous TM topologies had incorrectly left out one TM segment, which would correspond to segments 4 and 9 in the 9-TM segment model. We further analyzed the other members of this family with SFINX [ 8 , 9 ], and consistently found a 9-TM topology model with C-terminus in the cytoplasm. The conserved aspartic acid residues would be located in TM6 and TM7. As an example, the SFINX output for C36B1.12 and one of its human orthologs Q8TCT8 is shown in Fig. 1 . The overall topology is very similar between the putative worm protein and all of its orthologs in both human and mouse, as well as the other presenilin-like proteins. One difference, however, is that C36B1.12 and two of its human (Q8TCT8 and Q8IUH8) and mouse orthologs (Q9JJF9 and Q8BHP0) are predicted to have a N-terminal signal peptide, a feature that seems to be missing from the other presenilin-like proteins. F40F9.1 and F40F9.2 F40F9.1 and F40F9.2 are 48% identical to each other on the protein sequence level and they also appear to have similar membrane topologies. They are close in the genome (<100 bp apart), but on opposite strands. It has been shown that genes closer than 500 bp on opposite strands are likely to have a shared control region [ 22 ], which means that these two genes might be coexpressed. The expression patterns observed are indeed overlapping, although not to a full extent (see Table 2 ). One of the human orthologs (Q9BWQ8) identified has been shown to protect cells from Fas-mediated cell death [ 23 ], suggesting that F40F9.1 and F40F9.2 might be involved in apoptosis. sft-4 The sft-4 gene (C54H2.5) is highly conserved throughout evolution with orthologs in both vertebrates and non-vertebrates. All orthologous relationships are one-to-one with a high bootstrap support. The tree in Fig. 2C indicates that there exists a worm homolog (O45731) to sft-4 . O45731 (T02E1.7) was found to be 33% identical to sft-4 on the protein sequence level and it was also predicted to have a PF02077 (SURF4) domain. However, our RNAi screen indicates that there is no or little functional redundancy between the two genes, since sft-4 has a very strong RNAi phenotype, showing an almost complete larval arrest at stages L2–L3. The RNAi phenotype for T02E1.7 is wildtype according to previous studies [ 24 ]. Our gene expression analysis revealed a wide spread expression of the reporter construct for sft-4 (see Table 2 ). Taken together, these data suggests that sft-4 may play an essential role during development acting in many tissues. ZK721.1 ZK721.1 is most probably orthologous to the human genes Q9NXL6 and Q9Y357 (CGI-40 protein). The CGI-40 protein was found in a screen where novel human genes evolutionary conserved in C. elegans were identified [ 1 ]. The function of both CGI-40 and Q9NXL6 is unknown. ZK721.1 has several worm homologs, one of which is sid-1 (Q9GZC8). SID-1 has been identified as a protein that is required for systemic RNAi [ 25 , 26 ]. It was predicted to have eleven transmembrane (TM) regions and some of them have been experimentally verified [ 27 ]. The high number of TM regions suggests that SID-1 forms a channel. Double stranded RNA is thought to diffuse through this channel, leading to spreading of the RNA and hence, a systemic RNAi effect. This idea is also supported by the fact that no homolog of sid-1 has been found in Drosophila , which can explain the observed absence of systemic RNAi in this organism [ 28 , 29 ]. Our analysis of ZK721.1 predicts that it has ten TM regions, which makes it a likely candidate for forming a channel. The phylogenetic tree indicates that ZK721.1 is the best candidate ortholog to human genes Q9NXL6 and Q9Y357, whereas sid-1 is a probable outparalog to the human genes (see Fig. 2D ). The tree also supports the previous finding that there is no homolog to ZK721.1 or sid-1 in the Drosophila genome. We have observed a wildtype RNAi phenotype for ZK721.1, which is consistent with results from other studies [ 30 , 31 ]. Further analysis might reveal if ZK721.1 also is involved in systemic RNAi or if it has some other function. Four additional genes required for systemic RNAi have been reported ( rsd-2 , -3 , -4 and -6 ) [ 26 ], but none of them map to locus ZK721.1. C30H6.2, H13N06.5, T01D3.5, T11F9.2 and T28F3.3 SLC39A4 (Q9H6T8) was identified as one of three possible human orthologs to C30H6.2. The human gene has been implicated in the rare inherited condition acrodermatitis enteropathica, which results from a defect in the absorption of zinc [ 32 ]. It is believed that SLC39A4 might encode a zinc transporter responsible for intestinal absorption of zinc. Therefore, C30H6.2 may also be a zinc/metal transporter. The predicted human ortholog (Q92504) to H13N06.5 in C. elegans has been shown to be a zinc transporter localized to intracellular membranes [ 33 ]. Q92504 probably transports zinc out of the endoplasmic reticulum and other intracellular stores. The Drosophila ortholog to H13N06.5 is Catsup ( Catecholamines up , Q9V3A4), which encodes a negative regulator of tyrosine hydroxylase (TH) activity [ 34 ]. TH is a rate-limiting enzyme for production of dopamine in the brain. The Arabidopsis thaliana gene IAR1 (Q9M647) is also an ortholog to H13N06.5. It is proposed to be involved in auxin metabolism or response [ 35 ]. Interestingly, the mouse ortholog (Q31125) to H13N06.5 and IAR1 was shown to functionally substitute for the Arabidopsis gene. These data indicate that there is functional conservation among these orthologs and it is likely that H13N06.5, and possibly also T28F3.3, could play similar roles in the corresponding pathways in the worm. BIGM103 (Q9C0K1) was identified as one of four candidate human orthologs to T11F9.2. The human gene was found to be induced during the infection and inflammatory response. It was also shown to play a role in intracellular zinc ion accumulation and retention [ 36 ]. Consequently, it is possible that T11F9.2 might be an integral membrane zinc/metal transporter. The human ortholog to T01D3.5 has no known putative function. We observed a wildtype RNAi phenotype for T01D3.5 as well as for the other four PF02535 (ZIP) domain containing putative proteins (C30H6.2, H13N06.5, T11F9.2 and T28F3.3). This indicates that there might be some functional redundancy between these genes. Considering the information available, it is conceivable that T01D3.5 may also be a zinc/metal transporter. F08F1.7, T04A8.12 and ZK858.6 F08F1.7 and ZK858.6 show 47% identity on the protein sequence level and they have similar membrane topologies with a large N-terminal non-cytoplasmic region and nine transmembrane regions in the C-terminal part. F08F1.7 and one of the splice variants of ZK858.6 were predicted to have a N-terminal signal peptide. The phylogenetic analysis indicated that the human gene p76 (Q99805) is the ortholog to F08F1.7 and it appears to localize to endosomes [ 37 ]. The function of the probable human ortholog (Q92544) to ZK858.6 is unknown. Both F08F1.7 and ZK858.6 are predicted to be in operons, as is T04A8.12 [ 38 ]. F08F1.7 is probably in an operon with tth-1 (F08F1.8, O17389). TTH-1 is likely to belong to the PF01290 domain family (thymosin beta-4), which includes actin-binding proteins, implicating a possible role in cytoskeleton organization. ZK858.6 is predicted to be in an operon with ZK858.5 (Q94421) and ZK858.7 (Q94416). A PF05154 (TM2) domain with unknown function is likely to be present in ZK858.5. ZK858.7 might have a PF04189 domain (eIF3gamma; eukaryotic initiation factor 3, gamma subunit), suggesting that it could be involved in translation. T04A8.12 may be in an operon with T04A8.11 (Q22140) and T04A8.13 (Q22142). T04A8.11 might be a ribosomal protein, since it appears to have a PF00252 (ribosomal L16) domain. T04A8.13 was predicted to have a PF00646 (F-box) domain, which is known for forming structural complexes with other proteins. There are no matching Pfam-A domains for T04A8.12. The best candidate human ortholog to T04A8.12 is FRAG1 (fibroblast growth factor receptor activating gene 1, Q9UHJ9). FRAG1 seems to be ubiquitously expressed in adult human tissues and it has also been detected in several human tumor cell lines [ 39 ]. Our results from the gene expression analysis demonstrated that F08F1.7 is probably expressed in several tissues in C. elegans , suggesting an important biological role. Previous studies have shown that the human ortholog p76 (Q99805) is ubiquitously expressed [ 37 ]. There was no expression observed for ZK858.6 and for gene T04A8.12, we failed to generate a transgenic line. All three genes exhibited a wildtype RNAi phenotype. For F08F1.7 and ZK858.6, the wildtype phenotype could possibly be explained by functional redundancy between the two genes and a third C. elegans EMP70 domain containing protein Y41D4A.4 (Q95Y24). We found that Y41D4A.4 is homologous to F08F1.7 and ZK858.6, and most likely an ortholog to the human gene Q9HD45. Taken together, these findings point to that F08F1.7, ZK858.6 and T04A8.12 might play fundamental biological roles, and that they may be involved in processes such as cellular organization (F08F1.7) and translation (ZK858.6 and T04A8.12). F14F3.3 and R155.1 The human genes Q99908 (BB1) and Q96N66 are probable orthologs to F14F3.3. BB1 has been shown to be overexpressed in breast and bladder carcinoma [ 40 ], suggesting that it might have a role in tumor progression. The function of Q96N66 is unknown. The likely Drosophila ortholog to R155.1 is Nessy (Q9XYV9), a putative Hox gene target [ 41 ], indicating a possible developmental role. The Dpy (dumpy) RNAi phenotype detected for R155.1 could be due to some developmental/body size regulatory error in possibly the hypodermis and/or body wall muscle; tissues in which the gene is expressed according to our analysis. The best human ortholog candidate (Q92980) to R155.1 has not yet been functionally characterized. Y6B3B.10 Y6B3B.10 is most probably orthologous to the human gene P27544 and they both seem to belong to the PF03798 (LAG1) domain family. LAG1 is a longevity gene that was cloned from yeast [ 42 ]. Members of the LAG1 family are thought to be involved in determining lifespan. However, the phylogenetic tree revealed that Y6B3B.10 and its human and mouse ortholog (P27545) form a tight cluster in the tree, separated from the other LAG1 domain containing proteins, indicating that they may have evolved a slightly different function. Y6B3B.10 showed a wildtype RNAi phenotype and it appears to have an expression restricted to the pharyngeal muscle. D2013.10 D2013.10 is orthologous to Q15055 (human), Q8K1A5 (mouse) and Q9VX39 ( Drosophila ). Neither of these genes has any putative function assigned to them and they have no matching Pfam-A domains. D2013.10 is expressed in several tissues in C. elegans and it exhibits a wildtype RNAi phenotype. Discussion This study illustrates how bioinformatic and experimental analysis can be combined to elucidate putative gene function. We have predicted worm-human orthologs and performed an initial functional characterization of the worm genes. Since orthologs are likely to have the same biological function, a better understanding of the function of the human genes can be accomplished through analysis in C. elegans . The genes explored in this study were selected from a previous study [ 5 ] and they are all predicted to encode transmembrane proteins. Such proteins are attractive to study since many interesting receptors, channels, transporters and signaling proteins are found among them, making them likely to be involved in important regulatory processes in multicellular organisms. The number of transmembrane (TM) regions predicted for each protein, is similar to the number predicted for each cluster of putative TM proteins from the former study (± 1 TM region) [ 5 ]. For the putative proteins H13N06.5, F14F3.3 and ZK721.1, however, the difference is larger (+2-3 TM regions). This divergence could be due to Remm and Sonnhammer having performed predictions on a cluster and not on individual genes. In addition, they used only the program TMHMM [ 43 ] for analyzing membrane topology. TMHMM, when using default settings, may miss weak TM regions, leading to a possible underestimation of the number of TM segments. A better estimate of the true topology can be achieved through the use of several different prediction programs. In this study, we used the consensus of nine different methods provided by the SFINX tool [ 8 , 9 ] to assign membrane topology. We observed an RNAi phenotype for 11.8% (2 of 17) of the genes when using both strains N2 (wildtype) and rrf-3 (RNAi sensitive), respectively. This is in agreement with previous experiments, where 10.3% (N2) and 12.8% ( rrf-3 ) phenotypes have been detected [ 30 , 44 ]. The RNAi phenotypes for sft-4 are consistent with previous results [ 30 , 31 ]. However, the two groups have reported non-overlapping phenotypes, but in this screen we have observed all of them. The Lva (larval arrest) phenotype has also been reported from the genome wide screen using strain rrf-3 [ 44 ]. The Dpy (dumpy) phenotype for R155.1 has not been reported before. The gene was downregulated using RNAi by injection in a screen of chromosome III [ 45 ] and the phenotype was found to be wildtype. However, the focus of that analysis was to identify genes involved in cell division and therefore only a few post-embryonic phenotypes were scored. The Dpy phenotype observed is also low penetrant and relatively weak and could therefore be missed. Furthermore, differences in results from RNAi screens have been shown to exist. A 10–30% difference between experiments done in both different and in the same laboratories has been reported [ 44 ]. When analyzing expression patterns using transcriptional reporter fusions, one issue of concern is whether the pattern observed is the expression pattern of the native gene or not. Ectopic or lack of expression can occur if the putative promoter used does not include all the regulatory elements. Expression in several different cell types in the pharynx and in the posterior intestinal cells of young animals has been attributed to the use of incomplete promoters [ 46 ]. Another limitation when using extrachromosomal arrays is that analysis of expression in the germ line is not possible, due to germ line silencing. The putative promoter used in the transcriptional fusion for C36B1.12 is only 1 kb, due to the presence of an upstream gene. Therefore, it is possible that the intestinal expression seen predominantly in young worms and mostly located to posterior intestinal nuclei, is an artifact of the use of an incomplete promoter region [ 46 ]. If this is the case, C36B1.12 might be expressed exclusively in neurons in the head (see Fig. 3 ). This finding provides support to the idea that C36B1.12 and its three human orthologs encode neuronal functions. A possible consequence of this could be that they act in a fashion analogous to presenilin, or even that they could be involved in β-amyloid precursor protein processing. It would be interesting to study their role in nervous system development and function, and whether they are linked to neurological disorders The transcriptional fusion for T28F3.3 also showed a similar intestinal expression, apart from the specific expression in neurons in the head, ventral nerve cord, vulva and a weak expression in hypodermis. The putative promoter region used is only 0.8 kb, due to the presence of an upstream gene. Thus, the intestinal expression seen for T28F3.3 may once again be related to the use of an incomplete promoter region [ 46 ]. Two of the transcriptional fusions (for the genes F14F3.3 and ZK858.6) showed no expression of the reporter gene. This is unlikely due to the use of an incomplete promoter region, since the upstream region included was 2.9 kb and 3.0 kb, respectively. Instead, it might be caused by conditional gene expression, germline silencing or absence of the promoter:: gfp fusion from the inheritable extrachromosomal array [ 46 ]. For three of the genes in this study we failed to establish transgenic lines. Possible reasons for this could be that the injected DNA concentration was too low or that the sequence was toxic. In either case, the extrachromosomal array formed may not have been sufficiently large to be inheritable [ 46 ]. Out of the seventeen genes in this study, three are predicted to be in operons (18%). This is equivalent to the number of genes in the C. elegans genome that are believed to be in operons (15%). Whether C. elegans operons contain genes of related function or not is still unknown. There are, however, some indications that genes encoding proteins of fundamental biological importance might be clustered into operons. For example, genes for mitochondrial proteins have a strong tendency to be together in operons and also genes encoding splicing proteins [ 38 ]. Conclusions This study has shed some light upon the putative function of a few predicted worm-human orthologs. Our aim was to identify genes that could play a role in the nervous system and indeed we have been able to find eight genes that appear to be expressed in neurons. C. elegans is an excellent model organism for pursuing the functional characterization of these genes, considering its well mapped and relatively sophisticated nervous system. Investigating the function of orthologous proteins using a simple multicellular organism is a suitable approach for the possibility of learning more about the function of a gene not only in one species but also hopefully in several. This approach becomes even more valid as several genomes are being sequenced at the moment with additional ones already in the pipeline. With the enormous amount of data that these sequencing efforts are generating, it is very useful to be able to start delineating the gene function based on functional characterization of the ortholog in another species, before initiating studies in more complex organisms. Methods Membrane topology predictions The membrane topology was predicted with nine different methods, and the SFINX tool [ 8 , 9 ], was used to display the results. Eight membrane topology predictors were used: Phobius [ 47 ], TMHMM2.0 [ 48 ], TMHMM1.0 [ 43 ], PHDhtm [ 49 ], HMMTOP2.1 [ 50 ], HMMTOP1.0 [ 51 ], MEMSAT [ 52 ] and TOPPRED [ 53 ]. In addition, a Kyte-Doolittle hydrophobicity curve [ 54 ] was constructed for each putative protein sequence. Transmembrane regions were considered positive if they were predicted by a majority of the methods, or by four methods and having a supporting Kyte-Doolittle hydrophobicity curve. Phobius also predicts N-terminal signal peptides. The signal peptides predicted by Phobius were verified with SignalP1.1 [ 55 ]. Each program was used with default settings. Databases The Pfamseq database version 10.0 [ 56 ] was used for searching for homologous sequences. It is based on the Swiss-Prot 41.10 and SP-TrEMBL 23.15 databases. The Pfam database [ 11 ] version 11.0 [ 12 ] was used for domain assignments. Phylogenetic analysis The Pfamseq database [ 56 ] was searched for homologs using the Blastp 2.2.5 program [ 57 ] with default settings and with the putative worm proteins as query. Multiple alignments of full-length sequences were created using POA [ 58 ] with default settings. Gappy sequences and columns (>50% gaps) and redundant sequences (>99% identical) were removed. The program PHYLOWIN [ 59 ] with tree building method neighbor-joining [ 60 ] and PAM distance was used for constructing phylogenetic trees. Trees were also built with observed divergence and Poisson correction as distance methods, however, the results from that analysis are only discussed for genes where there were major differences in bootstrap support. If available, a yeast sequence was used as an outgroup. A total of 500 bootstrap tests were run on trees to assess the significance of the branching order. Only bootstrap values ≥ 50% were considered positive. Domain assignments Pfam-A domains were assigned using the Pfam database [ 11 , 12 ]. Pfam-B domains were not considered, since they are automatically generated and non-curated and therefore of lower quality. Nematode strains and culture conditions Maintenance and handling of C. elegans strains were as previously described [ 61 ]. Strains used were CGC N2 (wildtype) and CGC NL2099 ( rrf-3(pk1426) II) [ 19 ] (Caenorhabditis Genetics Center [ 62 ]). The rrf-3 mutant strain has an increased sensitivity to RNAi, also for neuronal genes [ 20 ], which otherwise are more refractory towards RNAi compared to other tissue types. Strain CB00907 ( dpy-5(e907) I) was used for generating the transgenic lines [ 63 ]. RNAi screening Generation and cloning of PCR products Total RNA extracted and purified from C. elegans using TRIzol (Invitrogen Life Technologies, Carlsbad, CA) was reverse transcribed using Reverse Transcription System (Promega, Madison, WI) and then PCR products were generated using Advantage™ 2 PCR Enzyme System (Clontech, Palo Alto, CA) with gene specific primers as well as primers for spliced leader 1 (SL1) and SL2 (Invitrogen Life Technologies): 95°C 60 s, 35 cycles of (95°C 30 s, 55°C 30 s, 68°C 4 min) followed by an additional extension at 68°C 4 min. See additional data file 1 for the primer sequences used for the RNAi studies. Products were ligated into linearized (XmaI) (New England Biolabs, Frankfurt am Main, Germany) and dephosphorylated L4440 vector (Fire Laboratory [ 64 ]) using Rapid DNA Ligation Kit (Roche, Mannheim, Germany) and transformed into JM109 E. coli bacterial strain (Promega). Colonies were screened using XmaI, correct colonies were grown in overnight cultures and DNA was extracted using QIAfilter Plasmid Kit (QIAGEN, Hilden, Germany). The vector with the insert was sequenced using ABI PRISM ® Big Dye™ Terminator Cycle Sequencing Ready Reaction Kits (Applied Biosystems, Foster City, CA). RNAi by feeding Strains N2 and rrf-3 were used for RNAi screening by feeding [ 30 , 65 ]. CGC [ 62 ] bacterial strain E. coli HT115(DE3) was transformed with the L4440 vector (Fire Laboratory [ 64 ]) containing the cloned gene fragment using standard methods. The vector contains an ampicillin (Amp) resistance and strain HT115 is tetracycline (Tet) resistant, so bacteria were selected on Amp (75 μg/ml) and Tet (12.5 μg/ml) plates. Single colonies were picked and grown in cultures of LB with Amp (60 μg/ml) and Tet (12.5 μg/ml) for 14–17 h. The bacterial solution was seeded onto NGM plates containing 1 mM IPTG and 25 μg/ml carbenicillin. Seeded plates were allowed to dry at room temperature. Eggs were prepared with standard bleaching method and transferred to the plates. N2 strain was incubated at 15, 20 and 25°C. rrf-3 was incubated only at 15°C and 20°C, since the strain has a temperature-sensitive decrease in broodsize [ 20 ]. The hatched worms and their progeny were scored for a number of different phenotypes [ 30 , 44 ]. The phenotypes assayed were: Adl (adult lethal), Bli (blistering of cuticle), Bmd (body morphology defect), Brd (low broodsize), Clr (clear), Dpy (dumpy), Egl (egg laying defect), Emb (embryonic lethal), Gro (slow post-embryonic growth), Him (high incidence of males), Lon (long body), Lva (larval arrest), Lvl (larval lethal), Mlt (molt defect), Muv (multivulva), Prz (paralyzed), Pvl (protruding vulva), Rol (roller), Rup (ruptured), Sck (sick), Sma (small), Ste (sterile), Stp (sterile progeny), Unc (uncoordinated). Emb was defined as greater than 10% dead embryos for N2 and greater than 30% dead embryos for rrf-3 . Ste and Stp required a brood size of fewer than ten for N2 and fewer than five for rrf-3 . Each postembryonic phenotype was required to be present among at least 10% of the analyzed worms. The experiment was ongoing for about 4 generations and the phenotypes were scored on a daily basis. A constant supply of transformed HT115 bacteria was ensured. The experiments were performed in duplicates at each temperature for each gene and worm strain. For the postembryonic phenotypes typically at least 20 worms per plate were scored. As a positive control the gene unc-22 ("twitchin") was used (Fire Laboratory vector pPD34.09 [ 64 ]). Empty L4440 vector was used as negative control. RNAi by injections The L4440 vector (Fire Laboratory [ 64 ]) containing the cloned gene fragment was linearized in two separate reactions using restriction enzymes NcoI and XhoI (New England Biolabs, Frankfurt am Main, Germany), respectively. The reactions were purified and single stranded RNA was synthesized using T7 RNA polymerase (Promega, Madison, WI). The two reactions were mixed and annealing was performed to produce double stranded (ds) RNA, which was subsequently purified. The dsRNA was injected undiluted into twelve young adult N2 hermaphrodites for each gene. The injected worms were put on individual plates and split between three different incubation temperatures (15, 20 and 25°C). Phenotypes were scored for both the injected worms and two subsequent generations on a daily basis. Phenotypes scored and criteria for scoring were the same as for the RNAi by feeding of strain N2. Generation of transgenic lines The transgenic lines were constructed at the Baille Laboratory, Simon Fraser University, Canada [ 66 ]. Transcriptional expression constructs for gonadal injection were generated using fusion PCR, also known as "PCR-stitching" [ 67 ]. Typically, about 3 kb of genomic DNA sequence immediately upstream of the predicted ATG initiator site, was used as the putative promoter (see supplementary material for primer sequences). When an upstream gene was within the 3 kb, the size of the putative promoter was adjusted downwards. For genes in operons, the sequence upstream of the first gene in the operon was used. The putative promoter was fused with another DNA fragment containing gfp (green fluorescent protein) and unc-54 3'UTR amplified from vector pPD95.67 (Fire Laboratory [ 64 ]). See additional data file 1 for the primer sequences used for generating the fusion PCR products. The resulting fusion PCR product was injected without purification into the gonad of young adult hermaphrodites of strain CB00907 at a concentration of 10 ng/μl together with 100 ng/μl dpy-5 (+) plasmid (pCeh361) in 1xTE buffer to generate an extrachromosomal array. Analysis of the expression patterns of the different transgenic lines was performed at the Vaz Gomes Laboratory, Karolinska Institutet, Sweden. Author's contributions AH carried out the membrane topology predictions, phylogenetic analysis, domain assignments, RNAi studies and analysis of gene expression in the transgenic worm strains. ES participated in the membrane topology predictions, phylogenetic analysis and domain assignments. DB contributed in making the transgenic worm strains. AVG participated in the RNAi studies and analysis of gene expression. All authors read and approved the final manuscript. Supplementary Material Additional File 1 Primer sequences used in RNAi and gene expression studies. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC533873.xml
544345	SCOPmap: Automated assignment of protein structures to evolutionary superfamilies	Background Inference of remote homology between proteins is very challenging and remains a prerogative of an expert. Thus a significant drawback to the use of evolutionary-based protein structure classifications is the difficulty in assigning new proteins to unique positions in the classification scheme with automatic methods. To address this issue, we have developed an algorithm to map protein domains to an existing structural classification scheme and have applied it to the SCOP database. Results The general strategy employed by this algorithm is to combine the results of several existing sequence and structure comparison tools applied to a query protein of known structure in order to find the homologs already classified in SCOP database and thus determine classification assignments. The algorithm is able to map domains within newly solved structures to the appropriate SCOP superfamily level with ~95% accuracy. Examples of correctly mapped remote homologs are discussed. The algorithm is also capable of identifying potential evolutionary relationships not specified in the SCOP database, thus helping to make it better. The strategy of the mapping algorithm is not limited to SCOP and can be applied to any other evolutionary-based classification scheme as well. SCOPmap is available for download. Conclusion The SCOPmap program is useful for assigning domains in newly solved structures to appropriate superfamilies and for identifying evolutionary links between different superfamilies.	Background Protein structure classifications are commonly used for studying structural and evolutionary relationships between proteins (namely remote homology inference), protein structure and function prediction, identification of potential functional residues and binding sites, understanding sequence/structure/function relationships in proteins, and as an aid in describing protein folds and families. Several structural classification schemes such as SCOP [ 1 ], CATH [ 2 ], and Dali Domain Dictionary [ 3 ] have been developed for the purpose of cataloguing all available protein structures. These databases are commonly used for studying structural and evolutionary relationships between proteins. Detecting remote homology between protein structures is a difficult task because of the challenge in differentiating between distant homologs and structural analogs. Several researchers have reported the inadequacy of various structural similarity measures for distinguishing homologous and analogous relationships [ 4 - 7 ]. Therefore, although the databases mentioned above are associated with automatic methods for identifying potential structural neighbors of a new protein query, they are often incapable of assigning domains to a unique position in the classification according to evolutionary relationships. Determining appropriate evolutionary relationships within a database is usually accomplished by expert manual analysis. Although manual classification of protein structures remains the gold standard, the necessity for reliable automatic tools that can reproduce the results of such a classification scheme becomes increasingly apparent as available databases continue to grow in size. Such tools must be capable of detecting homology between distantly related proteins while keeping false positives at a minimum. Available tools for assigning proteins to existing classification schemes use either structure-based or sequence-based comparison methods. Classification predictions from structure comparison tools like SSM [ 8 ], GRATH[ 9 ], and F2CS [ 10 ] are generally accurate to the fold or topology level but do not necessarily have evolutionary implications. Consequently, establishing homology between the query and the predicted neighbors often requires a more thorough examination. Classification assignments from sequence comparison tools such as SUPERFAMILY [ 11 ] can detect homology but often miss the more remote homologous relationships suggested by structural similarities. These tools are generally reliable for homology detection in easy to moderate cases but frequently produce many false positive results for more distant relationships. A strategy combining information from both sequence and structure comparisons would be expected to perform better than either method alone by exploiting the advantages of each approach. In this paper, we describe an algorithm developed to map domains within protein structures with their homologs in an existing classification scheme. The general strategy employed by this algorithm is to combine the results of several existing sequence and structure comparison tools in order to determine classification assignments. The comparison tools incorporated in the algorithm each utilize a different methodology for identifying homologous domains, and consequently, these tools have different advantages and limitations. An approach combining different methods of homology detection is expected to capitalize on the proficiencies of each comparison tool while the limitations of those tools are neutralized by the inclusion of other methods. Our algorithm, named SCOPmap, has been developed to map domains in protein structures to the SCOP database, which is a manually curated hierarchical classification scheme based on the structural and evolutionary relationships between proteins. SCOPmap assigns protein domains at the superfamily level, which is the broadest level of homology in the SCOP database. SCOPmap also performs assignments at the SCOP fold level when confident superfamily level assignments cannot be made. SCOPmap has two general applications. First, domains within newly solved protein structures can be identified and assigned to the appropriate SCOP superfamily. Second, SCOPmap can be used to find new links in SCOP by identifying potential evolutionary relationships between existing SCOP superfamilies. The strategy employed by this algorithm is not limited to SCOP and could be applied to any other similar database or classification scheme as well. We have evaluated the performance of SCOPmap on two test sets, each of which includes over 4500 protein domains. The first set is comprised of the proteins that are included in SCOP v1.63 but not in SCOP v1.61, while the second set contains the proteins that are included in SCOP v1.65 but not in SCOP v1.63. SCOPmap was able to correctly map greater than 94% of both test sets at the SCOP superfamily level. Comparison of SCOPmap results and SUPERFAMILY [ 11 ] results for the same test set indicates that SCOPmap performs better than SUPERFAMILY both in terms of overall correct assignments and in accurate definition of the domain boundaries of those assignments. We have analyzed SCOPmap's performance at both the SCOP superfamily and SCOP fold levels. We have also evaluated the performance of the individual comparison tools incorporated in the algorithm. Furthermore, we describe examples of difficult cases that are successfully mapped and investigate the reasons why some domains are not mapped automatically by our algorithm. Results Evaluation of SCOPmap performance Mapping of the tweaking set domains Results of SCOPmap performance on the tweaking set are shown in Table 1 (see Methods for description of tweaking and testing sets). Correct SCOP superfamily assignments were made for 87.8% of the tweaking set domains. For an additional 0.3% of the tweaking set domains, the superfamily assigned by SCOPmap is not the same as the SCOP-assigned superfamily. However, in each of these cases, the superfamily assigned by SCOPmap and the superfamily specified by SCOP are homologous. For example, SCOPmap assigns the 7-bladed β-propeller domain of an archael surface layer protein to a homologous SCOP superfamily of 6-bladed β-propellers [ 12 ]. Because the purpose of the SCOPmap is to assign domains at the broadest level of homology in the classification (i.e. the SCOP superfamily level), such cases are not considered false positives but instead reflect special cases in the SCOP database. 6.2% of the tweaking set domains were given no superfamily assignment by SCOPmap, but are domains that belong to SCOP superfamilies that are new in v1.63. Because such domains cannot be appropriately assigned to a superfamily that is represented in the library used by SCOPmap (v1.61 in this case), these are also considered correctly mapped (i.e. true negative assignments). Thus, a total of 94.3% of the tweaking set domains are correctly mapped by SCOPmap. The remaining 5.7% of the tweaking set are false negative assignments. These domains belong to superfamilies that exist in SCOP v1.61, but no superfamily assignment is made by SCOPmap. Mapping of the testing set domains Results of SCOPmap performance on the testing set (see Methods) are shown in Table 1 . Correct SCOP superfamily assignments were made for 91.2% of the testing set domains. In an additional 0.2% of the test set, the domain assignments given by SCOPmap are homologous to the superfamilies specified by SCOP. 3.1% of the tweaking set domains are given true negative assignments. These are cases in which the appropriate superfamily assignment is not a part of the library used by SCOPmap (based on SCOP v1.63 in this case), and no superfamily assignment is made by SCOPmap. Thus, a total of 94.5% of the testing set domains are given correct assignments by SCOPmap. 5.3% of the testing set domains are false negative assignments in which the domain belongs to a superfamily that is present in SCOP v1.63, but no superfamily assignment is made by SCOPmap. The remaining 0.2% of the testing set domains are given false positive assignments. False positive assignments in the testing set Because the score cutoffs used by SCOPmap's individual comparison tools were determined while considering domains from the tweaking set, those cutoffs were therefore influenced by the specific collection of domains in that set. Had a different test set been considered when establishing these cutoffs, it is likely that the score cutoffs would be slightly different. Thus, the few false positive assignments observed in the second test set are not unexpected. Furthermore, the number of false positive domain assignments made is higher than the number of incorrect hits between query and library domains that are accepted. Due to redundancy in the test set (e.g. often one structure contains several identical chains and therefore several identical domains), the 7 domains mapped incorrectly essentially reflect only 3 different examples of false positive assignments. Each incorrectly assigned domain has less than 10% sequence identity to the nearest library representative from the same SCOP superfamily. Furthermore, all of the false positive assignments are due to scores from the individual comparison tools which barely meet the cutoffs required for acceptance. Such cases reflect the influence that a few specific domains can have in determining the exact values of the minimum score threshold requirements. All incorrect assignments were made due to a hit accepted by one of the comparison tools that includes both sequence and structure components. For example, addiction antidote protein MazE from Escherichia coli (PDB code: 1mvf, chains D and E[ 13 ]; SCOP domains: d1mvfd_ and d1mvfe_) belongs to the Kis/PemI addiction antidote superfamily in SCOP and forms a pseudobarrel as a homodimer. SCOPmap incorrectly maps this protein to the "Transcription-state regulator AbrB, the N-terminal DNA recognition domain" superfamily in SCOP, which is a 2-layer α/β protein. This assignment is due to a hit found to the N-terminal DNA recognition domain of AbrB from Bacillus subtilis (PDB code: 1ekt [ 14 ]; SCOP domain: d1ekta_). Although the aligned regions of these two domains have the same secondary structure (an α-helix, a β-strand, and followed by a β-hairpin) and similar spatial arrangement, the overall topologies of these folds are highly dissimilar. This hit is accepted due to the 18 pairs of residues from the query and library representative which are equivalently aligned in pairwise alignments produced by PSI-BLAST (E-value = 55) and DaliLite (Z-score = 0.2). As the score cutoffs required by this comparison tool are E-value ≤ 100, Z-score > 0, and number of equivalent residue pairs ≥15, this particular query-library hit clearly falls just within the boundaries of the accepted score ranges. The nuclease domain of putative ATP-dependent RNA helicase Hef from Pyrococcus furiosus (PDB codes: 1j22, 1j23, 1j24, and 1j25 [ 15 ]; SCOP domains: d1j22a_, d1j23a_, d1j24a_, and d1j25a_), a member of the restriction endonuclease-like superfamily in SCOP, is incorrectly mapped to the FAD/NAD(P)-binding domain superfamily. This assignment is made because of a conservation pattern analysis hit to NADH-dependent ferredoxin reductase BphA4 from Pseudomonas strain KKS102 (PDB: 1d7y [ 16 ]; SCOP domain: d1d7ya2). Although the core of both the query and the library representative is an α/β domain containing a 5-stranded β-sheet, the overall topology is not similar. This query-library pair hit is accepted because of the matrix-based conservation score of 0.32, which is based on the structural alignment of these two domains by DaliLite (Z-score = 3.7), while the score cutoffs required by this comparison tool are matrix-based score ≥ 0.25 and DaliLite Z-score ≥ 2. Again, the scores for this hit fall near the boundaries of the accepted score ranges. The proteolytically-cleaved peptide C from bovine lysosomal α-mannosidase (PDB code: 1o7d [ 17 ]; SCOP domain: d1o7d.2) belongs to the galactose mutarotase-like superfamily in SCOP, but is incorrectly mapped to the "alpha-Amylases, C-terminal domain β-sheet domain" superfamily. This assignment is due to a hit identified by conservation pattern analysis to the C-terminal domain of neopullulanase from Bacillus stearothermophilus (PDB code: 1j0h [ 18 ]; SCOP domain: d1j0ha2). Although the core of lysosomal α-mannosidase peptide C and the C-terminal domain of neopullulanase each form a β-sandwich-like fold, the topologies of these folds are different. The COMPASS-based conservation score for this query-library pair (0.52) is based on the structural alignment of the two domains by DaliLite (Z-score = 4.6). These scores fall just within the required ranges for acceptance by the conservation pattern comparison method (COMPASS-based conservation score ≥ 0.5 and DaliLite Z-score ≥ 2). Comparison of tweaking and testing set results Table 1 shows that the SCOPmap results are comparable for the tweaking set and the testing set. SCOPmap performance on the two test sets are nearly equivalent: 94.3% (tweaking set) vs 94.5% (testing set) correct assignments; 5.7% (tweaking set) vs 5.3% (testing set) false negative assignments; and 0.0% (tweaking set) vs 0.2% (testing set) false positive assignments. The most significant apparent differences are in the results for the specific types of correct assignments: true positives with ranges accurate within 10 residues, true positives with ranges that are not accurate within 10 residues, and true negatives. These seemingly disparate results are predominantly reflections of inconsistencies in test set composition rather than in SCOPmap performance. More specifically, these variations are primarily due to the number of query domains that belong to new SCOP superfamilies. The most obvious consequence is the fraction of each test set given true negative assignments (6.2% in tweaking set, 3.1% in testing set), which is directly dependent on the fraction of each test set that belongs to new SCOP superfamilies. If domains from new SCOP superfamilies are ignored, the apparent disparity in SCOPmap boundary definition accuracy is reduced. For example, if the entire test sets are considered, there is a 2.3% difference in the number of domains correctly assigned whose ranges are accurate within 10 residues of the SCOP-defined boundaries. However, when considering only domains that can potentially be mapped correctly (i.e. domains that do not belong to new SCOP superfamilies), 86.8% of the tweaking set domains are correct assignments that are accurate within 10 residues, compared to 86.4% of the testing set domains. Similarly, 92.4% of all correctly assigned domains in the tweaking set are accurate within 10 residues, compared to 91.6% for the corresponding domains in the testing set. The comparable results are a reliable indication of the consistency of SCOPmap performance because the two test sets are of nearly equivalent difficulty. First, the two test sets include approximately the same fraction of trivial assignments: 73.7% of mappable domains in the tweaking set are assigned by gapped BLAST while 73.6% of mappable domains in the testing set are assigned by gapped BLAST, where "mappable" means the domain is both evolutionarily relevant and is a member of a SCOP superfamily that exists in the version of SCOP used as the library. Of the non-trivial mappable domains (i.e. mappable domains that are not assigned by gapped BLAST), the average sequence identity between the query domain and the closest library representative from the same SCOP superfamily is 29.2% in the tweaking set and 28.6% in the testing set. Fold level assignments Fold level assignments are attempted for regions of query chains at least 20 residues in length for which no superfamily assignment was made. Results are shown in Figure 1 . In the tweaking set (v1.61-v1.63 test set), fold level assignments are made for ~30% of the 545 SCOP-defined domains with no superfamily level assignment. 92% of these fold level assignments are correct. In the testing set (v1.63-v1.65 test set), fold level assignments are made for ~44% of the 414 SCOP-defined domains with no superfamily level assignment. Of these assignments, ~94% are correct. Similar to the superfamily level assignments, the apparent disparity in fold level assignments are due primarily to the relative composition of the two test sets rather than inconsistency in performance. There are two principal attributes of test set composition that result in improved fold level results. First, domains from new folds are typically given no fold level assignment by SCOPmap, so a smaller fraction of unmapped domains from new folds will result in a decreased number of domains for which no assignment is made. Second, because the structural similarity between two domains from the same superfamily is likely to be greater than that between two domains from different superfamilies within the same fold, a larger fraction of unmapped domains from existing superfamilies will result in an increased number of correct fold level assignments. Both of these attributes favor the testing set over the tweaking set (results not shown). This indicates that the testing set is less challenging in terms of fold level assignments, which is consistent with the improved results relative to the tweaking set (Figure 1 ). Although no fold level assignment is made in a large number of cases (~70% of tweaking set unmapped domains and ~56% of testing set unmapped domains), this result is not altogether unexpected for several reasons. First, as discussed above, a significant fraction of the unmapped domains in each set belong to new SCOP folds, so no appropriate fold level assignment exists among the set of library representatives. Next, the minimum Z-score cutoff required for making fold level assignments is strict in order to minimize false positive assignments. While Ortiz et al . report that MAMMOTH Z-scores greater than 5.25 are generally reliable for fold predictions [ 19 ], we find that a MAMMOTH Z-score of 10 is required for making reliable fold assignments. Although 45% of domains in the tweaking set from existing folds but without a fold assignment (171 of 380 domains) have at least one MAMMOTH hit to a representative of the appropriate fold with a Z-score between 5.25 and 10, results in this range are not used due to many occurrences of false positive assignments (data not shown). Conversely, because MAMMOTH Z-scores greater than 22 are sufficient for assignments at the superfamily level (see Methods), fold assignments are neither necessary nor made for query-library domain pairs with such overwhelming structural similarity. Furthermore, because query-library domain pairs with sufficient sequence similarity to be recognized by automatic methods are mapped at superfamily level, unmapped domains have very little sequence similarity to the corresponding library representatives. Consequently, fold assignments are made only for a rather limited set of queries: domains with extremely low sequence similarity as well as significant but not overwhelming structural similarity to library representatives. The false positive rates are nearly identical in the two test sets (~2.6%). In both sets, the false positive rate of fold level assignments is significantly higher for domains that belong to new SCOP folds compared to those from existing SCOP folds. For example, in the second testing set, 6 of the 86 domains that belong to new folds have incorrect fold level assignments (7.0%) while only 5 of the 328 domains from existing folds are given an incorrect assignment (1.5%). Because false positive hits are likely to fall just above the Z-score cutoff for fold level assignment, many false positives are ignored due to other hits found with better Z-scores, which are true positives in most cases. Thus, because domains that belong to existing SCOP folds should have significant structural similarity to at least one library domain (i.e. the library representative(s) of that particular SCOP fold), the negative effect of false positive hits to these domains is minimized in the false positive rate relative to that for domains from new SCOP folds. False positive fold level assignments are typically due to a query and library representative sharing similar but not identical topology. For example, the structure of riboflavin kinase (PDB code: 1n06 [ 20 ]; SCOP domain: d1n06b_) is a query in v1.61-v1.63 test set and belongs to a SCOP superfamily that is new to SCOP v1.63. Appropriately, no superfamily level assignment is made. The fold of riboflavin kinase is a n = 6, S = 10 β-barrel with strand order 163452, but SCOPmap assigns this domain to the double psi β-barrel fold in SCOP, which is an n = 6, S = 10 β-barrel with strand order 163425. In this case, the incorrect fold assignment is based on similarity of overall topology, but other false positive fold assignments occur when a region within a query domain and a region within a SCOP representative have similar topology despite overall dissimilarity of the folds. For example, the structure of the ε-subunit of the plasmid maintenance system (PDB code: 1gvn [ 21 ]; SCOP domain: d1gvna_) is another query in v1.61-v1.63 test set which also belongs to a new superfamily in SCOP v1.63. Again, no superfamily level assignment is made, as appropriate. The fold of the ε-subunit is a 3-helix up-and-down bundle with left-handed twist, but SCOPmap assigns this domain to a 4-helix up-and-down bundle fold. The three α-helices in the query domain and the last three α-helices of the SCOP representative have identical topology, similar lengths, and equivalent spatial orientation to each other. This false positive is a result of the query topology matching a region of a SCOP representative. The opposite case, when a region of the query domain is the same as the topology of an entire SCOP representative, occurs as well. For example, the structure of viral chemokine binding protein m3 (PDB code: 1mkf [ 22 ]; SCOP domain: d1mkfa_), a query in v1.61-v1.63 test set, belongs to a new fold in SCOP v1.63. Appropriately, no superfamily level assignment is made for this query. The fold of this domain is a 10-stranded β-sandwich with 6 β-strands in one sheet and 4 in the other. This domain is mapped at the fold level to an 8-stranded β-sandwich with 4 β-strands in each sheet. Although the overall folds of these two domains are different, 7 β-strands from each of these two β-sandwich folds have identical topology and mutual spatial arrangement. Unsurprisingly, correct fold assignments are made predominantly for typical globular proteins while no fold assignments are made for small protein or coiled coil folds. Outside of this observation, there are no recognizable trends suggesting types of folds for which assignments are more easily made. Furthermore, it should be noted that fold assignments are not our main goal. Rather, these assignments are a by-product of the comparison tools that are used for mapping at the superfamily level by SCOPmap. The purpose of making fold level assignments is merely to assist the user in further study of those domains which SCOPmap does not assign at the superfamily level. The fold level mapping strategy and score cutoffs have not been optimized to perform fold mapping with high sensitivity or low false positives. Performance of SCOPmap compared to SUPERFAMILY Overall performance SUPERFAMILY is another tool that attempts to assign domains within a query protein to the superfamily level of SCOP. It is the only package that we are aware of that meets our two requirements for direct comparison: the program performs a similar task and is available for download. The results of the performance of SUPERFAMILY relative to SCOPmap are shown in Table 1 . Overall, SCOPmap performs better than SUPERFAMILY. SUPERFAMILY correctly maps 91.4% of domains compared to the 94.3% assigned to the correct SCOP superfamily by SCOPmap. Furthermore, SCOPmap is much more proficient at defining accurate domain boundaries. SCOPmap delineates domain boundaries within 10 residues of the SCOP-defined boundaries for 81.4% of domains, while SUPERFAMILY performs as well in only 70.1% of cases. This difference is due partly to the use of MAMMOTH and DaliLite in our algorithm. However, the results of our algorithm when using only sequence comparison tools show that there is still a 6.5% advantage over SUPERFAMILY in terms of accurately defined ranges (Table 1 ). Thus, the inclusion of structure comparison methods is not solely responsible for the dramatic improvement in boundary definition. Presumably, a second predominant factor in the increased domain boundary accuracy is the strict coverage criteria for sequence comparison methods incorporated in SCOPmap. Table 1 shows the results of using only the BLAST, RPS-BLAST, PSI-BLAST, and COMPASS portions of our algorithm. This modified version of SCOPmap (henceforth referred to as the "sequence-only algorithm") was expected to perform similarly, if not better than, SUPERFAMILY. It was therefore surprising to observe significantly more false negative assignments by the sequence-only algorithm compared to the SUPERFAMILY algorithm (12.5% and 8.6%, respectively). Investigation of the 573 false negatives from the sequence-only algorithm indicates three general explanations for these missed assignments (data not shown). In ~47% of these cases (270 of 573 domains), there are no sequence comparison hits below the required E-value thresholds. Next, in ~17% of cases (97 of 573 domains), sequence hits that pass both the E-value and coverage criteria are found, but the domain is not assigned due to an unresolved choice between conflicting superfamilies. In the remaining 36% of cases (206 of 573 domains), sequence comparison hits to at least one superfamily representative are found that pass the required E-value cutoffs but fail the coverage criteria. These 206 domains correspond to ~4.5% of this test set and account for the difference in false negative rates between the sequence-only algorithm and SUPERFAMILY, which does not have a coverage requirement. Performance on non-trivial domain assignments Because nearly 70% of the domains can be mapped using only gapped BLAST (Table 3 ), the results of both SCOPmap and SUPERFAMILY are skewed in favor of trivial domain assignments. In order to evaluate the performance of these two programs on more challenging assignments, the results were re-tabulated excluding all domains assigned via gapped BLAST (Table 2 ). Here, SCOPmap assigns 81.6% of domains to the appropriate SCOP superfamily while SUPERFAMILY correctly maps 77.1% of domains, so SCOPmap's advantage in correctly assigned domains increases from 2.9% for all domains to 4.5% for only non-trivial assignments. SCOPmap's proficiency in domain boundary definition is also accentuated, as the difference in percent of domains with accurately defined domain boundaries increases from 11.3% for all domains (SCOPmap: 81.4%, SUPERFAMILY: 70.1%) to 12.8% for non-trivial assignments (SCOPmap: 42.8%, SUPERFAMILY: 30.0%). Thus, evaluating only the non-trivial assignments emphasizes the advantages of SCOPmap over SUPERFAMILY. Comparison of false negative assignments The false negative assignments made by SCOPmap (261 domains) and by SUPERFAMILY (395 domains) were compared in order to determine the degree of overlap between the two sets of unassigned domains. One might expect that a significant number of the false negative assignments would be shared by the two algorithms and would represent those cases that are too difficult to be confidently mapped by existing automatic comparison tools. Indeed, 205 domains are given false negative assignments by both SCOPmap and SUPERFAMILY. Therefore, of the 261 false negative assignments made by SCOPmap, only 56 domains (21%) are correctly mapped by SUPERFAMILY. 38 of these domains were correctly identified by at least one of the comparison methods used but were not assigned (due, for example, to an unresolved choice of superfamily assignment). Most of the remaining domains that were assigned by SUPERFAMILY but not identified by SCOPmap represent cases that are typically difficult for automatic methods: 8 are small disulfide-rich domains, 3 are relatively short domains (74, 75, and 126 residues) that are interrupted by very large insertions (290, 289, and 282 residues respectively), and 1 domain contains many short breaks in the sequence and structure. The few remaining examples are domains that could have been reasonably expected to be mapped by SCOPmap: E. coli succinate dehydrogenase subunit SdhC (PDB codes: 1nek [ 23 ], chain D and 1nen [ 23 ], chain D; SCOP domains: d1nekd_ and d1nend_) is a helical bundle protein that belongs to the succinate dehydrogenase/fumarate reductase transmembrane segment superfamily in SCOP, and the PKD-like domain of Methanosarcina mazei surface layer protein (PDB codes: 1l0q [ 12 ], chains A, B, C, and D; SCOP domains: d1l0qa1, d1l0qb1, d1l0qc1, d1l0qd1) is an immunoglobulin-like domain that belongs to the PKD domain superfamily in SCOP. Other than the low sequence identity between these queries and the library representatives of the corresponding SCOP superfamilies, there are no convincing arguments for why these assignments might not be made. In each of these cases, significant hits are found by the structure comparison tools used in SCOPmap: SdhC has a DaliLite Z-score of 8.7 to a library representative of its SCOP superfamily, and surface layer protein PKD-like domain has a MAMMOTH Z-score of 10.6 to the library representative of its SCOP superfamily. However, the limited sequence similarity between the query and representative domains results in insufficient BLOSUM scores to meet the required score cutoffs of these methods. Although these are consequently false negative assignments at the superfamily level, the correct fold level assignment was made in each of these last 6 cases. Conversely, approximately half of the false negative assignments made by SUPERFAMILY (190 of 395 domains) are correctly mapped by SCOPmap. Of these domains, ~54% are first identified by a sequence comparison tool in SCOPmap (gapped BLAST, RPS-BLAST, PSI-BLAST, or COMPASS), ~29% are first identified by a structure comparison tool (MAMMOTH or DaliLite), and the remaining ~17% are first identified by a method that combines both sequence and structure information (correlation of conservation patterns or the agreement of DaliLite alignments with gapped BLAST, RPS-BLAST, or PSI-BLAST alignments). Discussion Performance of individual comparison methods In order to assess the relative performance of the individual comparison tools used by SCOPmap, the number of assignments in the tweaking set gained by each additional comparison method was evaluated. The results are summarized in Table 3 . For each comparison tool, the number of domains first identified by that method was determined, and the percent of previously unassigned domains gained by that method was calculated. The comparison tools are listed in order of increasing sensitivity to distant homologs: sequence comparison methods (BLAST, RPS-BLAST, PSI-BLAST, and COMPASS), structure comparison methods (MAMMOTH and DaliLite), and finally comparison methods that incorporate both sequence and structure information (correlation of conservation patterns and agreement of DaliLite alignments with BLAST, RPS-BLAST, or PSI-BLAST alignments). Domains are included in the total count for only the least sensitive comparison tool that identified the hit. The most number of assignments are made by gapped BLAST and RPS-BLAST, which give 69.1% gain and 36.3% gain of previously unmapped domains, respectively. However, these assignments are among the easiest in the set. The average sequence identity between the query domain and the closest library representative of that superfamily is 80.1% for gapped BLAST assignments and 41.1% for RPS-BLAST assignments. Furthermore, these numbers are considerably inflated as a consequence of the surfeit of trivial assignments in the tweaking set (Figure 2 ). PSI-BLAST, MAMMOTH, and DaliLite each give between 10% and 20% gain of previously unmapped domains. The average sequence identities between the identified query domains and the library domains indicate that these assignments are neither trivial nor unusually difficult. The two structure comparison methods show similar overall performance by this assessment, although DaliLite does have the advantage over MAMMOTH both in number of assignments and percent gain as well as in difficulty of assignments made. This seemingly implies that comparison via MAMMOTH is an unnecessary step, and indeed, nearly all domain assignments made by MAMMOTH are also made by DaliLite (data not shown). However, MAMMOTH is both necessary for and proficient at determining potential hits by DaliLite. The pre-identification of potential hits drastically reduces the running time compared to comprehensive comparison of the query domains to all library domains by DaliLite. Furthermore, MAMMOTH is essential for making fold level assignments. The conservation pattern analysis and the calculation of agreement between DaliLite alignments and BLAST, RPS-BLAST, or PSI-BLAST alignments have 4.2% and 5.9% gain of previously unmapped domains, respectively. Although the numbers of additional assignments are among the lowest of any of the comparison tools, these two methods also make the most challenging assignments of any of the comparison tools included in SCOPmap. The average sequence identity between query domains and library representatives for assignments made first by these methods is less than 15%. Specific examples are discussed below. Thus, the general observation is that, as expected, those comparison tools more sensitive to distant homology typically make more challenging assignments, but with lower percent gains. The only clear exception to this trend is COMPASS. COMPASS has the lowest percent gain of any step at 3.3%, and the domains first identified by this method are only moderately difficult assignments (average sequence identity 27.2%). This is presumably due in part to the extremely strict E-value cutoff necessary for avoiding false positives (1 × 10 -10 ). Furthermore, of the four sequence comparison tools used in SCOPmap, COMPASS is most sensitive to remote homologs. Therefore, if the query-library domain pair has sufficient sequence similarity to be recognized by automatic methods, it is likely that the hit would also be identified by one of the less sensitive sequence comparison tools and consequently be accounted for earlier in Table 3 . SCOPmap performance on remote homologs Correctly mapped remote homologs The similarity of the tweaking set to the representative library domains is shown in Figure 2 (white bars). Nearly 50% of tweaking set domains are more than 70% identical to one of the library representatives from the same SCOP superfamily. Furthermore, 69.1% of the tweaking set domains can be correctly mapped by gapped BLAST (Table 3 ). Other domains, however, are more difficult to assign due to limited similarity of the query domain to the representative library domains. SCOPmap is able to make several such assignments, including nearly 300 domains with less than 20% sequence identity to the closest library domain from the same SCOP superfamily (black bars, Figure 2 ). One prevalent difficulty in making classification assignments by automatic methods is correctly assigning domains that have very limited sequence similarity to the library representatives. One such example of a difficult but correctly assigned domain is the N-terminal domain of mannitol 2-dehydrogenase from Pseudomonas fluorescens (PDB code: 1lj8 [ 19 ], N-terminal domain; SCOP domain: d1lj8a2). In SCOP, this domain belongs to the NAD(P)-binding Rossmann-fold domains superfamily. There are 90 representatives of this superfamily in the library, all of which have less than 10% sequence identity to the query domain. There are no BLAST, RPS-BLAST, PSI-BLAST, COMPASS, MAMMOTH, or DaliLite hits to these library representatives that pass both the required coverage and E-value or Z-score thresholds. Hits to three of the 90 superfamily representatives are identified by DaliLite: the N-terminal domain of glycerol-3-phosphate dehydrogenase from Leishmania mexicana (PDB code: 1evy [ 24 ], N-terminal domain; SCOP domain: d1evya2) with Z-score 6.9, the N-terminal domain of conserved hypothetical protein MTH1747 from Methanobacterium thermoautotrophicum (PDB code: 1i36 [ 25 ], N-terminal domain) with Z-score 6.3, and the N-terminal domain of lactate/malate dehydrogenase from Methanococcus jannaschii (PDB code: 1hye [ 26 ], N-terminal domain; SCOP domain: d1hyea1) with Z-score 6.4. Because of the poor BLOSUM scores calculated for the pairwise alignments given by DaliLite, none of these hits are accepted by the DaliLite comparison method. However, these relatively high Z-scores indicate that the DaliLite alignments are reliable enough for use in the comparison of conservation patterns method, and hits to two of these superfamily representatives are accepted based on correlation of conservation patterns: the N-terminal domain of glycerol-3-phosphate dehydrogenase (SCOP domain: d1evya2) has matrix-based conservation score = 0.26, and the N-terminal domain of conserved hypothetical protein MTH1747 (SCOP domain: d1i36a2) has matrix-based conservation score = 0.11. In both of these cases, approximately 75% of the most conserved positions in the query domain and in the library domain are equivalent (Figure 3c ). Furthermore, these most conserved positions are clustered around the nucleotide-binding sites, which are equivalent in these domains (Figure 3a,b ). The N-terminal domain of this query structure is therefore mapped to the NAD(P)-binding Rossmann-fold domain superfamily in SCOP based on the high degree of correlation between the conservation patterns of the query domain and these two superfamily representatives. Conformational differences between similar protein domains also result in challenging classification assignments for automatic structure comparison tools. One such example is the antimicrobial cathelicidin motif of protegrin-3 from Sus scofa (PDB code: 1lxe [ 27 ]; SCOP domain: d1lxea_). The crystal structure of this protein shows the domain in a swapped dimer conformation (Figure 4a ). The closest library representative to this query domain is cystatin from Gallus gallus (PDB code: 1cew [ 28 ]; SCOP domain: d1cewi_), which belongs to the cystatin/monellin superfamily in SCOP. This domain is a monomer in the crystal structure (Figure 4b ). The sequence identity between the query (cathelicidin motif of protegrin-3) and this library representative (cystatin) is approximately 19%. The hit between the query and this library representative is found by both the RPS-BLAST and DaliLite methods. However, the scores for these hits are relatively poor as a result of the low sequence identity and the conformational variation between the two domains. The scores for these comparisons (RPS-BLAST E-value = 16 and DaliLite Z-score = 2.4) fail the score cutoff criteria for these methods individually. Comparison of the alignments produced by these two methods, however, indicates that a significant portion of the domain is aligned equivalently by RPS-BLAST and DaliLite (Figure 4c ). Thus, based on the agreement of these two methods, the cathelicidin motif of protegrin-3 is correctly mapped to the cystatin/monellin superfamily of SCOP. Another common problem for many automatic comparison methods is the presence of large insertions or deletions in the query domain. This third example demonstrates the ability of the mapping program to correctly assign such cases. Monomeric isocitrate dehydrogenase from Azotobacter vinelandii (PDB code: 1itw [ 29 ]; SCOP domain: d1itwa_) belongs to the isocitrate/isopropylmalate dehydrogenase superfamily in SCOP. There are two representatives of this superfamily in the library, both of which have less than 15% sequence identity to the query domain. Furthermore, the query domain has an approximately 250-residue insertion relative to the superfamily representatives (Figure 5 ). There are no BLAST, RPS-BLAST, PSI-BLAST, or COMPASS hits to either library representative. Although the MAMMOTH hit to 3-isopropylmalate dehydrogenase from Salmonella typhimurium (PDB code: 1cnz [ 30 ]; SCOP domain: d1cnza_) is accepted with Z-score 22.2, the presence of the large insertion in the query results in an erroneous range definition by MAMMOTH (Figure 5c ). Comparison of the query to this same library representative by DaliLite identifies residues 164–397 as an insertion in this domain (Figure 5c ). Although SCOP assigns the entire chain of monomeric isocitrate dehydrogenase as one domain (residues 1–741), residues 150–404 are defined as an insert region. Thus, the DaliLite-based assignment made by SCOPmap (residues 2–163, 398–671) is a reasonably accurate domain definition. Domains without SCOPmap assignments at the superfamily level In 5.7% of the tweaking set, no superfamily assignment is made for domains that should belong to superfamilies that are included in SCOP v1.61. General explanations for these false negative assignments are summarized in Table 4 . Of the 261 unmapped domains, 19.2% percent (50 domains) are found by meeting the required score cutoffs of one or more of the comparison tools used, but these domains are not assigned due to a conflict with another domain identified in the same query chain. There are two ways in which this may happen: there may be an unresolved choice of superfamily assignment over a certain region of the query chain, or the boundary of one domain may erroneously extend over a second domain resulting in one domain being assigned while the another domain is missed. In the remaining 80.8% of unmapped domains, comparison of the query to the library domains do not pass the score cutoffs of any of the methods used. These domains typically have only limited structural similarity as well as less than 20% sequence identity to the library representatives. All domains that have greater than ~20% sequence identity to a library representative from the same SCOP superfamily but are not identified by any of the comparison tools used in SCOPmap are small protein domains less than 50 residues in length. Because automatic methods often perform poorly on small proteins, such cases are not unexpected. These unmapped small protein examples comprise only 0.2% of the tweaking set. Furthermore, the unmapped domains often have inserted or deleted structural elements relative to the library domains. The unmapped and unidentified domains fall into three general categories in terms of structural similarity to the library representatives. First, 33.3% of unmapped domains have very little structural similarity to the corresponding library domains. When the MAMMOTH scores for a query domain are insufficient for making superfamily assignments, these scores are used as an initial indicator of whether specific query-library domain pairs are likely to be assigned by DaliLite (see Methods). For these unmapped domains, the MAMMOTH scores to library domains are too poor to be identified even as potential hits. Next, there are a small number of cases (6.1% of unmapped domains) that have potential but unconfirmed structural similarity to library representatives. In these cases, one or more potential hits are identified by MAMMOTH, but DaliLite does not produce output for those pairs. This could mean that the DaliLite Z-score is less than zero for the given pair of domains, or that either the query domain, the library representative, or both could not be handled by DaliLite because, for example, the structure lacks recognizable secondary structure, contains only Cα coordinates, or is less than 30 residues in length, etc . Finally, the remaining 41.4% of unmapped domains have recognizable but insufficient structural similarity to the library representatives. For these domains, hits are found via DaliLite but the scores of the hits do not meet the required cutoffs. Because such scores cannot be confidently distinguished from false positives, no superfamily assignment is made. Since the inception of the SCOP database, the rapid growth in the number of available protein structures has resulted in a classification scheme that is not equally uniform in all parts. This is primarily apparent in overpopulated folds and superfamilies, such as TIM β/α-barrels, where intermediate relationships exist but are difficult to describe within the original SCOP classification scheme. These special cases in the SCOP database also contribute to the rate of false negative assignments by SCOPmap. In a later section, the conservative nature of SCOP is demonstrated by cases in which homologous proteins are assigned to different superfamilies. As a consequence of this attribute of the SCOP database, good hits via automatic comparison methods are sometimes found to multiple SCOP superfamilies. In some cases, SCOPmap is not capable of selecting one final assignment out of several correct choices. These 28 examples, which make up the unresolved choice of superfamilies category in Table 4 , account for less than 1% of the tweaking set but 10.7% of all false negative assignments. Conversely, there are also numerous instances in which the SCOP classification is quite liberal. Examples are rampant in the sections of the database that the authors describe as not a part of the proper SCOP classification, such as the low resolution structures and peptides classes. These classes are not included in the SCOPmap library and are therefore not considered by our algorithm. However, cases were also observed in the evolutionarily relevant multi-domain proteins class of SCOP. The multi-domain proteins class is problematic in the sense that it deviates from the format followed by the remainder of the SCOP database. Members of this class have not been classified at the domain level, and there is often wide variation in the size and domain composition of the entries. One such example was detected during the manual investigation of false negative assignments from the tweaking set. Reovirus polymerase λ3 (PDB code: 1n1 h [ 31 ]; SCOP domain: d1n1ha_) belongs to the DNA/RNA polymerases superfamily in the multi-domain proteins class of SCOP. The structural fold of domains in the DNA/RNA polymerases superfamily has been described as a "right-hand" configuration containing "palm", "fingers", and "thumb" subdomains. Domains in this superfamily, of which there are >200, typically include 2 or 3 subdomains of the "right-hand" fold. For example, Moloney murine leukemia virus (MMLV) reverse transcriptase (PDB code: 1mml [ 32 ]; SCOP domain: d1mml__), which is one of the representatives of this superfamily included in the v1.61 library, is a 265-residue fragment containing only the "palm" and "fingers" subdomains. Reovirus polymerase λ3, however, also includes a 380-residue N-terminal domain as well as a 377-residue C-terminal "bracelet" domain, in addition to the "palm", "fingers", and "thumb" subdomains. Thus, a 1267-residue, 3-domain protein (reovirus polymerase λ3) and a 265-residue, single domain fragment (MMLV reverse transcriptase) are classified equivalently at the superfamily level in SCOP. Naturally, such variations within the database are problematic for making appropriate classifications via automatic methods. Examples of false negative SCOPmap assignments Some superfamily assignments are missed due to extremely limited similarity between the query domain and the corresponding library representatives. One such example is Saccharomyces cerevisiae DNA-binding domain from transcription factor Ndt80 (PDB code: 1mnn [ 33 ]; SCOP domain: d1mnna_), which belongs to the p53-like transcription factors superfamily in SCOP. Members of this superfamily bind DNA through an s-type Ig fold. There are seven library representatives of this superfamily, all of which have less than 10% sequence identity with the query domain. There are no hits to these representatives found by BLAST, RPS-BLAST, or PSI-BLAST with E-value less than 100 or by COMPASS with E-value less than 1 × 10 -3 . Because the MAMMOTH hits to these representatives are very poor (Z-scores below 2.5), MAMMOTH finds neither accepted hits nor potential hits for comparison via DaliLite. Although the conserved core of this superfamily is observable by eye (Figure 6a ), the many inserted structural elements relative to the library representatives contribute to the poor performance of the automatic structural comparison methods. The DNA-binding function of this domain may have contributed to its inclusion in this superfamily by the SCOP authors. Superfamily assignments are also missed in cases where the similarity to library representatives is moderately significant but still insufficient for distinction from false positives. One such example is adaptor protein ClpS from E. coli (PDB code: 1lzw [ 34 ], chain A; SCOP domain: d1lzwa_) (Figure 6b ), which belongs to the ClpS-like superfamily in SCOP. The one representative of this superfamily in the library shares ~11% sequence identity with the query domain. BLAST, RPS-BLAST, and PSI-BLAST hits to this library representative are not found with E-values less than 100, and a COMPASS hit to the library domain is not found with E-value less than 1 × 10 -3 . Comparison of the query and library domain by MAMMOTH and DaliLite give more substantial results: a MAMMOTH Z-score of 10.4 with BLOSUM score -1.0 × 10 -2 for the pairwise alignment produced by MAMMOTH, and a DaliLite Z-score of 8.8 with BLOSUM score 4.5 × 10 -4 for the pairwise alignment produced by DaliLite. Unfortunately, these scores fall just below the required cutoffs for superfamily assignment via these methods. Thus, no superfamily assignment is made. However, the MAMMOTH Z-score does meet the fold level cutoff, so a correct fold assignment is made for this query domain. Additionally, technical shortcomings of automatic methods contribute to missed superfamily assignments. For example, δ-conotoxin TxVIA from Conus textile (PDB code: 1fu3 [ 35 ]; SCOP domain: d1fu3a_) is a 27-residue small protein that belongs to the omega toxin-like superfamily in SCOP. There are 21 library representatives of this superfamily, some of which share up to 40% sequence identity with the query domain. However, there are no hits to these representatives found by BLAST, RPS-BLAST, or PSI-BLAST with E-value less than 100 or by COMPASS with E-value less than 1 × 10 -5 . The MAMMOTH hits to these 21 representatives all have Z-scores well below 4. Furthermore, DaliLite cannot handle this protein due to the short length, thus precluding DaliLite comparisons with library representatives. Thus, despite significant sequence and structural similarity of δ-conotoxin TxVIA to several library representatives (Figure 6c ), no superfamily assignment is made due to the poor performance of automatic methods on small proteins. Finding new links between SCOP superfamilies: examples of homologs in different SCOP superfamilies identified by SCOPmap The thiamin phosphate synthase superfamily and the ribulose-phosphate binding barrel superfamily are one example of homologous SCOP superfamilies identified by SCOPmap. Both superfamilies have a TIM β/α-barrel fold. When thiamin phosphate synthase is used as the query, hits to 8 different members of the ribulose-phosphate binding barrel superfamily are identified. These hits are found by PSI-BLAST, COMPASS, DaliLite, and the agreement between pairwise alignments produced by DaliLite and by RPS-BLAST or PSI-BLAST. Because confident hits are identified by both sequence and structure comparison methods, the homology between the two superfamilies is considered reliable, despite the limited sequence identity (<20%). The structure of thiamin phosphate synthase and indole-3-glycerophosphate synthase, which is a representative of the ribulose-phosphate binding barrel superfamily, are shown in Figure 7a,b . The RPS-BLAST alignment (E-value 1 × 10 -10 ) (Figure 7c ) and the DaliLite alignment (Z-score 15.4) of these two proteins are similar: 101 pairs of residues (~40% of the proteins) are equivalently aligned by the two comparison tools. Furthermore, three phosphate-binding residues are in equivalent positions both spatially and in the sequences of these proteins (Figure 7 ). The homology between these two superfamilies has been previously reported [ 36 ]. The C-terminal domain of RNA polymerase alpha subunit and the DNA repair protein Rad51, N-terminal domain superfamilies are another pair of homologous superfamilies identified by SCOPmap. The domains in these two superfamilies have a 5-helix bundle structure (SAM domain-like fold), with one classic and one pseudo HhH motif as noted in SCOP. Members of both superfamilies have DNA-binding functions, and the observed or predicted DNA-binding surfaces are similar between the two superfamilies (Figure 7d,e ). The closest representatives from each of these two superfamilies share ~32% sequence identity with each other. When the C-terminal domain of RNA polymerase alpha subunit superfamily is used as the query, all three members of this superfamily find hits to the single member of the DNA repair protein Rad51, N-terminal domain superfamily. RPS-BLAST (E-value 0.002), COMPASS (E-values ~10 -16 ), and MAMMOTH (Z-scores ~9) identify these hits. The detection of both confident sequence and structure comparison hits further supports the link between these two superfamilies. The examples discussed here are two cases among many. The examination of the complete list of potential homologs from different SCOP superfamilies is in progress. Conclusions We have developed an algorithm for mapping domains within protein structures to an existing classification scheme. When applied to the SCOP database, this algorithm performs with ~95% accuracy (i.e. the correct superfamily assignment is made or no superfamily level assignment is made, as appropriate). SCOPmap produces better results than SUPERFAMILY, both in terms of overall correct assignments and in the definition of the domain boundaries of those assignments. Examination of difficult cases has demonstrated the ability of SCOPmap to make non-trivial assignments, including some domains that represent common problems associated with automatic comparison tools. SCOPmap is also capable of identifying potential evolutionary links between proteins from different SCOP superfamilies. SCOPmap should be useful to researchers interested in determining the SCOP classification of domains within newly solved protein structures. Furthermore, SCOPmap can be modified to perform similar mapping tasks within other protein classification databases. An additional potential use of the algorithm would be as an internal check in the preparation of new classifications or the maintenance and updating of existing classifications. Reliable methods for automatic updates to existing classification schemes become increasingly important with the rapid growth in sequence and structure database size. Methods Mapping strategy of the SCOPmap algorithm General strategy The purpose of SCOPmap is to assign domains within protein structures to the SCOP classification at the broadest level of homology, i.e. the SCOP superfamily level. The general strategy is to combine the results of several existing sequence and structure comparison tools to determine superfamily assignments as well as domain boundaries. Because the basis for identifying relationships between proteins varies between the different comparison tools, this combinatorial approach is expected to perform better than a single comparison tool alone. Furthermore, an approach utilizing multiple comparison tools is consistent with the conclusions reached by Novotny et al . from an analysis of several fold comparison servers [ 37 ]. There are three main steps in this mapping strategy. First, hits are identified between the query protein and proteins with known SCOP assignments using several existing comparison tools. Next, the results of those comparison tools are used to determine the appropriate SCOP superfamily level assignment for domains within the query. Assignments are made by a consensus-like method in which more reliable comparison tools are given preference. Finally, the algorithm uses the results of the comparison tools to define the boundaries of the domain assignments by identifying the longest non-overlapping segments. Library of representative SCOP domains A subset of SCOP domains with less than 40% identity to each other was downloaded from the ASTRAL [ 38 , 39 ] database. This set contains domains from the all alpha proteins, all beta proteins, alpha and beta proteins (a+b and a/b), multi-domain proteins, membrane and cell surface proteins and peptides, and small proteins classes of SCOP. Domains from the coiled coil proteins class were manually added to the library. In this paper, results using two different SCOP libraries are discussed. The library based on SCOP v1.61 contains 4813 domains from 1110 SCOP superfamilies, while the library based on SCOP v1.63 contains 5265 domains from 1232 superfamilies. Each library includes at least one representative of each SCOP superfamily. Set of representative query chains Input for SCOPmap is a list of PDB [ 40 ] identifiers. Each chain in these structures is considered as a separate query. The BLASTCLUST program (I. Dondoshansky and Y. Wolf, unpublished; ) is used for preliminary clustering of all chains at 95% sequence identity and 95% length coverage. A representative set of query chains is constructed from the first member of each BLASTCLUST cluster, excluding chains fewer than 20 residues in length. Chains less than 20 residues in length are designated as fragments and are ignored by SCOPmap. Mapping step 1: identifying hits between query and library domains using existing comparison methods The gapped BLAST [ 41 ], RPS-BLAST[ 42 ], PSI-BLAST [ 41 ], COMPASS [ 43 ], MAMMOTH [ 19 ], and DaliLite [ 44 ] tools are used in SCOPmap. The first four of these are sequence comparison tools and are listed in order of increasing sensitivity to remote homologs: a query sequence against a database of sequences (gapped BLAST), a query sequence against a database of profiles (RPS-BLAST), a query profile against a database of sequences (PSI-BLAST), and a query profile against a database of profiles (COMPASS). The two structure comparison tools used are the MAMMOTH and DaliLite algorithms. Additionally, SCOPmap includes two tools which incorporate elements of both sequence and structure comparisons: correlation of conservation patterns and the agreement of pairwise alignments produced by structure comparison tools (DaliLite or MAMMOTH) with those produced by sequence comparison tools (gapped BLAST, RPS-BLAST, or PSI-BLAST). Thus, similarities between proteins are identified using eight different comparison methods, which are described in detail below. Method 1) gapped BLAST [ 41 ]: query sequence against database of sequences Gapped BLAST is run for each representative query sequence against sequences of all chains from PDB structures in SCOP (37,007 sequences in SCOP v1.61; 41,066 sequences in SCOP v1.63). The criteria for an accepted BLAST hit are an E-value ≤ 0.005 and coverage of all but 10 residues at each end of both the query and database sequences. Hits are also accepted if the query and library sequences are at least 80% identical and all but 10 residues at each end of the query sequence are covered by the alignment, irrespective of E-value. Because the database sequences used for gapped BLAST are complete chains, the accepted hits are then converted from library chains to library domains according to the SCOP-defined domain boundaries of those library sequences. This conversion is not necessary for accepted hits from the other seven comparison methods since the library representatives in those methods are domains rather than complete chains. For all query chains with accepted BLAST hits, superfamily assignment is based solely on the BLAST results and no other comparison tools are used. All query chains with no BLAST hits passing the described criteria are submitted to each of the remaining methods. Method 2) RPS-BLAST [ 42 ]: query sequence against database of profiles RPS-BLAST is run for the query sequence against a database of profiles for the library of representative SCOP domains. Profiles were constructed for each library domain by running PSI-BLAST against the non-redundant database for 5 iterations or until convergence with an E-value cutoff of 0.005. The criteria for an accepted RPS-BLAST hit are an E-value ≤ 0.005 and coverage of all but 10 residues at each end of the library domain. Method 3) PSI-BLAST [ 41 ]: query profile against database of sequences A profile for the query sequence is constructed by running PSI-BLAST against the non-redundant protein database for 5 iterations or until convergence with an E-value cutoff of 0.001. This profile is subsequently used as an input for a PSI-BLAST search against a database of all SCOP domain sequences (42465 domain sequences in SCOP v1.61; 47013 domain sequences in SCOP v1.63). The criteria for an accepted PSI-BLAST hit are an E-value ≤ 10 -4 and coverage of all but 10 residues at each end of the SCOP domain database sequence. Method 4) COMPASS [ 43 ]: query profile against database of profiles The profiles for the query (constructed in the PSI-BLAST step) and the SCOP library domains (constructed in the RPS-BLAST step) are prepared for COMPASS by: 1) deleting all columns with gaps in the query sequence, 2) removing all sequences identical to the query, and 3) retaining only 1 copy of any sequences in the profile that have greater than 97% identity. COMPASS is then run for the query profile against each of the SCOP library domain profiles. Accepted COMPASS hits have an E-value ≤ 10 -10 and coverage of all but 10 residues at each end of the library domain. Method 5) MAMMOTH [ 19 ]: query structure against database of structures The query structure is compared to each library domain structure via MAMMOTH. For each query-library domain pair, the MAMMOTH Z-score (Z M ) and the normalized BLOSUM [ 45 ] score for the pairwise alignment made by MAMMOTH (BS M ) are calculated. MAMMOTH hits are accepted if they meet all of the following criteria: 1) Z M ≥ 4.0; 2) coverage of ≥50% of the library domain; 3) (BS M ≥ 0.3) or (BS M ≥ Z M -1/2 - 0.24) or (Z M ≥ 22.0). For hits meeting only the first two criteria, the COMPASS E-value (CE M ) is calculated for the two domains, with the alignment of the two profiles guided by the pairwise alignment made by MAMMOTH. Thus, additional accepted hits are identified that pass the following criteria: Z M ≥ 4.0, coverage of ≥ 50% of the library domain, and CE M ≤ 1.0. The cutoffs for accepted hits were determined based on the MAMMOTH Z-score (Z M ), BLOSUM score (BS M ), and COMPASS E-value (CE M ) of 106,310 randomly chosen pairs of SCOP domains from SCOP v1.61. Approximately 1/3 of these pairs of domains belong to the same SCOP superfamily while the remaining 2/3 of the pairs belong to different SCOP superfamilies. Method 6) DaliLite [ 44 ]: query structure against library structure comparisons Additional structure comparisons are performed for queries with a segment of 20 residues or longer that did not correspond to an accepted MAMMOTH hit. Query-library domain pairs for which BS M ≥ -0.01Z M - 0.03, Z M > 0, and the pairwise alignment made by MAMMOTH covered at least 40% of the library domain are identified. If more than 200 query-library domain pairs met these criteria, only the 200 query-library domain pairs with the highest Z M scores are selected. If no pairs meet these criteria, the 50 query-library domain pairs with the highest Z M scores are identified. The score cutoffs for selecting pairs for comparison via DaliLite were determined by evaluating the MAMMOTH Z-scores (Z M ) and BLOSUM scores (BS M ) for randomly chosen pairs of SCOP domains that pass the DaliLite score cutoffs (see below) but fail the MAMMOTH score cutoffs (see above). The threshold was chosen by determining the score cutoffs that would identify the most number of pairs passing the DaliLite cutoffs and the fewest pairs failing the DaliLite cutoffs, thereby maximizing the number of potential accepted hits while minimizing the overall computation time required. DaliLite structure comparison is performed for each of the selected query-library domain pairs, and the DaliLite Z-score (Z D ) and the normalized BLOSUM score for the pairwise alignment made by DaliLite (BS D ) are calculated. Hits are accepted if they meet one of the following sets of criteria: 1) Z D ≥ 4.0, BS D ≥ -0.01Z D + 0.15, and coverage of ≥50% of the library domain; 2) BS D ≥ 0.3 and coverage of ≥50% of the library domain; 3) Z D ≥ 14.0 and coverage of ≥50% of the library domain. The cutoffs for accepted hits were determined based on the DaliLite Z-score (Z D ) and BLOSUM score (BS D ) of 4000 randomly chosen pairs of SCOP domains from SCOP v1.61, where half of these pairs belong to the same superfamily and half of the pairs belong to different superfamilies. Method 7) CSV: correlation of conservation patterns Because homologous domains often have similar conservation patterns, the degree of correlation between the conservation patterns of two domains can be used for remote homolog detection. Distant homologs typically display drastically diminished overall sequence similarity. Thus, such cases of remote homology are more likely to be identified by conservation pattern analysis, which considers only the most conserved residues, rather than by typical sequence comparison methods, which are highly dependent on overall sequence similarity. Conservation scores for query-library domain pairs are calculated by two methods: using a conservation substitution matrix and using the COMPASS algorithm. The query-library domain pairs selected for conservation pattern comparison are determined based on the results of the DaliLite pairwise comparisons in the previous method. The correlation of conservation patterns are calculated for all query-library domain pairs with Z D ≥ 4.0, or for the 20 pairs with highest DaliLite Z-score (Z D ≥ 2.0 required) if no pairs have DaliLite Z-score ≥ 4. Only pairs for which the library domain profile (constructed for the RPS-BLAST step and modified for the COMPASS step) contains 5 or more sequences are considered. The AL2CO algorithm [ 46 ] (window size 3) is used to calculate the entropy-based conservation index for each position in the query profile and in the library domain profile. DaliLite-aligned positions scoring in the top 25% of either profile are selected, henceforth referred to as the chosen positions. Any two given positions from the profiles of the query and library domains can be compared to determine their similarity in terms of conservation patterns. The degree of correlation between those conservation patterns is referred to as the position-pair conservation score. For example, if both positions are highly conserved, the position-pair conservation score for that specific pair will be high. Conversely, if one position is highly conserved while the amino acid distribution in the other position is random, the position-pair conservation score will be low. In the first scoring system, position-pair conservation scores are determined based on the entropy-based conservation indices for the chosen positions with a conservation substitution matrix used as a scoring matrix. Then, the scoring matrix-based conservation score is calculated for the query-library domain pair by: CSV cons,D = [S n - S rand ]/ [(S 1 +S 2 )/2 - S rand ], where S n is the sum of position-pair conservation scores of the aligned query positions vs. library domain positions ("chosen positions" only, see above), S 1 is the sum of position-pair conservation scores of the chosen query positions against themselves (query positions vs. query positions), S 2 is the sum of position-pair conservation scores of the chosen library domain positions against themselves (library domain positions vs. library domain positions), and S rand is the sum of position-pair conservation scores of the chosen positions for all-against-all query positions vs. library domain positions normalized over length. A COMPASS-based conservation score is also calculated for each query-library domain pair. In this scoring system, a COMPASS-based position-pair score, which describes the similarity between any two given positions, is determined based on the methodology introduced in the COMPASS method [ 43 ]. Then, the COMPASS-based conservation score for the query-library domain pair is calculated by: CSV compass,D = [CS n - CS rand ]/ [(CS 1 +CS 2 )/2 - CS rand ], where CS n is the sum of COMPASS-based position-pair scores of the aligned query positions vs. library domain positions ("chosen positions" only, see above), CS 1 is the sum of COMPASS-based position-pair scores of the chosen query positions against themselves (query positions vs. query positions), CS 2 is the sum of COMPASS-based position-pair scores of the chosen library domain positions against themselves (library domain positions vs. library domain positions), and CS rand is the sum of COMPASS-based position-pair scores of the chosen positions for all-against-all query positions vs. library domain positions normalized over length. Conservation score hits are accepted if they meet one of the following sets of criteria: 1) CSV cons,D ≥ 0.1 and Z D ≥ 5; 2) CSV cons,D ≥ 0.25 and Z D ≥ 2; 3) CSV compass,D ≥ 0.4 and Z D ≥ 5; 4) CSV compass,D ≥ 0.5 and Z D ≥ 2. These cutoffs for accepting hits were determined based on the CSV cons,D scores, CSV compass,D scores, and DaliLite Z-scores of 4000 randomly chosen pairs of SCOP domains from SCOP v1.61. In cases for which the DaliLite program produces no output, conservation pattern analysis is performed using pairwise alignment produced by MAMMOTH instead of FSSP alignments. The conservation analysis is done for the query-library domain pairs that would have otherwise been submitted to the DaliLite algorithm for structural comparison (see above). Only those residue pairs in which the Cα atoms are located within 4Å, which are indicated by an asterisk (*) by the MAMMOTH algorithm, are considered. Again, a window size of 3 is used in the AL2CO program and only the top scoring 25% of positions are used for calculating the conservation scores. Matrix-based and COMPASS-based conservation scores are calculated as described above. Conservation score hits based on MAMMOTH alignments are accepted if they meet one of the following sets of criteria: 1) CSV cons,M ≥ 0.3 and Z M ≥ 4; 2) CSV compass,M ≥ 0.4 and Z M ≥ 4 These cutoffs for accepting hits were determined based on the CSV cons,M scores, CSV compass,M scores, and MAMMOTH Z-scores of 2000 randomly chosen pairs of SCOP domains from SCOP v1.61. Method 8) agreement of DaliLite or MAMMOTH alignments with gapped BLAST, RPS-BLAST, or PSI-BLAST alignments Remote evolutionary links between protein domains can be gleaned using a combination of sequence and structural information, even when neither of these methods alone is capable of providing convincing evidence for common descent. In this method, the degree of correlation between a pairwise alignment made by DaliLite and alignments made by the sequence comparison methods is determined so that DaliLite can be used to evaluate potential hits from BLAST, RPS-BLAST, or PSI-BLAST. For any query-library domain pair with Z D > 0 and BLAST, PSI-BLAST, or RPS-BLAST E-value ≤ 100, the number of correctly aligned residues (N ali ) in the sequence alignment is calculated using the DaliLite alignment as a reference. Hits are accepted for which Z D > 0, E-value ≤ 100, and N ali ≥ 15. These cutoffs were determined based on the DaliLite Z-scores, E-values, and number of equivalently aligned residues from 1000 randomly chosen pairs of SCOP domains from SCOP v1.61. If an error occurs while running DaliLite for the query domain, agreement of the MAMMOTH alignment and BLAST, RPS-BLAST, or PSI-BLAST alignments is instead calculated for the same potential hits. In these cases, hits are accepted for which Z M > 2.0, E-value ≤ 100, and N ali ≥ 15. These cutoffs were determined based on the MAMMOTH Z-scores, E-values, and number of equivalently aligned residues from 1000 randomly chosen pairs of SCOP domains from SCOP v1.61. Mapping step 2: assigning domains from query chains to SCOP superfamilies Accepted hits from the sequence and structure comparison methods are mapped onto the query chain and domains within the chain are then assigned to SCOP superfamilies. In cases where accepted hits from multiple SCOP superfamilies mapped to the same region of the query chain, SCOPmap attempts to choose only one correct SCOP superfamily assignment. If the overlap between two different SCOP superfamily representatives covers <50% of both domains, the conflict is resolved by the domain boundary definition (see "Mapping Step 3" below). Otherwise, SCOPmap attempts to determine which SCOP superfamily among the accepted hits is most likely to be the correct assignment. First, for each of two conflicting assignments, all accepted hits that overlap by at least 75% and are from the same SCOP superfamily are identified. For each set of accepted hits (one set corresponding to each of the conflicting SCOP superfamilies), the number of methods that identified accepted hits to that SCOP superfamily is determined. If one SCOP superfamily is found by more methods than the other SCOP superfamily, the assignment with hits from the greater number of methods is accepted as correct. If both SCOP superfamilies are identified by an equal number of methods, the priority of those methods is used to choose the correct SCOP superfamily. The methods are ranked by reliability, which was subjectively determined based primarily on the observed number of false positives accepted by a given method during SCOPmap development. Priority rankings are as follows: BLAST > RPS-BLAST or PSI-BLAST > MAMMOTH or DaliLite > COMPASS > conservation pattern correlation or agreement of DaliLite and sequence method alignments. If both SCOP superfamilies are found by methods with equivalent priorities, the Z-scores and E-values of the hits are evaluated. If only one of the two conflicting SCOP superfamilies has E-values from any sequence comparison method below 10 -10 or Z-scores (Z M or Z D ) above 14.0, that SCOP superfamily assignment is accepted as correct. If a SCOP superfamily assignment has still not been made, the domain assignments to that query chain are flagged as unresolved. Of the 4580 tweaking set domains (see Results), only 25 domains (0.5%) were unassigned due to unresolved choice between conflicting SCOP superfamilies. The results obtained by inverting the order of these two steps (e.g. first comparing E-values and Z-scores, and then considering priority rankings of the eight methods) were also evaluated. There were no cases where the inverted order gave additional correct assignments, and there was a small number of cases that could be resolved by the original strategy but not by the inverted strategy. Thus, the methodology described above is used for choosing between conflicting superfamily assignments. Mapping step 3: defining boundaries of domain assignments Domain boundary definitions are assigned by identifying the longest non-overlapping domain assignments, with priority given to assignments made by structure comparison methods. First, DaliLite is run for all query-library domain pairs found by MAMMOTH, and the DaliLite range is used in place of the MAMMOTH range unless there is an error in the DaliLite output. Then, ranges of accepted hits are given priority rankings based on which method determined the range of that hit. DaliLite ranges have highest priority, followed by MAMMOTH ranges, and then all sequence comparison method ranges. The longest non-overlapping segments with the highest priority rankings are then identified. A 3-residue cushion for overlap is allowed. Overlapping domains for which boundaries cannot be reconciled within 3 residues are flagged as unresolved. Of 4580 tweaking set domains, only 3 domains (0.1%) were unassigned due to unresolved domain boundary definition. Assignments at the SCOP fold level For query chains with a segment at least 20 residues in length which is not assigned to a SCOP superfamily, mapping at the SCOP fold level is attempted. In the SCOPmap algorithm, MAMMOTH is run comprehensively against the library of representative structures. Therefore, no additional comparisons must be made in order for fold level assignments to be determined. For this reason, MAMMOTH is used for fold level assignments rather than DaliLite, which is typically run against less than 5% of the library domains. The single criterion for potential SCOP fold assignment is a MAMMOTH Z-score > 10. Fold level assignments are made by selecting the hit to an unmapped region with the highest MAMMOTH Z-score (>10) that also covers at least 50% of the library domain. The fold level Z-score cutoff was determined based on the MAMMOTH Z-scores of 106,310 randomly chosen pairs of SCOP domains from SCOP v1.61. These same pairs of domains were used for determining the superfamily assignment cutoffs (see above). Approximately 2/3 of these pairs of domains belong to the same SCOP fold while the remaining 1/3 of the pairs belong to different SCOP folds. Description of test sets SCOPmap performance was evaluated on two separate test sets. The first set is comprised of the proteins that are included in SCOP v1.63 but not in SCOP v1.61. SCOPmap was run using a library based on the previous SCOP release (v1.61), and the SCOPmap domain assignments were compared to the SCOP-defined classification in subsequent SCOP release (v1.63). This set contains 5133 SCOP-defined protein domains, but analysis of SCOPmap performance is based only on the 4580 SCOP-defined domains with evolutionary relevance: 464 low resolution structure domains, 63 peptides, 21 designed proteins, and 5 domains that were later removed from the database are intentionally excluded. The first test set was used to establish whether the score cutoffs for the individual comparison tools used by SCOPmap were strict enough to avoid false positive assignments. After first running SCOPmap for this set of domains, a false positive rate of ~1.5% was observed. The score thresholds for some of the individual comparison tools were subsequently made more strict in order to avoid all false positive assignments in this set. For example, the E-value cutoff for PSI-BLAST was changed from 5 × 10 -3 to 1 × 10 -4 , and the E-value cutoff for COMPASS was adjusted from 1 × 10 -4 to 1 × 10 -10 . Because some of the domains in this set were considered while establishing the score thresholds, the first test set is more correctly described as a "tweaking" set rather than a testing set. This set was also used for comparison to SUPERFAMILY, for which the score threshold was also chosen specifically for the purpose of precluding false positive assignments. The recommended 0.02 E-value cutoff for SUPERFAMILY, which would allow for the correct assignment of only an additional ~1% of the tweaking set domains, was not chosen due to the 4.3% false positive rate it incurs. Instead, the E-value cutoff was set at 1 × 10 -5 , the maximum value for which no false positive assignments were observed. For this comparison, the SUPERFAMILY algorithm was used with the library of SAM [ 47 ] hidden Markov models based on SCOP v1.61. The second set of domains used to evaluate SCOPmap performance contains proteins included in SCOP v1.65 but not in SCOP v1.63. The second test set can be considered a true testing set. The testing set contains 5335 SCOP-defined protein domains, but only the 4941 SCOP-defined domains with evolutionary relevance were used for analysis of SCOPmap performance. Low resolution structures, peptides, and designed proteins were ignored. The library of SCOP representative domains used for mapping the queries in this set is based on SCOP v1.63. Using SCOPmap to identify homologs between SCOP superfamilies SCOPmap can also be used to identify potentially homologous proteins that belong to different SCOP superfamilies. Detection of such homologs is accomplished with a slightly altered strategy from the mapping algorithm described above. The modified algorithm evaluates one SCOP superfamily at a time by attempting to detect potential hits to SCOP domains belonging to other superfamilies via the comparison methods described above. A set of query domains is constructed from the domains that are currently included in that SCOP superfamily (based on SCOP v1.63). As in the original mapping algorithm, the query sequences are first clustered at high sequence identity to reduce the computational time. Next, each of the 8 comparison methods described above is employed for each representative query. In the original mapping strategy, queries for which accepted hits are detected via gapped BLAST are not submitted to any of the other comparison methods. However, in this modified strategy, all comparison tools are run for all representative queries, regardless of the results of the gapped BLAST step. The output is a list of all accepted hits from each of the comparison methods to SCOP domains that do not belong to the query superfamily. All hits to SCOP domains within the query superfamily are simply ignored and excluded from the output. Finally, manual analysis of potential hits was performed for selected examples in order to evaluate the significance of those hits and to determine whether an evolutionary link is likely to exist between the two SCOP superfamilies in question. Program availability The SCOPmap script and instructions for library construction are available for download at . SCOPmap results for representative PDB structures that are not included in the SCOP database are available here as well. Authors' contributions SC developed the code, tested program performance, analyzed the results, and drafted the manuscript. YQ contributed to code development. SSK determined score thresholds for the individual comparison tools used. LNK proposed many additional suggestions for improving algorithm performance. NVG conceived of the study and participated in its design and coordination. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544345.xml
517709	Combined analysis of expression data and transcription factor binding sites in the yeast genome	Background The analysis of gene expression using DNA microarrays provides genome wide profiles of the genes controlled by the presence or absence of a specific transcription factor. However, the question arises of whether a change in the level of transcription of a specific gene is caused by the transcription factor acting directly at the promoter of the gene or through regulation of other transcription factors working at the promoter. Results To address this problem we have devised a computational method that combines microarray expression and site preference data. We have tested this approach by identifying functional targets of the a 1- α 2 complex, which represses haploid-specific genes in the yeast Saccharomyces cerevisiae . Our analysis identified many known or suspected haploid-specific genes that are direct targets of the a 1- α 2 complex, as well as a number of previously uncharacterized targets. We were also able to identify a number of haploid-specific genes which do not appear to be direct targets of the a 1- α 2 complex, as well as a 1- α 2 target sites that do not repress transcription of nearby genes. Our method has a much lower false positive rate when compared to some of the conventional bioinformatic approaches. Conclusions These findings show advantages of combining these two forms of data to investigate the mechanism of co-regulation of specific sets of genes.	Background A bioinformatic approach to identifying cis-regulatory elements controlling transcription has become feasible with the availability of complete genome sequences and large scale expression data using high-throughput methods such as microarrays [ 1 , 2 ] and SAGE [ 3 ]. The expression data provides a list of genes whose expression is significantly modified under a particular condition. However, this data does not indicate whether these genes are direct targets of a particular transcription factor or if the changes in expression are the result of an indirect effect caused by altering the expression of other transcription factors that work directly at the promoter. Using information about sequence preference for binding of particular transcription factors, one can identify possible regulatory binding sites within a sequenced genome. However, this approach does not indicate if the sites are functional. We have therefore developed an algorithm that combines both of these approaches to distinguish between the direct and indirect targets that are regulated by a particular transcription factor. We have applied this methodology to study the transcriptional regulatory system that specifies cell mating-type in the yeast Saccharomyces cerevisiae [ 4 ]. Yeast have three cell types, haploid a and α cells, and the a / α diploid, that differ in their ability to mate and in the proteins they express. Cell mating-type is determined in part by α 2 and a 1, which are cell-type-specific proteins that are members of the homeodomain (HD) DNA-binding family. In an a / α diploid cell, α 2 binds with a 1 to form a heterodimer complex that represses transcription of haploid-specific genes [ 5 ]. The crystal structures of the α 2 HD binding DNA alone and in complex with a 1 have been solved, providing models for how these complexes bind DNA [ 6 , 7 ]. Biochemical and mutational analysis of each protein and their DNA-binding sites have defined the requirements for DNA recognition by this complex [ 8 - 10 ]. Genome-wide expression analysis has also been performed on each of the different cell types [ 11 ]. The combination of these resources has allowed us to develop and test algorithms to identify target sites for the a 1- α 2 complex. Previous work, using a relatively simple binding site search program identified targets for the α 2-Mcm1 complex, which represses a -cell-type specific genes in α and a / α cells [ 12 ]. The more advanced methods described in this paper have helped identify several novel targets of the a 1- α 2 complex that may be involved in cell-type specific processes. Interestingly, we identified several genes that are repressed in diploid cells but do not appear to be direct targets of the a 1- α 2 complex, suggesting that these genes are controlled by another transcriptional regulatory factor that is directly or indirectly regulated by the a 1- α 2 complex. We have also identified a number of a 1- α 2 target sites that do not repress adjacent genes. The combination of site preference and expression data is therefore a valuable tool to identify direct functional targets of a transcription factor or complex. Results Development of a search algorithm for targets of the a1- α 2 complex To generate an algorithm that combines microarray expression data and mutational analysis of binding sites, we first defined a scoring method that ranks gene expression data. We utilized the microarray expression data from Galitski and coworkers for gene expression in the a and α haploid and a/ α diploid cells, as well as various polyploids [ 11 ]. Since the a 1- α 2 complex should be absent in any of the homozygous a or α type polyploids ( a, aa , ..., α,αα ,..., etc.) we expect the expression of haploid-specific genes in these cells to be much higher than in cells that are heterozygous for the MAT locus ( a α , aa α , a αα , aa αα , etc.). Thus, one term in the scoring function rewards lower expression in heterozygous cell types compared to the homozygous cell types (see the Methods section for details). We expect that most of the haploid-specific genes will be expressed equally in both of a and α cell types. Consequently, we have introduced a second term in the scoring function that penalizes such differences in expression in the two haploid cell types. This scoring function would identify haploid-specific genes that are repressed in diploid cells, but would not indicate if these genes are direct targets of the a 1- α 2 repressor complex. To identify genes from this ranking that are directly repressed by the a 1- α 2 complex we used the available mutational data on the a 1- α 2-binding site [ 9 ]. In these experiments, the effects of single base pair mutations of the a 1- α 2 consensus binding site were measured by assaying their ability to repress transcription of a heterologous promoter and by electrophoretic mobility shift DNA-binding assays (EMSA). Under the assumption that the level of expression is proportional to how often that site is unoccupied, we used the effects of the single base mutations to estimate the parameters for the binding energy of sites with different bases at each position. We then used this information to search for potentially strong binding sites in the promoter regions (in practice, 800 bp upstream of the translation start site) of every gene in the genome. This search provides us with a list of genes with putative a 1- α 2 binding sites in the promoter, irrespective of functionality. Even under ideal conditions, either of the above lists of genes would not directly indicate a haploid-specific gene directly repressed by the a 1- α 2 complex. In addition, our accuracy is limited by the noise in the microarray expression data, as well as by simplifying assumptions made to utilize single-base mutational data to score arbitrary sequences. We therefore decided to rank the genes using a scoring system that takes into account expression patterns across different mating types, as well as the likelihood of finding a good a 1- α 2-binding site in the promoter region of the gene. To test the significance of these composite scores, we generated permuted data, which combined random promoters with the expression data. This analysis suggested that the top 10–15 predictions were significant. The results of experiments on the 24 genes with the highest combined scores from our analysis are shown in Table 1 . Table 1 Potential a 1- α 2 Binding Sites in Haploid-specific Genes ORF Gene Expression p-val a Binding p-val b Combined p-val a 1- α 2 ChIP c YDL227C HO 0.0006 0.0017 1.1e-6 + f YLR265C NEJ1 0.0003 0.0053 1.7e-6 + YBL016W FUS3 0.0001 0.0991 1.6e-5+ YOR212W STE4 0.0020 0.0082 1.7e-5 + YJR086W STE18 0.0008 0.0218 1.7e-5 + YHR005C GPA1 0.0005 0.0437 2.1e-5 + YDR103W STE5 0.0017 0.0298 5.2e-5 + YBR073W RDH54 0.0053 0.0116 6.2e-5 + YGR044C RME1 0.0009 0.0720 6.8e-5 + f YGL248W PDE1 0.0182 0.0040 7.3e-5 + YPL038W MET31 0.0292 0.0027 8.0e-5 + YDR088C SLU7 0.0303 0.0038 1.2e-4 - YGL052W 0.0117 0.0109 1.3e-4 - YJL157C FAR1 0.0013 0.1141 1.4e-4 + YPR122W AXL1 0.0091 0.0163 1.5e-4 + f YIL099W SGA1 0.0063 0.0267 1.7e-4 - YLR233C EST1 0.0226 0.0090 2.0e-4 - YKL182W FAS1 0.0578 0.0035 2.1e-4 - YMR053C STB2 0.0028 0.0884 2.5e-4 - YNL319W 0.0123 0.0222 2.7e-4 - YFR012W 0.0088 0.0125 2.8e-4 - YNL188W KAR1 0.0019 0.1890 3.6e-4 - YGL193C 0.0014 0.2654 3.8e-4 - YMR157C FMP39 0.1557 0.0026 4.0e-4 - YBR158W d AMN1 0.0037 0.0221 8.7e-5 + YCL066W e MATα1 0.1630 0.9510 1.5e-1 + f a The ranking of haploid-specific gene expression determined by analysis of microarray data [11]. b The ranking of potential a 1- α 2 target of haploid-specific genes. c A + indicates that the a 1- α 2 binds to the promoter by ChIP assay. d Identified in a search for sites with 1500 bp from the start of the ORF e Identified if remove the penalty for expression in one haploid cell type but not the other. f Identified as direct targets of the a 1- α 2 repressor complex in previous studies. a 1- α 2 binding to the promoter regions of genes identified in the search To evaluate the success of our computational algorithm for identifying direct targets of the a 1- α 2 repressor complex, we assayed binding by the complex to the identified promoters using chromatin-immunoprecipitation (ChIP) assays with polyclonal antibody directed against the α 2 protein. In α haploid and a/ α -diploid cells the α 2 protein combines with the MADS-box transcription factor Mcm1 to bind to elements in the promoters of a -specific genes to repress their transcription [ 13 , 14 ]. We therefore included in each PCR a primer set for the promoter region of the a -specific gene STE6 to serve as a positive control for the ability to ChIP α 2 in both haploid α and diploid a/ α cells. This primer set also allowed us to rule out the possibility that the predicted a 1- α 2 target genes were immunoprecipitating because of binding by the α 2-Mcm1 complex. The primer set for the YDL223C promoter, a gene not bound or repressed by the a 1- α 2 or α 2-Mcm1 complexes, was included in the reaction as a negative control for non-specific immunoprecipitation of the DNA. A gel displaying the results for a few of the promoters that were assayed is shown in Figure 1 and the data summarized for all the promoters that were predicted to be directly repressed by the a 1- α 2 complex is listed in Table 1 . In general, there is a very good correlation between the experimental data with the predictions based on our computational algorithm. Almost all of the high scoring genes were bound by the a 1- α 2 complex in vivo. Genes that had a combined p-value higher than the threshold of 1.5 × 10 -4 (which corresponds to choosing the top 15 of the list in Table 1 ) do not appear to be strongly bound by the complex. This p-value corresponds to roughly a probability of one in six thousand, indicating it is possible to get such combination by chance. Figure 1 ChIP assays of promoter fragments that are predicted to be targets of the a 1- α 2 complex. ChIP assays with antibody to the α 2 protein were performed on lysates from MAT a / MATα ( a / α ) and mat a Δ/ MATα (Δ/ α ) cells. Total chromatin (TC) and immunoprecipitated (IP) samples were subjected to multiplex PCR with primers flanking potential a 1- α 2 sites (labeled hsg ) in the indicated promoters. Primers for the promoter region of STE6 , an a -specific gene that is repressed by the α 2-Mcm1 complex in both cell types were used as a positive control for the ChIPs. Primers that hybridize in YDL223C , a gene that is not regulated by α 2, was used as the negative control. The presence of a band over background levels from the test promoter from a / α lysates but not mat Δ/ α indicates that the a 1- α 2 complex is specifically binding to the promoter. Assays for the GPA1, FAR1, NEJ1, RDH54, MET31, PDE1, AMN1/CST13 and KAR1 are shown. A summary of the all of the ChIPs performed is in Table 1. In comparison to higher eukaryotes, most yeast promoters are relatively small and contain activator or repressor binding sites within several hundred base pairs of the start site of the open reading frame (ORF) of the gene. However, there are a few genes, like HO , whose regulation is controlled by a region several Kb long. Consequently, we did a separate search looking for additional a 1- α 2 binding sites that are within a region 1.5 Kb upstream of the ORF. Most of the sites identified in this search were well above the threshold value and were not bound by the a 1- α 2 complex in ChIP assays (data not shown). However, the search identified one site upstream of the AMN1/CST13 gene that was a potential target site. The ChIP analysis verified this as a functional target site for the a 1- α 2 complex in vivo (Figure 1 , Table 1 ). Analysis of haploid-specific genes predicted not to be bound by a 1- α 2 Among the top 35 genes in the list ranked by haploid-specific expression score, more than half do not appear to contain an identifiable a 1- α 2 site by our analysis (lack of a significant binding site being defined as binding p-value greater than 2 × 10 -4 ) (Table 2 ). Although there were no apparent strong affinity a 1- α 2-binding sites in the promoters of these genes, it is possible that there are several weak affinity sites in the promoter that were not identified in the search. If present, these sites may work cooperatively to increase binding by the a 1- α 2 complex to the promoter and therefore repress transcription. In support of this model we have found that under some conditions weak affinity a 1- α 2 sites in the HO promoter have a role in repression of the promoter (Mathias and Vershon, unpublished). Promoters with a number of weak affinity sites may therefore be directly regulated by the a 1- α 2 complex. However, it is also possible that these genes are indirectly repressed by a 1- α 2, through its ability to repress expression of another transcription factor that is required for expression of the genes identified in the microarray. To distinguish between these possibilities we assayed for a 1- α 2 binding to these promoters by ChIP (Figure 2 and Table 2 ). As predicted from the site identification analysis, the a 1- α 2 complex did not appear to bind to most of these promoters. This result suggests that these genes are not directly repressed by the complex. The one exception to our predictions was that the a 1- α 2 complex appeared to weakly bind to the NEM1 promoter in the ChIP assays (Figure 2 ). Interestingly, NEM1 is downstream of GPA1 , a gene that is strongly repressed by the a 1- α 2 complex (Fig 1 ). Binding and repression by a 1- α 2 at GPA1 may help binding to weak sites in the NEM1 promoter. Alternatively, although we sheared the DNA used in the ChIP to an average of less than 500 bp, it is possible that there may have been some fragments that spanned the ~2 kb between the genes, thereby giving a positive result in the ChIP assay. Table 2 Haploid-specific Genes that Do Not Contain a 1- α 2 Target Sites ORF Gene Expression p-val a Binding p-val b Combined p-val a 1- α 2 ChIP c YIL117C PRM5 0.0011 0.3690 0.0004 - YLR159W 0.0015 0.4050 0.0006 - YBR051W 0.0022 0.4390 0.0010 - YLR080W EMP46 0.0024 0.6757 0.0016 - YIR039C YPS6 0.0025 0.7843 0.0020 - YCL014W BUD3 0.0027 0.3475 0.0009 - YPL189W GUP2 0.0030 0.3708 0.0011 - YJL077C ICS3 0.0032 0.9540 0.0030 - YML042W CAT2 0.0033 0.7564 0.0025 - YHR004C NEM1 0.0035 0.6396 0.0022 + YFR046C CNN1 0.0036 0.5712 0.0020 - YDR220C 0.0038 0.5919 0.0022 - YPL025C 0.0041 0.6844 0.0028 - YBR006W UGA2 0.0043 0.7684 0.0033 - YBR108W 0.0044 0.4390 0.0019 - YFL034W 0.0046 0.3287 0.0015 - YCL027W FUS1 0.0047 0.7170 0.0034 - YLR308W CDA2 0.0049 0.3424 0.0017 YGR014W MSB2 0.0050 0.6554 0.0033 YJL202C 0.0052 0.6416 0.0033 a The ranking of haploid-specific gene expression determined by analysis of microarray data [11]. b The ranking of potential a 1- α 2 target of haploid-specific genes. c A + indicates that the a 1- α 2 binds to the promoter by ChIP assay. Figure 2 ChIP assays of a 1- α 2 binding to promoter fragments that are not predicted to be targets of the complex. ChIP assays were performed as described in Figure 1. ChIP assays for a 1- α 2 binding to the FUS1 , BUD3 , NEM1 , CAT2 , YFL034W and GUP2 promoters are shown. Comparison with binding site identification by the weight matrix method We compare the performance of our algorithm with that of the weight matrix method [ 15 - 18 ]. In our study, we derived our parameters from a set of artificial sequences. Usually, the weight matrix has to be constructed from a set of known sites. We calculate the weight matrix for a 1- α 2 from regulatory elements upstream several known target genes: HO, GPA1, FUS3, AXL1, STE5, RME1 and MATα1 . As usual, one is faced with a choice of threshold weight matrix score for selecting putative sites in the yeast genome. For a stringent threshold that corresponds to the top 16 targets, we recovered all the genes, other than RME1 , used in construction of the weight matrix. However, we did not recover most of the other genuine targets identified, and verified, in this study. If we set the threshold to be lax enough to include RME1 , we obtained 55 candidate genes, including STE18 and RDH54, but still miss targets like STE4 . It is likely that most of the 55 putative targets are false positives, as evidenced by lack of haploid-specific regulation in the corresponding gene expression data. Overall, we find our method to be more successful than the weight matrix method. The use of mutational data as opposed to literature based data for sequence preference possibly accounts for part of the success (an advantage we may not have for some other transcription factors). However, much of our success has to do with cutting down of false positive rates by using microarray data judiciously. Analysis of all potential a 1- α 2 target sites in the genome Among the genes identified in the computational analysis, there is a good correlation between the presence of strong a 1- α 2-binding sites in their promoter region and repression in diploid cells. This raises the question of whether all a 1- α 2-binding sites function as repressor sites. To address this question we searched for all potential binding sites in the yeast genome. As expected, many of the best sites are in the promoters of known or previously identified haploid-specific genes (Table 1 ). However, we also identified a number of putative a 1- α 2-binding sites within ORFs (Table 3 ). To test if the a 1- α 2 complex is able to bind to these sites we performed electrophoretic mobility shift assays (EMSAs) with purified α 2 and a 1 proteins and radiolabeled oligonucleotides containing these sites (Fig 3A ). The a 1- α 2 complex bound to sites from the YKL162C , CDC25 , PRM8 , PRM9 , and URB1 ORFs with weaker affinity than to a strong binding site from the HO promoter, HO(10) . However, these sites did have slightly better binding affinity than to the HO(8) site, which we have shown is unable to repress transcription on its own (Mathias and Vershon, unpublished). Table 3 Potential a 1- α 2 target sites in ORFs ORF Gene Expression p-val a Binding p-val b Combined p-val a 1- α 2 ChIP c EMSA d YKL014C URB1 0.621 0.042 0.025 - 25× YGL053W PRM8 0.291 0.011 0.002 - 5× YAR031W PRM9 0.891 0.013 0.012 - 10× YKL162C 0.782 0.017 0.014 - 10× YLR310C CDC25 0.102 0.018 0.002 - 5× YPL061W ALD6 0.597 0.018 0.009 - YBR028C 0.793 0.046 0.036 - YOL022C 0.563 0.023 0.013 - YJR016C ILV3 0.248 0.025 0.006 - YBR218C PYC2 0.324 0.027 0.008 - YFR040W SAP155 0.234 0.165 0.038 - YMR269W 0.332 0.049 0.016 - YJL129C TRK1 0.297 0.083 0.024 - YCR053W THR4 0.607 0.073 0.044 - a The ranking of haploid-specific gene expression determined by analysis of microarray data [11]. b The ranking of potential a 1- α 2 target of haploid-specific genes. c A + indicates that the a 1- α 2 binds to the promoter by ChIP assay. d Relative binding affinity in fold decrease in affinity compared to the HO(10) binding site. Figure 3 a 1- α 2 binding in vitro and in vivo to sites in the ORF regions of the genome. (A) An EMSA of purified a 1 and α 2 proteins binding to a strong a 1- α 2 binding site, HO(10) (lanes 1–8) and sites from the YKL162C (lanes 9–13), CDC25 (lanes 14–18), a weak a 1- α 2 binding site from the HO promoter, HO(8) (lanes 19–23), PRM8 (lanes 24–28), PRM9 (lanes 29–33) and URB1 (lanes 34–38). The concentration of the a 1 protein was held constant at 1.4 × 10 -6 M (lanes 2 and 4–38) and mixed with 5-fold dilutions of the α 2 protein starting at 8.2 × 10 -8 M (lanes 3, 4, 9, 14, 19, 24, 29 and 34)). The EMSAs shown are phosphorimages of the gels. Lane 1 contains HO(10) probe alone, and lanes 2 and 3 contain 1.4 × 10 -6 M a 1 and 8.2 × 10 -8 M α 2 respectively. The fold repression by each site in the context of a heterologous promoter is shown below each gene/promoter. (B) ChIP assays for the genes YKL162C , CDC25 , PRM8 , PRM9 and URB1 are shown. ChIP assays were performed as described in Figure 1 Since the a 1- α 2 complex was able to bind to these sites with weak to moderate affinity in vitro, it is possible these sites may partially repress transcription on their own. To test this model, we cloned these sites into the context of the CYC1 promoter driving expression of a lacZ gene and measured the ability of the sites to repress transcription of the reporter in diploid cells [ 9 ]. The sites from the CDC25 and URB1 ORFs did not repress transcription of the reporter promoter in diploid cells (Fig 3A ). However, the site from PRM8 ORF, which showed the highest binding affinity among the sites found in ORF regions, weakly (2.8-fold) repressed the reporter promoter. This result indicates that this site can function as a repressor site in vivo if placed in the proper context. We next tested whether a 1- α 2 bound to these sites in the normal genomic context in vivo by ChIP assays. None of the sites in the ORF regions were bound by the a 1- α 2 complex (Fig 3B and Table 3 ). This result indicates that while they are competent for weak binding and repression in a heterologous promoter, they are unable to repress transcription in their normal genomic context. Our search also identified several potential a 1- α 2 binding sites in the promoter regions of genes that do not appear to be repressed in diploid cells (Table 4 ). Only the COX13 site had moderate binding affinity for the a 1- α 2 complex in the EMSAs (Fig 4A ). However, despite the relatively weak binding affinity of these sites, they were able to partially repress transcription of the reporter in diploid cells (Fig 4A ). In particular, the sites from the COX13 and REX2 promoters showed significant levels of repression. Interestingly, although these sites functioned as repressor sites in the context of the heterologous reporter, except for the COX13 promoter, most of these sites were not bound by the a 1- α 2 complex at their genomic locations by ChIP assays (Fig 4B ). These results suggest that the genomic context of most of these a 1- α 2 sites prevents binding by the complex. Table 4 Potential a 1- α 2 Target Sites in the Promoters of Non-Haploid-specific Genes ORF Gene Expression a p-val Binding p-val Combined p-val a 1- α 2 ChIP b YGL191W COX13 0.4111 0.0053 0.0021 + YPL099C FMP14 0.1622 0.0075 0.0012 YLR059C REX2 0.1532 0.0098 0.0015 +/- YDR212W TCP1 0.5694 0.0017 0.0010 YJL124C LSM1 0.0830 0.0120 0.0010 +/- YAR033W MST28 0.7419 0.0133 0.0099 - YMR015C ERG5 0.6021 0.0138 0.0083 - YMR291W 0.2659 0.0218 0.0058 - YPL188W POS5 0.2648 0.0230 0.0061 +/- YDR101C ARX1 0.3694 0.0298 0.0110 + YHR058C MED6 0.4507 0.0350 0.0157 YGL117W 0.5577 0.0370 0.0206 YIL027C KRE27 0.5762 0.0381 0.0220 a The ranking of haploid-specific gene expression determined by analysis of microarray data [11]. b A + indicates that the a 1- α 2 binds to the promoter by ChIP assay. A +/- indicates weak (2-fold) enhancement of the band in a / α cells by ChIP assay. Figure 4 a 1- α 2 binding in vitro and in vivo to putative binding sites in promoters associated with genes which are not expressed in a haploid-specific manner. (A) An EMSA of purified a 1 and α 2 proteins binding to a strong a 1- α 2 binding site, HO(6) (lanes 1–7), COX13 (lanes 8–11), REX2 (lanes 12–15), LSM1 (lanes 16–19) and FMP14 (lanes 20–23). The concentration of the a 1 protein was held constant at 1.4 × 10 -6 M (lanes 2 and 4–23) and mixed with 5-fold dilutions of the α 2 protein starting at 8.2 × 10 -8 M (lanes 3, 4, 8, 12, 16 and 20). The EMSAs shown are phosphorimages of the gels. Lane 1 contains HO(6) probe alone, and lanes 2 and 3 contain 1.4 × 10 -6 M a 1 and 8.2 × 10 -8 M α 2 respectively. The fold repression by each site in the context of a heterologous promoter is shown. (B) ChIP assays for COX13 , REX2 , LSM1 and FMP14 were performed as described in Figure 1. Discussion Genome-wide gene expression data using SAGE or DNA microarrays has provided a wealth of information on the regulation of genes under certain conditions or by specific transcription factors. The combination of this information with sequence analysis programs has enabled researchers to identify potential regulatory sites. For example, in a pioneering paper, Tavazoie et al. clustered expression data and used multiple local sequence alignment algorithms on the promoter regions of the co-clustered genes to discover regulatory motifs [ 19 ]. This approach has been further refined by using Bayesian networks to incorporate additional constraints regarding relative positions and the orientations of the motifs [ 20 ]. Another approach has been to break the genes into modules and perform module assignments and motif searches at the same time via an expectation maximization algorithm (as opposed to clustering first and finding motifs later) [ 21 , 22 ]. Although these approaches have worked well at identifying potential targets sites one drawback is that the expression patterns have to cluster well for these methods to work. For a small number of microarray experiments, this may always not be the case. A method that does not utilize clustering is a regression model based analysis to locate "words" in the promoter that correlates with modulation of expression [ 23 ]. However, this approach is restricted to retrieving functional consensus binding sites in the promoter regions and for transcription factors with low sequence specificity, this approach needs to be modified. Most of these approaches attack the difficult problem of what to do when relatively little is known about the regulatory system and sequence recognition by the protein. Consequently they develop pattern recognition algorithms that are essentially unsupervised. Our focus has been to take advantage, as much as possible, of knowledge about the biological system and use that information combined with expression analysis to identify potential target sites. The minor loss of generality of the tools resulting from such an approach is more than offset by its predictive power. To determine if the changes in expression of a specific gene are the result of a transcription factor working at the promoter we developed an algorithm that combines expression data with information on the binding site preference for a transcription factor. As a test for this algorithm we identified genes in yeast that are direct targets for regulation by the a 1- α 2 repressor complex. We also used this method to identify genes that are repressed in diploid cells but that are not direct targets of the complex, as well as functional a 1- α 2 binding sites that do not appear to repress transcription in their genomic context. The combination of these sets of findings has provided insight into the regulatory network and mechanism of repression by the a 1- α 2 complex. The primary goal of this study was to identify genes that are direct targets for repression by the a 1- α 2 complex. There are two major functional subsets among the a 1- α 2 target genes identified in this analysis (Table 1 ). One, not surprisingly, involves genes that are required for various processes in mating of the two haploid cell-types. These include components of the mating pheromone signal transduction pathway, such as GPA1 , STE18 , STE4 , and STE5, which are activated in response to the binding of pheromone from the other cell type [ 24 ]. This group also includes genes further down that pathway, such as FAR1 and FUS3 , which are required for cell-cycle arrest before mating. A number of these genes have previously been shown or suspected to be under the control of a 1- α 2 repressor complex [ 25 , 26 ]. Repression of these genes in diploid cells is biologically important because it prevents further mating by diploid cells. If diploid cells mate they would form triploids or higher ordered genomic polyploids, which are genetically unstable during meiosis and therefore detrimental to cell survival. The second subset of genes identified in the analysis is associated with mating type switching and recombination. The HO gene is a known target of the a 1- α 2 complex and its promoter contains 10 binding sites of varying affinity [ 25 ]. Repression of HO is essential in diploid cells because it prevents switching of one of the MAT loci to form homozygous a / a or α / α diploid cells. Although diploid in genomic content, cells homozygous for the MAT loci are competent to mate and therefore would form higher order genomic polyploids that are genetically unstable. We have also shown that NEJ1 , which is involved in non-homologous end-joining (NHEJ), is a direct target for the a 1- α 2 complex [ 27 , 28 ]. It has been proposed that that repression of the NHEJ pathway may promote homologous recombination and crossing over in diploid cells. In addition, we found that RDH54 , a gene involved in double-stranded DNA break repair, is a direct target for the a 1- α 2 complex [ 29 ]. This result is somewhat unexpected because RDH54 is required for meiosis and null mutants show significantly reduced spore viability. It is likely that the a 1- α 2 complex only partially reduces the level of expression of the gene and that diploid cells require a lower level of activity of the protein. We also identified several genes that fell outside of these two subsets. One is RME1 , which encodes a transcriptional repressor of IME1 , the master regulator of meiosis [ 30 - 32 ]. a 1- α 2-mediated repression of RME1 is required to allow cells to enter the meiotic pathway in diploid cells. Interestingly, we also found that PDE1 and MET31 are weakly, but reproducibly, direct targets for repression by the a 1- α 2 complex. The Pde1 protein is a low affinity cAMP phosphodiesterase that appears to have a role in response to stress and cell aging [ 33 ]. Repression of PDE1 in diploids may partially account for the difference of starvation response between haploids and diploids. Met31 is a zinc finger DNA-binding protein that activates genes involved in sulfur metabolism [ 34 ]. It is unclear why this gene would be a target for the a 1- α 2 complex. It is possible that the presence of an a 1- α 2 target site upstream of a gene that has lower expression in diploid cells was fortuitous and that these sites were not functional targets. However, if this was the case then there would be little pressure to conserve these binding sites through evolution. Several closely related species of yeast have been sequenced and comparison of the corresponding promoter regions has led to the discovery of conserved regulatory motifs [ 35 , 36 ]. Although lack of conservation does not imply non-functionality, significant conservation strongly argues for functionality of a putative regulatory element. To investigate this possibility, we performed a phylogenetic comparison to infer whether these sites are preserved among six sequenced Saccharomyces species using the PhyloGibbs program [ 37 ]. The program identified the a 1- α 2 binding site among a promoter set including many known haploid-specific genes ( HO, NEJ1, GPA1, STE4, and STE18 ). This analysis also showed that the a 1- α 2 binding sites in the RDH54, PDE1 and MET31 promoters are strongly conserved among multiple species, suggesting that these sites play an important functional role. Our analysis identified a number of haploid-specific genes that do not appear to be direct targets of the a 1- α 2 repressor complex (Table 2 ). Genes in this list do not contain a recognizable a 1- α 2-binding site and, with the exception of NEM1 , are not detectably bound by the a 1- α 2 complex in the ChIP assays. It is possible that a 1- α 2 indirectly turns off these genes by repressing an activator protein that is required for their expression. However, besides MET31 , there were no obvious genes coding for activator proteins that were direct targets of the a 1- α 2 complex. It is possible that the haploid-specific genes without a 1- α 2 sites are indirectly repressed through more complex mechanisms that involve repression of RME1 . We also identified potential a 1- α 2-binding sites in the genome that do not appear to repress expression of nearby genes. Although sites from the PRM8 , PRM9 , CDC25 , and LSM1 promoters appear to be moderate binding sites for the a 1- α 2 complex in vitro, ChIP and heterologous reporter assays showed these sites are neither bound by the proteins nor are functional repressor sites in vivo. Many of these sites lie in open reading frames of actively transcribed genes and so it is possible that transcription through the binding site or the chromatin structure of the region prevents high affinity binding by the complex. The model that the genomic context of these sites is important for their regulatory activity is further supported by our results that show that some of these sites, such as COX13 and REX2 , function as strong a 1- α 2 dependent repressor sites in the context of the heterologous promoter. Although a 1- α 2 complex is bound to the COX13 site in vivo it does not appear to repress transcription of this gene in diploid cell. Interestingly, this binding site is very close to the end of the coding region of IME4 , an inducer of meiosis that is expressed in diploid cells [ 38 ]. The IME4 gene is only expressed in diploid cells and it was thought that the a 1- α 2 complex may be indirectly activating its expression by repressing a repressor protein, such as RME1 . However, the fact that a 1- α 2 binds to the downstream region of this gene suggests that it may play a direct role in its expression. Our data shows that the algorithm we have developed is useful in sorting between direct and indirect targets of a transcription factor. Although we have used mutational data to define the binding site for the a 1- α 2 complex, in principal binding site sequences derived from site selection experiments may also be used. This analysis may also complement genome-wide ChIP studies to identify the target sites of the transcription factor. Conclusions In summary, we show that combining microarray data with motif analysis, lets us distinguish between the genes that are direct targets of a transcription factor and those that are modulated because of secondary effects. We get excellent agreement of the computational predictions with location analysis by ChIP experiments. We find most of the direct targets of a 1/ α 2 complex to be involved in the mating pathway, mating type switching, recombination and meiosis. We also found a few weak targets that are possibly involved in sensing and control of the metabolic state. We also see that the sites we predict solely based on single species data are often evolutionarily conserved in other species of Saccharomyces. Methods Combined scoring of genes from microarray data and mutational analysis We define a scoring algorithm that ranks gene expression patterns. For gene g , the score is given in terms of the expression in different types of cells ( a , α and a / α ) Score( g ) = sgn(( X a ( g ) + X α ( g ))/2 - X a / α ( g )) [( X a ( g )) + X α ( g ))/2 - X a / α ( g )] 2 - A ( X a ( g ) - X α ( g )) 2 . We initially used the logarithm of expression level of the gene g for the three cell types for the variables X t ( g ) with t indicating the type. We have since found that using the complete expression data for the polyploids is a better strategy. In the final results, shown in the paper, X a ( g ) is the average of the log expression for a , aa and aaa . Likewise, X α ( g ) is the average from α, αα and ααα . X a / α ( g ) comes from averaging log expression over aa α , a αα and aa αα . The polyploid averaged quantities tend to be less noisy (demonstrated, for example, by the quantities X a and X α being close to each other for the generic gene, which is not regulated by cell type. This, in turn, allows easier detection of genuine haploid-specific targets. An explanation of the purpose served by different terms in the overall score is described below. The first term scores well when expression in diploids is lower than the average expression in haploids. The second term penalizes the gene if the expressions in different types of haploids are very different. A is chosen to be large enough so that known a -specific genes and α -specific genes score worse than known haploid-specific genes, but not large enough to overwhelm the first term. The optimal A is about 10. Comparison of the performance of our algorithm for A = 1 and A = 10, shows that the biologically known sites almost always stay near the top but further down in the list the second choice is better. The exception is a special gene: MATα 1. Since MATα 1 is not present in the MAT a type cell, there is a penalty for expression patterns. Cumulative probability for any gene to have higher score than a gene g is P exprn ( g ), namely, fraction of genes with score higher than g. This scoring function would rank haploid-specific genes high but may not select out genes that are directly regulated by the a 1- α 2 repressor complex. In order to select for genes with an upstream region with a strong a 1- α 2 repressor, we used the binding site mutational data available [ 9 ]. In these experiments, repression of a heterologous promoter, incorporating single site mutations of a consensus binding site of the a 1- α 2 repressor, was measured. Under the assumption that the degree of repression is inversely proportional to how often that site is occupied, we derived the expression: 1/Repression ∝ [1 - 1/(1 + e βE ( S ) / z )] ≈ e βE ( S ) / z assuming near saturation of binding. The symbol z represents the fugacity and β is inverse of k B T , k B being the boltzman constant. The binding (free) energy is given by E ( S ) = Σ ib ε ib S ib , within the single base model [ 15 , 16 ]. The index i runs over the positions in the motif and b runs over the bases A, C, G, T. S ib is 1 or 0 depending upon whether the i- th base is b or not. The parameters ε ib represent effects of single base changes on the binding (free) energy. They are related to the weight matrix parameters [ 17 , 18 ] widely used to characterized variable motifs. Note that e βE ( S ) / z would more commonly be represented as ( K / [Protein])•exp(Δ G ( S )/ RT ) in the biochemistry literature [ 18 , 39 ]. The independent base model is only an approximation and mutations in nearest neighbor sites could produce effects that we cannot estimate from the existing data. There is a better separation between well-known sites and generic sequences if an extra penalty is added to the score for neighboring base pairs which both differ from the consensus. In this way every base different from consensus and neighboring another base different from consensus draws an additional penalty to the binding energy score. This parameter was set to be ln(2), by experience. Although this method prevented many false positives, it also penalized a few genuine candidates, such as the binding sites in the promoters of FAR1 and MATα 1. Thus, from the effect of the single base mutations, we estimated the parameters ε ib . Armed with these parameters, we found the probability P ( E \| ε , L ) that a random sequence of a certain length, L , would have a subsequence of binding energy greater than E . For gene g, the strongest site in the upstream region of length L would have binding energy E g . Low values of P binding ( g ) = P ( E g \| ε , L ), indicated the presence of a good binding site. The genes were ordered according to the lowest value of a combined p-value, P exprn ( g ) P binding ( g ), and then ranked as candidates for haploid-specific genes that are directly repressed by the a 1- α 2 complex. One of the issues in such studies is how to decide on how many of the top candidates are significant. This problem occurs even in solely, sequence based analysis as well [ 40 ]. In our study, we generated random combinations of expression p-values with scrambled binding p-values, so that we could choose a cutoff threshold by comparing the ordered p-values with the ordered "scrambled" p-values. Figure 5 plots these sorted p-values against average (log) sorted p-values for the random combinations. The comparison suggests that only about 10–15 top candidates in the list are significant. Weight matrix based search A weight matrix [ 15 - 18 ] search for binding sites was performed using a set of known sites [ 9 ] to construct the matrix. Each matrix entry, w ia , was set to log(( f ia + δ )/( P a + δ )), where f ia is the frequency with which base a appears in the i th position in the known sites, P a is the frequency with which base a appears in the promoters of genes and δ is a small number added to ensure that the weight matrix score is finite even when f ia = 0. Each subsequence, S ia , of length 20 in each promoter was assigned a weight matrix score Σ ia S ia w ia . After a threshold score is chosen, sites scoring above that threshold are declared to be binding sites. Automated primer generation An automated procedure for generating primers flanking a specified site in the genome sequence, σ , was implemented. To each pair of numbers, d u , and d d , representing primer distances upstream and downstream of the candidate binding site respectively, and primer lengths l u , and l d , a score is assigned via S ( d u , d d , l u , l d , σ ) = - Σ a k a ( P a ( d u , d d , l u , l d , σ ) - ) 2 where the P a are functions including the melting temperatures, distances of upstream and downstream primers from the candidate site, total length of the bound region, and the difference between the primer melting temperatures. The s are the preferred values of those functions. The k a are adjusted to reflect the relative importance of the parameters; for example it is more important that the difference in melting temperatures be close to zero than it is that the distance to the upstream primer match the distance to the downstream primer. Values of d u , d d , l u and l d are restricted to those whose corresponding primers have GG, GC, CG, or CC at the end nearest the candidate site. Primers are identified by selecting the values of d u , d d , l u and l d which maximize S . [A web-based interface to this algorithm is available at ] Chromatin immunoprecipitation Chromatin immunoprecipitation (ChIP) was carried out as described previously [ 41 ] with the following modifications. One liter of JRY103 ( MATα/MAT a ade2-1/ADE2 HIS3/his3-11,15 leu2-3,112/leu2-3,112 trp1-1/trp1-1 ura3-1/ura3-1 ash1Δ::LEU2/ash1Δ::LEU2 ) and JRY118 ( MATα/mat a Δ::TRP1 ade2-1/ADE2 HIS3/his3-11,15 leu2-3,112/leu2-3,112 trp1-1/trp1-1 ura3-1/ura3-1 ash1Δ::LEU2/ash1Δ::LEU2 ) cultures were grown to an A600 of 0.5 and treated with 1% formaldehyde for 20 min at RT on a rotating shaker at low speed. Cells were collected, washed 2X with cold 1XTBS. Equal volumes of cells were aliquoted into ten 1.5 ml microfuge tubes, washed once with 1.5 ml of cold 1X TBS. The pellets in each tube were resuspended with 400 μ l of lysis buffer (50 mM HEPES, pH 7.4, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% Na-Deoxycholate) plus 1 mM PMSF, 1 mM benzamidine, and 1X Protease inhibitor cocktail from Roche (Cat No. 1873580) and also manufacturer recommended concentration of protease inhibitor cocktail from SIGMA (Cat No., P 8215). To this 200 μ l of glass beads were added to each tube and lysed using a multitube vortexer at full speed for 30 min at 4°C. The lysate was transferred in a new tube and 400 μ l of lysis buffer was added and vortexed briefly. The lysates were centrifuged at 12,000 g for 10 min at 4°C and the supernatants were sonicated at 30% output for four 10 sec cycles with intermittent cooling on ice. The lysates were cleared by centrifugation at 12,000 g for 10 min and 1 mM PMSF was added to the samples. A 1/10 th volume aliquot was removed and frozen to be used as total chromatin control. The remaining sample was precleared by the addition of 25 μ l recombinant protein G-agarose beads, incubated while nutating for 30 min and the supernatant was collected after centrifugation at 12,000 for 5 min. 1 μ l of rabbit anti- α 2 antiserum (a gift from A. Johnson, UCSF) was added to each supernatant of the samples and incubated 12 h on a nutator at 4°C. To immunopreciptiate α 2 50 μ l of recombinant protein G agarose beads (Roche) was added to the samples and nutated for 90 minutes at 4°C. The protein G beads were pelleted, washed once in low salt buffer (0.1%SDS, 1% Triton X-100, 20 mM Tris pH8.0, 2 mM EDTA and 150 mM NaCl), once in high salt (composition same as lowsalt + 500 mM NaCl), once in LiCl buffer (0.25 M LiCl, 1% IGEPAL, 1XTE and 1% Na-Deoxycholate) and twice with 1XTE (pH8.0). The immunoprecipitated DNA was eluted twice with 250 μ l of elution buffer (1%SDS and 0.1 M NaHCO3) and the eluates were pooled (500 μ l final volume). To this 20 μ l of 5 M NaCl was added and incubated 12 h at 65°C. To remove the crosslinks, 10 μ l of 0.5 M EDTA, 20 μ l of 1 M Tris-HCl, pH 7.5 and 2 μ l of proteinase K (10 mg/ml) was added and incubated for 45 minutes at 45°C. The DNA samples were extracted once with Phenol:chloroform:Isoamylalcohol and the DNA was ethanol precipitated, washed once with 70% ethanol and resuspended in 50 μ l (IP) or 500 μ l (TC) TE. Purified DNA from the immunoprecipitated samples was subjected to multiplex PCR amplification with primers specific for the STE6 promoter as a positive control for the immunoprecipitation of α 2 and the YDL223C ORF as a negative control for nonspecific immunoprecipitation, along with the specific primers for candidate α 2- a 1 target sites. PCRs were carried out in 50 μ l containing 10 pmols of each primer, 0.2 mM dNTPs, 2 mM MgCl2, 1X Eppendorf Taq buffer, 0.5X Taq Master buffer and 2.5 U of Eppendorf Taq polymerase. The amplifications were carried out at 94°C for 1 min and 30 secs, followed by 25 cycles of 94°C for 30 secs, 52°C for 1 min, and 72°C for 30 secs and a final extension step of 7 min at 72°C. The PCR products were separated on 2.5% agarose gels. Electrophoretic mobility shift assays Oligonucleotides containing the predicted a 1- α 2 binding sites from within the ORFs of URB1 , PRM8 , PRM9 , YKL162C and CDC25 and the promoters of COX13, REX2, LSM1 , and FMP14 were synthesized, one strand was end-labeled with [ γ - 32 P]-ATP, and then annealed with excess cold complementary oligonucleotide. The HO(10 ) and HO(8 ) a 1- α 2 sites within Upstream Regulatory Sequence 1 (URS1) of the HO promoter were used as strong and weak binding sites respectively. The EMSA was performed as described previously [ 42 ], using a constant 1.4 μ M a 1 and five-fold titrations of α 2 starting at 82 nM in protein dilution buffer (50 mM Tris pH 7.6. 1 mM EDTA, 500 mM NaCl, 10 mM 2-mercaptoethanol, 10 mg/ml bovine serum albumin). β -galactosidase assays Oligonucleotides containing a 1- α 2 binding sites were synthesized with 5' overhangs to allow cloning into the Xho I site of pTBA23 (2 μ URA3 Amp r ), a reporter plasmid containing a CYC1-lacZ fusion [ 43 ]. Reporter constructs were transformed into JRY103 and JRY118 and the β -galactosidase activity was measured on three independent transformants, as described previously [ 14 ]. List of Abbreviations SAGE – Serial Analysis of Gene Expression HD – Homeodomain ORF – Open Reading Frame ChIP – Chromatin immunoprecipitation NHEJ – Non-Homologous End Joining TC – Total Chromatin IP – Immunoprecipitated sample EMSA – Electrophoretic Mobility Shift Assay PCR – Polymerase Chain Reaction PMSF – Phenyl Methyl Sulfonyl Fluoride EDTA – Ethylenediaminetetraacetic Acid Author's Contribution VHN carried out the ChIP experiments, and wrote parts of the first draft of the manuscript. RAO did the bioinformatics analysis, and wrote programs facilitating primer design. ARB did EMSA and beta-gal assays. JRM constructed the strains and contributed to the design of the ChIP experiment. AKV and AMS supervised and coordinated the computational and the experimental research as well as prepared the manuscript. All authors contributed to the manuscript and approved the final version. Figure 5 Significance of combined p-values. Natural logarithm of combined p-values for twenty permutations/scrambles generated, sorted and plotted (in blue) against average of log p-value. The genuine combined value is plotted in red.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517709.xml
509279	Molecular dissection of the human antibody response to the structural repeat epitope of Plasmodium falciparum sporozoite from a protected donor	Background The circumsporozoite surface protein is the primary target of human antibodies against Plasmodium falciparum sporozoites, these antibodies are predominantly directed to the major repetitive epitope (Asn-Pro-Asn-Ala) n , (NPNA) n . In individuals immunized by the bites of irradiated Anopheles mosquitoes carrying P. falciparum sporozoites in their salivary glands, the anti-repeat response dominates and is thought by many to play a role in protective immunity. Methods The antibody repertoire from a protected individual immunized by the bites of irradiated P. falciparum infected Anopheles stephensi was recapitulated in a phage display library. Following affinity based selection against (NPNA) 3 antibody fragments that recognized the PfCSP repeat epitope were rescued. Results Analysis of selected antibody fragments implied the response was restricted to a single antibody fragment consisting of V H 3 and V κ I families for heavy and light chain respectively with moderate affinity for the ligand. Conclusion The dissection of the protective antibody response against the repeat epitope revealed that the response was apparently restricted to a single V H /V L pairing (PfNPNA-1). The affinity for the ligand was in the μM range. If anti-repeat antibodies are involved in the protective immunity elicited by exposure to radiation attenuated P. falciparum sporozoites, then high circulating levels of antibodies against the repeat region may be more important than intrinsic high affinity for protection. The ability to attain and sustain high levels of anti-(NPNA) n will be one of the key determinants of efficacy for a vaccine that relies upon anti-PfCSP repeat antibodies as the primary mechanism of protective immunity against P. falciparum.	Background Malaria threatens public health in regions of the world where more than a third of the human population lives [ 1 , 2 ]. It has been shown that immunization with radiation-attenuated Plasmodium sporozoites, the infective stage of the malaria parasite, confers protective immunity [ 3 , 4 ]. The role of specific antibody in conferring protection was demonstrated with passive administration of murine mAbs directed against the major repeat epitope of the circumsporozoite (CS) protein [ 5 ] in a rodent model. The corresponding epitope of the human malaria parasite Plasmodium falciparum is contained within the repeat tetramer peptide (Asn-Pro-Asn-Ala) n , (NPNA) n [ 6 ]. In some studies of volunteers protected against malaria by immunization with radiation attenuated P. falciparum sporozoites, protected individuals had significant elevations of anti-repeat antibodies (>19 μg/ml) [ 7 ]. With the advent of recombinant combinatorial antibody technology [ 8 , 9 ] and phage display [ 10 - 13 ] it is possible to attempt to dissect the human antibody response against a wide range of pathogens. In order to further investigate the role of the human antibody response in P. falciparum sporozoite induced protection, a phage display library of antibody gene fragments isolated from the peripheral blood lymphocytes of such a protected donor (WR5) [ 7 ] was assembled. Recombinant antibodies against the PfCSP structural repeat (NPNA) 3 epitope were selected. Recognition was restricted to a single antibody designated PfNPNA-1, encoded by V H 3 and V κ I families. This restricted humoral response has implications for rational vaccine design and the potential use of this human monoclonal antibody to prevent P. falciparum infection. Methods RT-PCR of Immunoglobulin genes A human volunteer (WR5), who was previously exposed to the bites of γ-irradiated P. falciparum infected Anopheles mosquito's and subsequently shown to be protected against a non-irradiated parasite challenge, donated lymphocytes by leukophoresis five days after a booster challenge (appropriate informed consent was obtained) for details see Egan et al., [ 7 ]. The irradiated sporozoite immunization protocol was approved by the Naval Medical Research Institute's Committee for the Protection of Human Subjects in accordance with the US Navy regulation (SECNAVINST3900.39B) governing the use of human participants in medical research. Total RNA was extracted from 2 ml of packed cells using an RNA isolation kit (Stratagene, La Jolla, CA) with a modified protocol [ 9 ]. The equivalent of 2.5 μg total RNA template were used in each cDNA synthesis reaction using reverse transcriptase (Invitrogen, CA) with oligonucleotide oligo dT or 3 'HuVH (5'GCCCCCAGAGGTGCTCTTGGA-3', anneals in CH1 domain) following the instructions provided by the supplier. The genes encoding variable heavy (V H ) and the kappa chain (κ) were accessed by RT-PCR and combined by overlap extension PCR, resulting in shuffling of the V H and the V L domains. The V H PCR amplification was carried out with the cDNA template generated using the 3'HuVH primer. The V H domains were amplified using 5'HuVHA and 3'HuVH-Link 3' designed to anneal with the sequence corresponding to the first β-strand of the CH1 domain and overlap with the 5'HuVk primer. The κ chains were amplified using 5'HuVk and the 3'Hukappa primers. The V H and the κ chain PCR products were combined by overlap extension PCR using a V H flanking primer 5'HuVHB (to introduce a NheI site) and the 3'HuKappa primer. Oligonucleotide primer sequences 5'HuVk 5'-TATTAGCGGCCGCCCAACCAGCCATGGCCGAEFIJLOPETGACBCAGTCTCC-3' (where B=G+C+T, S=G+C, E = 50%A+33%C+17%T, F = 83%A = 17%G, I = 83%T+17%C, J = 50%T+33%C+17%G, L = 67%G+17%T+17%C, O = 67%T+17%A+17%C, and P = 83%G+17%C) 3'HuKappa 5'-TCCTGAAGCTTGACGACCTTCGATCTCTCCCCTGTTGAAGCTCTT-3' 5'HuVHA 5'-SAGGTGCAGCTGSTGSAGTCTGG-3' 5'HuVHlink3' 5'-GGCTGGTTGGGCGGCCGCTAATATGGAGGAGGGTGCCAGGGGGAAGAC-3' 3'HuVHB 5'-GTTTCGCTAGCGTAGCTCAGGCTSAGGTGCAGCTGSTGSAGTCTGG-3' The procedural steps are illustrated in Figure 1 . Figure 1 V H /κ library construction. A schematic diagram of the steps involved in constructing a V H /κ library from mRNA isolated from PBL. Cloning PCR fragments into pORFES and JC-M13-88 The PCR amplified V H /κ products were digested with restriction enzymes NheI and HindIII, and ligated into pORFES [ 14 ]. An aliquot of E. coli transformed with the ligation mixture was plated with and without carbenicillin selection, to determine the number of functional inserts. The V H /κ coding sequences are directionally inserted for expression between an OmpA leader peptide (to direct the polypeptide into the periplasm), and the β-lactamase. Functional full-length V H /κ β-lactamase fusion polypeptide is secreted into the periplasm. Bacteria harbouring plasmids conferring antibiotic resistance may be positively selected. The V H /κ coding insert may be readily transferred as a XbaI - HindIII fragment into the JC-M13-88 phage vector to display the insert polypeptide as a gpVIII fusion. The selected "functional" library of V H /κ inserts were excised from pORFES using XbaI and HindIII , ligated into pre-digested JC-M13-88 [ 4 ], and transformed into E. coli (XL1-Blue: Stratagene). Phage was produced overnight at 37°C in the presence of 1 mM IPTG, unless otherwise stated. A schematic outline of the vectors is shown in Figure 2 . Figure 2 Illustration of vectors pORFES, JC-M13-88 and pAbHIS. Phage panning The peptide (NPNA) 3 C (Chiron Mimotopes Peptide Systems, San Diego, CA.) was conjugated to BSA using Imject Activated Immunogen kit (Pierce, Rockford, IL) according to the manufacturers guidelines. ELISA plates (Dynatech Immunlon I, Alexandria, VA) were coated with BSA or (NPNA) 3 C-BSA and used in phage panning experiments essentially as described elsewhere [ 5 ]. To blocked antigen coated wells a total of 4 × 10 10 plaque forming units (pfu) of the phage library in dilution buffer (PBS pH 7.2, Tween-20 0.05%, BSA 0.1%, NaN 3 0.02%) was added (1 × 10 10 plaque forming units (pfu) per well). After 4 h the wells were washed and the bound phage were eluted by applying either 0.1 M glycine-HCl, pH2.2 or a solution of the free peptide (NPNA) 3 (~8 μM) dissolved in dilution buffer, for 15 min at ambient temperature. An aliquot of the phage elute was titered, and the remainder was used to propagate phage for further rounds of panning. The three-domain single chain antibody retains the kappa constant domain thus permits plaques filter lifts to be probed with anti-human kappa chain antibodies for immunodetection. V H and V L coding sequences were determined by sequencing of replicative form (rf) phage DNA prepared from κ-positive plaques, using the oligonucleotides primers: 3'Seq VH-JC130 (5'-CGGCCATGGCTGGTTGGGCGGCC-3') and 3'Seq VL-JC128 (5'TTCAACTGCTCATCAGATGGCGG-3'). Expression of PfNPNA-1 V H / k in E. coli The expression vector pAbHIS, was constructed by modification of pUC18. The β-galactosidase coding region was removed and XbaI - HindIII sites introduced upstream of a sequence encoding a six histidine tail. Insertion of V H /κ coding sequence selected by phage display as XbaI - HindIII fragment would result in the expressed polypeptide being secreted into the periplasmic space with a hexa-histidine tag. The plasmid pAbHIS was constructed by PCR modification of pUC18 using the primers PUCSpe-JC127(5'-TCATCATACTAGTAACGACACCCGCCAACACCC-3') and M13-JC118 (5'-AAGCTTATGATGTCTAGAGCTGTTTCCTGTGTGAA-3'). A pair of annealed oligonucleotides designed to encode a 6×His tag were ligated into the HindIII digested plasmid to complete pAbHIS. The selected PfNPNA-1 V H /κ gene was excised from the rf JC-M13-88 DNA by digestion with XbaI and HindIII and ligated into similarly digested pAbHIS. An additional 6×His-coding pair of oligonucleotides was ligated into the PfNPNA-1 V H /κ linker sequence as NotI-NcoI insert. The expression of PfNPNA-1 V H /κ in E. coli D29A1 cells at 25°C, and the isolation of bacterial periplasmic material was performed as described [ 16 ] with modifications; Dnase I n(1 μg/ml) and MgCl 2 (20 mM) were added, the bacterial suspension was incubated on ice for a further 20 min before final centrifugation step. The periplasmic extract was passed over Ni-NTA resin (Qiagen), washed and the PfNPNA-1 V H /κ was eluted with 300 mM imidazole. SDS PAGE and western blotting were used to asses purity and integrity of the expressed V H /κ polypeptide during the purification procedure (data not shown). Purified PfNPNA-1 V H /κ was quantified spectrophotometrically assuming an OD at 280 nm of 1 = 0.72 mg/ml protein. ELISA affinity and specificity determination ELISA Plates (Dynatech Immunlon I) were coated with (NPNA) 3 C-BSA (10 μg/ml). Dilutions of the peptide (NPNA) 3 were made in dimethyl formamide (DMF) before mixing with the PfNPNA-1 V H /κ diluted in PBST. Aliquots of 0.1 ml were added to duplicate wells, incubated for 2 h at 37°C. In all wells the final concentration of DMF was 1% (v/v). After washing 4 times with PBST, anti-human kappa chain alkaline phosphatase conjugate diluted 1:1000 in PBST was added and incubated as before. The wells were washed 4 × with PBST and rinsed 1× with PBS and substrate p -nitrophenyl phosphate was added, the absorbance was determined at 405 nm The binding of immune serum (WR5), non-immune serum and PfNPNA-1 V H /κ to R32tet32, recombinant hepatitis core containing (NANP) 4 peptide sequence and (NPNA)3C-BSA conjugate coated microtiter plate well was determined by ELISA essentially as described above. The serum(s) and the recombinant PfNPNA-1 V H /κ were diluted 1/16 and 1/10 respectively. Phage ELISA Phage at 1 × 10 12 pfu/ml in dilution buffer were applied (0.1 ml/well) to duplicate wells coated with (NPNA) 3 -C-BSA or BSA (10 μg/ml). After incubation at ambient temperature for 4 h, plates were washed with PBST. The bound phage was detected with sheep anti-M13 antibodies (5'-prime 3'-prime), followed by rabbit anti-sheep alkaline phosphatase antibodies in PBST added sequentially for 1 h at 37°C. Plates were washed and developed as described above. Indirect immunofluorescence assay (IFA) on P. falciparum sporozoites The PfNPNA-1 V H /κ was compared with a well-characterized murine monoclonal anti- Pf repeat antibody 2A10 [ 17 , 18 ] in IFA. All incubations were at 37°C in a humid container. Printed multiwell slides coated with Plasmodium falciparum NF54 strain sporozoites were either fixed in ice cold acetone for 10 min or used unfixed. Slides were first blocked with 4%BSA in PBS for 1 h. Antibodies diluted in PBST were applied for 2 h, then slides were washed 4× with PBS and fluoroscein-conjugated anti-human kappa chain or anti-mouse immunoglobulin (Sigma) was applied, diluted 1:25 in PBST. After 2 h slides were washed as above and mounted in SlowFade anti-fade reagent (Molecular Probes, Eugene, OR) and viewed by fluorescence microscopy. Other antibodies The murine mAb 2A10 [ 17 , 18 ] (IgG2b, κ), which recognizes the (NANP) 3 sequence of the P. falciparum CSP was provide as whole ascitic fluid (a kind gift from Dr P. Sinnis New York University). Concentration of the whole IgG was estimated using a standard antibody capture ELISA. Immune IgG (denoted (Vol-IgG) was purified from serum of the immune volunteer (WR5), donated at the time of lymphophoresis using Protein A Sepharose (Pharmacia) and quantified assuming OD at 280 nm of 1.0 represents 0.8 mg/ml IgG. Within the Vol-IgG, the proportion of (NPNA) 3 specific IgG with κ or λ light chains were determined by ELISA (data not shown). Results Library construction Sera from the protected individual (WR5) [ 7 ] contained antibodies against the PfCSP, which were predominantly IgG/κ and against the structural repeat peptide as determined by ELISA. Gene fragments encoding V H /κ single chain antibodies were amplified and assembled by PCR from cDNA derived from the peripheral blood lymphocytes of the immune donor WR5 (as outlined in Figure 1 ). The library of PCR amplified V H /κ sequences were inserted into pORFES [ 14 ] and an aliquot compared for number of functional inserts by selecting in the presence of either chloramphenicol (total transformation events) or chloramphenicol and carbenicillin (functional inserts). Approximately half of the initial library contained non-functional domains (data not shown). The remainder of the library was selected on 100 μg/ml carbenicillin, yielding a primary library of 1.3 × 10 6 members, these V H /κ sequences were transferred to the phage display vector JC-M13-88 [ 15 ] with ten fold over representation of the primary library. Panning Samples of the V H /κ-phage library were subjected to four rounds of panning on (NPNA) 3 C-BSA coated wells. Both the acid and peptide elution strategies yielded significantly greater numbers of phage after four cycles of panning on (NPNA) 3 C-BSA when compared to panning on BSA alone (Table 1 ). Analysis of fifteen individual phage after the fourth round of panning on (NPNA) 3 C-BSA eluted with free peptide revealed, twelve kappa positive phage, of these three clones (NP 04, 12, 13) were positive in the phage ELISA for binding to (NPNA) 3 C-BSA and were encoded by an identical sequence, henceforth denoted PfNPNA-1. Prior to panning ten kappa positive clones were randomly selected for sequencing (R 01-10; Table 2 ). The PfNPNA-1 V H and V L sequences were members of the V H 3 and V κ I families respectively and were not found amongst the random sampling of phage prior to panning. In an independent experiment with phage propagated at 30°C, but otherwise an identical panning procedure 12 out of 12 selected phage clones were identical to PfNPNA-1. Likewise, phage selected by acid elution and evaluated by ELISA for binding to (NPNA) 3 C-BSA were all identical to PfNPNA-1. Despite extensive sampling of phage that were positive in the phage ELISA for binding to (NPNA) 3 C-BSA (n = 25), only the PfNPNA-1 sequence was observed. Table 1 Phage panning experiments ELISA plates (Dynatech Immulon I) were coated with BSA or (NPNA) 3 C-BSA and used in phage panning experiments. To the blocked antigen coated wells a total of 4 × 10 10 pfu of the phage library in dilution buffer were added 1 × l0 10 pfu per well. After 4 h the wells were washed and phage eluted by applying either 0.1 M glycine-HCl pH 2.2 or a solution of the free peptide (~8 μM) (NPNA) 3 dissolved in dilution buffer for 15 min at ambient temperature. An aliquot of the phage eluate was titered and the output determined. Eluate after Coating antigen / Elution method (×l0 5 pfu)* panning rounds BSA / acid BSA /(NPNA) 3 (NPNA) 3 C BSA /acid (NPNA) 3 C BSA /(NPNA) 3 1 2.9 (0.38) 0.82 (0.032) 3.0 (0.34) 0.51 (0.024) 2 1.4(0.03) 0.24 (0.020) 3.5 (0.24) 1.5 (0.028) 3 1.2(0.06) 0.47 (0.020) 4.3 (0.024) 12 (0.68) 4 13 (0.70) 1.0(0.032) 170 (30) 370 (20) * Figures represent the mean of the total plaque forming units eluted by either acid or excess free peptide, after repeated panning against BSA or (NPNA) 3 C-BSA. Values for the standard deviation are shown in brackets (). Table 2 V H and V L assignments and alignment of CDR 3 sequences The selected (NP 04, 12, 13 designated Pf NPNA-1 bind to the repeat epitope), all other NP clones were randomly picked after the panning procedure and were subsequently shown not to be reactive with the repeat epitope. Non-selected (R01-10) were randomly picked from the library prior to initiating panning. The peptide sequence of the heavy and light chain complementarity-determining region 3 (CDR3) is shown below. V H /V L families, segments and the number of differences from germline segments were determined by using the V BASE sequence directory (Tomlinson, I. M., Williams, S. C., Corbett, S. J., Cox, J. P. L. & Winter, G., MRC Centre for Protein Engineering, Cambridge, UK) and the DNAPLOT alignment package (Müller, W. & Althaus, H.-H., Köln University) clone code* V H family V H Segment Differences from germline V H CDR3 V L family V L Segment Differences from germline V L CDR3 PfNPNAl VH3 DP46 10 DRDSSSYFDS VkI L12a 15 QQYNSYSGLT NP04, NP12, NP13 VH3 DP46 10 DRDSSSYFDS VkI L12a QQYNSYSGLT R01 VH1 4M28 †‡ 28(+6)* § - DSESVAQWRY VkIV DPK24 43 QQSLSPVWT R02 VH3 COS-3 ‡ 27 (+3)_ GVNWCSDY VkI DPK9 10 QQSYSTSWT R03 VH5 DP73 35 LYTSIYYFDS VkIV DPK24 7 QQYYSTPLT R04 VH3 DP46 8 DRVTNFWSGYFDY VkIII DPK22 13 QQYGSSPGFT R05 VH3 DP58 23 DSTVKTVTKMRYGLD V VkIII DPK22 8 QQYGSSPFT R06 VH1 4M28 † 12 DNYGDPGGGFDI VkIII DPK22 11 QQYGNSPRT R07 VH5 DP73 9 RFWFGELYDAFDI VkIV DPK24 16 HQYYSTPQT R08 VH5 DP73 34 LYTSIYYFDS VkIII DPK22 14 QQYGRSPWT R09 VH3 V3-21 † 34 DQGGGWSSEVDS VkIII Vg 5 QQRSNWPLT R10 VH1 DP7 ‡ 21 (+9)** ALYGHDAFDI VkI DPK4 12 PKYNSALHT NP02 VH3 DP47 36 ERPYDAFDS VkIII DPK22 23 QQYSTSPPMYN NP03 VH5 DP73 40 LYTSIYYFDS VkIII Vg 17 KQRSKWPPIT NP05 VH3 V3-48 14 EPRGAGTTLYFDY VkIII DPK22 22 QQYGGSPGYN NP08 VH4 4.30 † 18 DRGVSSGWTFDC VkII DPK16 32 MQLTAFPWT NP09 VH4 DP71 17 FRGGVAAGYDY VkIII DPK22 24 QHYRESCS NP10 VH4 DP78 29 DRVRVPYYYIDV VkIII DPK22 15 QQYGTSPYS NP11 VH3 VH3-8 † 12 DTTVTHYFDY VkI DPK9 21 QQSFSSPRT NP14 VH1 DP88 20 GPGATIHYYYMDV VkI DPK8 18 QQLDNYPLT NP15 VH5 DP73 36 LYTSIYYFDS VkIII DPK22 28 QQYGNSPPT Phage clones were either selected from the library at random (prefix R) or after four rounds of panning against (NPNA) 3 C-BSA, eluting with free (NPNA) 3 peptide (prefix NP). † The segment given the best DNAPLOT match, although the segment sequence has not been verified by duplication. ‡ Aligned after removal of the unusual sequence additions (see §, -, ‡, -, ). § Figure in brackets indicates an unusual sequence addition. - Sequence has two additional codons in CDR1. - Sequence has one additional codon in CDR1. * Sequence has three additional codons in CDR2. Expression and evaluation of the recombinant antibody fragment The PfNPNA-1 sequence was transferred to the expression vector pAbHIS (as outlined in Figure 2 . Purification of the V H /κ polypeptide was carried out on Ni-NTA agarose beads, yielding 0.5 mg of the 38 kDa V H /κ polypeptide/L bacterial culture. Fine specificity and affinity determination Anti-sporozoite activity of the PfNPNA-1 V H /κ molecule was clearly evident in an immunofluorescence assay (IFA) with P. falciparum sporozoites (Figure 3 ). The human single chain monoavalent antibody (panel A) was compared with a known in vitro protective whole murine antibody 2A10 (panel B). The murine antibody and the recombinant PfNPNA-1 V H /κ molecule both labelled the parasites. Figure 3 Indirect immunofluorescence assay (IFA) on Plasmodium falciparum sporozoites. Panel (A) PfNPNA-1 V H /κ, (B) 2A10 MAb. Competitive ELISA was carried out and the IC 50 value used to approximate the affinity of binding. Binding affinity of the monovalent PfNPNA-1 for (NPNA) 3 compared favourably with values previously reported for a panel of conventional murine monoclonal antibodies directed against the repeat epitope [ 18 ], which also have affinities in the μM range (Figure 4 ). Figure 4 Competition ELISA. Analysis of the fine specificity of the antibody PfNPNA-1 revealed weak binding to the repeat based [NVDP(NANP) 15 ] 2 , R32tet32 [ 19 ], whilst binding to the (NANP) 4 epitope contained within the hepatitis B virus nucleocapsid (C75CS2) [ 20 ] was strong. This activity profile pattern was mirrored in the protected donor serum (Figure 5 ). The very high binding observed with WR5 immune serum with the (NPNA) 3 C-BSA conjugate is probably due to the multivalent array of the capture ligand (i.e. multiple peptides coupled per BSA molecule), favouring more efficient retention of the antibody. Figure 5 Determination of specificity of PfNPNA-1. The binding of immune serum (WR5), non-immune serum and PfNPNA-1 V H /κ to R32tet32, recombinant hepatitis core containing (NANP) 4 peptide sequence and (NPNA)3C-BSA conjugate coated microtiter plate well was determined by ELISA essentially as described in Figure 4. The serum(s) and the recombinant PfNPNA-1 V H /κ were diluted 1/16 and 1/10 respectively. Discussion The recombinant antibody library construction differed from conventional antibody phage display library assembly [ 10 - 13 ], a pre-selection step was introduced to remove antibody inserts that were either; prematurely terminated, intact but did not translate well or were intact, translated well but failed to translocate into the bacterial periplasmic space, a prerequisite for functional display. Previously an approach towards developing a vector to select for fully intact functional sequences for antibody or peptide display had shown promise with model sequences [ 21 ], but had not been applied for large-scale random antibody library assembly. A "clean-up" vector, plasmid open reading frames expression secretion (pORFES) [ 14 ] was developed and used to remove these non-functional sequences. Up to 50% of the clones from the initial transformed library were non-functional. Some of the non-functional antibody fragments could in part be due to errors introduced during PCR amplification resulting in frame shifts. However it may be that some sequences either did not express well or did not translocate into the periplasmic space. Irrespective of the explanation, the size of the functional library was half of the total transformation events. An initial enhancement of the initial library by removing most non-functional inserts may at first appear to be a minor improvement. However, in conventional phage display the initial expansion of the library prior to panning results in a preferential growth of phage that do not make and display encoded inserts, moreover phage that lack an insert have a greater growth advantage. This results in a phage population that is greatly biased towards non-productive elements, which impacts directly on the panning efficiency. Incorporation of the pORFES step assured that only the functional (1.3 × l0 6 ) sequences were subsequently transferred to the phage display vector. Panning with a functionally enhanced library resulted in very efficient enrichment and recovery. Previously it had been demonstrated that manipulating the conditions of phage production results in modulation of the density of antibody display on phage [ 15 ]. The phage library was expanded using parameters that would result in either monovalent display (0-1 antibody/phage) or multivalent display (0–5 antibodies/phage) [ 15 ] prior to initiating panning. It was anticipated that a range of antibodies with varying affinities would be present in the library, and modulating antibody display on phage would permit capture antibodies with a range of affinities and sequence diversity. Induction of protective immunity against sporozoite challenge by exposure to radiation attenuated malaria sporozoite has been demonstrated in humans [ 4 , 7 , 22 ]. Protection is thought by most investigators to be primarily cellular in nature [ 23 ], but there is no question that antibodies with significant sporozoite neutralizing activity are elicited [ 22 ] and may play a role in protection. The antibody response is primarily directed against the repeat region of the PfCSP. Studies of subunit vaccines which induce antibodies only against the repeat region demonstrate that protective immunity can be induced in some individuals [ 24 , 25 ]. At the onset of this study it was proposed that the dissection of the anti- P . falciparum sporozoite antibody response by combinatorial antibody library phage display would permit individual selected antibodies to be evaluated for protective potential and the information generated could be used in vaccine design. In particular, attention was focused on antibodies against the structural motif (NPNA) n . Despite using two different strategies for the elution of repeat region peptide specific antibodies (acid and peptide specific) it would appear that the anti-structural repeat response by this protected individual is restricted to a single V H /V L combination observed in the panel of selected phage (n = 25). Sequencing of randomly picked phage prior to panning revealed that a diverse range of V H and V L families were represented in the library as shown in Table 2 . Moreover the PfNPNA-1 V H /V L was not represented in the sampling and was only detected after enrichment. Comparison of the monovalent PfNPNA-1 molecule with the conventional bivalent murine mAb, such as the in vitro inhibitory 2A10 against P. falciparum sporozoites indicates that they recognize the repeat epitope(s) with equivalent affinities [ 18 ]. The sequence revealed extensive somatic hyper mutations in both the V H and V L genes suggesting antigen driven affinity maturation. Based on these observations, PfNPNA-1 may be a good candidate to develop and evaluate as a protective antibody. Analysis of field samples in rural Gambia [ 26 ], Thailand [ 27 ] Indonesia [ 28 ] and Kenya [ 29 ], suggest that anti-sporozoite antibody is poorly developed under natural conditions of exposure and does not protect against clinical malaria. In contrast to exposure to P. falciparum sporozoites under natural conditions in the field, immunization with irradiated P. falciparum sporozoites induces in general higher levels of antibodies against the PfCSP repeats, and does induce sterile protective immunity [ 4 , 7 , 30 - 38 ]. In the study by Egan et al., 3 of the 4 volunteers were protected against challenge with P. falciparum sporozoites. The generally accepted explanation for the lack of protection in the one volunteer is that the volunteer did not receive an adequate immunizing dose of irradiated sporozoites (less than 1000 infective bites [ 4 , 7 ]). However, it is of interest that this non-protected volunteer (WR1, [ 22 ])had significantly lower levels of antibodies against the PfCSP repeat than did the protected volunteer who donated cells for this study (WR5, [ 22 ]) (2.4 μg/ml vs 50 μg/ml of specific antibody). This raises the question as to whether the antibodies are markers for adequate immunization or are actually major mediators of protection. Regardless, this anti-repeat response in this protected individual appeared to be restricted to a single antibody. This does not preclude that antibodies directed against non-repeat epitopes on PfCSP and other sporozoite proteins [ 39 ] play a role in protection. It is not possible to conclude that the response against the structural repeat epitope is restricted to a single antibody of moderate affinity, since only a single protected donor has been used in this study. One may speculate that in concordance with the argument put forward by Saul [ 40 ] that the inability to recover high affinity antibody, may reflect that high affinity antibodies may not be required for protection. Due to the repetitive nature of the antigen one can further speculate that only limited affinity maturation is required to obtain physiologically relevant efficacy. The restricted recovery of antibodies is unlikely to be a technical limitation on the phage technology since others have generated panels of very high affinity human antibodies against a range of antigens [ 13 ]. Very few examples of different approaches of generating human antibodies from immune donors are described in the literature, in particular when attempting to make antibodies against the same antigen. Currently it is not possible to fully understand the limitations of a technology. Using an alternative technology of engrafting immune human PBL's directly into SCID mice from donors vaccinated against anthrax vaccine adsorbed, boosting with protective antigen (PA), recovering immortalizing antibody-producing cells via conventional hybridoma technology [ 41 ] resulted in a panel of very high affinity potent neutralizing antibodies against anthrax toxin. Independently, an antibody phage display library from a similar (not identical) immune donor PBL's was constructed and panned against PA [ 42 ] also resulted in a panel of high affinity anti-anthrax PA antibodies. This would suggest that the methodology is not limiting. However in this example, unlike CSP, the PA antigen does not contain repeating epitopes. Further it is speculated that antibodies directed against the structural (NPNA) n repeat play a role in conferring protection against P. falciparum sporozoites in some of the protected volunteers and this protection may be associated with circulating levels of this specific antibody against the structural repeat. Efforts are being directed towards producing a fully human IgG based on the PfNPNA-1 V H and V L domains for further in vitro and in vivo evaluation. The use of a human monoclonal antibody as a preventive measure against P. falciparum malaria, would be independent of factors which hinder active vaccination, such as adjuvant effects, the requirement to be effectively presented in a diverse range of human leukocyte class I and II molecules, and immunlogical antagonism [ 43 , 44 ]. In practice, the utility of monoclonal antibodies as anti-infectious agents is often negated by the presence and or the inevitable emergence of variants with altered surface epitopes (in particular with viral targets). Fortunately, there has never been a P. falciparum isolate that does not contain the (NPNA) n repeats on the PfCSP [ 45 ], and the number of tandem array of repeats on the PfCSP reduces the likelihood of variants arising which evade antibody recognition. This would suggest that an effective antibody directed against the repeats would be effective against all P. falciparum. If this restricted antibody response to the repeat epitope plays a role in preventing P. falciparum infection, PfNPNA-1 may be a useful prophylactic agent. Moreover, if PfNPNA-1 is shown to be protective in passive immunization in humans or monkeys as previously demonstrated for anti- P. vivax CSP murine mAb, NVS3 [ 46 ], it would provide a template that could be used in defining the precise conformation of the structural repeat required for the induction of desired antibodies that can neutralize parasites. Conclusions Over the past 25 years the antibody response against the PfCSP repeat epitope has been pursued as a target for active vaccination, with encouraging results [ 47 ]. Our attempt to dissect the protective antibody response against the structural PfCSP repeat revealed that the response was restricted to a single V H /V L pairing, designated PfNPNA-1 encoded by V H 3 and V κ I families (with evidence of somatic mutations). The affinity for the ligand was in the μM range, which in the context of a whole antibody may be more than sufficient for retention on a polyvalent surface such as the P. falciparum CSP. It is speculated that the induction and the maintenance of high circulating levels of antibodies against the structural PfCSP repeat may be more important than intrinsic high affinity for the ligand for protection against P. falciparum infection. The absence of high affinity anti-repeat antibodies is in concordance with the expected response against a multivalent antigen (i.e. sporozoite surface). Under physiological conditions a whole IgG antibody and a multimeric ligand result in bivalent binding. Such complexes can have avidities estimated to be approaching the product of two independent monomeric interactions. In this case, the 1 × 10 -6 M monovalent affinity of PfNPNA-1 may approach a theoretical higher avidity (1 × 10 -12 M) in the context of a whole antibody. This implies that further affinity maturation either in vivo or in vitro may not necessarily increase physiological effectiveness of the whole IgG antibody. Public health officials have acknowledged the urgency for development of an effective anti- P . falciparum malaria vaccine. One of the key criteria of such a putative vaccine may be the induction and maintenance of high levels of anti-(NPNA)n antibodies. The fully human PfNPNA-1 IgG could be used as a positive control in evaluating sera from immunized donors, or possibly be developed as a prophylactic agent that could be used alone or in combination with various vaccination strategies. One immediate hurdle for the development of such an antibody as a prophylactic would be the anticipated high cost of commercial manufacture in mammalian cells. However, advances in alternative antibody production technology may one day provide some more cost effective solutions [ 48 , 49 ]. With the availability of an antibody phage display library constructed from a protected individual immunized via bites of irradiated P. falciparum infected Anopheles mosquitoes, it should be possible to further dissect the antibody response against "other" sporozoite antigens [ 39 ]. Authors' contributions JAC was the postdoctoral researcher on this project. WOR and SLH co-investigators. ASK was the PI and recipient of the Department of Army award. All authors read and approved the final manuscript Disclaimer The views and opinions expressed herein are those of the author and do not purport to reflect those of the U.S. Navy or the Department of Defense, Sanaria Inc or Avanir Pharmaceuticals Inc.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC509279.xml
520827	A weather-driven model of malaria transmission	Background Climate is a major driving force behind malaria transmission and climate data are often used to account for the spatial, seasonal and interannual variation in malaria transmission. Methods This paper describes a mathematical-biological model of the parasite dynamics, comprising both the weather-dependent within-vector stages and the weather-independent within-host stages. Results Numerical evaluations of the model in both time and space show that it qualitatively reconstructs the prevalence of infection. Conclusion A process-based modelling structure has been developed that may be suitable for the simulation of malaria forecasts based on seasonal weather forecasts.	Background The importance of climate as a driving force of malaria transmission has been known since the earliest days of research on this devastating parasitic disease. However, it is only with the advent of effective weather forecasting techniques that this knowledge may be implemented numerically. Seasonal climate forecasting (with up to six months lead time) has developed rapidly in recent years with a number of atmospheric climate modelling groups showing evidence of skill and reliability in their systems. Because of the chaotic nature of the atmosphere, seasonal forecasts are necessarily probabilistic. These probabilistic predictions are derived from multiple integrations of deterministic climate models. These models successfully predicted the onset and demise of the 1997/1998 El Nino event and its impact on weather in Africa [ 1 ]. That event in East Africa was associated with devastating malaria epidemics[ 2 ] and, consequently, the health community has shown an increasing interest in the use of seasonal forecasts for predicting epidemics of climate related diseases[ 3 ]. The DEMETER project was aimed to advance the concept of seasonal climate forecasts based on multi-model ensembles. The DEMETER coupled models and the DEMETER retrospective forecast (hindcast) integrations are described elsewhere [ 4 ]. The European Centre for Medium Range Weather Forecasting (ECMWF) second-generation global weather re-analysis data set ERA-40, is being used to test the accuracy ("skill") of the hindcasts. Central to the DEMETER project is an evaluation of the potential of seasonal climate forecasts for end-user communities, such as those concerned with agricultural output and malaria epidemic control[ 5 ]. ERA is thus being used as the "gold standard" for the weather forecasts, and in the research presented here is being used as a daily weather database for all Africa. The aim of the MALSAT group at the School of Tropical Medicine along with the Department of Geography of the University of Liverpool was the assessment of the methodological issues raised by driving a dynamic malaria model with seasonal climate forecasts. In this paper, the first phase is presented, namely the formulation of the model and the development of a dynamic mathematical model of malaria transmission, which can be driven by daily meteorological variables (rainfall and temperature, as provided by ERA-40). Additional research assessed the future risk of malaria epidemics in probabilistic terms[ 4 ]. Mathematical models of malaria transmission Mathematical models of malaria span nearly a century and are well established. Macdonald[ 6 ] reformulated the pioneering model of Ross[ 7 ] and identified mosquito vector longevity as the single most important variable in the force of transmission. Further modelling work established vectorial capacity as a practical means of assessing the effectiveness of control measures aimed at the vector[ 8 ] and many refinements in modelling technique have since been applied[ 9 , 10 ]. However, these models have, until recently, been dependent on the unrealistic assumption of quasi-static vector numbers[ 8 ] and unvarying parasite development rates. Where variation in mosquito numbers has been introduced [ 11 ], this was achieved using pseudo-climate, a seasonal variation in mosquito numbers, but not involving variations in vector and parasite development rates, and definitely not in relation to the climate that is actually experienced, i.e. changing weather[ 12 ]. In a new report[ 13 ], varying mosquito biting frequencies were indeed simulated, but not the co-varying mosquito and parasite weather dependent dynamics. Climate (as distinct from weather) models of malaria transmission have been developed in recent years to improve our understanding of the likely impact of climate change on malaria transmission. For example, Craig et al. [ 14 ] developed a fuzzy-logic climate-based distribution model which they suggest could be used to look at the impact of climate change on malaria transmission and, combined with population, morbidity and mortality data, to estimate the burden of disease and aid strategic control of malaria. Lindsay and Birley[ 15 ] used a simple mathematical model to look at the effects of temperature on the ability of Anopheles maculipennis to transmit Plasmodium vivax malaria. Martens [ 16 ] used a rules-based modelling approach to examine how climate change might affect global malaria transmission. Lindsay and Martens[ 17 ] used a similar model to look at the implications of climate change scenarios on highland malaria in Africa and, more specifically, in Zimbabwe. Hay et al. [ 18 ] analysed the potential effects of climate change on highland malaria, using a regression approach, and Rogers and Randolph[ 19 ] used a statistical model to determine that the global impact of climate change on malaria distribution will be minimal. The relationships identified and applied in the body of research on climate change and malaria transmission highlight the possibility of explicitly relating malaria transmission both spatially and temporally to climatic variables such as temperature, rainfall and (less clearly) humidity. It is, therefore, possible to use these relationships to drive the currently available models of malaria transmission, although, to-date, none of these models are designed to indicate temporal changes in transmission dynamics based on weather[ 20 ]. In order to be able to predict within-season and between-year variation in weather-related malaria risk, the model must be driven by varying weather. This paper describes in detail the development of a weather-driven dynamic mathematical malaria model, the final output of which is new infections in the human host. Preliminary results of its numerical evaluation in time and space are presented. The choice of a causal mathematical model rather than a statistical model is based on the knowledge that the former is better suited to extrapolations to novel situations (e.g. when interventions are introduced), and for investigating the non-linear impacts of short-lived changes in driving meteorological determinants. The present malaria model is designed to be used for two distinct but related functions (a) to determine the impact of weather variables on model output (malaria cases/infection) for given interventions and (b) to determine the impact of specific control interventions on model outputs by modifying model parameters. For the former function, the model can be driven with meteorological variables, from ground-based observations, satellite or modelled weather data, seasonal climate forecasts (and potentially) climate change scenarios. In the second function, the value of a malaria early warning system (MEWS) in terms of triggering earlier or scaling up intervention efforts (e.g. residual spraying) in epidemic years may be assessed. In both cases, the biological processes are included as a series of interlinked sub-models and thus represented as coupled delay differential equations. Each new item of knowledge (such as improvements in the structure of the dynamic equations, due to novel experimental or field data) may be immediately used in the model. New developments such as, for example, the ominous spread of drug resistance may also be immediately incorporated into the model, as, in this example, a reduction in the parasite clearance rate. The implications of all such changes may then be assessed as a quantitative amendment to the prediction. This is done independently for each location and, therefore, the model can be fitted to local conditions, where relevant data are available, or to regional parameters when such data are lacking. Sensitivity analysis can also be used to establish the relative importance of obtaining more accurate data on each parameter Thus, central to the analysis is the development of a new weather-dependent mathematical dynamic model which no longer attempts to calculate a single constant epidemiological per-case multiplication rate, but follows the temporal progress of the prevalence of infection within a population through seasons and years, To simulate the stochastic elements of the model, delay differential equations based on probabilistic transition between groups rather than on Monte Carlo modelling, as has been undertaken by Gu et al. [ 13 ], have been formulated.. Methods Data-source The ERA-40 weather reanalysis data set was chosen because it is the reference data for DEMETER and can provide daily estimates of a range of potentially significant weather variables for the whole globe. This data set was prepared by ECMWF and consists of weather reanalysis data for the whole globe for 40 years (1960–2000) and builds on previous reanalysis data ERA-15 which had been used in an earlier analysis of multi-model ensembles seasonal forecasts . These three- to six-hourly data are stored on the ECMWF site and extracted by local software at the site by the user. Weather variables were extracted from the database as gridded data at one degree (~111 km.) and twelve-hour resolution for the African region. Ideally, daily averaged temperature, accumulated rainfall and humidity would have been the variables. However, reanalysis-based daily average temperature correlated poorly with station data (which is more representative of typical local conditions), and daily minimum temperature proved even worse, and humidity too was poorly modelled. Thus, the analysis used surface (2 m above ground level) daily maximum temperature, offset by -5°C (to roughly represent mean temperature) and total rainfall estimates as input variables. As accumulated rainfall (puddles etc.) is more important than daily rainfall, R d (dekadal rainfall), the sum of the previous 10 days of rainfall., were used. There are four malaria parasites ( Plasmodium spp.), which cause disease in humans. The focus here is on Plasmodium falciparum , as it is the principal life-threatening parasite species and is most common in Africa. The temporal resolution of the model was based on the nocturnal activity of the vector and the fact that empirical vector observations are usually made at no better than daily time-resolution. Therefore, the simulation time-step is a single day (24 hours). It was assumed that the dynamics at the grid-points used do not interact significantly and can be treated as independent, as the distance between grid-points is far greater than the normal mosquito flight distance of roughly 1 to 2 km [ 21 - 23 ]. To simulate large-range transmission by human movement (due to a migrating workforce for example), a small, constant influx of infected people is assumed for each grid point. Biological model Human malaria disease is caused in the individual by an infective mosquito biting a non or semi-immune human. After some two weeks the first gametocytes are produced, independent of ambient temperature. A second mosquito biting the infected human thereafter may ingest gametocytes, which after fertilization pass through the gut wall, develop and ultimately produce sporozoites which become infective when they migrate to the mosquito salivary glands. This process is ambient temperature-dependent. As transmission is less dependent on the number of parasites than on the infective status of the carriers, human and arthropod, only the infection and infectiousness status of the carrier populations are simulated (see Table 1 and 2 ). Modelling the vector population The most common vector of falciparum malaria in Africa is Anopheles gambiae (s.l.) [ 24 , 25 ]. As the female mosquito needs to feed on blood to enable ovum development, its entire life cycle must be modelled. The blood may come also from other mammals, such as cattle (which are not Plasmodium hosts), and the mosquito's anthropophilic tendency is an important factor in establishing the intensity of transmission. While the anthropophily varies between regions, at this stage it is assumed constant. The male does not bite and, therefore, does not transmit the disease, and as there are always sufficient males to impregnate the mature females, there is no need to simulate the males' dynamics. The female life is divided into two major parts: the immature stages (egg, larval and pupae), and the mature stage, where onset of maturity is defined as the time of the first flight, which is shortly followed by the first bite. The importance of this division is twofold. First, the immature mosquitoes do not participate in the infection cycle and are, thus, basically in a waiting period, which limits rapid vector population growth. Second, the survivorship (defined as the probability to survive 24 hours) and development rate (part of stage completed in 24 hours) have a different dependence on weather conditions for mature and immature mosquitoes. A schematic representation of the mosquito life cycle is presented in Fig. 1 . Figure 1 Schematic presentation of the life cycle of Anopheles gambiae (s.l.). Immature mosquitoes progress at temperature-dependent rate m . They are liable to die at daily rate 1-s. Upon completion of immature process they form mature mosquitoes which begin a gonotrophic cycle with progress rate P R . They are liable to die at a rate of 1-a per day. New mosquitoes are being imported with rate trickle2. Each mosquito as oviposition lays gR d eggs. Immature populations Hitherto, the immature population dynamics have not been involved explicitly or clearly in malaria modelling. The immature forms are water-bound and are thus totally dependent on the existence of water bodies. High temperatures in breeding sites and evaporation (resulting in elimination of puddles, following the cessation of rain) are generally lethal (see below). Unfortunately, reports giving quantified relationships of temperature and rainfall/humidity dependence of the mosquito dynamics are in short supply, although some data are available [ 26 ]. Their further publication would assist the development of models. Eggs are posited by mosquitoes in pools. As a mosquito must find water to reproduce, the oviposition rate is roughly assumed to be proportionate to both the ovipositing mosquito number and to the dekadal (ten-daily) rainfall R d filling the pool. Thus a mosquito's probability to oviposit, and for the larvae to survive, is proportional to the amount of water it finds. The inclusion of hydrology and soil typemay improve the understanding of the connection between rainfall and breeding-sites. This dependence on rainfall is valid for areas in which the surface water is dependent only on rainfall. However, in constant mosquito habitats such as stable pools or rice-fields, lack of rain may not limit growth of larvae, but natural predators may curb the growth of larval populations. As the movement between the human location (where a bloodmeal is taken) and the ovipostion site is dangerous and energy consuming, this distance is a major factor in determining the probability of oviposition. Future numeric data for this would be of great value. There is a shortage of detailed information on the varied survivorship and stage-dependent developmental progress in natural habitats as functions of weather conditions. There are, however, certain limited sources for derivation of information[ 27 , 28 ]. Jepson et al. [ 28 ] measured the length of each of the three stages (egg L e , larvae L 1 and pupae L p ) of An. gambiae s.l. in 11 different breeding habitats (typically sunlit pools), in which the daytime water temperature was measured. Maturation rate, m, is defined as the fraction of the total immature stage covered in a single day, and is the inverse of the sum of the duration of the immature stages[ 29 , 30 ] m = 1/(L e +L 1 + L p ) (1). Thus the maturation rate is a function of temperature for these stages. Even though it is known that high temperatures are detrimental to larvae survival, there are no published numerical data (such as Ndegwa et al. [ 31 ] found for Trypanosoma Congolese) which would allow the introduction of this element. The most important cause of larval weather-attributable death is probably desiccation[ 32 ]. However, in some circumstances eggs can survive for weeks without water[ 33 ], and so an immature mosquito rainfall-dependent daily mortality rate, actually resulting in total clearance of the population, is not used in the model. Lack of rain will cause the numbers to be reduced in any case, as above. The overcrowding of immature mosquitoes may result in significant differences in both larval/pupal Survival and also in body size (and thus survival probability of emerging adults). Larval and purpal predation in well-established pools (as opposed to transitory puddles) has significant effects on population development [ 34 ]. The model does not currently account for these factors, but assumes a fixed per diem survival rate, s, of 90% [ 30 ]. This will be amended, by the aid of new data being collected now, for further model development. The immature population is, thus, simulated as a set of ν virtual boxes with populations I(n) (n = 1..ν; each box representing the inverse of ν, the length of the immature phase in days). At each simulated day the whole population of each box is multiplied by the per diem survival rate σ and moved on by mν boxes (with m, as above, the per diem fractional maturation rate). New eggs (box number 1) are laid as a fraction of the number of ovipositing adult females (as shall be discussed). Thus, at each time step, t, the immature population at stage s, I(s,t), has the dynamics: I(s+mn,t+1) = σ I(s,t) (2). Maturing pupae (reaching age n and above) are removed and enter the mature mosquito dynamics according to: i.e. all immature mosquitoes within one day of maturation (at stages above ν(1-m)) will become mature mosquitoes the next day, if they survive(s). Mature mosquito dynamics Mature mosquitoes begin their adult life with their first, nuptial flight, during which fertilization occurs. In the model, a small constant trickle (trickle2) of young uninfected mosquitoes (representing a new imported population) is added to the population of maturing mosquitoes. Afterwards, the development of the fertilized eggs requires the intake of protein, i.e. blood meals. Although sometimes more than one initial blood meal is required before egg maturation can occur, and fed mosquitoes that are unable to develop mature eggs are best described as pre-gravid[ 35 ], this fraction is neglected at this stage of the model, and all blood meals are assumed to allow, at least potentially, egg maturation. At this level of model development, situations where blood meals are interrupted or when partially fed mosquitoes complete their meal on a second host are ignored. The rate of development of each brood of eggs in a vector is dependent on temperature and, to a lesser degree, on external humidity (probably as a result of the stress caused by a harsher dry environment on the vector). Detinova[ 36 ] detected a "degree-day" dependence of the time for the preparation of a brood in An. maculipennis (the gonotrophic cycle, G c ) and hence also of the time for biting, which may be expressed as G C = 1+D d /(T-T c ) (4); where D d is the number of degree-days required, T c the threshold beneath which development halts and T is the daily average temperature. Both D d and T c are dependent on humidity. In highly humid conditions, D d = 37 from Detinova's data. In the tropics G C is typically about three days depending on temperature. The temperature dependence of An. gambiae is assumed similar until further data become available. As temperature is not assumed constant (on the contrary, the model is interested in its variation), a daily progress rate (part of gonotrophic cycle covered in one day): P R = 1/G C is calculated. This assumes that the temperature dependence of the rate is constant throughout the gonotrophic cycle, which is implicit in the degree-day concept. The completion of a cycle may be established when the sum of daily P R values reaches 1. 37 "boxes" (corresponding to the 37 degree days) are constructed, between which the mosquitoes progress in steps of P R reduced by multiplication by the survival rate (which shall be elaborated below). At the end of a gonotrophic cycle (upon arrival at box 37), each mosquito oviposits and then begins a new cycle. The success of oviposition is dependent on the existence of water-bodies, and hence on dekadal rainfall, and we assume that each ovipositing female lays γR d viable eggs, where γ is a constant. The following day these eggs begin the immature mosquito cycle (as above). It seems that the survivorship of mosquitoes is only weakly dependent on their age[ 6 , 9 , 37 - 39 ], in spite of some conflicting evidence[ 40 ]. The stage most dangerous to the adult mosquito in this model is the feeding stage, consisting of the approach to the mammal for the bite, the duration of the bloodmeal (with the corresponding irritation to the mammal) and the escape to a resting point afterwards. This risk is likely to be increased in unfavourable weather conditions (high temperatures and low humidity), but this has not been investigated yet. Thus, in the present model, the survival of the mosquito per gonotrophic cycle is a constant, a, independent of the duration of the cycle [ 15 ]. Estimates of the constant typically vary between 0.4–0.6, and are bound more tightly as 0.48–0.54 by some groups [ 41 - 43 ]. The per diem survival is thus calculated by P = α 1/G c . As G C is weather-dependent, so is the daily survival. It was assumed that survivorship is independent of the infective state[ 6 , 44 ], even though there are some reports that being infected is harmful to the mosquito. Combining these, it may be possible to write for δφ, the daily change in the total number of mosquitoes (N m ) is the difference between the new mosquitoes maturing (and not dying in the period) and the fraction of the mature mosquitoes dying (the daily cycle completion rate 1/G C multiplied by the death rate 1-α): As mentioned, the parasite within-vector dynamics is superimposed on the mosquito dynamics. Mosquitoes are assumed to bite human hosts randomly (independent of their infective status) and thus the proportion of infectious humans (H i ) rather than non-infectious (H n ) bitten is the human infectious ratio r = H i /(H i +H n ) (6). Non-random biting by mosquito vectors is well described[ 45 ] and this could be incorporated in the model at a future stage. The preference for human biting over cattle is described by the human blood index (B, the proportion of bites on humans, of total bites), which is high (0.6+) for anthropophilic An. gambiae s.s. (even though the tendency varies between strains and regions) and generally much lower for zoophilic Anopheles arabiensis [ 46 ]. Of course, when cattle are far more abundant than humans, the effective B would be reduced. A fraction χ of mosquitoes that bite infective humans become themselves infected and thus the mosquito infection per bite probability is M IP = χ B r (7) It is generally assumed that infected mosquitoes stay so for life. The sporogonic cycle (S C ) (the process of fertilization of the macrogametocyte, formation of the oocyst, ookinete, penetration of the midgut and then the subsequent development of the sporozoites which dwell in the salivary glands) for P, falciparum lasts 111 degree-days above 18°C[ 36 ]. The daily sporogonic progress (in degree days) is thus S R = 111/S C . The infectivity of a specific mosquito over its lifetime is dependent on the number of bites it makes after the completion of a sporogonic cycle following the first bite of an infective host. To combine the gonotrophic and sporogonic processes each of the 37 box-stages of the gonotrophic cycle are sub-divided into 112 sub-sections, numbered 0 to 111, representing progress in degree-days. The 0 subsection reflects an uninfected mosquito. The mosquito population is governed by the following dynamics. An infected mosquito sub-population, M(s,S s ,t), at stage s of the gonotrophic cycle and at stage S s of the sporogonic cycle (in sporogonic-cycle degree days) at time t (in calendar days) progresses each day by gonotrophic rate P R and by the sporogonic rate S R : M(s+P R S s +S R ,t+1) = pM(s,S s ,t) (8). A finite fraction (1-p) of the mosquito population which dies and thus does not make the transition. Upon completion of the gonotrophic cycle, the process restarts. Upon the completion of the sporogonic cycle the mosquito remains at the infectious stage. If the mosquito is not infected at biting, it remains uninfected throughout the gonotrophic cycle: M(s+P R, 0,t+1) = pM(s,0,t) (cycle without infection); (9a) but upon biting an infectious human, an uninfected mosquito has a finite probability of either becoming infected M(P R ,S R ,t+1) = pM(0,0,t)M IP (new infection) (9b) or not: M(P R ,0,t+1) = pM(0,0,t)(l-M IP ). (9c) Mosquitoes may arrive at the uninfected biting stage M(0,0,t+1) by two processes, either just after maturation or else by completing an uninfected gonotrophic cycle: New eggs are laid by mosquitoes completing a gonotrophic cycle: This means that all mosquitoes located less than P R from the end of the gonotrophic cycle will oviposit. Their number must be summed over all infection states 0...111. As discussed above, the average brood size is dependent on rainfall by a multiplicative constant γ. In the present report all the initial mosquitoes were non-infected. Modelling the infected host population Obviously, the focus is on the infected host population dynamics, which reflects the diseased population. The simulation of this population is based on the following assumptions: All hosts and mature mosquitoes are equivalent except for their infection status. Acquired immunity is not accounted for. Thus, the model reflected the prevalence of malaria infection in the population unless the population is largely non-immune in which case it reflects the prevalence of malaria disease. Immune individuals are assumed to be potential carriers, even though not at personal risk. This issue is contended within the modelling community, and future models will allow for both possibilities. The crude human death rate is taken as low enough to be unimportant over the time scale. Specifically, the malaria-induced death rate does not influence transmission patterns. Newly infected patients are not infectious for two weeks, during the intra-hepatic phase of the disease and the early erythrocytic stage, before gametocytaemia rises sufficiently for significant transmission: H(h+l,t+l) = δH(s,t) (13>h>l) (12) Malaria clearance is a slow process. Patients may become uninfected at a constant rate (first order process). The rate selected (δ= 0.97 per day, based on MacDonald's work[ 6 ]) enables 90% of the population to clear their infection after 80 days, but other values of lower clearance do not show significantly different results. Super-infection is not accounted for, as for low- to medium-level transmission this effect is of secondary importance, although this may be built into subsequent models. The initial population is assumed to be non-infected, but new infections are being introduced at a constant low rate ( trickle = 0.01 into a population of 100 every four days). This has the technical benefit of not changing the total population considerably over the simulation period, but limiting the influence of initial conditions found when a large infected population is assumed, or the dying out of the disease after a short initial dry period. This reflects the constant pressure of low-scale transmission by migration (infected migrant workers or troops for instance). This would be quite similar to the case of static communities with low-level external contact [ 47 ]. A human bitten by an infectious mosquito may become infected. The finite probability for a bite by an infectious mosquito causing human infection is integrated into the constant c of the reverse process. There is a probability, H IR , of a human being bitten by an infectious mosquito each day. This rate depends on the abundance of infectious mosquitoes and of human hosts [ 48 ]. There are three components of the human infectious population at time t+1, H(14,t+1): (i) Individuals remaining so from time t, (ii) those who complete the hepatic latent period and (iii) new imports: H(14,t+1) = (H(14,t)+H(13,t))(δ) + trickle (13a). A host may be uninfected at time t+1, by either (1) remaining uninfected with probability (1-H IR ), or else (2) being an infected host (S(H(s,t))) and clearing his/her infection with probability (1-δ): H(0,t+1) = (1-H IR )H(0,t)+(1-δ)S(H(s,t)) (13b) A newly infected host begins the latent phase: H(1,t+1) = H IR H(0,t). (13c) The described modelling process was used to establish the fit of the model to a time series of clinical data from Hwange District, Matabeleland, Zimbabwe [ 49 ]. Due to the strong dependence of such a local clinical data-set on local weather conditions, the model was driven by station weather data, taken from a CD obtained from NCDC ( ) using Victoria Falls weather station (WMO ID 678430). Weather values for days with missing data (some 10% of all days) were filled by averaging data from adjacent days. Using the steps described above, the model was run using the ERA-40 weather reanalysis for every grid point covering the African continent (a rectangle from 37°N 18 W to 35 S 52 E, with a mask for areas covering the ocean) over the time period 1987–2000. To allow a spin-up period for the model weather data for 1987 was run twice, while not storing the daily output for the first run, but allowing the first run's output of host, parasite and mosquito situation at the end of the year to be the initial conditions for the "real" run. This allows the simulation of more climatologically realistic starting conditions. Variations in the length of the spin-up period gave similar results. The average prevalence and incidence for the period is then established. Next, the variation of incidence during the period from interannual means was calculated, and hence the standard deviation of annual incidence. This value serves as an estimate of the extent of anomalous malaria, thus reflecting epidemics, beyond holoendemicity and seasonal variation. Results Development of immature Anopheles gambiae s.l Fig. 2 presents the rate of maturation of larvae using the data from Jepson et al. [ 28 ]. The x-axis is the average temperature in Celsius and the y-axis is the fraction of the whole larval stage covered in a single day at the given temperature. The straight line is the best-fit (with standard errors) Figure 2 Larvae maturation. Rate of development of larvae as a fraction of the complete development cycle as a function of the water temperature in Celsius. Data based on that of Jepson et al . Line best fit by least squares. X-axis average water temperature, Y-axis: rate of development (in 1/days) as the reciprocal of length of cycle. m = 0.011 (± 0.001) T-0.2 (± 0.26) (1/day) (14) Note that the report is for water temperature in shallow pools, which may be significantly higher than the ambient temperature. A few points suggest themselves. To begin with, the intercept with the X-axis is around 18°C. Even though the variation for this value is large, it suggests a lower limit for larval development. Beyond this point, it seems that the assumption in the theoretical methodology of linearity of the development rate with temperature is justified. New data being collected may allow a more thorough validation. Using the proposed rain-dependent daily survivorship (S) and the length of cycle (1/ m ), the per-cycle rain-dependent survivorship is simply S 1/ m . The survivorship for the immature stage, by temperature for different values of S, is depicted in Fig. 3 . Not surprisingly, larval development increases with rain and temperature. The temperature-linked increase in survival is, however, limited by rainfall. This interdependence of the influence of the two climatic factors limits the regions and times of vector abundance and, hence, also the transmission of malaria. As weather is not constant in reality, numerical integration of the process is required. Figure 3 Larvae survivorship. The probability of a new Anopheles gambiae larva surviving to maturity as function of ambient temperature for different values of per diem survivorship. X-axis temperature in °C. Y-axis the probability of completing development until maturity. Lines from top to bottom daily survivorships of 0.9, 0.8 and 0.7 respectively. Development of mature Anopheles gambiae s.l To stress the importance of the mature-stage dynamics temperature dependence, Figs. 4 and 5 depict the three processes: biting, development of sporozoites within the vector and the vector survival probability as a function of time, for two constant temperatures (assuming humid conditions), 28°C and 19°C. The main points that can be seen from the figures are discussed in turn. At time 0 a female mosquito bites an infected human and begins egg-production, concluding with oviposition. At the end of this process, the gonotrophic cycle, it will bite again and so on as long as it survives. Meanwhile, the mosquito's survival proability drops. Thus the number of mosquitoes which may survive to become infectious from the initial bite at time 0 is decreasing. Meanwhile, the parasites acquired by that initial bite are developing, a process, which may last many gonotrophic cycles. When this process is complete, any surviving mosquitoes become infectious. Thus, the transmission probability is the sum of all survival probabilities after the completion of a sporogonic cycle (when the ascending line reaches 100%). At T = 28°C the sporogonic cycle is completed within less than 12 days, and thus at the next bite over 5% of initial mosquitoes will survive, thus infecting an uninfected human. At T = 19°C the sporogonic cycle lasts for months, and the survival probability of mosquitoes by then is extremely small. This explains the strong transmission in tropical regions and the lack of transmission in temperate zones. Figure 4 Vector dynamics and probability of transmission. The biting cycle (periodic spikes, arbitrary scale), vectorial probability of survival by day (descending line, left axis) and fraction of sporogonic cycle completed (rising line, right axis) for constant temperature. (Ambient temperature 28°C). Figure 5 Vector dynamics and probability of transmission. The biting cycle (periodic spikes, arbitrary scale), vectorial probability of survival by day (descending line, left axis) and fraction of sporogonic cycle completed (rising line, right axis) for constant temperature (Ambient temperature 19°C) Fig. 6 shows the fit of the model simulation (based on station data for 1995–1998) to clinical data from Hwange, Zimbabwe[ 49 ]. The following graphs are presented: the rain time series, the incidence as calculated by the model and the number of cases of malaria disease as reported by district. In this case we see the main peaks of the 1996–7 epidemic expressed in the model results. It can de seen that the malaria is driven, both according to the model and according to the clinical reports, by the intense rainfall. The rainfall is, however, a local station set, while the malaria clinical results represent a district, in which there was a certain level of prevention and vector control, as well as treatment of cases, all of which prevent the fast exponential increase in case number and predicted by a model, well into the rainy season. Thus, one would expect the model results plot to be spikier than the clinical report. This is, in fact, a general aspect of process-based models, which predict exponential growth of prevalence. The modelling of intervention is an issue under ongoing research. To assess the location of epidemic regions, Fig. 7 presents a map of the anomalous malaria incidence according to the present model for the fourteen years 1987–2000. This was achieved by calculating the interannual epidemic incidence. The anomalies, calculated as standard deviation of annual incidence for each grid-point, are presented in the map. The intensity of spots is proportional to the value of the standard deviation in absolute terms. The regions for which malaria prevalence is usually high (averaging 20%) are blotted out, to differentiate between endemic and epidemic zones. This figure focuses on a different aspect of malaria distribution than usual. The regions in which MEWS will have the greatest benefit are those in which malaria variation is largest. The most pronounced regions are the fringes of endemic transmission. This includes especially the Sahel in West Africa, wide regions of East Africa, ranging deep into Somalia, and in Southern Africa most of Zimbabwe and part of Namibia. Figure 6 Rainfall, reported cases and modelled cases for Hwange, Zimbabwe 1995–1998. Left Y-axis: Case prevalence according to model and according to clinical reports for Hwange, Zimbabwe years 1995–1998 51 . Model results are rescaled. Right Y-axis: rainfall in mm, using station data (NCDC). Blue line Rainfall. Dashed red line Modelled cases. Solid line Reported cases. Discussion The weather-based dynamic malaria model has been driven here using a reanalysis-climate data set, which is considered to represent the actual historic meteorology (at the appropriate temporal and spatial scale) for the range of climatic variables predicted by the hindcasts in the DEMETER multi-model system. In addition, this reanalysis-based dataset is used to derive the initial conditions of the hindcasts and verification dataset for DEMETER hindcasts. Thus an assessment of the value of the biological malaria model driven by ERA weather data represents the potential value of a "perfect" forecast of climatic conditions to the prediction of malaria epidemics. The assessment of this ability of the weather-driven model to describe interannual variability in malaria infection rates given a perfect forecast, is the basis for a realistic assessment of the benefit associated with the use of operational (and, therefore, unvalidated and imperfect) seasonal climate forecasts in the context of a malaria early warning system. The dynamic approach claims that total knowledge of the initial state and of the equations, as well as of the external driving forces, allows total knowledge of all future states. However, in real situations, this precise knowledge is often lacking. In many cases, bifurcations of the motion in phase-space due to minute perturbations of initial conditions, may result in large differences in the result, i.e. chaos. Chaotic behaviour may, however, be limited by the application of saturation limits to certain model parameters (such as a self-limiting proliferation rate). In addition, the knowledge of the equations themselves is often limited, being typically a linearization of behaviour of empirical data. Thus, there is considerable uncertainty in the result. The dynamic, deterministic method may, however, be used numerically, experimentally, by using small steps so as to validate the stability of the equations. For this reason, the improvement of the understanding of both the ability of weather forecasts to predict weather and of biological models to predict disease, presents a path to the understanding of probabilistic solutions to non-linear epidemic prediction problems. The capability of seasonal climate forecasts to predict anomalous seasonal climate conditions has improved considerably over the last decade. In particular, the El Niño/La Niña cycle has been correlated with extreme weather conditions throughout the globe[ 50 ]. This cycle has been repeatedly related to malaria epidemics. If the existence of an anomalous season can be predicted with skill, and if the relation between anomalous weather and localized malaria epidemics can be determined from time series analysis of a number of events[ 51 ], one may assess the probability of unusual increases in malaria transmission resulting from anomalous seasonal climates. This stage is, however, limited to reconstructions. The good fit in Fig. 2 suggests that the rate of maturation of larvae is linear with temperature. As the measurements were not from a controlled experiment, but from naturally occurring pools, the fit is surprisingly good. Fig. 3 shows the relation between surviving progeny and weather, using rainfall-determined oviposition and the dependence of maturation on temperature. This requires high temperatures (20+ °C) and at least moderate rainfall (10–20 mm/dekad (10 day period)). These conditions suit the regions of known habitation of An. gambiae s.l. In some regions, the existence of permanent waterbodies, such as slow-flowing rivers, lakes or swamps may provide suitable breeding sites, and thus make up for the shortage in rainfall, as far as the larvae are concerned. Even though the mature mosquito survival will be considerably reduced by the low humidity[ 52 ], malaria incidence during dry seasons will be possible[ 53 ]. These effects are not yet implemented in this model. The requirement for high temperatures for malarial transmission is further illustrated by the plots in Figs. 4 and 5 . Even though mosquitoes may well survive and multiply during the summer in temperate regions, they may not become vectors for transmission of falciparum malaria, unless the temperature remains in the high 20s (°C) for considerable periods. For this reason falciparum malaria is associated heavily with tropical regions, while in pre-eradication malarious Europe P.. vivax was the dominant malaria parasite species. Further improvements should be added to the model. These may be achieved as new relevant numerical information becomes available on the biological processes, which were here handled somewhat heuristically. For example, the relationship between rainfall and larvae survivorship was simplified due to the lack of data. Laboratory and field, meteorological and entomological data may establish its true form. The relative importance of the various parameters, assessed by sensitivity analysis by variation of individual parameters is of great interest. This will allow field scientists to focus their efforts in establishing the values of the most critical parameters. In a parallel paper this has been presented. However, a full multivariate evaluation is still underway, using novel parallel computing methods. The model currently ignores both antiparasitic immunity (immunity to infection) and antitoxic immunity (immunity to disease) – the sharp distinction between which may not exist in reality[ 54 , 55 ]. Neither of these forms of immunity is relevant in areas where 'true' epidemics result from climate anomalies, as it is widely assumed that in these areas the population is largely non-immune and that severe morbidity and mortality due to malaria may occur in all age groups. An age-stratified model is being undertaken at present, but its verification will require considerable input from new targeted longitudinal studies. However, where the model may be used to predict the impact of an intervention amongst a semi-immune or immune population then both forms of immunity may be significant factors in determining the transmission dynamics of malaria infection[ 56 ]. Thus, in endemic areas there will be a need to separate the population between adults and under-fives, and collect suitable data. This work is underway at present. Numerical evaluations of the model in both time and space show that it has a good first order approximation to the prevalence of infection across the continent. It captures well both the seasonality and interannual variability of infection at the test site in Zimbabwe. Note that there is a large level of under-reporting of clinical cases (due to lack of access to health services) and also over-diagnosis of malaria, which often confound correlations (Barnish, personal communication) [ 57 , 58 ]. The model, when run with the following inputs (Table 1 ), is able to capture the gross spatial dynamics of malaria transmission across the African continent (Fig. 7 ). The model represents well both the endemic stable areas of transmission[ 14 ], as the shaded regions where average prevalence of infection is above 20%, and the epidemic zones in red, which are not detected by more standard methods. Of course, as with both these previous models, it is clear that there are inherent limitations in trying to fit one model to the entire African continent. In the two examples cited this has been overcome by splicing together two separate models. The development of strain-specific datasets of entomological time-series will provide the basis for separate dynamic models for areas of varying relative abundances of anopheline species. This is one of the reasons for the northern limit of malaria epidemics as depicted being somewhat conservative. Table 1 Values of parameters used B 0.5 D d 37 degree days H 14 calendar days M IP 0.5 N 37 S C 111 degree days α 0.5. χ 0.5. γ 1.0 δ 0.9716. ν 14. Figure 7 Spatial epidemicity of malaria. Interannual standard deviation of incidence of infection as determined from the model run with daily ERA-40 data for the period 1987–2000. Regions with average prevalence rates of >20% (stable malaria) are shaded. Note that since the results are annual averages the values are likely to be less than those recorded from point prevalence surveys during the period of peak malaria transmission. Stronger variation, both between seasons and interannual, may be expected. Thus malaria transmission will continue further north than suggested, but will also be less stable. There are other points to consider. The most obvious discrepancy between our results and those of previous workers is that of high prevalence rates being predicted for Somalia – an area where malaria transmission in normally low. The results are created from a relatively short time series and may be disproportionately influenced by the anomalously high rainfall recorded in 1997/1998 in this area, and in East Africa in general, which corresponded with unusually high rates of transmission[ 3 ]. The malaria epidemic in southern Africa 1996–7, resulting from the heavy rains during the 1995–6 and 1996–7 rainy seasons in Zimbabwe and its region are reflected by the strong anomaly there. The epidemic character of the East African highlands is well represented too. The southward spread of actual malaria cases, as opposed to model cases, is limited by large-scale prevention schemes in South Africa. Obviously, the crude resolution of the weather data driving the model (approx 111 km 2 ) means that the model can only represent the most marked changes in transmission potential over large geographic regions – appropriate to seasonal climate forecasts. This is expressed in the poor spatial resolution as regards Madagascar, in which the spatial variation of prevalence is not represented. Despite the limitations, which may be overcome in specific regions by using daily weather data at a finer spatial resolution formed by downscaling (a process under considerable meteorological research), the model is capable of describing in general terms the spatial, seasonal and inter-annual variability of malaria transmission in Africa. The usage of T max -5 as a surrogate for daily temperature was an ad hoc attempt to create a single value for a wide spectrum varying in time and place. The diurnal variation is, for example, usually far greater in dry regions than in humid conditions. Our attempt was to use a single value which bears relation to something measurable, namelyT max . In fact, the two-metre model temperature value, was has been used may not have been the ideal, as larval development is closer to ground, but that would have required detailed soil data, which was beyond the scope of this research. The usage of reanalysis data is obviously inferior to high quality station data. Convective precipitation is highly local and is not well correlated to the averaging required by meteorological reanalysis. For example, the heavy rains in north-western Zimbabwe (end 1995) were under-expressed in ERA-40, partly due to the scale of the reanalysis data, and thus the local malaria incidence using that data is too low. The integration of the data in a reanalysis model in itself makes numerous physical assumptions. In some cases, the result is poor quality of even some large-scale processes, such as the El-Niño of 1997 in East Africa. High quality station data at the pan-African level is, however, not available. The stations are irregularly dispersed, and not all produce complete data sets, having many missing days. As the model requires daily values, interpolation over large areas with varying orographic characteristics is required with the inherent uncertainty this brings. In addition, the relation between mosquito habitat microclimate and station data too is unknown. High temporal and spatial resolution of weather data will improve the modelling attempts. Among other issues, the continuous measurement of weather data at malaria epidemiological and entomological research and surveillance sites now introduced will allow development of coupled malaria and meteorological data sets, which will be more effective for future analysis. Development of seasonal climate forecasting tools over large geographic areas, however, will remain for some time partly dependent on verification by reanalysis. Thus, malaria epidemic seasonal early warning will also be linked to this imperfect data source, though perhaps improved by novel downscaling methods. Conclusions This paper presents a first step in the preparation of a weather-driven dynamical model of malaria transmission, for use with both observed weather data and seasonal climate forecasts. The model incorporates the stages of the malaria vector and their dependence on temperature and rainfall, and part of the within-host parasite population dynamics. Some of these elements lack concrete theoretical and empirical development, requiring further input. Further work, under work at present, will enable the employment of such a model in the prediction of outbreaks based on skilful weather forecasts. Authors' contributions MBH formulated the mathematical model, prepared the code and ran the program. APM lead the applications model work package in DEMETER.. Both authors read and approved the final manuscript. Table 2 List of symbols used in text B Human blood index, the preference of a mosquito to bite humans and not other animals D d Length of gonotrophic cycle in degree days G c Length of gonotrophic cycle in days H Hepatic stage in days H IR Human new infection rate H i Number of mosquitoes biting infected humans in a day H n Number of mosquitoes biting uninfected humans in a day H(s,t) Human population at stage s of the development of infection at time t. s = 0 symbolises an uninfected host. I(s,t) Immature mosquito population at stage s of maturation cycle (in degree days) at time t L e , L 1 , L p Length of egg, larval and pupal stage of mosquito maturation in days M(s,S s ,t) Mature mosquito population which is at stage s of gonotrophic cycle and stage S s of sporogonic cycle at time t. M IP Infection probability of a single mosquito for each bite m Maturation rate of larvae in reciprocal days N Number of sections into which the gonotrophic cycle was divided. p Mosquito population daily survival rate P R Fraction of gonotrophic cycle covered in one day r Fraction of infected humans out of total human population R d Dekadal (ten daily) rainfall in mm R 0 Single case multiplication factor: number of secondary cases induced per case s Dummy variable representing stage of development in degree days S C Length of sporogonic cycle in degree days S R Daily progression of sporozoites in degree days S S Stage of sporogonic cycle in degree days. S S = 0 represents an uninfected mosquito T c Threshold temperature for gonotrophic or sporogonic cycle α Fractional per-gonotrophic cycle survival of mosquito. χ Fraction of mosquitoes biting infective humans that become themselves infected. γ Ratio of brood of each ovipositing mosquito to rainfall. δ Fractional per diem survival rate of human infection. Hence 1-d is the daily infection clearance rate. φ Total number of mosquitoes. ν Number of sections into which the larval cycle was divided. σ Per diem larval survival rate	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC520827.xml
517721	Long acting β2 agonists for stable chronic obstructive pulmonary disease with poor reversibility: a systematic review of randomised controlled trials	Background The long acting β2-agonists, salmeterol and formoterol, have been recommended, by some, as first line treatment of stable chronic obstructive pulmonary disease (COPD). We reviewed evidence of efficacy and safety when compared with placebo or anticholinergic agents in patients with poorly reversible COPD. Methods After searching MEDLINE, EMBASE, HealthSTAR, BIOSIS Previews, PASCAL, ToxFile, SciSearch, the Cochrane Library, and PubMed, as well as Web sites, selected journals, reference lists, and contacting drug manufacturers, two reviewers independently screened reports of randomised controlled trials of parallel or crossover design lasting four weeks or longer and including patients with a forced expiratory volume in one second (FEV1) ≤ 75% of predicted, a ratio of FEV1 to forced vital capacity (FVC) ≤ 88% of predicted, and < 15% improvement from baseline FEV1 after a dose of a β2 agonist. We included trials comparing salmeterol or formoterol with placebo or with ipratropium bromide and reporting one of these outcomes: lung function; exercise capacity; quality of life scores; dyspnea; exacerbations; rescue inhaler use; incidence of tachycardia, hypokalemia, or dry mouth. Two reviewers assessed the quality of included reports using the Jadad scale and allocation concealment, and abstracted data. Results Twelve trials satisfied our inclusion criteria; eight were high quality (Jadad score >2) and four were low quality (≤ 2). The adequacy of allocation concealment was unclear in all of them. We did not perform a meta-analysis due to differences in trial design and how outcomes were reported. Two trials comparing salmeterol with ipratropium did not detect differences; one trial comparing formoterol and ipratropium described greater improvement with formoterol in morning PEFR (15.3 versus 7.1 l/min, p = 0.040). Of twelve trials comparing long acting β2 agonists with placebo, six reported no improvement in exercise capacity, eleven reported improvements in FEV1 lung function (one reported no improvement), six reported less rescue inhaler usage (one reported no difference) and five reported improved dyspnea scores (two reported no improvement). Differences in quality of life were detected in one salmeterol trial ; however, two salmeterol, and one formoterol trial reported no differences. Adverse effects of interest were not reported. Conclusion In terms of clinical outcomes and safety, we could not find convincing evidence that salmeterol and formoterol have demonstrated advantages to ipratropium, a less expensive drug, for patients with stable COPD and poor reversibility. Compared to placebo, we found evidence of reduced rescue inhaler usage and improved spirometric outcomes without a significant impact on quality of life or exercise capacity.	Background Bronchodilators are the primary agents used to manage chronic obstructive pulmonary disease (COPD). They modestly improve forced expiratory volume in one second (FEV1) and reduce dynamic hyperinflation; breathlessness may lessen and exercise tolerance increase despite little improvement in spirometric measurements [ 1 ]. The bronchodilators currently available for COPD include β2 agonists (e.g., salbutamol and salmeterol), anticholinergics (e.g., ipratropium bromide) and methylxanthines (e.g., theophylline). According to the Canadian guidelines for the treatment of stable COPD [ 2 ] first line treatment consists of ipratropium, two to four doses three to four times daily, plus a short acting β2 agonist administered on an "as needed" basis. If the patient uses substantial amounts of short acting β2 agonists, or if the symptoms are greater at night than in the early morning, a long acting β2 agonist (salmeterol or formoterol) is added twice daily. However, recently some have recommended the latter as first line agents for stable COPD [ 3 ], [ 4 - 7 ] potentially replacing the less expensive ipratropium [ 4 ], [ 5 - 8 ]. Several trials have demonstrated the usefulness of salmeterol and formoterol for the management of COPD[ 8 ], [ 9 - 11 ]. According to a 1998 meta-analysis [ 12 ], in patients with non-reversible COPD these agents produce small increases in FEV1; however, these changes alone may not correlate highly with symptom relief [ 13 ]. The authors of the meta-analysis suggested that these drugs be prescribed only for patients who find they provide definite clinical improvement: reduced breathlessness or better exercise capacity. All three trials in the meta-analysis [ 10 , 14 - 16 ]. compared long acting β2 agonists with placebo. Since then, other studies of these agents in COPD, including comparisons with anticholinergics [ 8 , 11 ] have appeared in the literature. Canadian provincial drug plan managers have noted a substantial increase in the use of salmeterol and formoterol in recent years, an observation supported by data from International Medical Services Canada, which collects information on Canadian patterns of drug prescribing and estimates use: between 1997 and 2001, the use of salmeterol and formoterol for COPD increased 1,150% and 1,975%, respectively, whereas the use of ipratropium for COPD decreased by 37% [ 17 ]. In light of the new trials and the recent changes in prescribing practices, we undertook a systematic review to evaluate the efficacy and safety of long acting β2 agonists when compared with placebo or anticholinergic agents in patients with stable, poorly reversible, COPD. Methods Searching We obtained published literature and conference abstracts for this document from two separate sources: (1) search results from the CCOHTA's published health technology review "Long-acting β2-agonists for maintenance therapy of stable chronic obstructive pulmonary disease: a systematic review"; and, (2) search results from CCOHTA's ongoing clinical review on long-acting β 2 -agonists for maintenance treatment of stable chronic obstructive pulmonary disease in mixed population. The first search was performed on MEDLINE ® , EMBASE ® , HealthSTAR, BIOSIS Previews ® in June 2001 using a sensitive search strategy. The second search performed in December 2002 had a more focused search strategy and included PASCAL, SciSearch and ToxFile databases in addition to MEDLINE ® , EMBASE ® and BIOSIS Previews ® databases. As designed in the search strategy, this search captured all the studies included in CCOHTA's published review as well as some additional trials published since previous search date. Search details for both searches can be found in Appendix 2 [see Additional file 2 ]. Regular alerts have been established on these databases to capture new studies and are ongoing in 2004. Parallel searches were performed and updated in PubMed and the Cochrane Library. In addition, we periodically searched Web sites of clinical trial registries and health technology assessment (HTA) and related agencies. Google™ and other search engines were used to retrieve conference abstracts of major respiratory associations. We also hand searched selected journals and documents in the library of the Canadian Coordinating Office for Health Technology Assessment and the bibliographies of retrieved reports. As well, we contacted the Canadian offices of the manufacturers of salmeterol and formoterol for nonconfidential information on unpublished studies. Selection Two reviewers (D R H and V K S) worked independently on these phases of the study. Disagreements were resolved by discussion and consensus; a neutral third party (M B) was consulted when necessary. The reviewers evaluated the 504 unique citations by reviewing titles and abstracts, discarding those deemed irrelevant (e.g., case reports, review articles, and studies unrelated to the use of β2 agonists for maintenance treatment of stable COPD). They then selected all reports of randomised controlled trials (RCTs) comparing salmeterol or formoterol with placebo or an anticholinergic agent, with or without the additional use of short acting β2 agonists. No restrictions were placed on dosage, but the trials had to be of either parallel or crossover design, have lasted four weeks or longer, and have included patients that met each of the following criteria. • Non-asthmatic. • Stable COPD: no infections, exacerbations, or hospitalizations in the past month. • FEV1 ≤ 75% of predicted. • Ratio of FEV1 to forced vital capacity (FVC) ≤ 88% of predicted. • After a dose of a short or long acting β2 agonist < 15% improvement in FEV1. Since bronchodilators are much more efficacious in asthma than in COPD, including patients with asthma would have influenced the findings. It may be difficult to determine whether chronic airflow obstruction with relatively large responses to short acting β2 agonists represents COPD with reversibility or asthma with incomplete reversibility. A suggestive feature in the differential diagnosis of COPD is irreversible airflow limitation.[ 18 ] To better reflect this and to minimize the chance of including patients with asthma, we excluded those trials in which the average FEV1 response to a bronchodilator was greater than or equal to 15%. In addition, the trials had to have investigated one of the following outcomes. • Lung function, including FEV1 and peak expiratory flow rate (PEFR). • Exercise capacity: six minute or shuttle walking test. • Health related quality of life (QoL). • Dyspnea, including symptom diary scores. • Exacerbations of COPD. • Rescue use of salbutamol, a short acting β2 agonist. • Adverse effects, including tachycardia, hypokalemia, and dry mouth. Validity assessment The reviewers independently scored the quality of the included trial reports using a five-point scale described by Jadad [ 19 ], which assigns zero to two points each for appropriateness of randomization and double blinding and zero to one point each for reporting on withdrawals and dropouts; low scores are associated with exaggerated estimates of benefit. Concealment of allocation to treatment was also categorized as adequate, inadequate, or unclear. Data abstraction The reviewers independently recorded characteristics of the trials and patients, as well as details of the interventions and outcomes. When outcome data were available only graphically, each reviewer estimated values, and the means of the two estimates were reported. Quantitative data synthesis When possible, we calculated mean differences with 95% confidence intervals (CIs) for continuous outcomes and odds ratios (ORs) with 95% CIs for binary outcomes for individual trial data using Statistics with Confidence software [ 20 ], We used intention-to-treat data when available and otherwise end point data for patients completing the trials. Qualitative data were recorded descriptively. We had intended to do a meta-analysis, pooling data on outcomes of interest. This approach is useful when the samples of individual studies are too small for detection of an effect and when results from several trials disagree in magnitude and direction of effect [ 21 ]. However, it is only appropriate when the trials are clinically homogeneous. We found that even commonly measured outcomes, such as FEV1, could not be combined by meta-analysis because of differences in how they were reported. For example, in the six trials comparing salmeterol with placebo, FEV1 was reported as a mean change in percent predicted[ 16 ], a mean change overall[ 15 ], a mean difference between trial arms[ 10 ], no difference (without data)[ 22 ], baseline and overall FEV1 (after 24 hrs without medication)[ 8 ] and as an 0 to 12 hour area-under-the-curve (FEV1-AUC) function[ 23 ] We were not successful in obtaining more data from study authors. We also had concerns about the meta-analysis of data from trials of parallel and crossover design[ 24 ] and differences in spirometry protocols including allowable medications. Therefore, we decided on a best evidence synthesis approach [ 25 ] instead. Results Trial Flow Both reviewers agreed to tentatively accept 35 of the 58 potentially relevant reports. After further evaluation one reviewer disagreed with including 14 of the 35, which resulted in a moderate level of agreement (Kappa = 0.58; 95% CI 0.39 to 0.78). Discussion revealed that this difference related primarily to confusion surrounding interpretation of one of the criteria for eligibility, and ultimately the other reviewer agreed to reject the disputed reports. The reviewers then independently selected the same nine reports [ 8 , 10 , 14 - 16 , 22 , 23 , 26 , 27 ] for final inclusion. Figure 1 illustrates the study selection process. The updated search strategy identified 24 additional potentially relevant reports. Of these, four reports were independently selected based on the inclusion criteria. There were no disagreements. Appendix 1 [see Additional file 1 ] presents all of the 69 reports excluded with reasons. Figure 1 Flow diagram of RCT screening and selection procedure. Process through which reports were selected from those potentially relevant. RCT = randomised controlled trial; ROAD = reversible obstructive airways disease; FEV1 = forced expiratory volume in one second. Study characteristics The 13 reports were of 12 trials, all funded by manufacturers of the drugs. One report [ 14 ] describes outcomes in a subset of patients fully described in another report [ 10 ]. One reports was a conference abstract [ 22 ]; the other twelve reports were journal articles. Duplicate reports were used as a source of supplementary information. Based on the reports, eight of the trials [ 8 , 15 , 16 , 23 , 26 - 29 ] were of high quality (score > 2) and four[ 10 , 22 , 30 , 31 ] of low quality (score ≤ 2). Concealment of the allocation sequence was unclear from all of the trial reports. The reviewers agreed completely about quality. Table 1 presents details of the trials and patients. Table 1 Characteristics of included randomised, double blind, controlled trials of long acting β 2 agonists in maintenance therapy for chronic obstructive pulmonary disease (COPD) First author, year of publication, design Trial quality Patients meeting inclusion criteria Interventions Outcomes investigated Notes Ulrik, 1995[16] Crossover 3 66 current smokers with FEV 1 of 1–2 L (< 60% of predicted) and FEV 1 /FVC < 60% of predicted. FEV 1 of <15% or 300 ml after salbutamol Salmeterol (50 μg twice daily) or placebo for 4+4 weeks; no crossover washout. FEV 1 , PEFR, daytime and night-time symptom scores, rescue use of salbutamol. Two week run in. Methylxanthines, corticosteroids (short oral courses) allowed. Newman, 1996[22] (abstract) Crossover 2 42 patients with mean FEV 1 of 0.93 L (35% of predicted) and no response to oral steroids. Salmeterol (100 μg twice daily) or placebo for 8+8 weeks. FEV 1 , FVC, six minute walk test and Borg dyspnoea assessment,[26] daytime and night-time symptom scores, rescue use of salbutamol, proportion of days unable to perform normal activity, incidence of adverse events and COPD exacerbations. Two week run in. Salbutamol rescue allowed. Grove, 1996[15] Crossover 3 29 patients with FEV 1 25%–75% of predicted and 5%–15% reversibility with 200 μg of salbutamol. Salmeterol (50 μg twice daily) or placebo for 4+4 weeks; one1 week crossover washout. FEV 1 , FVC, TLC, RV, 6 minute walk test and exertion on Borg scale, oxygen uptake. At least one week run in. Inhaled corticosteroids, anticholinergics, oral theophylline allowed. Boyd, 1997[10] Parallel 2 674 patients with FEV 1 ≤ 70% and FEV 1 /FVC ratio ≤ 60% of predicted and 5%–15% reversibility of FEV 1 with 400 or 800 μg of salbutamol. Salmeterol (50 or 100 μg twice daily) or placebo for 16 weeks. FEV 1 , six minute walk test and Borg dyspnoea assessment, daytime and night-time symptom scores, rescue use of salbutamol. Two week run in. Medications other than β 2 agonists allowed. Jones, 1997[14] Parallel 2 283 patients with FEV 1 ≤ 70% and FEV 1 /FVC ratio ≤ 60% of predicted; 5%–15% reversibility of FEV 1 with 400 or 800 μg of salbutamol. Salmeterol (50 or 100 μg twice daily) or placebo for 16 weeks. HRQoL with SGRQ27 and SF-36[28]. Two week run in. Medications other than β 2 agonists allowed. Mahler, 1999[8] Parallel 3 145 patients with FEV 1 ≤ 65% and FEV 1 /FVC ratio ≤ 70% of predicted; ≤ 15% reversibility of FEV 1 with short acting β 2 agonist; grade 1 baseline severity of breathlessness. Salmeterol (42 μg twice daily) or ipratropium bromide (36 μg four times daily) or placebo for 12 weeks. FEV 1 AUC, six minute walk test, daytime and night-time symptom scores, dyspnoea on BDI and TDI,[29] supplemental use of salbutamol, HRQoL on CRDQ,[30] COPD exacerbations. Run in six hours to three days. Prednisone (≤ 10 mg) or equivalent or inhaled corticosteroids allowed. Rennard, 2001[23] Parallel 3 179 patients with FEV 1 ≤ 65% and FEV 1 /FVC ratio ≤ 70% of predicted; ≤ 12% reversibility of FEV 1 with salbutamol; score ≥ 1 on MMRC five point dyspnoea scale. Salmeterol (42 μg twice daily) or ipratropium (36 μg four times daily) or placebo for 12 weeks. FEV 1 and FVC AUC, dyspnoea on BDI and TDI, six minute walk test and Borg dyspnoea assessment, symptom scores, QoL on CRDQ, COPD exacerbations. Corticosteroids, inhaled and oral (< 10 mg/d), allowed. Rossi, 2002[27] Parallel 3 418 patients with FEV 1 < 70% and FEV 1 /FVC ratio ≤ 88% of predicted; < 15% reversibility of FEV 1 with short acting β 2 agonist; grade 1 baseline severity of breathlessness. Formoterol (12 or 24 μg twice daily) or placebo or oral slow release theophylline for 12 months. FEV 1 AUC. Inhaled corticosteroids and rescue use of salbutamol allowed. Stahl, 2002[26] Parallel 3 183 patients with FEV 1 < 60% and FEV 1 /FVC < 70% of predicted; < 12% reversibility of FEV 1 after single dose of formoterol. Formoterol (18 μg twice daily) or ipratropium (80 μg three times daily) or placebo for 12 weeks. FEV 1 , FVC, PEFR, shuttle walking test, morning and evening symptom scores, HRQoL on SGRQ. Inhaled corticosteroids at constant doses and rescue use of short acting β 2 agonists allowed. Gupta, 2002[29] Parallel 4 33 patients with FEV 1 < 60 % predicted and FEV 1 /FVC ≤ 70%; reversibility <12 % improvement of FEV 1 after 400 μg salbutamol Salmeterol (50 μg twice daily) or placebo twice daily for 8 weeks FEV 1 , FVC, six minute walk test, HRQoL on SF-36[28], dyspnoea on BDI, patient self-assessment, and rescure inhaler usage Two week run in period. Patients required to take beclomethasone 400 μg twice daily and ipratropium 20 μg four times daily. Mahler, 2002[30] Parallel 2 158 patients with FEV 1 < 65 % predicted and FEV 1 /FVC ≤ 70%; reversibility <12 % improvement of FEV 1 after 400 μg salbutamol Salmeterol (50 μg twice daily) or placebo twice daily for 24 weeks FEV 1 , morning PEF, dyspnoea on BDI and TDI; rescue salbutamol use; HRQoL on CRDQ [30]; symptoms on CBSQ Randomization stratified by reversibility. Calverly, 2003[28] Parallel 5 733 patients with FEV 1 25–70% predicted and FEV 1 /FVC ≤ 70%; reversibility <10 % of predicted FEV 1 after salbutamol Salmeterol (50 μg twice daily) or placebo twice daily for 52 weeks FEV 1 , FVC, relief medication, symptom scores, night-time awakenings, exacerbation rates, HRQoL on SGRQ Two week run in and two week follow up Hanania, 2003[31] Parallel 2 163 patients with FEV 1 < 65% predicted but > 700 ml (or if ≤ 700 ml > 40 % predicted) and FEV 1 /FVC < 65%; reversibility < 12 % of predicted FEV 1 after salbutamol Salmeterol (50 μg twice daily) or placebo twice daily for 24 weeks FEV 1 , morning PEF, dyspnoea on BDI and TDI; rescue salbutamol use; HRQoL on CRDQ [30]; symptoms on CBSQ, exacerbation rates (all severities) Randomization stratified by reversibility AUC = area under the curve; BDI = baseline dyspnoea index;[29] CBSQ = chronic bronchitis symptom questionnaire; [42] CRDQ = chronic respiratory disease questionnaire;[30] FEV1 = forced expiratory volume in one second; FVC = forced vital capacity; HRQoL = health related quality of life; MMRC = Modified Medical Research Council; PEFR = peak expiratory flow rate; RV = residual volume; SF-36 = Medical Outcomes Study Short Form 36;[28] SGRQ = St. George's Respiratory Questionnaire;[27] TDI = transition dyspnoea index;[29] TLC = total lung capacity. Data synthesis Comparative efficacy of long acting β2 agonists and anticholinergic agents Three trials [ 8 , 23 , 26 ] that compared long acting β2 agonists and anticholinergic agents were identified. Two 12 week trials compared salmeterol, ipratropium, and placebo [ 8 , 23 ].; however, only one trial [ 8 ] reported data for FEV1 and transition dyspnea index (TDI) scores for the subset of patients that met our inclusion criteria, and the data were presented graphically. No significant differences (p > 0.05) between the salmeterol and ipratropium groups were observed in the change in FEV1 from baseline, in TDI scores, or in the rescue use of salbutamol [ 8 ]. In a 12 week trial [ 26 ] formoterol produced significantly greater improvement in morning PEFR from baseline to endpoint than ipratropium (15.3 versus 7.1 l/min, p = 0.040). However, the differences between the active treatment groups were not significant (p > 0.05) for percent predicted FEV1 (13% versus 7%, p > 0.05), percent predicted FVC (8% versus 8%, p > 0.05), improvement in breathlessness score (-0.21 versus -0.29, p > 0.05), or improvement in the St. George's Respiratory Questionnaire (SGRQ) total score (0.0 versus -0.5, p > 0.05). Data on adverse effects of interest, including tachycardia, hypokalemia, and dry mouth, were not available from the reports. Comparative efficacy of long acting β2 agonists and placebo Ten trials [ 8 , 10 , 14 - 16 , 22 , 23 ], [ 28 - 31 ] had salmeterol and placebo treatment arms; the other two [ 26 , 27 ]. had formoterol and placebo arms. Table 2 and the following text summarize outcome data only for the patients that met our inclusion criteria. Table 2 Selected results First author FEV 1 Symptom scores (lower is better) Salmeterol versus placebo Ulrik[16] No significant differences in reversibility of percent predicted FEV 1 with treatment. Mean (SE): 2.7% (0.4) versus 3.4% (0.4). Significant differences in median (range) symptom scores during treatment. Daytime (scale 0–5): 1.0 (0–3.4) versus 1.8 (0.1–4.0). Night-time (scale 0–4): 0.9 (0–3.4) versus 1.6 (0.1–4.0). Newman[22] No significant differences in measurements with treatment (data not reported). Symptoms significantly reduced during salmeterol compared with placebo treatment. Scale and scores not reported. Grove[15] Significant differences one and six hours after single dose and six hours after four weeks of treatment. Mean change: 120 versus 10 ml after four weeks. Boyd[10] Significant differences in improvement with treatment. Mean difference (95% CI): for salmeterol 50 μg versus placebo 97.80 (55.6 to 139.99) ml; for salmeterol 100 μg versus placebo 117.60 (67.88 to 167.32) ml. Significant difference in distribution of median daytime and night-time symptom scores between active treatment and placebo groups (CI 0.0 to 0.0 in all cases) but not between active treatment groups. Daytime (scale 0–5): baseline, 2 in each group; from week 5, 1 in active treatment groups and 2 in placebo group. Night-time (scale 0–4): baseline, 1 in placebo and salmeterol 50 μg groups and 0 in salmeterol 100 μg group; from week 1, 0 in salmeterol 50 μg group and no change in other groups. Jones[14] (Presented QoL results for subset of patients described in Boyd[10].) Gupta[29] A mean increase in predose FEV 1 of 170 ml (distibution not reported) for salmeterol vs. a mean decrease of 20 ml (distribution not reported) for placebo after 8 weeks. Both salmeterol and placebo produced significant improvemnts in BDI scores, however the magnitude of increase was greater vs. placebo (3 vs. 1); 100% patients treated with salmeterol reported decreased cough and dyspnea vs. 69% (11/16) of placebo recipients Mahler 2002 [30] A mean increase of 80 ml (95%CI 35 to 125) for salmeterol vs. mean decrease of -8 ml (95%CI: -53 to 37) for placebo. Two-hour post-dose FEV 1 mean increase of 175 ml (95%CI: 116 to 234) vs. mean increase of 28 ml (95%CI: -17 to 73) Mean increase of 0.5 (SE 0.4) in TDI for salmeterol recipients and 0.4 (SE 0.3) for placebo recipients. Not clinically or statistically significant. Calverly [28] A mean increase in predose FEV 1 of 25 ml vs. a mean decrease of -38 ml (P < 0.05) in salmeterol and placebo recipients. Smaller difference for two-hour post-dose FEV 1 (data not reported). Mean scores for cough (scale 0–3); breathlessness (scale 0 to 4); sputum production (scale 0 to 3); sputum colour (scale 0 to 4): salmeterol: cough 1.36 (SE0.03); breathlessness 1.59 (0.03); sputum production 1.30 (0.03) and colour 1.35 (0.03) vs. placebo: cough 1.44 (0.03); breathlessness 1.66 (0.03), sputum production 1.34 (0.03) and colour 1.36 (0.03). Hanania[31] A mean increase of 26 ml (95%CI: -27 to 79) for salmeterol vs. mean increase of 19 ml (95%CI: -26 to 64) for placebo. Two-hour post-dose FEV 1 mean increase of 119 ml (95%CI: 70 to 168) vs. mean increase of 71 ml (95%CI: 24 to 118) The magnitude of TDI responses was less in non-reversible vs. reversible patients. (Data are not reported) Salmeterol versus ipratropium versus placebo Mahler[8] Significant differences between active treatment and placebo groups but not between active treatment groups. Peak improvements with treatment: 155, 165, and 24 ml, respectively. No significant differences in change of mean daytime symptom score with treatment. No significant differences in TDI except between ipratropium and placebo groups at week 8. After 12 weeks, mean TDI 0.35, 0.98, and 0.48, respectively. Rennard[23] Significant differences between active treatment and placebo groups but not between active treatment groups. FEV 1 AUC 0–12 hour responses significantly greater with salmeterol and ipratropium than with placebo (data not reported). Formoterol versus placebo Rossi[27] Significant differences in estimated difference in FEV 1 AUC 0–12 hour responses: between formoterol 12 μg and placebo groups, 145 ml; between formoterol 24 μg and placebo groups, 141 ml. (Individual values for treatment groups not available.) Formoterol versus ipratropium versus placebo Stahl[26] Significant differences in improvement in percent predicted FEV 1 between active treatment and placebo groups but not between active treatment groups: 13%, 7%, and 6%, respectively. Significant differences between active treatment and placebo groups in change from baseline in breathlessness (scored 0 to 4 morning and evening). Means: -0.21, -0.29, and 0.0, respectively. CI = confidence interval; SE = standard error. Lung function FEV1 As table 2 shows, the changes in FEV1 from baseline to endpoint differed significantly (p < 0.05) between the salmeterol and placebo groups in eight of ten trials and between the formoterol and placebo groups in two trials. FVC Five trials[ 15 , 23 , 26 , 28 , 31 ] reported on this outcome. In one 4+4 week (4 weeks then crossover then 4 additional weeks) trial [ 15 ] the increase in FVC was significantly greater with salmeterol than with placebo six hours after a single dose (200 versus 30 ml, 95%CI for difference: 40 to 290) but not after four weeks of treatment (150 versus 130 ml, 95%CI for difference: -180 to 220). In one 12 week study [ 23 ] the change in FVC was significantly greater (p < 0.001) for salmeterol (and ipratropium) than for placebo on day 1, there was no loss of response during treatment, and after four weeks the morning predose values were significantly greater in the patients treated with either active drug (data not reported). In the other 12 week trial [ 26 ] the percent predicted FVC was significantly increased by the end of formoterol treatment, compared with placebo treatment, by 8% versus -0.4% (p = 0.02). In one 52 week trial).)[ 28 ], the change in mean FVC measured 12 hours after treatment was 86 ml greater (p = 0.004) in salmeterol recipients. The difference in mean change in FVC at 52 weeks was 200 ml between groups. In an 8 week trial[ 29 ], the mean increase in FVC was 280 ml in the salmeterol group compared to a fall of 8 ml in the placebo group (p < 0.05). PEFR Three trials[ 16 , 26 , 28 ].).) reported on this outcome. In one four week trial [ 16 ] salmeterol treatment compared with placebo treatment produced a mean treatment difference in morning values of 12 l/min (238 compared with 226 l/min, 95%CI for difference: 6 to 17; p < 0.001); a statistical difference for evening values was not detected [242 (95%CI: 222 TO 262) and 237 (95% CI: 217 to 257) l/min for salmeterol and placebo, p > 0.1]. The diurnal variation was significantly lower during salmeterol treatment, at 3 (95%CI: -0.9 to 6.9) versus 11 (95%CI: 7.1 to 14.9) l/min; however, the mean treatment difference was only 7 (95%CI: 3 to 11) l/min. In the other trial, lasting 12 weeks [ 26 ], the change in morning PEFR was significantly greater by the end of formoterol (or ipratropium) treatment compared with placebo treatment: 15.3 versus -0.9 l/min (p < 0.001). In one 52 week trial).)[ 28 ], the change in mean PEF values differed significantly (p < 0.0001) for salmeterol treatment, at 257 l/min (95%CI: 253 to 261) versus placebo, at 242 l/min (95%CI: 238 to 246). Exercise capacity Results (but not always data) for six minute walk tests were reported from six trials [ 8 , 10 , 15 , 22 , 26 , 29 ]. None of the trials found statistically significant differences between salmeterol and placebo therapy, although one 12 week trial[ 8 ] found that at week 10 the patients receiving ipratropium walked a mean of 14 (95%CI: 0.3 to 27.7) yards farther in six minutes than those receiving placebo; there were no differences in prewalk or postwalk breathlessness between the treatment groups. The only other trial reporting data [ 15 ] found a median (interquartile range) distance in six minutes of 450 (371–491) m for placebo recipients and 425 (392–473) m for salmeterol recipients; the difference was not reported to be significant, but the patients receiving salmeterol (50 μg twice daily) perceived significantly less exertion by the end of treatment, as measured on the Borg scale [median (interquartile range) 0.5 (0–1) for salmeterol versus 1 (0–2) for placebo, p = 0.004] [ 32 ]. A 16 week trial [ 10 ] found a significant (p < 0.05) reduction in postwalk breathlessness (three or more points on the 10 point Borg scale) after eight and 16 weeks of 50 μg but not 100 μg of salmeterol twice daily, compared with placebo (OR 0.62 [95% CI 0.42 to 0.91]). Similarly, an 8+8 (8 weeks then crossover then 8 additional weeks) week trial [ 22 ] did not detect a significant (p > 0.05) difference in postwalk breathlessness between patients receiving 100 μg of salmeterol and those receiving placebo. One study comparing formoterol, ipratropium, and placebo [ 26 ] reported mean changes in walking distance from baseline to endpoint, measured with the shuttle walking test, of 19.2, 17.5 and 5.1 m, respectively; the differences were not significant (p > 0.05). Dyspnea In several trials [ 10 , 16 , 22 , 26 , 28 - 31 ] the patients assessed symptom severity every day, generally using ordinal scales. One 12 week trial comparing salmeterol, ipratropium, and placebo [ 8 ] measured the severity of dyspnea at baseline with a multidimensional baseline dyspnea index (BDI) and changes in severity every two weeks with a TDI [ 33 ]. As table 2 shows, some differences during treatment with an active drug as compared with placebo were significant and others were not. Rescue use of a short acting β2 agonist In five of six trials salmeterol treatment was associated with less salbutamol use than was placebo treatment [ 10 , 16 , 22 , 28 , 29 ] In one trial (4+4 weeks) [ 16 ] the median numbers (range) of daytime rescue doses were 1.7 (0–6.1) and 2.6 (0–7.9), respectively, and the median numbers of night-time doses 0 (0–4.2) and 0.3 (0–5.0). In a 52 week trial.[ 28 ], the median number of rescue inhalations per day was 2 for both salmeterol and placebo recipients, but these groups were statistically different (p = 0.028). Another trial[ 29 ] reported the mean number of doses of rescue salbutamol was significantly lower during treatment in salmeterol recipients (0.59, 95%CI: 0.30 to 0.88) versus placebo recipients (1.75, 95%CI: 1.33 to 2.17). In one 12 week trial comparing salmeterol, ipratropium, and placebo [ 8 ] no significant difference was observed in additional bronchodilator use between the placebo and active drug groups. Quality of life HRQoL was evaluated in four trials[ 14 , 26 , 28 , 29 ] In a salmeterol study [ 14 ] a subset of a larger patient group was asked to complete the disease-specific SGRQ [ 34 ] and the Medical Outcomes Study Short Form 36 (SF-36) [ 35 ] at baseline and after 16 weeks of treatment. The SGRQ has three components: distress due to respiratory symptoms, effects of disturbances on mobility and physical activity, and psychosocial impact of the disease; negative changes represent improvement. Data from 283 patients (95 in the placebo group and 94 in each salmeterol group) were analysed; data for others were excluded because of noncompletion of one or both questionnaires at 16 weeks or inability to meet quality control criteria or both. Salmeterol 50 μg (but not 100 μg) twice daily was associated with significantly greater improvement in mean (standard deviation) SGRQ scores from baseline to endpoint than was placebo: -6.8 (13.2) versus -1.4 (11.7) for total score and -8.0 (17.6) versus 0.0 (15.7) for impact score. No significant differences between placebo and either dose of salmeterol were observed in any of the domains of the SF-36 except for "role-emotional": these scores were significantly worse for recipients of salmeterol 100 μg than for recipients of placebo. In the 52 week study).)[ 28 ], health status was assessed with the SGRQ. The adjusted mean score was not statistically different in salmeterol recipients, at 45.2 (95%CI: 44.4 to 46.0) versus placebo recipients, at 46.3 (95%CI: 45.3 to 47.2). In an 8 week study[ 29 ] the magnitude of improvement for salmeterol versus placebo recipients rated on an SF-36 scale was significantly greater for the dimensions of "general health" (p = 0.008), "health change" (p = 0.026); physical functioning" (p = 0.008) and "vitality energy and fatigue" (p = 0.008) In the trial comparing formoterol, ipratropium, and placebo [ 26 ] HRQoL was also evaluated with the disease specific SGRQ. Of the 183 patients, 144 completed the assessment; reasons for not doing so were not reported. The changes from baseline to endpoint in total score were negligible in all three groups, at 0.0, -0.5, and 1.5, respectively. COPD exacerbations Three trials[ 10 , 22 , 28 ] reported on this outcome; only one trial).)[ 28 ] defined "COPD exacerbation" as episodes that required antibiotics or corticosteroids but not hospital admission; these occurred at a mean rate of 0.54 exacerbations/patient/year in salmeterol recipients and 0.76 exacerbations/patient/year in placebo recipients (p = 0.0003). In one 16 week trial [ 10 ] the numbers (and proportions) of patients having exacerbations among those receiving salmeterol 50 or 100 μg twice daily or placebo were 75 (33%), 91 (42%), and 98 (43%), respectively. In an 8+8 week trial [ 22 ] there were fewer exacerbations during treatment with salmeterol 100 μg twice daily than during treatment with placebo (p = 0.065); data were not presented. Adverse Effects Data on adverse effects of interest were not available from the reports. Discussion We identified thirteen reports of twelve randomized controlled trials describing the effect of administering the long acting β2-agonists, salmeterol and formoterol, to patients with poorly reversible COPD. It is not clear from the reports whether the twelve selected trials had sufficient power to detect significant differences between treatment and control groups in the various subjective and objective outcome measures. Since data were not pooled for meta-analysis, we were not able to conduct a sensitivity analysis based on the quality of trial reporting. Accordingly, we cannot comment on the possible influence of quality on the effect size of the outcome measures. Clinical heterogeneity among the trials limited assessment of the overall effect of the interventions. Since we did not perform a meta-analysis, statistical heterogeneity was not an issue. We selected reports that met our inclusion criteria, regardless of publication status, language and trial quality using a systematic research methodology; this approach has been shown to minimize potential selection and publication bias and lead to more reliable conclusions[ 36 ] We made every effort to conduct our review and report its results with the highest rigour. A potential limitation of our research is that we did not seek trials comparing long acting β2-agonists marketed outside of Canada (e.g., bambuterol) or those trials comparing long acting β2-agonists to agents other than ipratropium and placebo (e.g., short-acting β2-agonists, methylxanthines). Similarly, we excluded those trials in which the FEV1 response to a bronchodilator was not reported or greater than 15%. Thus, our results may not be generalisable to the greater population of patients who can be currently defined as having COPD[ 18 ]. We plan to include a greater number of comparators and a broader population in an upcoming analysis[ 37 ] Our results are similar to those of an earlier review [ 12 ] that identified three placebo controlled trials included in our review, but there are two important differences. In the earlier review FEV1 endpoint data from the placebo and salmeterol groups in two crossover trials [ 15 , 16 ] were pooled; the weighted mean differences were not significant. We preferred to analyse net improvement in FEV1 (the difference from baseline to endpoint), as we felt that it more accurately reflected the impact of maintenance therapy. In addition, no trials comparing long acting β2 agonists and ipratropium were available at the time of the earlier review. Another review has recently been published.[ 38 ] However, these authors restricted their search to MEDLINE and failed to identify a clinical trial comparing formoterol with ipratropium.[ 26 ] As a consequence, the evidence describing the use of formoterol versus ipratropium is limited to a single trial.[ 39 ] In contrast, we opted to exclude this trial after identifying two trials because roughly 40% of patients exhibited partial reversibility of FEV1 (15% to 80%) to an inhaled dose of 200 mcg salbutamol at baseline. We are in agreement with the authors' summary of the evidence surrounding salmeterol versus ipratropium. We believe our findings are in accord with current guidelines, such as the GOLD guidelines, that suggest bronchodilators should be prescribed according to individual patient responses. However, policymakers with limited health service resources need to be aware of an identifiable sub-population of patients with poorly reversible COPD for which long acting β2-agonists may result in reduced efficiency (cost-effectiveness). Our research also suggests clinical investigators of COPD trials should stratify trial participants into groups for which outcomes may consistently differ. Of the trials identified, four[ 23 , 30 , 31 , 40 ] used this approach. Outcome information from patients with poor reversibility was also analyzed in an abstract[ 41 ] of an excluded trial but not in the published report.[ 39 ] We were unable to ascertain sufficient details surrounding this analysis to add it to our findings. Conclusions In terms of clinical outcomes and safety, we could not find convincing evidence that salmeterol and formoterol have demonstrated advantages to ipratropium, a less expensive drug, for patients with stable COPD and poor reversibility. Compared to placebo, we found evidence of reduced rescue inhaler usage and improved spirometric outcomes without a significant impact on quality of life or exercise capacity. Competing interests Donald Husereau, Vijay Shukla, Michel Boucher, and Shaila Mensinkai have no competing interests to declare. CCOHTA is an independent, nonprofit health research agency funded by the federal, provincial, and territorial governments of Canada. Robert Dales sits on the advisory committees for GlaxoSmithKline (makers of the long acting β2 agonist, salmeterol) and Boeheringer Ingelheim (makers of the anticholinergic agent, ipratropium bromide). Authors' contributions DH edited and prepared the final manuscript for publication. VS led development of the research protocol, supervised the literature review, and summarized results. DH and VS were responsible for reviewing articles, judging their relevance, assessing their quality, and extracting data. MB assisted in developing the research protocol and in conflict resolution during study selection. RD assisted in developing the research protocol and provided clinical expertise. SM designed and conducted the electronic searches and provided expertise in the area of information science. All authors either wrote sections or critically reviewed drafts of this article. Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 Appendix 1. A list of reports considered in this review but excluded. Click here for file Additional File 2 Appendix 2. Search strategies including databases, time horizons and subject headings/keywords used to locate trials. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517721.xml
545073	MAB21L2, a vertebrate member of the Male-abnormal 21 family, modulates BMP signaling and interacts with SMAD1	Background Through in vivo loss-of-function studies, vertebrate members of the Male abnormal 21 (mab-21) gene family have been implicated in gastrulation, neural tube formation and eye morphogenesis. Despite mounting evidence of their considerable importance in development, the biochemical properties and nature of MAB-21 proteins have remained strikingly elusive. In addition, genetic studies conducted in C. elegans have established that in double mutants mab-21 is epistatic to genes encoding various members of a Transforming Growth Factor beta (TGF-beta) signaling pathway involved in the formation of male-specific sensory organs. Results Through a gain-of-function approach, we analyze the interaction of Mab21l2 with a TGF-beta signaling pathway in early vertebrate development. We show that the vertebrate mab-21 homolog Mab21l2 antagonizes the effects of Bone Morphogenetic Protein 4 ( BMP4 ) overexpression in vivo , rescuing the dorsal axis and restoring wild-type distribution of Chordin and Xvent2 transcripts in Xenopus gastrulae. We show that MAB21L2 immunoprecipitates in vivo with the BMP4 effector SMAD1, whilst in vitro it binds SMAD1 and the SMAD1-SMAD4 complex. Finally, when targeted to an heterologous promoter, MAB21L2 acts as a transcriptional repressor. Conclusions Our results provide the first biochemical and cellular foundation for future functional studies of mab-21 genes in normal neural development and its pathological disturbances.	Background The male-abnormal 21 ( mab-21 ) gene was first characterized in Caenorhabditis elegans as part of the combinatorial genetic code affecting morphogenesis of the sensory rays, a male-specific sense organ located in the tail and involved in copulation. In that context, early studies [ 1 ] identified a hypomorphic mab-21 allele as a dominant enhancer of egl-5 , a loss-of-function mutation affecting the nematode homolog of the Drosophila homeotic gene AbdominalB . Subsequent studies [ 2 ] demonstrated that mab-21 (+) plays a key role in the formation of the nematode male tail. Homozygous mab-21 mutations act cell-autonomously on the identity of a sensory ray, leading to alterations in anteroposterior identities akin to a homeotic transformation. Furthermore, they act non-cell-autonomously on the binary switch between hypodermal and neuroblast specification. Finally, hypomorphic mab-21 mutants exhibit pleiotropic changes affecting movement, body shape and fecundity, suggesting that mab-21 (+) also plays key developmental roles outside the male tail region. The mab-21 gene encodes a 41 kD basic protein with no significant homologies to any other published proteins. Homologs of the mab-21 gene have been identified in many other species, from Drosophila to humans. The first vertebrate counterparts of mab-21 were isolated in amniote genomes: a human homolog, MAB21L1 , was cloned during a systematic search for transcripts carrying trinucleotide repeats [ 3 ] and potentially involved in the pathogenesis of neuropsychiatric disorders. Interestingly, recent physical mapping data have located Mab21l1 within a chromosomal deletion discovered in a patient with autism and language deficit secondary to an auditory processing abnormality [ 4 ]. Our group, in parallel with others, identified two murine mab-21 genes ( Mab21l1 , Mab21l2 ) [ 5 - 8 ] and the human MAB21L2 gene, and showed that they encode nuclear proteins [ 6 ] that are 94% identical, 98% homologous to each other. In the mouse, Mab21l1 and Mab21l2 are expressed in largely overlapping territories [ 5 - 7 ], suggesting the possibility of a high degree of functional redundancy. More recently, only one vertebrate member of the mab-21 family was isolated in Xenopus laevis ( Xmab21l2 ) [ 9 ], (and our unpublished data: GenBank entry AF040992) and Danio rerio [ 10 , 11 ] ( Zmab21l2 ). Mab21l2 is an early marker of the tectum as well as primitive eye field, optic cup and retina in anamniotes, in agreement with observations previously made in mouse embryos. To address the in vivo roles of mab-21 genes in vertebrates, loss-of-function studies have been carried out by different groups in Xenopus and mouse embryos. Lau et al. [ 9 ] interfered with the functions of Xmab21l2 in embryogenesis by applying antisense DNA and double-stranded RNAi techniques. As a result, they reported a high frequency of embryos arrested at late gastrula/early neurula, as well as a significant incidence of neural tube closure defects in tadpoles. Likewise, Wong and Chow [ 12 ] cultured mouse embryos in the presence of antisense oligos specific for Mab21l1 and Mab21l2 , and reported that both treatments caused a sharp increase in the incidence of defective axial turning, incomplete notochord formation, and neural tube closure defects. However, Mab21l2 -specific antisense oligos were more potent and more effective than Mab21l1 -specific ones, suggesting that Mab21l2 may play a more irreplaceable role in early development than Mab21l1 . As said, mab-21 genes mark early stages of eye development in various species. Likely due to redundancy in the vertebrate mab-21 gene family a knock-out mouse carrying a null mutation of Mab21l1 featured an isolated, cell-autonomous defect in lens placode development, providing a novel and valuable model for the study of various eye defects in humans [ 13 ], but no overt retinal abnormalities. Conversely, Mab21l2 has been found to participate crucially in zebrafish retina formation acting as a downstream effector of Rx2 [ 14 ]. Some of these important advances point to a critical role for the mab-21 gene family as a whole in early embryogenesis on one hand, and in various aspects of neural development on the other. However, our knowledge on the molecular mechanisms through which mab-21 genes operate remains inadequate at best. Interestingly, by genetic analysis of loss-of-function mutations in C. elegans , Morita et al . have proposed mab-21 as a downstream target of a TGF-beta superfamily signaling pathway involved in sensory ray identity, the Small/Male-tail-abnormal pathway. This pathway is initiated by the secreted ligand CET-1 [ 15 ], also known as DBL-1 [ 16 ], and regulates body growth and male tail development. The CET-1/DBL-1 ligand transmits its signal through two receptor serine threonine kinases, DAF-4 and SMA-6, which in turn regulate the activity of the nuclear transducers SMA-2, SMA-3, and SMA-4. Hypomorphic mutations of mab-21 are epistatic to cet-1 pathway mutations affecting ray pattern formation ( cet-1 , sma-2 , sma-3 , sma-4 ), while they do not affect body size in double mutants [ 15 ]. Reportedly, this interaction did not reflect regulation of mab-21 gene transcription by the sma signaling pathway. The above genetic studies prove the biological relevance of the mab-21 – TGF-beta interaction, but stop short of addressing several key questions. What is the molecular counterpart of the genetic interaction described in C. elegans between mab-21 and the sma pathway? Is this interaction conserved in vertebrate development, and what TGF-beta superfamily molecules does it involve? Does it play a relevant role in the establishment of vertebrate dorsoventral (DV) polarity? If so, at what level do MAB-21 and TGF-beta pathways converge into a single regulatory cascade? In the present study, through a gain-of-function strategy, we analyze the molecular interplay between MAB21L2, a vertebrate homolog of C. elegans MAB-21, and TGF-beta superfamily signaling molecules. In particular, we focus on the signaling pathway initiated by BMP4, a ligand structurally related to CET-1/DBL-1, that regulates DV axis formation and numerous key aspects of neural development in chordata . Results Xmab21l2 expression in development It is well established that the BMP4 pathway is essential for proper formation of the dorsoventral axis; overexpression of BMP4 in Xenopus embryos can lead to the formation ventralized embryos, i.e. embryos devoid of axial structures, whereas BMP4 depletion can lead to dorsalized embryos, i.e. embryos exhibiting an expansion of dorsal mesoderm [ 17 - 20 ]. At tailbud stages, the expression of Xmab21l2 is restricted to the primitive eye field, optic recess, optic cup and retina (excluding the choroid fissure), to the tectum and dorsal neural tube, and to branchial arches (Figure 1A–E ). The early-onset and sustained expression of the MAB21L2 protein in oocyte-, blastula, gastrula, tailbud and tadpole lysates suggests that Mab21l2 is expressed both maternally and zygotically, during and after gastrulation (Figure 1F ). Figure 1 Expression of Mab21l2 gene and protein in Xenopus laevis development. A) Double whole mount in situ hybridization for Xmab21l2 (blue) and Xkrox20 (brown). At the end of neurulation (dorsal view of a st. 19 embryo is shown) Xmab21l2 expression (arrows) is restricted to the eye primordium (bottom arrowhead) and midbrain (top arrowhead). Xkrox20 labels the third and fifth rhombomeres (arrows). B) Frontal view of a stage 24 embryo. Xmab21l2 expression (blue, arrows) is restricted to the eye (left, solid arrow) and midbrain (right, solid arrow), clearly posterior to the forebrain marker Emx2 expression domain (brown, empty arrow). C–E: Sections obtained from late tadpole embryos were wholemount-hybridized with Xmab21l2 . Three transverse sections, from anterior to posterior, show Xmab21l2 expression in the retina and lens (Ey), in the branchial arches (BA), midbrain (arrow in C), hindbrain (arrow in D), and in the dorsal neural tube, (SC in E). F: Western blot analysis of MAB21l2 protein expression in Xenopus oocytes (stage 0) and Xenopus embryos (stages 3–23). ME: day 12 mouse embryo (positive control). Protein concentrations were quantitated through a Bradford assay and loaded on gel in equal amounts. The experiment shows maternal expression of MAB21l2 (41 kD, arrow), the levels of which decrease around blastula stage to resume zygotically thereafter. In particular, stage 11 gastrulae are positive for the MAB21l2 protein. Mab21l2 coexpression rescues the dorsal axis in BMP4-injected Xenopus embryos To determine whether Xmab21l2 and BMP4 have a synergistic or antagonistic interaction in vivo , we took a gain of function approach utilizing Xenopus laevis embryos. Preliminary RT-PCR experiments conducted on gastrula lysates injected at the one-cell stage with BMP4 mRNA had shown that, as previously reported by others in studies conducted in the nematode, Xenopus Mab21l2 expression is not controlled by BMP4 signaling (not shown). Prior to addressing the nature of the interaction between Xmab21l2 and BMP signaling molecules, we analyzed the effects of isolated Xmab21l2 overexpression. Embryos were injected at the one-cell stage with 800 pg in vitro -transcribed Xmab21l2 alone. We analyzed the phenotype of injected tailbud embryos as well as the expression of dorsal markers at gastrulation. A significant percentage of injected embryos (48%, N = 200) featured various degrees of dorsalization/anteriorization (DAI 6–8), with a well formed, sometimes enlarged head, a broad cement gland, a short and/or bent axis (Figure 2B ). We prepared embryos for histological examination by sectioning them along the longitudinal plane, both sagittally and horizontally. Histological analysis of horizontal sections revealed many tadpoles featuring an enlarged notochord (nc, fig. 2D ), suggestive of a possible defect in convergent extension. Finally, wholemount in situ hybridization experiments conducted on gastrulae with markers of DV polarity revealed an increased percentage of embryos (50%, N = 22) with expanded and/or enhanced Chordin gene expression (blue signal, fig. 2F ). Perturbation of Xvent2 gene expression was sporadic, with 5 out of 21 Xmab21l2 -injected embryos featuring a reduced Xvent2 expression domain (not shown). Our results are in sufficient agreement with those previously obtained by others [ 9 ], confirming that Xmab21l2 overexpression produces neither signs of complete dorsalization, such as Janus twins or radial eyes [ 21 ], nor axis duplication. Figure 2 Isolated overexpression of Xmab21l2 produces partially dorsalized embryos. A, C, E: control-injected embryos; B, D, F: Xmab21l2 -injected embryos. A–D: tailbud-tadpole stages; E, F: gastrula stage. In B: external appearance of Xmab21l2 -injected embryos. D: horizontal section of an Xmab21l2 -injected section stained with orange-G/aniline blue. Note enlarged notocord (nc). F: expanded and enhanced Chordin expression in Xmab21l2 -injected gastrulae (blue signal). To determine whether Xmab21l2 can compensate the effects of BMP4 overexpression, Xenopus embryos were injected at the one-cell stage with in vitro -transcribed BMP4 mRNA alone (0.6 ng, n = 111), or coinjected with Xmab21l2 mRNA (0.8 ng, n = 83; 1.6 ng, n = 101). We scored the formation of dorsal structures in tadpole stage embryos by using the dorsoanterior index [ 21 ]. A complete loss of dorsal structures (DAI = 0) (top in Figure 3A ) was observed in 11.4% of the embryos after BMP4 overexpression (0.6 ng) (Figure 3B ). Coinjection with 0.8 or 1.6 ng Xmab21l2 drastically reduced this class (0% and 3.1%, respectively). Likewise, whilst complete dorsal structures (DAI = 5) (Figure 3A , bottom) were observed in only 34.1% of BMP4 -injected embryos, coinjection with 0.8 or 1.6 ng Xmab21l2 increased this class significantly (65.9% and 59.4%, respectively), suggesting that Xmab21l2 antagonizes BMP4 in its ability to ventralize embryonic mesoderm in vivo (see Figure 3B for details). Figure 3 Xmab21l2 rescues dorsal structures in BMP4 injected embryos. (A) As previously described, injection of 0.6 ng of BMP4 mRNA gave rise to ventralized embryos. The most severely ventralized phenotype of BMP4 injection, called bauchstück , was scored with a dorso-anterior index (DAI) of 0 (top), whereas normally dorsalized embryos were assigned a DAI of 5 (bottom). (B) Embryos were injected with 0.6 ng BMP4 (n = 111), with 0.6 ng BMP4 plus 0.8 ng Mab21l2 (n = 83), or with 0.6 ng BMP4 plus 1.6 ng Mab21l2 (n = 101). In each group, we scored the percentage of complete ventralization (DAI = 0, grey bars) and normal dorsal axis formation (DAI = 5, black bars) in embryos that completed development. Intermediate classes are omitted in the plot. Coinjection of embryos with Mab21l2 significantly increased the percentage of correctly dosalized embryos (see text for details). Statistical analysis was conducted using the Chi square algorithm (1 df ). : p = 0.0005; : p = 0.0087. Next, by wholemount in situ hybridization, we analyzed the expression of molecular markers of dorsal ( chordin , Figure 4A–C ) and ventral ( Xvent2 , Figure 4D–F ) mesoderm in wildtype embryos, as well as embryos injected with BMP4 alone or coinjected with Xmab21l2 . Whilst the expression domain of chd was sharply reduced in 88% of BMP4 -injected embryos, 47% of the embryos co-injected with Xmab21l2 showed wild type chd expression (Figure 4B,C,G ). Likewise, 64% of the BMP4 -injected embryos showed a dorsal expansion of Xvent2 expression, that was rescued in 83% of the embryos co-injected with Xmab21l2 (Figure 4E,F,H ). Figure 4 Xmab21l2 restores the normal expression of Xvent2 and Chordin in BMP4 -injected embryos. Chordin (A–C) and Xvent2 (D–E) expression were analyzed by whole mount in situ hybridization. Embryos are shown in a vegetal view, dorsal to the top, ventral to the bottom. (A, D) Wild type expression of Chordin (A) and Xvent2 (D) in stage 10.5 gastrulae; (B, E) embryos injected with 1.2 ng of BMP4 mRNA alone showed reduced Chordin expression (B) (22/25) and expanded Xvent2 expression (E) (16/25); (C, F) embryos co-injected with 1.2 ng of BMP4 and 1.6 ng Xmab21l2 mRNA showed a rescue of wild type Chordin (C) (17/36) and Xvent2 expression (F) (24/29). (G) Percentage of embryos exhibiting wild type (black bars) or reduced (grey bars) Chordin gene expression in embryos injected with 1.2 ng BMP4 alone, and in embryos coinjected with 1.6 ng Mab21l2 . Coinjection of Mab21l2 significantly increased the number of embryos exhibiting a wild type Chordin expression pattern; : p = 0.004. (H) Percentage of embryos exhibiting wild type (black bars) or expanded (grey bars) Xvent2 gene expression in embryos injected with 1.2 ng BMP4 and in embryos coinjected with 1.6 ng Mab21l2 . Again, Mab21l2 significantly increased the number of embryos exhibiting a wild type Xvent2 expression pattern; *: p = 0.00044. Statistical analysis was conducted using the Chi square algorithm (1 df ). MAB21L2 interacts with SMAD1 in vitro and in vivo The antagonistic interaction between Mab21l2 and BMP4 may be either interpreted as evidence of a molecular interaction between MAB21L2 and some BMP signaling transducer, or alternatively suggest that MAB21L2 and some component of the BMP signaling pathway compete for a common cofactor. It is well established that the DNA-binding protein SMAD1, once phosphorylated at the plasma membrane by the BMP receptor heterotetramer, assembles with the receptor-independent protein SMAD4, enters the nucleus, establishes interactions with various cofactors, and acts by regulating the transcription of a host of target genes [reviewed in [ 22 ]]. Because vertebrate MAB-21 proteins are mainly or exclusively distributed in the nucleus [ 6 ], we investigated the possibility of a molecular interaction between MAB21L2 and SMAD proteins. In order to obtain in vitro evidence of this interaction, P19 cells were either treated with BMP4 or left untreated, and subsequently lysed. P19 cell extracts were then incubated with a resin-bound His-ZZ-tagged form of MAB21L2, or with a control His-ZZ-coupled resin. Eluates were analyzed by immunoblotting with a polyclonal anti-SMAD1 antibody. The experiment revealed a clear interaction between synthetic His-ZZ-MAB21L2 and endogenous SMAD1 that was absent when P19 lysates were incubated with the control resin (Figure 5A ). Although BMP4 treatment seemed to enhance it considerably, the interaction was not entirely BMP4-dependent, suggesting that conformational changes secondary to receptor-mediated phosphorylation may not be strictly required for SMAD1 to interact with MAB21L2. Figure 5 Mab21l2 interacts with Smad1 in vitro and in vivo. (A) Lanes 1, 2: immunoblotting of total lysates from P19 cells, mock- and Smad1 -transfected, respectively. P19 cells do not express detectable levels of endogenous SMAD1. Lanes 3–8: Affinity chromatography (pull-down) experiment. High-stringency eluates from untransfected P19 were separated by SDS-PAGE and transfered onto nitrocellulose filters. The blots were immunolabeled using an anti-SMAD1 antibody. Lanes 3–6: the in vitro interaction is not strictly dependent on activation of BMP signaling: the pull-down experiment was performed with 20 μl (3, 4) and 40 μl of P19 cell lysates (5, 6). Lysates came from P19 cells treated (4, 6) or untreated (3, 5) with BMP4. Lanes 7, 8: an His-ZZ-MAB21L2 protein (HisMab) synthesized in E. coli (lane 8) was coupled to a sepharose-Ig resin and incubated with cell lysates of P19 cells treated with BMP4. As a negative control, an in-vitro synthesized His-ZZ incubated with the same P19 lysates fails to pull down a 53 kD band. Arrows: SMAD1 (53 kD). (B) direct interaction between SMAD1 and MAB21L2; no direct interaction between SMAD4 and MAB21L2. An His-ZZ-MAB21L2 protein (+Mab) synthesized in E. coli (lane 2) was coupled to a sepharose-Ig resin and incubated with in vitro -translated SMAD1 (lanes 2), SMAD4 (lanes 4), and SMAD1 + SMAD4 (lanes 6). Negative controls were represented by His-ZZ-coupled resins (-Mab) incubated with the same in vitro -synthesized proteins (lanes 3, 5, 7). Lanes 1 and 8 contain in vitro -synthesized SMAD1 and SMAD4, respectively. (C) BMP4-dependent in vivo co-immunoprecipitation of flag-SMAD1 and myc-MAB21L2 in P19 cells. Cells were mock-transfected, transfected with flag-SMAD1 and/or with myc-MAB21L2, as indicated, and either treated with BMP4 or left untreated. In lanes 3–7, cell lysates were immunoprecipitated with a monoclonal anti-flag antibody. Cell lysates in lanes 1, 2 and immunoprecipitates in lanes 3–7 were gel fractionated and transfered to nitrocellulose filters. Blots were immunolabeled with an anti myc antibody. Arrow (42 kD) points to a band in lanes 2, 3) corresponding to myc-MAB21L2. (D) in vivo co-immunoprecipitation of flag-SMAD1 and myc-MAB21L2 in stage 11 Xenopus embryos, facilitated by BMP4 overexpression. Embryos were water-injected, injected with flag-Smad1 and/or with myc-Mab21l2 RNA, as indicated, and coinjected with BMP4 where indicated. In lanes 3–7, cell lysates were immunoprecipitated with a monoclonal anti-flag antibody. Embryo lysates in lanes 1, 2 and immunoprecipitates in lanes 3–6 were gel fractionated and transfered to nitrocellulase filters. Blots were immunolabeled with an anti myc antibody. Arrow (42 kD) points to a band in lanes 2–4) corresponding to myc-MAB21L2. These results prompted more questions, such as whether the interaction between MAB21L2 and SMAD1 is direct and whether it is dependent on prior formation of a SMAD1-SMAD4 complex. Once again, we resorted to affinity chromatography, incubating in vitro -translated, 35 S-Met-labeled SMAD1 and SMAD4 with a resin containing His-ZZ-MAB21L2. This experiment suggested that the interaction with SMAD1 is direct, and that MAB21L2 does not assemble with SMAD4 directly but only through SMAD1, indicating that the formation of a SMAD4-MAB21L2 complex is mediated by SMAD1 (Figure 5B ). Indirectly, this experiment also showed that MAB21L2 does not obviously compete with SMAD4 to assemble with SMAD1. Finally, we investigated whether the in vitro interaction observed between MAB21L2 and SMAD1 could be replicated in vivo . P19 cells co-transfected with myc-MAB21L2 and flag-SMAD1 were either treated with BMP4 or left untreated. In parallel, Xenopus laevis embryos were injected at the four-cell stage into the animal pole of each blastomere with myc-Mab21l2 and flag- Smad1 mRNAs, and coinjected with either BMP4 mRNA or H 2 O. The embryos were grown until mid-gastrula stage (st. 10.5/11). Both cell- and embryo lysates were subjected to immunoprecipitation, using an anti flag monoclonal antibody, whereas an anti-myc antibody was utilized for immunodetection in Western blotting. Experiments conducted in transfected P19 cells showed that SMAD1 coprecipitates with MAB21L2, and that the interaction is enhanced by activation of BMP signaling, which is required for nuclear localization of receptor-dependent SMADs (Figure 5C ). Likewise, MAB21L2-SMAD1 co-precipitation took place especially, albeit not exclusively, in BMP4-coinjected embryos (Figure 5D ). Interaction of the two proteins in the absence of exogenous BMP4 can likely be explained by the abundance of endogenous BMP2/4 ligands in stage 11 gastrulae. When targeted to an heterologous promoter, MAB21L2 acts as a strong transcriptional repressor Because in vivo MAB21L2 acts as a BMP4 antagonist, and because BMP4 signaling is transduced in the nucleus by SMAD transactivators, we asked whether MAB21L2 has any intrinsic repressor functions. To address this point, we generated fusion transcripts adding the GAL4 DNA binding domain (residues 1–147) [ 23 ], either to the N- or C-terminus of MAB21L2. For as yet unexplained reasons, the latter fusion protein proved extremely unstable and was discarded. Conversely, the N-terminal fusion protein is stable and was shown to bind DNA by gel-retardation assay (Figure 6A ). Thus, we were able to assess its activity in cotransfection assays, using as a reporter system the GAL4 target promoter (5XUAS) fused to luciferase. Luciferase activity in Mab21l2 -cotransfected COS7 cells was compared to the baseline reporter activity scored in cells transfected with the GAL4 DBD alone (G-D) or with a transcriptionally inactive GAL4 DBD fusion protein, (GAL4HOXB3NT, G-N) [ 24 ]. Reporter activity was clearly downregulated in Gal4-Mab21l2 -transfected cells only, suggesting that MAB21L2 may function as a transcriptional repressor or co-repressor (Figure 6B ). The same luciferase assay was performed in P19 cells, co-transfected with plasmid DNA, G-D, and G-M, revealing a 12-fold downregulation of 5XUAS-luc in G-M-transfected cells (not shown). This confirmed the evidence of a strong repressor activity of MAB21l2 when targeted to DNA. Figure 6 When targeted to a heterologous promoter, Mab21l2 behaves as a strong transcriptional repressor. (A) An in vitro -translated chimeric protein containing the GAL4 DNA binding domain fused to the N-terminus of MAB21L2 was used in an electrophoretic mobility shift assay. An end-labeled ds-DNA containing the GAL4 nucleotide binding site (5XUAS) was incubated in native conditions with in vitro translated GAL4-MAB21L2 (G-M) and separated electrophoretically by nondenaturing PAGE. As negative controls, we used free oligonucleotides (F) and rabbit reticulocyte lysates (RRL). As a positive control we used a previously tested GAL4 fusion protein, GAL4-D9NT (G-D9) [40]. The arrow indicates a bandshift specific for GAL4-MAB21L2 and absent in negative controls. u: unbound. (B) 5XUAS-luc transcription is clearly downregulated in cells transfected with GAL4-MAB21L2 (G-M) vs. untransfected COS7 cells or same cells transfected with GAL4 DBD, G-D) alone (: p = 0.0019). No downregulation of 5XUAS-luc activity was observed in COS7 cells transfected with a fusion of GAL4 and the N-terminal domain of the HOXB3 protein, which possesses no transcriptional activation or repression activity [24] (ns: not significant). Statistical analysis was conducted using the T test method (two tails). Discussion Mab-21 genes have been isolated in several species and have been shown, through genetic and epigenetic loss-of-function experiments, to play key roles in various developmental processes, ranging from gastrulation and neural tube closure to eye and lens formation [ 2 , 9 , 10 , 12 , 13 ]. However, no biochemical data have been available to identify the regulatory cascade(s) in which these elusive factors are involved. While epistatic analysis has proposed that C. elegans mab-21 undergoes inhibitory regulation by TGF-beta superfamily signals [ 15 ], the molecular nature of this regulation remained obscure. Likewise, no information has been published as to the conservation and relevance of the sma/mab-21 interaction in vertebrate development. Our results, obtained in an anamniotic model system, the Xenopus embryo, and in murine embryonic carcinoma cells, indicate that MAB21L2 interacts functionally with SMAD1, a nuclear transducer of BMP2/4/7 signaling. Overexpression of Mab21l2 complements the effects of BMP4 overexpression, in keeping with the epistatic interactions observed in nematodes [ 15 ]. Thus, genetic interactions first described in pseudocoelomates appear to be strongly conserved in evolution. As the nuclear localization of MAB21L2 suggested, our pull-down and immunoprecipitation results indicate that MAB21L2 is a new interactor of the receptor-activated transducer SMAD1. Whilst the interaction between MAB21L2 and SMAD1 appears to be direct, our results do not seem to favor a direct contact between MAB21L2 and SMAD4. The interaction of MAB21L2, a transcriptional repressor, with nuclear transducers of BMP signaling offers a possible explanation for the results of loss-of-function experiments conducted by other investigators in Xenopus laevis and mouse embryos. The increasing levels of MAB21L2 protein observed starting at midblastula transition may be required to compensate mesoderm-ventralizing signals triggered by BMP2/4. In the absence of MAB21L2 genes, completion of gastrulation may be hampered, resulting in various abnormalities of dorsoventral polarity [ 9 , 12 ]. Results obtained by others have demonstrated that SMAD complexes can be turned into negative regulators of transcription through a physical interaction with transcriptional repressors [ 25 ]. Our results indicate that MAB21L2 possesses a considerable transcriptional repressor activity. However, the antagonistic interactions between MAB21L2 and BMP signaling may not necessarily depend on the formation of transcriptional regulation complexes. Further experiments will be required to characterize the mechanism(s) of transcriptional regulation mediated by MAB21L2. Importantly, the identification of Mab21l2 downstream genes will make it possible to investigate mab-21 – BMP4 interactions from the opposite standpoint, i.e. by looking at the modulatory effect of BMP signaling on the expression of Mab21l2 targets. Although our results are more directly relevant to understanding the functions of Mab21l2 in early embryogenesis, the gene is clearly expressed at high levels and in a tightly localized fashion in neurulation and morphogenesis as well, likely contributing to neural tube and eye formation. Indeed, because of redundancy in the mab-21 gene family, many of the early effects of mab-21 gene mutation may be masked, uncovering these genes' roles in later stages of development and postnatal life. In late developmental processes, well beyond the end of gastrulation, MAB-21 proteins may prolong their dynamic interaction with BMP signaling transducers. As mentioned, a cell-autonomous requirement for Mab21l1 in lens formation has been recently documented in knock-out mice [ 13 ] that develop otherwise normally in the nervous system. In the same mice, genetic redundancy has hampered the analysis of mab-21 gene function in retinal development. However, antisense studies have shown that inhibition of mab-21 leads to disruption of the retinal anlage, where the gene is strongly expressed in a territory overlapping with the Bmp4 and Xvent2 expression domain. Recent studies have shown that BMP signaling plays a relevant role in guiding morphogenesis along the dorsoventral axis of the chick [ 26 ] and Xenopus [ 27 ] retina. In this context, it would be interesting to determine if mab-21 genes act by balancing BMP signals in the establishment of retinal polarity. Mab21l1 and Mab21l2 are prominently expressed in the embryonic midbrain, both dorsally and ventrally, and in prosomere 1 [ 5 , 6 ]. Roles exerted by BMP signaling molecules in these territories are starting to emerge, and appear relevant to the induction and patterning of dorsal mesencephalic structures (roof plate) [ 28 ], and, more ventrally, to the differentiation of dopaminergic precursors [ 29 ]. Throughout the hindbrain and spinal cord, Mab21l2 is expressed mostly in dorsal territories. In cerebellar development, it is expressed in the cerebellar plate and, later on, in granule cell neurons [ 5 ]. BMP signaling molecules are coexpressed with Mab21l2 at several of those sites, and their developmental roles have been at least partially addressed: as an example, granule cell specification can be induced by BMP7 overexpression in the dorsal and ventral rhombencephalon [ 30 ]. BMP7 is homologous to BMP2 and BMP4, and its signal can be transduced by SMAD1 [ 31 ]. Likewise, in the spinal cord Mab21l2 is expressed in a dorsal territory, where BMP2/4 signaling plays a key role in the specification of neuronal identities alongside the dorsoventral axis of the neural tube [ 32 - 34 ]. Finally, in postmigratory neural crest, the interplay between BMP2/4 and Mash1 maintains competence for neuronal differentiation [ 35 ]. In neural crest derivatives, including the branchial arches, Mab21l2 may thus participate in specifying neuronal and non-neuronal cell fates by interacting with BMP signals. Conclusions Our results provide a biochemical and molecular foundation for future studies of mab-21 gene function in vertebrate systems, demonstrating that MAB21L2 interacts functionally with the BMP4 signalling pathway and physically with its best characterized nuclear transducer, SMAD1. Furthermore, we show that MAB21l2 can act as a powerful transcriptional repressor when targeted to an heterologous promoter. Further work is clearly required to determine if MAB21L2 binds DNA, what its binding specificity is, or if it only exerts its effect in a DNA-binding-independent fashion. More broadly, additional studies are required to clarify the role of mab-21 genes in the context of BMP signaling, and their likely function as a molecular switch linking different regulatory pathways in development and disease. Methods Image acquisition and processing Images were obtained either by traditional photography (slides 1–3) or by electronic scanning of autoradiography films. All images were processed through the Adobe Photoshop 7.0 software or through the Adobe Illustrator 10 software. Isolation of Xmab21 cDNA and constructs The Xenopus laevis homolog of mouse Mab21L2 was isolated by screening a Xenopus stage 28–30 lambdaZapII library (Courtesy of Richard Harland) using the murine homolog as a probe. The 2.8 Kb clone isolated contained the full length Xmab21l2 cDNA (GenBank AF040992) as well as 5' and 3' regions. The coding region of Xmab21 was excised using NotI and BalI (1.3 Kb) and subcloned into pBluescript for whole mount in situ hybridization, or BamHI-StuI (1.1 Kb) and subcloned into pT7TS for overexpression in embryos and into pCDNA3 for cell transfection. Embryos All animal experimentation was conducted according to the stipulations of the Institutional Animal Care and Use Committee, San Raffaele Scientific Institute. In order to obtain embryos, Xenopus females were primed 1000 U of human chorionic gonadotropin (Profasi HP 5000, Serono) the night before collection. Ovulated eggs were fertilized with testis homogenates and allowed to develop in 0.1X MMR (1X MMR is 0.1 M NaCl, 2.0 mM KCl, 1.0 mM MgCl2, 2.0 mM CaCl2, 5 mM HEPES, pH7.4). Jelly coats were removed in 3.2 mM DTT, 0.2 M Tris pH 8.8. Embryos were staged according to Nieuwkoop and Faber [ 36 ] and fixed in MEMFA [ 37 ] for in situ hybridization. Histology For histological examination, wt or injected and stained embryos were fixed in MEMFA, embedded in wax, cut into 10 μm sections, dried onto slides, dewaxed with xylene and dehydrated with alcohol. Sections were rehydrated and stained with an orange G solution (2 g orange G, 8 ml glacial acetic acid, 100 ml water) and subsequently with an orange G-aniline blue solution (2 g orange G, 0.5 g aniline blue, 8 ml glacial acetic acid, 100 ml water), and dehydrated [ 36 , 38 ]. Whole-mount in situ hybridization Whole-mount in situ hybridization was performed on staged embryos as described by [ 37 ]. The antisense or control sense-strand were generated from the following linearized plasmids: pBS-Xmab21 (antisense linearized NotI, trancribed from T7; sense linearized HindIII, transcribed from T3), Engrailed 2, Krox20, Emx2 (Maria Pannese), CS2+-Chordin (a gift of Stefano Piccolo), Xvent1, Xvent2 (a gift of Christof Niehrs). In single whole-mount hibridizations the probe was labeled with digoxigenin. Digoxigenin-labeled probes were immunolabeled with alkaline phosphatase-conjugated anti-digoxigenin antibody and with the BM Purple substrate (Boehringer). In the double whole-mount one probe was digoxigenin-labeled the second was labeled with fluorescein-UTP. Each probe was used at a concentration of 1 g/ml. Sequential detection with alkaline phosphatase-conjugated anti-fluorescein antibodies (1:8000) and anti-digoxigenin (1:2000) was done. The first color reaction was revealed with NBT/BCIP, while for the second color reaction the Vector Black kit II (Vector Laboratories) was used. Selection of peptide and production of Mab-21 antisera in rabbits A 14aa C-terminal stretch from the mMab21l2 protein was chosen as immunogenic peptide to obtain antisera in rabbits. Peptide was synthesized, lyophilized and coupled to Keyhole limped hemocyanin (KLH). Coupled peptide was used for immunization of two rabbits. Immune and preimmune serum were controlled by ELISA using the same peptide coated at different serial dilutions. Immune serum was then purified by affinity chromatography: briefly, 20 mg peptide was coupled on CNBr-Sepharose-4B resin (Pharmacia), according to the standard procedure suggested. 50 ml antiserum diluted in 1:1 in 1X PBS was applied on the peptide/CNBr-Sepharose-4B over night at 4°C. After washing the resin with 50 ml PBS and 50 ml 3 M NaCl, bound antibodies were eluted using 100 mM Glycine pH 1.8. Amount, specificity and purity of the antiserum was tested by the Bradford assay, ELISA and SDS-PAGE/Coomassie staining. Purified antibody was tested for specificity in western blot on total protein extracts from cos-7 cells transfected with mMab21l2-myc fusion protein or mock vector, and on total brain extract from mouse embryos. One single band of 41 KDa was detected by anti Mab21l2 antibody. The same 41 kD band was detected by WB in lysates from transfected COS-7 cells, and was missing in untransfected cells. Cross-reactivity of anti mouse Mab21l2 antibody with the Xenopus Mab21l2 protein wastested by Western blot on total extracts from X. laevis embryos. The antibody did not work successfully in IHC or IP experiments, suggesting that it fails to recognize the native epitope. Western blotting Protein concentrations in extracts were quantitated through Bradford assays. Extracts were gel-fractionated by denaturing polyacrylamide gel electropohoresis SDS-PAGE. Gels were transfered onto nitrocellulose filters. Even loading was confirmed by Ponceau staining of nitrocellulose filters. Imunostaining was conducted as described [ 39 ]. Embryo injections The vector Xmab21-T7TS was linearized with Eco RI and transcribed with the T7 RNA polimerase, N-MycMab21L2-CS2+ was linearized with Asp718 and transcribed with the SP6 RNA polymerase. The Flag-Smad1 was excised from the human construct Flag-Smad1-CMV5 (Courtesy of J. Massagué) and subcloned into the vector pCS2+. The Flag-Smad1-pCS2+ vector was linearized with Asp718 and transcribed with SP6. BMP4 expression constructs were a kind gift of N. Ueno. Capped mRNA was synthesized using the Ambion Message Machine kit according to manufacturer's instructions. Injections were performed in 4% Ficoll in 1X MMR. Embryos were injected animally at the one-cell or two-cell stage into one or both blastomeres. The amount of mRNA injected is given in the text. Luciferase activity assay COS7 cells were maintained in DMEM Medium (Gibco Brl) supplemented with 10% FBS (Euro Clone), 2 mM L-glutamine (Euro Clone), 100 IU/ml penicillin, and 100 É g/ml streptomycin. Calcium phosphate transfection was performed in a 60 mm Petri dish with different amounts of the expression constructs, 5 É g of reporter plasmids, and 100 ng of pRL-TK (Promega) as an internal control. 18 h after transfection the cells were washed twice with PBS1X and the medium was replaced with DMEM Medium supplemented with 10% FBS, 2 mM L-glutamine (Euro Clone), 100 IU/ml penicillin, and 100 É g/ml streptomycin. Cells were harvested 36 h after the transfection, lysed, and assayed for luciferase activity (Dual Luciferase Reporter Assay System Kit, Promega). All luciferase assays were performed in triplicate. Immunoprecipitations P19 cells were maintained in MEM alpha medium (Gibco Brl) supplemented with 10% FBS (Euro Clone), 2 mM L-glutamine (Euro Clone), 100 IU/ml penicillin, and 100 μg/ml streptomycin. Calcium phosphate transfection was performed in a 100 mm Petri dish with 10 μg of mycMab21l2 expression construct and 10 μg of flagSmad1 expression construct. 24 h after transfection the medium was replaced with MEM Alpha Medum supplemented with 1% FBS 2 mM L-glutamine (Euro Clone), 100 IU/ and 40 ng/ml BMP4 (R&D Systems). Cells were harvested 36 h after the transfection with TEN Solution (40 mM Tris pH7.5; 1 mM EDTA; 150 mM NaCl) and centrifugated for 15 minutes at 5000 RPM. Cell pellets were frozen at -80°C o/n, thawed at RT and resuspended in 5 volumes of Extraction Buffer (10 mM Hepes pH 7.9; 400 mM NaCl; 0,1 mM EGTA, 5% glicerol). The samples were centrifugated at 34000 RPM for 30 minutes at 4°C and the supernatants were harvested. The concentration of the proteins was determined by BCA assay (Pierce). Xenopus laevis lysates were obtained in the same way starting from injected embryos harvested by centrifugation. After preclearing with 50 μl of protein G-sepharose o/n at 4°C, lysates were incubated with 25 μg of M2 anti-Flag monoclonal antibody (Sigma) and 50 μl of protein G-Sepharose o/n at 4°C. The resin was washed twice with Dilution Buffer (10 mM Tris-Cl pH8; 140 mM NaCl; 0.025% NaN 3 ; 0.1% triton), once with TSA solution (10 mM Tris-Cl pH8; 140 mM NaCl; 0.025% NaN 3 ) and once with 50 mM Tris-Cl pH 6.8. The protein complexes were eluted by addition of Sample buffer (Tris-Cl 125 mM pH 6.8; 0.1 M 2-mercaptoethanol; 2% SDS; 20% glycerol; 25 mg/ml Bromphenol Blue) followed by boiling for 5 minutes and separated on an SDS-polyacrylamide gel followed by anti-myc (1 μg/ml of clone n°9E10) immunoblotting. The secondary antibody was a goat anti-mouse antibody HRP-conjugated (1:30000 Bio-Rad). The blots were developed with the Supersignal West Pico reagent (Pierce). Purification of His-ZZ fusion proteins The protocol was based on manufacturer's recommendations for Ni-NTA Agarose (Qiagen). His-ZZ fusion protein expression in bacterial cultures (37°C, BL21 E. coli ) was induced by β-D-thiogalactopyranoside (IPTG) (0.4 mM for His-ZZ-Mab21; 1 mM for His-ZZ and for His-ZZ-Smad1) when the optical density (OD600) reached 0.7–0.9. Cells were harvested after 3 hours, resuspended in lysis buffer (50 mM NaH 2 PO 4 ; 300 mM NaCl; lysozyme 2 mg/ml), and sonicated. The purification protocol was performed with Ni-NTA Agarose (Quiagen) beads and the elution with 250 mM Imidazole (Sigma). His-ZZ-Mab21l2 fusion protein pull-down assays 120 μg of His-ZZ-Mab21l2 fusion protein was incubated 1 h at 4°C with 25 μl of IgG-Sepharose beads (Amersham). The beads were recovered by centrifugation and washed ten times with 0.4 M KCl. The beads were incubated 1 h with 250 μg of precleared lysates of P19 treated and untreated with 40 ng/ml of BMP4. The beads were washed 5 times with 0.2 M KCl; the protein complexes were eluted by addition of Sample buffer followed by boiling for 5 minutes, and separated on an SDS-polyacrylamide gel for anti-Smad1 (1:2000 Santa Cruz) immunoblotting. The secondary antibody was a goat anti-rabbit HRP-conjugated antibody (Bio-Rad). The blots were developed with the Supersignal West Pico reagent (Pierce). His-ZZ-Smad1 fusion protein pull-down assays 120 μg of His-ZZ-Smad1 fusion protein was incubated 1 h at 4°C with 25 μl of IgG-Sepharose resin (Amersham). The beads were recovered by centrifugation and washed ten times with 0.4 M KCl. The beads were incubated 1 h with 250 μg of precleared lysates of COS7 cells transfected with mycMab21l2. The beads were washed 5 times with 0.2 M KCl; the protein complexes were eluted by addition of Sample buffer,boiled for 5 minutes and separeted on a SDS-polyacrylamide gel for anti-myc immunoblotting. His-ZZ-Mab21l21 fusion protein pull-down assays with in-vitro translated proteins 120 μg of His-ZZ-Mab21l2 fusion protein was incubated 1 h at 4°C with 25 μl of IgG-Sepharose beads (Amersham) in Binding buffer (50 mM Tris pH8; 50 mM KCl; 5 mM MgCl 2 ; 1 mM DTT; 0.2% NP-40, 10% glycerol). The beads were recovered by centrifugation and washed with Binding buffer. In-vitro translated, 35 S-methionine-labeled Smad1 and Smad4 were prepared with the TNT coupled transcription/translation system (Promega). 45 μl of the TNT reaction were mixed with the His-ZZ-Mab21l2 bound to the IgG-Sepharose beads. The mixture was incubated for 2 hours at 4°C in Binding buffer. The Sepharose-protein complex was washed four times with Wash buffer (50 mM Tris pH7.5; 150 mM NaCl; 1 mM EDTA; 0.2% NP-40), eluted by addition of Sample buffer followed by boiling for 5 minutes and separated on SDS-polyacrylamide gels. Gels were fixed and stained with 50% methanol, 10% acetic acid and 0.1% Comassie Blue for 20 minutes. After the incubation with Sodium salicylate 1 M (Fluka) for 20 minutes to improve the radioactive signal the gels were dried and exposed to Kodak X-Omat film o/n at -80°C. Authors' contributions DB did in situ hybridizations and in vivo overexpression experiments. AB did pull-downs, immunoprecipitations and luciferase assays. RL designed peptides and produced an anti MAB-21 polyclonal antibody. VZ produced constructs for, and carried out DNA-binding assays and luciferase assays. In addition he helped design and evaluate all cellular assays. GGC designed and coordinated the experimental work. All authors read and approved the final manuscript	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545073.xml
509286	How can Health Behavior Theory be made more useful for intervention research?	Background The present paper expresses the author's views about the practical utility of Health Behavior Theory for health behavior intervention research. The views are skeptical and perhaps even a bit exaggerated. They are, however, also based on 20-plus years of in-the-trenches research focused on improving health behavior practice through research. Discussion The author's research has been theoretically driven and has involved measurement of varying variables considered to be important theoretical mediators and moderators of health behavior. Regretfully, much of this work has found these variables wanting in basic scientific merit. Health Behavior Theory as we have known it over the last 25 years or so has been dominated by conceptualizations of behavior change processes that highlight cognitive decision-making. Although much of health behavior practice targets what people do rather than what they think, the logic of focusing on thoughts is that what people think about is the key to what they will do in the future, and that interventions that can measure and harness those processes will succeed to a greater extent than those that do not. Unfortunately, in the author's experience, the premise of cognitive theories has fallen short empirically in a number of ways. The cognitive schemata favored by most health behavior theories are difficult to measure, they do not predict behavioral outcomes very well, there is little evidence that they cause behavior, and they are hard to change directly. Summary It is suggested that health behavior researchers reconsider their use of these theories in favor of models whose variables are more accessible to observation and experimental manipulation and that most importantly have strong empirical support.	Background The author has been conducting research on behavioral treatment of obesity for about 25 years. During that time, the dominant conceptual models guiding intervention development have been cognitive behavior models that have their origin in psychological theory. Those most often cited include the Health Belief Model [ 1 ], Protection Motivation Theory [ 2 ], Subjective Expected Utility Theory [ 3 ], the Theory of Reasoned Action [ 4 ], Social Cognitive Theory [ 5 ], and the Transtheoretical Model [ 6 ]. All of these theories are concerned with how people make behavioral choices and the general idea is that people decide what to do based on the extent to which they expect that their choices will produce results that they value. Much of the content of the theories is concerned with factors that may affect value/expectancy calculations. As summarized by Weinstein in a comparative review of four social psychological theories [ 7 ], variables thought to influence value/expectancy judgments include such factors as perceived rewards of current behavior, self-efficacy, normative beliefs, motivation, and the perceived consequences of not changing behavior. Weinstein's summary is illustrative of the fact that Health Behavior Theory has tended to be particularly interested in understanding people's motivation to change behavior rather than ability to change. Moreover, motivation is thought to be the result of a relatively complex, but logical, interpretation of large quantities of information about self and environment. The theories that Weinstein reviewed deal almost exclusively with behavioral decision processes in people's minds. They have few if any terms relating to how information gets into peoples minds or how subsets of it receive more or less attention. Broader health behavior theories such as Social Cognitive Theory or the Transtheoretical model have addressed issues and variables outside the person to a greater extent, but the fundamental interest in and belief in psychological variables as the key force in determining health behavior remains. The implications of the focus of health behavior theory on psychological determinants of behavioral decision-making for my own research area of interest, obesity treatment, are several. One is the inclusion of measures of psychological characteristics in most research protocols (e.g., assessment of behavioral intentions, self-efficacy, perception of barriers to change, perception of social support, and outcome expectations). A second is the inclusion of treatment elements that specifically target psychological perceptions and processes independent of the diet and physical activity behaviors that actually produce weight change (e.g., how to deal with emotional eating, how to deal with the frustration of lapses and relapses, and how to talk to yourself to increase self-motivation). A third is the belief that psychological reactions to treatment experiences themselves are very important and deserve independent attention. Common behavioral prescriptions for weight-loss goals and frequency of self-weighing are exemplary (i.e., recommending infrequent weighing to prevent discouraging feedback about progress and encouraging smaller and thus "more attainable" behavior and weight-loss goals in the belief that they will be more motivating). The problem with the emphasis on cognitive variables in weight-control research is that they have so far failed to meet fundamental scientific criteria for empirical verification. Thus, they also have not led to a better understanding of the weight-loss process, have not improved our ability to predict weight-loss outcomes, and have not led to improvement in treatment methods. In some cases it is even arguable that they have made treatment worse. I will illustrate these problems with results from my own research. Discussion Like most behavioral researchers in the obesity area, I have attempted to measure elements of health behavior theory in every obesity intervention project I have ever conducted. I have assessed weight-loss goals, behavioral and weight-loss self-efficacy, psychological well-being, perceived barriers to diet and physical activity change, stages-of-change, and perceived social support. How well have empirical examinations of these factors fared as predictors of success in weight control? Self-efficacy We have examined the predictive value of self-efficacy assessments in several of our studies and describe the results from three of these here in more detail [ 8 - 10 ]. In the first study, self-efficacy was assessed at baseline, posttreatment, and one year later in 85 men participating in a 15-week weight-loss program [ 8 ]. The self-efficacy instrument had subscales for emotional states (e.g., anxiety) and situations (e.g., eating away from home). Higher baseline self-efficacy on both subscales was associated with greater weight loss in treatment and at 1- and 2-year follow-up. Emotional self-efficacy at posttreatment did not predict weight loss at 1- or 2-year follow-up. Situational self-efficacy at posttreatment predicted weight loss at 1-year but not 2-year follow-up. The second study examined mood and situational self-efficacy in 55 men and 58 women before and after a 16-week weight-loss treatment with a 1-year follow-up [ 9 ]. Women had lower pretreatment self-efficacy than men. Self-efficacy was predictive of weight loss and maintenance in men but not in women. Change in self-efficacy over time was positively related to weight change in women but not in men. The third study examined predictors of weight change over a 2-year period in 460 men and 1172 women who received a low-intensity weight-loss intervention delivered through their HMO [ 10 ]. The self-efficacy measure was the WEL questionnaire. Men again were found to have higher baseline self-efficacy than women. Self-efficacy did not predict weight change in men but was positively, though weakly, related to weight change at 6 months only in women. Our overall conclusion from the analyses described above, as well as others not pursued in as great detail, is that self-efficacy is a weak predictor of weight loss and is inconsistent across study populations and gender. It tends to increase with weight loss. However, treatment-induced increases in efficacy are not predictive of longer-term weight-loss success. Barriers to Adherence We have also attempted to measure barriers to adherence to weight-control behaviors in many of our studies [ 11 - 14 ]. The instruments used for this have typically been formatted similarly to efficacy questionnaires in that people are asked to indicate how difficult they find situational, knowledge, and motivational challenges to achieving diet and exercise changes. The findings in these studies have been quite consistent. Baseline assessments of perceived barriers to behavior change are not predictive of weight change. Weight loss is associated with reported decreases in perceived barriers. Treatment-induced change in perceived barriers are not predictive of future weight change. In other words, barrier perceptions as we have measured them do not appear to have pragmatic significance. Weight Goals Goal-setting has long been of interest to health behavior theory and in recent years has attracted attention in weight-loss research when it was realized that most people who enter weight-loss treatments want to lose a lot more weight than is realistic given the potency of current weight-loss methodologies [ 15 ]. When asked to describe weight losses they deem to represent "dream, happy, acceptable, and disappointing," many individuals in treatment fail to reach even "disappointing" weight losses even though in objective medical terms the results are positive. Based on the argument that failure to reach gratifying weight-loss goals leads to psychological distress that lowers weight self-efficacy and undermines weight-loss efforts, it has become popular to recommend counseling in weight-loss treatments specifically targeting the lowering of weight-loss goals. The theoretical argument is that excessive outcome expectations undermine behavioral efforts. We have now completed three sets of formal analyses examining whether weight goals are predictive of weight-loss success. In one of these analyses the relationship between weight-loss goals, weight-loss goal attainment, and long-term (30 months) weight-loss attainment and psychological well-being were assessed in 69 men and 61 women participating in an intensive behavioral treatment program [ 16 ]. Results indicated that weight-loss goals were unrealistically high on average and that lower goals were more likely to be reached. Nevertheless, weight-loss goals did not predict either short- or long-term weight losses and were not associated with elevated psychological distress. Two more recent analyses we have conducted looking at weight-loss goals as predictors of success have produced similar results [Linde JA, Jeffery RW, Levy RL, Pronk NP and Boyle RG, unpublished data [ 17 ]]. Weight-loss goals either did not predict weight loss at all or were slightly positively related to weight-loss success. Perceived Social Support Perceived social support is another psychological factor thought to influence health behavior decision-making. We have measured social support in a variety of ways in our studies, ranging from single-item questions to multipaged assessments attempting to differentiate among informational, instrumental, and emotional support. The results, unfortunately, have closely paralleled those we have seen with other assessments of barriers to adherence. Assessments of social support prior to treatment do not predict weight loss. Average reports of social support tend to parallel weight loss itself. When people lose weight they report more social support. When they regain, they report less. In other words, perceptions of social support are not predictive of success in weight-loss treatments. Frequency Weight Self-monitoring Self-monitoring of health behavior is incorporated into many health behavior theories, usually as part of a person's assessment of achieved outcomes. Although self-monitoring is usually considered a positive element in the adoption of health behavior, in obesity treatment frequent self-monitoring of weight has tended to be down-played or even discouraged on the grounds that disappointing results (i.e., less than desired weight change) may undermine motivation. This is another example in which health behavior theory may have indirectly led to incorrect treatment recommendations. In weight-loss treatments, active discouragement of frequent self-observation of weight has become popular based on the premise that more frequent weighting will cause psychological stress and lower self-efficacy. Recently, we have examined the relationship between frequency of self-weighing and body weight in both clinical and population samples and have found, somewhat to our surprise, that frequency of self-weighing is one of the strongest single predictors of body weight cross-sectionally, and change in the frequency of self-weighing is one of the strongest predictors of weight change [Linde JA, Jeffery RW and French SA, unpublished data]. The direction of predictions, however, is opposite that derived from theory. People who weigh themselves more weigh less and are more successful in losing weight. Stage-of-Change A final failure of current health behavior theory to prove useful in weight-control research is a recent examination of the relationship between a stage-of-change measure adopted from Prochaska and short- and long-term weight loss [ 18 ]. Categories of precontemplation, contemplation, preparation, and action were defined based on questions about weight-loss intentions and recent weight-loss attempts. Despite a large sample size, excellent follow-up rates, and well-measured objective outcomes, we were unable to demonstrate that staging algorithms recommended by proponents of the Transtheoretical Model could predict weight-loss outcomes. Experimental Modification of Expectations Our most recent effort to utilize health behavior theory in obesity intervention research is a study that attempted to examine the effectiveness of experimentally-induced outcome expectancies on weight loss [Finch EA, Linde JA, Jeffery RW, Rothman AJ and King CM, unpublished data]. Obese men and women participated in an 8-week weight-loss program with 18-month follow-up in which they were assigned to one of two expectancy groups. The optimistic group was told that focusing exclusively on the positive benefits of weight loss would be valuable in ensuring that they remained motivated in their weight-loss efforts and was given assignments during weekly group sessions and homework between sessions to reinforce this optimistic mindset. A "balanced" expectancy group received the instructions that focusing on both the positive and negative aspects of weight loss, a balanced approach, would be most conducive to maintaining weight-loss motivation. This group also received assignments to reinforce their message. Results of this study indicated that the expectation induction was successful initially but difficult to maintain in the face of real weight-loss experience. We were also unable to show that experimentally-induced expectations influenced weight-loss success. Summary and Conclusion To summarize the findings described above, I have had considerable difficulty over the last 25 years in confirming that the psychosocial variables favored by health behavior theory are of much value for obesity intervention research. They do not predict weight loss well, either as mediators or moderators. There is little evidence to support the idea that targeting them for intervention improves weight-loss outcomes. It is, of course, arguable that the weak findings relating to health behavior theory variables are due in large part to methodological weaknesses, either in measurement tools and/or their frequency of measurement. I would argue, however, that 25 years is long enough to wait for improved methods and that it is time to look elsewhere for variables that better predict weight-change outcomes and that, therefore, may form a better basis for improving future treatments. Implication for Weight-Loss Treatment Given the lack of success finding support for cognitive mediators of behavior change in weight loss, one might surmise that progress in improving weight-loss interventions over the last 20 years must have been dreary indeed. Somewhat surprisingly, however, that is not the case. In fact, the short-term (6 to 12 months) success of weight-loss treatments has approximately doubled over that time and several variables have been identified that reliably enhance treatment outcomes. It has been clearly shown experimentally that increasing treatment length [ 19 ], prescribing low-energy intakes [ 20 ], prescribing high-energy expenditure [ 21 ], using a deposit contract and group-based reward systems [ 22 ], and simplifying adherence to diet through meal substitutes [ 23 ] and exercise by providing exercise equipment [ 24 ] all improve initial weight loss. From a theoretical perspective, however, one thing is noteworthy about these successful innovations. Although not incompatible with health behavior theory, none of them are specifically derived from cognitive decision-making models. Indeed, health behavior theory does not include variables like these in its models. Where Do We Go From Here? The argument above about the practical limitations of many popular theories of health behavior is not meant to be a call to abandon theory. Behavior scientists have amassed much useful information about the principles underlying human behavior that should be valuable for health behavior interventions. Much is known about human perception, learning, motivation, and responsiveness to environmental opportunities and contingencies. Health behavior intervention lies at the interface between people and their environment. Interventionists change aspects of the environment (cues, information, behavioral contingencies) with the intention of producing changes in how people behave. What is needed to advance health behavior intervention is theory that addresses relationships between modifiable aspects of the environment and behavior. There is no doubt that cognitive processes are involved in these relationships. However, the extent to which current theories capture this is questionable. Data now available suggest that easily obtainable information about people's cognitive processes adds little to our ability to predict the results of interventions. Thus, it may be wise to pay more attention to applied theories like classical behavior theory [ 25 ], communications theory [ 26 ], and learning theory [ 27 ] than to those coming out of the social cognitive traditions. Competing interests None declared.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC509286.xml
546208	The tumor suppressor Scrib interacts with the zyxin-related protein LPP, which shuttles between cell adhesion sites and the nucleus	Background At sites of cell adhesion, proteins exist that not only perform structural tasks but also have a signaling function. Previously, we found that the Lipoma Preferred Partner (LPP) protein is localized at sites of cell adhesion such as focal adhesions and cell-cell contacts, and shuttles to the nucleus where it has transcriptional activation capacity. LPP is a member of the zyxin family of proteins, which contains five members: ajuba, LIMD1, LPP, TRIP6 and zyxin. LPP has three LIM domains (zinc-finger protein interaction domains) at its carboxy-terminus, which are preceded by a proline-rich pre-LIM region containing a number of protein interaction domains. Results To catch the role of LPP at sites of cell adhesion, we made an effort to identify binding partners of LPP. We found the tumor suppressor protein Scrib, which is a component of cell-cell contacts, as interaction partner of LPP. Human Scrib, which is a functional homologue of Drosophila scribble , is a member of the leucine-rich repeat and PDZ (LAP) family of proteins that is involved in the regulation of cell adhesion, cell shape and polarity. In addition, Scrib displays tumor suppressor activity. The binding between Scrib and LPP is mediated by the PDZ domains of Scrib and the carboxy-terminus of LPP. Both proteins localize in cell-cell contacts. Whereas LPP is also localized in focal adhesions and in the nucleus, Scrib could not be detected at these locations in MDCKII and CV-1 cells. Furthermore, our investigations indicate that Scrib is dispensable for targeting LPP to focal adhesions and to cell-cell contacts, and that LPP is not necessary for localizing Scrib in cell-cell contacts. We show that all four PDZ domains of Scrib are dispensable for localizing this protein in cell-cell contacts. Conclusions Here, we identified an interaction between one of zyxin's family members, LPP, and the tumor suppressor protein Scrib. Both proteins localize in cell-cell contacts. This interaction links Scrib to a communication pathway between cell-cell contacts and the nucleus, and implicates LPP in Scrib-associated functions.	Background At the heart of structural and functional integrity of multicellular entities is the ability of each and every cell of it to successfully integrate signals arising from soluble factors, cell-substratum adhesion and cell-cell adhesion [ 1 ]. Correct processing of these signals allows appropriate cellular growth, differentiation, and tissue morphogenesis, but malfunctions often lie at the basis of pathologies such as tumor growth and metastasis. At sites of cell adhesion, more and more proteins are being identified that not only play a role in maintaining cell shape and motility but that, in addition to these structural functions, are also implicated in signaling events. Because of this dual function, these proteins have to interact, via multiple binding motifs, with components of both the actin cytoskeleton and signaling pathways that regulate e.g. gene expression. A protein that may play a role in these processes is the LPP (Lipoma Preferred Partner) protein [ 2 ]. LPP is a member of the zyxin family of LIM domain proteins, which consists of five members: zyxin [ 3 ], TRIP6 (Thyroid Receptor Interacting Protein 6) [ 4 ], LPP [ 2 ], ajuba [ 5 ] and LIMD1 (LIM Domain containing 1) [ 6 ]. LPP has three LIM domains (zinc-finger protein interaction domains) at its carboxy-terminus, which are preceded by a proline-rich pre-LIM region containing a number of protein interaction domains (Fig. 1A ). LPP has been shown to localize at sites of cell adhesion, such as focal contacts, which are membrane attachment sites to the extracellular matrix, and cell-cell contacts. However, apart from its localization in cell adhesion sites, this protein has also been shown to localize transiently in the nucleus. Because of its structural features and its characteristic to shuttle between the nucleus and the cytoplasm, LPP has been proposed to be a scaffolding protein involved in signal transduction from sites of cell adhesion to the nucleus. Figure 1 Schematic representation of human LPP and Scrib proteins. (A) Schematic representation of the human LPP protein. Human LPP contains a proline-rich pre-LIM region followed by three tandem LIM domains. In its pre-LIM region, LPP harbors one nuclear export signal, two VASP binding sites and one α-actinin binding site. Furthermore, LPP has two regions in common with its family member TRIP6 and one region with its family members zyxin, TRIP6 and LIMD1. At its carboxy-terminus, LPP has a binding site for Scrib. (B) Schematic representation of the human Scrib protein human Scrib is a 1630 amino acid protein that contains 16 leucine-rich repeats (LRR) followed by 2 domains that are specific for the LAP family of proteins (LAP-specific domain (LAPSD) a and b), and 4 PDZ domains. The corresponding position of the mouse Scrib prey clone that was isolated in a yeast-two hybrid screen using LPP as bait is indicated. The amino acid sequence of the PDZ domains of mouse Scrib was compared to that of human Scrib, and percentage identity is indicated for each PDZ domain. Originally, we identified the LPP gene, as being the preferred translocation partner of the HMGA2 gene in a subgroup of lipomas, which are benign tumors of adipose tissue [ 2 ]. In these tumors, HMGA2 / LPP fusion transcripts are expressed and identical fusion transcripts have also been found in a subgroup of pulmonary chondroid hamartomas [ 7 ], in a parosteal lipoma [ 8 ], and in a soft tissue chondroma [ 9 ]. In a case of acute monoblastic leukemia, the LPP gene acts as translocation partner of the MLL gene and the tumor expresses MLL / LPP fusion transcripts [ 10 ]. All tumor-specific fusion transcripts that are expressed in the above mentioned tumors encode similar LPP fusion proteins containing AT-hooks (DNA binding domains) of the HMGA2 or MLL proteins followed by LIM domains of LPP. These fusion proteins are mainly expressed in the nucleus [ 11 ]. At cell adhesions, the LPP protein interacts with α-actinin and VASP (vasodilator-stimulated phosphoprotein) via its pre-LIM region that contains an α-actinin binding site located near its N-terminus and two VASP-binding ("FP 4 ")-motifs (Fig. 1A ) [ 11 , 12 ]. In the nucleus, LPP has transcriptional activation capacity in reporter gene assays suggesting that it is involved in the regulation of gene expression [ 11 ]. The nucleocytoplasmic distribution of this protein involves a nuclear export signal (NES) that also resides in the pre-LIM region (Fig. 1A ) [ 11 ]. Recently, we have shown that the LIM domains of LPP cooperate to target the protein to focal adhesions, and that the linker between LIM domains 1 and 2 plays a pivotal role in this targeting [ 13 ]. When overexpressed in the cytoplasm of cells, these LIM domains deplete endogenous LPP and vinculin from focal adhesions suggesting a role for LPP in focal adhesion assembly [ 13 ]. Recently, LPP was found to be highly expressed in smooth muscle [ 14 , 15 ], and a role for LPP in regulating cell motility was proposed [ 14 ]. In an effort to learn more about the molecular function of LPP, we performed a yeast two-hybrid screening experiment to identify potential LPP-interacting proteins. Here, we report that LPP interacts with Scrib, a member of the LAP (leucine-rich repeat and PDZ domain) family of proteins [ 16 ]. Scrib is a functional homologue [ 17 ] of Drosophila Scribble, a tumor suppressor that plays a role in the regulation of cellular adhesion, cell shape and polarity [ 18 , 19 ]. In follow-up of the results of the yeast two-hybrid screening, we have performed various experiments to find out whether the observed interaction also occurs in mammalian cells and have substantiated this interaction in vitro and in vivo. Furthermore, we have studied whether or not the Scrib protein plays a role in the subcellular targeting of LPP. Results Screening for LPP-interacting proteins by yeast two-hybrid In a previous study [ 13 ], we showed that the LIM domains of LPP are the major units for targeting LPP to focal adhesions. LIM domains are cysteine- and histidine-rich domains that form two zinc fingers capable of mediating protein-protein interactions [ 20 , 21 ]. However, the protein(s) that is/are responsible for the targeting of LPP to focal adhesions, i.e. protein(s) that bind(s) to the LIM domains of LPP, are not yet known. To identify protein binding partners of the LIM domains of LPP, we performed a yeast two-hybrid screening experiment. We made use of a yeast two-hybrid system that is based on transcriptional activation of two reporter genes HIS3 and LacZ whose expression is driven by upstream GAL4 DNA-binding sites. Because all three LIM domains of LPP cooperate to target LPP to focal adhesions [ 13 ], we initially focused on a screening using a bait that contained all three LIM domains. Unlike in mammalian cells, where we have shown that the three LIM domains of LPP have transcriptional activation capacity [ 11 ], this bait, although well expressed, did not activate the reporter genes in yeast cells (results not shown). This is similar to what has been found for zyxin's LIM domains [ 22 ], but in contrast to what has been found for the three LIM domains of TRIP6 that do activate reporter genes in yeast [ 22 ]. However, the bait containing all three LIM domains of LPP appeared to be very sticky since thousands of yeast colonies were obtained in which both reporter genes were activated. In an effort to reduce background activity, we deleted the first LIM domain, or the first and the second LIM domain, in the bait, leaving the two most carboxy-terminal, or the most carboxy-terminal LIM domain(s) intact, respectively. These deletions completely abolished all background activity making these baits the baits of choice to perform a library screening. Here, we report about the screening that was performed with the bait containing only the most carboxy-terminal LIM domain of LPP. As described before [ 13 ], we showed that the third LIM domain of LPP only has a very weak targeting capacity for focal adhesions. This makes it very unlikely that, by using this bait, we would pick up a protein that targets LPP to these structures, which was the initial goal of our studies. Indeed, our screening did not reveal any focal adhesion binding partners of LPP, however, in stead, we found another very interesting LPP-interacting protein as will be outlined in the following sections. A mouse embryonal cDNA library was screened using a bait (pGBT9-LPP WT ) containing the third LIM domain and carboxy-terminus of human LPP (amino acids 531–612). Among ~1.0 × 10 6 yeast cotransformants (Leu + and Trp + ), 56 clones were His + of which 23 were LacZ + too. PCR analysis of these His + /LacZ + clones, using prey-specific insert-flanking primers, revealed that 21 of the 23 obtained clones, contained a prey-construct having a 2 kb cDNA insert (results not shown). Subsequent fragmentation of the obtained 2 kb PCR products, representing the cDNA inserts of the prey-constructs, using the HaeIII restriction enzyme (frequent cutter), indicated that all 21 isolated prey-constructs, having a 2 kb insert, were identical. The 2 kb cDNA insert of one representative prey-construct was completely sequenced and the sequence was submitted to the NCBI database (Genbank accession no. AF271735). A BLAST (Basic Local Alignment Search Tool)-search revealed that this mouse prey-construct encoded an amino- and carboxy-terminally truncated protein comprising four PDZ domains that was almost identical to the human Scrib protein (Fig. 1B ), indicating that the prey-construct represented mouse Scrib. The Scrib protein contains a set of 16 leucine-rich repeats (LRRs) near its amino-terminus and four PDZ (PSD-95, Discs large, ZO-1) domains distributed throughout the remainder of the protein (Fig. 1B ). The partial mouse Scrib protein, expressed by the prey-construct, corresponded to amino acids 709 – 1242 of human Scrib (Fig. 1B ). Further analysis indicated that the isolated prey-construct, which was named pACT2-mScrib, activated the HIS3 and LacZ reporter genes of the yeast only in the presence of pGBT9-LPP WT , identifying pACT2-mScrib as a true positive (Table 1 , upper three rows). Table 1 Interaction of LPP with Scrib in the yeast two-hybrid system BAIT PREY HIS LACZ pGBT9 pACT2-mScrib - - pGBT9-LPP WT pACT2 - - pGBT9-LPP WT pACT2-mScrib + + pGBT9-LPP S609A pACT2 - - pGBT9-LPP S609A pACT2-mScrib + + pGBT9-LPP T610A pACT2 - - pGBT9-LPP T610A pACT2-mScrib - - pGBT9-LPP D611A pACT2 - - pGBT9-LPP D611A pACT2-mScrib + + pGBT9-LPP L612A pACT2 - - pGBT9-LPP L612A pACT2-mScrib - - Yeast cells (CG-1945), cotransformed with a bait and a prey as indicated, were selected on medium containing 5 mM 3-AT, lacking Trp, Leu and His. Yeast colonies were tested for the expression of β-galactosidase. + indicates strong positive interaction; - indicates no interaction. LPP binds to the PDZ domains of Scrib via its C-terminal tail Since the pACT2-mScrib prey-construct contained four PDZ domains, and since PDZ domains are one of the most commonly found protein-protein interaction domains in organisms from bacteria to humans [ 23 ], it was most likely that Scrib would bind to LPP via its PDZ domains. The LPP-bait that was used to screen the library was pGBT9-LPP WT containing the third LIM domain and carboxy-terminus of human LPP. Although PDZ domains have been shown to bind LIM domains [ 24 ], binding to carboxy-terminal peptides appears to be the typical mode of interaction [ 25 ]. The common structure of PDZ domains comprises six β strands (βA-βF) and two α helices (αA and αB), which fold in an overall six-stranded β sandwich [ 25 ]. The binding specificity of PDZ domains is critically determined by the interaction of the first residue of helix α B (position αB1) and the side chain of the -2 residue of the C-terminal ligand. This forms the basis for PDZ classification [ 25 ]. Since all four PDZ domains of Scrib contain a histidine at position αB1, they are classified as class I PDZ domains. Therefore, based on what has been demonstrated for this subclass of PDZ domains [ 25 , 26 ], the carboxy-terminal sequence of Scrib target proteins is predicted to require a hydrophobic amino acid (h) at the 0 (carboxy-terminus) position, and a serine (S) or threonine (T) at the -2 position. Theoretically, the carboxy-terminus of the LPP protein, being -STDL, thus completely fulfils the criteria for binding to the PDZ domains of Scrib. To evaluate these predictions experimentally and to demonstrate that the binding of LPP to Scrib is specific, we performed yeast two-hybrid experiments using pGBT9-LPP WT as well as pGBT9-LPP S609A , pGBT9-LPP T610A , pGBT9-LPP D611A and pGBT9-LPP L612A as bait. The last four baits are identical to pGBT9-LPP WT except for a point mutation to alanine, respectively introduced at serine 609 (-3 position), threonine 610 (-2 position), aspartate 611 (-1 position) and leucine 612 (position 0). As prey, we used pACT2-mScrib. As summarized in Table 1 , this alanine-scan mutant analysis identified threonine 610 (-2 position) and leucine 612 (0 position) of LPP as being essential for binding to Scrib indicating a PDZ domain-mediated specific interaction between Scrib and the carboxy-terminus of LPP. Additional yeast two-hybrid analysis showed that LPP did not interact with Erbin, PICK1, PSD-95, Syntenin, CASK, or AF6 PDZ domains, as summarized in Table 2 . Table 2 Interaction of LPP with PDZ domains of proteins different from Scrib BAIT PREY HIS -HIS pGBT9-LPP WT Scrib-PDZ + + Erbin-PDZ + - PICK1-PDZ + - PSD95-PDZ + - Syntenin-PDZ + - CASK-PDZ + - AF6-PDZ + - Yeast cells, cotransformed with pGBT9-LPP WT and a pACT2-prey as indicated, were selected on medium lacking Leu and Trp, and either containing His or no His with 5 mM 3-AT. + indicates growth of yeast transformants; - indicates no growth of yeast transformants. LPP interacts with Scrib PDZ domains in mammalian cells We verified the Scrib-LPP interaction, which was identified in yeast cells, in mammalian two-hybrid experiments. Doing the assay in mammalian cells rather than in yeast cells, provides a more physiological environment: proteins are more likely to be in their physiological configuration, i.e. appropriately folded and modified posttranslationally, etc. Interaction between bait- and prey-proteins in a mammalian two-hybrid assay takes place in the nucleus. For an accurate performance of this assay, this means that bait- and prey-proteins should be localized in the nucleus. In contrast to the yeast assays, where we used partial bait-proteins, we wanted to use full length bait-proteins in the mammalian assay. However, since LPP contains a nuclear export signal (NES) (amino acids 117–128) in its pre-LIM region [ 11 ], we used bait-proteins in which this NES had been deleted. To verify whether deletion of the NES in LPP induced nuclear accumulation of the bait-proteins that were used in the mammalian two-hybrid assay, we introduced wild-type and mutated LPP-bait-proteins in 293T cells. While pM-LPP WT , containing GAL4-fused full length wild-type human LPP with an intact NES, was excluded from nuclei, pM-LPP dNESWT , containing GAL4-fused full length human LPP with a deletion of the NES, was accumulating in the nuclei of the cells (results not shown). These results indicate that deletion of the NES in the LPP bait proteins used in this study indeed induce nuclear accumulation of these proteins. To perform the mammalian two-hybrid experiments, we used as baits: pM-LPP dNESWT , containing full length human LPP with a deletion of its NES, and pM-LPP dNEST610A and pM-LPP dNESL612A , which are identical to pM-LPP dNESWT except for a point mutation to alanine introduced at threonine 610 (position -2) and leucine 612 (position 0), respectively. As determined in the yeast two-hybrid assay, each of the threonine 610 and leucine 612 residues is critical for the interaction with Scrib. As prey-protein, we used pSNATCH-hScrib PDZ containing a part of the human Scrib protein (amino acids 669–1233) encompassing all four PDZ-domains. As summarized in Fig. 2 , the interaction between wild-type full length LPP and Scrib PDZ domains resulted in high levels of luciferase reporter activity. These high levels dropped to background levels when pM-LPP dNEST610A or pM-LPP dNESL612A were used as baits in combination with Scrib as prey. The "background" levels of luciferase that were detected when pM-LPP-baits were used in combination with pSNATCH (empty prey-vector) as prey, are due to the intrinsic transcriptional activation activity of the LPP protein [ 11 ]. Figure 2 Scrib interacts with LPP in mammalian cells. pM-bait- and pSNATCH-prey-constructs were cotransfected into 293 cells in the combination indicated, together with a GAL4-regulated luciferase reporter and a CMV-β-galactosidase internal control. Cell lysates were assayed for luciferase activity 18–24 hours after transfection. Relative luciferase activity is reported as the average of three independent duplo experiments (with standard error). These results indicate that LPP binds to Scrib PDZ domains and that this binding is abolished when amino acids at position 0 or -2 are mutated. Development and characterization of Scrib antibodies To analyze expression and intracellular distribution of Scrib in cultured cells, we prepared a Scrib-specific antibody (Scrib-472), as described in the Methods section. The Scrib-472 antibody recognized a protein of an apparent molecular mass of more than 200 kDa in a number of different cell extracts (Fig. 3A ). Scrib was easily detected in the epithelial cell lines 293 and MDCKII, in the fibroblast cell line CV-1, and also in the T lymphocyte cell line Jurkat (Fig. 3A ). These results indicate that our antibody recognizes Scrib-proteins of different species, being human (Jurkat and 293), monkey (CV-1) and dog (MDCKII). The Scrib-472 antibody also reacted with an Xpress-hScrib fusion protein produced in 293T cells transfected with the corresponding DNA (Fig. 3B ). In Fig. 3B , lane 2, which depicts a Western analysis of untransfected 293T cell lysate with Scrib-antibodies, no band of endogenous Scrib is seen. Longer exposure, however, did show a band indicating that Scrib is expressed in these cells, however, 293T cells express much lower levels of endogenous Scrib as compared to 293 cells (our unpublished observations). The Scrib protein was migrating slower in SDS gels than would be expected from its theoretically calculated molecular mass (175 kDa). Possible explanations include anomalous migration per se, and posttranslational modifications. To investigate whether the Scrib-472 antibody not only recognizes denatured Scrib protein on Western blots but also is capable of detecting Scrib in fixed cells, MDCKII cells were grown to confluency on glass coverslips and stained with the Scrib-472 antibodies. From previous studies, it is known that Scrib is localized in cell-cell contacts [ 17 ]. As shown in Fig. 3C , the Scrib-472 antibody indeed is capable to detect native Scrib in cell-cell contacts in fixed cells. Figure 3 Characterization of anti-Scrib antibodies. (A) Total cell extracts were prepared from the following cell lines: human embryonic kidney epithelial cells (293) (lane 1), dog normal kidney epithelial cells (MDCK) (lane 2), human T lymphocytes (Jurkat) (lane 3), and African green monkey kidney fibroblast cells (CV-1) (lane 4). Approximately 30 μg of protein from each extract was analysed by SDS-PAGE and Western blotting with the Scrib-472 antibodies. The position of molecular markers are as shown. (B) Total cell extracts of 293T cells, either not transfected (lane 2), or transiently transfected with Xpress-hScrib that is composed of the full length human Scrib protein fused to an Xpress-epitope-tag at its amino-terminus (lanes 1 and 3) were analyzed by SDS-PAGE and Western blotting with an anti-Xpress antibody (lane 1) or with the Scrib-472 antibody (lanes 2 and 3). The position of molecular markers are as shown. (C) MDCKII cells, grown on glass coverslips, were fixed and stained with Scrib-472-antibodies. Immunofluorescence was visualized by epifluorescence microscopy. Scrib is not localized in focal adhesions in CV-1 and MDCKII cells, and is dispensable for targeting LPP to these structures We have shown before that LPP is localized in cell-cell contacts [ 11 ] and also for human Scrib, it was shown that it is localized in these structures [ 17 ] (also shown in Fig. 3C and 5 , upper right panel). Since LPP is not only localized in cell-cell contacts but also in focal adhesions [ 11 , 13 ], we investigated whether also Scrib had the ability to localize at these structures. For this purpose, we used two different cell lines: the epithelial cell line MDCKII and the fibroblast cell line CV-1. However, in contrast to LPP, Scrib could not be detected in focal adhesions as shown by staining CV-1 cells with Scrib-472 antibodies (Fig. 4 , upper left panel). Identical results were obtained in MDCKII cells (results not shown). Focal adhesions were indeed present, as these structures could be stained using vinculin antibodies used as a marker for focal adhesions (CV-1 cells: Fig. 4 , upper right panel; MDCKII cells: results not shown). If Scrib had been present in focal adhesions, we would have detected it there, because, as shown in Fig. 3A , Scrib is highly expressed in CV-1 as well as in MDCKII cells, and as shown in Fig. 3C , Scrib-472 antibodies are able to detect Scrib in its native conformation in fixed cells. Moreover, a hScrib-GFP protein expressed in CV-1 or MDCKII cells was never detected in focal adhesions (results not shown) but was localized in cell-cell contacts (MDCKII cells: Fig. 5 , lower left panel). The nature of the nuclear staining observed in CV1-cells stained with the Scrib-472 antibody (Fig. 4 , upper left panel) is aspecific, as it is also obtained with the corresponding pre-immuneserum. In addition, nuclear staining was never obtained when an hScrib-GFP protein was transiently overexpressed in these cells (results not shown). Nuclear staining was also not detected in MDCKII cells as shown in Fig. 3C and 5 , upper right panel. These results indicate that, in contrast to LPP, which is localized both in focal adhesions and in cell-cell contacts in CV-1 and MDCKII cells, Scrib is only localized in cell-cell contacts but not in focal adhesions in these cells. Figure 4 Scrib is not localized in focal adhesions in CV-1 cells, and is dispensable for targeting LPP to these structures. Upper panels: CV-1 cells, grown on glass coverslips, were double labelled with Scrib-472 antibodies (left panel) and anti-vinculin antibodies (right panel) used as a marker for focal adhesions. Lower panels: CV-1 cells were transiently transfected with wild-type human LPP (left panel), or LPP with a mutated carboxy-terminus (T610A) (right panel), as GFP-fusions. GFP-fluorescence was visualized by epifluorescence microscopy. Figure 5 Scrib and LPP are localized in cell-cell contacts but are dispensable for targeting each other to these structures. Upper panels: MDCKII cells, grown on glass coverslips, were double labelled with anti-LPP antibodies (left panel) and anti-Scrib antibodies (right panel). Lower panels: MDCKII stable cell lines, expressing GFP-fusion proteins containing wild-type human LPP (upper left panel), LPP with a mutated carboxy-terminus (T610A) (upper right panel), human wild-type Scrib (lower left panel), or Scrib with a deletion of all its PDZ domains (lower right panel), were grown on glass coverslips (Scrib) or on Transwell-Clear polyester membranes (LPP). GFP-fluorescence was visualized by epifluorescence microscopy (Scrib) or by confocal microscopy (LPP). As deduced from these results, we hypothesized that Scrib was not involved in targeting LPP to focal adhesions. Indeed, evidence for this hypothesis was obtained by transfecting CV-1 cells with a construct expressing GFP-LPP WT containing full length wild-type LPP, or GFP-LPP T610A , which is identical to GFP-LPP WT except for a point mutation to alanine introduced at threonine 610 , which abolishes binding to Scrib. No difference in focal adhesion localization could be detected between wild-type and mutated GFP-LPP fusion proteins (Fig. 4 , lower panels). Scrib and LPP are dispensable to target each other to cell-cell contacts Since Scrib and LPP both localize in cell-cell contacts [ 11 , 17 ] (Fig. 5 , upper panels), we investigated whether Scrib was responsible for targeting LPP to cell-cell contacts. For this, we made stable MDCKII cell lines expressing wild-type and mutated GFP-coupled forms of the LPP protein, of which the mutant form is not able to bind anymore to Scrib. However, as shown in Fig. 5 , lower panels, LPP proteins that could not bind to Scrib anymore were still able to localize in cell-cell contacts in a similar way as their wild-type counterparts. These results indicate that Scrib is not responsible for targeting LPP to cell-cell contacts. We next investigated whether LPP was responsible for targeting Scrib to cell-cell contacts. To look into this aspect, we made stable MDCKII cell lines expressing either wild-type full length Scrib-GFP or a mutated Scrib-GFP protein lacking all four PDZ domains (deletion of amino acids 725–1227). However, both the full length Scrib-GFP protein as well as the mutated form lacking all four PDZ domains localized equally well in cell-cell contacts (Fig. 5 , lower panels). These results indicate that the PDZ domains of Scrib are dispensable for targeting the protein to cell-cell contacts, and as a consequence LPP is not necessary to locate Scrib in cell-cell contacts. In summary, these results indicate that LPP and Scrib are dispensable to target each other to cell-cell contacts. There is a direct interaction between the carboxy-terminus of LPP and the PDZ domains of Scrib To further assess the binding between LPP and Scrib, we investigated whether there is a direct interaction between these two proteins. For this, we performed GST pull-down experiments. In vitro translated full length Scrib was tested for binding with glutathione beads, which were coupled with GST-LPP-LT WT , GST-LPP-LT L612A , or GST alone. GST-LPP-LT WT contains 40 amino acids of the pre-LIM region, the three LIM domains, and the wild-type carboxy-terminal tail of human LPP. GST-LPP-LT L612A is identical to GST-LPP-LT WT except for a point mutation to alanine introduced at leucine 612 (position 0). All GST-fusion proteins as well as GST alone were expressed well in E. coli (Fig. 6A ). As shown in Fig. 6B , Scrib interacted specifically with the wild-type LPP protein but not with its mutated form, GST-LPP-LT L612A or with GST alone. These results indicate that there is a specific and direct interaction between LPP and Scrib. Figure 6 Direct interaction between the carboxy-terminus of LPP and the PDZ domains of Scrib. (A) GST fused to either wild-type LPP (40 amino acids of the pre-LIM region, the three LIM domains and the tail), or a similar LPP molecule with a mutated carboxy-terminus (L612A) and GST alone were expressed in E. coli , purified and analyzed by SDS-PAGE and Coomassie Blue staining. All proteins were expressed well. Protein markers are as indicated. (B) In vitro synthesized [ 35 S]-methionine-labelled full length Scrib was incubated with immobilized GST or with either one of the above-described GST fusion proteins and allowed to interact over night at 4°C. After extensive washing, bound proteins were eluted in sample buffer, separated by SDS-PAGE and visualized by autoradiography. The amount of synthesized protein loaded as a reference on the gel corresponds to 10% of the input used in each binding experiment. (C) All four PDZ domains of Scrib (amino acids 616 – 1490), either wild-type or mutated as indicated, were synthesized in vitro and [ 35 S]-methionine-labelled. these labelled proteins were incubated with immobilized GST or with GST-LPP-LT WT and allowed to interact over night at 4°C. Bound proteins were eluted in sample buffer, separated by SDS-PAGE and visualized by autoradiography. The amount of synthesized protein loaded as a reference on the gel corresponds to 10% of the input used in each binding experiment. To further investigate the requirements in the Scrib protein for binding to LPP, we performed additional GST pull-down experiments. From our previously described experiments (yeast and mammalian two-hybrid), it was clear that the PDZ domains of Scrib bind to LPP. These findings were confirmed by using GST pull-down: as shown in Fig. 6C , upper panel, a portion of the Scrib protein encompassing all four PDZ domains was efficiently pulled down by GST-LPP-LT WT . To find out which of the four PDZ domains of Scrib was responsible for the observed interaction with LPP, we mutated the PDZ domains of Scrib, one at the time, by destroying their carboxylate binding loop (LG → AE), and tested how efficiently these mutated proteins were pulled down by GST-LPP-LT WT . From the results, which are presented in Fig. 6C , we can conclude that all four PDZ domains of Scrib more or less contribute to the binding to LPP, but that PDZ 3 is most important, since binding to GST-LPP-LT WT was almost completely abolished when the carboxylate binding loop of this PDZ domain was destroyed. Scrib can target LPP to an ectopic location in vivo through its PDZ domains Evidence for an in vivo interaction between Scrib and LPP was obtained by performing mitochondrial targeting experiments. We tested if Scrib was sufficient to recruit LPP to an ectopic location in living cells. The membrane anchor of the ActA sequence has been shown previously to be sufficient to target proteins expressed in mammalian cells to the surface of mitochondria [ 27 , 28 ]. This ectopic localization allows testing ligand recruitment in vivo. For this purpose, we generated a chimera named Xpress-hScrib-mito made up by an Xpress-epitope tag fused to the amino-terminus of human full length Scrib and linked in frame to the membrane anchor of the Listeria monocytogenes protein ActA (mito). Expression of this construct was confirmed by Western blotting with the use of an anti-Xpress antibody (results not shown). CV-1 cells were transiently transfected with Xpress-hScrib-mito and full length wild-type or carboxy-terminally mutated LPP green fluorescent protein fusions. Cells were stained with an anti-Xpress antibody and examined by fluorescence microscopy. In all transfected cells, the Xpress-hScrib-mito chimera localized to mitochondria, as shown in Fig. 7 , left upper and middle panels. As shown in Fig. 7 , upper right panels, wild-type LPP can be recruited to Xpress-hScrib-mito on mitochondria. This recruitment of LPP to Xpress-Scrib-mito-coated mitochondria was completely abolished when the carboxy-terminus of LPP was mutated (Fig. 7 , middle panels). Figure 7 Scrib can recruit LPP to an ectopic location in vivo through its PDZ domains. CV-1 cells were transiently co-transfected with Xpress-hScrib-mito or Xpress-hScribdPDZ-mito, and GFP-fusions of wild-type full length human LPP, or LPP with a mutated carboxy-terminus (T610A). Xpress-hScrib-mito and Xpress-hScribdPDZ-mito are composed of the human full length Scrib protein with or without its PDZ domains, respectively, which is fused to an Xpress-epitope-tag at its amino-terminus, and to an ActA-derived mitochondrial membrane anchor at its carboxy-terminus, Cells were stained with an anti-Xpress antibody to detect Xpress-Scrib(dPDZ)-mito. Immunofluorescence and GFP were visualized by epifluorescence microscopy. The focal adhesion localization of the GFP-LPP proteins is not visible in these pictures because a focal plane corresponding to mitochondrial staining is shown. To investigate the importance of the PDZ domains of Scrib in this recruitment of LPP, we deleted all four PDZ domains (amino acids 724–1192) from Xpress-hScrib-mito (=Xpress-hScribdPDZ-mito) and tested whether this PDZ-less protein still was able to recruit LPP to mitochondria. As shown in Fig. 7 , lower panels, Xpress-hScribdPDZ-mito lost its ability to recruit LPP to mitochondria. These results indicate that Scrib can recruit LPP to an ectopic location in vivo, and that the PDZ domains of Scrib are an absolute requirement for this activity. As mentioned above, the Xpress-hScrib-mito chimera localized to mitochondria in all cells that expressed this protein. However, LPP, which was co-expressed, was only recruited to mitochondria in a small fraction of these cells. This issue will be further addressed in the Discussion section. Discussion In the course of our studies of chromosomal aberrations in benign tumors, we have previously discovered the LPP gene as being rearranged in certain subtypes of these tumors [ 2 ], and identified the LPP protein as a member of the zyxin family of proteins [ 11 ]. In this study, we report that LPP specifically interacts with Scrib. We provide evidence that this interaction is mediated by the carboxy-terminus of LPP on the one hand, and the PDZ domains of Scrib on the other hand. Futhermore, we show that Scrib is not necessary for targeting LPP to focal adhesions, and that Scrib and LPP are dispensable to target each other to cell-cell contacts. Scrib is a member of the LAP (LRR (leucine-rich repeat) and PDZ (PSD-95/Discs-large/ZO-1)) family of membrane-associated proteins that play a role in the regulation of cell polarity [ 16 ]. LAP family members have been identified in mammals (Erbin, Densin-180, Lano, and Scrib) [ 29 - 32 ], in Caenorhabditis elegans (LET-413) [ 33 ], and in Drosophila melanogaster (Scribble) [ 18 ]. LAP proteins contain a set of leucine-rich repeats (LRRs) at their amino-terminus, and either four (Scrib and Scribble), one (Erbin, Densin-180 and LET-413) or no (Lano) copies of the PDZ domain. A specific characteristic of these proteins are the LAP-specific domains (LAPSa and b), which are located carboxy-terminally of the LRRs [ 34 ]. Most information regarding the function of Scrib comes from studies in Drosophila melanogaster . Drosophila Scribble was identified as being required for the apical confinement of polarity determinants in epithelia [ 18 ]. Mutations in Scribble cause aberrant cell shapes and loss of the monolayer organization in embryonic epithelia. Scribble is localized in septate junctions and loss of Scribble function results in the misdistribution of apical proteins and adherens junctions to the basolateral cell surface. Subsequent studies in Drosophila provided evidence that Scribble is a tumor suppressor and cooperates with two other tumor suppressors, Lethal giant larvae (Lgl) and Discs-large (Dlg) to regulate cell polarity and growth control [ 19 ]. Recently, these three tumor suppressors were shown to regulate cell size and mitotic spindle asymmetry in Drosophila neuroblasts [ 35 ]. The role of Scribble in tumorigenesis was further supported by the discovery that Scribble mutants cooperate with oncogenic Ras or Notch to cause neoplastic overgrowth of the eye disc [ 36 ], and that cooperation between oncogenic Ras and inactivation of Scribble leads to metastatic behavior [ 37 ]. Additional studies in Drosophila implicate Scribble in the regulation of synaptic plasticity and synaptic vesicle dynamics [ 38 , 39 ], and show that Scribble is essential for olfactory behavior in Drosophila [ 40 ]. As for mammalian Scrib, little information is available at the moment. Relating to the control of cell polarity and proliferation, human Scrib was found to be a functional homologue of the Drosophila scribble protein [ 17 ]. Polarity defects and tumorous overgrowth of Scribble-mutant flies are rescued by human Scrib predicting an important role for human Scrib in the suppression of mammalian tumorigenesis. Further support for this hypothesis, was obtained by the fact that human and mouse Scrib are targeted for degradation by high-risk papillomavirus E6 proteins [ 32 , 41 ]. Human papilloma viruses cause papillomas or warts on skin, genital tissues, and the upper respiratory tract, and high-grade lesions progress to carcinomas at a high frequency. The high-risk subgroup of human papilloma viruses detected in these lesions have been causally linked to the development of over 90% of uterine cervical carcinomas, the second leading cause of cancer-related deaths among women world-wide. High-risk papilloma virus E6 proteins direct Scrib for degradation by directly binding to the PDZ-domains of Scrib. In this regard, it is noteworthy that we observed a remarkable aspect regarding the expression levels of Scrib in a number of mammalian cell lines. As already mentioned before (Fig. 3 ), we noticed that 293T cells expressed much lower levels of Scrib as compared to 293 cells. 293T cells are derived from 293 but, in contrast, these cells stably express Simian Virus 40 largeT antigen. SV40 large T is a powerful oncoprotein capable of transforming a variety of cell types [ 42 ]. Its transforming activity is attributed to its binding and manipulation of the function of certain key tumor suppressors and cell regulatory proteins such as retinoblastoma and p53. However, certain factors that contribute to its full transformation potential are not yet completely understood. We hypothesize that large T induces the downregulation of Scrib expression, and that Scrib contributes to the transformation potential of SV40 large T. In addition to its role as a tumor suppressor, Scrib was also implicated in the regulation of planar cell polarity, a role that is not established for Drosophila Scribble [ 43 ], and it was shown that disruption of Scrib is the causal factor for the severe neural tube defects that occur in the circletail mouse [ 44 ]. Disruption of neural tube closure leads to a group of disorders termed neural tube defects, which are one of the commonest causes of congenital malformation and lethality in humans. The most severe form of neural tube defect is craniorachischisis, in which almost the entire brain and spinal cord remain open. Craniorachischisis comprises 10–20% of human neural tube defects, and is caused by a failure to initiate neural tube formation at the start of neurulation. Circletail is one of only two mouse mutants that exhibit craniorachischisis. The fact that Scrib was identified as the gene that was mutated in this mouse attributes an important role for Scrib in development [ 44 ]. We show here that Scrib is expressed equally well in very different cell types, such as Jurkat cells, which are human T lymphocytes, epithelial cells such as 293 and MDCKII cells, and in fibroblasts such as CV-1 cells. As described above, the function of Scrib and its Drosophila ancestor Scribble have been mainly addressed in epithelial cells. To our knowledge, nothing is known yet about the function of Scrib in other cell types such as lymphocytes and fibroblasts. We show here that LPP specifically binds to and partially co-localizes with Scrib in cell-cell contacts of epithelial and fibroblastic cell lines. Previous studies have shown that PDZ domain proteins play an important role in the targeting of proteins to specific membrane compartments and in the assembly of these proteins into supramolecular complexes [ 25 ]. Therefore, we investigated whether Scrib was essential to localize LPP in cell-cell contacts. However, as demonstrated by these experiments, Scrib is not necessary to target LPP to these structures. These findings are similar to what has been found for targeting of zyxin family members to focal adhesions. Recently, zyxin and TRIP6 were shown to interact with members of the p130 Cas family of signal transducers, which are focal adhesion components [ 22 ]. This interaction is primarily mediated by the LIM domains of zyxin and TRIP6. One specific function associated with the LIM domains of zyxin family members is targeting to focal adhesions. Despite this feature of the zyxin family LIM domains, and despite their interaction with p130Cas, it was shown that p130Cas is not required for focal adhesion localization of zyxin and TRIP6 [ 22 ]. We also investigated whether LPP was responsible for targeting of Scrib to cell-cell contacts. However, as demonstrated by our experiments, also this appeared not to be the case. In fact, our results indicate that all of the PDZ domains of Scrib are dispensable for targeting the protein to cell-cell contacts. For epithelial cells, these results are in agreement to what has been published in the course of our investigations by Legouis and Jaulin-Bastard et al . [ 45 ], who have shown that a point mutation of a specific proline residue that is located at position 305 in LRR number 13 of human Scrib is enough to abolish membrane localization. Taken into account that PDZ domains vary in their range and stringency of specificity [ 25 ], it is not excluded that LPP might bind to other PDZ domains than the ones of Scrib. Concerning Scrib, to date, three other proteins have been described that bind to the PDZ domains of Scrib: as mentioned above, the high-risk human papillomavirus E6 protein [ 32 ] interacts with the PDZ domains of human Scrib, whereas the GUKH (guanylate kinase holder) protein was shown to bind to the PDZ domains of Scribble at Drosophila synapses [ 38 ], and very recently, mammalian Scrib was shown to form a tight complex with the βPIX exchange factor at neuronal presynaptic compartments [ 46 ]. These findings raise the possibility that different binding partners of the Scrib PDZ domains, including LPP, can compete with each other for binding to Scrib, and as such play a role in processes in which Scrib is involved. In this regard, it is worth mentioning that the binding of LPP to Scrib appears to be regulated. In our mitochondrial targeting experiments (Fig. 7 ), we noticed that the full length wild-type LPP-protein was not targeted to Scrib-coated mitochondria in all cells. In fact, in the majority of these cells, full length wild-type LPP was not recruited by Scrib. We hypothesize that the binding of LPP to Scrib is regulated by an intra- or intermolecular interaction of LPP, as a result of which the carboxy-terminal tail is hidden in such a way that it is not available anymore for binding to Scrib. One piece of information that supports this hypothesis is the observation that, in contrast to full length LPP, the carboxy-terminal region containing only the LIM-domains and the tail but lacking the pre-LIM region, was efficiently recruited to Scrib-coated mitochondria in nearly 100% of the cells examined while carboxy-terminally mutated versions were not recruited (our unpublished results). Our observations are similar to what has been reported for the binding of zyxin to the tumor suppressor warts/LATS1. In in vitro binding experiments, it was demonstrated that parts of the zyxin protein containing LIM domains 1 and 2 efficiently bind to warts/LATS1 while the full length protein does not bind [ 47 ]. Based on these findings, the authors speculated that the LIM1/2 domains are masked in full-length zyxin, and that intramolecular and/or intermolecular modifications may regulate the interaction between zyxin and warts/LATS1. Conclusions Taken together the fact that LPP shuttles between cell adhesion sites and the nucleus [ 11 , 48 ], and the evidence that we have provided here that Scrib interacts with LPP, establishes that Scrib is connected to the communication pathway between cell adhesion sites and the nucleus of which LPP is an important element, and suggests that LPP is implicated in Scrib-associated cellular events. Methods Plasmid constructs The GFP-LPP construct was described before [ 11 ]. A construct expressing Xpress-hScrib-mito was made by cloning the coding region of human Scrib with a mutated stop codon in the pcDNA3.1/His vector (Life Technologies) followed by inserting a DNA fragment encoding the membrane anchor of ActA (LILAMLAIGVFSLGAFIKIIQLRKNN; a kind gift of Evelyne Friederich, Centre de Recherche Public-Santé, Luxembourg) behind the mutated stop codon. All amino acid changes in Scrib and LPP were made, using the QuikChange™ Site-Directed Mutagenesis Kit (Stratagene) according to the supplier's protocols. All synthetic mutations, ligation sites and PCR-amplified regions were verified by sequencing. Protein expression was checked by Western blotting. Construction and sequencing of a full-length human Scrib cDNA The KIAA0147 partial cDNA clone was kindly provided by Takahiro Nagase (Kazusa DNA Research Institute, Japan). In order to obtain full-length 5'-cDNA sequences encoding human Scrib, RNA-linker mediated 5'-RACE (RLM-RACE) was performed according to published protocols [ 49 ] using RNA isolated from HEK293 cells. The RLM-anchor primer sequence is: 5'-GGGCATAGGCTGACCCTCGCTGAAA-3'. The gene-specific primers are 1) 5'-CACGTCCAGCTCCACCAGCTGCATG-3' and 2) 5'-GAAGTTGGCCACCTCGGGAGGCAAC-3' (nested). This allowed us to construct a composite cDNA of about 5.1 kb which was completely sequenced (Genbank accession no. AF240677). Yeast two-hybrid system The Matchmaker Two-Hybrid System 2 was used (Clontech). All experiments were performed in the yeast reporter strain CG-1945. Bait-constructs were made using the vector pGBT9 (Clontech). The prey-constructs pACT2-AF6, pACT2-Erbin, pACT2-PICK1, and pACT2-PSD95 were kindly provided by Jean-Paul Borg (INSERM, Marseille, France), and were described in Audebert and Navarro et al., (AF6, Erbin, PICK1) [ 46 ] and in Saito et al., (PSD-95) [ 31 ]. The prey-constructs pACT2-Syntenin and pACT2-CASK were kindly provided by Pascale Zimmermann (University of Leuven & VIB, Belgium). An oligo(dT)- and randomly primed prey-cDNA library constructed with mRNA from 12.5 day embryonic mice using pACT2 as vector [ 50 ] was kindly provided by Kristin Verschueren and Danny Huylebroeck (University of Leuven & VIB, Belgium). The prey-library was screened as follows: yeast strain CG-1945, containing a HIS3 and a lacZ reporter gene under the control of promoters containing GAL4-binding sites, was transformed with 66 μg of bait-DNA and 33 μg of prey-library-DNA using a LiAc high efficiency transformation protocol [ 51 ]. Transformants were grown for 10 days at 30°C on triple selective (lacking Trp, Leu and His) synthetic dropout (SD --- ) agar plates containing 5 mM 3-AT (Sigma). Transformed His + yeast colonies were restreaked on new SD --- agar plates and grown for another 1 to 2 days. Colony-lift filter assays were performed for the qualitative measurement of β-galactosidase activity according to standard protocols. Cell culture, stable cell lines and transfections Cell lines used included CV-1 (ATCC CCL-70), HEK293 (ATCC CRL-1573), 293T (HEK 293 cells expressing the SV40 T-antigen), Jurkat (ATCC TIB-152), and MDCK strain II (Dog normal kidney epithelial cells). Jurkat cells were grown in RPMI 1640 (Life Technologies) supplemented with 10% fetal bovine serum. All other cell lines were grown in DMEM/F12 (1:1) (Life Technologies, Inc.) supplemented with 10% fetal bovine serum. Cells were cultured at 37°C in a humidified CO 2 incubator. Transient transfections were performed using FuGene™ 6 Transfection Reagent (Boehringer Mannheim) according to the supplier's instructions. Cells were incubated at 37°C for 18–24 hours before analysis. Stable MDCK strain II cell lines were made expressing wild-type and carboxy-terminally mutated human GFP-LPP proteins, wild-type full length Scrib-GFP, or Scrib-GFP lacking all four PDZ-domains. Transfection of MDCK cells was performed using Lipofectamine 2000 Reagent (Life Technologies) according to the manufacturer's instructions. Transfected cells were selected in medium containing 250 μg/ml G418 (Life Technologies), and resistant colonies were isolated 10–14 days later. Individual clones were screened for expression of the respective GFP fusion proteins by Western blotting using a rabbit polyclonal anti-GFP antibody (Tebu Bio). Mammalian two-hybrid system Bait-constructs were made using pM-vectors [ 52 ], prey-constructs were made in the pSNATCH-vector [ 53 ]. 24 hours upon seeding, semi-confluent HEK293 cells on 24-well plates were transiently cotransfected with 100 ng DNA of a bait-construct, 100 ng DNA of a prey-construct, 250 ng DNA of a luciferase reporter construct and 50 ng of CMV-β-galactosidase DNA (internal control for transfection efficiency). The reporter construct contains the gene encoding the firefly luciferase enzyme, which is under the control of a minimal promoter containing five consecutive GAL4-binding sequences (kindly provided by W. Schaffner and D. Escher, Zürich, Switzerland). Cell lysates were prepared 18 to 24 hours after transfection and assayed for luciferase activity as described previously [ 11 ]. In vitro transcription/translation and GST pull-down assays All in vitro translation reactions were carried out using the TNT T7 Quick Coupled Transcription/Translation System (Promega) following the manufacturer's instructions. For GST pull-down assays, bacterial expression constructs were made using pGEX-5X vectors (Amersham-Pharmacia Biotech) directing the synthesis of glutathione S-transferase (GST) fusion proteins containing wild-type or mutated forms of human LPP. These fusion proteins were purified according to manufacturer's instructions and verified by SDS-PAGE. GST fusion proteins or GST alone, bound to glutathione-agarose beads, were incubated with in vitro synthesized [ 35 S]-methionine-labelled full length human Scrib protein, or a portion of the human Scrib protein encompassing all four PDZ domains (amino acids 616–1490) (wild-type or mutated) in NENT 100 buffer (100 mM NaCl, 20 mM Tris-HCl pH = 7.6, 1 mM EDTA, 0.1% NP-40, protease inhibitors). This mixture was tumbled overnight at 4°C. Subsequently the beads were washed 5 times in 500 μl NENT 100 buffer, resuspended in 25 μl SDS-PAGE sample buffer and incubated at 95°C for 5 minutes. Proteins were separated by SDS-PAGE and interacting Scrib was detected by autoradiography. Scrib-specific antiserum and commercial antibodies The Scrib-specific polyclonal antiserum Scrib-472 was prepared by Eurogentec by immunization of rabbits with a keyhole limpet hemocyanin (KLH) coupled peptide 1612 CSSRRPVRPGRRGLGPVPS 1630 (19 C-terminal AA of human Scrib). For the detection of endogenous LPP in MDCKII cells, a LPP-specific monoclonal antibody was used (BD Biosciences, Transduction Laboratories). For the detection of GAL4-fusion proteins in immunocytochemistry, a rabbit polyclonal anti-GAL4 DNA-binding domain antibody (Tebu Bio) was used. Vinculin was detected in cells with a monoclonal anti-vinculin antibody (Sigma, clone hVIN-1), Xpress-tagged proteins with a monoclonal anti-Xpress antibody (Life Technologies). Fluorescently-tagged Alexa-antibodies (Molecular Probes) were used as secondary antibodies for immunofluorescence detection. SDS-PAGE and Western blotting Eukaryotic cell extracts were prepared by harvesting the cells in PBS (phosphate buffered saline), and subsequent lysis of the cell pellets in SDS-PAGE sample buffer (60 mM TRIS-HCl pH = 6.8, 12% glycerol, 4% SDS, 5% β-mercapto-ethanol). Protein concentrations in cell extracts were determined using BCA Protein Assay Reagent (Pierce) according to the manufacturer's instructions. 30 μg of proteins of each cell extract were loaded on 5% SDS-polyacrylamide gels. After size-separation, proteins were electrophoretically transferred to PROTEAN Nitrocellulose Membranes (Schleicher and Schuell). ECL Western blotting was performed using Western Lightning Chemiluminescence Reagent Plus (Perkin Elmer Life Sciences) according to the supplier's instructions. GFP-fluorescence and indirect immunocytochemistry CV-1 or 293T cells seeded on glass coverslips, and MDCKII cells seeded on glass coverslips or on Transwell-Clear polyester membranes (0.4 μm, Costar) were fixed in 4% formaldehyde for 20 minutes at room temperature followed by three washes in PBS containing 0.1 mM CaCl 2 and 0.1 mM MgCl 2 (PBS ++ ). For GFP-fluorescence, slides were mounted in vectashield mounting medium (Vector Laboratories, Inc.) and analyzed on a Zeiss Axiophot fluorescence microscope equipped with an RT slider SPOT camera (Diagnostic Instruments, Inc.) using SPOT RT Software v3.4, or by confocal microscopy (MRC-1024 Laser Scanning Confocal Imaging System; Bio-Rad). For indirect immunocytochemistry, after fixation, quenching was performed by incubating the cells for 10 minutes at room temperature in PBS ++ containing 50 mM NH 4 Cl. Cells were then permeabilized with 0.4% Triton-X-100 for 5–10 minutes at room temperature. Unspecific binding was blocked with 0.5% Blocking Reagent (Roche) in PBS ++ for 30 minutes at room temperature. Subsequently, the slides were incubated with primary antibodies for 1 hour at room temperature. After washing the cells three times in PBS ++ , bound primary antibodies were detected with fluorescently labelled secondary antibodies (Molecular Probes) for 30 minutes at room temperature. Following three washes in PBS ++ , slides were mounted and analyzed as described for GFP-fluorescence. Authors' contributions MMRP designed the study, performed the yeast and mammalian two-hybrid experiments, as well as the mitochondrial targeting experiments, made the LPP-stable cell lines, performed the confocal microscopy with these cell lines, and prepared the manuscript. SMPM constructed most of the DNA constructs, made the Scrib-stable cell lines, performed the epifluorescence microscopy with these cell lines, and carried out the Western blotting experiments. PA performed the GST pull down experiments, and contributed to the characterization of the Scrib-472 antibodies. TAYA contributed to the establishment of a full length Scrib cDNA clone, and to the writing of the manuscript. EJ contributed to the development of the Scrib-472 antibodies, to the establishment of a full length Scrib cDNA clone, and to the funding of the project. WJMVDV contributed to the writing of the manuscript and to the funding of the project.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546208.xml
524499	Molecular regulation of pancreatic stellate cell function	Until now, no specific therapies are available to inhibit pancreatic fibrosis, a constant pathological feature of chronic pancreatitis and pancreatic cancer. One major reason is the incomplete knowledge of the molecular principles underlying fibrogenesis in the pancreas. In the past few years, evidence has been accumulated that activated pancreatic stellate cells (PSCs) are the predominant source of extracellular matrix (ECM) proteins in the diseased organ. PSCs are vitamin A-storing, fibroblast-like cells with close morphological and biochemical similarities to hepatic stellate cells (also known as Ito-cells). In response to profibrogenic mediators such as various cytokines, PSCs undergo an activation process that involves proliferation, exhibition of a myofibroblastic phenotype and enhanced production of ECM proteins. The intracellular mediators of activation signals, and their antagonists, are only partially known so far. Recent data suggest an important role of enzymes of the mitogen-activated protein kinase family in PSC activation. On the other hand, ligands of the nuclear receptor PPARγ (peroxisome proliferator-activated receptor γ) stimulate maintenance of a quiescent PSC phenotype. In the future, targeting regulators of the PSC activation process might become a promising approach for the treatment of pancreatic fibrosis.	Review Excessive production of connective tissue molecules forming the extracellular matrix (ECM) is a pathological process relevant to diseases of many organ systems, including liver, lung, kidney, bowel and pancreas. The resulting fibrosis frequently leads to a progressive loss of specific organ functions. In the past two decades, fibrogenesis has been intensively studied by a large number of laboratories, and a great deal of scientific information has been accumulated regarding the pathogenesis of fibrosis in various organs. Until a few years ago, pancreatic fibrosis, however, remained an exception: although known for a long time as a central pathological feature of both chronic pancreatitis and pancreatic cancer [ 1 , 2 ], its cellular and molecular basics remained obscure. This situation has changed significantly since the identification of a fibroblast-like cell type in the pancreas with close similarities to hepatic stellate cells (HSCs; also called Ito cells) [ 3 , 4 ], the predominant source of ECM in the fibrotic liver [ 5 , 6 ]. In the meantime, it has become increasingly clear that these stellate cells of the pancreas (named pancreatic stellate cells; PSCs) are the principle effector cells in pancreatic fibrosis. In the following sections, I will focus on (I) the current understanding of the role of PSCs in fibrogenesis, (II) extracellular signals involved in PSC activation, (III) intracellular mediators of activation signals in PSCs, (IV) future directions of research, and (V) activated PSCs as a target for antifibrotic therapies. Pancreatic stellate cells and their role in pancreatic fibrogenesis Both chronic pancreatitis and pancreatic cancer are accompanied by an organ fibrosis [ 1 , 2 ]. The progressive replacement of pancreas-specific tissue by ECM-rich connective tissue leads to the development of an exocrine and endocrine insufficiency of the gland. So far, specific therapies to prevent, retard or even reverse this process are not available. Fibroblast activation has been reported to be a common event in pancreatitis already more than a decade ago [ 7 - 9 ], but the basic matrix producing cell type in the pancreas remained to be identified. In 1997, Saotome et al. [ 10 ] described the isolation of periacinar fibroblast-like cells from human pancreas. The cells displayed some characteristics of activated myofibroblasts, e.g. expression of α-smooth muscle actin (α-SMA) and synthesis of ECM proteins. One year later, Bachem et al. [ 3 ] and Apte et al. [ 4 ] found that vitamin A-storing cells resembling hepatic stellate cells can be isolated from human and rat pancreas. In the healthy organ, PSCs comprise about 4% of all pancreatic cells and show a periacinar distribution. They can be identified by the presence of retinoid-containing cytoplasmic lipid droplets and by immunostaining for cytoskeletal proteins such as desmin and glial fibrillary acidic protein [ 4 ]. In culture, pancreatic stellate cells readily grow [ 4 ] and change from a quiescent phenotype to a myofibroblast-like cell expressing α-SMA and producing large amounts of the ECM proteins collagen type I and III, fibronectin as well as laminin [ 3 ]. This activation process is accompanied by a loss of the characteristic retinoid-containing fat droplets [ 3 , 4 ]. Together, these in vitro data gave rise to the hypothesis that PSCs might play a pivotal role in pancreatic fibrogenesis. In the meantime, this hypothesis has been supported by the results of several in vivo studies using experimental models of pancreatic fibrosis: Infusion of trinitrobenzene sulfonic acid (TNBS) into the pancreatic duct of rats causes a pancreatic necroinflammation followed by fibrosis [ 11 ]. In TNBS-treated rats, areas of pancreatic fibrosis colocalized with α-SMA-positive cells, suggesting the presence of activated PSCs. Furthermore, dual staining techniques indicated that these α-SMA-positive cells were the main source of collagen in the fibrotic pancreas [ 12 ]. Importantly, very similar data were obtained when pancreatic tissue from patients with chronic pancreatitis was analyzed [ 12 ]. Another well-established model of fibrogenesis involves the administration of a single intravenous dose of dibutyltin dichloride (8 mg/kg body weight), resulting in the development of a chronic pancreatitis associated with fibrosis [ 13 ]. Time course studies of DBTC-induced chronic pancreatitis revealed an early activation of PSCs that preceded development of fibrosis [ 14 ]. In mice, repeated intraperitoneal application of supraphysiological cerulein doses causes a pancreatic injury and, subsequently, fibrosis [ 15 , 16 ]. In agreement with the data mentioned above, collagen gene expression was colocalized to PSCs [ 16 ]. Overexpression of transforming growth factor-beta (TGF-β) 1 in transgenic mice has been shown to be associated with increasing deposition of ECM in the pancreas. In parallel with the development of fibrosis, the number of PSCs in the pancreas increased [ 17 ]. Recently, it has also been suggested that PSCs contribute to regeneration early after acute necrotising pancreatitis in humans [ 18 ]. Together, in vitro and in vivo data suggest that PSCs are essentially involved in the development of pancreatic fibrosis. Extracellular signals involved in pancreatic stellate cell activation Based on the results of various recent studies, extracellular factors involved in PSC activation may be divided into two major groups: (I) cytokines/growth factors [ 3 , 19 - 22 ] and (II) ethanol and its metabolites, most of all acetaldehyde [ 23 ]. Cytokines stimulating PSC activation include platelet-derived growth factor (PDGF) [ 3 , 19 , 21 , 22 ], the TGF-β family members TGF-β1 [ 3 , 19 , 21 , 22 ] and activin A (24), TGF-alpha [ 3 , 22 ], basic fibroblast growth factor [ 3 , 22 ], tumor necrosis factor-α (TNF-α) [ 22 ], interleukin (IL)-1 [ 20 ] and IL-6 [ 20 ]. While TGF-β1 efficiently promotes ECM synthesis [ 3 , 19 , 21 , 22 ], PDGF is considered to be the most effective mitogen [ 22 ]. Furthermore, PDGF also enhances the migratory capacity of PSCs [ 25 ]. Potential sources of cytokines stimulating PSC activation in the inflamed pancreas are, for example, activated macrophages (secreting TGF-β1) [ 26 ], platelets (containing PDGF and TGF-β1) [ 21 ], and possibly acinar cells (expressing, among other cytokines, TNF-α [ 27 ], IL-1 and IL-6 [ 28 ]). Importantly, PSCs themselves are capable of synthesizing cytokines such as TGF-β1 [ 29 , 30 ], activin A [ 24 ] and IL-1 [ 31 ]. These observations suggest the existence of autocrine loops that may contribute to the perpetuation of PSC activation after an initial exogenous signal, thereby promoting the development of fibrosis. Recent studies have also implicated the pancreatic renin-angiotensin system [ 32 , 33 ] in pancreatic fibrogenesis. Thus, application of the angiotensin-converting enzyme inhibitor lisinopril [ 34 ], as well as the angiotensin II receptor antagonist candesartan [ 35 ], suppressed pancreatic inflammation and fibrosis in an animal model of spontaneously occurring chronic pancreatitis, Wistar Bonn/Kobori rats. In angiotensin II receptor type 1a-deficient (AT1a(-/-)) mice, pancreatic fibrosis induced by repeated episodes of acute pancreatitis (following cerulein injections) was found to be attenuated [ 36 ]. In vitro , angiotensin II (ATII) stimulates PSC proliferation [ 37 , 38 ] and induces cell contraction [ 38 ]. Cytokines that act as antagonists of PSC activation have not been systematically studied so far. Recently, it has been shown that IFN-α protects hepatic stellate cells from lipid peroxidation by enhancing biological activities against oxidative stress, resulting in an inhibition of activation [ 39 ]. Furthermore, antiproliferative effects of IFN-α [ 40 ], IFN-β and IFN-γ [ 41 ] on HSCs have been reported. On the other hand, IFN-α also inhibits spontaneous apoptosis of activated HSCs [ 40 ]. The effects of interferons on pancreatic fibrogenesis remain to be characterized. Although it is known for a long time that chronic pancreatitis, associated with fibrosis, is a serious complication of alcohol abuse, the pathogenesis of alcoholic pancreatitis still remains to be fully elucidated [reviewed in [ 42 ]]. In recent studies, the question has been addressed how long-term alcohol consumption is linked to PSC activation and fibrosis. It has been proposed that the profibrogenic effects of ethanol are in part mediated by PSC-activating proinflammatory cytokines released during episodes of alcoholic pancreatitis (associated with necroinflammation) [ 43 ]. Furthermore, in vitro data suggest that ethanol directly acts on PSCs and induces activation [ 23 ]: Cultured PSCs respond to ethanol application by increased α-SMA expression and collagen synthesis. Stimulatory effects of ethanol were detectable both in already activated and still quiescent PSCs. The cells express alcohol dehydrogenase, indicating that they are capable of ethanol oxidation and generation of its metabolite acetaldehyde. Very likely, induction of oxidant stress in PSCs contributes to the profibrogenic effects of ethanol [ 23 ]. Although the exact chain of events linking ethanol abuse to pancreatic inflammation and PSC activation remains to be described, it is likely that both direct and indirect (cytokine-mediated) effects of ethanol on PSCs are involved in the development of pancreatic fibrosis. Intracellular transduction of activation signals In the past two years, analysis of signal transduction pathways regulating PSC function has become a focus of attention. As detailed below, identification of signaling molecules that play a crucial role in PSC activation is a promising approach for the development of therapeutic strategies to inhibit pancreatic fibrosis. It is therefore envisaged that the systematic elucidation of signaling pathways in PSCs will also be one of the most important issues for future research. So far, research regarding intracellular signaling in PSCs has focused on two main aspects: the role of enzymes of the mitogen-activated protein kinase (MAPK) family and the transcriptional control of PSC activation. MAPKs are a family of serine/threonine specific protein kinases with a wide range of biological functions in the regulation of fundamental cellular processes, including gene expression, proliferation and cell survival/apoptotic cell death [ 44 - 46 ]. In mammalian cells, three major MAPK families (extracellular signal-regulated kinases [ERKs], c-Jun N-terminal kinase [JNK] and p38) have been identified [ 45 ], and all of them have recently been studied with respect to the regulation of PSC activation. The best-characterized ERKs, ERK 1 and 2, are activated through a well-established pathway (induced by many growth factors) that involves, among several other cytosolic proteins, the small G-protein Ras and the serine/threonine-specific protein kinase Raf-1 [ 45 ]. In the process of PSC activation induced by sustained culture, ERK 1/2 activation is an early event that precedes exhibition of a myofibroblastic phenotype [ 47 ]. The strong PSC mitogen PDGF induces an activation of ERK 1/2, and inhibition of signaling through the Ras-Raf-ERK signaling cascade attenuates PSC proliferation [ 47 - 49 ]. It has also been shown that exposure of PSCs to ethanol and acetaldehyde is accompanied by a fast [ 50 ] and long-lasting [ 51 ] ERK 1/2 activation. The other two major MAP kinase pathways, involving JNK and p38, are well-established mediators of signals induced by pro-inflammatory cytokines and cellular stressors (e.g., oxidant stress, UV irradiation) [ 52 ]. In PSCs, both JNK and p38 are activated in response to ethanol/acetaldehyde exposure [ 50 , 51 ]. Inhibition of p38 enzymatic activity interferes with ethanol-induced myofibroblastic transdifferentiation of PSCs [ 51 ]. The p38 signaling pathway has also been implicated in the mediation of the mitogenic PDGF effect and in the induction of PSC activation induced by sustained culture [ 53 ]. Incubation of freshly isolated PSCs with the JNK inhibitor SP600125 attenuates proliferation of the cultured cells [ 54 ]. MAP kinase pathways have also been shown to be involved in ATII signaling in PSCs [ 37 , 38 ]. Together, these data support the hypothesis that MAPKs are key mediators of activation signals in PSCs. Two other intracellular signal transduction pathways that have recently been studied regarding their role in PSC activation are the phosphatidylinositol 3 (PI 3)-kinase and the Rho-Rho kinase (ROCK) pathway. The results suggest that PI 3-kinase activity is required for PDGF-stimulated PSC migration but not proliferation [ 49 , 55 ]. The Rho-ROCK pathways was shown to be involved in the activation process of PSCs in vitro by regulating the actin cytoskeleton [ 56 ]. Cytokine and growth factor receptors exert their effects on the expression of target genes through signaling cascades that regulate the activity of a characteristic set of transcription factors. Recently, the group of the author has analyzed the activation profiles of activator protein (AP)-1 [ 57 , 58 ], signal transducer and activator of transcription (STAT) 3 [ 59 ] and nuclear factor (NF)-κB [ 60 , 61 ] in the course of PSC activation induced by sustained culture. AP-1 and NF-κB displayed an earlier maximum of DNA binding activity than STAT3 [ 62 ]. Further experiments revealed that phenotypic transition of PSCs towards myofibroblasts was accompanied by characteristic changes of AP-1 complex composition (increase of the JunD content relative to the one of JunB) [ 62 ]. DNA binding of AP-1 in PSCs is induced by PDGF, suggesting AP-1 activation as an important step in the process of PSC activation [ 47 ]. In the transduction of TGF-β receptor-derived signals into the nucleus, Smad transcription factors play a central role [ 63 , 64 ]. Studies by Ohnishi and co-workers revealed that TGF-β1 stimulated PSC activation (indicated by increased α-SMA expression) in a Smad2-dependent manner, while Smad3 was required for TGF-β1-induced growth inhibition [ 65 ]. Interestingly, exogenous TGF-β1 increased TGF-β1 mRNA expression in PSCs through an ERK-dependent but Smad2/3-independent pathway. Together, these data suggest distinct roles of Smad2-, Smad3- and ERK-dependent pathways in TGF-β1 regulation of PSC functions. Based on recently published data on HSC biology [ 66 ], it can be hypothesized that Smad7, a negative regulator of TGF-β signaling, might act as a transcriptional inhibitor of PSC activation, but so far experimental evidence has not been presented. Recent studies have implicated the nuclear hormone receptor peroxisome proliferator-activated receptor γ (PPARγ) in the inhibition of stellate cell activation in liver [ 67 - 69 ] and pancreas [ 69 ]: The PPARγ ligands 15-deoxy-Δ12,14-prostaglandin J 2 and troglitazone (an antidiabetic drug of the thiazolidinedione group) act as antagonists of PSC activation in vitro that decrease cell proliferation and expression of α-SMA [ 70 ]. In Wistar Bonn/Kobori rats, troglitazone attenuates pancreatic inflammation and fibrosis [ 71 ]. The antifibrotic effect of the drug, however, was found to be in part mediated via a PPARγ-independent mechanism [ 72 ]. Thus, the precise role of PPARγ in pancreatic fibrogenesis remains to be elucidated in further studies. Open questions with respect to PSC biology and pathology While the role of activated PSCs in pancreatic fibrosis is well established, the physiological functions of their quiescent precursors are less well understood. Importantly, PSCs are not only a source of ECM but also of matrix-degrading enzymes of the MMP (matrix metalloproteinases) family and their inhibitors (tissue inhibitors of matrix metalloproteinases, TIMPs). Thus, PSCs have been shown to secrete MMP-2, MMP-9 and MMP-13 and to express TIMP-1 and TIMP-2 [ 73 ]. It appears therefore likely that PSCs participate in the regulation of matrix turnover in the healthy pancreas. The embryonic origin of PSCs still remains to be determined. Very recently, Seaberg et al. [ 74 ] reported the clonal identification of multipotent precursors from adult mouse pancreas that generate neural and pancreatic lineages, including β-like cells and pancreatic stellate cells. With regard to PSC biology, one implication of this pioneer study is that PSCs share with exocrine and endocrine pancreatic lineages a common progenitor cell. Kruse et al. [ 75 ] have described the isolation and culture of undifferentiated pancreatic cells, capable of extended self-renewal and spontaneous differentiation into cells of all three germ layers. The relationships between these cells, which were described as stellate-like cells, and PSCs are currently unknown and should be further studied using clonal cell populations. Until now, the physiological consequences of vitamin A-storage in PSCs remain unclear. It has recently been shown by the group of the author that the vitamin A derivate all-trans retinoic acid has complex effects on PSC function and acts, at least in part, as an antagonist of the activation process [ 76 ]. It is, therefore, tempting to speculate that retinoic acids, through the binding to their nuclear receptors and the regulation of gene expression, are involved in the maintenance of a quiescent PSC phenotype. In this scenario, the loss of retinoids in the course of PSC activation might be not an epiphenomenon but an essential prerequisite. In the past, research regarding PSC biology has almost exclusively focused on the molecular basics of the activation process. However, given that participation in regeneration after pancreatic injury is an important function of activated PSCs, it is apparent that a disturbance of the inactivation or elimination of activated PSCs, rather than PSC activation itself, is the pathological process that leads to fibrosis. So far, it has not been systematically studied whether activated PSCs are capable of returning into a quiescent stage after fulfilling a repair function. Alternatively, elimination by apoptosis might be important in terminating the wound-healing response after pancreatic injury [ 77 ]. Finally, work on the complex relationships between PSCs and pancreatic tumor cells is still in its infancy. Very likely, activation of PSCs not simply accompanies tumor progression but plays an active role in this process. Thus, it has been recently been shown that pancreatic cancer growth and progression is accelerated through complex functional interactions between carcinoma cells and PSCs [ 78 ]. Furthermore, the increased deposition of connective tissue in pancreatic carcinoma was suggested to be the result of a paracrine stimulation of PSCs by cancer cells [ 79 ]. Interestingly, TGF-β1-transfected pancreatic tumor cells have been demonstrated to induce a rich stroma after orthotopical transplantation in the nude mouse pancreas [ 80 ]. Considering the established role of TGF-β1 in PSC activation [ 3 , 19 , 21 , 22 ], it appears likely that the cytokine is a key effector in tumor-associated pancreatic fibrosis. It is easy to predict that the further analysis of PSC activation in pancreatic cancer will be an important research area in the future. Studies on PSC biology are still hampered by the limited availibility of primary cells. Possibly, recently established pancreatic stellate cell lines [ 81 , 82 ] will be helpful in overcoming this problem. Therapeutic implications Given that activated PSCs are the principle effector cells in pancreatic fibrosis, targeting PSCs might become a promising therapeutic approach. Principle strategies that can be envisaged include an interruption/reversion of the activation process as well as an elimination of activated PSCs, e.g. through an induction of apoptosis. So far, potential antifibrotic drugs have been mainly tested in models of liver fibrosis (reviewed in [ 83 ]). The existence of common mechanisms in the development of liver and pancreatic fibrosis (particularly, the key role of activated stellate cells), however, suggests that at least some of these drugs may also be effective inhibitors of fibrogenesis in the pancreas. In this regard, the efficiency of substances interfering with the action of stellate cell mitogens (e.g., PDFG), or cytokines stimulating ECM synthesis (especially TGF-β), should be tested in animal models of pancreatic fibrosis. The inhibitory effects of an angiotensin-converting enzyme inhibitor [ 34 ], as well as an ATII receptor antagonists [ 35 ], on pancreatic fibrosis need to be further evaluated. Interesting candidates are also cytokines that display inhibitory effects on hepatic stellate cell activation, such as interferons. As described above, studies on the regulation of PSC activation at the intracellular level have identified key mediators of stimulatory and inhibitory signals. Targeting molecules such as PPARγ, MAP kinases, PI 3-kinase, or Smad proteins might become an important approach for the treatment of pancreatic fibrosis in the future. Further progress in the development of antifibrotic therapies can be expected from the ongoing elucidation of the molecular principles of PSC activation. Conclusions PSCs play a crucial role in pancreatic fibrogenesis (Figure 1 ). Ethanol metabolites and cytokines such as PDGF and TGF-β are key activators of PSCs. The intracellular regulation of PSC activation is incompletely characterized. MAP kinase signaling cascades are involved in the transduction of activation signals, while PPARγ ligands induce a quiescent PSC phenotype. The recent progress in the understanding of the cellular and molecular basics of pancreatic fibrosis will facilitate the development of therapeutic strategies to inhibit pancreatic fibrosis. Figure 1 Pancreatic stellate cell activation in chronic pancreatitis and pancreatic cancer. Pancreatic stellate cells are activated by profibrogenic mediators, such as ethanol metabolites and cytokines/growth factors. Perpetuation of stellate cell activation under persisting pathological conditions results in pancreatic fibrosis.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC524499.xml
526259	Use of and attitudes to a hospital information system by medical secretaries, nurses and physicians deprived of the paper-based medical record: a case report	Background Most hospitals keep and update their paper-based medical records after introducing an electronic medical record or a hospital information system (HIS). This case report describes a HIS in a hospital where the paper-based medical records are scanned and eliminated. To evaluate the HIS comprehensively, the perspectives of medical secretaries and nurses are described as well as that of physicians. Methods We have used questionnaires and interviews to assess and compare frequency of use of the HIS for essential tasks, task performance and user satisfaction among medical secretaries, nurses and physicians. Results The medical secretaries use the HIS much more than the nurses and the physicians, and they consider that the electronic HIS greatly has simplified their work. The work of nurses and physicians has also become simplified, but they find less satisfaction with the system, particularly with the use of scanned document images. Conclusions Although the basis for reference is limited, the results support the assertion that replacing the paper-based medical record primarily benefits the medical secretaries, and to a lesser degree the nurses and the physicians. The varying results in the different employee groups emphasize the need for a multidisciplinary approach when evaluating a HIS.	Background Hospital information systems (HIS) and Electronic Medical Records (EMRs) are considered prerequisites for the efficient delivery of high quality health care in hospitals. However, a large number of legal and practical constraints influence on the design and introduction of such systems [ 1 ]. Hence, many EMR implementation projects do not aim at introducing the EMR and eliminating the paper-based counterpart in one step [ 2 ]. As a start, the EMR is introduced along with its paper-based counterpart, and both are kept updated. In such environments, health care workers have to deal with a hybrid electronic and paper-based solution. This probably limits the use of EMR [ 2 ]. Furthermore, errors are prone to develop due to cumbersome maintenance of the medical record information in dual storage media [ 3 ]. In Norway and in other countries, most hospital EMR projects have not passed beyond this phase [ 1 ] Aust-Agder Hospital is the first hospital in Norway to eliminate the paper-based medical record, using a widespread [ 2 ] and commercially available HIS in combination with scanning technology. In a recent report, we have evaluated the EMR part of the HIS in this hospital [ 4 ], discussing the views of the physicians only. However, to get a more complete picture of the impact of the system, its use by employees other than physicians needs to be evaluated. Both medical secretaries and nurses are important users of a HIS, utilizing both the EMR and the administrative part of the system. The medical secretaries work as transcriptionists, receptionists and coordinators of patient logistics and communication, and the nurses have their own documentation and administrative routines. The elimination of the paper-based medical records is a radical change in the work routines in the hospital organization. To assess the impact of this change on the organization, the EMR system may be described from the perspectives of three important employee groups separately. In this report, we have used questionnaires and interviews to assess how often medical secretaries, nurses and physicians use the HIS system for essential tasks, how easily these tasks are performed using the system, and how satisfied the hospital employees are with it. Methods The hospital The investigation was performed in a 410-bed community hospital in Aust-Agder county, Norway. The hospital serves a population of 102,000, caring for 18,600 inpatients and 74,000 outpatients per year (1998). The patients are admitted by primary care physicians external to the hospital and followed up by the hospital physicians. The hospital comprises of departments for psychiatry, general surgery, internal medicine, orthopaedics, gynecology, ear, nose and throat and ophthalmology. Well funded, and with a strong commitment by the hospital administration, the hospital staff began implementation of DIPS 2000 ® , a commercially available combined EMR and hospital administrative system in March 2000. In April 2001, all except the psychiatric department started to scan documents. From this date, all new patient data was channeled into the EMR in these departments, either as electronic text and data or as scanned documents. The HIS was available in 1100 terminals throughout the hospital, except for the inpatients' rooms. The transition to HIS was administered by a project group, which had been recruited from the hospital staff. The group worked in conjunction with the IT department and the HIS vendor, and was also responsible for communicating with and training the users. The group regularly held series of mandatory hands-on training classes adapted to each profession (3–8 h in total). However, a substantial proportion of the users never attended the classes, particularly the physicians. To reach these users, a task force of medical secretaries was trained and employed during the first month after implementation of the HIS for ambulant training in the wards. Further support was provided by a network of super users (the most experienced users) among the ward staff. The EMR The patient data in the EMR part of the HIS is either stored as searchable text and numbers or as document images. The former, called "regular electronic data", essentially consists of the chronological, text-based medical record integrated with lab data in numerical form and textual radiology reports (fig 1 ). The latter is divided by structure into two categories, as follows: Upon admittance or consultation, the documents in the old paper-based medical records are scanned into the system as digital images in TIFF format. Each image contains all the sheets of one main section of the paper-based record, and hence corresponds to a whole document group (groups A-J in fig 1 ). These images are called "scanned multiple documents". Upon patient discharge, various paper sheets accumulated during the stay (e.g. the medical treatment form, printouts from diagnostic devices) are scanned, dated and labeled by document type singularly (fig 1 ). The resulting images are called "scanned single documents". In summary, the patient data is stored as regular electronic data, scanned multiple documents and scanned single documents. They all appear in the hierarchical list in the "medical record explorer" window (fig 2 ), but are treated separately in this paper, due to their difference in structure, indexation and functionality. The user interface of the HIS system is identical to all types of users, although medical secretaries, nurses and physicians often utilize different parts of system. The survey A questionnaire previously used in a national survey of hospital physicians [ 2 ] was modified for this study. The original questionnaire contained sections regarding frequency of use of an EMR system or HIS for specified tasks, user satisfaction with the system as a whole [ 5 ] as well as detailed aspects of it [ 6 ], and availability of computers. To make the questionnaire applicable to medical secretaries and nurses, new versions of the section regarding frequency of use of the HIS were developed. In collaboration with the authors, 3–6 representatives from the medical secretaries and the nurses identified work tasks for the questionnaire each in two 2-hour group sessions, using recently developed detailed work-flow charts as templates (not shown). The identified tasks were then reduced to 23 and 19 tasks supported by the HIS, respectively (see appendix A). The questionnaire was reviewed in similar sessions by representatives from the physicians. As a result, one new task was added to the physicians' questionnaire, and four tasks not supported by the HIS were removed. For all professions, a new section was added, containing questions about ease of performing each task using the system. The survey was conducted during February–April 2002, and 85 medical secretaries, 235 nurses and 80 physicians in the medical, surgical and other somatic wards received the questionnaire. Of these, 79 medical secretaries (93%), 172 nurses (73%) and 70 physicians (88%) responded, giving a total response rate of 81% (321/400). We used Teleform™ for data acquisition and SPSS 11.0 for Windows™ for statistical analysis. In addition to the survey, one of the authors interviewed 8–12 representatives of each profession for 0.5–2 hours. Comments on advantages and disadvantages of the system in all relevant work tasks were noted and summarized Results The medical secretaries used the HIS routinely for most of their tasks defined in the questionnaire. This stands in contrast to the nurses and the physicians (fig 3 ). The number of tasks with a median response of "always or almost always" was highest for the medical secretaries (15 out of 23 tasks, 65%), and lowest for the nurses (4 out of 19 tasks, 21%). The medical secretaries reported that all of the defined tasks were performed more easily than before the HIS was introduced (i.e. median response for ease of performing the task was "increased" or "significantly increased", in 23 out of 23 tasks, fig 4 ). In comparison, the number of tasks more easily performed was much lower for the nurses and the physicians (respectively 9 [47%] and 7 [37%] out of 19 individual tasks). The medical secretaries were much more satisfied with the use of the HIS than the nurses and physicians, both when assessing the detailed aspects of it and the system as a whole. The detailed aspects of the HIS was assessed in twelve questions related to the factors content, accuracy, format, user friendliness and timeliness [ 6 ]. The parts of the HIS that contained scanned document images and regular electronic data were assessed separately. The medical secretaries were equally satisfied with both parts of the HIS (fig 5 ). This stands in contrast to nurses and in particular the physicians, who were less satisfied, particularly with the part containing the scanned document images. The difference between the professions was significant in all factors regarding the scanned document images (ANOVA p < 0.001), and in all factors except accuracy regarding the regular electronic data (fig 5 , (ANOVA p = 0.001 to 0.04, p = 0.07 for factor 'accuracy'). In addition to the detailed aspects, the user satisfaction with the HIS as a whole was assessed (fig 6 ). The medical secretaries gave significantly more positive responses than the nurses and the physicians in all of the five questions in this section (Kruskall-Wallis p = 0.05 in question 2, p < 0.001 in the remaining four questions). However, the majority of each profession gave positive answers in all of these questions. To summarize all results regarding user satisfaction, the system seems to be well adapted to the work of medical secretaries but leave nurses and physicians less satisfied. Partly explaining the differences in user satisfaction, the physicians reported more frequent problems related to availability of the HIS than the medical secretaries and the nurses (fig 7 , Kruskall-Wallis p < 0.001 in all questions). The most frequently reported problems among the physicians occurred daily or weekly, and consisted of various software and hardware-related problems, the system working too slowly, and lack of computers where the clinical work was being done. Such problems were not frequently reported among the medical secretaries, except problems with the systems working too slowly (42% daily or weekly, 32/77). In the interviews, the perceived advantages and disadvantages of the HIS were discussed. Both nurses and physicians in the medical ward found that patient data were more accessible when stored electronically than when stored on paper, in particular regarding lab test data. However, the nurses were still using pen and paper when documenting their activities. The medical secretaries found that generation, handling, fetching and delivery of paper documents and logistics of paper-based patient records had diminished dramatically. The generation of written text had become considerably easier. On the other hand, the scanning process had become an additional burden and was considered time consuming. Overall, handling of paper documents was considered additional work whenever the documents appeared. Discussion In this hospital, we have found that the medical secretaries use the HIS more extensively for their tasks than the nurses and the physicians. Also, they are much more satisfied with the HIS. Medical secretaries reported that they use the HIS routinely for most of the tasks defined in the questionnaire (fig 3 ). A simple explanation is that their tasks generally are smaller in scope and have a smaller and more easily defined range of needed information types than that of the nurses and physicians (See appendix A). Hence, the medical secretaries' tasks should be more easily supported by computers than the nurses' and the physicians' tasks. The particular inefficiencies of certain paper-based routines (e.g. regarding task 6, 15, 18 and 19) readily demonstrates the usefulness of computer support [ 7 ]. Unlike the work of nurses and physicians, the work of medical secretaries is stationary, avoiding the difficulties in providing an efficient mobile work environment. In addition, each medical secretary typically is assigned a computer, while nurses and physicians usually have to share a limited number of them (fig 7 , question C). Another possible reason for the difference in usage pattern could be difference in computer literacy. However, the usage patterns were not consistent with the limited differences found in self-reported computer literacy (data not shown), and the amount of in-house training of medical secretaries and physicians was principally equal. The medical secretaries reported that all of the tasks in their questionnaire are more easily performed (fig 4 ). The results from the interviews identify the elimination of the paper-based medical record as a major contributor to this, as several manual paper routines have disappeared (e.g. searching for a lost paper-based medial record or sorting the contents of a medical record) or are replaced by more efficient computer functions (e.g. transferring new lab data to doctors for review). Furthermore, having the administrative functions integrated with the EMR means that a substantial selection of structured demographic, clinical and administrative data is concurrently available to the users of the HIS. This makes several tasks more efficient for the medical secretaries (e.g. sending standard letters to patients in waiting lists). The results are supported by the fact that the number of medical secretaries in the hospital has been reduced by 15 since the onset of the HIS project (Bjørn Engum, personal communication Sept 2003). Not surprisingly, the medical secretaries were more satisfied with the system than the nurses and the physicians (figs 3 and 4 ). This agrees with the results of Sittig [ 8 ] and Lee [ 9 ], who both found that user satisfaction was strongest correlated to questions regarding how easily the work was done. On the other hand, when comparing the user satisfaction scores to the reference data of Doll & Torkzadeh [ 6 ], the median user satisfaction score of the medical secretaries lies between the 20 th and 30 th percentile of the reference data set. This suggests that there is room for improvement of the EMR system regarding the medical secretaries as well as the others. Unlike the nurses and the physicians, the medical secretaries were equally satisfied with the scanned document images as that of the regular electronic medical record. The most likely reason is that the document images are not very often used by the medical secretaries, particularly the document images scanned in sections (data not shown). The disadvantages of the document images, for instance that they can not be searched, therefore seem to affect the user satisfaction of nurses and physicians to a stronger degree than that of the medical secretaries. The use of the HIS by medical secretaries, nurses and physicians may to some degree be compared at a task-by-task level when the tasks are equally worded. In these tasks, work roles seem to explain the differences. For instance, the tasks "Reviewing the patient's problems" (tasks 1) and "Seek out specific information from patient records" (task 2) appeared in all questionnaires. Of all the respondents, only the physicians had a significant proportion finding that these tasks were more difficult to perform than before (figure 4 ). A possible reason is that the physicians, in order to perform these tasks as they saw fit for their work role, more often needed to search the scanned document images extensively. When examining the task "Order clinical biochemical laboratory analyses" (task 6 for nurses, task 7 for physicians), the nurses both use the HIS more frequently for this task and find the task more easily to perform than the physicians. However, many Norwegian physicians find that order entry is a task better performed by others [ 10 ], reducing the motivation for learning the new system. This way, understanding work roles in the given context appears necessary to interpret the results. A secondary finding in this study was that the physicians reported frequent computer-related problems, much more frequent than that of medical secretaries and nurses (fig 7 ). This may be due to escalated demands on computing power, system stability and availability. Without the paper-based medical record, the EMR is taken into full use and the real demands of supporting the physicians' information processing are revealed. The high reported frequency of computer-related problems may partly explain the overall lower user satisfaction of the physicians, as well as the relatively high proportion of physicians finding certain tasks more difficult to perform (task 1 and 2, fig 4 ). An observational study could elaborate on these relationships, focusing on what kinds of computer problems are the least tolerable to the physicians. Limitations of the study In the questionnaire, we do not know how often each task is carried out (using the HIS or not) or how long it takes, which means that demanding tasks might be outnumbered by the less demanding ones. Furthermore, the list of tasks supported in some way by the system may not be complete, and the list does not cover the full range of conceivable tasks suited for support by any given HIS. However, given that the tasks defined for each group cover important parts of their information-related work, a cautious comparison of general patterns of use between groups of hospital employees is possible. Conclusion Evaluation of a HIS in a hospital that has eliminated the paper-based medical record reveals considerable differences in user satisfaction and reported use of the system among medical secretaries, nurses and physicians. Although the basis for reference is limited, the results seem to support the claim that replacing the paper-based medical record primarily benefits the medical secretaries, and to a lesser degree the nurses and the physicians. Inspired by Aust-Agder Hospital, two of 22 other Norwegian hospitals using the same system (as of Aug 2002) are about to eliminate the paper-based medical record, making a future comparison between hospitals possible. When assessing the effects of a HIS on a hospital organization by asking users, the multidisciplinary nature of health care provision should be reflected in the selection of hospital employees that participate in the evaluation. Competing interests The authors declare that they have no competing interests. Authors' contributions HL, THK and AF planned the investigation. HL and THK developed the questionnaires for the medical secretaries and nurses. THK organized the administration of the questionnaires, and HL scanned and analysed the data. HL, THK and AF wrote the manuscript jointly. Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 Appendix A: Task lists The three lists of tasks as they appear in the questionnaire developed for the medical secretaries, nurses and physicians, respectively. Click here for file Additional File 2 An English translation of the questionnaire used for the physicians in the survey. It is meant for review purposes. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC526259.xml
534113	A survey of relationship between anxiety, depression and duration of infertility	Background A cross sectional study was designed to survey the relationship between anxiety/depression and duration/cause of infertility, in Vali-e-Asr Reproductive Health Research Center, Tehran, Iran. Methods After obtaining their consents, 370 female patients with different infertility causes participated in, and data gathered by Beck Depression Inventory(BDI) and Cattle questionnaires for surveying anxiety and depression due to the duration of infertility. This was studied in relation to patients' age, educational level, socio-economic status and job (patients and their husbands). Results Age range was 17–45 years and duration and cause of infertility was 1–20 years. This survey showed that 151 women (40.8%) had depression and 321 women (86.8%) had anxiety. Depression had a significant relation with cause of infertility, duration of infertility, educational level, and job of women. Anxiety had a significant relationship with duration of infertility and educational level, but not with cause of infertility, or job. Findings showed that anxiety and depression were most common after 4–6 years of infertility and especially severe depression could be found in those who had infertility for 7–9 years. Conclusions Adequate attention to these patients psychologically and treating them properly, is of great importance for their mental health and will improve quality of their lives.	Background The impact of infertility on the psychological well being of couples involved has been the object of increasing attention in recent years. It cannot be denied that infertility is a deeply distressing experience for many couples [ 1 ]. In latter part of twentieth century, psychogenic cause was an accepted topic in infertility until when diagnostic abilities improved [ 2 ]. Edelmann et al (1985) found that infertility has a significant effect on psychological factors. Some authors have paid attention to the fact that health problems, loss of self-esteem, feeling akin to mourning, threat, sexual distress, depression, guilt, anxiety, frustration, emotional distress and marital problems are all associated with infertility [ 3 ]. We aimed to examine prevalence and severity of anxiety/depression in relation to duration/cause of infertility in Iranian infertile women. Several studies have demonstrated anxiety has a detrimental effect on fertility [ 4 ] and that reduction of anxiety increases pregnancy rate [ 5 , 6 ]. Other researches failed to support a relationship between anxiety and infertility [ 7 ] Lapane et al (1995) indicated that depression could play an important role in the pathogenesis of infertility [ 8 , 9 ]. Infertility sometimes is accompanied by existential crises and emotional tensions such as anxiety, interpersonal problems, and suppressed anger, unsatisfactory in interpersonal, frustration, inferiority feeling, depression, rejected feeling and unconscious guilt feeling. Those couples with a history of failure in Assisted Reproductive Technique (ART) have shown personality maladjustment [ 10 ]. Overall percentage of psychological problem in infertile couples ranges between 25 and 60% [ 11 ]. One study has demonstrated that 74.6% patients reported changes in their mood [ 12 ]. Psychological difficulties of infertile patients are complex and influenced by a number of factors such as gender differences, cause and length of infertility. Freeman et al (1987) found that half of their sample of infertile couples described infertility as the most upsetting experience of their lives, whereas 80% of the sample reported by Mahlstedtet et al (1987) described their experience of infertility to be either stressful or very stressful [ 1 ]. Duration of infertility increases of stress [ 13 ]. Depression and anxiety were improved in infertile women as their age and duration of infertility increased [ 14 ]. Long lasting infertility and unsuccessful treatment cycles intensifies stress and psychopathologic problems especially depression [ 15 , 16 ]. Although, some studies showed that there is no relation between duration of infertility and depression or psychological factors [ 17 ]. Another study showed those who had 2–3 years infertility had more depression / anxiety than those who had this problem for a year or more than 6 years. Peak of depression could be seen during third year of infertility. After six years there will be a reduction in psychological symptoms in women. During first three years, infertility is accompanied by signs such as anxiety, depression, loss of self-esteem, impotence and maladjustment of marital status. After 3 years, optimistic attitude would change to despair and at last there will be some emotional changes to adopt a child or live without one, thereafter. Those who have social support, positive personal characteristics, and have a satisfactory life with their spouse show no signs of anxiety/depression [ 18 ]. Since most literature on psychological aspects of infertility is from developed countries it was thought that a study from a developing country with a different culture might contribute to existing knowledge on the topic. We aimed to examine prevalence and severity of anxiety/depression in relation to duration of infertility in Iranian infertile women. The results of this study, which included counseling and couple-therapy, are being prepared for infertile couples. Methods The subjects were 370 infertile women who were referred to Vali-e-Asr Reproductive Health-Research Center between January 2001 and January 2002 for treatment of their infertility problems. In this group mean (± SD) age was 28 (± 5.37) and mean (± SD) duration of infertility was 6.36 (± 4.18). Among them 293 (79.2%) women were housewives and the others were working outside. A gynecologist evaluated patients and then they were visited by a psychologist and were informed of the study purposes. Diagnosis of male factor is based on WHO criteria 1999. Menstrual history and mid-luteal progesterone level considers ovulatory factor, diagnosis of endometriosis is done by lapascopy, tubal factor is diagnosed by HSG and cervical facto is detected in Post Coital Test. After obtaining oral consent from each patient, data were collected using BDI [ 19 , 20 ], and Cattle [ 21 ]. inventories: Beck Depression Inventory (BDI) The test used was a translated and validated Persian version of Beck's depression Inventory. A full 21-items BDI was administered. This scale is a widely used measure for intensity of depression. Each item describes a specific behavioral manifestation of depression. Scores on each item can range from 0, indicating no depressive symptomatology, to 3, indicating a severe level of symptomatology. Total scale scores can thus range from 0 to 63. Scores of 17 or above it indicates of a clinically significant depression. The classification of depression scores involves: 1. 0–16 (without depression) 2. 17–27 (mild depression) 3. 28–34 (moderate depression) 4. 35–63 (severe depression) Cattle Inventory The Cattle inventory is a 40 items self-report measure of anxiety. This test was a translated and validated Iranian version of Cattle's Inventory. Scores can range from 0 to 80, with scores of 28 or above demonstrate anxiety. Classification of anxiety scores involves: 1. 0–27 (without anxiety) 2. 28–40 (moderate anxiety) 3. 41–49 (neurotic anxiety) 4. 50–80 (severe anxiety) For each patient the following data were recorded: age, cause and duration of infertility, education and job. Data were analyzed by using statistical SPSS. The relationship between continuous and binary explanatory variables with Beck and Cattle scores were assessed using spearman's rho and unpaired t-test, respectively. In addition, the relationship between categorical responses and explanatory variables were evaluated using chi-square test. For descriptive purposes, we presented frequency tables. Results Three hundred seventy infertile women were considered in this cross-sectional study. The results of Beck and Cattle inventories showed that 40.8% and 86.8% of these women had depression and anxiety symptoms, respectively (Table 1 ). Table 1 Prevalence of depression and anxiety Depression (Beck) Frequency Percent Anxiety (Cattle) Frequency Percent Normal 219 59.2% Normal 49 13.2% Mild 96 25.9% Moderate 141 38.1% Moderate 37 10% Neurotic 117 31.6% Severe 18 4.9% Severe 63 17% Total 370 100% Total 370 100% Mean scores of depression and anxiety by age groups are shown in figure 1 . Depression and anxiety were more severe in 21–25 years and under 20 years respectively, although there was no significant difference between age groups. Figure 1 Mean score of depression/anxiety by age groups In next step, we compared the prevalence of depression (depression group is consisted of mild, moderate and severe depressive women) in all categories of infertility causes. Chi 2 statistic for this 2 × 9 table (chi-square = 20.643 P = 0.014) showed that prevalence of depression is not equal in these categories. Same analysis for the prevalence of anxiety (anxiety group is consisted of moderate, neurotic and severe anxietic women) in different groups of infertility causes showed no significant difference between them (chi-square = 7.491 P = 0.485) (Table 2 ). Table 2 Frequency of depression and anxiety by cause of infertility Causes of Infertility Percent of Depression Percent of Anxiety Oligo-astheno-terato spermia 24.6% 86.2% Azospermia 31.6% 80.7% Ovulatory 48.0% 85.7% Endometriosis 20.0% 80.0% Uterus 52.2% 82.6% Tubal 50.0% 90.5% Habitual abortion 33.3% 100% Unexplained 56.5% 95.7% Male & Female (Both) 49.0% 93.9% P-value 0.014 0.485 chi-square = 20.643 The correlation between Beck and Cattle scores based on Spearman's Rho was 0.707 (P < 0.001), which shows a significant relation between depression and/or anxiety scores. Then, we checked correlation coefficient between Beck and Cattle scores with quantitative variables like age, education (in years) and especially the duration of infertility. Duration of infertility showed a significant relation with both Beck (r = 0.15, P = 0.004) and Cattle (r = 0.157, P = 0.002) scores. Based on duration of infertility 31(29.2%), 51 (42.1%), 30 (46.8%) and 39 (49.3%) patients had depression in different groups, but there was no significant relationship between duration of infertility and depression (P-value = 0.106) (Table 3 ). Table 3 Frequency and rate of depression and anxiety based on infertility duration. 1–3 years 4–6 years 7–9 years >10 years All Freq. Rate Freq. Rate Freq. Rate Freq. Rate Freq. Rate Depression stage (P = 0.106) Normal 75 70 70 57.8 34 53.8 40 50.6 219 59.2 Mild 22 20.5 34 28.1 16 25 24 30.3 96 25. Moderate 5 4.6 12 9.9 8 12.5 12 15.1 37 10. Severe 4 3.7 5 4.1 6 9.3 3 3.7 18 4.8 Anxiety Stage (P = 0.048) Normal 24 22.6 11 9 6 9.3 8 10.1 49 13.3 Moderate 41 38.6 51 42.1 24 37.5 25 31.6 141 38.1 Neurotic 29 27.3 36 29.7 24 37.5 28 35.4 117 31.6 Severe 12 11.3 23 19 10 15.6 18 22.7 63 16.9 Also 82 (77.3%), 110 (90.9%), 58 (90.6%) and 71 (89.8%) of patients had different stages of anxiety and there was a significant relation between anxiety and duration of infertility (P-value = 0.048) (Table 3 ). Educational level had a significant and negative relation with these two scores, but age showed no significant effect on depression and/ or anxiety (Table 4 ). Table 4 Correlation between Beck and/or Cattle scores with duration of infertility, women's age and women's education Beck scores Cattle scores Spearman's Rho P- value Spearman's Rho P- value Duration of Infertility 0.150 0.004 0.157 0.002 Women's Age -0.052 0.315 -0.044 0.395 Women's Education (in years) -0.319 < 0.001 -0.156 0.003 Finally, we tested the relation between depression and/or anxiety scores with women's job. Results of unpaired t-test showed significant difference between depression scores of housewives and employees (t = 9.179, P = 0.003). Anxiety and depression were observed more in homemakers comparing to outside employees though we found no significant effect of women's job on Cattle scores (t = 2.943, P = 0.087) (Table 5 ). Table 5 Relation between depression and/or anxiety scores and women's job Women's job Mean SD P-value Beck Housewife 16.3549 10.2261 0.003 Employee 12.4286 9.7014 Cattle Housewife 6.3857 2.084 0.087 Employee 5.9091 2.4878 t = 9.179 **t = 2.943 Discussion Patients participating in this study were from different geographical areas in Iran. The finding of this study provides information about frequency and severity of anxiety/depression in order to duration of infertility in childless women. The prevalence of psychiatric morbidity specially depression and/or anxiety in infertile patients have been assessed in several countries, for example Jones et al (1993) found that, there was mild to moderate depression in 28.3% of infertile women, moderate to severe depression in 7.2% and 1.2% had most severe depression based on BDI [ 7 ]. Another study showed that 67% of infertile women suffered from anxiety [ 1 ] and the same studied by Oddens et al (1999) reported that 24.9% had depressive disorders [ 22 ]. Anxiety were investigated in 130 infertile women in China, the results showed that different levels of mental pressure were found in 83.8% of infertile women, and moderate or severe types in 25% [ 23 ]. There was depression and/or anxiety disorder in 33% (Hong Kong), in 32 % (Scotland) of infertile women [ 16 , 22 ]. The overall percentage of depression disorder in infertile women ranges between 24 and 36% and also anxiety disorder ranges between 67 and 84%. Our study showed 40.8% depression and 86.8% anxiety in infertile women. Consistent with our research, Iranian infertile women show higher rates of anxiety and/or depression than the other countries. In Islamic and eastern countries such as Iran, family status especially childbearing is very important and valuable. Having a child stabilizes family and increases marital satisfaction. In our culture and society, negative attitudes to infertility are so throbbing. Having a child is psychologically or effectively, a vital factor for women, and the absence of children may cause marital problems such as divorce or even second marriage especially in Islamic societies which it is possible for men to marry with more than one woman. Intervention of relatives especially husband's family, negative attitude and behavior of surroundings (family, friends, neighbors, etc.) causes psychological problems for infertile women. Generally infertile women experience negative social consequences including marital instability, stigmatization and abuse. Infertility can have a serious effect on both psychological well being and social status of women in our country. The most common age for depression and/or anxiety in our study was 21–25 years. In this study, anxiety and/or depression had negative correlation with education. In other words, with the raise of education level, anxiety and/or depression decrease. Results of different studies about relationship of age and education with anxiety and/or depression were not similar. Age and education level have no significant relationship with depression and/or anxiety [ 24 ]. Another study showed that there was positive correlation between them [ 18 ]. In such closed societies as some parts of our country, education and job may be the lone gate leading women to joyful aspects of their life other than maternity. This is why education plays a considerable role in decreasing their depression/anxiety. Having a job may reduce stress from In Vitro fertilization (IVF) [ 25 ]. In our study, anxiety and/or depression were observed more in housewives (vs. outside employees). It seems being at work outside home decreases psychological signs of anxiety and depression. Based on previous researches [ 26 - 28 ], infertile women showed higher rates of psychiatric symptoms than their partners, especially in female and unexplained factors. Women are necessarily more deeply involved in treatment procedures and it is normal for them to be more affected. One of the characteristics of infertile couples is that women are habitually more affected by the situation of infertility than men [ 29 ]. Based on our study depression is more common in "unexplained cause" group comparing to other causes and anxiety is more common in "endometriosis" group comparing to other causes, our results is similar to other studies [ 2 , 3 , 6 , 9 , 13 ]. Anxiety and/or depression increases with duration of infertility [ 30 ]. A study demonstrated that women who had experienced infertility for a long or medium range of time presented a significantly lower state of anxiety [ 31 ] and there was a trend of decreasing psychological stress with lengthening of infertility time. Based on depression scales, infertile patients who had infertility for an intermediate to a long time showed less symptoms than those who are in their first stage of their problem [ 32 ] but other studies showed that psychological distress in infertile women increase with time [ 15 ] and depression peaks between the second and third year of infertility and does not return to normal range until after 6 years of infertility [ 18 ]. In our study, women with lower stages of depression and anxiety can be seen during 1–3 years of infertility, but during 4 through 6 years after infertility diagnosis their signs become more prominent, especially severe depression has the most common frequency during 7–9 years. It was shown that the first three years (1–3) anxiety and/or depression is in its lowest limit and after 4 to 9 years it becomes worse. It seems that our results are completely different comparing to other countries. It may show that having a child is very important for our people, especially our women, therefore women show higher and longer emotional reactions and psychiatric symptoms lasts longer in comparison to other countries. In conclusion it can be suggested that psychological interventions especially in 4–9 years of infertility may prevent the surge in depression/anxiety and could presumably lead to increased pregnancy rates. Authors' contribution FM contributed to design and management of the study. NA contributed to study design and data gathering. The initial idea of this study was by MMA. NK contributed in design, writing of the manuscript and analysis of this research. FZ contributed in design, writing of the manuscript and analysis of this research. MS contributed in edition and revision of the article. MJ contributed in edition and revision of the article. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC534113.xml
524472	Components of Coated Vesicles and Nuclear Pore Complexes Share a Common Molecular Architecture	Numerous features distinguish prokaryotes from eukaryotes, chief among which are the distinctive internal membrane systems of eukaryotic cells. These membrane systems form elaborate compartments and vesicular trafficking pathways, and sequester the chromatin within the nuclear envelope. The nuclear pore complex is the portal that specifically mediates macromolecular trafficking across the nuclear envelope. Although it is generally understood that these internal membrane systems evolved from specialized invaginations of the prokaryotic plasma membrane, it is not clear how the nuclear pore complex could have evolved from organisms with no analogous transport system. Here we use computational and biochemical methods to perform a structural analysis of the seven proteins comprising the yNup84/vNup107–160 subcomplex, a core building block of the nuclear pore complex. Our analysis indicates that all seven proteins contain either a β-propeller fold, an α-solenoid fold, or a distinctive arrangement of both, revealing close similarities between the structures comprising the yNup84/vNup107–160 subcomplex and those comprising the major types of vesicle coating complexes that maintain vesicular trafficking pathways. These similarities suggest a common evolutionary origin for nuclear pore complexes and coated vesicles in an early membrane-curving module that led to the formation of the internal membrane systems in modern eukaryotes.	Introduction The ability to sharply curve membranes was a defining event in the evolution of early eukaryotes, allowing the formation of endomembrane systems ( Blobel 1980 ). In modern eukaryotes, these systems have become elaborate internal membranes, such as the Golgi apparatus, the endoplasmic reticulum (ER), and the nuclear envelope (NE). To date three major kinds of transport vesicles, distinguished by the compositions of their protein coat complexes, have been shown to traffic between these internal membranes and the plasma membrane: First, the clathrin/adaptin complexes are responsible for endocytosis and vesicular trafficking between the Golgi, lysosomes, and endosomes; second, the COPI complex mediates intra-Golgi and Golgi-to-ER trafficking; and lastly, the COPII complex supports vesicle movement from the ER to the Golgi (reviewed in Kirchhausen 2000a , 2000b ; Boehm and Bonifacino 2001 ; Bonifacino and Lippincott-Schwartz 2003 ; Lippincott-Schwartz and Liu 2003 ). The NE is contiguous with the ER and delineates the nucleus. It is made of an inner and outer membrane that together form a barrier between the nucleoplasm and the cytoplasm. This barrier is perforated by nuclear pore complexes (NPCs), which form pores between the inner and outer NE membranes by stabilizing a sharply curved section of connecting pore membrane. NPCs are approximately 50-MDa octagonally symmetric cylinders that function as the only known mediators of nucleocytoplasmic exchange; while permitting the free flow of small molecules, they restrict macromolecular trafficking to selected cargoes that are recognized by cognate transport factors. NPCs are found in all eukaryotic cells and are composed of a broadly conserved set of proteins, termed nups (reviewed in Rout and Aitchison 2001 ; Bednenko et al. 2003 ; Rout et al. 2003 ; Suntharalingam and Wente 2003 ; Fahrenkrog et al. 2004 ). Although the nups have been fully cataloged for both yeast (Saccharomyces) ( Rout et al. 2000 ) and vertebrates ( Cronshaw et al. 2002 ), there is currently little information concerning their origin and evolution. To this end, protein structures are helpful because it is easier to recognize similarities in structure than in sequence, especially for distantly related proteins. Thus, we have characterized the structures of seven proteins forming a core building block of the NPC, termed the yNup84 subcomplex in Saccharomyces and the vNup107–160 subcomplex in vertebrates. These structures reveal how the nuclear pore complex could have evolved from organisms with no analogous transport system. Results The yNup84/vNup107–160 subcomplex has a molecular weight of approximately 600 kDa and has been shown in yeast to be flexible ( Siniossoglou et al. 1996 ; Siniossoglou et al. 2000 ; Lutzmann et al. 2002 ), presenting a considerable challenge to conventional experimental methods for structure determination; thus, we used a computational approach that relies on a database of experimentally determined structures ( Marti-Renom et al. 2000 ). We first focused on the component nups of the yNup84 subcomplex: ySeh1, ySec13, yNup84, yNup85, yNup120, yNup133, and yNup145C, whose corresponding vertebrate homologs are, respectively, vSec13 l, vSec13R, vNup107, vNup75, vNup160, vNup133, and vNup96 ( Siniossoglou et al. 1996 ; Fontoura et al. 1999 ; Siniossoglou et al. 2000 ; Cronshaw et al. 2002 ; Lutzmann et al. 2002 ; Boehmer et al. 2003 ; Harel et al. 2003 ; Walther et al. 2003 ; Loiodice et al. 2004 ). For putative domains in each of these nups, we first applied two threading programs to assign structure folds based on similarity to known protein structures (templates) ( Marti-Renom et al. 2000 ) (see Materials and Methods ). The corresponding sequence-structure alignments were refined and used to generate three-dimensional models of the nup domains, followed by evaluation of the models. Our analyses predicted that every nup in the yNup84/vNup107–160 subcomplex consists of a β-propeller domain, an α-solenoid domain, or both ( Figure 1 ; Table 1 ). β-propellers contain several blades arranged radially around a central axis, each blade consisting of a four-stranded antiparallel β-sheet; α-solenoid domains are composed of numerous pairs of antiparallel α-helices stacked to form a solenoid ( Figure 1 ) ( Neer et al. 1994 ; Andrade et al. 2001a ; Andrade et al. 2001b ). While we have not defined the precise details of each domain, such as the exact shapes and numbers of propeller blades and solenoid repeats, the overall fold assignments for each nup are clear. These predictions indicate that yNup84, yNup85, and yNup145C all mainly consist of an α-solenoid domain, whereas yNup120 and yNup133 contain both an amino-terminal β-propeller and a large carboxyl-terminal α-solenoid region. Both ySec13 and ySeh1 are predicted to be almost entirely single-domain β-propellers of six and seven blades, respectively. These latter two proteins fall into the well-conserved class of tryptophan/aspartic acid (WD) repeat-containing β-propeller proteins. For both proteins, homology with the WD-repeat β-propellers has been reported previously ( Saxena et al. 1996 ; Siniossoglou et al. 1996 ; Yu et al. 2000 ) and is confirmed here. Figure 1 Ribbon Representation of Nup Models β-sheets (β-propellers) are colored cyan and α-helices (α-solenoids) are colored magenta. Gray dashed lines indicate regions that were not modeled. Arrowheads indicate the positions of high proteolytic susceptibility (see Figures 2 and 3 ). Table 1 Nup84 Subcomplex Proteins are Composed of Two Fold Types A list of the best scoring models for domains in the proteins of the Nup84 subcomplex in yeast. For Nup84, Nup85 and Nup145C, about 200 amino-terminal residues were not modeled. However, secondary structure predictions, hydropathy profiles, and threading of the yeast proteins and their homologs suggest that most of the unmodeled portion of these proteins also adopt the solenoid fold. For Nup120 and Nup133, we were unable to model, respectively, 133 and 299 amino-terminal residues. Secondary structure predictions suggest extensions or variations to the typical β-propeller and the α-solenoid folds a Percentage identity between the aligned sequence of the nup and its template b Z-score of the comparative model based on the alignment indicated by percentage identity (number of residues) ( Melo et al. 2002 ) (Tables S1–S6) We support our fold assignments using four considerations ( Figure 2 ; Tables 1 and S1–S7 ). First, both fold assignment programs returned their predictions with highly significant scores ( Tables S1–S7 ), and they predominantly assigned only the two predicted folds out of the approximately 1,000 different known fold types ( Tables S1–S7 ) ( Orengo et al. 1997 ). Moreover, while there are numerous variations corresponding to different proteins within each predicted fold type, the two different methods used for fold recognition often selected the same template proteins ( Tables S1–S7 ). Second, the evaluation of the atomic model for each nup was statistically significant when compared against the best models generated for random sequences of identical amino acid composition and length; all the nup models were at least six standard deviations away from the mean score of the random models ( Figure S1 ; Tables 1 and S1–S7 ) ( Melo et al. 2002 ). Third, secondary structure predictions from amino acid sequences alone indicate that all seven nups consist mainly of repetitive structures that largely match the secondary structures observed in their corresponding three-dimensional models ( Figure 3 and Figure S2 ). The agreement ranges from 58% to 87% of the residues for a three-state assignment (helix, strand, other). This agreement is the maximum possible level of consistency, given the approximately 75% accuracy of the secondary structure prediction methods ( Koh et al. 2003 ). Figure 2 Proteolytic Domain Map of the Yeast Nup84 Subcomplex Proteins Immunoblots of limited proteolysis digests for Protein A-tagged versions of each of the seven nups in the yNup84 subcomplex. Each protein is detected via its carboxyl-terminal tag; thus, all the fragments visualized are amino-terminal truncations (except for the full length proteins, which are indicated by arrowheads). The fragments of the Asp-N and Lys-C protease digests depicted in Figure 2 are labeled with letters (A, B, C…) that correspond to those in Table 2 , and the terminal Protein A fragments are labeled with an X (the Protein A tag is resistant to proteolysis). The sizes of marker proteins are indicated in kilodaltons (kDa) to the right of the gel. Figure 3 Predicted Secondary Structure Maps of the Nup84 Subcomplex Proteins Thin horizontal lines represent the primary sequence of each protein; secondary structure predictions are shown as columns above each line for β-strands (β-propellers; cyan) and α-helices (α-solenoids; magenta). The height of the columns is proportional to the confidence of the secondary structure prediction ( McGuffin et al. 2000 ). The modeled regions are indicated above each sequence by horizontal dark bars, corresponding to the models in Figure 1 . Proteolytic cleavage sites are identified by small, medium, and large arrows for weak, medium, and strong susceptibility sites, respectively. Where necessary, uncertainties in the precise cleavage positions are indicated above the arrows by horizontal bars. Table 2 Proteolytically Sensitive Sites of yNup84 Subcomplex Proteins Listed are the sites in each of the yNup84 complex proteins most sensitive to the two proteases, as shown in Figure 2A a Fragments labeled as in Figure 2 A b Molecular weight of carboxyl-terminal fragments, containing 26-kDa Protein A tag, was calculated on gel scans using NIH Image c The amino acid residues adjacent to the cleavage site are indicated, where “D” designates an aspartic acid and “K” designates a lysine d Amino acid residue positions indicated for full-length Nup145 Finally, we provide direct biochemical evidence in support of our fold assignments, using proteolytic mapping of domain boundaries and loop locations in the seven nups (see Figure 2 ). Tagged nups were purified from yeast extracts and incubated with the endoproteinases Asp-N (which hydrolyzes peptide bonds at the amino side of aspartic acid) or Lys-C (which hydrolyzes peptide bonds at the carboxylic side of lysines) while still attached to the magnetic beads via their proteolytically resistant tags. After digestion, proteolytic fragments that remained attached to the beads were separated by SDS-PAGE, and cleavage sites were determined either by molecular weight estimation of the fragments or by amino-terminal Edman sequencing ( Table 2 ). The regions predicted to form β-propellers were, as expected, extremely resistant to proteolysis (see Figure 2 ) ( Kirchhausen and Harrison 1984 ; Saxena et al. 1996 ). On the whole, the predicted α-solenoid regions were also resistant to proteolysis, although less so than the β-propellers. However, the major cleavages were found toward the end of the predicted α-solenoid domains, even in the most susceptible nup (yNup133). Strikingly, the strongest cleavages generally occurred in the border regions between the predicted domains, as is particularly evident for yNup133 and yNup120 ( Figure 3 ). Hence, in every case, the regions that we predicted to form compact folded structures were proteolytically resistant, and the predicted linkers between these domains were proteolytically sensitive. This correlation provides support for all seven of our structural models. In addition, circular dichroism and Fourier transform infrared spectra reported for Nup85 are in agreement with our predictions, indicating a composition characteristic of α-solenoids (approximately 50% α-helical, 23% loops, 5% turns, and 10% β-sheet) ( Hirano et al. 1990 ; Denning et al. 2003 ). We expect our findings will spur efforts to determine the detailed atomic structures of nups; the rapid proteolytic domain mapping and molecular modeling techniques we have utilized here should aid these efforts. Having established the domain folds for the yNup84 subcomplex, we also assigned domain folds in its vertebrate (i.e., human) and plant (i.e., Arabidopsis ) homologs. All seven nups from both human and Arabidopsis yielded identical domain fold assignments to their yeast counterparts ( Table S7 ), despite low primary sequence conservation among them ( Suntharalingam and Wente 2003 ). These findings indicate that the overall architecture of the yNup84/vNup107–160 subcomplex has been preserved throughout the eukaryotes. Hence, the yNup84/vNup107–160 subcomplex, which contributes nearly one-quarter of the mass of the NPC, is composed in the main of repetitive β-propellers and α-solenoids; taken together with other repetitive domain nups (such as the FG repeat nups), this suggests that a significant percentage of the NPC's bulk is composed of protein repeats ( Rout and Aitchison 2001 ; Suntharalingam and Wente 2003 ). To gain insight into the function and origin of the yNup84/vNup107–160 subcomplex, we asked whether there are other known subcomplexes that share similar compositions and fold arrangements. A search of the entire SwissProt/TrEMBL database for entries that contain an amino-terminal β-propeller followed by an α-solenoid revealed that this specific architectural combination is absent from both bacteria and archaebacteria, and is found only in eukaryotic proteins, whose role (where known) is as components either of coated vesicles or of the yNup84/vNup107–160 subcomplex. Thus, the clathrin heavy chain, a major component of clathrin-coated vesicles, appears remarkably similar in domain architecture ( ter Haar et al. 1998 ; Kirchhausen 2000b ) to both yNup120/vNup160 and yNup133/vNup133. All three proteins are composed of an amino-terminal β-propeller followed by an extended α-solenoid. Proteolysis of assembled clathrin cages leads to the release of an amino-terminal fragment of 52–59 kDa ( Kirchhausen and Harrison 1984 ). This result is similar to our domain mapping results, where the proteolysis of yNup120 and yNup133 resulted in amino-terminal fragments of 45 kDa and 60 kDa, respectively. Strikingly, one component of the yNup84/vNup107–160 subcomplex, ySec13/vSec13R, is also a known vesicle-coating protein. Similarly, ySeh1/vSec13L, a close homolog of ySec13/vSec13R, is also associated with both the yNup84/vNup107–160 subcomplex and the vesicle-coating proteins ( Siniossoglou et al. 1996 ; Kirchhausen 2000b ; Cronshaw et al. 2002 ; Gavin et al. 2002 ; Harel et al. 2003 ). Together, these results point to an intimate connection between vesicle-coating complexes and the yNup84/vNup107–160 subcomplex. In clathrin-coated vesicles, clathrin is attached via its amino-terminal domain to an adaptin complex. There are four types of adaptin complexes, all made of two large subunits that wrap around two small subunits. The bulk of each large subunit is made of an α-solenoid trunk ( Figure 4 ) ( Collins et al. 2002 ; Evans and Owen 2002 ). Similarly, the bulk of yNup84/vNup107, yNup85/vNup75, and yNup145C/vNup96 are also composed of α-solenoid trunks. Hence, the yNup84/vNup107–160 subcomplex resembles the clathrin/adaptin complex, in that the clathrin-like yNup120/vNup160 and yNup133/vNup133 are attached to the adaptin-like proteins yNup84/vNup107, yNup85/vNup75, and yNup145C/vNup96. This resemblance is further strengthened by our observation that the preferred templates for modeling the α-solenoid domains in the yNup84/vNup107–160 subcomplex were derived from proteins in vesicle coating complexes ( Figure S1 ; Tables S1–S7 ). Figure 4 The Nup84 Complex and Coated Vesicles Share a Common Architecture A diagram showing the organization of the clathrin/AP-2 coated vesicle complex is shown at left; the positions of clathrin and the adaptin AP-2 large subunits (α, β2 plus “ear” domains) and small subunits (σ, μ) are indicated. β-propeller regions are colored cyan, α-solenoid regions are colored magenta, and sample ribbon models for each fold are shown in the center. The variants of each fold that are found as domains in major components of the three kinds of vesicle-coating complexes and the yNup84 subcomplex are listed on the right. The -N and -C indicate amino-terminal and carboxyl-terminal domains, respectively. The classification of these domains is based on X-ray crystallography data (clathrin, α-adaptin, β2-adaptin [PDB codes 1gw5, 1bpo, 1b89 ( ter Haar et al. 1998 ; Collins et al. 2002 )]), by the detailed homology modeling presented here (yNup84 complex proteins; ySec13 also in Saxena et al. [1996] ), or by sequence homology or unpublished secondary structure prediction and preliminary analyses (COPI I (sec31) complex proteins [ Schledzewski et al. 1999 ], Sec31). Our analyses showed that the yNup84/vNup107–160 subcomplex and all three major classes of vesicle coating complexes can be linked together through their common architecture. As summarized in Figure 4 , these similarities include both previously reported relationships (e.g., between the clathrin/adaptin complexes and the COPI complexes) ( Schledzewski et al. 1999 ), and previously unsuspected relationships (e.g., between the COPII component Sec31 [ Salama et al. 1997 ; Shugrue et al. 1999 ; Belden and Barlowe 2001 ; Boehm and Bonifacino 2001 ; Lederkremer et al. 2001 ] and clathrin). The common architecture of the yNup84/vNup107–160 subcomplex and all three major classes of vesicle-coating complexes suggests that all of these complexes have common function in curving membranes. There is, in fact, circumstantial evidence for a role of the yNup84/vNup107–160 subcomplex in the establishment and maintenance of pore membrane curvature. Members of this complex, when disrupted in yeast, cause the uniformly distributed NPCs to cluster into patches in the plane of the NE ( Siniossoglou et al. 1996 ; Siniossoglou et al. 2000 ; Ryan and Wente 2002 ; Teixeira et al. 2002 ), suggesting that impairment of yNup84 subcomplex function results in a suboptimal interaction of the NPC with its surrounding nuclear membranes. Discussion As shown here, protein structure modeling is particularly useful in uncovering potential evolutionary and functional relationships that are refractory to classical approaches based on comparison of protein sequences alone. Our results show that clathrin/adaptin complexes, COPI complexes, COPII complexes, and the yNup84/vNup107–160 subcomplex all share a common molecular architecture. This commonality could have arisen by either convergent or divergent evolutionary pathways. In a convergent pathway, β-propeller and α-solenoid folds could have been independently utilized by both NPCs and vesicle-coating complexes at different stages of eukaryotic evolution. This possibility is supported by the high abundance of both fold types in eukaryotic genomes (which could potentially make their fusion in proteins or complexes relatively frequent) ( Yanai et al. 2002 ) and the low sequence similarities between proteins of the NPC and vesicle coating complexes (which may suggest that they are not related). In a divergent pathway, NPCs and vesicle-coating complexes share these folds because both complex types could have originated from a common ancestor. In this scenario, a single “protocoatomer” would have been the progenitor for numerous vesicle coating complexes, as well as the yNup84/vNup107–160 subcomplex. Several lines of evidence support this latter hypothesis. First, the most confident models of the yNup84/vNup107–160 subcomplex proteins are based on structures of coated vesicle proteins ( Figure S1 ; Tables S1–S7 ). Second, the particular arrangement of an amino-terminal β-propeller followed by an α-solenoid appears to be unique to components of either vesicle coating complexes or of the yNup84/vNup107–160 subcomplex ( Protocol S1 ). Third, the overall composition of both complex types is similar, being mainly composed of proteins containing comparable distributions of β-propellers and α-solenoids ( Figure 4 ). Fourth, both vesicle coating complexes and NPCs apparently share a common function: the bending and stabilizing of curved membranes. Fifth, the yNup84/vNup107–160 subcomplex actually contains bona fide vesicle coat components, Sec13 and Seh1. In light of these considerations, we favor the “protocoatomer” hypothesis, in which the NPCs and vesicle-coating complexes arose by a process of divergent evolution. The lack of detectable sequence similarity between the proteins in the yNup84/vNup107–160 subcomplex and the coated vesicles is not surprising. Sequence comparisons of α-solenoid- and β-propeller-containing proteins suggest that these folds arose just before or around the time of the origin of eukaryotes, then rapidly duplicated and diversified ( Cingolani et al. 1999 ; Smith et al. 1999 ; Andrade et al. 2001b ). Both folds consist of repetitive structures, so the functional constraints on an individual repeat are weak, compared with the whole fold domain. It has been proposed that the robustness of these folds with respect to changes in their sequences permits their component repeats to individually lose their sequence similarity, eventually allowing the proteins they comprise to drift into new functions ( Malik et al. 1997 ; Smith et al. 1999 ; Andrade et al. 2001a ; Andrade et al. 2001b ). Moreover, the lack of detectable sequence similarity for members of the same fold family is not necessarily an indicator of convergent evolution; obvious sequence similarities are often lost during long periods of evolution (e.g., FtsZ and tubulin or MreB and actin [ Amos et al. 2004 ]). The divergent pathway is also consistent with the conservation among members of the syntaxin family (key components of the vesicular transport machinery), which points to a similar early origin and rapid diversification of the eukaryotic endomembrane system ( Dacks and Doolittle 2002 ; Dacks and Field 2004 ). Based on these observations, we propose a single evolutionary origin for the structures maintaining both the endomembrane systems and the nucleus ( Figure 5 ) over models suggesting separate or even endosymbiotic origins for these structures. Figure 5 Proposed Model for the Evolution of Coated Vesicles and Nuclear Pore Complexes Early eukaryotes (left) acquired a membrane-curving protein module (purple) that allowed them to mold their plasma membrane into internal compartments and structures. Modern eukaryotes have diversified this membrane-curving module into many specialized functions (right), such as endocytosis (orange), ER and Golgi transport (green and brown), and NPC formation (blue). This module (pink) has been retained in both NPCs (right bottom) and coated vesicles (left bottom), as it is needed to stabilize curved membranes in both cases. The current protocoatomer hypothesis posits that a simple coating module containing minimal copies of the two conserved folds evolved in protoeukaryotes as a mechanism to bend membranes into sharply curved sheets and invaginated tubules ( Figure 5 ). The ability to so manipulate cell membranes represented a major evolutionary innovation that allowed, among other possibilities, the elaboration of internal membranes, phagotrophy, and endosymbiosis ( Maynard Smith and Szathmâary 1997 ); the importance of this ability is underscored by the presence of numerous types of membrane-curving devices in modern eukaryotes. As with clathrin, the flexibility of the α-solenoid in this simple module enabled the formation of curved membranes of various sizes. In addition, the α-solenoid repeat structure, together with the repeats in the β-propeller fold, provided the coating module with a large binding area. These features allowed the membrane-curving module to polymerize and form a coat, as well as to interact with other membrane-associated proteins. The endomembranes and their membrane-coating modules subsequently evolved to become more elaborate and specialized, with the partitioning of different functions into separate, interconnected compartments such as the ER, the Golgi, and the nucleus ( Figure 5 ), each with their own specialized set of coating modules. In conclusion, we suggest that the progenitor of the NPC arose from a membrane-coating module that wrapped extensions of an early ER around the cell's chromatin. In this primitive NE, the coating modules would have originally formed the sharply curved membrane, creating large and freely permeable pores ( Figure 5 ). These pores then closed to form the selectively permeable NPCs of modern eukaryotes ( Rout et al. 2003 ). In doing so, they retained at their core a coating module as a relic of their evolutionary origins. This module, the yNup84/vNup107–160 subcomplex, may still serve to curve and stabilize the nuclear pore membrane in modern eukaryotes; as such, it would function as a key scaffold to form the NPCs, the portals of the nucleus. Our findings could thus provide an explanation for the origin of the nuclear pore complex (which until now has been a mystery) and may fill a significant gap in our understanding of the evolution of eukaryotes. Materials and Methods Only two domains in the seven nups have their folds assigned by sequence comparison to proteins of known structure ( Saxena et al. 1996 ; Siniossoglou et al. 1996 ). Therefore, to assign folds for as many target domains comprising the yNup84/vNup107–160 subcomplex as possible, we applied a structure-based approach consisting of iterative detection of potential template structures, their alignment to the target sequence, model building, and model assessment ( Marti-Renom et al. 2000 ). Secondary structure was predicted from sequence by the PredictProtein ( Rost 1996 ) and PSI-Pred ( McGuffin et al. 2000 ) servers. Detection of potential template structures For each of the seven yeast nups and representative homologs, potentially related known structures were detected by the mGenThreader ( McGuffin and Jones 2003 ) and FUGUE ( Shi et al. 2001 ) web servers ( Tables S1–S7 ). Several other servers gave similar results (unpublished data). To find out whether or not mGenThreader frequently identifies the β-propeller and α-solenoid folds as false positives, we randomly selected 20 sequences of known structure from each one of the structural classes and submitted them to mGenThreader. Using the same parameters as in our analysis of the nups, only two of these 140 sequences were incorrectly predicted to contain β-propeller or α-solenoid folds (unpublished data). Thus, we estimate the false positives rate for the nup fold assignments based on mGenThreader alone to be approximately 1%–2%. Alignment of the matched target-template pairs The matches obtained in the previous step provided an operational definition of a domain. They were either accepted or refined by manual and automated alignment. Manual realignment relied on sequence conservation and secondary structure predictions by PROF ( Rost 1996 ) and PSI-PRED ( McGuffin et al. 2000 ). The automatic realignments were obtained by SALIGN ( Marti-Renom et al. 2004 ) and T-Coffee ( Notredame et al. 2000 ). In the last iteration, the alignments and the models were refined by MOULDER, a genetic algorithm method for iterative alignment, model building, and model assessment ( John and Sali 2003 ). Model building For each alignment, an all-atom model was built by comparative modeling based on satisfaction of spatial restraints as implemented in MODELLER ( Sali and Blundell 1993 ). Model assessment. The fold assignment, alignment, and model building were repeated by varying the domain boundaries, target sequences for modeling, template structures, and their alignments. The aim was to improve model assessment by statistical potentials of ProsaII ( Sippl 1993 ) and DFIRE ( Zhou and Zhou 2002 ), and by a composite model evaluation criterion ( Melo et al. 2002 ; John and Sali 2003 ). The only importance of explicit model building in this analysis was to provide another semi-independent way to validate the fold assignments: If a model was assessed to have the correct fold, the initial fold assignment must have been correct. Beyond that, the models were not used. Domain combination search. To search for proteins that resemble the domain architecture of clathrin, we queried MODBASE ( Pieper et al. 2004 ), our relational database of annotated comparative protein structure models, and Superfamily ( Gough et al. 2001 ), a database of HMM-based structural assignments. Both databases assign folds to all available protein sequences that match at least one known protein structure. We first searched for any protein sequences that were matched to both β-propeller and α-solenoid structures. We used the broadest definitions of the β-propeller folds (b.66, b.67, b.68, b.69, b.70, for 4-, 5-, 6-, 7- and 8-bladed β-propellers, respectively) and α-solenoid folds (a.118) from the SCOP database (v1.65) ( Lo Conte et al. 2002 ). In MODBASE, we found 95 proteins predicted to contain both β-propeller and α-solenoid domains ( Protocol S1 ). Of these 95 proteins, 37 passed the following filters, ensuring clathrin-like characteristics: they had to be 800–2,000 residues long, the amino-terminal β-propeller domain had to be followed by a carboxyl-terminal α-solenoid domain, the β-propeller and α-solenoid domains each had to span at least 35% of the total length, and no other domain could be more than 25% of the total length. All of the 37 proteins were from eukaryotes. Their functions were assigned either as clathrin or unknown in the Swiss-Prot/TrEMBL database ( O'Donovan et al. 2002 ). Similar results were obtained by querying the Superfamily database ( Gough et al. 2001 ). Proteolytic domain laddering. Magnetic beads (2.8 μm Dynabeads M-270 Epoxy [#143.02; Dynal, Oslo, Norway]) were conjugated to rabbit IgG (#55944; ICN Biochemicals, Costa Mesa, California, United States) according to the manufacturer's instructions. Yeast cells carrying PrA-tagged versions of nups were grown and harvested as described previously ( Rout et al. 2000 ). Cell pellets were frozen in liquid nitrogen and homogenized to a fine powder in a motorized grinder (#RM100; Retsch, Haan, Germany) continuously cooled with liquid nitrogen. The cell powder was thawed on ice and ten volumes of extraction buffer (20 mM HEPES [pH 7.4], 1.0% Triton X-100, 0.5% sodium deoxycholate, 0.3% sodium N-lauroyl-sarcosine, 0.1 mM MgCl 2 , 1 mM DTT, 1:500 protease inhibitor cocktail [#P-8340; Sigma, St. Louis, Missouri, United States]) were added to cells and homogenized at 4 °C with a Polytron (Kinematica, Littau-Luzerne, Switzerland). The cell lysate was clarified by centrifugation (2,000 g for 5 min at 4 °C). The magnetic beads were added to the extract to a ratio of about 8 × 10 9 beads per g of cells. After incubation for 1 h at 4 °C, the beads were magnetically recovered. The beads were washed, resuspended in 50 μl of reaction buffer (according to the manufacturer's specifications), and Asp-N (#1420488; Roche, Basel, Switzerland) or Lys-C (#1420429; Roche) proteinase was added to give a weight ratio of 1:200 of proteinase to the tagged nup. After incubation at different time points at 37 °C, bead aliquots were removed and washed, and tagged fragments were eluted with 0.5 M NH 4 OH containing 0.5 mM EDTA. The eluant was vacuum-dried, resuspended in SDS-PAGE sample buffer, and separated on a 4%–12% bis-Tris gel (Invitrogen, Carlsbad, California, United States). Proteins were then either transferred electrophoretically to nitrocellulose or PVDF and probed with HRP-rabbit IgG (#011–0303-003; Jackson ImmunoResearch, West Grove, Pennsylvania, United States), or analyzed by amino-terminal Edman sequencing ( Fernandez et al. 1994 ). Supporting Information Figure S1 Model Score Versus Length The graphs plot the assessment score of the model (Melo Z-score) ( Melo et al. 2002 ) versus the model size, for the "non-MOULDER" models in Tables S2–S6 . The red circles indicate the entries in Table 1 in the main text of the paper. Because the Z-score depends on the number of residues in the model, the smallest model with the highest Z-score was considered most significant. (87 KB DOC). Click here for additional data file. Figure S2 Agreement between Predicted and Modeled Secondary Structure The secondary structure predicted from sequence by PROF ( Rost and Liu 2003 ) and PSI-Pred ( McGuffin et al. 2003 ) is compared to the secondary structure observed in the three-dimensional models presented in Table S1 (“…” represents regions that are not modeled). The numbers above the predicted secondary structures correspond to the confidence score returned by the servers. Current secondary structure prediction methods based on multiple alignments correctly predict the secondary structure state for 70%–80% of residues (in a three-state prediction) ( Eyrich et al. 2001 ). Since the random prediction would predict only about 30% of the residues correctly, the fact that our predictions match the assignments at 58%–87% level is highly suggestive, supporting our fold assignments. A representative example, Nup85, is shown here. For the visualization of all the Nups, see the additional information web page ( http://salilab.org/damien/NPC/ ). (47 KB DOC). Click here for additional data file. Protocol S1 List of Proteins Modeled as β-Propeller and α-Solenoid Domains in ModBase (42 KB DOC). Click here for additional data file. Table S1 Modeling Results for Yeast Nup84 Complex Proteins I (yNup133) (491 KB DOC). Click here for additional data file. Table S2 Modeling Results for Yeast Nup84 Complex Proteins II (yNup133) (101 KB DOC). Click here for additional data file. Table S3 Modeling Results for Yeast Nup84 Complex Proteins III (yNup133) (115 KB DOC). Click here for additional data file. Table S4 Modeling Results for Yeast Nup84 Complex Proteins IV (yNup133) (132 KB DOC). Click here for additional data file. Table S5 Modeling Results for Yeast Nup84 Complex Proteins V (yNup133) (124 KB DOC). Click here for additional data file. Table S6 Modeling Results for Yeast Nup84 Complex Proteins (yNup133) (93 KB DOC). Click here for additional data file. Table S7 Modeling Results for Human and Plant Nup84 Complex Proteins (yNup133) (144 KB DOC). Click here for additional data file. Accession Numbers Uniprot ( Apweiler et al. 2004 ) accession numbers ( http://www.pir.uniprot.org ) for proteins discussed in this paper are as follows. Yeast: ySeh1 (P53011), ySec13 (Q04491), yNup84 (P52891), yNup85 (P46673), yNup120 (P35729), yNup133 (P36161), and yNup145C (P49687). Human: vSec13 l (Q96EE3), vSec13R (P55735), vNup107 (P57740), vNup75 (Q9BW27), vNup160 (Q12769), vNup133 (Q8WUM0), and vNup96 (P52948).	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC524472.xml
539351	Crystal structure of subunit VPS25 of the endosomal trafficking complex ESCRT-II	Background Down-regulation of plasma membrane receptors via the endocytic pathway involves their monoubiquitylation, transport to endosomal membranes and eventual sorting into multi vesicular bodies (MVB) destined for lysosomal degradation. Successive assemblies of E ndosomal S orting C omplexes R equired for T ransport (ESCRT-I, -II and III) largely mediate sorting of plasma membrane receptors at endosomal membranes, the formation of multivesicular bodies and their release into the endosomal lumen. In addition, the human ESCRT-II has been shown to form a complex with RNA polymerase II elongation factor ELL in order to exert transcriptional control activity. Results Here we report the crystal structure of Vps25 at 3.1 Å resolution. Vps25 crystallizes in a dimeric form and each monomer is composed of two winged helix domains arranged in tandem. Structural comparisons detect no conformational changes between unliganded Vps25 and Vps25 within the ESCRT-II complex composed of two Vps25 copies and one copy each of Vps22 and Vps36 [ 1 , 2 ]. Conclusions Our structural analyses present a framework for studying Vps25 interactions with ESCRT-I and ESCRT-III partners. Winged helix domain containing proteins have been implicated in nucleic acid binding and it remains to be determined whether Vps25 has a similar activity which might play a role in the proposed transcriptional control exerted by Vps25 and/or the whole ESCRT-II complex.	Background Endosomal compartments receive membrane bound cargo from both the biosynthetic and the endocytic pathways. Receptor downregulation by endocytosis includes transport to early endosomes and either recycling or sorting into late endosomes. The latter have the morphological characteristics of multivesicular bodies (MVB) [ 3 ] that can undergo homotypic fusion or heterotypic fusion with lysosomes, which deliver MVB cargo for proteolytic degradation [ 4 ]. In addition to receptor downregulation, MVB formation has been implicated in antigen presentation [ 5 ] and in the release of enveloped viruses [ 6 , 7 ]. Gene deletion and inactivation studies in yeast have identified 17 proteins that directly affect MVB formation (yeast class E compartment) by resulting in aberrant endosomal/vacuolar morphology [ 4 ]. All proteins are required for vacuolar protein sorting (VPS) into the class E compartment and are recruited to endosomal membranes from the cytosol in order to assemble into three ESCRT ( E ndosomal S orting C omplexes R equited for T ransport ) complexes that function in MVB formation [ 8 - 11 ]. Receptor mono-ubiquitinylation has been shown to serve as a signal to enter the MVB pathway [ 12 ]. Initial recognition of ubiquitinated cargo by Vps27 recruits the ubiquitin binding protein Vps23 [ 11 , 13 ], which in turn leads to the assembly of the multi-protein complex ESCRT-I (VPS23, VPS28, and VPS37) [ 10 ]. ESCRT-I subsequently recruits ESCRT-II, composed of Vps22, Vps25, and Vps36, which in turn activates ESCRT-III subcomplexes [ 8 , 9 ]. Assembly of ESCRT-III at the endosome initiates the sorting and concentration of ubiquitinated cargo; ubiquitin is removed and Vps4, an AAA-type ATPase, dissociates the ESCRT complexes concomitantly with membrane invagination and budding of vesicles into the lumen of the endosome [ 4 ]. Two recent crystal structures of a core of the ESCRT-II complex reveal a trilobal complex, containing two copies of Vps25, one copy of Vps22 and the C-terminal region of Vps36. Each subunit is composed of two winged helix domains and an N-terminal region of Vps25 interacts with Vps22 and Vps36 [ 1 , 2 ]. Although ESCRT-II is essential for the MVB pathway, since cells missing ESCRT-II components fail to localize ESCRT-III to late endosomes [ 8 , 9 ] the complex has also been found "moonlighting" in the nucleus. The human and rat homologues of ESCRT-II were originally identified as the EAP complex ( ELL Associating Protein ; Vps22/EAP30; Vps25/EAP20; Vps36/EAP45), associated with the RNA polymerase II elongation factor ELL in the nucleus [ 14 , 15 ]. Consistent with a role in transcriptional control, yeast Vps22 (or SNF8) as well as Vps25 and Vps36 have been implicated in glucose-dependent gene expression control [ 15 , 16 ]. To date, it is not clear whether the role of ESCRT-II in MVB formation is independent of its function as a transcriptional activator or whether both processes are linked. Here, we report the crystal structure of full-length yeast Vps25, composed of two homologous winged-helix domains. Results and discussion Structure of Vps25 The structure of Vps25 was solved by single wavelength anomalous diffraction (SAD) using selenomethionine-derivatized crystals. Vps25 consists of two homologous winged helix domains as detected by the program GRATH that are arranged in tandem (Figure 1A ). Winged helix folds are compact alpha/beta structures with secondary structure elements arranged in a typical order (H1-S1-H2-H3-S2-W1-S3- W2optional ) [ 17 ], which fold into a mostly helical part followed by a twisted anti-parallel beta-sheet and two large loops (wings, W). The fold of Vps25 deviates slightly from the canonical fold. The N-terminal domain 1 (residues 1 to 126) contains two additional N-terminal 3/10 helices, implicated in the interaction with either Vps22 or Vps36 [ 1 , 2 ], followed by the canonical helix 1 and strand 1. It lacks canonical helix 2, which instead folds into a large disordered loop followed by strands 3 and 4 that connects to helix 2 (at the corresponding position of canonical helix 3). Strands 5 and 6 then form, together with strand 1, a twisted anti-parallel beta-sheet with wing W1 protruding from the structure (Figure 1A and Figure 2 ). Domain 1 also lacks wing W2, as in the cases of winged helix domain containing transcription factors E2F4 and DP2 [ 18 ]. Strand 6 flows directly into domain 2, which also has a canonical winged helix fold except for the absence of wing W2 (Figure 1A and Figure 2 ). Domains 1 and 2 are tightly packed against each other and their C alpha atoms can be superimposed with an r.m.s. deviation of 3.4 Å (Figure 1B ), confirming their structural relatedness. The domain interface is dominated by van der Waals contacts including conserved and non conserved residues Trp44, Phe122, Leu104, Leu124, Trp125 in domain 1 and Leu128, Trp131, Met168, Pro169 and Leu172 in domain 2 (Figure 2 ). Figure 1 Vps25 contains two winged helix domains arranged in tandem. (A) Ribbon diagram of Vps25; the two domains are shown in orange and yellow. Secondary structure elements are labeled. The major missing loop region connecting strands 1 and 3 is indicated by a dashed line. (B) Superposition of the Calpha positions of the N- and C-terminal domains (residues 23 to 48 and 85 to 101 with corresponding C-terminal domain residues; r.m.s.d. 3.4 Å). Note that the positions of helices 1/3 and helices 2/5 as well as wing positions W1 match up well. Figure 2 Structure based sequence alignment of Vps25. Sequences aligned using S. cerevisiae Vps25 (gene bank #CAA89632) and Vps25 orthologues from H. sapiens (#BE386260), D. melanogaster (#AAF59066) and from C. elegans (#T26073). Identical residues are shown on red background, similar residues are drawn in red and sequence similarity is underlined by blue boxes. Secondary structure elements are shown. Disordered regions in the Vps25 structure are indicated by dashed lines. Structural comparision of unliganded Vps25 and Vps25 in complex with Vps22 and Vps36 (ESCRT-II) Two recent crystal structures of the ESCRT-II core reveal trilobal structures with head to tail interactions of one copy of Vps25 with Vps22 and the other copy of Vps25 with Vps36 at the center. In both cases a conserved proline rich N-terminal region of Vps25 (Figure 2 ) together with conserved Arg83 mediate key interactions [ 1 , 2 ]. Therefore it was of interest to analyse whether Vps25 undergoes any conformational changes upon participation in ESCRT-II complex formation. Superposition of the C alpha atoms with one copy of Vps25 from either ESCRT-II complex structure ([ 1 , 2 ]; pdb entries 1U5T and 1W7P) revealed r. m. s. displacements of 1.2/1.2 Å (residues 3 to 51), 1.5/1.7 Å (residues 74 to 155) and 2.3/2.9 Å (residues 159–199) respectively. The major changes are confined to both wings W1 and W2 indicating their conformational flexibility (Figure 3 ). In contrast, the conserved N-terminal segment, which is implicated in Vps22 and Vps36 interactions shows no substantial changes (Figure 3 ). Figure 3 Comparison of unliganded and liganded Vps25. Superposition of unliganded Vps25 (red) with Vps25 from both ESCRT-II structures [1, 2] (blue, pdb code 1U5T chain C; green, pdb code 1W7P chain B). The peptide backbones are shown as coils. Vps25 is shown in the same orientation as in figure 1A. The position of Arg83 is indicated by an arrow. In the unliganded Vps25 structure, this helical segment constitutes the 1192 Å 2 dimerization interface of two identical Vps25 dimers present in the asymmetric crystal unit. The dimer contact is mediated by hydrophobic residues Pro5, Pro6, Val7, Phe10, Pro11, and Pro12, which is similar to the contact region described for Vps25 interactions with Vps22 and Vps36 [ 1 , 2 ]. In the Vps25 structure Arg83 does not participate in dimerization but hydrogen bonds to Thr15 instead of forming salt bridges with either Vps36 Asp548 or Vps22 Asp214 as observed in the ESCRT-II complex [ 1 , 2 ]. Arg83 locates to a beta hairpin (strand 4; Figure 2 ) in the unliganded form of Vps25. Although the position of Arg83 is unchanged in all Vps25 structures (Figure 3 ) the position of the preceding loop region varies which might be due to differences in secondary structure assignment [ 1 , 2 ]. Therefore Vps25 seems to dock as a rigid body onto either Vps22 or Vps36 upon ESCRT-II complex formation. Although we do not detect Vps25 dimer formation in vitro , a dimeric form of Vps25 might be stabilized through other unknown interactions. Structural homology of Vps25 with nucleic acid binding winged helix domains Analysis of the full-length structure with DALI [ 19 ] revealed seven structural homologues displaying nucleic acid binding winged helix domains with a Z score above 5 for Vps25 domain 1. The top two hits were the selenocysteine-specific elongation factor fragment (PDB 1lva, Z score 6) and double-stranded RNA specific adenosine deaminase (ADAR) Z-alpha domain (PDB 1qbj, Z score 5.5). Winged helix family members interact with nucleic acids mostly via the "specificity helix" that binds to the major groove of the DNA with two flanking loops contributing to DNA interactions [ 17 ]. Superposition of Vps25 domain 1 onto the winged helix domain of E2F-4 bound to DNA [ 18 ] matching the "specificity helices" (Vps25 helix H2) revealed a potential fit with only minor clashes at the helix H1 loop region (data not shown). A potential nucleic acid interaction of Vps25 might be interesting in light of the described role of Vps25 and the other ESCRT-II subunits in glucose-dependent gene regulation [ 15 , 16 ] and complex formation with RNA polymerase II elongation factor ELL [ 14 , 15 ], although no biochemical data exist so far to support such a proposed function. Vps25 participates in protein complex formation The ESCRT-II complex assembles at the endosomal membrane downstream of ESCRT-I and recruits ESCRT-III subcomplexes [ 8 - 10 ]. Consistent with such a sequential assembly, further ESCRT-II interactions of Vps25 have been described, namely with Vps28 (ESCRT-I) and with Vps20 (CHMP6; ESCRT-III) [ 7 , 20 ]. Surface electrostatic potential maps of Vps25 reveal a negatively charged surface within domain 2 that is characterized by a patch of conserved residues such as Glu153, Glu170 and Tyr152 (Figure 4A and Figure 2 ). Tyr152 is also part of the highly conserved domain 2, helix 4 (Figure 2 ). Domain 2 is the outer domain of Vps25 in the ESCRT-II complex and this region would thus be freely accessible for potential interaction(s) with Vps28 or Vps20. Similarly, basic residues (Lys99 and Arg23) potentially implicated in nucleic acid recognition are part of a conserved patch on domain 1 (Figures 4B and 2 ). Figure 4 Surface charge distribution of Vps25. (A) Surface potential representation of Vps25 with regions where electrostatic potential <-10 k B T are red, while those >+10 k B T are blue (k B , Boltzmann constant; T, absolute temperature). (B) Horizontal rotation (180°). Exposed residues are labeled for orientation. Note that one face of the molecule carries a mainly negative charge (A) while the other one carries a mainly positive charge (B). Vps25 contains additional features, which are unique to S. cerevisiae , as evidenced from multiple sequence analysis [ 15 , 16 ]. Vps25 orthologues have a shorter strand 2 to strand 3 connection (19 residues), whose sequence is composed of mostly charged residues and is disordered in our structure as well as in the ESCRT-II structures [ 1 , 2 ]. Furthermore, domain 1 wing W1 is shorter (7 residues) (Figure 2 ), which might indicate S. cerevisiae unique protein-protein interaction sites. Conclusions Clear evidence suggests that ESCRT-II recruitment is involved in MVB formation leading to plasma membrane receptor downregulation [ 4 ]. On the other hand ESCRT-II seems to play a role in transcription regulation [ 15 ]. Similarly, other ESCRT components such as Tsg101 ( T umor s usceptibility g ene ; Vps23; ESCRT-I) and members of the CHMP protein family (ESCRT-III; Ch romatin M odifying P rotein ; Ch arged M ultivesicular body P rotein ) are also found to act in the nucleus as well as in the cytosol and at endosomal membranes [ 21 - 23 ]. Interestingly, both Vps25 and Vps36 have been implicated in regulating stress and pheromone response pathways [ 24 ] and pheromone receptor Ste2 is downregulated via the endosomal pathway [ 12 ]. Similarly, SNF8 (Vps 22; EAP30), Vps36 and Vps25 are all directly involved in derepression of glucose-repressed genes, which might be linked to sorting of sucrose receptors via the endosomal pathway [ 15 , 25 ]. Protein sorting into MVB involves monoubiquitylation of cargo, which is recognized by ESCRT members. ESCRT-II Vps36 contains an ubiquitin binding NZF zinc finger motif that is necessary for protein sorting [ 26 ]. Therefore, ESCRT-II complexes may sense the turnover of specific ubiquitylated receptors at the endosomal membrane together with other unknown signals. As ESCRT-II only transiently associates with endosomal membranes [ 9 ] a signal within the MVB process might induce nuclear localization of ESCRT-II, where it could stimulate gene expression leading to up or down regulation of specific membrane receptors. Methods Protein expression, purification and crystallization Full length yeast Vps25 DNA (gene bank #CAA89632) was cloned into expression vector pETM30 (EMBL, Protein Expression Facility) and the Vps25 GST fusion protein was expressed in E. coli BL21 codon+ cells. For purification, cell pellets from 6 liter cultures were lysed in 150 mls of buffer A (50 mM Tris-HCl, pH 8.5, 200 mM NaCl, 0.2 mM DNaseI, 2 mM β-ME, 2 complete EDTA-free protease inhibitor tablets (Pierce)) and 0.1 mg/ml lysozyme for one hour on ice. The cell lysate was cleared by centrifugation and loaded onto a GST-sepharose (Pharmacia) column. The column was extensively washed with buffer B (50 mM Tris pH 8.5, 200 mM NaCl) and Vps25 fusion protein was eluted with buffer B containing 5 mM reduced glutathione. GST was then removed by TEV cleavage (w/w; 1:200) at 4°C overnight. His-tagged GST and TEV were subsequently both removed on a Ni 2+ chelating sepharose column. Vps25 was further purified on a superdex75 column (Pharmacia) in buffer C (50 mM Tris 8.5, 200 mM NaCl, 2 mM βME). Selenomethione-labeled Vps25 was produced using standard procedures and purified as described above. Crystallization conditions for Vps25 (7 mg/ml) were first determined by screening 600 conditions using a Cartesian crystallization robot. Initial conditions were refined using the hanging drop method, and the final crystallization condition (100 mM Na cacodylate pH 6.5, 200 mM Mg or Ca acetate, 5–7% glycerol, and 15–18% polyethylene glycol 8000) produced rectangular- and wedge-shaped selenomethionine-labeled Vps25 crystals in the same drop. Native Vps25 crystallized initially only with rectangular morphology and wedge-shaped crystals were produced by microseeding with the original SeMet crystals. For cryogenic data collection, the crystals were equilibrated in 25% glycerol and flash cooled in a gaseous nitrogen stream at 100 K. Crystallization produced rectangular crystals that belong to space group P422 with unit cell dimensions a = b = 78 Å, c = 54 Å and diffract to 3.2 Å resolution. However, all data sets collected from these crystals proved to be almost perfectly merohedrally twinned. The second crystal form, wedge-shaped, belonged to space group P2 1 2 1 2 1 with unit cell dimensions as indicated (table 1 ), contained 4 molecules per asymmetric unit, diffracted X-rays to 3.1 Å resolution and was used for structure solution. Table 1 Data Collection and Refinement. Crystal VPS25-SEMET VPS25-NATIVE Space Group P2 1 2 1 2 1 P2 1 2 1 2 1 Wavelength 0.97914 0.931 Unit Cell (Å) a 53.44 53.36 b 124.11 123.66 c 139.48 140.30 Resolution (outer shell) (Å) 100-3.20 (3.31-3.20) 100 - 3.10 (3.21-3.10) Total Reflections (outer shell) 111407 (10016) 65477 (6064) Unique Reflections 29230 16330 Completeness (%) (outer shell) 98.9 (93.1) 93.1 (91.3) Rmerge (outer shell) 0.090 (0.335) 0.053 (0.307) Average I/sigma (outer shell) 12.3 (4.9) 20.9 (4.9) Phasing Number of Se Sites 14 SOLVE FOM 0.351 RESOLVE FOM (ncs) 0.694 Refinement Resolution (outer shell) (Å) 25.0-3.10 (3.29-3.10) Number of reflections (test set) 16404 (790) R factor 0.275 Free R factor 0.327 Number of protein/solvent atoms 5760/16 Average B factor (Å 2 ) 51.3 Rms deviation bond lengths (Å) 0.009 Ramachandron Mol A Mol B MolC MolD Most favored Additionally favored 83.8 15.7 79.5 20.5 82.1 17.9 67.2 32.8 Data Collection Native data for Vps25 were collected at the European Synchrotron Radiation Facility (ESRF) beamline ID14-EH3 and data from SeMet-labeled crystals were collected at the ESRF beam line ID29 at three wavelengths (table 1 ). Data were processed and scaled with XDS [ 27 ]. Phasing and refinement Significant radiation damage had occurred for data collected at the inflection and remote wavelengths, therefore only data collected at the peak wavelength (table 1 ) were used for SAD phasing. ShelXD [ 28 ] was used to find 14 out of 16 selenium sites, which were further refined with SOLVE [ 29 ]. Four-fold non-crystallographic symmetry was imposed on the sites in addition to solvent flattening with RESOLVE [ 30 ]. Phasing statistics are listed in table 1 . The initial model was built with O [ 31 ] guided by the SeMet positions and clear tryptophan (7 per mol) and tyrosine (8 per mol) densities followed by refinement with CNS [ 32 ]. Strict four-fold NCS and phases were initially kept throughout the initial chain-tracing and refinement. During model building it was observed that molecules A and B and molecules C and D are arranged in the same dimer configuration and strict NCS was changed to restrained NCS during refinement. The packing also indicated tight interactions between molecules A, B, and C while molecule D showed only very few crystal contacts yet formed the "bridge" between two-dimensional layers formed by molecules A, B and C. The electron density maps for molecules A, B, and C were clear and well defined, while electron density for molecule D was poorly defined for side chains and loops. The model was improved by alternating cycles of model building and conjugate gradient minimization and restrained individual B-factor refinement using CNS [ 32 ]. The final coordinates were refined against the native dataset (30 to 3.1 Å) using the MLHL maximum likelihood target with the RESOLVE phases as constraint and retaining the original test set reflections. In the final stage of refinement, a maximum likelihood target and model phases alone were used. The final model lacks two to five flexible loops (molecule mol A, residues 56–72, 114–115, 156–157; mol B, residues 53–73, 157–158; mol C, residues 57–72, 155–158; mol D, residues 19–21, 55–73, 107–120, 156–160, 185–186). Accordingly, mol D is poorly defined (43 residues missing out of 204). The final R factor and R free (0.275/0.327) reflect missing residues and the poor model for molecule D. The model exhibits otherwise overall good stereochemistry with no outliers in the Ramachandran plot as defined in PROCHECK (table 1 ) [ 33 ]. The coordinates have been deposited in the RCSB Protein Data Bank accession code 1XB4 [PDB:1XB4]. Structure analysis Figures were generated using coordinates of molecule C with programs MOLSCRIPT [ 34 ], Raster 3D [ 35 ], ESPript [ 36 ], GRASP [ 37 ] and PyMOL . Sequences were aligned using Clustalx [ 38 ]. Secondary structure elements were assigned using the program DSSP [ 39 ]. The buried surface was calculated with CNS [ 32 ] and the program LSQMAN was used for superpositioning of C-alpha positions [ 40 ]. Authors' contributions W.W. conceived of the study, and participated in its design, coordination and writing of the manuscript. W.W. expressed, purified and established initial crystallization conditions and participated in data collection. A.K.W. carried out data collection, structure solution and refinement and participated in writing of the manuscript. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC539351.xml
535543	Psychometric performance of an assessment scale for strain in nursing care: The M-NCAS	Background Multiple instruments exist to measure dementia behaviors, but the nursing staff perspective on those behaviors and their level of burden has not been well measured. The goal of this study was to examine the psychometric performance of the Modified Nursing Care Assessment Scale (M-NCAS), a 28-item nurse rating of burden associated with care for institutionalized individuals with dementia. Nurses rate items in terms of extent to which the behavior or characteristic is present ("attitude" domain), and extent to which it is a burden ("strain" domain). Methods Data from 282 patients enrolled in a 12-week, double-blind, randomized clinical trial comparing risperidone treatment to placebo was used to evaluate M-NCAS item performance, internal consistency reliability, and construct validity. Empirical subscales were identified via exploratory factor analysis (EFA). Results Four poorly-performing items were deleted from further analyses. EFA identified 3 "attitude" subscales and 5 "strain" subscales. Cronbach's alphas were 0.65 and above. Correlation with the Cohen-Mansfield Agitation Inventory and the BEHAVE-AD, clinical ratings of dementia behaviors, were low to moderate. Conclusion The M-NCAS provides a valid and reliable means of obtaining care burden ratings from formal caregivers in long-term care, and provides a method for evaluating dementia interventions from the perspective of nursing staff.	Background Caring for the institutionalized dementia patient can be challenging for nursing staff [ 1 , 2 ]. Persons with dementia often exhibit such disturbing behaviors as pacing, use of inappropriate language, acting out, refusal of necessary care, hallucinations, and delusions. Nursing staff caring for persons with dementia may have difficulty forming relationships with their patients, and the stress associated with such care may contribute to staff turnover. Nursing staff turnover in nursing homes is high, ranging from 40%–96% annually in the United States [ 3 ]. The societal costs of nursing staff turnover (particularly in the current environment of nursing staff shortages) are great, and can be expected to increase as the number of institutionalized older adults increases. Studies of disturbing or distressing behaviors among institutionalized dementia patients demonstrate that these behaviors occur frequently, and are a source of great frustration to the nursing staff [ 4 - 9 ]. The prevalence of dementia in nursing homes, estimated to be close to 50% in the U.S. [ 10 ] and over 70% in Australia [ 11 ], coupled with the behaviors associated with dementia, combine to make caring for patients with dementia a particular challenge to nursing staff. Measuring this burden can provide another means of evaluating outcomes of dementia treatment in the nursing home setting. There are measures relevant to the burden of paid/formal caregivers (i.e., nursing staff caring for dementia patients), but most focus on such overt behaviors as "forgetting what day it is" or "pull away" only (e.g., the Memory and Behaviors Problems Checklist [ 12 ] and the Resistiveness to Care Scale [ 2 ]). In their comprehensive examination of formal caregiver stress and cognitive impairment, Novak and Chappell [ 9 ] used an extensive individual face-to-face interview that lasted approximately one and a hours – a form of the Burden Interview [ 13 ] modified for institutional caregivers and Maslach's Burnout Inventory [ 14 ]. In their earlier work, they also asked 5 dementia-specific questions [ 15 ]. The Strain in Nursing Care Assessment Scale (NCAS) [ 9 , 10 ] was developed to address the need for a more comprehensive measure of nursing burden – specific to the long-term care setting – that includes sources of burden beyond overt behaviors. It is based on the conceptualization of burden as deriving from patient behaviors, as well as patient characteristics as perceived by the nurse carer. Unlike solely behavior-based measures completed by nurses, the NCAS addresses other aspects relevant to the provision of nursing care, such as nursing staff's perceptions regarding the meaningfulness of resident lives and the residents' level of gratefulness for care. The NCAS demonstrated the ability to capture changes in nurses' rating of difficulty regarding dementia patients' characteristics in a year-long study of a care intervention [ 16 ]. The Modified Strain in Nursing Care Assessment Scale (M-NCAS) was adapted from the NCAS developed and used in Swedish long-term care facilities [ 16 , 17 ]. The M-NCAS includes more items than the original instrument to capture the presence and impact of additional behaviors not represented in the original instrument, thus providing an even more comprehensive measure of nursing burden. The M-NCAS was used in a clinical trial of an atypical antipsychotic, risperidone, for treatment of nursing home patients with behavioral and psychological signs and symptoms in dementia. This preliminary investigation describes the psychometric evaluation of the M-NCAS using data from the clinical trial; thus, it is a post-hoc assessment of reliability and validity. Assessment of psychometric properties included exploration of item performance, subscale assignment, and assessment of reliability and validity based on the available data from the clinical trial. Methods Design and Procedures The psychometric evaluation used data from the baseline assessment of a multi-site, double-blind, placebo-controlled trial of risperidone in the treatment of behavioral and psychological symptoms among institutionalized patients with dementia. Fourteen sites throughout Australia and New Zealand recruited patients from the various nursing homes associated with each site. Patients were included in the clinical trial if they were 55 years of age or older, residing in a nursing home environment for one month, with a DSM-IV diagnosis of either dementia of the Alzheimer's type with behavioral disturbance, vascular dementia with behavioral disturbance, or mixed dementia, had scores less than or equal to 23 on the Mini-Mental Status Exam (MMSE) [ 18 ], and met aggressive item score criteria on the Cohen-Mansfield Agitation Inventory (CMAI) [ 19 ]. Aggressive item score criteria was defined as a score of at least 4 on one aggressive item, a frequency score of 3 on at least two aggressive items, or a frequency score of 2 on at least three aggressive items or two aggressive items occurring at a frequency of 2 and one at a frequency of 3 on the CMAI. These criteria were designed by the trial investigators to limit the sample to moderate to severe dementia patients having recognizable behavioral disturbances. Patients were required to have a carer who was able and committed to assisting the subject to comply with medication intake and the trial protocol, and who was willing to provide the information required at assessment interviews. Nursing home patients were excluded if they had medical or neurological conditions other than dementia which diminished cognitive function, dementia secondary to alcoholism, a diagnosis of major psychiatric comorbidity, substance abuse, tardive dyskinesia, clinically-uncontrolled medical conditions or laboratory abnormalities, a history of adult seizures, an administration of a depot injection or a long-acting neuroleptic within two treatment cycles of selection, hypersensitivity to neuroleptic treatment, or a history of failure to respond to risperidone of four weeks duration or participation in clinical trials with investigational drugs within the past four weeks. All patients or their appropriate representatives provided written informed consent, and the research protocol was approved by the appropriate institutional review board and state guardianship boards where required. Nurse carers were identified for each patient participant. Nursing Care Assessments The Behavioral Pathology in Alzheimer's Disease Rating Scale (BEHAVE-AD), the CMAI, and the M-NCAS were administered to the nurse carers of the individual patients by the investigator or site coordinator at baseline, week 4, week 8, and week 12 (trial endpoints). Nurse carers could have more than one patient participating in the trial, and they completed questionnaires for each individual patient. Modified Nursing Cares Assessment Scale (M-NCAS) The M-NCAS is was adapted from the original NCAS instrument developed in Sweden, and contains 32 items based in part on the 21 original items of the NCAS. Additional items were selected based on comments made by long-term care nurses regarding their experience caring for dementia patients with dementia in long-term care facilities, and added to make the instrument more comprehensive. The M-NCAS contains 32 items, with two domains per item: one addresses the occurrence and intensity of the behavior, in which the staff members indicate the extent to which they agree that the patient exhibits the behavior (the "attitude" domain); and one which addresses staff rating of the difficulty of coping with the behavior (the "strain" domain). Responses are measured on a four-point Likert-type scale (ranging from Agree to Don't Agree for the Attitude domain, and Very Easy to Very Difficult for the Strain domain). Lower scores are better for both domains. Domain total and subscale scores are calculated separately; total and subscale scores are calculated as an average of the individual item scores. The M-NCAS was translated from the original Swedish to English using standard forward and backward translation techniques. For subscale analyses, if < 50% of the scale items were missing, the scale was scored with the mean score of the non-missing items for that individual used to impute a score for the missing items. If > 50% of the items were missing, no scale score was calculated; the subscale score was considered missing. Total scores were calculated only if < 50% of the items were missing. Missing item criteria was based on that recommended by Ware et al. [ 20 ]. Cohen-Mansfield Agitation Inventory (CMAI) The CMAI is a 29-item scale specifically developed to assess agitated and disruptive behaviors, as well as care-related problems, occurring in demented subjects residing in nursing homes. The scale measures the frequency of behaviors, and systematically assesses agitation. The scale's 29 activities are rated on a 7-point scale indicating the frequency of a particular activity (range; never to a few times per hour). The activities are organized into 4 subscales: physical/aggressive, physical/non-aggressive, verbal/aggressive, and verbal/non-aggressive. The total aggression score is the sum of the physical and verbal aggression subscales. The total non-aggression score is the sum of the verbal and physical non-aggression subscales. A recall period of one week was used. The CMAI was used to assess the construct validity of the M-NCAS. Behavioral Pathology in Alzheimer's Disease Rating Scale (BEHAVE-AD) The BEHAVE-AD is a 25-item instrument designed to assess the severity of behavioral disturbances in subjects with dementia, based upon the caregiver's observation and report of the subject's behavioral problems. It consists of 7 subscales: paranoid and delusion ideation, hallucinations, activity disturbances, aggressiveness, diurnal rhythm disturbances, affective disturbances, and anxieties and phobias. Symptoms are rated on a scale from 0 (absence of the symptom) to 3 (greatest severity). A global item is included, in which a single judgment is made as to how troubling or dangerous the subject's behavior has been to the caregiver. A recall period of one month was used. The BEHAVE-AD was used to assess the construct validity of the M-NCAS. Data Analysis and Psychometric Evaluation We evaluated the item performance, scaling characteristics, reliability, and construct validity of the M-NCAS using baseline assessment data. Item Performance and Scaling Characteristics Item performance was examined to identify the presence of items that may reduce the instrument's ability to detect changes over time, and to discriminate between groups. The characteristics of individual M-NCAS items examined were mean, minimum and maximum responses, percent missing, floor and ceiling effects, and item-total correlations. These data were used to determine if some items should be deleted for subsequent analyses (e.g., item-to-total correlations below 0.40 [ 21 ]). Exploratory factor analysis was performed to evaluate the underlying subscale structure of the M-NCAS. Reliability Internal consistency reliabilities of total and subscale scores were estimated using Cronbach's coefficient α. Data to examine test-retest reliability was not available. Construct Validity Correlations were determined for total and subscale scores for attitude and strain domains and the CMAI physical and verbal aggression and non-aggression subscales. Correlations were also calculated between M-NCAS scores and the total score, seven subscales, psychotic symptoms subtotal and global rating question of the BEHAVE-AD, and the CGI-S and CGI-C scores. Pearson correlations were used unless otherwise specified. Correlations were expected to be low to moderate (0.20 ≤ r < 0.40). All statistical analyses were conducted using SAS statistical software version 8.0. A significance level of 0.05 and 2-tailed tests were used unless otherwise noted. No adjustments were made for multiple comparisons. Results Two hundred and eighty-two patients had evaluable baseline data for the psychometric evaluation. The majority of patients were female (72.4%) and Caucasian (98.4%); 11.5% were between the ages of 65 and 74, while 87.1% were over age 74. Most subjects had a diagnosis of Alzheimer's Disease (59.1%), while 28.3% had vascular dementia and 12.5% had mixed dementia. Nurse carers were primarily Caucasian. Further demographic data and information about the number and type of carers were not available. Item Performance Items and their distributional characteristics were reviewed. See Table 1 for a listing of items, mean scores, and floor (% of responses at 1) and ceiling effects (% of responses at 4). The attitude domain contained several items with high floor and ceiling effects. Item 12 had 79% of responses at floor – the highest among all items. Items 19 and 28 had the highest percentages of responses at ceiling (76% and 77%, respectively). Only one person had any missing data at the baseline visit. Examination of item correlation to total and subscale scores on the attitude domain demonstrated low or inverse correlations for four items (Item 5: tries to influence others in order to maintain control of his/her life; Item 12: is submissive; Item 27: patterns of behaviors you can foresee; and Item 19: has little control over his/her difficult behavior) (data not shown). Table 1 Items in the M-NCAS Attitude Strain Description Mean (SD) Floor (N, %) Ceiling (N, %) Mean (SD) Floor (N, %) Ceiling (N, %) Item 1 R Seems to behave in a completely aimless way 3.16 (1.08) 1 (45, 15.96%) 4 (144, 51.06%) 2.64 (0.93) 1 (39, 13.83%) 4 (50, 17.73%) Item 2 R Is anxious 3.3 (0.97) 1 (28, 9.93%) 4 (160, 56.74%) 2.79 (0.88) 1 (26, 13.83%) 4 (59, 20.92) Item 3 R Is unpredictable 3.2 (1.1) 1 (50, 50.0%) 4 (169, 59.9%) 2.86 (0.92) 1 (26, 9.22%) 4 (74, 26.24%) Item 4 Does things for a reason 2.86 (1.19) 1 (56, 45.0%) 4 (127, 45.0%) 2.57 (0.96) 1 (49, 17.38%) 4 (47, 16.67%) Item 5 1 Tries to influence others in order to maintain control of his/her own life 2.904 (1.3) 1 (74, 26.24%) 4 (154, 54.6%) 2.22 (1.06) 1 (92, 32.62%) 4 (41, 14.54%) Item 6 Is calm 3.17 (1.04) 1 (13, 4.61%) 4 (165, 58.5%) 2.76 (0.92) 1 (28, 9.93%) 4 (64, 22.70%) Item 7 R Is apathetic/seems to have limited emotions 2.54 (1.32) 1 (109, 38.65%) 4 (101, 35.82%) 2.48 (0.92) 1 (47, 16.67%) 4 (37, 13.12%) Item 8 R Is selfish 2.40 (1.32) 1 (116, 41.41%) 4 (94, 33.33%) 2.32 (0.99) 1 (69, 24.47%) 4 (39, 13.83%) Item 9 Is rewarding to work with 2.37 (1.20) 1 (87, 30.85%) 4 (82, 29.08%) 2.55 (0.97) 1 (46, 16.31%) 4 (51, 18.09%) Item 10 Is grateful for what is done for him/her 2.49 (1.18) 1 (71, 25.18%) 4 (89, 31.56%) 2.30 (0.94) 1 (61, 21.63%) 4 (33, 11.70%) Item 11 R Is paranoid 2.40 (1.29) 1 (111, 39.36%) 4 (89, 31.56%) 2.43 (1.04) 1 (65, 23.05%) 4 (52, 18.44%) Item 12 R Submissive/gives in to everything done to him/her 1.47 (0.93) 1 (222, 78.72%) 4 (15, 5.32%) 2.97 (0.89) 1 (20, 7.09%) 4 (88, 31.21%) Item 13 R Is attention seeking 2.57 (1.31) 1 (101, 38.82%) 4 (105, 37.23) 2.48 (1.00) 1 (56, 19.86%) 4 49, 17.38%) Item 14 R Is manipulative 1.97 (1.21) 1 (158, 56.03%) 4 (51, 18.09) 2.16 (0.96) 1 (82, 29.08%) 4 (29, 10.28%) Item 15 R Is ungrateful for the care he/she receives 2.23 (1.18) 1 (113, 40.07%) 4 (57, 20.21%) 2.30 (0.96) 1 (66, 23.40%) 4 (35, 12.41%) Item 16 R Is frightened/vulnerable 2.77 (1.18) 1 (72, 25.53%) 4 (109, 38.65%) 2.56 (0.90) 1 (38, 13.48%) 4 (41, 14.54%) Item 17 R Is lonely 2.76 (1.20) 1 (65, 23.05%) 4 (112, 39.72%) 2.49 (0.87) 1 (40, 14.18%) 4 (32, 11.35%) Item 18 R Has to concentrate exclusively on his/her needs in order to survive 2.36 (1.30) 1 (115, 40.93%) 4 (90, 32.03%) 2.40 (0.96) 1 (58, 20.64%) 4 (38, 13.52%) Item 19 R Has little control over his/her difficult behavior 3.59 (1.30) 1 (14, 4.97%) 4 (214, 75.89%) 3.16 (0.77) 1 (10, 3.55%) 4 (100, 35.46%) Item 20 R Is deliberately difficult 2.06 (1.17) 1 (138, 48.94%) 4 (46, 16.31%) 2.58 (1.00) 1 (52, 18.44%) 4 (53, 18.79%) Item 21 Tries to maintain some independence 2.33 (1.27) 1 (108, 38.30%) 4 (89, 31.56%) 2.56 (0.83) 1 (34, 12.06%) 4 (29, 10.28%) Item 22 Knows what he/she wants and stands up for his/herself 2.30 (1.25) 1 (103, 36.53%) 4 (86, 30.50%) 2.79 (0.85) 1 (28, 9.93%) 4 (51, 18.09%) Item 23 Seems to experience the normal range of emotions 2.83 (1.24) 1 (63, 22.34%) 4 (134, 47.52%) 2.59 (0.83) 1 (30, 10.64%) 4 (33, 11.70%) Item 24 Friendly 2.05 (1.03) 1 (94, 33.33%) 4 (48, 17.02%) 2.21 (0.84) 1 (58, 20.57%) 4 (17, 6.03%) Item 25 R Needs someone close by all the time/is demanding 2.57 (1.31) 1 (104, 36.88%) 4 (101, 35.82%) 2.58 (0.96) 1 (43, 15.25%) 4 (53, 18.79%) Item 26 R Has an empty life 2.81 (1.17) 1 (60, 21.28%) 4 (110, 39.01%) 2.56 (0.92) 1 (41, 14.54%) 4 (42, 14.89%) Item 27 R Has patterns of behavior you can foresee 3.00 (1.12) 1 (53, 18.79%) 4 (125, 44.33%) 2.75 (0.85) 1 (24, 8.51%) 4 (50, 17.73%) Item 28 R Is stubborn/resistive 3.68 (0.71) 1 (12, 4.26%) 4 (218, 77.31%) 3.22 (0.78) 1 (12, 4.26%) 4 (113, 40.07%) Item 29 R Is aggressive/ hostile 3.44 (0.91) 1 (26, 9.22%) 4 (181, 64.18%) 3.18 (0.84) 1 (16, 5.67%) 4 (113, 40.07%) Item 30 Has a meaningful life 3.17 (1.00) 1 (25, 8.87%) 4 (145, 53.19%) 2.59 (0.86) 1 (32, 11.35%) 4 (38, 13.48%) Item 31 Compliant/voluntarily co-operative 3.10 (1.01) 1 (9, 3.19%) 4 (150, 53.19%) 2.87 (0.80) 1 (16, 5.67%) 4 (58, 20.57%) Item 32 R Gives no job satisfaction 2.49 (0.91) 1 (43, 15.25%) 4 (37, 13.12%) 2.49 (0.91) 1 (43, 15.25%) 4 (37, 13.12%) 1 Italics indicate items deleted from instrument in subsequent analyses . R Item is reverse-coded for the attitude total and subscale scores. Attitude ranges from 1 (agree) to 4 (don't agree) and Strain ranges from 1 (very easy) to 4 (very difficult) Lower scores are considered better for both dimensions . Examination of any floor or ceiling effects on the strain domain demonstrated no outliers. Item-to-item correlations appeared acceptable, as did item-to-total correlations, all generally above 0.30. Factor Analysis With the number of factors unspecified, no clear factor pattern was discernible. Next, a six-factor solution for both dimensions (strain and attitude) was examined, based on the suitability of six factors for the original instrument [ 16 ]. Examination of the oblique rotated factor loadings demonstrated substantial item overlap. Results of subsequent factor analyses indicated that the best solution was a three-factor solution for the attitude domain, and a five-factor solution for the strain domain. Examination of the three-factor solution for the attitude dimension indicated that three items (item 19 has little control over his/her difficult behavior, item 27 has patterns of behavior you can foresee, item 7 is apathetic/seems to have limited emotions) did not perform well on the factor analysis – either overlapping substantially with another factor, or producing factor loadings less than 0.30. Based on overall item analysis results, including floor and ceiling effects, item-to-item and item-to-total correlation results, and a review of factor analysis results, four poorly-performing items were deleted from the attitude domain: item 5 tries to influence others in order to maintain control of his/her own life; item 12 submissive/gives in to everything done to him/her; item 19 has little control over his/her difficult behavior; and item 27 has patterns of behavior you can foresee. The removal of these items resulted in three distinct subscales for the attitude domain, demonstrating good approximation of simple structure. The three factors, all with variable loadings of 0.30 or above were "Attention-seeking" (e.g., "is manipulative", "needs someone close by all the time/is demanding"), "Autonomy" (e.g., "does things for a reason", "is apathetic/seems to have limited emotions"), and "Difficulty" (e.g. "is unpredictable" and "is friendly"). Items were reverse-coded as appropriate. To ensure inter-domain consistency, we removed the same four items from the strain dimension, thus resulting in an acceptable five-factor solution for this domain. The five factors, all with variable loadings of 0.30 or above, are "Affect" (e.g., "is calm", 'is anxious"), "Job satisfaction" (e.g., "is grateful for what is done for him/her", "gives no job satisfaction"), "Neediness" (e.g., "is selfish", "is manipulative"), "Predictability" (e.g., "knows what he/she wants and stands up for his/herself", "is aggressive/resistive"), and "Self-direction" (e.g., "is frightened/vulnerable", "seems to experience the normal range of emotions"). See Tables 2 and 3 for a list of specific items in each domain. Table 2 Item-to-subscale correlations for the attitude domain of the M-NCAS (28 items) at baseline Item Attention Seeking Autonomy Difficulty Item 2: Is anxious 0.573 0.073* 0.117 Item 6: Is calm 0.492 0.194 0.338 Item 11: Is paranoid 0.569* -0.059 0.288 Item 13: Is attention seeking 0.673 -0.203 0.174 Item 14: Is manipulative 0.588 -0.199 0.343 Item 16: Is frightened/vulnerable 0.425 0.124 -0.078* Item 17: Is lonely 0.568* 0.074 -0.009* Item 18: Has to concentrate exclusively on his/her own needs in order to survive 0.546 -0.167 0.256 Item 25: Needs someone close by/is demanding 0.706 -0.041* 0.129 Item 26: Has an empty life 0.462 0.305 0.286 Item 1: Seems to behave in a completely aimless way 0.089* 0.479 0.0750* Item 4: Does things for a reason -0.084 0.615 0.0043 Item 7: Is apathetic/seems to have limited emotions 0.037 0.514* 0.222 Item 21: Tries to maintain some independence 0.020* 0.652 0.115* Item 22: Knows what he/she wants and stands up for his/herself -0.141 0.696 -0.013* Item 23: Seems to experience the normal range of emotions -0.044* 0.601 -0.194 Item 30: Has a meaningful life 0.170 0.397 0.187 Item 3: Is unpredictable 0.164 0.082* 0.387 Item 8: Is selfish 0.391* -0.073 0.523 Item 9: Is rewarding to work with 0.159 0.241 0.651 Item 10: Is grateful for what is done for him/her 0.010 0.326* 0.630 Item 15: Is ungrateful for the care he/she receives 0.279 0.127 0.656 Item 20: Is deliberately difficult 0.259 -0.071* 0.482 Item 24: Friendly 0.076* 0.251 0.628 Item 28: Is stubborn/resistive 0.090* -0.012* 0.482 Item 29: Is aggressive/hostile 0.114* 0.023* 0.533 Item 31: Compliant/voluntarily co-operative 0.172 0.225 0.595 Item 32: Gives no job satisfaction 0.196 0.066* 0.525 All correlations significant at p < 0.05, except as indicated Table 3 Item-to-subscale correlations for the strain domain of the M-NCAS (28 items) at baseline Item Affect Job Satisfaction Neediness Predictability Self Direction Item 1: Seems to behave in a completely aimless way 0.727 0.447 0.409 0.348 0.488 Item 2: Is anxious 0.705 0.393 0.472 0.397 0.567 Item 3: Is unpredictable 0.694 0.441 0.479 0.452 0.306 Item 4: Does things for a reason 0.675 0.378 0.407 0.462 0.347 Item 6: Is calm 0.715 0.541 0.445 0.538 0.393 Item 7: Is apathetic/seems to have limited emotions 0.702 0.533 0.444 0.444 0.539 Item 9: Is rewarding to work with 0.572 0.777 0.475 0.536 0.422 Item 10: Is grateful for what is done for him/her 0.523 0.840 0.560 0.490 0.516 Item 15: Is ungrateful for the care he/she receives 0.520 0.828 0.601 0.521 0.513 Item 24: Friendly 0.406 0.728 0.477 0.500 0.515 Item 32: Gives no job satisfaction 0.549 0.807 0.536 0.580 0.615 Item 8: Is selfish 0.485 0.560 0.764 0.467 0.386 Item 11: Is paranoid 0.428 0.447 0.627 0.419 0.387 Item 13: Is attention seeking 0.483 0.432 0.809 0.389 0.471 Item 14: Is manipulative 0.442 0.482 0.784 0.389 0.381 Item 18: Has to concentrate exclusively on his/her own needs in order to survive 0.504 0.497 0.753 0.465 0.592 Item 20: Is deliberately difficult 0.414 0.510 0.672 0.537 0.450 Item 25: Needs someone close by all the time/is demanding 0.469 0.483 0.717 0.450 0.597 Item 21: Tries to maintain some independence 0.476 0.477 0.495 0.738 0.494 Item 22: Knows what he/she wants and stands up for his/herself 0.433 0.485 0.540 0.773 0.455 Item 28: Is stubborn/resistive 0.515 0.498 0.452 0.815 0.383 Item 29: Is aggressive/hostile 0.487 0.510 0.436 0.823 0.358 Item 31: Compliant/voluntarily co-operative 0.557 0.623 0.464 0.780 0.525 Item 16: Is frightened/vulnerable 0.476 0.423 0.413 0.383 0.710 Item 17: Is lonely 0.518 0.489 0.550 0.456 0.803 Item 23: Seems to experience the normal range of emotions 0.532 0.536 0.457 0.537 0.725 Item 26: Has an empty life 0.445 0.557 0.581 0.396 0.806 Item 30: Has a meaningful life 0.429 0.476 0.435 0.403 0.793 Following the removal of these 4 items, item-to-subscale correlations were examined for all subscales. In all cases, items correlated moderately to highly with the subscales to which they were assigned via the factor analysis. See Tables 2 and 3 . Reliability The internal consistency reliability (Cronbach's alpha) for the attitude and strain total scores was good – 0.79 and 0.95, respectively. Internal consistency reliability for subscales in both domains was excellent in general. See Table 4 . Table 4 Internal consistency reliability Scale Cronbach's α Attitude Domain Total Score 0.790 Autonomy 0.646 Attention Seeking 0.759 Difficulty 0.776 Strain Domain Total Score 0.945 Affect 0.795 Job Satisfaction 0.856 Neediness 0.856 Predictability 0.845 Self Direction 0.826 Construct Validity Table 5 summarizes the correlations between the M-NCAS attitude total and subscale scores and the CMAI total aggression, total non-aggression, and subscale scores. In general, correlations were low to moderate [ 22 ]. The highest correlation was between attention-seeking and verbal non-aggression (r = 0.68; p < 0.01). Table 5 Correlations 1 between the CMAI and the M-NCAS attitude and strain domains CMAI Scales M-NCAS Scales Physical Aggression Physical Non-Aggression Verbal Aggression Verbal Non-Aggression Total Aggression Total Non-Aggression Attitude Domain Total Score 0.286 0.079 0.325* 0.467* 0.336* 0.306* Attention Seeking 0.060 0.142 0.202 * 0.675* 0.107 0.463* Autonomy 0.145 0.011 0.016 -0.074 0.129 -0.031 Difficulty 0.370* -0.005 0.390* 0.242* 0.426* 0.125 Strain Domain Total Score 0.290* 0.141 0.273* 0.344* 0.324* 0.286* Affect 0.330* 0.211 0.193* 0.246* 0.337* 0.286* Job Satisfaction 0.320* 0.0189 0.333* 0.210 0.367* 0.125 Neediness 0.150* 0.128 0.222 0.458* 0.190* 0.338* Predictability 0.297* 0.166* 0.258* 0.154* 0.327* 0.204 Self Direction 0.151* 0.060 0.143 0.302* 0.170* 0.204* significant at p < 0.01; 1 Pearson product-moment correlations Table 5 also summarizes the correlations between the strain total and subscale scores and the CMAI total aggression, total non-aggression, and subscale scores. Correlations were generally low to moderate, with only one correlation exceeding 0.40 (verbal non-aggression and neediness; r = 0.46, p < 0.01). Table 6 presents the correlations between the M-NCAS attitude total and subscale scores and the BEHAVE-AD. With the exception of the activity disturbance subscale, most correlations between the BEHAVE-AD subscales and the attitude total score were statistically significant. In general, the attention-seeking attitude subscale was related to BEHAVE-AD scores, but the autonomy subscale was not. Results were mixed for the difficulty subscale, with a correlation of 0.47 to the aggressiveness BEHAVE-AD subscale. Table 6 Correlations 1 between the BEHAVE-AD and the M-NCAS BEHAVE-AD Scales M-NCAS Scales Total BEHAVE-AD Paranoid and delusional ideation Hallucination Activity disturbance Aggressiveness Diurnal rhythm disturbance Affective disturbance Anxiety and phobias Attitude Domain Total Score 0.380 0.272* 0.159* 0.078 0.386* 0.182* 0.225* 0.354* Attention Seeking 0.478* 0.421* 0.158* 0.091* 0.269* 0.185* 0.373* 0.556* Autonomy -0.059 -0.176* 0.044 0.021 -0.032 0.101 -0.015 -0.041 Difficulty 0.268* 0.207* 0.104 0.038 0.471* 0.079 0.053 0.128 Strain Domain Total Score 0.397* 0.281* 0.134* 0.146* 0.425* 0.209* 0.156* 0.353* Affect 0.360* 0.195* 0.137 0.243* 0.362* 0.191* 0.169 0.286* Job Satisfaction 0.265* 0.161* 0.074 0.061 0.431* 0.121 0.061 0.210* Neediness 0.383* 0.325* 0.140 0.084 0.327* 0.234* 0.165* 0.372* Predictability 0.331* 0.246* 0.107 0.169* 0.437 0.144 0.028 0.217* Self Direction 0.310* 0.224* 0.093 0.066 0.241* 0.164* 0.213* 0.369* *p < 0.01; 1 Pearson product-moment correlations Table 6 also presents results for the M-NCAS strain domain and BEHAVE-AD score correlations. The total and most of the strain subscale scores were significantly related to BEHAVE-AD total and subscale scores. Discussion The M-NCAS demonstrated good psychometric properties based on an analysis of baseline data collected during a clinical trial of risperidone versus placebo, as reported in this preliminary investigation. Item performance, particularly floor and ceiling effects (together with item-total correlation data), suggested that four items on the attitude domain were performing poorly from a psychometric standpoint. Two items performed poorly on item correlations as well as factor analysis, while item 12 had a high floor effect as well as poor correlations. The remaining 28 items still capture sufficiently comprehensive data on the domains of interest, based on content review and on the empirical performance of subscales derived from them. The collection of additional data would enhance the confidence of conclusions regarding M-NCAS performance. In general, the M-NCAS demonstrated excellent internal consistency reliability, with only the autonomy subscale producing a Cronbach's alpha below 0.70. Cronbach's alpha values of 0.70 or greater are considered suitable for use in the analyses of group comparisons. Supporting construct validity of the M-NCAS were the moderate and significant correlations to the CMAI and BEHAVE-AD. Correlation with these measures was not expected to be high, given the more expansive focus of the M-NCAS relative to the CMAI and BEHAVE-AD. These results support the use of the M-NCAS for the collection of valid, reliable, and comprehensive information on the burden experienced by nurse carers when caring for dementia patients in long-term care settings. The instrument can assist in staff management by identifying nurse carers with the greatest levels of burden – ideally so that remedial action could be taken, either through extra support for staff members or through the shifting of caseloads from particularly difficult patients for nurses at risk for attrition. There are several limitations to this study. Of major concern is the lack of information on the nurse carers themselves. No data was available as to the type of carer, the duration they had cared for the patient, and their sociodemographic characteristics. The inability to characterize the respondents to the M-NCAS is a limitation in interpreting generalizability of results. No control was attempted for nurse raters, so the possibility exists that measurement properties of the M-NCAS differed across nurse raters – therefore skewing the results. The extent to which rater effects limit the validity of results is likely to be minimal, however, given that multiple subjects were being rated by each nurse rater. Future examination of between-subject properties would be helpful. The sample size for the study is small, relative to the standard recommendations regarding the number of subjects per item for factor analysis. However, we set criteria for relatively moderate to high factor loadings. Of note is that the final recommendations for the measure length are closer to that guideline. Because this was a post-hoc analysis of clinical trial data, the instruments selected for use in assessing construct validity may not have been optimal, in that they did not assess nursing strain per se. However, they did assess the propensity of exhibiting certain behaviors – as does the M-NCAS. Additional data on stability across time (test-retest reliability) is desirable for a full psychometric description of the M-NCAS, and was not available from this study. Finally, this sample of nurse carers was almost exclusively Caucasian. Therefore, the generalizability of these results to nurse carers of different ethnicities, and in different locales is limited. Conclusions The M-NCAS enables the detection of the presence or absence of specific behaviors similar to checklists (the attitude scale), but extends that information by providing a rating of the degree of burden of each aspect rated. It possesses good psychometric properties for use with nurse carers working with Alzheimer's patients in long-term care facilities. The M-NCAS provides a unique approach to identifying both positive and negative behaviors, and to quantifying the amount of stress felt by carers as a result of these behaviors. The nursing staff perspective on residents with dementia is unique [ 23 ], and the M-NCAS exploits this perspective by capturing the aspects of residents beyond overt behaviors. List of Abbreviations Behavioral Pathology in Alzheimer's Disease Rating Scale (BEHAVE-AD) Cohen-Mansfield Agitation Inventory (CMAI) Exploratory Factor Analysis (EFA) Mini-Mental Status Exam (MMSE) Modified Strain in Nursing Care Assessment Scale (M-NCAS) Strain in Nursing Care Assessment Scale (NCAS) Authors' contributions LK participated in the data analysis design, data interpretation, and drafting of the manuscript. LF participated in the design, data interpretation, and drafting of the manuscript. GC performed the statistical analyses. MR conceived of the study and participated in data interpretation. HB conceived of the study and participated in both data collection and interpretation. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535543.xml
535572	Brain versus Machine Control	Dr. Octopus, the villain of the movie "Spiderman 2", is a fusion of man and machine. Neuroscientist Jose Carmena examines the facts behind this fictional account of a brain- machine interface	“Dr. Octopus,” the villain that terrorizes the city in the most recent film of the popular Spider-Man comic, is the ultimate characterization of a brain–machine interface (BMI) on the big screen. In Spider-Man 2, the brain is that of nuclear physicist Dr. Otto Octavius, who dreams of harnessing nuclear fusion. The machine is a harness of four mechanical arms designed with tentacle-like flexibility, gripping and vision capabilities, and an artificial intelligence module that gives them some autonomy. The interface between the machine and the brain is at the spinal cord level, with an “inhibitor chip” to prevent the artificial intelligence module in the mechanical arms from taking over Octavius's brain. Controlling this mechanical device with his own thoughts, Octavius is able to manipulate hazardous materials during his fusion experiments. However, things go terribly wrong during the exhibition of one of these experiments: the mechanical arms fuse to Octavius's body while the inhibitor chip is disabled, resulting in the machine gaining partial control of his brain. Unable to subvert the machine to his will and conscience, Octavius, together with the BMI, becomes the villainous Dr. Octopus. At the end of the movie, in a flicker of sanity and heroism, Octavius dramatically sacrifices his life as the only way to terminate the evil machine and save the world. Although Dr. Octopus is a fictional character, a figment of a vivid imagination, audiences are fascinated by the fact that he is a human BMI. BMIs straddle the worlds of fact and fiction. While the entertainment industry has focused primarily on applications for augmenting cognitive and sensorimotor function, as seen in Star Trek, Firefox, and many other science-fiction scenarios, the scientific community has targeted clinical applications, such as neuroprostheses for restoring motor function after traumatic lesion of the central nervous system. The current BMI approach is based on the idea that a human user could enact voluntary motor intentions through a direct interface between his brain and an artificial actuator in virtually the same way that we see, walk, or grab an object with our own natural limbs. Proficient brain control of an external device or actuator should be achievable through training using any combination of visual, tactile, or auditory feedback. As a result of long-term use of the BMI, the brain should be able to “incorporate” (or adapt to) the artificial actuator as an extension of its own body. With these goals in mind, the last five years have witnessed a dramatic increase in BMI-related studies in academic institutions around the world. Subjects have learned to utilize their brain activity for different purposes, ranging from electroencephalogram- and electrocorticographic-based systems ( Wolpaw et al. 2002 ; Leuthardt et al. 2004 ), in which human subjects control computer cursors, to multielectrode-based systems, in which nonhuman primates control the movements of cursors and robots to perform different kinds of reaching and grasping tasks ( Serruya et al. 2002 ; Taylor et al. 2002 ; Carmena et al. 2003 ; Musallam et al. 2004 ). These examples of what could be called the first generation of BMIs have something in common: they have been exclusively controlled by neural signals. Even with BMIs that use neural activity recorded with invasive electrodes to yield higher bandwidth and thus allow for the execution of more complex tasks, it remains unclear whether the quality of the signal will ever suffice for a patient to freely, safely, and effectively control a prosthetic arm to perform daily tasks. For instance, the level of motor skill required for dexterous finger manipulation is outstanding. Planning paths and avoiding obstacles while reaching and grasping in unconstrained environments requires similarly fine motor control. Thus, realistic motion through a complex environment with a BMI is extremely challenging and, perhaps, not feasible with the relatively low bandwidth (∼10 Hz) of current BMIs. Even if significant improvements are made in the algorithms used to decode neural activity by, for example, incorporating knowledge from neurophysiological experiments of how motor signals that underlie movements are encoded in the brain, current BMI bandwidth still may not be sufficient to reach the performance level an injured patient would desire. What does this mean for second-generation BMIs? We may find some inspiration in Dr. Octopus. The fictional BMI in Spider-Man 2 is innovative in the sense that it is a hybrid system that incorporates both neuronal and artificial control signals. It makes perfect sense to take advantage of the fields of engineering (control theory) and artificial intelligence to build better BMIs—part brain and part robot. In principle, these hybrid BMIs would allow a patient to accomplish a task more efficiently than those relying on neuronal signals alone. For example, in a common task such as reaching for and grasping a glass of water, a hybrid BMI would be fed with both brain and machine control signals; the intention of movement would be decoded directly from neuronal signals, leaving obstacle avoidance and grasping stabilization to the artificial control module of the system. Such a module would get inputs from sensors embedded in the robot, and would produce a control signal that would fuse with the neuronal control signal to augment the final output command. What ratio of neuronal versus artificial signal would be needed for optimal control of a BMI? In the movie, Octavius's crisis is a severe unbalance in favor of machine control. Science fiction aside, we see the more realistic potential problems of having a physical device gaining autonomous control. Technically, this could be analyzed as too much gain in the artificial control signal, which, in a realistic scenario, would likely result in oscillating behavior, jerky grasping, etc. Hence, safeguarding measures (characterized in the movie as the inhibitor chip in Octavius's brain stem) would be needed to avoid dangerous situations when a chronic neuroprosthesis freely interacts with the real world. For both science and science fiction, the question is the same. Brain and machine: which one gets the power?	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535572.xml
509292	Nicotine's Defensive Function in Nature	Plants produce metabolites that directly decrease herbivore performance, and as a consequence, herbivores are selected for resistance to these metabolites. To determine whether these metabolites actually function as defenses requires measuring the performance of plants that are altered only in the production of a certain metabolite. To date, the defensive value of most plant resistance traits has not been demonstrated in nature. We transformed native tobacco (Nicotiana attenuata) with a consensus fragment of its two putrescine N-methyl transferase (pmt) genes in either antisense or inverted-repeat (IR pmt ) orientations. Only the latter reduced (by greater than 95%) constitutive and inducible nicotine. With D 4 -nicotinic acid (NA), we demonstrate that silencing pmt inhibits nicotine production, while the excess NA dimerizes to form anatabine. Larvae of the nicotine-adapted herbivore Manduca sexta (tobacco hornworm) grew faster and, like the beetle Diabrotica undecimpunctata, preferred IR pmt plants in choice tests. When planted in their native habitat, IR pmt plants were attacked more frequently and, compared to wild-type plants, lost 3-fold more leaf area from a variety of native herbivores, of which the beet armyworm, Spodoptera exigua, and Trimerotropis spp. grasshoppers caused the most damage. These results provide strong evidence that nicotine functions as an efficient defense in nature and highlights the value of transgenic techniques for ecological research.	Introduction Plants produce many secondary metabolites, of which some are thought to function as direct defenses against pathogens and herbivores by reducing their performance, survival, and reproduction. Numerous plant allelochemicals with antiherbivore properties are classified according to their mode of action (e.g., toxins, antifeedants, antidigestive proteins, etc.) ( Bennett and Wallsgrove 1994 ) and have been used in agriculture to control insect pests ( Hedin 1991 ). The fact that a secondary metabolite reduces herbivore performance does not by itself demonstrate that the endogenously expressed metabolite functions defensively in the plant's natural environment ( Bell 1987 ), because the evolutionary interaction between herbivores and their host plants may have reduced the defensive efficacy of the metabolite. Phytophagous insects have evolved various strategies to cope with allelochemicals ( Karban and Agrawal 2002 ) and tend to tolerate, or even co-opt, plant defenses for their own defenses ( Wink and Theile 2002 ). Pharmacological studies demonstrating a resistance effect of metabolites applied to plants or artificial diets ( Yamamoto et al. 1968 ; Bowers and Puttick 1988 ), and studies using heterologously expressed genes in agricultural systems ( Carozzi and Koziel 1997 ; Hilder and Boulter 1999 ), represent a first step in evaluating the defensive function of a secondary metabolite. The interpretation of these studies is confounded by both the altered ecological context in which the resistance is measured and the altered chemical milieu, which is also known to influence the defensive function of a metabolite. Stronger evidence for resistance effects of allelochemicals arises from studies establishing correlations between plant resistance against herbivores and the genetically variable accumulation of secondary metabolites ( Berenbaum et al. 1986 ; Shonle and Bergelson 2000 ) or from studies demonstrating the defensive role played by a suite of elicited metabolites ( Orozco-Cardenas et al. 1993 ; Baldwin 1998 ; Halitschke and Baldwin 2003 ). Ideally, the benefits of a putative defense trait should be determined in plants differing only in a single gene that controls the expression of a resistance trait and are otherwise identical ( Bergelson and Purrington 1996 ). To date, studies measuring resistance of “near isogenic” lines with altered metabolite accumulations ( Jackson et al. 2002 ) provide the strongest evidence for their resistance, but these lines, which are created by repetitive backcrossing, are likely to differ in many loci linked to the target locus, which may also affect resistance. Such problems of genetic linkage have been overcome through the use of genetic transformation to explore the fitness effects of herbicide resistance ( Bergelson et al. 1996 ; Purrington and Bergelson 1997 ) and pathogen resistance ( Tian et al. 2003 ) in field populations of Arabidopsis . In this study, we use transgenic silencing to alter a single putative resistance trait—the production of nicotine—and thereby establish its contribution to plant resistance in the field. The pyridine alkaloid nicotine is one of the best-studied putative plant resistance traits. Because it can interact with the acetylcholine receptors in the nervous systems of animals, nicotine is extremely toxic to most herbivores and, consequently, was one of the first insecticides used to control pests in agriculture ( Schmeltz 1971 ). Evidence for the resistance value of nicotine arises from the agricultural practice of using nicotine sprays and genotypes of cultivated tobacco differing in nicotine levels ( Jackson et al. 2002 ). Although nicotine is widely toxic, insects adapted to nicotine-producing plants have evolved resistance to this alkaloid ( Glendinning 2002 ). The tobacco specialist Manduca sexta (tobacco hornworm) tolerates doses of nicotine that are fatal to unadapted herbivores but grows more slowly on high-nicotine diets ( Appel and Martin 1992 ; Wink and Theile 2002 ). Other studies suggest that M. sexta might even be better defended by dietary nicotine against its parasitoid, Cotesia congregata, which suffers higher mortality when parasitizing larvae fed on high- rather than low-nicotine diets ( Barbosa et al. 1986 ; Thorpe and Barbosa 1986 ). Thus, the coevolutionary arms race between nicotine-producing plants and their adapted herbivores may have reduced the defensive value of nicotine. In the native tobacco species Nicotiana attenuata and N. sylvestris, nicotine is the most abundant alkaloid. Elicitation of N. attenuata with jasmonic acid methyl ester (MeJA) in its native habitat increases nicotine content, which is correlated with enhanced plant fitness when plants are attacked ( Baldwin 1998 ). However, herbivore attack and MeJA elicitation (as well as the plant's endogenous jasmonic acid cascade [ Halitschke and Baldwin 2003 ]) regulate many resistance traits, including trypsin protease inhibitors (TPIs), diterpene glycosides, and volatile emissions involved in indirect defense. Hence, nicotine is only one of a suite of putative defense traits elicited by herbivore attack, and its specific role remains to be determined. In laboratory trials, resistance benefits of nicotine production against M. sexta larvae were established using transgenic N. sylvestris plants silenced in their nicotine biosynthesis by antisense expression of putrescine N-methyl transferase (PMT). Plant consumption and the performance of M. sexta larvae were negatively correlated with constitutive nicotine levels in laboratory feeding trials ( Voelckel et al. 2001 ); whether this result applies to plants in their natural habitat is unclear. To examine the resistance effect of nicotine, we transformed N. attenuata with inverted-repeat pmt (IR pmt ) and antisense pmt constructs and found that only IR pmt plants had strongly reduced nicotine content. We characterized the defense and growth phenotypes of two independently transformed homozygous IR pmt lines and found that measured direct and indirect defenses did not differ from those of the wild-type (WT) plants, except for a dramatic reduction (greater than 95%) of MeJA-elicited and constitutive nicotine production and an increase in anatabine content. In pulse-chase experiments with D 4 -nicotinic acid (NA) ethyl ester, we demonstrated that the increased anatabine likely results from a dimerization of the NA that would normally have been used in nicotine biosynthesis. In feeding trials, M. sexta larvae preferred and grew faster on IR pmt than WT leaves. We transplanted WT and IR pmt plants into N. attenuata 's native habitat in southwestern Utah and elicited a subset with MeJA. Several naturally occurring herbivore species attacked and damaged unelicited IR pmt plants more than unelicited or elicited WT and elicited IR pmt plants. These results demonstrate that nicotine functions as an effective resistance trait under natural conditions. Results/Discussion IR pmt Constructs Silence Nicotine Production Nicotine accumulation was not reduced in most of the independent lines transformed with antisense pmt constructs (25 lines of pNATPMT1 and six lines of pCAMPMT1) compared to WT ( Figure 1 A). None of the five lines with lower nicotine accumulation in the T 1 screen had nicotine levels lower than those of WT in the homozygous T 2 generation. In contrast, 29 of 34 independently transformed lines with the IR pmt construct pRESC5PMT had dramatically reduced constitutive and MeJA-induced nicotine accumulations ( Figure 1 B). The suppression of nicotine accumulation was stable during plant development and when plants were grown in the glasshouse or in the field in Utah. Clearly, inverted-repeat constructs are more efficient at silencing the expression of endogenous genes, as has been previously described ( Wesley et al. 2001 ). Figure 1 Comparison of Antisense and Inverted-Repeat Silencing of pmt Nicotine content (mean of 5–6 plants/line) normalized to mean of WT of unelicited (control) N. attenuata plants and plants 5 d after elicitation with 150 μg of MeJA per plant from independent lines transformed with (A) antisense pmt constructs and (B) an IR pmt construct. In contrast to the 31 lines transformed with the antisense pmt construct, 29 of the 34 IR pmt lines had dramatically reduced constitutive and MeJA-induced nicotine levels. T, terminator; P, promoter; I, spliceable intron; arrow, 950-bp consensus fragment of pmt1 and pmt2 . For details of transformation constructs see Protocol S1. Genomic and Transcriptional Characterization Two homozygous T 2 IR pmt lines (108 and 145) with reduced nicotine levels were further characterized. Southern blot analysis using a probe hybridizing to the selective marker in the IR pmt construct demonstrated that both lines contained a single insertion ( Figure S1 ). Transformation with a pRESC transformation vector allowed the transferred DNA (T-DNA) and flanking DNA at the insertion site to be recovered from the plant genomic DNA. These experiments demonstrated that the T-DNA integrated into the N. attenuata genome at a single site in each line, since all sequenced clones from a line (108, n = 4; 145, n = 5) contained the same flanking sequence (see Figure S1 and Protocol S1 ). Transcripts of the pmt genes in the two lines were significantly reduced to approximately 10% of the constitutive and MeJA-induced WT mRNA levels ( Figure 2 A), demonstrating that the targeted genes were successfully silenced. Figure 2 PMT Transcript and Alkaloid Levels of IR pmt Lines Mean (± SE) relative PMT mRNA transcript levels in the roots (A), and leaf levels of (B) nicotine and (C) anatabine, in two independent lines of IR pmt -transformed (108 and 145) and WT N. attenuata plants. Elicited (150 μg of MeJA) and unelicited (control) plants were harvested at 10 h for transcript (A) and at 4 d for alkaloid (B and C) quantification. Both IR pmt lines had significantly reduced PMT transcript and nicotine but featured anatabine not present in WT plants. Lowercase letters signify differences at p ≤ 0.01, Bonferroni corrected ([A] n = 3, ANOVA: F 2,12 = 12.55; [B] n = 8–10, ANOVA: F 2,50 = 135.4; [C] n = 8–10, ANOVA: F 2,50 = 39.611]. n.d., not detected. Metabolic Consequences of pmt Silencing in N. attenuata Consistent with the observed silencing of pmt transcripts, the constitutive and induced nicotine levels in transformed plants of both lines were dramatically reduced to 3%–4% of the levels found in WT plants ( Figure 2 B). All 29 IR pmt lines with reduced nicotine levels accumulated the alkaloid anatabine, which was not detected in WT plants. Constitutive and MeJA-induced total (nicotine, anabasine, and anatabine) alkaloid contents of the two IR pmt lines were about one-half and one-third of the WT levels, respectively, of which anatabine comprised 30% and 23% ( Figure 2 C). Levels of anabasine representing 20% of the constitutive and 8% of the MeJA-elicited total alkaloid contents in WT plants were unchanged in IR pmt plants ( Figure S2 ). Elevated anatabine levels were also found in recently published studies with antisense pmt transformation of N. tabacum; elevated anatabine levels did not affect transcript levels of other genes encoding enzymes involved in alkaloid metabolism ( Chintapakorn and Hamill 2003 ). Anatabine consists of a pyridine and a piperideine ring. Both are likely derived from NA, which is also the precursor of the pyridine ring of nicotine ( Leete and Slattery 1976 ). Disrupting nicotine biosynthesis at the formation of the pyrrolidine ring by silencing PMT activity might cause an oversupply of the NA used in the biosynthesis of anatabine. Feeding the roots of hydroponically grown MeJA-elicited WT plants with NA ethyl ester resulted in formation of anatabine at levels of about a third of the total alkaloids (nicotine and anatabine) ( Figure 3 ); in the IR pmt lines, anatabine constitutes 98% of the total alkaloids. Feeding plants with D 4 -NA ethyl ester results in the formation not only of D 4 -nicotine and D 4 -anatabine but also of D 8 -anatabine, demonstrating that the last integrates two D 4 -NA units. When these experiments are conducted with WT plants, about half of the anatabine is labeled, suggesting that the unlabeled half was formed from endogenous unlabeled NA. In addition, about one-fourth of the WT nicotine was D 4 -nicotine. In IR pmt plants, in contrast, only traces of D 4 -nicotine were found, but one-third of the anatabine was either D 4 - or D 8 -labeled. In summary, exogenously supplied NA is taken up by the roots of N. attenuata plants and used in alkaloid biosynthesis, and an oversupply of NA results in the formation of anatabine. These results support the hypothesis that the silencing of pmt disrupts nicotine biosynthesis, causing an oversupply of NA and the subsequent formation of anatabine. Figure 3 Alkaloid Biosynthesis and the Consequences of a NA Oversupply Biosynthesis scheme and proportion of unlabeled (M + ) and labeled (M + +4, M + +8) nicotine and anatabine in the leaves of two independently transformed N. attenuata IR pmt lines (108 and 145) and WT plants 5 d after elicitation with 150 μg of MeJA per plant. Plants were grown in hydroponic solutions and supplied with either unlabeled or D 4 -ring-labeled NA ethyl ester (1 mM) 24 h after elicitation ( n = 3 or 4). The oversupply of NA resulted in the formation of anatabine even in WT plants from both labeled exogenous and unlabeled endogenous NA pools. IR pmt plants did not differ from WT plants in any other measured secondary metabolite or growth parameter. Constitutive or MeJA-induced levels of caffeoylputrescine, chlorogenic aid, rutin ( Figure S2 ), TPI activity, or the release of cis -α-bergamotene ( Figure S3 ) in IR pmt -transformed plants did not differ from those of WT plants. Rosette-stage and elongation-stage growth in individual pots in both the glasshouse and the field ( Figure S4 ) did not differ between WT and IR pmt lines, and transformed lines were not visually or morphologically distinguishable from WT plants. Hence, the IR pmt plants represent an ideal construct with which to examine the ecological consequences of nicotine production. Effects of Nicotine Silencing on N. attenuata Herbivores M. sexta larvae reared on IR pmt plants in the glasshouse gained significantly more mass and changed instars faster than larvae reared on WT plants ( n = 17–20; ANOVA: p < 0.01, p WT-PMT108 < 0.02, p WT-PMT145 < 0.01). The differences were comparable to those observed for M. sexta larvae reared on nicotine-enriched artificial diets ( Parr and Thurston 1972 ; Appel and Martin 1992 ) or on nicotine-enhanced WT ( Baldwin 1988 ) or antisense- pmt –transformed N. sylvestris plants ( Voelckel et al. 2001 ). Two-thirds of freshly eclosed M. sexta larvae, given the choice between leaf material from WT or IR pmt (108) plants, preferred to initiate feeding on the latter ( n = 43; Chi 2 = 6.7, p < 0.01). Such behavior suggests that nicotine plays an important role in determining feeding sites of M. sexta larvae , as has been suggested in a study with cultivated tobacco ( Kester et al. 2002 ). While the relative toxic effects of anatabine and nicotine remain unstudied, these results are likely to underestimate the influence of nicotine on M. sexta choice and performance, because IR pmt plants had enhanced levels of anatabine. Since secondary metabolism is known to be sensitive to environmental parameters that differ between glasshouse and field conditions (e.g., UV-B influence; Caldwell et al. 1983 ), nicotine, anatabine, and TPI levels of WT and IR pmt plants grown in the field plantation were analyzed: they were found not to differ from plants grown under laboratory conditions ( Figure 4 A). A M. sexta feeding choice test evaluating the larvae's choice between field-grown WT and IR pmt plants ( n = 57; Chi 2 = 7.74, p < 0.01) verified the results described above for the same experiment conducted with glasshouse-grown plants. Thus, the phenotype of glasshouse-grown IR pmt plants was not altered by growth under field conditions. In addition, choice tests with field-collected D. undecimpunctata, which was observed colonizing only IR pmt plants in the field plantation, revealed that 77% of these beetles preferred the nicotine-deficient IR pmt leaf material over WT ( n = 35; Chi 2 = 10.31, p < 0.001). Another beetle species observed occasionally on WT plants, Trichobarus mucorea, does not distinguish between WT and IR pmt leaf material in choice tests ( n = 19; Chi 2 = 0.05, p = 0.8). Figure 4 Herbivore Damage to IR pmt and WT N. attenuata Plants in Nature (A) Leaf alkaloids (nicotine and anatabine) and TPIs 7 wk after transplantation ( n = 6). Mean (± SE) percentage total leaf area damaged by (B) all herbivores and (C) only by Spodoptera exigua on WT N. attenuata plants and plants transformed with an IR pmt construct (108) that were either untreated (solid lines) or elicited (dotted lines; asterisk) with MeJA 7 d after plants were transplanted into a field plot in a native habitat. Differences between 108 and WT, 108, and WT are significant at p ≤ 0.05 ( n PMT = 36, n WT = 50, n PMT* = 28, n WT* = 27; [B] ANOVA: F 3,822 = 5.73, p = 0.001; [C] ANOVA: F 3,822 = 4.6, p = 0.004). Plants of the nicotine-deficient transformed line 108 suffered significantly higher leaf area damage than did WT plants, but when line 108 was elicited, leaf damage by all herbivores was reduced to WT levels. In the field plantation, IR pmt plants lost significantly more leaf area to herbivores than did WT plants ( Figure 4 B), demonstrating that nicotine indeed functions as a direct resistance trait of N. attenuata in its native habitat. Over a period of 16 d, IR pmt plants exposed to naturally occurring herbivores lost 16% of their total leaf area to herbivores, an amount that is more than double the amount of damage incurred by WT plants. In order to meet compliance requirements described in the Code of Federal Regulations (7CFR340.3c) for the introduction of organisms altered through genetic engineering, flowers were removed as they matured, and therefore we could not directly measure the fitness consequences of this greater herbivore load. However, in other experiments with N. attenuata plants grown in natural populations, leaf area damage is negatively correlated with capsule number ( Baldwin 1998 ; Kessler and Baldwin 2004 ), suggesting that the strongly enhanced herbivore damage of the nicotine-deficient IR pmt plants translates into a fitness loss. IR pmt plants were attacked by a variety of insect herbivores. About half of the total herbivore damage resulted from S. exigua feeding ( Figure 4 C). One-third of the total herbivore damage was damage from grasshoppers of the genus Trimerotropis, which followed the same general pattern of distribution as S. exigua damage, but the differences between unelicited IR pmt and WT plants were not significant. The damage caused by Epetrix hirtipennis was variable but significantly higher for unelicited IR pmt compared to WT plants (ANOVA: F = 2.81, df = 3, p = 0.04, p PMT-WT < 0.05). MeJA elicitation significantly reduced the damage of IR pmt plants to levels found on WT plants, suggesting that MeJA treatment elicits defense traits that are as efficient as the constitutive levels of nicotine in protecting plants. MeJA elicitation of N. attenuata plants is known to induce a diverse suite of transcriptional responses and secondary metabolites including TPIs, phenolics, flavonoids, phenolic putrescine conjugates, diterpene sugar esters, and volatile organic compounds ( Halitschke and Baldwin 2003 ; Roda and Baldwin 2003 ), some of which apparently function as resistance traits. Which component of this complex suite of elicited metabolites is as effective as nicotine remains to be determined. It should be noted that the overall amounts of leaf area lost to herbivores was relatively low during the field experiments. Only 5% of the canopy area was lost from control and MeJA-elicited WT plants. In previous experiments ( Baldwin 1998 ), fitness differences were observed between control and MeJA-elicited WT plants in populations that had lost approximately 40% of their canopy area to herbivores. Altogether, these results provide direct evidence for the defensive value of nicotine. In a field trial, we established that a native tobacco, which produces large amounts of nicotine, is better defended against its natural herbivores than are nicotine-deficient transformants of the same genetic background. This is likely mediated by the reduction of herbivore performance and by the fact that these phytophagous insects prefer low-nicotine diets. In contrast to studies demonstrating genetic correlations between the production of secondary metabolites and herbivore resistance ( Berenbaum et al. 1986 ; Shonle and Bergelson 2000 ), the resistance effects established in this study can be directly attributed to the altered traits. The fact that the silencing of one enzyme in the nicotine biosynthetic pathway redirects metabolite flux, resulting in the accumulation of an apparently less toxic alkaloid, anatabine, underscores the importance of characterizing single-gene transformants for secondary effects. Conclusion Plant secondary metabolites are widely accepted as essential components of a plant's direct defenses against its natural enemies, but unambiguous proof has been lacking, mainly because of the difficulty of altering the expression of single traits in plants and testing the consequences of these manipulations under natural conditions. Transformation technology has provided biologists with the ability to manipulate and study the ecological consequences of single-gene manipulations. To date, the technology has largely been used for the heterologous expression of resistance genes (e.g., Bacillus thuringiensis d-endotoxin ) in agricultural systems (see Tian et al. [2003] for an elegant exception), and therefore has provided little evidence for the defensive value of endogenously expressed traits against a plant's native herbivore community. The scientific value of transgenically silencing endogenous genes in native plants to understand the ecological function of particular genes has been undermined by the polarized attitudes towards the use of genetically modified organisms in agriculture. Transgenic down-regulation of nicotine demonstrates that N. attenuata is under relentless herbivore pressure. Disabling this resistance trait, even in a year of low herbivore abundance, results in a large increase in opportunistic herbivory and supports the conclusion that secondary metabolites play an important role in explaining why the earth is largely green ( Hairston et al. 1960 ). Materials and Methods Plant material and transformation N. attenuata Torr. ex Watson (synonymous with N. torreyana Nelson and Macbr.; Solanaceae) grown from field-collected seeds ( Baldwin 1998 ) and inbred 11 or 14 generations were used for transformation and all experiments. Seed germination and the Agrobacterium tumifaciens (strain LBA 4404)–mediated transformation procedure are described in Krügel et al. (2002) . In order to silence the expression of the two N. attenuata pmt genes, plants were transformed with pCAMPMT1 and pNATPMT1 vectors, which contain a gene fragment of pmt1 (which has 95% identity to pmt2 ) in an antisense orientation, and pRESC5PMT, which contains the pmt gene fragment twice in an inverted orientation separated by intron 3 of the Flaveria trinervia gene pyruvate orthophosphate dikinase (pdk) (for vector construction and plasmids see Figure S5 and Protocol S1 ). T 1 plants were screened for hygromycin resistance ( hygromycin phosphotransferase II gene of the vector pCAMBIA-1301) and constitutive and induced nicotine accumulation; homozygosity was determined by resistance screening of the T 2 plants. Two independently transformed homozygous IR pmt lines (108 and 145) were further characterized by Southern blot analysis and by the rescuing of the transformation vector from genomic DNA into Escherichia coli to identify copy number and insertion site of the T-DNA (see Figure S1 and Protocol S1 ). PMT mRNA accumulation and secondary metabolites. Transformed lines (108 and 145) and WT plants were grown in 1-l hydroponic vessels in a climate chamber as described in Hermsmeier et al. (2001) , and 4-wk-old rosette-stage plants were treated (elicited) on the first two fully expanded (source) leaves with 150 μg of MeJA per plant applied in 20 μl of lanolin paste, or left untreated. Approximately 200 mg of young roots was harvested and frozen in liquid nitrogen 10 h after elicitation, and RNA was extracted with Tri Reagent (Sigma, Taufkirchen, Germany) according to the manufacturer's instructions ( n = 3/line/treatment). PMT transcript accumulation was analyzed by real-time PCR (ABI PRISM 7000; Applied Biosystems, Darmstadt, Germany). cDNA was generated from 20 ng of RNA with MultiScribe reverse transcriptase (Applied Biosystems), and amplified using the qPCR core reagent kit (Eurogentec, Searing, Belgium) and a probe and primers that were gene-specific (for sequences see Figure S6 ). For analysis of secondary metabolites, leaves growing one node above the sink-source transition leaf and young root tissue were harvested 4 d after elicitation ( n = 8–10/line/treatment). Samples were analyzed by HPLC for alkaloids, caffeoylputrescine, chlorogenic acid, and rutin ( Keinänen et al. 2001 ; Halitschke and Baldwin 2003 ). A peak occurring in IR pmt alkaloid extracts but not in extracts of WT N. attenuata was collected and identified by nuclear magnetic resonance imaging as anatabine (for spectra and method, see Protocol S1 ). To determine whether a NA oversupply was responsible for the formation of anatabine in the transformed lines, we supplied 4-wk-old plants with either unlabeled or D 4 -NA ethyl ester (1 mM) in their hydroponic solution 24 h after MeJA elicitation ( n = 4/line/treatment). After 4 d, the treated leaf was harvested and extracted as above, but analyzed by LC/MS/MS to detect incorporation of the deuterium into nicotine and anatabine (for instrument conditions, see Protocol S1 ). To examine the release of cis -α-bergamotene in the transformed lines compared to WT, volatiles from hydroponically grown plants ( n = 3–5/line/treatment) enclosed in open-top volatile collection chambers were collected for an 8 h period starting 24 h after MeJA elicitation of the first two source leaves, and analyzed by GC/MS ( Halitschke et al. 2000 ). TPI activity in the MeJA-treated leaf 3 d after elicitation was analyzed in plants ( n = 5/line/treatment) by radial diffusion activity assay ( van Dam et al. 2001 ). M. sexta performance and feeding choice In the glasshouse, 2-wk-old seedlings were planted individually into 2-l pots with potting soil (C 410; Stender, Schermbeck, Germany) at 26–28 °C under 16-h supplemental light from Philips Sun-T Agro 400- or 600-W Na lights. For analysis of performance, newly eclosed M. sexta larvae (North Carolina State University, Raleigh, North Carolina, United States) were placed on the first-stem leaf of 8-wk-old WT and IR pmt (108 and 145) plants and allowed to feed for 14 d. Larval mass was recorded at 8, 10, 12, and 14 d. The first feeding choice of M. sexta was determined by placing newly eclosed larvae in the center of a 3-cm–diameter cup containing, on opposite sides, 1.5-cm 2 WT and IR pmt (108) leaf pieces and recording the leaf on which larvae started feeding ( n = 44). Resistance of WT and IR pmt plants to herbivores in the natural habitat. In a field plantation (15 m × 18 m; GPS: lat 37°08′45′′N, long 114°01′12′′) in N. attenuata 's natural habitat in southwest Utah, transformed IR pmt (108) and WT plants were exposed to naturally occurring herbivores dispersing from adjacent populations. To allow for spatial heterogeneity, plants were transplanted in a paired design (with 0.3 m and 1.5 m between plants of a pair and between pairs, respectively) in which plants were matched for equal rosette diameters. Plants were grown in soil (Potting Mix; Miracle-Gro, Marysville, Ohio, United States) for 5 wk after germination ( Krügel et al. 2002 ), and were transplanted into the field plot (10 columns by 15 lines) in their 3.8-l pots. Seven days after transplantation, 30 WT and IR pmt plants were elicited with 150 μg of MeJA per plant applied in 20 μl of lanolin paste to the two youngest rosette leaves. Starting 4 d after transplantation, each plant was examined for damage and insects (including predators and eggs) every other day for 14 d. Damage amount was estimated as a percentage of the total leaf area, and the characteristic damage caused by caterpillars, beetles, grasshoppers, myrids, and leafhoppers was noted separately. The most abundant herbivores observed in the field plantation during the release were S. exigua , Trimerotropis spp., E. hirtipennis, and D. undecimpunctata. M. sexta and M. quinquemaculata occurred in the season only rarely, and no eggs were laid in the plantation during the 14 d. As plants began to elongate and produce flowers, they were examined daily, and all flowers were removed before opening and anthesis to meet the performance standards in the Code of Federal Regulations (7CFR340.3c). Consequently, direct fitness measures were unobtainable in this experiment. For analysis of alkaloids and TPIs under field conditions, leaf samples of WT and IR pmt plants in the plot ( n = 6) were taken 7 wk after transplantation and frozen (dry ice). To determine if the herbivore phenotype of IR pmt plants observed in glasshouse-grown plants was retained in plants grown under natural light conditions, the M. sexta choice experiment was repeated. The first feeding choices of freshly eclosed M. sexta larvae (North Carolina State University) and of adults of field-collected D. undecimpunctata and Trichobarus mucorea (Chrysomelidae and Curculionidae) found on N. attenuata were determined as described above. Supporting Information Figure S1 Copy Number of T-DNA in the Two Studied IR pmt Lines (A) Southern blot analysis of two independently transformed N. attenuata IR pmt lines (108 and 145) and WT plants. Genomic DNA (15 μg) from individual plants of the three genotypes and the plasmid used for transformation pRESC5PMT (4 ng) were digested with EcoRV and blotted onto nylon membranes ( Winz and Baldwin 2001 ). The blot was hybridized with a PCR fragment of the hygromycin phosphotransferase II gene from pCAMBIA-1301, which is specific for the selective marker on the T-DNA and signifies one insertion in each of the two lines. (B) Ethidium bromide staining of the DNA revealed an overload of the DNA of the IR pmt lines and therefore loading of the WT was controlled by stripping and rehybridization with a PMT probe, which clearly revealed the endogenous pmt1 and pmt 2 genes described ( Winz and Baldwin 2001 ) (unpublished data). (6.3 MB TIF). Click here for additional data file. Figure S2 Secondary Metabolite Levels in the Studied IR pmt Lines Inverted-repeat silencing of pmt did not change the levels of (A) anabasine, (B) caffeoylputrescine, (C) chlorogenic acid, and (D) rutin (mean ± standard error [SE]) in two independently transformed N. attenuata lines (108 and 145) compared to WT plants. Plants were harvested 4 d after receiving one of four treatments: untreated control (Con), wounding (W), wounding and regurgitate application (W+R), and application of 150 μg of MeJA per plant applied in a lanolin paste. Plants were treated at the first two fully expanded (source) leaves and wounding was performed by generating three rows of puncture wounds on each leaf side using a pattern wheel. Subsequently, 10 μl per leaf of either water or M. sexta regurgitate diluted 1:1 (v:v) was dispersed over the puncture wounds ( n = 8–10). (179 KB PPT). Click here for additional data file. Figure S3 Proteinase Inhibitor and Volatile Emission of the Studied IR pmt Lines Levels of (A) TPI and (B) cis- α-bergamotene emission (mean ± SE) in two independently transformed N. attenuata IR pmt lines (108 and 145) did not differ from WT plants 4 d (for TPI) and 10 h (for cis- α-bergamotene) after receiving one of four treatments (as described for S2): untreated control (Con), wounding (W), wounding with additional regurgitate application (W+R), and MeJA elicitation. IS, internal standard. (73 KB PPT). Click here for additional data file. Figure S4 Growth Parameters Under Glasshouse and Field Conditions of the Studied IR pmt Lines N. attenuata plants transformed with an IR pmt construct (108 or 145) did not differ in (A) stalk length [ n PMT = 43, n WT = 57, n PMT* and n WT* = 28] and (B) rosette diameter [ n = 8] from WT grown under either field (A) or glasshouse (B) conditions. Plants in (A) were untreated or elicited (*) with MeJA 7 d after plants were transplanted into a field plot in a native habitat. (98 KB PPT). Click here for additional data file. Figure S5 Transformation Vectors This figure shows plasmids used for the generation of N. attenuata lines with reduced levels of two PMTs due to posttranscriptional gene silencing. Both (A) pCAMPMT1 (10.7 kb) and (B) pNATPMT1 (9.7 kb) allow the synthesis of pmt antisense RNA. (C) pRESC5PMT (12.4 kb) was used for the synthesis of pmt RNA capable of forming an inverted repeat. Functional elements: bla , beta-lactamase gene from plasmid pUC19; hptII , gene for hygromycin resistance from pCAMBIA-1301; LB and RB, left and right border of T-DNA; nptIII , aminoglycoside phosphotransferase of type III from Streptococcus faecalis ; ori ColE1, origin of replication from pUC19; ori pVS1, origin of replication from plasmid pVS1; P CaMV and T CaMV , 35S promoter and terminator of cauliflower mosaic virus; pdk i3 , intron 3 of pdk; pmt1 , gene fragment of pmt1 (95% identical with N. attenuata pmt2 ); P NOS and T NOS , promoter and terminator of the nopaline synthase gene; repA pVS1, replication protein gene from pVS1; sat-1 , nourseothricin resistance gene; staA pVS1, partitioning protein gene from pVS1. Displayed restriction sites mark the borders of functional elements, which are displayed in gray if on the T-DNA and in black if outside the T-DNA. (56 KB PPT). Click here for additional data file. Figure S6 PMT Sequences and TaqMan Probe Nucleotide sequences of N. attenuata pmt1 and pmt2 mRNA ( Winz and Baldwin 2001 ) aligned with ClustalW. Primers and probe (underlined) used for real-time PCR of pmt mRNA are highlighted and bold. (396 KB TIF). Click here for additional data file. Protocol S1 Molecular and Analytical Methods (58 KB DOC). Click here for additional data file. Accession Numbers GenBank accession numbers for the genes discussed in this paper are bla from puc19 (L09137), hygromycin phosphotransferase II from pCAMBIA-1301 (AF234297), pdk (X79095), pmt1 (AF280402), and pmt2 (AF280403).	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC509292.xml
517937	Non-Invasive monitoring of diaphragmatic timing by means of surface contact sensors: An experimental study in dogs	Background Non-invasive monitoring of respiratory muscle function is an area of increasing research interest, resulting in the appearance of new monitoring devices, one of these being piezoelectric contact sensors. The present study was designed to test whether the use of piezoelectric contact (non-invasive) sensors could be useful in respiratory monitoring, in particular in measuring the timing of diaphragmatic contraction. Methods Experiments were performed in an animal model: three pentobarbital anesthetized mongrel dogs. The motion of the thoracic cage was acquired by means of a piezoelectric contact sensor placed on the costal wall. This signal is compared with direct measurements of the diaphragmatic muscle length, made by sonomicrometry. Furthermore, to assess the diaphragmatic function other respiratory signals were acquired: respiratory airflow and transdiaphragmatic pressure. Diaphragm contraction time was estimated with these four signals. Using diaphragm length signal as reference, contraction times estimated with the other three signals were compared with the contraction time estimated with diaphragm length signal. Results The contraction time estimated with the TM signal tends to give a reading 0.06 seconds lower than the measure made with the DL signal (-0.21 and 0.00 for FL and DP signals, respectively), with a standard deviation of 0.05 seconds (0.08 and 0.06 for FL and DP signals, respectively). Correlation coefficients indicated a close link between time contraction estimated with TM signal and contraction time estimated with DL signal (a Pearson correlation coefficient of 0.98, a reliability coefficient of 0.95, a slope of 1.01 and a Spearman's rank-order coefficient of 0.98). In general, correlation coefficients and mean and standard deviation of the difference were better in the inspiratory load respiratory test than in spontaneous ventilation tests. Conclusion The technique presented in this work provides a non-invasive method to assess the timing of diaphragmatic contraction in canines, using a piezoelectric contact sensor placed on the costal wall.	Background Non-invasive monitoring of respiratory function is an area of increasing research interest, resulting in the appearance of new monitoring devices [ 1 ]. At present, the most utilised non-invasive method for continuous quantitative monitoring of breathing pattern is respiratory inductive plethysmography. This technique allows the study of various breathing pattern parameters such as respiratory frequency, but it is based on an averaged measurement of the whole thoracic-abdominal movement. On the other hand, the use of mouthpieces or face masks in pneumotacography, influences in the tidal volume and respiratory frequency [ 2 - 4 ]. Other systems like piezoelectric contact sensors measure the system acceleration when placed on body surfaces. In previous works we have shown that the beginning and end of diaphragmatic contraction can be determined by inflexion points in the thoracic cage motion signal acquired with a contact sensor [ 5 - 7 ]. The purpose of this study is to evaluate a non-invasive method to study the timing of the diaphragmatic function, using an animal model (dogs). Accordingly, the present study was designed to test whether the use of contact (non-invasive) piezoelectric sensors, placed on the dogs' costal wall, could be useful in monitoring the diaphragm contraction period in different respiratory conditions, comparing it with other physiological signals such as transdiaphragmatic pressure, diaphragm length measured by sonomicrometry, and respìratory airflow. Diaphragm contraction time is expected to be very close to inspiratory time, which is one of the most utilized parameters in the studies of breathing pattern under experimental or clinical conditions. Methods Three mongrel dogs (15–20 kg) were surgically instrumented under general anesthesia given via a femoral vein catheter (pentobarbital sodium, 25 mg/kg). Respiratory flow was recorded with a Fleisch pneumotachograph. Diaphragm shortening was measured via two piezoelectric crystals (Sonomicrometer, Triton Tech. Inc., m. 120), as described in [ 8 ]. The diaphragm was exposed by a midline abdominal incision, and the two piezoelectric crystals were implanted along the rib diaphragm fibres. An anterior midline incision was made in the neck to allow the left C5 and C6 phrenic nerve roots to be isolated. Motion of the thoracic cage surface was recorded by a piezoelectric contact sensor (HP 21050A) positioned on the costal wall and fixed to the skin by an elastic band. Maximal deflection of the accelerometer following a unilateral phrenic nerve electrical pulse was measured on the 6–7 intercostal space area of the rib cage, where the diaphragmatic fibres are directly apposed to its inner surface, thereby minimizing the distance between the accelerometer and the muscle. Transdiaphragmatic pressure was measured in the usual way as the difference between gastric and esophageal pressures, each recorded with the conventional balloon-catheter technique [ 9 ]. The electromyogram of the diaphragm (EMGdi) was recorded with two (parallel) 10 mm long single filament copper wires (1 mm in diameter) attached 20 mm apart on a semi-rigid plastic plate, as described in [ 10 ]. Measurements were made at a similar level of anaesthesia (corneal reflex just suppressed). All animals were in supine position during the study, and spinal anaesthesia was applied as a means to isolate diaphragmatic function by eliminating the activity of the intercostal muscles. To that effect, with the animal lying down with its head and neck raised, a hyperbaric Tetracaine solution (Sigma) was injected into the subarachnoid space at the lumbar level (bolus of 1 ml). The injection of tetracaine was halted when intercostals EMG activity was abolished below intercostals spaces 3–4 on both sides of the cage (as determined by needle electrodes recording from the parasternals). With the head and neck elevated, the animal was then turned over to the supine position. All dogs performed spontaneous ventilation before the use of spinal anaesthesia (SVN), and spontaneous ventilation (SVW) and respiration with an inspiratory resistive load (ILW) with spinal anaesthesia. The duration and respiratory frequency of the tests varies in function of the dog analyzed. Table I shows the number of cycles and the duration of the respiratory tests performed by the three dogs. Table 1 Number of cycles and duration of the respiratory tests in the three dogs. Spontaneous ventilations without anaesthesia (SVN) Spontaneous ventilations with anaesthesia (SVW) Inspiratory load with anaesthesia (ILW) No. cycles Duration (s) No. cycles Duration (s) No. cycles Duration (s) DOG1 45 180 40 160 159 400 DOG2 19 180 29 140 194 600 DOG3 11 60 8 30 46 220 All analogical signals were amplified (HP 8802A), filtered and digitised with a 12 bit A/D system at a sampling rate of 4 kHz. Inspiratory airflow (FL), diaphragm length (DL), thoracic cage motion (TM), transdiaphragmatic pressure (DP) and diaphragmatic electromyography (EMGdi) signals were decimated at a new sampling rate (FL, DL, DP: 100 Hz ; TM: 200 Hz; EMGdi: 1200 Hz) and digitally filtered (FL, DL, DP: DC-40 Hz ; TM: DC-80 Hz; EMGdi: 10–480 Hz). The sampling frequencies and filter bands were selected to be adapted to the frequency content of the signals. Figure 1 shows a typical strip-chart recording of the five signal acquired, during 8 seconds (2 respiratory cycles) of the ILW respiratory test of the second dog. Figure 1 Respiratory Signals. Example of the five signals acquired during 8 seconds (two respiratory cycles) during the inspiratory load with spinal anaesthesia (ILW) respiratory test of the second dog. In order to detect the initial and final diaphragm contraction times ( t i , t f ), the integral of TM signal and the first derivative of the DL, FL and DP signals were computed. Initial contraction time is detected when these signals reach 10 % of the maximum. In a similar way, final contraction time is detected when the signals reach 10 % of their minimum. We also computed contraction period ( T C = t f - t i ). The EMGdi signal was included initially in the study but later was rejected because the postinspiratory activity present in this signal hindered the detection of the end of the diaphragm shortening contraction time (as seen in Fig. 1 ). One representative experimental record of TM, DL, FL and DP signals is shown in Fig. 2 . Furthermore, the integral of the TM signal is presented, as well as the first derivative of the DL, FL and DP signals. Figure 2 Diaphragmatic contraction time detection. Morphologies of thoracic cage motion signal and its integral (a), diaphragmatic length (b), respiratory airflow (c) and transdiaphragmatic pressure (d) and their first derivatives. Vertical lines mark initial and final instants of diaphragmatic contraction. Statistical Analysis Differences between contraction periods obtained with DL signal and the other three signals were summarized by the mean (MEAN), standard deviation (STD). DL signal was used as a goldstandard (reference signal) for muscle shortening, since it is a direct measure of the reduction of the diaphragm muscle length, contrary to flow and trandiaphragmatic pressure which are more remote markers of muscle contraction. Relationships between contraction periods obtained by means DL and TM signals, DL and FL, and DL and DP were analyzed by diverse statistical methods. Different correlation coefficient were calculated: the Pearson correlation coefficient (r), the intra-class correlation coefficient (or reliability coefficient: R) [ 11 ], the slope of the linear regression line (p), and the Spearman's rank-order correlation coefficient (r). Furthermore the Bland-Altman method for agreement analysis was performed [ 12 ]; in this graphical method the differences between two measures or techniques are plotted against the averages of the two techniques. Results The obtained results of the comparison of contraction period estimated with the DL signal (a direct measure of diaphragm shortening) with the contraction periods estimated with the TM, FL and DP signals, are shown in Tables 2 , 3 and 4 , and in Figs. 3 and 4 . Table 2 Differences between contraction period measured with diaphragmatic length (DL) and contraction period measured with thoracic motion (TM), respiratory airflow (FL) and transdiaphragmatic pressure (DP). DL vs. TM DL vs. FL DL vs. DP MEAN (s) SD(s) MEAN - 2SD (s) MEAN + 2SD (s) MEAN (s) SD(s) MEAN - 2SD (s) MEAN + 2SD (s) MEAN (s) SD(s) MEAN - 2SD (s) MEAN + 2SD (s) Dog 1 (SVN) 0.15 0.05 0.05 0.26 -0.25 0.08 -0.39 -0.01 -0.19 0.07 -0.32 -0.05 Dog 2 (SVN) 0.03 0.03 -0.02 0.09 -0.00 0.04 -0.09 0.08 0.01 0.03 -0.05 0.06 Dog 3 (SVN) 0.04 0.03 -0.01 0.10 0.03 0.04 -0.05 0.11 -0.00 0.04 -0.07 0.07 Dog 1 (SVW) -0.00 0.03 -0.05 0.06 0.02 0.03 -0.04 0.08 0.02 0.04 -0.05 0.10 Dog 2 (SVW) 0.03 0.03 -0.03 0.08 -0.04 0.04 -0.12 0.04 -0.02 0.02 -0.06 0.02 Dog 3 (SVW) -0.03 0.02 -0.08 0.02 -0.00 0.03 -0.06 0.05 -0.02 0.05 -0.11 0.08 Dog 1 (ILW) 0.09 0.02 -0.04 0.13 0.02 0.02 -0.02 0.06 0.03 0.02 -0.00 0.07 Dog 2 (ILW) 0.04 0.02 -0.00 0.08 -0.01 0.05 -0.10 0.08 0.01 0.02 -0.02 0.04 Dog 3 (ILW) -0.01 0.02 -0.04 0.02 -0.01 0.02 -0.04 0.03 0.04 0.02 -0.00 0.08 MEAN: Mean difference; SD: standard deviation; MEAN ± 2SD: limits of agreement of the Blond and Altman analysis; SVN: spontaneous ventilation without anaesthesia respiratory test, SVW: spontaneous ventilation with anaesthesia respiratory test; ILW: inspiratory load with anaesthesia respiratory test. Table 3 Correlation coefficients between contraction period measured with diaphragmatic length (DL) and contraction period measured with thoracic motion (TM), respiratory airflow (FL) and transdiaphragmatic pressure (DP). DL vs. TM DL vs. FL DL vs. DP r R p ρ r R p ρ r R p ρ Dog 1 (SVN) 0.61 -0.44 0.43 0.58 0.12 -0.82 0.71 0.23 0.14 -0.74 0.059 0.15 Dog 2 (SVN) 0.91 0.77 1.00 0.84 0.72 0.73 0.65 0.81 0.90 0.89 0.76 0.87 Dog 3 (SVN) 0.77 0.39 0.68 0.79 0.63 0.45 0.72 0.65 058. 0.56 0.38 0.53 Dog 1 (SVW) 0.62 0.62 0.60 0.64 0.54 0.39 0.42 0.55 0.37 0.23 0.32 0.42 Dog 2 (SVW) 0.69 0.47 0.80 0.66 0.51 0.15 0.74 0.51 0.85 0.69 0.94 0.83 Dog 3 (SVW) 0.60 0.23 0.55 0.58 0.59 0.62 0.71 0.56 0.26 0.24 0.46 0.32 Dog 1 (ILW) 0.99 0.92 0.99 0.99 1.00 0.99 1.00 0.99 1.00 0.98 0.99 0.99 Dog 2 (ILW) 0.97 0.79 1.04 0.96 0.91 0.85 1.31 0.93 0.98 0.97 1.00 0.96 Dog 3 (ILW) 0.98 0.97 0.94 0.98 0.98 0.97 1.01 0.97 0.97 0.86 0.99 0.97 r: Pearson correlation coefficient; R: reliability coefficient; p: slope of the linear regression line; ρ : Spearman's rank-order correlation coefficient; SVN: spontaneous ventilation without anaesthesia respiratory test, SVW: spontaneous ventilation with anaesthesia respiratory test; ILW: inspiratory load with anaesthesia respiratory test. Table 4 Differences and correlation coefficients between contraction period measured with diaphragmatic length (DL) and contraction period measured with thoracic motion (TM), respiratory airflow (FL) and transdiaphragmatic pressure (DP), for all the respiratory cycles analysed. r R p ρ MEAN(s) SD(s) MEAN - 2SD (s) MEAN + 2SD (s) DL vs TM 1 0.98 0.95 1.01 0.98 0.055 0.050 -0.045 0. 156 DL vs FL 1 0.94 0.93 0.86 0.91 -0.021 0.080 -0. 180 0. 138 DL vs DP 1 0.96 0.96 0.86 0.94 0.001 0.064 -0. 127 0. 129 DL vs TM 2 0.99 0.96 0.96 0.98 0.044 0.040 -0.032 0.125 DL vs FL 2 0.99 0.99 0.97 0.98 -0.001 0.040 -0.080 0.077 DL vs DP 2 0.99 0.99 0.94 0.99 0.018 0.026 -0. 034 0.070 1 Including all the respiratory cycles analysed; 2 Without the respiratory cycles corresponding to the spontaneous ventilations without anaesthesia test. r: Pearson correlation coefficient; R: reliability coefficient; p: slope of the linear regression line; ρ : Spearman's rank-order correlation coefficient; MEAN: Mean difference; SD: standard deviation; MEAN ± 2SD: limits of agreement of the Blond and Altman analysis; SVN: spontaneous ventilation without anaesthesia respiratory test, SVW: spontaneous ventilation with anaesthesia respiratory test; ILW: inspiratory load with anaesthesia respiratory test. Figure 3 Relationships and Bland and Altman analysis. Relationship and Bland and Altman plot of the contraction period estimated with thoracic motion (TC-TM), respiratory airflow (TC-FL) and transdiaphragmatic pressure (TC-DP) signals versus contraction period estimated with diaphragmatic length signal (TC-DL), for the three dogs in the spontaneous ventilations without anaesthesia (SVN), spontaneous ventilations with anaesthesia (SVW) and inspiratory load with anaesthesia (ILW) respiratory tests. The solid black continuous line is the identity function (desired relationship). Each dot represents a respiratory cycle. Figure 4 Relationships and Bland and Altman analysis (all cycles together). Relationship and Bland and Altman plot of the contraction period estimated with thoracic motion (TC-TM), respiratory airflow (TC-FL) and transdiaphragmatic pressure (TC-DP) signals versus contraction period estimated with diaphragmatic length signal (TC-DL), for all the respiratory cycles analyzed in the three dogs and the three respiratory tests. The solid black continuous line is the identity function (desired relationship). Each dot represents a respiratory cycle. The cycles corresponding to the spontaneous ventilation without spinal anaesthesia respiratory test of the Dog 1 are encircled in the Bland and Altman plot. In the Fig. 3 is shown graphically the relationship among the periods of contraction obtained by means the different estimation methods, for each animal and each respiratory test. This relationship is showed in two formats: a plot of the data with the line of equality (all points would lie in this line), and a Bland and Altman plot [ 12 ] with the mean difference and agreement limit lines. In Table 2 , are shown the values of the mean difference (MEAN), the SD of the difference, and the agreement limits of the Bland and Altman plots. The mean error obtained with the three indirect measures of diaphragm shortening were lower than 0.1 seconds and the SD of the difference was lower than 0.05 seconds, except in the SVN test of Dog 1 (in this test it has been observed a great variability in the values obtained by means the four signal analyzed, as is could be seen in the first graph of Fig. 5 ). Furthermore the Bland and Altman agreement limits are always lower than 0.13 seconds. Figure 5 Diaphragmatic contraction period monitoring. Contraction period (TC) estimated with thoracic cage motion (MT: thin solid line), diaphragm length (DL: thick solid line), respiratory airflow (FL: dotted line) and transdiaphragmatic pressure (DP: dashed line) signals versus duration of the respiratory tests in seconds, for the three dogs in the spontaneous ventilations without anaesthesia (SVN), spontaneous ventilations with anaesthesia (SVW) and inspiratory load with anaesthesia (ILW) respiratory tests. Table 3 shows values of the correlation coefficient (r), the reliability coefficient (R), the slope of the linear regression line (p), and the Spearman's rank correlation coefficient between DL and TM contraction periods, between DL and FL contraction periods, and between DL and DP contraction periods for the three dogs for SVN, SVW and ILW respiratory tests. The relationships in the ILW respiratory test were nearly linear (r > 0.91), with reliability coefficients indicating a high reliability in the measurements (R > 0.79), slopes of the linear regression line very close to equality line (except in the flow of the second dog), and Spearman rank-order coefficient showing a strong link between the variables analyzed (r > 0.93). In the SVN and SVW respiratory tests (except for the SVN test of the first dog), results showed a moderate relationship, but, in general, correlation coefficients estimated in TM signal were better than estimations in FL and DP signal. Relationship between the contraction period estimated with TM signal and contraction period estimated with DL signal showed Pearson correlation coefficients between 0.60 and 0.91, reliability coefficients between 0.23 and 0.77, slope of the linear regression lines between 0.55 and 1, and Spearman's rank-order correlation coefficient between 0.58 and 0.84. In Fig. 4 and Table 4 are shown the results obtained analyzing all the respiratory cycles together. In both correlation analysis and Bland-Altman plots it is seen that respiratory cycles corresponding to the SVN test of Dog 1 (marked with a circle in Fig. 4 ) have different behaviour than the rest. For that reason in the first 3 rows of Table 4 it are shown the parameters obtained with all the cycles and in the last 3 rows are shown the parameters obtained excluding the respiratory cycles of the SVN test of the first dog. Excluding these respiratory cycles, the contraction period estimated with the TM signal tends to give a lower reading than the measure made with the DL signal, with a mean of 0.04 seconds (0.06 without the exclusion), a standard deviation of 0.04 seconds (0.05 without the exclusion), and limits of agreement between -0.03 and 0.12 seconds (between -0.05 and 0.16 without the exclusion). These results are slightly worse than the results obtained from the comparison with the FL signal or the DP signal. The correlation coefficients are very similar in the three signals. Finally, Figure 5 shows the evolution of the diaphragm contraction periods estimated with the four signals throughout the SVN, SVW and ILW respiratory tests for the three dogs studied. It is seen that, in all cases, the behaviour of contraction time estimated with the four signals is very similar (although the MEAN difference of the SVN test of the first dog is unsatisfactorily great, as seen in Table 2 ). Discussion In the present work we have compared the rib cage motion recorded by surface sensors with the changes in diaphragm activity registered by sonomicrometry, transdiaphragmatic pressure and airflow recorded in dogs during spontaneous and inspiratory load breathing. Diaphragmatic time contraction measured with a surface sensor has a good correlation with the rest of signals, especially during the inspiratory load test. In a recent study we observed that the beginning and the end of diaphragmatic contraction were indicated with inflexion points in the thoracic cage motion signal, acquired with a piezoelectric contact sensor placed on the costal wall of the thorax [ 6 , 7 ]. An algorithm was implemented and validated to detect the initial and final instants of diaphragm contraction, and the results were compared with the direct measurement of the diaphragmatic muscle length changes made by sonomicrometry [ 8 ]. In the present work we have compared the contact sensor signal with other transducer signals to test the monitoring capacity of contact sensors in different respiratory patterns. Three different tests have been studied: spontaneous ventilation before the use of spinal anaesthesia, spontaneous ventilation (SVW) after spinal anaesthesia, and breathing through a resistive inspiratory load (ILW). Spinal anaesthesia was used in the present study as a means to isolate diaphragmatic activity by eliminating the activity of the intercostals muscles. In this way, TM, FL and DP signals are directly related with the contraction of diaphragm muscle in the SVW and ILW respiratory tests (as well as DP signal). In the SVN test the morphologies of TM, FL and DP signals are influenced by the activity of intercostal muscles. However, time differences between contraction periods measured in SVN and SVW respiratory tests were very similar (except for the SVN test of the first dog). This could be explained taking into account that during spontaneous breathing the activity of intercostals and respiratory accessory muscles is minimal [ 6 , 7 ]. A close relationship between the different methods of diaphragm contraction time estimation has been found during the inspiratory load test. However, during spontaneous ventilation, the correlation was lower than that obtained in the resistive load test. This difference could be explained by the fact that during spontaneous ventilation, the motion of the respiratory system is very low and in consequence, the signal recorded by the contact sensor (and in general, all the respiratory signals) has not sufficient intensity to detect with the same precision the beginning and ending of diaphragmatic activity. Besides, during inspiratory resistive loading, there is a marked distortion of the rib cage (in particular when the intercostal muscles are paralyzed) and this produces a more marked thoracic-cage motion signal. Another important finding of this work was that time contraction differences between signals were less than 0.1 s for spontaneous ventilation and inspiratory load (except in the irregular case of the SVN respiratory test of the first dog), which indicates that the application of contact sensors constitutes an indirect way to detect the diaphragmatic activity. This could be seen in the graphs of Fig. 5 , in which we observed that practically in all cases the behavior of the diaphragmatic contraction time estimated with the four signals is very similar, being appropriate for diaphragmatic contraction period monitoring. The TM signal has the advantage with respect to other signals in that it is non-invasive and does not affect the breathing pattern [ 2 - 4 ]. Therefore, it is suitable for continuous monitoring of breathing pattern parameters such as respiratory frequency and diaphragmatic contraction time (which is very close to inspiratory time). Thus, the usefulness of TM signal for non-invasive diaphragmatic contraction period monitoring is demonstrated. A particular case is the SVN test of the first dog. In this case the breathing was very weak, causing that beginning of the contraction was very slow and irregular. This generated difficulties to determine the beginning of the contraction, provoking great differences in the contraction time estimated with the four signals. Nevertheless, the evolution of contraction time throughout the respiratory tests is very similar. The contraction of the costal diaphragm acts to displace the rib cage through its insertions at the costal margins and by changing the pressure on the inner surface of the rib cage in the area of apposition. The crural diaphragm is not inserted on the rib cage, but it is considered to have an action through the central tendon. Therefore, in spontaneous respiration the crural part has an important inflationary action on the lower rib cage [ 13 ], and although we only registered costal muscle length changes, we believe that the contact sensor shows the activity of both diaphragm components. Finally, to acquire the diaphragm shortening signal, it has been necessary to surgically isolate their diaphragms. The effect of diaphragm isolation could be to favour the operation of the piezoelectric contact sensor to measure diaphragmatic contraction timing. This should be kept in mind when extrapolating these results to human. Conclusions The technique presented in this work represents a non-invasive method to assess the timing of diaphragm contraction in dogs. We believe that in the future this technique could provide a new potentially useful method for non-invasive respiratory timing monitoring in humans. Competing interests None declared Authors' contributions JAF, JM, RJ and AG conceived the study, and participated in its design and coordination. JAF and AG designed and conducted the experiments. AT participated in the design of the study and performed the signal processing and statistical analysis. BG and JG provided advice on analysis of the data and manuscript writing. JAF, AT and RJ wrote the first draft of this manuscript. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517937.xml
449846	Activation-Induced Cytidine Deaminase Initiates Immunoglobulin Gene Conversion and Hypermutation by a Common Intermediate	Depending on the species and the lymphoid organ, activation-induced cytidine deaminase (AID) expression triggers diversification of the rearranged immunoglobulin (Ig) genes by pseudo V (ψV) gene- templated gene conversion or somatic hypermutation. To investigate how AID can alternatively induce recombination or hypermutation, ψV gene deletions were introduced into the rearranged light chain locus of the DT40 B-cell line. We show that the stepwise removal of the ψV donors not only reduces and eventually abolishes Ig gene conversion, but also activates AID-dependent Ig hypermutation. This strongly supports a model in which AID induces a common modification in the rearranged V(D)J segment, leading to a conversion tract in the presence of nearby donor sequences and to a point mutation in their absence.	Introduction Immunoglobulin (Ig) genes are further diversified after V(D)J rearrangement by gene conversion, hypermutation, or a combination of the two. Surprisingly, even closely related species employ different strategies: mice and humans use exclusively hypermutation ( Milstein and Rada 1995 ), whereas rabbits, cows, and pigs use mainly gene conversion ( Butler 1998 ). The balance between the two phenomena can also shift during differentiation: for example, chicken B-cells first develop their Ig repertoire by gene conversion in the bursa ( Reynaud et al. 1987 ; Arakawa and Buerstedde 2004 ) and later fine tune it by hypermutation in splenic germinal centers ( Arakawa et al. 1996 ). All three B-cell specific activities of Ig repertoire formation—gene conversion ( Arakawa et al. 2002 ), hypermutation, and isotype switch recombination ( Muramatsu et al. 2000 ; Revy et al. 2000 )—require expression of the activation-induced cytidine deaminase (AID) gene. Whereas it was initially proposed that AID is an mRNA editing enzyme ( Muramatsu et al. 1999 ), more recent studies indicate that AID directly modifies DNA by deamination of cytosine to uracil ( Di Noia and Neuberger 2002 ). However, the cytosine deamination activity must be further regulated, because only differences in the type, the location, or the processing of the AID-induced DNA modification can explain the selective occurrence of recombination or hypermutation in different species and B-cell environments. Based on the finding that certain AID mutations affect switch recombination but not somatic hypermutation, it was suggested that AID needs the binding of a cofactor to start switch recombination ( Barreto et al. 2003 ; Ta et al. 2003 ). Analysis of knockout mutants of the chicken B-cell line DT40 indicate that the RAD54 gene ( Bezzubova et al. 1997 ) and other members of the RAD52 recombination repair pathway are needed for efficient Ig gene conversion ( Sale et al. 2001 ). Most interestingly, disruption of RAD51 paralogs reduces Ig gene conversion and induces hypermutation in the rearranged light chain gene ( Sale et al. 2001 ), suggesting that a defect in DNA repair by homologous recombination can shift Ig gene conversion to hypermutation. Valuable insight into complex recombination processes has been gained by the genetic and biochemical analysis of reaction intermediates ( Haber 1998 ). Since sequence information needs to be copied from the donor to the target at some stage of Ig gene conversion, we reasoned that the deletion of the donor sequences might arrest the reaction and allow the recovery of an intermediate. Here we report that ablation of pseudo V (ψV) donors activates AID-dependent Ig hypermutation in DT40 cells. This shows that Ig gene conversion and hypermutation are competing pathways derived from the same AID-initiated intermediate. Furthermore we propose ψV knockout DT40 as an ideal model system to approach the molecular mechanism of Ig hypermutation and as a new tool for in situ mutagenesis. Results Targeted Deletion of ψV Donor Sequences in the Rearranged Light Chain Locus Two ψV knockout constructs were made by cloning genomic sequences that flank the intended deletion end points, upstream and downstream of a floxed guanine phosphoribosyl transferase (gpt) cassette ( Arakawa et al. 2001 ). Upon targeted integration, the first construct, pψVDel1-25, deletes all pseudogenes (ψV25 to ψV1), whereas the second construct, pψVDel3-25, deletes most pseudogenes (ψV25 to ψV3) ( Figure 1 A). A surface IgM–positive (sIgM[+]) clone, derived from DT40 Cre1 AID –/– cells ( Arakawa et al. 2002 ) by transfection and stable integration of a floxed AID–internal ribosome entry site (IRES)-green fluorescent protein (GFP) transgene, was chosen for the transfection of the ψV knockout constructs. This AID-reconstituted clone, named AID R , has the advantage that the appearance of deleterious Ig light chain mutations can be easily detected by the loss of sIgM expression, and that GFP-marked AID expression can be shut down after tamoxifen induction of the Cre recombinase transgene inherited from DT40 Cre1 ( Arakawa et al. 2002 ). Figure 1 ψV Gene Deletion (A) Physical map of the rearranged Ig light chain locus in the chicken B-cell line DT40 and the ψV knockout constructs. The locus contains a total of 25 ψV genes upstream of the functional V segment. The strategy of knocking out ψV genes by the targeted integration of the pψVDel1-25 and the pψVDel3-25 constructs is shown. Only the relevant EcoRI sites are indicated. (B) Southern blot analysis of wild-type and knockout clones using the probe shown in (A) after EcoRI digestion. The wild-type locus hybridizes as a 12-kb fragment, whereas ψV partial and ψV – loci hybridize as 7.4-kb and 6.3-kb fragments, respectively. (C) AID status. The AID gene was amplified by PCR to verify the presence or absence of the AID cDNA expression cassette. Following transfection of the ψV knockout constructs into the AID R clone, mycophenolic acid–resistant clones containing targeted deletions of the rearranged light chain locus were identified. These primary ψV knockout clones contain two floxed transgenes, the inserted gpt marker gene in the rearranged light chain locus and the AID-IRES-GFP gene of the AID R progenitor clone. Since the gpt gene might perturb the adjacent transcription or chromatin configuration, the primary ψV knockouts were exposed to a low concentration of tamoxifen and then subcloned by limited dilution. In this way, secondary ψV knockout clones could be isolated that lacked either only the gpt gene (AID R ψV – and AID R ψV partial ) or the gpt gene together with the AID-IRES-GFP gene (AID –/– ψV – and AID –/– ψV partial ). The disruption of ψV genes in the rearranged light chain locus and the excision of the AID overexpression cassette were confirmed by Southern blot analysis ( Figure 1 B) and PCR ( Figure 1 C), respectively. Increased Loss of sIgM Expression after Deletion of ψV Genes in AID-Positive Clones To estimate the rates of deleterious Ig mutations, sIg expression was measured by fluorescence-activated cell sorting (FACS) after 2 weeks' culture for 24 subclones each of the DT40 Cre1 , AID R , DT40 Cre1 AID –/– , and ψV knockout clones ( Figure 2 ). Analysis of the controls with the intact ψV locus revealed an average of 0.52% and 2.27% sIgM(–) cells for the DT40 Cre1 and AID R subclones respectively, but only 0.08% for the DT40 Cre1 AID –/– . Previous analysis of spontaneously arising sIgM(–) DT40 variants demonstrated that about a third contained frameshift mutations in the rearranged light chain V segment that were regarded as byproducts of the Ig gene conversion activity ( Buerstedde et al. 1990 ). This view is now supported by the finding that the AID-negative DT40 Cre1 AID –/– clone, which should have lost the Ig gene conversion activity, stably remains sIgM(+). Most interestingly, subclones of the AID-positive ψV knockout clones (AID R ψV partial and AID R ψV – ) rapidly accumulate sIgM(–) populations, whereas subclones of the AID-negative ψV knockout clones (AID –/– ψV partial and AID –/– ψV – ) remain sIgM(+) ( Figure 2 ). This suggests that the deletion of the pseudogenes dramatically increases the rate of deleterious light chain mutations in AID-expressing cells. Figure 2 sIgM Expression Analysis of Control and ψV Knockout Clones (A) FACS anti-IgM staining profiles of representative subclones derived from initially sIgM(+) clones. (B) Average percentages of events falling into sIgM(–) gates based on the measurement of 24 subclones. Replacement of Ig Gene Conversion by Hypermutation in the Absence of ψV Donors To analyze the newly identified mutation activity, the rearranged light chain VJ segments of the ψV knockout clones were sequenced 5–6 weeks after subcloning. A total of 135 nucleotide changes were found in the 0.5-kb region between the V leader and the 5′ end of the J-C intron within 95 sequences from the AID R ψV – clone ( Figure 3 , upper reference sequence). In contrast to the conversion tracts seen in wild-type DT40 cells, almost all changes are single base substitutions, and, apart from a few short deletions and dinucleotide changes, mutation clusters were not observed. The lack of conversion events in AID R ψV – , which still contains the ψV genes of the unrearranged light chain locus, confirms that Ig gene conversion recruits only the ψV genes on the same chromosome for the diversification of the rearranged light chain gene ( Carlson et al. 1990 ). No sequence diversity was found in a collection of 95 light chain gene sequences from the AID –/– ψV – clone ( Figure 4 A; Table 1 ), indicating that AID is required for the mutation activity. Figure 3 Ig Light Chain Sequence Analysis of the ψV Knockout Clones Mutation profiles of the AID R ψV – and AID R ψV partial clones. All differences identified in different sequences in the region from the leader sequence to the J-C intron are mapped onto the rearranged light chain sequence present in the AID R precursor clone. Mutations of the AID R ψV – and AID R ψV partial clones are shown above and below the reference sequence, respectively. Deletions, insertions, and gene conversion events are also indicated. Hotspot motifs (RGYW and its complement WRCY) are highlighted by bold letters. Changes displayed in the same horizontal line are not necessarily derived from the same sequence. Figure 4 Mutation Profiles of Hypermutating Cell Lines (A) Percentages of sequences carrying a certain number of mutations. Each untemplated nucleotide substitution is counted, but gene conversions, deletions, and insertions involving multiple nucleotides are counted as single events. PM, point mutation; GC, gene conversion; D, deletion; I, insertion. (B) Hotspot preferences of untemplated nucleotide substitution mutations. Mutations occurring within a hotspot motif (RGYW or its complement WRCY) are shown by percentages. The hotspot preference was statistically significant ( p < 0.05) by the standard difference test. (C) Patterns of nucleotide substitutions within sequences from ψV and the XRCC3 knockout clones. Nucleotide substitutions as part of gene conversion events are excluded. The ratios of transitions (trs) to transversions (trv) are also shown. Table 1 Mutation Profile a Point mutations that are not templated by cis-pseudogene donors b Number of gene conversion tracts c From sorted IgM(–) d Templated point mutation in ψV-positive clones cannot be distinguished from short conversion tracts Sequences derived from the AID R ψV partial clone occasionally display stretches of mutations that can be accounted for by the remaining ψV1 and ψV2 ( Figure 3 , lower reference sequence). Nevertheless, the majority of AID R ψV partial mutations are single untemplated base substitutions as seen with the AID R ψV – cells ( Figure 4 A; Table 1 ). Only three base substitutions, which are possibly PCR artifacts, were found in 92 sequences of the AID –/– ψV partial clone, confirming that both the gene conversion and the mutation activities of AID R ψV partial are AID-dependent. The New Mutation Activity of the ψV Knockout Clones Closely Resembles Somatic Hypermutation The Ig mutation activity discovered in the ψV knockout clones with a predominance of single nucleotide substitutions suggests that somatic hypermutation had replaced Ig gene conversion. There is, however, a difference between the nucleotide substitutions in the AID R ψV partial and AID R ψV – clones and Ig hypermutations in germinal center B-cells: The clones show very few mutations in A/T bases and a preference for transversion mutations, and among transversions, a preference for G-to-C and C-to-G changes ( Figure 4 ). Ig hypermutations are typically localized within 1 kb of the transcribed gene sequence, with preferences for the complementary determining regions (CDRs) of the V(D)J segments, whereas no or few mutations are present in the downstream C region ( Lebecque and Gearhart 1990 ). To investigate whether the mutations in the AID R ψV – clone follow a similar distribution, sequence analysis was extended to the promoter region and the J-C intron of the rearranged light chain gene ( Figure 5 ). Although mutations are found close to the promoter and in the intron downstream of the J segments, the peak incidence clearly coincides with the CDR1 and CDR3, which are also preferred sites of gene conversion in DT40 (unpublished data). Approximately half of all point mutations fall within the RGYW (R = A/G; Y = C/T; W = A/T) sequence motif or its complement WRCY (see Figure 4 B), known as hotspots of Ig hypermutation in humans and mice. Figure 5 Mutations within Unsorted and Sorted sIgM(–) cells Distribution of nucleotide substitutions within genomic sequences from unsorted AID R ψV – cells and within cDNA sequences from sorted IgM(–) AID R ψV – cells. The numbers of mutations are counted for every 50 bp, and are shown together with the corresponding physical maps of the light chain genomic locus or the cDNA sequence. It was previously reported that the deletion of RAD51 paralogs induces Ig hypermutation in DT40 cells ( Sale et al. 2001 ). To compare the hypermutation activity in the ψV gene-negative and RAD51 paralog-negative backgrounds, the XRCC3 gene was disrupted in the DT40 Cre1 clone, and the rearranged VJ genes were sequenced 6 weeks after subcloning. The mutation spectrum of the XRCC3- deficient clone was similar to that of the AID R ψV – clone (see Figure 4 C) and to what was previously reported for the XRCC3 knockout ( Sale et al. 2001 ). Nevertheless, the mutation rate in the new XRCC3 mutant was about 2.5-fold lower than in the AID R ψV – clone, and there was a clear slow-growth phenotype of the XRCC3 mutant compared to wild-type DT40 and the AID R ψV – clone (unpublished data). To identify the mutations responsible for the loss of sIgM expression in the AID R ψV – clone, 94 light chain cDNAs from sorted sIgM(–) cells were amplified and sequenced. Although one short insertion and five deletions were detected in this collection ( Table 1 ), 89% of the 245 total mutations are single-nucleotide substitutions within the VJ segments and only few mutations were observed in the C segment ( Figure 5 ). Surprisingly, only about 10% of the sequences contained a stop codon or a frameshift, suggesting that the lack of sIgM expression is mainly caused by amino acid substitutions that affect the pairing of the Ig light and heavy chain proteins. Ig Locus Specificity of Hypermutation It has been reported that high AID expression in fibroblasts ( Yoshikawa et al. 2002 ) and B-cell hybridomas ( Martin and Scharff 2002 ) leads to frequent mutations in transfected transgenes. To rule out the possibility that the pseudogene deletions had induced a global hypermutator phenotype, the 5′ ends of the genes encoding the B-cell -specific marker Bu-1 and the translation elongation factor EF1α were sequenced for the AID R ψV – clone. Only a single 1-bp deletion was found within 95 sequences of the Bu-1 gene, and only two single nucleotide substitutions within 89 sequences of EF1α ( Table 1 ). As these changes most likely represent PCR artifacts, this further supports the view that the hypermutations induced by the ψV deletions are Ig-locus-specific. Discussion These results demonstrate that the deletion of the nearby pseudogene donors abolishes Ig gene conversion in DT40 and activates a mutation activity that closely resembles Ig hypermutation. The features shared between this new mutation activity and somatic hypermutation include (1) AID dependence, (2) a predominance of single nucleotide substitutions, (3) distribution of the mutations within the 5′ transcribed region, (4) a preference for hotspots, and (5) Ig gene specificity. The only differences between the mutation activity induced by loss of ψV and Ig hypermutation in vivo are the relative lack of mutations in A/T bases and a predominance of transversion mutations in the ψV knockout clones. However, these differences are also seen in hypermutating Epstein Barr virus–transformed B-cell lines ( Bachl and Wabl 1996 ; Faili et al. 2002 ) and DT40 mutants of RAD51 paralogs ( Sale et al. 2001 ), indicating that part of the Ig hypermutator activity is missing in transformed B-cell lines. Interestingly, the rate of Ig hypermutation in the AID R ψV – clone seems higher than the rate of Ig gene conversion in the DT40 Cre1 progenitor. An explanation for this could be that some conversion tracts are limited to stretches of identical donor and target sequences and thus leave no trace. The ratio of transversion to transition was lower for the AID R ψV partial clone (see Figure 4 ). Although we can only speculate about the cause of this difference, it might be due to the correction of point mutations by mismatch correction of one or more sites in gene conversion tracts. The induction of Ig hypermutation by the blockage of Ig gene conversions supports a simple model explaining how hypermutation and recombination are initiated and regulated ( Figure 6 ). Initiating the events is a modification of the rearranged V(D)J segment that is either directly or indirectly induced by AID. The default processing of this lesion in the absence of nearby donors or of high homologous recombination activity leads to Ig hypermutation in the form of a single nucleotide substitution ( Figure 6 , right). However, if donor sequences are available, processing of the AID-induced lesion can be divided into a stage before strand exchange, when a shift to Ig hypermutation is still possible, and a stage after strand exchange, when the commitment toward Ig gene conversion has been made ( Figure 6 , left). Whereas completion of the first stage requires the participation of the RAD51 paralogs, the second stage involves other recombination factors, such as RAD54. Figure 6 A Model of the Regulation of Ig Gene Conversion and Hypermutation This difference in commitment explains why disruptions of the RAD51 paralogs not only decrease Ig gene conversion, but also induce Ig hypermutation ( Sale et al. 2001 ), whereas disruption of the RAD54 gene only decreases Ig gene conversion ( Bezzubova et al. 1997 ). The model also predicts that low cellular homologous recombination activity prevents Ig gene conversion even in the presence of conversion donors. Such a low homologous recombination activity might be the reason why human and murine B-cells never use Ig gene conversion despite the presence of nearby candidate donors in the form of unrearranged V segments and why chicken germinal center B-cells have shifted the balance from Ig gene conversion to Ig hypermutation ( Arakawa et al. 1998 ). The AID R and the ψV knockout DT40 clones are a powerful experimental system to address the role of trans-acting factors and cis-acting regulatory sequences for Ig gene conversion and hypermutation. Compared to alternative animal or cell culture systems, it offers the advantages of (1) parallel analysis of Ig gene conversion and hypermutation, (2) conditional AID expression, (3) easy genome modifications by gene targeting, (4) normal cell proliferation and repair proficiency, and (5) Ig locus specificity of hypermutation. The ability to induce gene-specific hypermutation in the DT40 cell line might also find applications in biotechnology. One possibility is to replace the chicken antibody coding regions with their human counterparts and then to simulate antibody affinity maturation from a repertoire that continuously evolves by Ig hypermutation. Materials and Methods Cell lines DT40 Cre1 , which displays increased Ig gene conversion due to a v- myb transgene and contains a tamoxifen-inducible Cre recombinase, has been described previously ( Arakawa et al. 2001 ). DT40 Cre1 AID –/– was generated by the targeted disruption of both AID alleles of DT40 Cre1 ( Arakawa et al. 2002 ). AID R was derived from DT40 Cre1 AID –/– after stable integration of a floxed AID-IRES-GFP bicistronic cassette, in which both AID and GFP are expressed from the same β-actin promoter. AID R ψV – was derived from AID R by transfection of pψVDel1-25 (see Figure 1 A). Stable transfectants that had integrated the construct into the rearranged light chain locus were then identified by locus-specific PCR. Targeted integration of pψVDel1-25 results in the deletion of the entire ψV gene loci starting 0.4 kb upstream of ψV25 and ending 1 bp downstream of ψV1. AID R ψV partial was produced in a similar way as was AID R ψV – , by transfection of pψVDel3-25, which, upon targeted integration, leads to a partial deletion of the ψV loci starting 0.4 kb upstream of ψV25 and ending 1 bp downstream of ψV3. Cell culture and electroporation were performed as previously described ( Arakawa et al. 2002 ). XRCC3 –/– was derived from DT40 Cre1 by deleting amino acids 72–170 of the XRCC3 gene following transfection of XRCC3 knockout constructs. Clones that underwent targeted integration were initially identified by long-range PCR, and the XRCC3 deletion was then confirmed by Southern blot analysis. Ig reversion assay Subcloning, antibody staining, flow cytometry, and quantification of sIgM expression has been described previously ( Arakawa et al. 2002 ). All clones used in the study were sIgM(+) because of the repair of the light chain frameshift of the original Cl18(–) variant ( Buerstedde et al. 1990 ) by a gene conversion event. PCR To minimize PCR-introduced artificial mutations, PfuUltra hotstart polymerase (Stratagene, La Jolla, California, United States) was used for amplification prior to sequencing. Long-range PCR, RT-PCR, and Ig light chain sequencing were performed as previously described ( Arakawa et al. 2002 ). The promoter and J-C intron region of Ig light chain plasmid clones were sequenced using the M13 forward and reverse primers. Bu-1 and EF1α genes were amplified using BU1/BU2 (BU1, 5′-GGGAAGCTTGATCATTTCCTGAATGCTATATTCA-3′; BU2, 5′-GGGTCTAGAAACTCCTAGGGGAAACTTTGCTGAG-3′) and EF6/EF8 (EF6, 5′-GGGAAGCTTCGGAAGAAAGAAGCTAAAGACCATC-3′; EF8, 5′-GGGGCTAGCAGAAGAGCGTGCTCACGGGTCTGCC-3′) primer pairs, respectively. The PCR products of these genes were cloned into the pBluescript plasmid vector (Stratagene) and were sequenced using the M13 reverse primer. Supporting Information Accession Numbers The GenBank ( http://www.ncbi.nlm.nih.gov/ ) accession numbers of the genes discussed in this paper are as follows. AID (NM_009645; NM_020661), RAD54 (GGU92461), RAD52 (U01047), Bu-1 (X92865), and EF1α (NM_204157).	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC449846.xml
526254	Renal outcome in adults with renal insufficiency and irregular asymmetric kidneys	Background The commonest cause of end-stage renal failure (ESRF) in children and young adults is congenital malformation of the kidney and urinary tract. In this retrospective review, we examine whether progression to ESRF can be predicted and whether treatment with angiotensin converting enzyme inhibitors (ACEI) can delay or prevent this. Methods We reviewed 78 patients with asymmetric irregular kidneys as a consequence of either primary vesico-ureteric reflux or renal dysplasia (Group 1, n = 44), or abnormal bladder function (Group 2, n = 34). Patients (median age 24 years) had an estimated GFR (eGFR) < 60 ml/min/1.73 m 2 with at least 5 years of follow up (median 143 months). 48 patients received ACEI. We explored potential prognostic factors that affect the time to ESRF using Cox-regression analyses. Results At start, mean (SE) creatinine was 189 (8) μmol/l, mean eGFR 41 (1) ml/min 1.73 m 2 , mean proteinuria 144 (14) mg/mmol creatinine (1.7 g/24 hrs). Of 78 patients, 36 (46%) developed ESRF, but none of 19 with proteinuria less than 50 mg/mmol and only two of 18 patients with eGFR above 50 ml/min did so. Renal outcome between Groups 1 and 2 appeared similar with no evidence for a difference. A benefit in favour of treatment with ACEI was observed above an eGFR of 40 ml/min (p = 0.024). Conclusion The similar outcome of the two groups supports the nephrological nature of progressive renal failure in young men born with abnormal bladders. There is a watershed GFR of 40–50 ml/min at which ACEI treatment can be successful at improving renal outcome.	Background Nearly half the children and young adults who develop end-stage renal failure (ESRF) have asymmetric irregularly shaped kidneys [ 1 ]. This appearance, often referred to as bilateral renal scarring, is frequently associated with vesico-ureteric reflux (VUR) and sometimes with a history of urinary tract infection (UTI). It is generally a consequence of congenital malformations of the kidneys and urinary tract and is variously described as `reflux nephropathy' or `chronic pyelonephritis.' Such patients fall into two broad groups. Firstly, there is a group who appear to have normal bladders without outflow obstruction and normal calibre ureters when not micturating, described as having either primary VUR or primary renal dysplasia. Secondly, there is a group with some form of bladder outflow dysfunction which causes a secondary VUR and dilated upper urinary tracts, of which a posterior urethral valve (PUV) in males is the most common cause. The primary group have a bimodal presentation. Commonly they present in childhood with UTI; the rest present in early adult life with renal insufficiency and often with no preceding history of UTI [ 2 - 6 ]. Traditionally the diagnosis was made by recognising the characteristic appearance of calyceal clubbing and irregular `scarring' of the kidney on intravenous urography (IVU) [ 7 , 8 ]. With significant renal insufficiency, however, these changes can be impossible to see clearly by IVU [ 2 ], and the irregular, asymmetrical kidney is more sensitively visualised by 99 m Tc-dimercaptosuccinic acid (DMSA) renography [ 9 , 10 ]. In this adult population a micturating cysto-urethrogram (MCU) frequently will not show evidence of VUR as reflux usually ceases spontaneously in childhood [ 2 , 4 , 5 ]. In fact, the finding of VUR is a weak predictor of renal damage in children admitted with an UTI [ 11 ]. The appearance of proteinuria and progressive renal failure indicates glomerular capillary hypertension (glomerular hyperfiltration) and progressive focal and segmental glomerulosclerosis (FSGS) [ 12 , 13 ]. Risk factors for patients with reflux nephropathy developing progressive renal failure after childhood are proteinuria, renal insufficiency, bilateral scarring of the kidneys and hypertension [ 2 , 4 , 5 ]. Patients with congenital bladder outflow obstruction and secondary reflux, however, have usually been excluded from such outcome studies, and very little has been published from a nephrological perspective about their long-term outcome. In this retrospective observational review, from a large, single centre nephro-urological practice, we have examined the natural history and progression to ESRF of patients with primary and secondary reflux with asymmetric irregular kidneys and moderate to severe renal insufficiency. We have tested the null hypothesis of no difference in renal outcome between patients with primary and secondary reflux. Methods Patients Patients with bilaterally scarred kidneys and glomerular filtration rate (GFR) 15–60 mls/min/1.73 m 2 were identified from a review of the records of outpatients and of patients receiving renal replacement therapy at the Renal Unit of the Middlesex Hospital (UCL Hospitals Trust). Most patients had been referred, as adolescents, from the nephrology and urology clinics at the Great Ormond Street Hospital for Children. All patients had renal scarring confirmed by DMSA or 99 m Tc-mercaptoacetyltriglycine (MAG-3) renography, although most patients had undergone extensive investigations. For inclusion in this study, patients had: • an isotopic 51 Cr-edetic acid (EDTA) GFR < 60 ml/min/1.73 m 2 ; or estimated GFR < 60 ml/min/1.73 m 2 • apparently stopped growing and with a steady body weight (so that plasma creatinine could be used to estimate serial GFRs), and • data for at least 5 years of follow up. Patients specifically excluded from this study were those with bladder exstrophy, neuropathic bladders, or any form of urinary diversion (conduit or reservoir). In our analysis, the patients were divided into two broad groups: Group 1: those with normal calibre ureters and normal bladders (Primary group) Group 2: those with megaureters, hydronephrosis and abnormal bladders (Secondary group). Data Glomerular filtration rate (GFR) was estimated by single exponential analysis of the plasma clearance of 51 Cr-edetic acid (EDTA) following a single intravenous injection with blood samples taken after 2 and 4 hours [ 14 ]. Plasma (PCr) and urine creatinine concentration were measured by the Jaffe technique using an autoanalyser (Chemlab Instruments, Hornchurch, UK) and urinary protein (Uprot) by turbidometric assay following precipitation with trichloroacetic acid. GFR was estimated by different formulae and compared with the isotopic GFR. The Jelliffe formulae (I and II) [ 15 , 16 ] have been shown to approximate most closely low values of GFR when compared with the inulin clearance [ 17 ]. Jelliffe I (ml/min/1.73 m 2 ) [ 15 ]:- "(100 × 88/ PCr μmol/l) - 12" for males; and "(80 × 88/creatinine μmol/l) - 7" for females. Jelliffe II (ml/min/1.73 m 2 ) [ 16 ]:- "(98–0.8(age-20)) × 88/ PCr μmol/l" for males. (Jelliffe II × 0.9 for females). [The original formulae used creatinine mg/dl. Conversion to μmol/l introduces the factor 88]. We found that a mean of these 2 formulae gave closer approximations to measured isotopic GFR than the other methods. We have termed this mean value `estimated GFR' (eGFR). All measured values of 51 Cr EDTA GFRs were compared with the eGFR calculated using the contemporary value of plasma creatinine. 151 values of corrected isotopic GFR (ml/min/1.73 m 2 ) from a range of 12–60 ml/min/1.73 m 2 were found to have no significant bias (-0.34 ml/min/1.73 m 2 with a 95% CI of -1.19 to 0.51 ml/min/ 1.73 m 2 ) and an agreement within limits of -11.0 to 10.3 ml/min/1.73 m 2 . Normal bladder: patients presenting after adolescence were considered to have a normal bladder if they had no bladder outflow symptoms, a normal urine flow rate (> 15 ml/sec) and no residual urine volume seen by ultrasound after voiding. Declining renal function: the rate of progression of CRF, -delta GFR (`-ΔGFR'), was calculated as the rate of change of eGFR and is shown as ml/min/year. Proteinuria: was initially measured as the amount of protein (g) in a 24-hour urine collection. Since 1995 proteinuria was more commonly measured on a random (spot) sample of urine at clinic visits with the proteinuria expressed as mg protein/mmol creatinine (normal laboratory range 0–13 mg/mmol). As all 24 hour urine data (Uprot) included creatinine excretion (mmol/24 hours) we have been able to calculate protein/creatinine ratios (Up/Cr). Using the data from 161 separate 24-hour collections, we assessed how proteinuria in g/day predicts protein/creatinine ratios. Paired values ranged from 0.1–8.6 g/day and 8–700 mg protein/mmol creatinine, with a high correlation (r = 0.90). The regression equation was [Up/Cr = 90.3 × Uprot 0.94 ]. Hypertension was defined as either blood pressure consistently > 140/85 mmHg, or patients receiving blood pressure lowering therapy. End-stage renal failure (ESRF): was taken to be the date when the patient began dialysis. To calculate changes in renal function with time the eGFR was assumed to be 8 ml/min at this time. Outcome: Renal outcome was defined as having reached ESRF or not at last review. ACEI therapy Since June 1986 some patients were started on ACEI therapy when anti-hypertensive therapy was required. In addition, some anti-hypertensive regimens were changed to ACEI therapy. A small group with blood pressure <140/85 were started on ACEI therapy because of increasing proteinuria. Some patients never received ACEI because ESRF was reached before the use of ACEIs in renal insufficiency had become routine. Patients, who were started on ACEI for hypertension, were advised to restrict their salt intake and the initial aim was for a blood pressure of ≤ 130/70. If ACEI alone did not lower the blood pressure to the target a diuretic was added. The latter was not always possible in Group 2 patients who might already have features of hypovolaemia secondary to their renal tubular pathology. No patient ever received any immunosuppressive drug. Data collection Only data from the start and end of the study are presented from all 78 patients. Of the 48 patients who were treated with ACEI, we also report data on 28 of the patients (for whom it was available for at least 18 months before and 48 months after the introduction of ACEI.) at the start of ACEI therapy and again at 2 years after start of therapy. `Statistical Analysis' Means were compared within groups by paired samples t-test, and between groups by Wilcoxon rank-sum test. The time scale was time since birth, since the setting was of a congenital disease eventually leading to renal failure. It was checked graphically that there was no obvious pattern to loss to follow up over time. Renal outcome (reaching ESRF or not) was compared between groups using Kaplan-Meier survival plots. Outcomes were quantified as the median survival outcome in months (with 95% Confidence Intervals [CI]). This is equivalent to the median time for 50% of the group to reach ESRF. Differences in renal outcome over time were tested using log-rank tests. For graphical examination of the proportionality, continuous variables were grouped into approximate tertiles. The Cox proportional hazards model was chosen for further analysis. Continuous variables were centred on the mean. Protein/urine creatinine ratios were log-transformed for further analyses. Treatment with ACEI was entered as a time-changing variable to take account of the variable start of treatment after referral. Univariable regression models estimated crude (unadjusted) effects of the prognostic variables. Since it is possible that the effect ACEI depends on the GFR at start of treatment (entered as continuous variable), especially in patients with moderate to severe renal insufficiency, an interaction between those variables was fitted. Variables were entered in the multivariable model in a stepwise forward fashion. Analyses were performed using SPSS for Windows v10.1. and Stata 8 (Stata Corporation, Texas, USA). Results Demography Data from 78 patients, who were first seen between December 1969 and February 1988, were analysed. Demographic details are presented in Table 1 . The age of patients at the start of the study period ranged from 15–49 years (median 23 years) with one lady aged 65 years. Table 1 Demographic Details At Entry (n = 78). Data shown are means (SE); Dates (month/year) are medians. Because isotopic GFRs were not always performed this data is not shown in Table, but 26 Group 1 patients had a mean (SE) isotopic-GFR of 41.2 (2.1) ml/min/1.73 m 2 with a contemporaneous mean eGFR of 44.2 (2.3) ml/min/1.73 m 2 , and the 26 Group 2 patients with an isotopic GFR of 40.2 (2.3) had an eGFR 41.5 (2.0) ml/min 1.73 m 2 . Group 1 (Primary reflux) Group 2 (Secondary reflux) TOTAL Total 44 34 78 Male:Female 22:22 32:2 54:24 Date at start (median) 8/1988 6/1986 2/1987 Age (year) 28.9 (1.6) 22.2 (1.4) 26 (1.1) Creatinine (μmol/l) 178 (9.0) 203 (13) 189 (7.7) eGFR (ml/min) 42.3 (1.8) 40.3 (2.1) 41.4 (1.3) Proteinuria (g/d) 1.63 (0.19) 1.83 (0.33) 1.72 (0.2) Protein/creatinine (mg /mmol) 136 (14) 154 (27) 144 (14) Hypertension 18 (41%) 5 (15%) 23 (29%) Treatment with ACEI 32 (73%) 16 (47%) 48 (62%) Treatment date (median) 4/1992 8/1993 11/1992 Total months of follow up 145 (11) 143 (9) 144 (7) Group 1 (n = 44; 50% female) were patients who had primary VUR or primary renal dysplasia. Twenty two patients (71% female) presented in childhood with a UTI. In each case, MCU performed at a median age of 7 years showed reflux which was either bilateral (82%) or unilateral (18%). In contrast, the remainder (n = 22) presented at a median age of 24 years. Only 32% were female and they almost invariably presented either with hypertension after starting a contraceptive pill or with complications during pregnancy. Only 2 had had a MCU performed and neither showed VUR. Group 2 patients (n = 34; 6% female) had the following diagnoses: PUV (n = 15), prune belly syndrome (n = 2), single dysplastic kidney with megaureter (n = 2), renal dysplasia with abnormal bladder function (n = 2), bilateral megaureters (n = 1), megacystis and megaureters (n = 4), and finally a group (n = 8) in whom the initial diagnosis (pre-1979) had included "bladder neck obstruction". Six of these had megacystis and megaureters and might now be termed `pseudo-prune belly syndrome'. Renal outcome By Group: The median survival time of Group 1 versus Group 2 was compared by log rank test: 231 months (95% CI: 153–309) vs. 162 months (95% CI: 135–189) respectively, p = 0.35. There was no evidence for a major difference in renal outcome between these patients with primary and secondary reflux. Thus for subsequent analyses the data from all 78 patients was combined. By eGFR: We compared renal outcome for all patients after they were stratified by initial eGFR into four groups (15–30, 31–40, 41–50, and 51–60 ml/min/1.73 m 2 ) (see Fig 1 ). Of 18 patients with eGFRs >50 ml/min, only 2 (11%) reached ESRF: however, eGFR still declined in the other 16 by 1.30 ml/min/yr and mean proteinuria rose from 35 to 55 mg/mmol creatinine, despite 12 patients (75%) receiving ACEI. Figure 1 Renal outcome stratified for eGFR (ml/min/1.73 m 2 ) at start. eGFR 51–60 vs 41–50: p = 0.17; eGFR 41–50 vs 31–40: p = 0.004; eGFR 31–40 vs 15–30: p = 0.041. By Proteinuria: We compared renal outcome for all patients after they were stratified by initial proteinuria into three groups (0–99, 100–199, and ≥ 200 mg/mmol) (Fig 2 ). The great significance of proteinuria is emphasised further by the observation that seven patients with proteinuria ≥ 200 mg/mmol reached ESRF despite initial eGFRs ≥ 40 ml/min. In contrast, none of 19 patients with proteinuria <50 mg/mmol creatinine at start developed ESRF after a median follow up of 160 months (range 87–227). Figure 2 Renal outcome stratified for proteinuria (mg/mmol) at start. 10 – 99 vs 100–199 mg/mmol: p = 0.009; 100–199 vs >200 mg/mmol: p = 0.002. Proteinuria increased with time and declining function in both Groups (Table 2 ) but levels were consistently higher in Group 2 compared with Group 1 patients (Table 1 ). There was a strong correlation between rate of loss of function and proteinuria at start (R = 0.63, p < 0.0001), and end of study: (R = 0.69). Table 2 Creatinine, eGFR, proteinuria and ACE-I stratified by renal function at outset. Data are medians (range). Proteinuria : b) vs c) p = 0.06, c) vs d) p = 0.031; -Δ eGFR † : b) vs c) p= 0.14, c) vs d) p = 0.012; Total -ΔGFR is the rate of change of function in ml/min/yr from start to last follow up; Rx ACEI is the percentage of patients receiving ACEI treatment Renal Function groups N= Creatinine eGFR Proteinuria -Δ eGFR Total -Δ eGFR post-ACEI Rx ACEI μmol/l ml/min/1.73 m 2 mg/mmol ml/min/yr ml/min/yr a) 15–30 ml/min 16 295 (220–450) 24 209 (57–680) 2.94 (0.55–5.7) 2.63 (1.51–4.3) 38% b) 31–40 ml/min 14 198 (160–233) 36 200 (71–275) 3.05 (0.95–7.4) 1.7 (0.9–3.45) 50% c) 41–50 ml/min 30 160 (130–185) 46 100 (10–276)* 1.71 (0.44–8.21) † 1.68 (0.66–7.85) 77% d) 51–60 ml/min 18 130 (115–153) 55 38* (10–250) 1.34 (0.24–3.41) † 1.76 (0.24–3.73) 72% ACEI treatment 48 patients commenced ACEI therapy at a median of 48 months (range 0–311) after the start of the study, by which time their median eGFR had fallen from 46 (range 15–60) to 36 (10–60) ml/min/1.73 m 2 . Effect on renal outcome Table 3 shows the results of both the univariable and multivariable analyses. For every variable entered into the model the assumption of proportionality of hazards was met. It was notable, given the small number of patients of our sample, that we were able to detect an interaction between treatment with ACEI and renal function at treatment start. The effect of ACEI was estimated to have its main effects just above a GFR of 40 ml/min. In the crude, as well as in the full model, neither sex nor type of reflux seem to have a significant effect upon time to ESRF since referral. At entry to study, both the amount of proteinuria and eGFR were important prognostic variable towards ESRF in crude as well as adjusted analyses. There was one 65 year old lady. Refitting the final model omitting this record did not affect the estimates. Table 3 Estimated crude and adjusted hazard ratios for incidence of ESRF in all patients. full model includes all variables, since analyses were conducted on the age-scale, effects are taking account of current age; interaction parameters (95%CI): crude model: -0.088 (-0.162,-0.014); p = 0.019 full model: -0.093 (-0.174,-0.012); p = 0.024; # effect of 100 mg/mmol creatinine = (displayed hazard ratio) 0.7 effect of 200 mg/mmol creatinine = (displayed hazard ratio) 1.4; ## effect of 10 ml/min/1.73 m 2 decrease = (displayed hazard ratio) 2 effect of 15 ml/min/1.73 m 2 decrease = (displayed hazard ratio) 3 Hazard ratios Estimated effects of categories/unit crude 95%CI p-value Adjusted 95%CI p-value Gender female 1.00 1.00 male 1.17 (0.55, 2.50) 0.677 0.55 (0.21,1.39) 0.205 Type of reflux primary 1.00 1.00 secondary 1.28 (0.65, 2.55) 0.478 1.73 (0.73, 4.06) 0.211 Proteinuria per 50 mg/mmol proteinuria increase # 1.71 (1.33, 2.20) <0.001 1.50 (1.17, 1.91) 0.001 eGFR per 5 ml/min/1.73 m 2 decrease ## 1.38 (1.16, 1.64) <0.001 1.29 (1.05, 1.58) 0.016 ACE inhibitor at 30 ml/min/1.73 m 2 0.53 (0.22, 1.29) 0.162 0.67 (0.27, 1.67) 0.393 at 35 ml/min/1.73 m 2 0.34 (0.12, 1.00) 0.051 0.42 (0.14, 1.26) 0.121 at 40 ml/min/1.73 m 2 0.22 (0.06, 0.84) 0.027 0.27 (0.07, 1.04) 0.058 at 45 ml/min/1.73 m 2 0.14 (0.03, 0.73) 0.02 0.17 (0.03, 0.91) 0.039 Effect on rate of progression and proteinuria We examined the effect of ACEI on 28 patients with deteriorating function for whom data was available for at least 18 months before and 48 months after the introduction of ACEI. ACEI reduced the rate of loss of renal function during the first 24 month period of follow-up, but the benefit was greater in the subsequent follow-up period with the rate slowing from a median of -1.86 before ACEI to -1.48 ml/min/yr (p = 0.007). ACEI treatment was associated with a reduction in proteinuria after 24 months of therapy. However, proteinuria had increased at last follow up owing to loss of the anti-proteinuric effect in half the group (Fig 3 ). Figure 3 Effect of ACEI on Proteinuria. Time points are 1) start of study, 2) at begin of ACEI therapy, 3) 2 years after begin ACEI; 4) at end of study. Proteinuria* at ACEI vs +2 years post-ACEI ; p < 0.0001. Analysis of 30 patients who did not receive an ACEI provides indirect support of benefit from this treatment. The median rate of loss of renal function for 25 of these patients who started with a eGFR ≤ 50 ml/min was 2.80 ml/min/yr and if initial proteinuria was ≥ 50 mg/mmol, (n = 23) the rate of loss of renal function was 3.0 ml/min/yr. Blood pressure At the start of the study 18 (41%) Group 1 patients and 5 (15%) Group 2 patients were hypertensive. Group 2 patients tended to be normotensive and some were started on an ACEI for proteinuria. At ESRF or last follow up, all patients were on conventional anti-hypertensive therapy or ACEI, except for two Group 2 patients. Discussion There is a consensus that patients with renal insufficiency and proteinuria have progressive renal failure, that the rate of decline of function is proportional to the magnitude of the proteinuria, and that angiotensin antagonists both slow the rate of progression and reduce proteinuria [ 18 - 20 ]. Our data show that this nephro-urological group of patients is no exception. While VUR patients with abnormal bladders almost invariably present in early childhood, patients with normal bladder function have a bimodal presentation. In one series from New Zealand, 42 patients (36 adults) had ESRF with reflux nephropathy. Many had presented with advanced renal insufficiency, hypertension, and proteinuria, and only 22% of males and 58% of females had a history of UTI. Similarly, in our study, VUR had been proven in 50% of patients with primary reflux nephropathy and these patients (71% female) had almost invariably presented with an UTI in childhood (median age 7.0 years). In contrast, the other half (32% female) presented at a median age of 24 years with advanced disease and persistent reflux was not demonstrated in the 2 patients who underwent a MCU. Similar to the New Zealand experience [ 5 , 21 , 22 ], nearly all our women presented with hypertension after starting a contraceptive pill or with complications during pregnancy, whereas the men were found to have proteinuria, hypertension or renal insufficiency – usually on routine investigation. Although reflux nephropathy is frequently viewed as a disease of little girls with recurrent UTIs [ 7 ], our data confirms findings of others regarding the late presentation of adults with asymmetric irregular kidneys. In a UK series from Newcastle, only 9% presented under 20 years of age [ 2 ], in Australia 22% under 15 years [ 4 ], and from Italy 10% under 12 years [ 6 ]. Furthermore it is clear from published reviews [ 2 , 4 - 6 ] and dialysis programmes [ 5 , 23 ], that there is no female preponderance at ESRF. In 1978 Kincaid-Smith and her colleagues [ 12 ] reported that progressive renal failure with primary VUR was very unlikely unless proteinuria was in excess of 1.0 g/day (equivalent to 100 mg protein/mmol creatinine). In a subsequent report, in which 147 such patients were followed for a mean of 6.9 years, renal function deteriorated in 37% and 14% progressed to ESRF. Proteinuria, elevated creatinine and hypertension at presentation were associated with relative risks (RR) of 25, 24 and 4.5 respectively for the development of progressive renal failure [ 4 ]. In an Italian study [ 6 ], 80 patients were followed for a mean of 5.6 years and retrospectively stratified into those with stable renal function and those with slowly or rapidly progressive renal failure. For those with progressive nephropathy, there was no difference in initial renal function but proteinuria was much greater in the rapidly progressive group. Loss of function was unusual with creatinine ≤ 1.7 mg/dl (150 μmol/l) and inevitable above that concentration [ 6 ]. In a UK study from Newcastle, proteinuria and renal insufficiency (plasma creatinine >130 μmol/l) were present from presentation in 21% and 13% of 125 patients respectively, and with time (mean 5.9 years) a further 21% developed proteinuria and 22% renal insufficiency. In all the 16 patients with progressive renal failure, the decline was linear. In a subsequent report, progressive renal failure did not develop in 138 adult patients with normal function at the start (plasma creatinine < 90 μmol/l) [ 3 ]. Our data support the established relationship between the risk of progressive nephropathy and renal insufficiency with proteinuria, but suggests a watershed range for renal function as a predictor of outcome. When the eGFR exceeds 40–50 ml/min/1.73 m 2 nephropathy rarely progresses, but disease progression is invariable when function is worse. Nakashima et al. similarly reported that an isotopic GFR less than 49 ml/min predicted decline to ESRF [ 24 ]. The other determining factor of poor renal outcome is proteinuria, and we find that deterioration can be expected when proteinuria exceeds 50 mg/mmol (0.5 g/d). In a Japanese study, serial biopsy samples from patients with reflux nephropathy confirmed the close association between the degree of renal scarring, the extent of the glomerular pathology, and proteinuria [ 25 , 26 ]. Once extensive glomerular sclerosis was present there was conspicuous glomerular hypertrophy which correlated with increasing proteinuria. This is consistent with our findings that proteinuria increased as renal function declined. The prognostic importance of proteinuria is emphasised by our observation that 6 of the 25 (24%) patients with proteinuria ≥ 200 mg/mmol developed ESRF despite initial eGFRs exceeding 40 ml/min. On the other hand, the survival outcome benefit of ACEI treatment was most conspicuous when patients with proteinuria = 100 mg/mmol were compared (p < 0.00001). In a 10-year follow up study of 52 children, randomised to medical or surgical management of severe bilateral VUR (grades III-IV) between 1985–1989, progressive renal failure developed in only 4 children (2 from each group) all of whom had GFRs at or below 40 ml/min/1.73 m 2 at outset [ 27 ]. Despite the few long-term studies of adult patients with primary VUR and reflux nephropathy [ 2 - 5 ], there is almost no renal outcome data on adult patients born with abnormal bladder function (Group 2) [ 28 ]. It had been our clinical impression that those with secondary reflux did less well, but although there was a trend for Group 2 patients to do less well this was not statistically significant, although a difference might emerge if large numbers were studied. Progressive renal damage due to congenital outflow tract obstruction may be averted by urological intervention. This is not, however, always successful and even treatment in utero may not prevent progressive renal damage [ 29 ] and the development of ESRF. Despite correction of urethral obstruction, 30% of boys with PUV develop ESRF by the age of 15 years and this may be due to continuing bladder dysfunction [ 28 ]. Most boys born with abnormal bladders who develop ESRF have posterior urethral valves, but in our series 35% had presented with megacystis/megaureters. Before 1975, this was attributed to bladder neck obstruction and treated by bladder neck surgery but subsequently it has been determined that most of this group are born with gross bilateral VUR. The dilated bladders and ureters are attributed to the constant recycling of refluxed urine, sometimes exacerbated by the nephrogenic polyuria [ 30 ] although urodynamic studies often show high voiding pressure suggestive of detrusor/sphincter dyssynergia. The concept that progressive renal failure is often `nephrogenic' in origin, rather than urological, is supported by our data which show that adults with abnormal bladders do not behave significantly differently to those with primary reflux – although the numbers are relatively small. Nevertheless, if renal function deteriorates in a urological patient in the absence of proteinuria, then some other cause (such as obstruction) must be sought. Apart from small numbers, the limitations of this study are the usual ones in observational research with effect estimates that are potentially confounded. We believe, however, that it would not now be possible to design a prospective study in which ACEI therapy was with held from one group. In this study, ACEI therapy improved renal outcome although non-ACEI patients were generally from an earlier period and not necessarily seen in a specialist clinic. The analysis demonstrated a long-term benefit for ACEI both in slowing the rate of loss of function and in reducing proteinuria. However, ACEI therapy appeared to be ineffective if started when the eGFR was already = 30 ml/min. We have studied a group of patients whose renal pathology is not immunological and whose residual functioning tissue is not homogenously distributed. The pathophysiology is similar to experimental sub-total nephrectomy which results in proteinuria, progressive glomerulosclerosis and renal failure [ 13 , 31 ]. Like the experimental models, we have found that there is a watershed level of GFR above which ESRF is unlikely and below which it is invariable, and that the most important harbinger of poor outcome is proteinuria. Why, however, one patient with a GFR of 45 ml/min should have no proteinuria and do well, while another with a GFR of 50 and 2 g/day of proteinuria does badly, remains to be determined. Long term studies [ 27 ] and follow up observations on outcome [ 23 ] in patients with reflux nephropathy confirm no benefit from anti-reflux surgery. This is consistent with our current view that progressive nephropathy is nephrogenic rather than urological in origin. Conclusions Patients with a GFR of <60 ml/min need careful follow up and we would recommend that anyone with increasing proteinuria, or proteinuria >50 mg/mmol (0.5 g/d) is started on an angiotensin antagonist to reduce proteinuria and slow the rate of progression of the renal failure. Competing interests GHN has been a consultant and lectured for MSD regarding both ACE inhibitors and angiotensin II receptor blockers. G Thomson, D Nitsch, RG Woolfson, JO Connolly, and CRJ Woodhouse have no conflict of interest. Authors' contributions GN was responsible for the study design, management of patients, and writing of the report. GT collected clinical data from medical records; DN contributed to analyses and writing of the report; RW and JC contributed to patient management and writing of the report. CW was responsible for the continuing urological supervision of the patients and writing of the report. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC526254.xml
539360	Modification of the mean-square error principle to double the convergence speed of a special case of Hopfield neural network used to segment pathological liver color images	Background This paper analyzes the effect of the mean-square error principle on the optimization process using a Special Case of Hopfield Neural Network (SCHNN). Methods The segmentation of multidimensional medical and colour images can be formulated as an energy function composed of two terms: the sum of squared errors, and a noise term used to avoid the network to be stacked in early local minimum points of the energy landscape. Results Here, we show that the sum of weighted error, higher than simple squared error, leads the SCHNN classifier to reach faster a local minimum closer to the global minimum with the assurance of acceptable segmentation results. Conclusions The proposed segmentation method is used to segment 20 pathological liver colour images, and is shown to be efficient and very effective to be implemented for use in clinics.	Background Segmentation is an important step in most applications that use medical image data. For example, segmentation is a prerequisite for quantification of morphological disease manifestations and for radiation treatment planning [ 1 , 2 ], for construction of anatomical models [ 3 ], for definitions of flight paths in virtual endoscopies [ 4 ], for content-based retrieval by structure [ 5 ], and for volume visualization of individual objects [ 2 ]. A Number of algorithms based on approaches such as histogram analysis, regional growth, edge detection and pixel classification have been proposed in other articles of medical image segmentation. In recent years, Artificial Neural Networks (ANNs) have been proposed as an attractive alternative solution to a number of pattern recognition problems. In our previous works [ 6 ], we have explored the potential of a Special Case of Hopfield Neural Network (SCHNN) in segmenting cerebral images obtained using the Magnetic Resonance Imaging (MRI) technique. Hopfield network for the optimization applications consists of many interconnected neuron elements. The network minimizes an energy function of the form: where N is the number of neurons, V k is the output of the k th neuron, I k is the bias term, and T kl is the interconnection weight between the k th and l th neurons. The energy function used in the segmentation problem is slightly different from the one defined by Hopfield and the arguments are given in [ 7 ]. The results that have been obtained in [ 6 ] were preferable to those obtained using Boltzmann Machine (BM) and the conventional ISODATA clustering technique. Also, in [ 8 ] we have shown that SCHNN is also able to make crisp segmentation of pathological liver colour images. However, during our study attempt to improve the segmentation process, we found that SCHNN segmentation results depend strongly on some parameters in the energy function formulating the classification problem. A summery of this study follows. Methods The segmentation problem of an image of N pixels is formulated in [ 8 ] as a partition of the N pixels among M classes, such that the assignment of the pixels minimizes a criterion function. The SCHNN classifier structure consists of a grid of N × M neurons with each row representing a pixel and each column representing a cluster. The network classifies the image of N pixels of P features among M classes, in a way that the assignment of the pixels minimizes the following criterion function: where R kl is the Mahalanobis distance measure between the k th pixel and the centroid of class l , R kl is also equivalent to the error committed when a pixel k is assigned to a class l . The index n in is the power or weight of the considered error in the energy function of the segmentation problem, and V kl is the output of the kl th neuron. N kl is a N × M vector of independent high frequency white noise source used to avoid the network being trapped in early local minimums. The term c ( t ) is a parameter controlling the magnitude of noise which is selected in a way to provide zero as the network reaches convergence. The minimization is achieved by using SCHNN and by solving the motion equations satisfying: where U kl is the input of the k th neuron, and μ ( t ) is a scalar positive function of time, used as heuristically motivated stopping criterion of SCHNN, and is defined as in [ 6 ] by: β ( t ) = t ( T s - t ) (4) where t is the iteration step, and T s is the pre-specified convergence time of the network which has been found to be 120 iterations [ 6 ]. The network classifies the feature space, without teacher, based on the compactness of each cluster calculated using Mahalanobis distance measure between the k th pixel and the centroid of class l given by: where X k is the P-dimensional feature vector of the k th pixel (here P = 3 with respect to the RGB color space components), is the P-dimensional centroid vector of class l , and Σ l is the covariance matrix of class l . The segmentation algorithm is described as follows [ 8 ]. Step 1 Initialize the input of the neurons to random values. Step 2 Apply the following input-output relation, establishing the assignment of each pixel to only and only one class. Step 3 Compute the centroid and the covariance matrix Σ l of each class l as follows: where n l is the number of pixels in class l , and the covariance matrix is then normalized by dividing each of its elements by . Step 4 Update the inputs of each neuron by solving the set of differential equations in (2) using Eulers approximation: Step 5 if t < T s , repeats from Step 2 , else terminated. For this study, a total of 20 liver tissue sections were provided by the pathological division of National Cancer Center in Tokyo. These sections were taken using needle biopsy, stained with hematoxylin and then magnified with an optical microscope. Figure 1 shows a true RGB color image of liver tissue of 768 × 512 pixels. We have used the above described SCHNN classifier with the image components in the R.G.B color space. The number of classes is fixed to five based on medical information. These classes are the contour of the image, the cell's nuclei, the cytoplasm, the fibrous tissues, and the class of both blood sinus and fat cells. Figure 1 A sample of pathological liver colour image in true colour (Red, Green, and Blue). The cells nuclei are represented by a circular shape in dark purple colour, the cytoplasm regions are coloured purple, the circular objects in white represent the fat cells, and the remaining objects in wave shape and white colour represent the fibrous tissues and blood sinus. The contour of the image is black. Figure 2 shows the curves of SCHNN energy function during the segmentation of the sample shown in Figure 1 with T s values between 30 and 120 iterations. Similar curves were obtained for the rest of the images of the dataset. As it is illustrated in Figure 2 . The curve corresponding to T s = 120 iterations gives the optimal solution, the same as it is with MRI data [ 6 ]. Figure 2 SCHNN energy function curves during the segmentation of the sample shown in Figure 1 using different values of the pre-specified convergence time Ts. In order to study the effect of the weight of the Mahalanobis distance R kl in the cost function (2), we have provided a simple modification to the above algorithm as follows: Step 1 Use the same random initialization N × M matrix, as input of the neurons, when minimizing the energy function (1) with different error's weight n . This condition is added to the algorithm in order to make sure that the random field does not have any effect on the generated results. Step 2 trough Step 5 remain the same. Results Figure 3 shows different curves of the optimization of the energy function of the segmentation of the sample shown in Figure 1 using SCHNN with the above modification (Step 1) with respect to different values of the variable n in equation (2). As aforementioned, the pre-specified convergence time of SCHNN is fixed to T s = 120 iterations. However, we can clearly see from Figure 3 that with a higher value of n in Equation (2), the same convergence point or a close position is reached in half the time of the one reached with n = 2 and T s = 120 iterations. So, this raises the following question: what is the type of relation between the variable n in (2) and the pre-specified convergence time T s ? Figure 3 Shows different curves of the optimization of the energy function of the segmentation, of the sample in Figure 1, by considering different values of the variable n in equation (2) and with pre-specified convergence Time Ts = 120 iterations. Before answering this question, it is essential to know at this level what is the best value of n that corresponds to the optimum solution with T s = 120 iterations. From Figure 4 , it can be seen that n = 6 gives the optimum solution with T s = 120. Similar figures to Figure 3 and Figure 4 were obtained with the rest of the images in the dataset. Figure 4 This curve is extracted from Figure 3 , it connects the convergence values of the energy function of the segmentation problem, of the sample in Figure 1, by considering different values of the variable n in Equation (2), with the same random initialization matrix for all n values, and with a pre-specified convergence time Ts = 120 iterations. Discussion Analysis of the pre-specified convergence time effect In order to study the effect of the pre-specified time, we repeated the above experiments with different Ts values. We realized that each value of Ts corresponds to a value of n , in Equation (2). When both ( Ts and n ) used together they give a local optima in the energy landscape of SCHNN. Figure 5 shows the curves linking the convergence values of SCHNN with respect to the value of n in Equation (2) that are obtained with Ts values 120, 60, and 30. We realized that the curves corresponding to Ts = 120 and Ts = 60 intersect in their optimum solutions obtained with n = 6, and the two curves are similar when n is in the range 5–10. However, the curve corresponding to Ts = 30, shows higher error at convergence of all values of n . Figure 5 Curves of the energy function of SCHNN at convergence with respect to values of the variable n in Equation (2) and for different pre-specified convergence time Ts. The green and red curves correspond to Ts = 60 and Ts = 120, respectively, are almost identical when n is in the range 5–10. Analysis of the SCHNN random initialization effect In order to see the effect of the random initialization on the results of the algorithm described in section 3, we have executed the same algorithm with different initialization matrices and the curves of the convergence values of SCHNN corresponding to these initializations are shown in Figure 6 . As it is clear from the curves, in Figure 6 , the random initialization does not have effect on the variable n in (2) when it takes the value of six where SCHNN gives an optimum and acceptable results that agree with the pathological experts point of views. However, with other values of n, the random initialization may affect the solution of the problem, or in other words, may affect the error of the SCHNN at convergence as shown in Figure 6 . Figure 6 Curves of the energy function of SCHNN at convergence with respect to values of the variable n in Equation (2) and for different initialization matrices. Conclusions We analyzed the effect of considering the mean-square error in formulating the segmentation problem of multidimensional medical images. We have shown, empirically, that considering an integer power equal to six, of the error in the energy function of the problem, helped SCHNN to converge twice as fast as the same optimal solution obtained with the mean-square error algorithm. This result is promising to make our segmentation method useful for a Computer Aided Diagnosis (CAD) system for liver cancer and the like. In our future work, we will study deeply the effect of the random initialization and its effect on the segmentation result and on the SCHNN classifier. Competing interests The author(s) declare that they have no competing interests. Authors' contributions Rachid carried out the theoretical study, the sequence alignment and drafted the manuscript. Mohammed participated in the design of the study and performed the analysis and helped to draft the manuscript. All authors read and approved the final manuscript. Figure 7 Segmentation result of the sample in Figure 1, obtained using SCHNN in optimizing equation (1) with n = 2, and a pre-specified convergence time Ts = 120 iterations. The cells nuclei are represented by a circulate shape with white colour, surrounded by the red regions representing the cytoplasm of the cells, fat cells are coloured blue, and the fibrous tissues and blood sinus are coloured green. Figure 8 Segmentation result of the sample liver pathological image in Figure 1, obtained using SCHNN in optimizing equation (2) with n = 6, and a pre-specified convergence time Ts = 120 iterations. Figure 9 Segmentation result of the sample liver pathological image in Figure 1, obtained using SCHNN in optimizing equation (2) with n = 6, and a pre-specified convergence time Ts = 60 iterations. Figure 10 Segmentation result of the sample liver pathological image in Figure 1, obtained using SCHNN in optimizing equation (2) with n = 12, and a pre-specified convergence time Ts = 30 iterations Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC539360.xml
533881	Protective role of Scoparia dulcis plant extract on brain antioxidant status and lipidperoxidation in STZ diabetic male Wistar rats	Background The aim of the study was to investigate the effect of aqueous extract of Scoparia dulcis on the occurrence of oxidative stress in the brain of rats during diabetes by measuring the extent of oxidative damage as well as the status of the antioxidant defense system. Methods Aqueous extract of Scoparia dulcis plant was administered orally (200 mg/kg body weight) and the effect of extract on blood glucose, plasma insulin and the levels of thiobarbituric acid reactive substances (TBARS), hydroperoxides, superoxide dismutase (SOD), catalase (CAT), glutathione peroxidase (GPx), glutathione-S-transferase (GST) and reduced glutathione (GSH) were estimated in streptozotocin (STZ) induced diabetic rats. Glibenclamide was used as standard reference drug. Results A significant increase in the activities of plasma insulin, superoxide dismutase, catalase, glutathione peroxidase, glutathione-S-transferase and reduced glutathione was observed in brain on treatment with 200 mg/kg body weight of Scoparia dulcis plant extract (SPEt) and glibenclamide for 6 weeks. Both the treated groups showed significant decrease in TBARS and hydroperoxides formation in brain, suggesting its role in protection against lipidperoxidation induced membrane damage. Conclusions Since the study of induction of the antioxidant enzymes is considered to be a reliable marker for evaluating the antiperoxidative efficacy of the medicinal plant, these findings suggest a possible antiperoxidative role for Scoparia dulcis plant extract. Hence, in addition to antidiabetic effect, Scoparia dulcis possess antioxidant potential that may be used for therapeutic purposes.	Background The neurological consequences of diabetes mellitus in the Central Nervous System (CNS) are now receiving greater attention. Cognitive deficits, along with morphological and neurochemical alterations illustrate that the neurological complications of diabetes are not limited to peripheral neuropathies [ 1 ]. The central complications of hyperglycemia also include the potentiation of neuronal damage observed following hypoxic/ischemic events, as well as stroke [ 2 ]. Glucose utilization is decreased in the brain during diabetes [ 2 ], providing a potential mechanism for increased vulnerability to acute pathological events. Oxidative stress, leading to an increased production of reactive oxygen species (ROS), as well as lipidperoxidation, is increased in diabetes [ 3 ] and also by stress in euglycemic animals [ 4 ]. Similarly, oxidative damage in rat brain is increased by experimentally induced hyperglycemia [ 5 ]. Under experimental conditions, hyperglycemia dramatically increases neuronal alterations and glial cell damage caused by temporary ischaemia [ 6 ]. Several lines of evidence indicate that the modified oxidative state induced by chronic hyperglycemia [ 7 ] may contribute to nervous tissue damage: free radical species impair the central nervous system, attacking neurons and schwann cells [ 8 ] and the peripheral nerves [ 9 ]. Due to their high polyunsaturated lipid content, schwann cells and axons are particularly sensitive to oxygen free radical damage: lipidperoxidation may increase cell membrane rigidity and impair cell function. Increases in superoxide production are observed in the serum of Type 1 diabetic patients and was reduced with improved glycemic control [ 10 ]. Lipidperoxidation products are also increased in the brains of Type 1 diabetic rats [ 11 ] and Type 2 diabetic mice [ 8 ]. Diabetes and stress mediated increases in oxidative stress, as well as decreases in antioxidant activity, may make the brain more vulnerable to subsequent pathological events. Nowadays, the use of complementary/alternative medicine and especially the consumption of botanicals have been increasing rapidly worldwide, mostly because of the supposedly less frequent side effects when compared to modern western medicine [ 12 ]. Scoparia dulcis L (Scrophulariacae), a folk-medicinal plant known as sweet broomweed, has been used as a remedy for diabetes mellitus in India [ 13 ] and for hypertension in Taiwan [ 14 , 15 ]. A number of active principles from Scoparia dulcis include scoparic acid A, scoparic acid B and scoparic acid D [ 16 ], scopadulcic acid A and B, scopadulciol [ 17 ] and Scopadulin [ 18 ] that have been identified as contributor to the observed medicinal effect of the plant. Among them, scopadulcic acid B (SDB) and scopadulciol (SDC) were found to be unique biomolecules with inhibitory effects on replication of herpes simplex virus type 1 (HSV-1) [ 16 ], gastric proton pump and bone resorption stimulated by parathyroid hormone (PTH) [ 18 ]. In addition, SDB showed antitumour promoting activities [ 17 ]. Because of their unique carbon skeleton and many sided biological activities, they were paid much attention as chemical synthetic targets by organic synthetic chemists. In a previous study, Nath (1943) studied the antidiabetic effect of Scoparia dulcis and obtained a glycoside, amellin from fresh plant and reported that it brought relief in other complications accompanied with diabetes (ie., pyorrhoea, retinopathy, joint pain, susceptibility to cold etc.) within a very short period [ 19 ]. Administration of Scoparia dulcis to STZ diabetic rats led to reduction in blood glucose [ 20 ]. In Recent studies on this plant, we have demonstrated a defective metabolism of lipid peroxides in tissues (liver, kidney and brain) of STZ diabetic rats [ 21 ] for 3 weeks treatment. Since increases in oxidative stress are associated with both long standing diabetes and stress, the present investigation was to assess the antioxidant efficacy of Scoparia dulcis in STZ diabetic rats after 6 weeks treatment and the effect produced by Scoparia dulcis was compared with Glibenclamide. Methods Animals Adult male albino Wistar rats (8 weeks), weighing 180–200 g bred in the Central Animal House, Rajah Muthiah Medical College, Annamalai University, were used. All animal experiments were approved by the ethical committee (Vide. No: 73, 2002), Annamalai University and were in accordance with the guidelines of the National Institute of Nutrition, Indian Council of Medical Research, Hyderabad, India. The animals were fed ad libitum with normal laboratory pellet diet (Hindustan Lever Ltd., India) and water. Animals were maintained under a constant 12 h light and dark cycle and at an environmental temperature of 21–23°C. Drugs and chemicals All the drugs and biochemicals used in this experiment were purchased from Sigma Chemical Company Inc., St Louis, Mo, USA. The chemicals were of analytical grade. Plant material Whole plants of Scoparia dulcis L. (40 – 60 cm in height) were collected from Neyveli, Cuddalore District, Tamil Nadu, India in September 2001. The plant was identified and authenticated at the Herbarium of Botany Directorate in Annamalai University. A voucher specimen (No.3412) was deposited in the Botany Department of Annamalai University. Preparation of Scoparia dulcis plant extract (SPEt) Five-hundred grams of fresh whole Scoparia dulcis plants were extracted with 1.5 l of water by the method of continuous hot extraction at 60°C for 6 h according to Jain (1968) and the filtrate was concentrated at 40°C to constant weight in a rotavapor apparatus (Buchi Labortechnik AG, Switzerland). The residue collected (yield 31 g) were thick paste, green in color and gumaceaous in nature and stored at -20°C, when needed the extract was dissolved in sterile water and used in the investigation [ 22 ]. Induction of experimental diabetes STZ, freshly prepared in 10 mmol/l citrate buffer, pH 4.5, was injected to experimental animals (30 rats) intraperitoneally at a dose of 45 mg/kg body weight [ 23 ]. 48 h after STZ administration, rats with moderate diabetes having glycosuria and hyperglycemia (i.e with blood glucose of 200 – 300 mg/dl) were taken for the experiment. Experimental design In the experiment, a total of 30 rats (18 diabetic surviving rats, 12 normal rats) were used. The rats were divided into 5 groups of 6 rats each. Group 1: Normal rats. Group 2: Normal rats given Scoparia dulcis plant extract (SPEt) (200 mg/kg body weight) in aqueous solution daily using an intragastric tube for 6 weeks. Group 3: Diabetic control rats. Group 4: Diabetic rats given SPEt (200 mg/kg body weight) [ 21 ] in aqueous solution daily using an intragastric tube for 6 weeks. Group 5: Diabetic rats given Glibenclamide (600 μg/kg body weight) in aqueous solution daily using an intragastric tube for 6 weeks [ 24 ]. All doses were started after 48 h STZ injection. No detectable irritation or restlessness was observed after each drug or vehicle administration. No noticeable adverse effect (i.e., respiratory distress, abnormal locomotion and catalepsy) was observed in any animals after the drug administration. Blood samples were drawn at weekly intervals till the end of study (ie. 6 weeks). At the end of 6 th week, all the rats were killed by decapitation (Pentobarbitone sodium) anaesthesia (60 mg/kg). Blood was collected in two different tubes (i.e.,) one with anticoagulant – potassium oxalate and sodium fluoride for plasma and another without anticoagulant for serum separation. Plasma and serum were separated by centrifugation. Whole Brain was immediately dissected out, washed in ice cold saline to remove the blood. The brains were weighed and 10% tissue homogenate was prepared with 0.025 M Tris – HCl buffer, pH 7.5. After centrifugation at 200 rpm for 10 min, the clear supernatant was used to measure thiobarbituric acid reactive substances (TBARS), hydroperoxides and GPx activity. For the assay of SOD, CAT, GST and GSH, the brains were weighed and 10% homogenate was prepared with 0.2 M, phosphate buffer pH 8.0. After centrifugation, the clear supernatant was used for the assay of enzyme activities. Biochemical analysis Estimation of blood glucose and plasma insulin Blood glucose was determined by the O-toluidine method [ 25 ]. Plasma insulin was assayed by ELISA, using Boeheringer-Mannheim Kit with a Boeheringer analyser ES300. Estimation of lipid peroxidation Lipid peroxidation in brain was estimated colorimetrically by thiobarbituric acid reactive substances TBARS and hydroperoxides by the method of Niehius and Samuelsson [ 26 ] and Jiang et al. [ 27 ], respectively. In brief, 0.1 ml of tissue homogenate (Tris-Hcl buffer, pH 7.5) was treated with 2 ml of (1:1:1 ratio) TBA-TCA-HCl reagent (thiobarbituric acid 0.37%, 0.25 N HCl and 15% TCA) and placed in water bath for 15 min, cooled. The absorbance of clear supernatant was measured against reference blank at 535 nm. 0.1 ml of tissue homogenate was treated with 0.9 ml of Fox reagent (88 mg butylated hydroxytoluene (BHT), 7.6 mg xylenol orange and 9.8 mg ammonium ion sulphate were added to 90 ml of methanol and 10 ml 250 mM sulphuric acid) and incubated at 37°C for 30 min. The color developed was read at 560 nm colorimetrically. Hydroperoxides was expressed as mM/100g tissue. Assay of catalase (CAT) and superoxide dismutase (SOD) CAT was assayed colorimetrically at 620 nm and expressed as μmoles of H 2 O 2 consumed/min/mg protein as described by Sinha [ 28 ]. The reaction mixture (1.5 ml, vol) contained 1.0 ml of 0.01 M pH 7.0 phosphate buffer, 0.1 ml of tissue homogenate (supernatant) and 0.4 ml of 2 M H 2 O 2 . The reaction was stopped by the addition of 2.0 ml of dichromate-acetic acid reagent (5% potassium dichromate and glacial acetic acid were mixed in 1:3 ratio). SOD was assayed utilizing the technique of Kakkar et al. [ 29 ] based on inhibition of the formation of nicotinamide adenine dinucleotide, phenazine methosulfate and amino blue tetrazolium formazan. A single unit of enzyme was expressed as 50% inhibition of NBT (Nitroblue tetrazolium) reduction/min/mg protein. Determination of glutathione peroxidase (GPx) and reduced glutathione (GSH) GPx activity was measured by the method described by Rotruck et al. [ 30 ]. Briefly, reaction mixture contained 0.2 ml of 0.4 M Tris-HCl buffer pH 7.0, 0.1 ml of 10 mM sodium azide, 0.2 ml of tissue homogenate (homogenised in 0.4 M, Tris-HCl buffer, pH 7.0), 0.2 ml glutathione, 0.1 ml of 0.2 mM hydrogen peroxide. The contents were incubated at 37°C for 10 min. The reaction was arrested by 0.4 ml of 10% TCA, and centrifuged. Supernatant was assayed for glutathione content by using Ellmans reagent (19.8 mg of 5,5'-dithiobisnitro benzoic acid (DTNB) in 100 ml of 0.1% sodium nitrate). GSH was determined by the method of Ellman [ 31 ]. 1.0 ml of supernatant was treated with 0.5 ml of Ellmans reagent and 3.0 ml of phosphate buffer (0.2 M, pH 8.0). The absorbance was read at 412 nm. Glutathione peroxidase activity was expressed as μg of GSH consumed/min/mg protein and reduced glutathione as mg/100g of tissue. Determination of glutathione-S-transferase (GST) The GST activity was determined spectrophotometrically by the method of Habig et al. [ 32 ]. The reaction mixture (3 ml) contained 1.0 ml of 0.3 mM phosphate buffer (pH 6.5), 0.1 ml of 30 mM 1-chloro-2, 4-dinitrobenzene (CDNB) and 1.7 ml of double distilled water. After pre-incubating the reaction mixture at 37°C for 5 min, the reaction was started by the addition of 0.1 ml of tissue homogenate and 0.1 ml of glutathione as substrate. The absorbance was followed for 5 min at 340 nm. Reaction mixture without the enzyme was used as blank. The activity of GST is expressed as μmoles of GSH-CDNB conjugate formed/min/mg protein using an extinction coefficient of 9.6 mM -1 cm -1 . Estimation of protein Protein was determined by the method of Lowry et al. [ 33 ] using Bovine Serum Albumin (BSA) as standard, at 660 nm. Statistical analysis All data were expressed as mean ± SD of number of experiments (n = 6). The statistical significance was evaluated by one-way analysis of variance (ANOVA) using SPSS version 7.5 (SPSS, Cary, NC, USA) and the individual comparison were obtained by Duncan's Multiple Range Test (DMRT). A value of p < 0.05 was considered to indicate a significant difference between groups [ 34 ]. Results Table 1 shows the level of blood glucose and plasma insulin in normal and experimental groups. The level of blood glucose was significantly increased whereas the level of plasma insulin was significantly decreased in diabetic control rats. Oral administration of SPEt and glibenclamide to diabetic rats significantly reversed all these changes to near normal levels. Table 1 Effect of Scoparia dulcis on blood glucose and plasma insulin in normal and experimental rats Groups Fasting blood glucose (mg/dl) Plasma insulin (μu/ml) Normal 84 ± 4 a 12 ± 1 a Normal + SPEt (200 mg/kg) 77 ± 3 a 15 ± 1 b Diabetic control 270 ± 15 b 4 ± 0.3 c Diabetic + SPEt (200 mg/kg) 98 ± 3 c 11 ± 4 d Diabetic + Glibenclamide (600 μg/kg) 114 ± 9 d 9 ± 0.5 e Values are given as mean ± SD from 6 rats in each group. Values not sharing a common superscript letter differ significantly at p < 0.05 (DMRT); Duncan Procedure; Ranges for the level: 2.95, 3.09, 3.20, 3.22. Table 2 illustrates markers of lipidperoxidatioon namely, TBARS and hydroperoxides from brain of normal and experimental rats. The levels of TBARS and hydroperoxides were significantly increased in diabetic control rats. Administration of SPEt to diabetic rats significantly decreased the levels of lipidperoxidative markers. Treatment of normal rats with SPEt did not show significant changes in lipidperoxidation. The effect produced by SPEt was significant than glibenclamide. Table 2 Change in the levels of brain TBARS and hydroperoxides in normal and experimental rats Groups TBARS (mM/100g tissue) Hydroperoxides (mM/100g tissue) Normal 1.10 ± 0.08 a 113.20 ± 4.10 a Normal + SPEt (200 mg/kg) 0.90 ± 0.05 a 108.70 ± 2.20 b Diabetic control 1.85 ± 0.07 b 130.90 ± 1.50 c Diabetic + SPEt (200 mg/kg) 1.18 ± 0.06 c 117.22 ± 3.26 d Diabetic + Glibenclamide (600 μg/kg) 1.32 ± 0.06 d 120.40 ± 4.05 e Values are given as mean ± SD for 6 rats in each group. Values not sharing a common superscript letter differ significantly at p < 0.05 (DMRT). Duncan procedure, Range for the level 2.95, 3.09, 3.20, 3.22. For studying the effect of SPEt on antioxidant status, the activities of enzymic antioxidants SOD, CAT, GPx, GST and non-enzymic antioxidant GSH were measured (Table 3 and 4 ). The activities of enzymic and the levels of non-enzymic antioxidant were significantly decreased in diabetic control rats. They presented significant increases in diabetic rats treated SPEt. Administration of SPEt to normal rats increased the antioxidants levels with no significant differences. The effect produced by SPEt was comparable with that of glibenclamide. Table 3 Changes in activities of catalase and superoxide dismutase in brain of normal and experimental rats Groups Catalase (Units A /mg protein) Superoxide dismutase (Units B /mg protein) Normal 3.12 ± 0.29 a 7.75 ± 0.38 a Normal + SPEt (200 mg/kg) 4.00 ± 0.20 b 7.05 ± 0.28 a Diabetic control 0.86 ± 0.05 c 5.17 ± 0.30 b Diabetic + SPEt (200 mg/kg) 2.75 ± 0.20 d 7.32 ± 0.46 c Diabetic + Glibenclamide (600 μg/kg) 1.98 ± 0.11 e 6.32 ± 0.30 d Values are given as mean ± SD for 6 rats in each group. Values not sharing a common superscript letter differ significantly at p < 0.05 (DMRT). Duncan procedure, Range for the level 2.95, 3.09, 3.20, 3.22. A – μmole of H 2 O 2 consumed/minute. B – One unit of activity was taken as the enzyme reaction, which gave 50% inhibition of NBT reduction in one minute. Table 4 Changes in activities of glutathione peroxidase, glutathione-S-transferase and the levels of reduced glutathione in brain of normal and experimental rats Groups Glutathione peroxidase (Units A /mg protein) Glutathione-S-transferase (Units B /mg protein) Reduced glutathione (mg/100g tissue) Normal 3.4 1± 0.20 a 5.62 ± 0.28 a 35.19 ± 2.21 a Normal + SPEt (200 mg/kg) 3.80 ± 0.18 a 5.90 ± 0.28 a 37.12 ± 2.14 a Diabetic control 1.01 ± 0.05 b 0.81 ± 0.02 b 15.20 ± 1.44 b Diabetic + SPEt (200 mg/kg) 2.62 ± 0.15 c 2.10 ± 0.13 c 26.02 ± 2.01 c Diabetic + Glibenclamide (600 μg/kg) 1.97 ± 0.12 d 2.04 ± 0.13 c 25.50 ± 2.10 c Values are given as mean ± SD for 6 rats in each group. Values not sharing a common superscript letter differ significantly at p < 0.05 (DMRT). Duncan procedure, Range for the level 2.95, 3.09, 3.20, 3.22. A – μg of GSH consumed/min. B – μmoles of CDNB – GSH conjugate formed/min. Discussion This work is one of a series of studies showing that chronic hyperglycemia causes an imbalance in the oxidative status of the nervous tissue and that the resulting free radicals damage the brain through a peroxidative mechanism. The STZ diabetic rat serves as an excellent model to study the molecular, cellular and morphological changes in brain induced by stress during diabetes [ 7 ]. Under normal conditions, the generation of free radicals or of active species in the brain, as in other tissues, is maintained at extremely low levels [ 4 ]. Diabetes also contributes to cerebrovascular complications, reductions in cerebral blood flow, disruption of the blood brain barrier and cerebral edema [ 5 ]. All of these neurochemical and neurophysiological changes ultimately contribute to the long-term complications associated with diabetes, including morphological abnormalities, cognitive impairments and increased vulnerability to pathophysiological event [ 6 ]. In the present study, treatment with aqueous extract of Scoparia dulcis showed significant antihyperglycemic activity. The antihyperglycemic activity of this plant may be, at least in part, through release of insulin from the pancreas in view of the measured increase in the plasma insulin concentrations. Earlier studies in this lab have demonstrated a defective metabolism of lipid peroxides in other tissues of diabetic animal [ 35 , 36 ]. TBARS and hydroperoxides (lipid peroxidative markers) showed high lipidperoxidation. This may be because; the brain contains relatively high concentration of easily peroxidizable fatty acids [ 37 ]. In addition, it is known that certain regions of the brain are highly enriched in iron, a metal that, in its free form, is catalytically involved in production of damaging oxygen free radical species [ 38 ]. It has been suggested that free radical species responsible for STZ toxicity is the hydroxyl radical, formed via the metal catalyzed Haber-weiss reaction or Fenton reaction. In this process, the ferric iron is reduced by superoxide, with subsequent oxidation of ferrous iron by H 2 O 2 forming hydroxyl radical: Fe 3+ + O 2 •- → Fe 2+ + O 2 Fe 2+ + H 2 O 2 → Fe 3+ + OH* + OH - The destruction of superoxide radical or H 2 O 2 by SOD or CAT would ameliorate STZ toxicity, as would substances able to scavenge the hydroxyl radical [ 39 , 40 ]. Vulnerability of brain to oxidative stress induced by oxygen free radicals seems to be due to the fact that, on one hand, the brain utilizes about one fifth of the total oxygen demand of the body and on the other, that it is not particularly enriched, when compared with other organs, in any of the antioxidant enzymes. Relatively low levels of these enzymes may be responsible in part for the vulnerability of this tissue [ 41 ]. The altered balance of the antioxidant enzymes caused by decrease in CAT, SOD, GPx, GST and GSH activities may be responsible for the inadequacy of the antioxidant defenses in combating ROS mediated damage. The decreased activities of CAT and SOD may be a response to increased production of H 2 O 2 and O 2 by the autoxidation of glucose and non-enzymatic glycation [ 5 ]. These enzymes have been suggested as playing an important role in maintaining physiological levels of oxygen and hydrogen peroxide by hastening the dismutation of oxygen radicals and eliminating organic peroxides and hydroperoxides generated from inadvertent exposure to STZ [ 42 ]. Treatment with SPEt increased the activity of enzymes and may help to control free radicals, as Scoparia dulcis has been reported to be rich in alkaloids and terpenoids [ 16 - 18 , 43 , 44 ], well-known antioxidants, which scavenge the free radicals generated during diabetes. The increase in SOD activity may protect CAT and GPx against inactivation by O 2 •- anions as these anions have been shown to inactivate CAT and GPx [ 45 ]. Under in vivo conditions, GSH acts as an antioxidant and its decrease was reported in diabetes mellitus [ 46 ]. We have observed significant decrease in GSH levels in brain during diabetes. The decrease in GSH levels represents increased utilization due to oxidative stress [ 47 ]. The depletion of GSH content may also lower the GST activity [ 48 ]. Depression in GPx activity was also observed brain of diabetic rats. GPx has been shown to be an important adaptive response to condition of increased peroxidative stress [ 46 ]. The increased GSH content in the brain of the rats treated with SPEt and glibenclamide may be a factor responsible for inhibition of lipidperoxidation. The elevated level of GSH protects cellular proteins against oxidation through glutathione redox cycle and also directly detoxifies reactive oxygen species generated from exposure to STZ [ 48 ]. The significant increase in GSH content and GSH dependent enzymes GPx and GST in diabetic rats treated with SPEt indicates an adaptive mechanism in response to oxidative stress. Significantly lower levels of lipid peroxides in brain of SPEt treated diabetic rats and increased activities of enzymic and non-enzymic antioxidants in brain suggest that the extract reduce oxidative stress by quenching free radicals. Terpenoids and alkaloids were reported to have free radical scavenging activity and antioxidant capacity in diabetes [ 49 , 50 ]. SPEt was reported to be rich in an alkaloid-6-methoxybenzoxazolinone [ 51 ] and terpenoids such as scoparic acids A, B, C and scopadulcic acids A and B [ 16 - 18 ], which may be responsible for scavenging free radicals liberated by STZ and thus enhance both enzymic and non-enzymic antioxidants in diabetic rats treated with SPEt. Any compound, natural or synthetic with antioxidant properties that might contribute towards the partial or total alleviation of this damage may have a significant role in the treatment of diabetes mellitus. The antioxidant responsiveness mediated by Scoparia dulcis may be anticipated to have biological significance in eliminating reactive free radicals that may otherwise affect the normal cell functioning. The disfunctioning of these antioxidant enzymes has been implicated in several disorders including rheumatoid arthritis, reperfusion injury, cardiovascular diseases, immune injury as well as diabetes mellitus [ 52 ]. It may be concluded that in diabetes, brain tissue was more vulnerable to oxidative stress and showed increased lipidperoxidation. The above observation shows that the aqueous extract of Scoparia dulcis plant possesses antioxidant activity, which could exert a beneficial action against pathological alterations caused by the presence of free radicals in STZ diabetes. Conclusions The brain exhibits numerous morphological and functional alterations during diabetes. Oxidative stress, a factor implicated in the pathogenesis of diabetic complications may contribute towards some of these alterations. Treatment of diabetic rats with Scoparia dulcis plant extract significantly decreased the lipidperoxidation and significantly increased the antioxidant status. Since the study of induction of the antioxidant enzymes is considered to be a reliable marker for evaluating the antioxidant efficacy of the medicinal plant, these findings are suggestions of possible antioxidant role played by Scoparia dulcis plant extract in addition to its antidiabetic effect. Competing interests The author(s) declare that they have no competing interests Authors' contributions LP – supervised the design and co-ordination of the study ML – Practically conducted the design of the study and drafted the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC533881.xml
535557	CLOE: Identification of putative functional relationships among genes by comparison of expression profiles between two species	Background Public repositories of microarray data contain an incredible amount of information that is potentially relevant to explore functional relationships among genes by meta-analysis of expression profiles. However, the widespread use of this resource by the scientific community is at the moment limited by the limited availability of effective tools of analysis. We here describe CLOE, a simple cDNA microarray data mining strategy based on meta-analysis of datasets from pairs of species. The method consists in ranking EST probes in the datasets of the two species according to the similarity of their expression profiles with that of two EST probes from orthologous genes, and extracting orthologous EST pairs from a given top interval of the ranked lists. The Gene Ontology annotation of the obtained candidate partners is then analyzed for keywords overrepresentation. Results We demonstrate the capabilities of the approach by testing its predictive power on three proteomically-defined mammalian protein complexes, in comparison with single and multiple species meta-analysis approaches. Our results show that CLOE can find candidate partners for a greater number of genes, if compared to multiple species co-expression analysis, but retains a comparable specificity even when applied to species as close as mouse and human. On the other hand, it is much more specific than single organisms co-expression analysis, strongly reducing the number of potential candidate partners for a given gene of interest. Conclusions CLOE represents a simple and effective data mining approach that can be easily used for meta-analysis of cDNA microarray experiments characterized by very heterogeneous coverage. Importantly, it produces for genes of interest an average number of high confidence putative partners that is in the range of standard experimental validation techniques.	Background The availability of genome sequences from several model organisms, including humans, and of high-throughput technologies to study gene function is dramatically changing the approach to biological problems. In particular, the consolidated reductionist gene-by-gene strategy is being replaced by a 'modular approach', in which several genes are studied simultaneously to gather a more comprehensive picture of the many different cellular processes [ 1 ]: in living organisms, the majority of gene products are part of intricate molecular circuits, composed of physical, functional and regulatory interactions. In higher eukaryotes, the study of gene function is further complicated by the alternative use of transcriptional units, frequently resulting in the production of proteins with different or even antagonistic activities from the same genes [ 2 , 3 ]. It is well recognized that one of the most important and widespread mechanisms used by cells to regulate functional modules is the coordinate transcriptional and/or post-transcriptional modulation of mRNA levels of the interacting genes. Therefore, DNA microarrays represent a fundamental tool to unravel biological complexity on a genome-wide scale. Information concerning the expression of thousands of genes, and also of different transcripts from the same gene, can be obtained in a single experiment, and the relationships among gene expression patterns can be studied systematically [ 4 ]. The extensive use of this technology by hundreds laboratories has resulted in the production of an enormous amount of data, many of which have been deposited in public databases [ 5 , 6 ]. Besides being useful to other researchers to confirm the published results, the deposited datasets can be used as a substrate for new analysis, aimed at discovering functional modules by searching for related expression profiles. Recent studies have shown that, if the expression of two or more genes is consistently related throughout many independent microarray datasets, the genes display a significant degree of functional similarity [ 7 , 8 ]. However, if this approach were applied to predict physical and functional relationships, a very high number of false positives would still be expected. A first method that can be used to reduce the number of false positives is to consider only co-expression links that are consistent among many different experimental datasets [ 7 ]. Nevertheless, even when the co-expression of two genes is reproducibly observed under a certain number of experimental conditions, this does not imply necessarily that they are functionally related. For instance, extensive meta analysis of microarray data across different species has revealed that neighboring genes are likely to be co-expressed, even though they are not functionally related in any obvious manner [ 9 , 10 ]. Phylogenetic conservation has been recently proposed as a very strong criterion to identify functionally relevant co-expression links among genes [ 11 ]. Significant co-expression of two or more orthologous genes across many species is very likely due to selective advantages, strongly suggesting a functional relation. In fact, the comparison of data across species as distant as Homo sapiens , Saccharomyces cerevisiae , Drosophila melanogaster and Caenorhabditis elegans was very effective in identifying new genes involved in core biological functions [ 11 ]. Although extremely specific, this multi-species approach would be unable to identify the relationships among genes involved in more specialized biological processes. Since regulatory regions diverge much more rapidly than coding sequences [ 12 , 13 ], a similar approach would be predicted to succeed even when comparing expression patterns in more closely related species, such as mice and humans. In this case, the possible loss of specificity would be strongly compensated by the increased sensitivity in the identification of functional links related to mammalian-specific gene modules. This possibility has not been so far explored. Additionally, when using microarray data to establish significant correlations among gene expression profiles, almost invariably the information obtained with probes covering different gene portions is averaged [ 14 ]. Though useful in many cases to reduce the experimental noise, this procedure could result in a significant loss of information in the case of genes expressing different isoforms with distinct expression patterns [ 15 ]: on one hand the isoform-specific expression profiles would not be detected; on the other hand, the average expression profile would be artificial and non-informative. In this study we describe CLOE ( C oexpression-based L inking of O rthologous E STs) a new data mining method for the identification of transcripts showing evolutionary conserved co-expression in cDNA microarray datasets. This approach is based on the pairwise comparison of data from two species. The predictive capability of the method was proved by comparing human with mouse data. Our results show that CLOE is a valuable tool for biologists that can be used to identify putative partners for genes of interest and/or to predict some of their functional properties. Results and discussion The top percentiles of expression similarity ranked lists obtained with human-mouse orthologous ESTs pairs are strongly enriched of orthologous ESTs The aim of our method is to use the available microarray expression data to identify high- confidence putative partners for genes of interest. The basic assumption is that, if two or more genes are part of a functional module, conserved between two species, they will be likely co-expressed in both species. In contrast, if the co-expression of two genes in one species has no functional meaning, it should not be conserved in the other. A flow chart of the method is given in Figure 1 . In summary, after finding representative EST clones for the gene of interest in cDNA microarray datasets of both species, we order all the ESTs in each dataset according to similarity of their expression pattern with that of the chosen ESTs. We then extract the orthologous pairs found in a given top percentage of the ranked lists. Moreover, to obtain a functional characterization of the identified putative partners, we analyze the co-expressed orthologous pairs for overrepresentation of Gene Ontology (GO) keywords [ 16 ]. Although in principle our method could be applied to every pair of organisms for which cDNA microarray data are available, we decided to compare the human and mouse datasets contained in the Stanford Microarray Database (SMD) [ 6 ] (2803 experiments for 74588 EST probes and 145 experiments for 37521 ESTs, respectively, data downloaded in Jan. 2004). The first reason for doing so is that this comparison is particularly relevant in the perspective of identifying mammalian-specific gene modules. The second is that, considering the widely different number of experiments and the relatively short phylogenetic distance between the two species, this represents a particularly severe test. As a first proof of the method's effectiveness, and in order to empirically determine a reasonable default cutoff for obtaining the final list of candidates, we analyzed whether genes with the highest ranks in the single organism lists are actually enriched of orthologous sequences. To this aim, we randomly chose 100 orthologous gene pairs represented in both the human and mouse datasets and selected, for each one, the most representative EST (i.e. the probes with the highest number of experiments in each dataset). We then generated the respective ranked lists, subdivided them in 1% rank intervals and analyzed the number of orthologous pairs in corresponding rank intervals. As a control, we performed the same analysis on an equal number of randomly chosen (and hence non-orthologous) human-mouse EST pairs. The analysis was repeated three times with essentially identical results. As shown in Figure 2 , compared to the control, a strong average enrichment of orthologous pairs was observed in the top 1% rank interval (p = 1.6·10 -94 , chi square test). The difference was still very significant in the 2% rank interval, even though with a much higher p value (p = 1.3·10 -10 ), but was not detectable below that threshold. Interestingly, a slight enrichment was also observed in the last rank interval (average number of orthologous pairs equal to 7.8 for CLOE and 5.7 for random lists, p = 2.2·10 -7 ). The latter observation is consistent with the previously noted fact that negative correlations tend to be less common and significant than positive correlations [ 7 ]. Based on these results, we chose a top 1% cutoff for all the following analysis. Predictive value of CLOE compared to single organisms and multiple species co-expression analysis We next investigated the effectiveness of our approach, by comparing it to single and multiple organisms co-expression analysis. To address this point, we analyzed the ability of the three methods to predict known physical and functional interactions among mammalian proteins. Protein-protein complexes have begun to be determined on a genome-wide scale only for Saccharomyces cerevisiae [ 17 ], Drosophila melanogaster [ 18 ] and Caenorhabditis elegans [ 19 ], but no comparable datasets have been so far published for mammalians, making it impossible to perform a systematic comparison. Therefore, we focused on three supramolecular structures, which have been analyzed by different proteomic strategies at a high level of detail: the centrosome [ 20 ] (110 proteins), the post-synaptic density [ 21 ] (105 proteins) and the TNF-alpha/NFkB signalosome [ 22 ] (128 proteins). For the single organism and CLOE approaches, the analysis was restricted to proteins represented in the SMD by at least one human and one murine EST probe. These corresponded to 62, 67 and 97 ESTs pairs, respectively, covering on average 66% of the proteins found in these complexes. In contrast, only 37% of these genes were represented in the multiple species network, thus confirming that the previous two methods can be applied to a number of genes much higher than the latter. The average number of candidates produced by CLOE for each analyzed protein was approximately 17, which represents a strong reduction if compared to the single organism approach (746 and 375 for the human and mouse datasets, respectively). On the other hand, the average percentage of CLOE links that correspond to a documented protein-protein interaction was 6.6 %, i.e. approximately 5 times higher than that obtained with the single organism method (Table 1 ). Significantly, the predictive value of human-mouse CLOE was very similar to that obtained by the multiple species co-expression network (Table 1 ). Since considering only the proteomically-identified interactions could lead to a strong underestimation of the positive results, as low affinity and purely functional interactions would be completely excluded, we decided to evaluate the predictive power of the three different methods respect to a less stringent functional index. To this aim, we first determined which GO keywords represent the best annotation of the three complexes, by identifying the ones that are significantly overrepresented in the annotation of the respective proteins. Then, every predicted candidate partner obtained with the three methods for all analyzed proteins was considered as a true positive if it is annotated to at least one of the overrepresented keywords of the corresponding complex. The results of this analysis are summarized in Table 2 . Interestingly, even though also in this case our approach and the multiple species comparison gave, on average, a higher percentage of compatible predictions, this was not dramatically different from the single-organism method. These results strongly suggest that, compared to the single organism approach, the highly reduced number of candidate partners produced by multiple organism co-expression analysis and CLOE is strongly enriched of genes characterized by more stringent functional relationships. Conclusions We have shown that CLOE represents a very flexible and effective data mining approach to infer a list of putative partners and the potential functions for genes of interest. It can be easily used for meta-analysis of cDNA microarray experiments characterized by very heterogeneous coverage, producing significant results even when data from two species as close as mouse and human are analyzed. Compared to single organisms co-expression analysis, it strongly reduces the number of potential partners for genes of interest, producing a list of targets that is highly enriched in physically interacting proteins. On the other hand, compared to multiple species co-expression analysis, it retains a comparable specificity, but can find candidate partners for a greater number of genes. Since the number of candidate partners obtained by this analysis is, on average, in the range of standard experimental validation techniques, we believe CLOE represents a useful tool for the exploration of gene function. Methods Definition of orthologous ESTs The first step of our procedure is the identification of orthologous ESTs in the two datasets. Although many different methods could be used to this purpose, we relied on the InParanoid algorithm [ 23 ], which is ideally suited for the identification of orthologous sequences between two species. The results shown were obtained using the pre-computed release 2.6 of the InParanoid database [ 24 ]. ESTs were linked to InParanoid clusters through their UniProt codes [ 25 , 26 ], and their association to UniGene identifiers [ 27 , 28 ]. Choice of representative probes for the gene of interest The procedure has been implemented for the analysis of cDNA microarray datasets, such as those contained in the SMD [ 6 , 29 ]. However, it could be adapted, with minor modifications, to the analysis of Affymetrix datasets. Ratiometric data for the different organisms are not subjected to any further normalization, and downloaded as log-transformed (base 2) ratios. Within these datasets, the number of ESTs representing a given transcription unit, as well as the number of valid experiments for each EST, are highly variable. While this feature would pose serious problems, if one should attempt to average the data of all the probes belonging to the same gene, it may offer extremely valuable information when every EST is considered independently, since each clone explores the expression properties of a particular group of exons, in a particular combination of experimental situations. Non-correlated or anti-correlated expression between two well-represented EST probes belonging to the same gene would strongly suggest that they correspond to alternatively expressed transcripts. For this reason, we decided not to merge the data of probes belonging to the same gene, but to treat separately every EST probe in the datasets. The choice of the most representative EST for the genes of interest represents a particularly critical aspect of our procedure. If interested to a single gene, the more exhaustive solution to this problem would be to generate and evaluate a list of candidate partners for every possible pairwise combination of ESTs probes. However, this would greatly complicate the analysis should one be interested in analyzing the potential partners of many genes. An alternative possibility is to focus, for every Unigene cluster, on the most representative ESTs, i.e. the probes with the highest number of experiments. For these reasons, the decision about what ESTs to analyze is left to the end user. Our implementation of the method can accept as input both a UniGene cluster ID or the results of a BLAST search performed with the sequence of interest against the EST database. In both cases, it retrieves a list of all probes found in the two datasets for the orthologous UniGene clusters. To help the user decide which ESTs to analyze, the program provides basic information about all the EST probes representing the gene of interest in the two datasets. Moreover, to help the user identify the most representative EST probe in each dataset, i.e. the probe with the highest number of experiments, it also provides the number of valid data points for every probe. Finally, to assess the redundancy of the information provided by the different probes, i.e. whether they represent different experiments and display similar/different expression profiles, the program calculates, for every pair of probes belonging to the same UniGene cluster, the number of common experiments and the Pearson correlation coefficient between their expression profiles. Identification of orthologous sequences coexpressed in both species After finding representative EST clones for the gene of interest in both species, we calculate, for each one, the Pearson correlation coefficient ( r ) with every other EST in the respective dataset. The raw r is the normalized for the number of common data points ( n ) obtained for the analyzed ESTs. This is done by multiplying r for , since the statistical significance of r is a function of the product · r . Such normalization is particularly important when, as in our case, n has a very wide range of variation. The ESTs are then ranked by decreasing normalized score. Finally, a user-defined top percentage of the two ranked lists is compared to identify those ESTs that are associated to the same InParanoid cluster ID. A non-redundant list of the positive InParanoid clusters, sorted by the average highest rank obtained in the two organisms, represents the main output of the program (see Table 3 for an example). Clearly, the choice of the cutoff top rank percentage represents a critical parameter, which may strongly influence the number of identified candidates. The empirical determination of an average best cutoff is reported in the results. Functional characterization of the co-expressed orthologous clusters After obtaining a list of putative partners for the genes under study, we analyze their functional characterization according to the GO vocabulary [ 16 , 30 ]. This is very useful to obtain new insight about the putative functional properties of the gene of interest. GO terms are associated to ESTs through the corresponding UniProt identifiers. For each list of candidates, we compute the prevalence of all GO terms among the annotated ESTs, and the probability that such prevalence would occur in a randomly chosen set of ESTs of the same size. We always consider a gene annotated to a GO term if it is directly annotated to it or to any of its descendants in the GO graph. For a given GO term t let K(t) be the total number of ESTs annotated to it in the first organism dataset that have an orthologous sequence in the second organism dataset, and k(m, t) the number of ESTs annotated to it in the final list S(m). If J and j(m) denote the number of orthologous ESTs in the dataset and in S(m) respectively, such probability is given by the right tail of the appropriate hypergeometric distribution: where As an example, the results of the analysis performed on the output shown in Table 4 . A similar strategy was used in all other cases where GO keywords overrepresentation test were performed. Availability The programs used for this work are publicly available at the URL: . This page contains the following files: programs.zip (the program files and the corresponding General Public License); readme.txt (detailed instructions for using our program). Authors' contributions MP: development of most of the software, execution of the described analysis. PP: development of the routines used to calculate the normalized Pearson, supervision of the statistical analysis. LS: assessment of the biological significance of results. FD: supervision of the project, manuscript writing.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535557.xml
524480	Tools for loading MEDLINE into a local relational database	Background Researchers who use MEDLINE for text mining, information extraction, or natural language processing may benefit from having a copy of MEDLINE that they can manage locally. The National Library of Medicine (NLM) distributes MEDLINE in eXtensible Markup Language (XML)-formatted text files, but it is difficult to query MEDLINE in that format. We have developed software tools to parse the MEDLINE data files and load their contents into a relational database. Although the task is conceptually straightforward, the size and scope of MEDLINE make the task nontrivial. Given the increasing importance of text analysis in biology and medicine, we believe a local installation of MEDLINE will provide helpful computing infrastructure for researchers. Results We developed three software packages that parse and load MEDLINE, and ran each package to install separate instances of the MEDLINE database. For each installation, we collected data on loading time and disk-space utilization to provide examples of the process in different settings. Settings differed in terms of commercial database-management system (IBM DB2 or Oracle 9i), processor (Intel or Sun), programming language of installation software (Java or Perl), and methods employed in different versions of the software. The loading times for the three installations were 76 hours, 196 hours, and 132 hours, and disk-space utilization was 46.3 GB, 37.7 GB, and 31.6 GB, respectively. Loading times varied due to a variety of differences among the systems. Loading time also depended on whether data were written to intermediate files or not, and on whether input files were processed in sequence or in parallel. Disk-space utilization depended on the number of MEDLINE files processed, amount of indexing, and whether abstracts were stored as character large objects or truncated. Conclusions Relational database (RDBMS) technology supports indexing and querying of very large datasets, and can accommodate a locally stored version of MEDLINE. RDBMS systems support a wide range of queries and facilitate certain tasks that are not directly supported by the application programming interface to PubMed. Because there is variation in hardware, software, and network infrastructures across sites, we cannot predict the exact time required for a user to load MEDLINE, but our results suggest that performance of the software is reasonable. Our database schemas and conversion software are publicly available at .	Background MEDLINE is a large biomedical bibliographic database that is well known to users around the globe. It contains over 12 million citations from over 4,600 journals. MEDLINE is a rich source of biomedical text that lends itself well to research on text mining, information extraction, and natural language processing in biomedical domains. The usual way in which users query MEDLINE is through PubMed, the web-based interface and search engine provided by the National Library of Medicine (NLM) [ 1 ]. PubMed allows individuals to conduct searches directly by entering search terms on web pages and viewing results, and supports software-based queries across the Internet with programming utilities offered by the NLM [ 2 ]. Because we were interested in developing custom-made programs that query MEDLINE, the programming utilities offered by the NLM were an obvious choice to consider. However, due to risks of server overload, the NLM places limits on the number of queries that a user can send in a given time interval, and requests that large-volume queries be done on nights or weekends [ 3 ]. By contrast, a local version of MEDLINE gives software developers greater control over how they use the data, and facilitates the development of customizable interfaces. In this report, we describe the design and implementation of the database schema and database loading tools we have built to enable others to produce similar systems at their sites. The entire content of MEDLINE is available as a set of text files formatted in XML (eXtensible Markup Language) [ 4 ]. The NLM distributes these files at no cost to the licensee, but the files are large and not easily searched without additional indexing and search tools. For example, in the 2003 release of MEDLINE, there are 396 files (which cover citations through 2002), and the total uncompressed size of these files is 40.8 gigabytes (GB). Although it is relatively inexpensive to store 40.8 GB of data, it is not easy to manipulate data of that magnitude without good software support. Relational databases are a natural choice for storing MEDLINE because they are able to handle large amounts of data, offer built-in approaches to query optimization, and enable the developer to create indexes. Additionally, the standard query language for relational databases, SQL (Structured Query Language), enjoys widespread familiarity and can be integrated with text-database queries in some commercial systems. Alternatives to relational databases are XML-based databases, which have recently emerged as another option for storing information transmitted in XML format.XML databases may exist as standalone databases or as add-ons to relational systems. The MEDLINE data set would be an excellent test of the capabilities of these databases because of its size and complexity. We focused on relational databases because they are currently more ubiquitous and standardized, and interested users are more likely to be comfortable with relational database technology. In the remainder of this report, we describe the software tools we developed for converting MEDLINE in XML files to MEDLINE in a relational database, and provide a few sample queries that demonstrate the flexibility of the resulting system. Implementation Database schema The NLM provides a DTD (Document Type Definition) that defines the structure of data in the MEDLINE XML files [ 5 - 7 ]. From this DTD, we designed a relational database schema. Although developers of MEDLINE at the NLM maintain their own version of MEDLINE in a relational database, the schema they use is not directly applicable to our purposes, because their implementation contains tables and data that are used for maintenance and that are not relevant to external users. Thus, it was appropriate for us to design our own schema based on the specific content of the XML files, as defined by the DTD. Other groups currently license the same MEDLINE XML files and may have implemented all of MEDLINE in a relational database, but if so, their database schemas are not well publicized and were not available. There are multiple ways in which one can design a schema from the same DTD, because DTD elements and attributes can be mapped to tables and fields in different ways. Certain design decisions may favor speed at loading time, and others may favor speed and ease of use at query time. Loading records associated with 12 million citations into a database is very time consuming, and the time can be minimized if lookups to the database are minimized during loading. In general, we aimed to minimize lookups even if that meant repeating information in the database. As developers often do in the design of data warehouses, we chose to de-normalize the schema in order to improve read-only query performance, which is the typical data access pattern in our workload. The typical table contains a PubMed identifier (PMID) in one column, and data related to that PMID in the remaining columns. Figures 1 and 2 show original representation of content from a portion of the DTD and a corresponding table that follows the typical table structure. Our development team included a group of researchers from the University of California at Berkeley and another group from Stanford University. We shared similar goals in that we all wanted to load MEDLINE into a relational database, but because we were in two different departments at two different institutions, we had different project constraints and timelines. Thus, our groups were loosely associated in the software development process, but not closely integrated, and therefore, the original schema that we shared diverged. The result was three MEDLINE schemas and three software variants: One schema was used with Java code developed at Berkeley, another schema was used with Berkeley's code modified to run at Stanford, and the third schema was used with Perl code developed at Stanford. Here we describe the underlying design that influenced all three of the schemas. (The schema used for the Java program did not include information from DTD elements DataBankList and AccessionNumberList. This has been corrected in the most recent version of the software available on the Berkeley website.) The main table in the schema is medline_citation . The medline_citation table contains the PMID as the primary key and has additional columns that correspond to single-valued elements in the DTD, where the values of those elements depend on the PMID. The medline_abstract table is similar in that it has a PMID as the primary key and columns of data that depend on the PMID. Since document abstracts are larger than the other data types, we placed them in a separate table. However, since abstracts are stored as CLOBs (Character Large Objects), they are not stored in the same pages as the rest of the fields in the medline_abstract table. Therefore, in a more recent implementation, we removed the medline_abstract table from the schema, and added the abstract_text field as a CLOB in the medline_citation table. This change reduces the number of tables by one, and eliminates the need for a join between the medline_citation and medline_abstract tables. Some tables in the schema have more than one row corresponding to the same PMID. Columns in these tables map to multi-valued elements in the DTD. Examples are the table medline_keyword_list , which stores multiple values of keyword for a given PMID, and medline_gene_symbol_list , which stores multiple values of gene_symbol for a given PMID. The element Article in the DTD has a one-to-one relationship between an article and a PubMed identifier. Rather than giving Article its own table, we put single-valued data from Article into the table medline_citation . To keep track of the name of the file from which data are read for a given citation, we added the field xml_file_name to the medline_citation table. This field does not correspond to any element in the DTD structure, but allows the database administrator to go back to the original XML file if necessary to find the original source of the data. We could have stored each author only once in a table of its own, and assigned each author a unique integer primary key to serve as an author identifier. An author is represented by a combination of values in fields for last name, forename, first name, middle name, initials, suffix, affiliation, and collective name. Another table would have stored the set of author identifiers associated with each PMID, and because integer joins are fast, this design would have facilitated rapid search for all PMIDs associated with a given author, by joining the author table with the table of author identifiers and citations. However, there are several drawbacks to this approach. Generating integer primary keys during loading requires that either a lookup be done to see if each author of each citation already exists or not (35 million lookups), or all authors and primary keys must be kept in memory. The former approach is very time consuming during loading; the latter approach strains memory resources. In addition, regardless of how primary keys are managed during loading, it is not possible to determine algorithmically if two different representations of one author are actually the same author, or if one representation is actually two different authors. We therefore avoided generating unique primary keys and repeated all eight fields representing the author for every citation occurrence of that author. Figure 3 shows relationships among the tables. The table medline_journal is a parent of thirteen other tables (it contains the primary key pmid , which is used as a foreign key by the other tables). One of the other tables, medline_mesh_heading , is a parent of medline_mesh_heading_qualifier . Multiple qualifiers can be associated with each MeSH heading for a given citation. Parsing and loading software We implemented three versions of software that parses and loads MEDLINE. The first was Java MedlineParser, which was developed at Berkeley [see additional file 1]. The second was the same Java code, modified to run at Stanford. The third was Perl ParseMedline, which was developed at Stanford. All versions of the software perform two basic tasks: (1) they parse the XML files to collect data, and (2) they load the data into the database. Figure 4 shows the steps involved. Data can be loaded as they are collected, or can be written out to disk initially, and loaded later. All three versions offer these two options to the user. Document parsing is processor intensive, data insertion is disk intensive, and if needed, the two tasks can be executed at different times to accommodate other demands on the server. There are two types of application programming interfaces (APIs) for parsing XML files – the tree-based DOM (Document Object Model) and the event-based SAX (Simple API to XML) [ 8 ]. We chose the latter. A DOM parser organizes data from the XML file into a tree of nodes, and requires that the entire document be read in and stored in memory prior to writing out any data. Thus, the DOM parser is impractical for large documents whose data do not fit in memory. The SAX parser, however, receives data through a stream, and recognizes the beginning or end of a document, element, or attribute in an event-driven manner. It writes out data as it proceeds through the parsing process, and there is no need for the entire document to fit into memory. In XML MEDLINE, one document is a single XML-formatted MEDLINE file, and in the 2003 release, the majority of files range in size from 60 to 142 megabytes (MB). Using the DOM parser would put great demand on resources. In addition, the SAX parser is faster because it does not need to create an entire XML tree structure, map that structure to the program's data structures, and then throw out the original tree. Instead, it creates its own data structures as events are handled. The Java version uses the Java SAX parser to parse the XML files, and JDBC (Java Database Connectivity) to communicate with the database. The Perl version uses the Perl SAX parser, and Perl DBI (Database Interface) to communicate with the database. We provide additional detail about the Java implementation here. The SAX parser requires the developer to write code that specifies the data model for objects in the domain. The data model is an object model that represents tables in the schema. The SAX parser also requires code that listens for SAX events and that maps elements – or nodes in the XML tree – to the object model. We created two main classes upon which our code is based: GenericXMLParser and NodeHandler. GenericXMLParser is responsible for generating events when nodes that correspond to objects in the object model are encountered in the document, and NodeHandler provides the event listener. Together, these two classes form a generic approach to reading in XML data and writing out those data to tables; they are independent of the DTD or table structure. As the parser processes the document, it decides how to handle the semantics of data at each node and determines whether to store parsed data at that node or to delegate the event to a child handler. For each node that corresponds to a table, there is a handler class that extends NodeHandler. A handler defines metadata for the node, and encodes any non-standard behavior at that node. An example of metadata is shown in Table 1 . Metadata include column names for the table and an XML element associated with each column name. An XML element is represented by a concatenation of the name of the element that holds the data value, and elements higher up in the element stack up to the node that corresponds to the object, or table. This concatenated name gives a unique representation of the element that holds the data. Finally, the data type is given for each column. The column names and data types match those specified in the database schema. Since NodeHandler and GenericXMLParser are generic, they can be used to write similar parsers for other XML documents. We have, for example, used these classes to write a parser for MeSH (Medical Subject Headings) XML files, which are also distributed by the NLM. The MeSH files are small compared to MEDLINE. MeSH 2003 comes in three XML files that total less than 600 MB. An optional feature is validation. XML files provided by the NLM are validly formatted, but we provide additional checks to ensure that all element tags in the XML data file have been handled by the parser and that all data have been inserted into the database. A developer who is extending the software to cover new tables can use this feature to ensure that metadata definitions are correct in classes that extend NodeHandler. Choice of Relational Database Management Systems In the course of our work, we applied the software tools we were developing to three different relational database products. Our Berkeley team initially experimented with PostgreSQL, since PostgreSQL is an open-source relational database and is freely available and modifiable. For the final implementation, however, we chose IBM's DB2 8.1 over PostgreSQL because we found that it could load our data more efficiently and because DB2 has a text-search extender. Our Stanford team used Oracle 9i, which like DB2, offers word-based indexing of text fields. Word-based indexing is essential to support keyword search of MEDLINE titles and abstracts. Hardware configurations At Berkeley, we used a Pentium IV Intel Xeon 2.00-GHz dual-processor system, with 1 GB of random access memory (RAM). It had an Integrated Drive Electronics (IDE) hard disk with a rotational speed of 7200 revolutions per minute (RPM). At Stanford, we used a Sun Fire V880 server configured with four 750-MHz processors, 8 GB of RAM, and storage-area-network (SAN) storage for the relational database. We also used a Sun Enterprise 3500 server with eight 400-MHz processors and 4 GB of RAM for reading input files and writing intermediate output files in the Perl version. Results and discussion In this section, we describe loading time and disk-space utilization for the three implementations, followed by examples of queries, emphasizing differences between our system and PubMed. The first implementation used the Java software, run on an Intel system (Linux), using IBM's DB2 database management system. The second implementation also used the Java software, run on a Sun server (SunOS), using Oracle's 9i database-management system. The third implementation used the Perl program, run on networked Sun servers, also using Oracle 9i. Table 2 summarizes our results. Loading time and disk space utilization It took 76 hours (3 days and 4 hours) for the Berkeley group to run Java MedlineParser to load MEDLINE, and 196 hours (8 days and 4 hours) for the Stanford group to do so in Oracle. There were numerous differences between the two systems, and it was not possible to test each variable independently. Therefore, we present our data as a range of possibilities, and recognize that other users will have systems that are not the same as either of ours. We believe that differences in processor speed, memory, disk read-write efficiency, and optimization methods employed in commercial database-management systems may have affected loading times. In addition, the code diverged slightly after the initial transfer of code from Berkeley to Stanford, with the main difference being that the Berkeley version used CLOBs for abstracts, whereas the Stanford version used text fields truncated to 4000 characters (size limit imposed on VARCHAR datatype). The Stanford run was also slower because a log file was generated, whereas this feature was turned off in the Berkeley run. The space requirement for the DB2 instance of MEDLINE at Berkeley was 46.3 GB, of which 10.4 GB are consumed by the abstract text CLOBs, 18.1 GB by the other tables, and 17.8 GB by indexes. The space requirement for the Oracle instance of MEDLINE at Stanford was 37.7 GB. The difference in size is primarily due to differences in the number of records that were loaded. The Berkeley group loaded data from XML files that included all of 2002 (early 2003 release) but also included additional files through April 2003. The Stanford group loaded data from 2002 XML files only. Berkeley parsed and loaded 500 input files (44.4 GB); Stanford parsed and loaded 396 input files (40.8 GB). The Stanford group used Perl ParseMedline to load an additional instance of MEDLINE. Parsing and loading of this instance of the database took place in a two-stage process. In the first stage, Perl ParseMedline parsed the XML files and wrote the data to disk in comma-separated-value files. To reduce processing time, the 396 XML input files were divided into 8 sets of about 50 files each, and sets were processed in parallel. The maximum time required for processing one set was 31 hours (1 day and 7 hours). The output comma-separated-value files required 25.6 GB of disk space. In the second stage, the Stanford team loaded data from the comma-separated-value files into the Oracle database using SQL*Loader, a data loading tool provided by Oracle. This stage took 33 hours (1 day and 9 hours) and used 31.6 GB of space in the Oracle database. This version used less space than the other two primarily because it had less indexing and fewer key constraints. Relaxation of constraints is reasonable because the data are well curated by the NLM, and we can count on data in the XML files released by the NLM to be of high quality. The total time required to parse and load the files in this two-stage process is the sum of the time required to parse the largest file if all files are processed in parallel (first stage) and the time required to load the resulting comma-separated-value files into the database (second stage). Alternatively, if the input files are parsed in series, the time for the first stage would be the sum of the input-file processing times. In our case, we overlapped the runs in a way that was convenient for us, given space and user constraints, and therefore mixed the parallel and serial approach. The overall time for our first stage was 99 hours (4 days and 3 hours); adding this time to the second stage gave a total time of 132 hours (5 days and 12 hours). Given the length of time to process each of our eight batches, we can estimate a lower limit of 64 hours (2 days and 16 hours), and an upper limit of 253 hours (10 days and 13 hours), if we had run the files completely in parallel or in series, respectively. Sample queries Certain queries that cannot be done easily through the PubMed application programming interface (API) can be done in a single SQL query to our relational database. In this section, we show the results of a several sample queries, run on a version of MEDLINE that contains citations through April 2003. Timing data is presented in terms of "cold" caches and "warm" cache. The cold cache represents the worst case for timing, assuming the database server has just been restarted and the buffer pool is empty. The warm cache represents the best possible performance: running the same query a second time. A typical timing number should fall somewhere between the two; hence these times represent the range of expected times to run the sample queries. A very simple query is one that retrieves all PMIDs in MEDLINE, where pmid is a column in table medline_citation (Table 3 ). Although typical users of PubMed would not be interested in such a query, we are managing MEDLINE as a complete database, and need to have access to all PMIDs. Running this query on the Berkeley implementation took 12 minutes and 26 seconds. Many articles in MEDLINE are assigned terms from MeSH. Another capability of this system that distinguishes it from PubMed is the ability to rank order journals according to how many articles those journals have published that have been assigned a particular MeSH term. In the query shown below (Table 4 ), the number of publications indexed with the MeSH term (or descriptor_name ) "Leukemia" is shown for each journal (where medline_ta is the title abbreviation of a journal). The result of this query is a table consisting of journals paired with number of publications (Table 5 ); note that the query does not normalize for the fact that some journals have been publishing for more years than others, and publish more articles than others. This query ran in 4 minutes with a cold cache, and in 20 seconds with a warm cache. SQL includes the "LIKE" operator which allows for partial matches. By modifying the query above to change the fifth line to read "WHERE msh.descriptor_name LIKE 'Leukemia%'," we change the query to match all MeSH terms that begin with "Leukemia". The query would thus include terms such as "Leukemia, Subleukemic" and "Leukemia, Feline." This results in dramatically more results, although the rank ordering is not all that different (Table 6 ). This query ran in 4 minutes with a cold cache, and in 46 seconds with a warm cache. MeSH terms are organized into a hierarchy, and each MeSH term has associated with it one or more descriptor tree numbers that indicate its place in the hierarchy. The Berkeley group developed additional code to parse MeSH XML data files (which can be downloaded from the NLM website [ 9 ]), and added MeSH tree data to the MEDLINE database. Using the additional functionality provided by the MeSH hierarchy, we can modify the query above to rank order journals according to how often they have articles that have been assigned the MeSH term under a certain tree number, thus eliminating the sensitivity to different spellings of related concepts that was shown in the queries above. In MeSH, a child tree number shares its leftmost digits with its parent tree number, and differs in its three rightmost digits Therefore, the SQL "LIKE" operator can be used to find a MeSH term and its descendants, as shown in the query below (Table 7 ). The MeSH tree number for "Leukemia" is "C04.557.337". The results of this query are shown in Table 8 . This query ran in 4 minutes with a cold cache, and in 41 seconds with a warm cache. The DB2 version of the system implementation makes use of the text index that is incorporated into the RDBMS system, using the operator "CONTAINS" which is not part of standard SQL. The following query asks how many papers in the last three years of MEDLINE have been published by authors with affiliations at Berkeley or Stanford (Table 9 ). This yields the results in Table 10 . This query ran in 2.5 minutes with a cold cache, and in 7 seconds with a warm cache. When we ran a similar query to determine the number of articles published by Berkeley, Stanford, MIT, Yale, and Harvard, the increase in time to run the query was minimal. This modified query ran in 3 minutes and 35 seconds with cold cache, and in 15 seconds with warm cache. Thus, SQL makes it easy to quantify and rank order results, and does not require a post-processing step as would be necessary with queries to the PubMed API. Similarly, results retrieved from previous queries can be stored directly in the same database, and reused in later queries by simply joining MEDLINE tables with user-created tables. Again, the power of SQL may alleviate the need for a post-processing step. Instead of writing custom code to integrate results from the current query to PubMed with results from previous queries, the user could use SQL joins to integrate current and previous results. Although our system offers capabilities that the PubMed API does not, we point out that PubMed offers functionality that is not available in our system. For example, the "Related Articles" feature found in PubMed is not available, and links to full text are not available. Also, PubMed provides a user interface that is more intuitive than SQL for an end user who is not a database expert, and will be preferred by users who simply want to look up an article. The value of our system is that it offers greater flexibility for innovative software developers who want to experiment with novel techniques for searching biomedical text, and for system developers who want to build systems of which MEDLINE is a component. If such developers want to offer their systems to end users (e.g., biologists, clinicians, or the lay public), they will need to create more intuitive user interfaces. With direct access to the underlying database, developers can create interfaces that are specifically designed to serve the needs of their particular users. Conclusions In this work, we developed highly customizable Java parsing code and a relational database schema that others may be interested in using or modifying. We developed software that uses the Java programming language and the SAX parser to parse XML-formatted MEDLINE files and load the data into a relational database. We loaded one copy into DB2 and another into Oracle, using our Java tool. We also created a similar tool in Perl. The Perl code is less flexible and not as readily extensible as the object-oriented code of our Java software, but the functionality offered by the resulting database implementations is very similar. Differences in loading time among the three installations of MEDLINE were due to a multiple factors, including differences in processors, disk storage devices, memory, operating systems, database-management systems, methods implemented in the software, and choices made by the user. Factors that affected disk-space utilization included the fact that one group loaded more data than the other, abstracts were stored as CLOBs at one site and as truncated text at the other, and indexes differed. Other groups will have system setups that differ from ours, and may make their own modifications to the code that affect their loading times. By presenting data on three examples, we have demonstrated a range of performance results as a guide to what other users might expect at their sites. Future work includes adding functionality to update the system to new versions of MEDLINE, and to accommodate MEDLINE update files. The Stanford group has begun to use MEDLINE to extract drug-gene relationships from the literature, and the Berkeley group used the system, augmented with data from MeSH and LocusLink, to compete in the TREC 2003 genomics track competition [ 10 ]. As we continue to use these systems for research purposes, we are likely to identify alternative approaches that offer enhancements and improvements over the current design. We encourage others who work in similar areas to contribute to the open-source effort. An updated version of our Java code accepts MEDLINE XML input files released in early 2004 that conform to the latest DTD (November 2003). The open-source code for this most current version of MedlineParser is available at . Availability and requirements Project name: Java MedlineParser Project web page: Operating system: Platform independent Programming language: Java Other requirements: Java 1.4.1 or higher, JAXP, relational database, and JDBC driver appropriate for the particular target database License: None Any restrictions on use by non-academics: None Project name: Perl ParseMEDLINE Project web page: Operating system: Platform independent Programming language: Perl Other requirements: Perl 5.8 or higher to handle MEDLINE Unicode data (if writing directly to database), or earlier version of Perl (if writing to comma-separated-value files first), Perl modules DBI and XML::Parser::PerlSAX, relational database, and Perl database driver appropriate for the particular database (e.g., DBD::Oracle) License: None Any restrictions on use by non-academics: None Authors' contributions DO, GB, and AS, developed the MEDLINE database schemas. GB and AS designed and implemented the Java MedlineParser. GB and AS ran MedlineParser to install MEDLINE in DB2 at Berkeley. DO ran MedlineParser to install MEDLINE in Oracle 9i at Stanford. DO developed Perl ParseMedline and ran it to install the second version of MEDLINE at Stanford. DO and GB were primary authors of the article, and the remaining authors added their contributions to the manuscript. MH supervised the work at Berkeley; RB supervised the work at Stanford.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC524480.xml
526297	Prospective study of clinician-entered research data in the Emergency Department using an Internet-based system after the HIPAA Privacy Rule	Background Design and test the reliability of a web-based system for multicenter, real-time collection of data in the emergency department (ED), under waiver of authorization, in compliance with HIPAA. Methods This was a phase I, two-hospital study of patients undergoing evaluation for possible pulmonary embolism. Data were collected by on-duty clinicians on an HTML data collection form (prospective e-form), populated using either a personal digital assistant (PDA) or personal computer (PC). Data forms were uploaded to a central, offsite server using secure socket protocol transfer. Each form was assigned a unique identifier, and all PHI data were encrypted, but were password-accessible by authorized research personnel to complete a follow-up e-form. Results From April 15, 2003-April 15 2004, 1022 prospective e-forms and 605 follow-up e-forms were uploaded. Complexities of PDA use compelled clinicians to use PCs in the ED for data entry for most forms. No data were lost and server log query revealed no unauthorized entry. Prospectively obtained PHI data, encrypted upon server upload, were successfully decrypted using password-protected access to allow follow-up without difficulty in 605 cases. Non-PHI data from prospective and follow-up forms were available to the study investigators via standard file transfer protocol. Conclusions Data can be accurately collected from on-duty clinicians in the ED using real-time, PC-Internet data entry in compliance with the Privacy Rule. Deidentification-reidentification of PHI was successfully accomplished by a password-protected encryption-deencryption mechanism to permit follow-up by approved research personnel.	Background The ability of medical researchers to obtain and store electronic clinical data was complicated by requirements of the Patient Privacy Rule of the Health Insurance Portability and Accountability Act ("HIPAA") of 1996 in title 45 of the Federal Register, parts 160, subparts A and E of part 164 [ 1 - 3 ]. The HIPAA creates a conflict for investigators. The law specifies that 18 data elements, known as protected health identifiers (PHI), that could be used to identify the patient, must be adequately protected from disclosure. However, to allow follow-up, the investigator usually must collect PHI. In the most conservative interpretation of the Privacy Rule, investigators must obtain written informed consent and written authorization to collect PHI. In the hectic setting of the emergency department, the step of obtaining written authorization can bias the data sample [ 4 ]. The Privacy Rule does allow PHI to be collected without written authorization if the institutional Privacy Board grants waiver of authorization. Waiver of authorization requires special handling of PHI. Existing electronic data collection methods are limited in their ability to centralize data in a fashion that expedites data sharing while remaining in compliance with HIPAA. For example, commercial spreadsheets that run on Windows ® do not mandate user identification, do not partition and encrypt sensitive data, and do not maintain a record and audit trail of use. Accordingly, we developed a comprehensive electronic system for clinicians to capture clinical research data from the bedside using commercially available hardware and data upload over the Internet. The system was programmed with multiple security steps and authentication procedures to maintain data security and privacy. We tested the hypothesis that real-time clinical data can be obtained from clinicians in multiple hospitals using electronic data collection stored in an off-site server, under waiver of Authorization, while remaining in compliance with the Privacy Rule. This study represents the development and implementation phase of an ongoing multicenter study to collect prospective and follow-up clinical data from patients undergoing evaluation for pulmonary embolism in the emergency department. The specific aims of this study were to: 1. Test the feasibility of real-time, electronic data collection on personal digital assistants and personal computers in the emergency department setting in two hospitals. 2. Test if the system would correctly upload protected health information (PHI) in a secure and encrypted fashion, but allow follow-up to be performed by selected individuals using password-protected access to PHI. Methods Human subjects and Institutional approval Patients were enrolled from two hospitals in Charlotte, NC: Carolinas Medical Center Main and Carolinas Medical Center University. The clinical protocol was approved under waiver of informed consent and waiver of authorization by the Carolinas HealthCare Institutional Review Board and the Institutional Privacy Board in accordance with the published guidelines of the Department of Health and Human Services (DHHS) [ 5 ], which were reviewed by Annas [ 6 ]. Because of institutional sensitivity about maintaining compliance with the privacy rule, this project required intensive planning and due diligence. Over a 6-month period, the authors scheduled several meetings with the Director of Privacy in Clinical Research, the hospital's Assistant Vice President of Privacy, and the Director of Information Security to discuss the protocol and methods. These individuals had oversight for privacy issues for both hospitals. Then, to facilitate the process of gaining assistance and approval from the Information Systems Department in implementing the technical aspects (software deployment and firewall access) at both hospitals, we obtained a letter of approval from each of these individuals to physically show to technical support personnel. All patients in this study underwent evaluation for pulmonary embolism. The method of selection and diagnosis have been described previously [ 7 ]. For each patient, two electronic data forms ("e-forms") were (or will be) completed. The first was a prospective e-form that was completed in real-time in the emergency department by the clinician in charge of the patient's care. The second e-form encoded follow-up data, and was completed 45 days or more after the prospective form. The follow-up e-form was completed by one of two research associates. This study was non-interventional. System overview This system was designed to allow data to be transferred from multiple sites and stored on one server using technical requirements described in part 160 and 164 of the Privacy Rule. Figure 1 shows a schematic of the overall system structure with the hypothetical participation of four sites. According to published recommendations of the DHHS, the overarching requirement for collection of databases under waiver of authorization is to ensure de-identification of data. The DHHS specifies that this can be done either by the "safe harbor" approach, which entails removal, or the "statistical probability" method, which for practical purposes, incorporates data encryption/de-encryption techniques. The present system uses the statistical probability method, whereby the PHI data are subjected to 128-key bit encryption prior to upload on the server, but are linked to a non-PHI unique identifier (e.g., CMC0001) that allows joining of non-PHI data with PHI data for the purpose of conducting follow-up (see the star in the middle of the schematic in Figure 1 ). This step allows a research coordinator with the appropriate login and password to access de-encrypted (re-identified) data. In the final step, an FTP protocol was used to download the non-PHI follow-up data together with the correct prospective data for each patient. Both the prospective and follow-up data were exported in table form, one row per patient. In summary, authorized research personnel from each site had password-protected access to PHI of patients from their hospital only, while unauthorized personnel could access non-PHI study data via an FTP. An example of the latter would occur in Figure 1 if the site PI from hospital 1 were interested in viewing research data collected at hospital 3. The description of the individual elements of the system that follows is presented in the order that the study was conducted. Data entry form structure The trigger for data entry was the decision to order a diagnostic test to rule out pulmonary embolism in a symptomatic emergency department patient. Patient data were entered on the prospective e-form. The e-forms were programmed using hypertext markup language (HTML) in conjunction with active server pages (ASP) and Standard Query Language (SQL). The prospective and follow-up e-forms are shown in Figures 2 and 3 . The prospective e-form contained a total of 70 fields for data entry including text strings, pull-downs, and click portals. The explicit definition of each field was provided by embedded text that could be viewed by mouse click over an adjacent question mark. These terms are defined in the Table 1 . When the user executed an e-form upload, the server-side ASP code queried data fields for presence of an entry and validity of the entry. For parametric data, such as heart rate, the side-code query interrogated whether numerals had been entered and whether the number fell within a defined range. For example, the heart rate entry had to fall within 21 and 200 beats per minute. (If the investigator encountered a patient with a parameter outside of the allowed range, he or she could click an email link to notify the study administrator, who could override the system to make the entry.) Likewise, if the form contained a missing field, or a nonsensical entry from keystroke error (e.g. a heart rate of "t3"), the server would not load the form, and an error message directed the clinician to the field requiring correction and highlighted the erroneous entry in red shading. Once the field was corrected, the form could be uploaded. To test for data validity in the prospective forms uploaded by clinicians, two authors independently examined each of 70 fields for all patients uploaded. We evaluated for blank cells, nonsensical character entry, or numeric entry that fell outside the predefined ranges. Real-time data entry Forms were completed by attending physicians (N = 22), resident physicians (N = 20), and physician assistants (N = 6) in two emergency departments while the patient of interest was still in the emergency department. Prior to study deployment, each clinician was individually trained in a 10-minute session by the study principal investigator using a pre-defined protocol, and each clinician received a follow-up letter that summarized the training session. Data forms could be accessed in the emergency department using designated Internet-connected personal computers within patient care areas, or could be completed using individually owned personal digital assistants using the Pocket PC ® operating system followed by synchronized upload to the hosted server. All clinicians owned a compatible PDA. The authors and staff assistants provided technical support to assist clinicians in the process of downloading the prospective e-form to their PDAs and uploading completed e-forms to the study's central server using commercial software (Microsoft ActiveSynch ® v. 3.5). The clinicians were shown that prospective data entry forms could also be accessed through a URL hyperlink that was posted on the desktop of all Internet-connected computers in both emergency departments. When the user clicked the hyperlink, this action routed the user through the firewall directly to the hosted server for this study. All computers ran Windows 95 or higher, with ethernet connection to a T3 44.736 Mbps channel. Upon opening the first web page, the user viewed a list of clinician names (Figure 4 ). The clinician then chose his or her name and opened a new, blank prospective e-form. No login was required to access or upload the form, but the central server was programmed to accept uploaded e-forms only from Internet provider addresses of the computers in the two emergency departments. When each new, complete prospective e-form was uploaded to the hosted server, the server encoded the e-form with a unique identification number bearing the initials of the hospital where data originated, the sequence number and clinician who entered the data. (e.g., CMC023JAK). Privacy controls Multiple methods were used to ensure that protected health information would not be subject to unauthorized access, viewing or hijacking. When clinicians entered data in the emergency department, the server polled the form for inactivity exceeding 30 minutes, at which time the page would automatically close without being uploaded. We anticipated scenarios where a clinician would enter data on a prospective form that would need to be revised as a result of updated information (e.g., access to additional medical records, or arrival of family). To allow for such editing, the clinician could re-access any prospective form for a period of 60 minutes after initial upload, provided that the Internet provider address of the computer was the same as the computer from which the form originated. After 60 minutes, the prospective e-form could be altered only by a study administrator. All data were transferred using secure socket link (SSL) protocol. The central server (Win2000 OS) was located off-site at a large commercial web hosting facility (NTT/Verio Inc). Upon upload, all fields containing any of the 18 elements of that constitute PHI were subjected to 128 key-bit encryption. Data were stored in relational tables. To allow data analysis for research purposes, the study PI could access stored by file transfer protocol and exported into a format compatible with commercial software (e.g., Microsoft Access ® , Seattle WA). However, all PHI data fields were remained encrypted. Follow-up data entry Patients were then followed prospectively to determine outcome at 45 days. The follow-up data were entered by an IRB and privacy board approved, designated research coordinator. Because follow-up mandated access to PHI data, a separate web page was developed to allow the study coordinator to have administrative access to the necessary data. The research coordinator would type the appropriate URL address and then view a login page (Figure 5 ). The research coordinator could obtain password-protected access to the list of all uploaded prospective data forms (Figure 6 ), and upon mouse click of the "Follow-up Patient" button, the follow-up form was displayed with the required data to assist in follow-up, including patient name, medical record number, social security number, and telephone number (see top of Figure 2 ). This system thus allowed upload of prospective and follow-up e-forms from multiple hospitals, while research coordinators with IRB and privacy board approval could view the minimum PHI required to perform follow-up at their hospital. Research coordinators could not view PHI from other hospitals. However, using a password-protected file transfer protocol, the central study PI could view the non-PHI clinical data input by all participating hospitals, without access to PHI from any hospital. The information required to populate the follow-up e-form required the research coordinator to perform a standardized review of a comprehensive medical record database maintained by the hospital. The first database was a central electronic record storing system where laboratory and radiology results and any transcriptions of dictated clinician notes and optical scanned images could be found for the entire hospital network. This allowed the evaluation of any return visit to the hospital system, (inpatient, ED, clinic or other outpatient visit) to determine if the patient had any of the outcomes of interest in the follow-up form. If no follow-up data were available within the hospital system to prove the patient was alive at 45 days, then the follow-up protocol required query of the public social security master death index to determine if a death certificate had been filed for the patient. Finally, if no valid follow-up were documented by electronic database search, we then attempted to contact the patient through a previously described, stepwise procedure, consisting of a mailed questionnaire, followed by a telephone call, if necessary [ 7 ]. When all of the data required for the follow-up form were obtained and input into the form, the research coordinator would complete the form and press the "check form for completeness" button. This would activate a system to ensure that all necessary follow-up data were entered. For example, every patient had to have a valid 45-day follow-up, either in the form of a documented follow-up to a clinic, telephone follow-up with the patient, or confirmation of patient death within 45 days. Results The system was fully implemented on April 15, 2003. As of April 15, 2004, prospective data forms were uploaded from 1022 patients evaluated for acute pulmonary embolism. Prospective data have been entered by 42 clinicians and 6 physician assistants from two hospitals in Charlotte, NC. The primary technical barrier to implementation was the process of loading and using the prospective e-forms on a personal digital assistant. All 48 clinicians required individual help and training, of over one hour each to show them how to install the e-form on their PDAs. This impedance was compounded by real-time difficulties associated with stylus use on a small PDA screen, followed by difficulties with uploading to the website from the PDA led to abandonment of this method of data entry. Out of 48 clinicians, only 6 successfully uploaded more than one e-form from the PDA. Only 12 of 1022 uploaded e-forms originated from a PDA. The primary technical barrier to implementation on the personal computers included maintaining the URL icon on desktops in the ED (it was occasionally removed by unknown persons). This problem was solved by the permanent link on the hospital's intranet home page. In two separate instances, clinicians reported that they had populated the e-form, attempted to submit, and for unknown reasons, were unable to upload the e-form, and they had to reenter the data and resubmit the e-form. The server has maintained a log of all successive e-forms uploaded by each clinician. No uploaded prospective forms have been lost or deleted. The side-server system was designed to prevent e-form upload with missing or erroneous data. To examine if this system properly, two observers reviewed the eight parametric field entries (age, heart rate, respiratory rate, systolic blood pressure, pulse oximetry, height, weight and temperature) that were key-entered by clinicians for 1022 patients. Validity required that both observers agree that the entry was a real number within the prespecified range of the parameter. In 44/8176 fields of 12 patients (0.6%), two observers deemed the parametric field entry to be useless for statistical analysis. Stated another way, these data would be coded as missing after data cleaning were completed. However, no categorical data were missing or erroneous. As a result, 1010/1022 (98.8%) of prospective e-forms had usable data in all 70 fields. Ninety-four percent of all 1022 patients have reported the site hospital to be their hospital of choice. Follow-up forms (Figure 3 ) have been completed and uploaded on 605 of 1022 patients. Using existing hospital-approved login and authentication procedures, research personnel were able to access necessary databases from their home computers. Thus, using their personal Internet connection and private telephone line, the research associates were able to complete follow-up forms from home. Follow-up forms were completed in an average of 20 ± 12 minutes. Follow-up has revealed that all prospective e-forms were authentic, and each was completed on an emergency department patient who underwent at least one clinical test for pulmonary embolism. No bogus forms were detected during follow-up to date. This demonstrated a low likelihood of an unauthorized person generating a spoofed form on one of the designated computers in the ED treatment area, given that the automatic control system would not allow a form to be uploaded until all 70 fields are completed. All uploaded prospective and follow-up data were obtained by the central PI using file transfer protocol and were inputted into a spreadsheet without difficulty. Figure 7 shows a screenshot illustrating a partial view of the downloaded study data, including the appearance of the PHI fields after encryption as well as unencrypted data. The purpose of this figure is to demonstrate that the central PI could have access to necessary study data from all sites while remaining blinded to PHI data. The "study ID" field represents the unique identifier used to re-identify PHI data. Query of the server log revealed no evidence of website hijacking or other intrusion. The server computer which houses the study database and runs the web application uses the Windows ® Server 2003 operating system. The only means of electronic access to the server is via hypertext transfer protocol (HTTP) and file transfer protocol (FTP). Both of these system services log all requests made to their ports. An example log entry is shown in the appendix. Discussion The step of obtaining written Authorization to comply with HIPAA can impart a selection bias in registries intended to study acute disease processes [ 4 ]. In section 164.512(i), the Privacy Rule allows for waiver of Authorization when the "research could not practicably be conducted without the waiver" and the "use of the PHI involves no more than a minimal risk to the privacy of individuals." The present report tests a system designed to collect clinical data in real-time from patients with acute diseases at multiple hospitals, including a mechanism to facilitate follow-up, while protecting the privacy of the participants. The first research objective was to determine if PDAs would represent an efficient and secure mechanism for clinicians to record real-time data at the bedside in the emergency department setting. Our experience in this phase I, two-hospital study demonstrated that PDAs presented unexpected complexities that eroded our enthusiasm. The clients were clinicians with variable levels of technical sophistication. Despite our use of a relatively standard process, clinicians found it difficult to download the e-form from the website onto their PDAs, and many needed help from the study authors. Clinicians frequently forgot to bring their PDA devices to work, and during the one-year course of this study, 10 of 48 clinicians bought new PDA devices. Clinicians consistently reported difficulty with the small screen size and data entry with a stylus. Unfortunately, we did not quantify this opinion using a structured survey. We believe this represents the first published experience at using PDAs to collect research data in the emergency department setting. Our results are somewhat less positive than other studies that have reported the use of hand-held computers to maintain clinical databases [ 8 , 9 ]. However, Lu and colleagues previously recognized similar barriers to physician use of PDAs [ 10 ]. We emphasize that our protocol was preplanned, adequately budgeted, and technically supported to disseminate the e-form via the PDA. Unfortunately, we did not perform preplanned measurements to explain this failure. We cannot conclude inferiority of the PDA versus other methods (e.g., paper forms or PC platform) for data collection accuracy inasmuch as we did not compare key quality index data (e.g., comparison rate of compliance, missed data, key errors, lost forms) between methods. Thus, we can only explain the failure of the PDA mechanism in the broad terminology of "it lacked feasibility." The second objective was a relatively complex task intended to determine if the system would allow prospective and follow-up data collection over the Internet in compliance with the requirements of the Privacy Rule. From a functional standpoint, we sought to determine if the system would allow us to protect the data fields that needed to be protected, but allow the non-sensitive data to be accessed by study personnel who did not have local IRB approval. This was accomplished while maintaining strict security standards at each step of data transfer (see Figure 1 ). Data were uploaded from designated Internet provider addresses via secure socket link protocol and stored in a database on an offsite hosted server that was protected by several layers. No study data could be accessed without a password. Further, the system mandated specific password-protected access to PHI only by IRB- and privacy board-approved individuals at each hospital. This mechanism was designed facilitate the acquisition of patient follow-up data at participating hospitals. However, the central study PI could download the study data of interest via a separate password-protected file transfer protocol, but the PHI data were encrypted (see Figure 7 ). Because the PHI data were stored on the server after 128 keybit encryption, even in the event of unauthorized data access (hacking), the hijacker would be unable to view the PHI. Although a large number of commercial systems are available for storing clinical data, most are designed and marketed explicitly for the billing process. In contrast, from the perspective of research, relatively little has been published on the design and implementation of a web-based system to allow collection of clinical data in a multicenter trial design [ 11 , 12 ]. We believe this is the first report of successful web-based clinical data collection under waiver of Authorization and in compliance with CFR 45, parts 160 and 164. This phase I project was limited to two hospitals in the same city, both covered under the same IRB. However, we submit that the system is ready to be expanded to other hospitals in the second phase of the study. This system was designed to be a reasonably comprehensive tool to obtain key information about the beliefs of a clinician at the time of test ordering. Here, we refer to the clinician's beliefs as what they thought were the values of certain specific clinical data that are commonly used to estimate the pretest probability of pulmonary embolism. To capture these beliefs in real time, the system cannot default to a retrospective review of the patient's chart, or having the clinician complete the form after a shift. Within the emergency department setting, the flow of knowledge is dynamic for each patient. As a consequence of time urgency, emergency clinicians often must decide to order expensive imaging tests based upon limited, changing, and sometimes erroneous information. Occasionally, clinical information becomes updated after an expensive radiological test has been done (e.g., a family member arrives with new information, or medical records arrive from another facility by facsimile). Accordingly, the data collection instrument must accurately capture the information that the clinician uses to motivate his or her test ordering behavior, rather than to collect data after the test results have returned, and more complete medical records may have arrived. This report represents a phase I study. For the second phase, we will deploy this system to 10 US hospitals to allow collection of data from 5000 patients. The ultimate goal of this project is to collect a large, multicenter database, as the substrate for a mathematical model to generate a pretest probability of pulmonary embolism based upon beliefs of many clinicians. Conclusions Research data can be successfully collected, entered and uploaded to a hosted server by emergency physicians working in different emergency departments, and in compliance with the Privacy Rule. Use of server side controls to test for data validity ensured that 98.8% of uploaded forms contained complete data usable for statistical analysis. The PHI data were successfully encrypted and deencrypted using password access to allow follow-up at a later date. Server log query demonstrated no evidence of intrusion or data loss, suggesting that data were securely stored. Abbreviations ASP – Active server pages ED – Emergency department FTP – File transfer protocol DHHS – Department of Health and Human Services HIPAA – Health Insurance Portability and Accountablity Act HTML – Hypertext markup language HTTP – Hypertext transfer protocol IRB – Institutional Review Board PDA – Personal digital assistant PHI – Protected health identifiers PI – Principal investigator SQL – Standard query language RUL – Uniform resource locator Competing interests Jeffrey A. Kline is primary author on a US copyright certificate "Electronic Form ("e-form") For Secure Collection of Clinical Data Via the Internet" and inventor on a US patent (pending) that describes the present work. Jeffrey A. Kline and Charles L. Johnson own stock in BreathQuant Medical Systems. Appendix Example of a log entry for the HTTP system service 208.10.156.69 – [02/Mar/2004:23:55:55 +0000] "POST /pestudy/peadmin.asp HTTP/1.1" 302 1349 " " "Mozilla/4.0 (compatible; MSIE 6.0; Windows NT 5.1)" The IP address of the HTTP requesting browser is 208.10.156.69 and the resource requested is the ASP file . For the FTP system service this would be a typical log entry 208.10.156.69 – generic [06/Aug/2003:17:00:14 +0000] " [22786]USER generic FTP" 331 0 "-" "-" The IP address of the HTTP requesting browser is 208.10.156.69 and the USER identity that was logged in was " generic" . Both these log files are scanned once a week to look for any suspicious requests. To date there have been no identified intrusion or hijack attempts on the specific study database. As an HTTP responding server on the public Internet, the web (HTTP) server program does receive many well identified "virus spreading" requests which it denies and whose denials are logged. One such "virus related" request common to all web logs today is 61.100.6.181 – [01/May/2004:11:30:52 +0000] "GET /scripts/nsiislog.dll HTTP/1.1" 404 3806 This example virus was developed to attack a vulnerability which existed in Microsoft Web Servers but was eliminated by a security update for their web (HTTP) server. The highlighted 404 code denotes that the request was denied. As security updates from Microsoft become available the study web server is updated. At present there are no known HTTP vulnerabilities which would allow an unauthorized user to gain access to files on the server. Authors' contributions JAK conceived the study, obtained funding and participated in the system design, data collection and drafted the manuscript. CLJ carried out all technical programming, participated in system design and drafting the manuscript. WBW participated in idea conception, system design, data collection and drafting the manuscript. MSR participated in data collection and drafting the manuscript. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC526297.xml
517923	Chronic arthritis in children and adolescents in two Indian health service user populations	Background High prevalence rates for rheumatoid arthritis, spondyloarthopathies, and systemic lupus erythematosus have been described in American Indian and Alaskan Native adults. The impact of these diseases on American Indian children has not been investigated. Methods We used International Classification of Diseases-9 (ICD-9) codes to search two Indian Health Service (IHS) patient registration databases over the years 1998–2000, searching for individuals 19 years of age or younger with specific ICD-9-specified diagnoses. Crude estimates for disease prevalence were made based on the number of individuals identified with these diagnoses within the database. Results Rheumatoid arthritis (RA) / juvenile rheumatoid arthritis (JRA) was the most frequent diagnosis given. The prevalence rate for JRA in the Oklahoma City Area was estimated as 53 per 100,000 individuals at risk, while in the Billings Area, the estimated prevalence was nearly twice that, at 115 per 100,000. These rates are considerably higher than those reported in the most recent European studies. Conclusion Chronic arthritis in childhood represents an important, though unrecognized, chronic health challenge within the American Indian population living in the United States.	Background As a group, the rheumatic diseases of childhood represent one of the most common chronic disease conditions in children [ 1 ]. These illnesses have global distribution [ 2 - 5 ], but little information exists regarding either prevalence or phenotypic expression of these diseases in children in any population other than North American and European whites [ 6 ]. Aggarwal and colleagues [ 7 ] have reported their experience with juvenile rheumatoid arthritis (JRA) on the Indian subcontinent, and their findings suggest that the patterns of disease reported in Europe and North America are not seen in that population. Most conspicuous in Aggarwal's study was the relative rarity of the pauciarticular form of JRA, in contrast to European and North American populations, where that subtype accounts for 50 to 75% of the cases [ 3 , 8 - 11 ]. The relative rarity of pauciarticular JRA in non-European populations has been documented in children of Kuwait [ 5 ], Turkey [ 12 ], Thailand [ 13 ], Japan [ 14 ] and South Africa [ 15 ] as well as in African American children in Detroit [ 16 ]. Further evidence that examining prevalence rates of rheumatic diseases in specific populations may be informative comes from studies of systemic lupus erythematosus (SLE). Studies from Great Britain, for example, indicate that the prevalence rates for SLE in people of Afro-Caribbean descent may be 4–8 times higher than that in Caucasians [ 17 , 18 ]. Prevalence rates of rheumatic disease in North American Indian/First Nations populations have been reported in small studies from single tribes. From these studies, significantly higher prevalence rates for rheumatoid arthritis (RA) have been reported in adults from tribes living in the Great Lakes region [ 19 ], the Pacific Northwest [ 20 ], the Southwest [ 21 ] and Canada [ 22 - 24 ]. To our knowledge, a comprehensive survey of rheumatic diseases affecting children, adolescents, and young adults has not been reported in any non-Caucasian population. Because both our clinical experience here in Oklahoma suggests that rheumatic diseases in children may also be more prevalent in the American Indian population compared with Caucasians, we undertook a search of the Oklahoma City Indian Health Service (IHS) user population databases in order to develop prevalence estimates of rheumatic diseases in American Indian children and adolescents. We performed the same queries from the database in the Billings Area IHS office as a basis of comparison. Methods Populations in the billings and Oklahoma City areas The Oklahoma City Area IHS serves a population of 291,288 individuals, most of whom reside in Oklahoma, with small numbers living in the neighboring states of Kansas and Texas [ 25 ]. IHS services are limited to members of federally recognized tribes, 39 of whom have tribal headquarters in Oklahoma. The 39 federally recognized tribes [ 26 ] represent people from multiple Native cultures including Eastern Woodlands tribes (e.g., Cherokee, Delaware, Seneca), Southeastern tribes (e.g., Creek, Choctaw, Chickasaw, Seminole), Southwestern tribes (e.g., Apache), as well as tribes who have long been resident on the southern Great Plains (e.g., Kiowa, Comanche, Southern Cheyenne). Tribal membership is determined by the tribes themselves and may or may not include specific blood quantum requirements for membership. Historical factors, the absence of reservations, and the proximity of European-descended people in Oklahoma has resulted in significant admixture between Native and Caucasian populations in many parts of the Oklahoma City Service Area. The Billings Area IHS serves a population of 72,591 individuals. The tribes in this service area, a significant proportion of whom live on 8 reservations located in Montana and Wyoming, consist largely of northern plains tribes (e.g., Crow, Sioux, Blackfeet). In both Areas, the population is younger than the population of the United States as a whole, with 40 percent of the population 19 years of age or younger [ 27 ]. Database search International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9) codes were used to search the Oklahoma City and Billings Area IHS National Patient Information Reporting Systems and Patient Registration user databases [ 28 ] over a three-year period (1998–2000) to identify individuals with rheumatic diseases. Outpatient data over this three-year period was gathered for the user population ≤ 19 years of age and included: patient chart number (with the chart number scrambled to protect patient identity), date of birth, sex, date of visit and diagnosis (the IHS database allows at least 9 diagnoses to be recorded on any given patient). The codes used and the diagnoses denoted by those codes are listed in Table 1 . Data were downloaded into a standard database (Microsoft Excel), which was then searched to eliminate duplicate records and to sort patients for each year on the basis of age, sex, and diagnosis. A case was defined as any person who was 19 years of age or younger on January 1,1998, January 1, 1999 and January 1, 2000 and whose diagnoses included at least one of the entities listed in Table 1 . The population at risk was defined as the number of individuals ≤19 years of age within the IHS user population (i.e., eligible individuals who have used the IHS facilities at least one time in three years) [ 29 ]. The user population (i.e., people who actually used IHS services) may differ from the IHS service population, which includes all individuals who are eligible to receive IHS services. Estimation of disease prevalence was based on three assumptions: (1) that the diagnoses recorded were, in fact, accurate; (2) that the population at risk did not change significantly over the three-year period; (3) that the three-year mortality rate for the diseases of interest was no greater in the IHS user population than for all races in the United States. For the Oklahoma City Area, accuracy of the IHS database information was assessed by matching the IHS identification number of known JRA cases followed at the Children's Hospital of Oklahoma with the same number in the IHS patient databank to be sure that known patients were identified and coded accurately. Results Rheumatoid arthritis/juvenile rheumatoid arthritis Rheumatoid arthritis and juvenile rheumatoid arthritis (RA/JRA; #ICD-9 # 714.0, 714.30) were the most frequent rheumatic disease diagnoses recorded in individuals ≤ 19 years of age in both IHS areas. In the Oklahoma City Area, we identified 62 individuals (45 females and 17 males) with these diagnoses. Assuming a population at risk of 117,409 (i.e., individuals 19 years of age or younger; source: IHS Headquarters office, data processing services unit, Albuquerque, NM), this gives a crude prevalence rate of 53 cases per 100,000 at risk. The prevalence rate for Billings was considerably higher. The 714.0 and 714.3 codes identified 33 individuals in the Billings Area. Based on a population at risk of 28,724, the prevalence estimate for Billings was 115 per 100,000 at risk (see Table 2 ). The age distribution of affected children, adolescents, and young adults in both areas differed from what has been reported in studies from predominantly European or European-descended populations (Figures 1A and 1B ). There is a distinct biphasic distribution of JRA prevalence by age in Caucasians, with peaks in the late preschool years and in early adolescence [ 30 , 31 ]. Data from both IHS databases show a distinct peak at age 5–12 years with proportionately smaller numbers of patients in the preschool and early adolescent age groups. In both Areas, peaks in late adolescence and early adulthood are observed, consistent with our observation that rheumatoid arthritis is a disease of young adults in this population (Mauldin et al , manuscript in preparation). The absence of a prevalence peak in the preschool years may reflect the almost complete absence of children with monoarticular or pauciarticular JRA in the IHS user population, consistent with previous studies in non-European populations [ 5 , 7 , 8 , 13 , 15 , 32 , 33 ]. We found no individuals in the database with the ICD-9 code commonly used to denote pauciarticular-onset JRA (#ICD-9 # 714.32). In the Oklahoma City Area we found a single child (a 12 year old female) diagnosed with monoarthritis (#ICD-9 # 714.33). Included in the above analyses are individuals who may not fit established criteria for a diagnosis of JRA. Since we did not examine age at onset, it is impossible to know whether a 19 year old identified as having rheumatoid disease was age 7 or age 17 at disease onset. JRA diagnosis criteria stipulate that patients must have disease onset at 15 years of age or younger [ 34 ]. When data were analyzed to include only children 15 years of age or younger, we identified 35 patients (23 females, 12 males) in the Oklahoma City Area and 21 patients in the Billings Area (9 females, 12 males) with a diagnosis of JRA. Based on the populations at risk of 87,936 (Oklahoma City) and 21,777 (Billings) this yields an estimated prevalence rate of 40 per 100,000 in the Oklahoma City Area and 96 per 100,000 in the Billings Area. Both of these estimates are more than twice the prevalence derived from a recent European study (14.8 per 100,000) [ 35 ]. In order to test the integrity of the IHS database in the Oklahoma City Area, we matched known cases of JRA followed at the Children' Hospital of Oklahoma (CHO; n = 15) by IHS identification number with patients in the database. All 15 of the children followed at CHO were identified within the database and correctly identified by subtype. We did not have access to patient records in the Billings Area, but do have access to databases at other IHS facilities. At a large facility in the Aberdeen Area (comprising the states of North Dakota, South Dakota, Nebraska, and Iowa and serving a patient population very similar to that served in the Billings Area) we identified 20 children with JRA using the same search strategy as that used in for the Billings and Oklahoma City Areas. Subsequent chart review demonstrated that 17 of these children had strong clinical evidence to confirm the diagnosis of JRA, while diagnosis could not be supported or excluded in the other three. The estimated prevalence for JRA in the population served by this facility calculates to 236/100,000, within the same order of magnitude by considerably higher than the Billings Area estimate. Spondyloarthropathy The second most common diagnoses identified in each Area database were the group of illnesses collectively denoted spondyloarthropathies (#ICD-9 # 720.0, 720.1, 720.2, 720.8, 720.89). Although these illnesses are sometimes viewed as distinct entities, they share sufficient common features that allow them to be grouped for purposes of this analysis. These common features include: (1) male sex preponderance; (2) arthritic involvement of the axial skeleton (e.g., sacroiliac joints); (3) extra-articular musculoskeletal involvement (e.g., bursitis, enthesitis); and (4) extra-articular (e.g., ocular, genito-urinary) inflammation. In both white [ 36 - 38 ] and American Indian patients [ 39 - 43 ], the human class I histocompatibility complex antigen HLA-B27 constitutes a strong risk factor [ 44 - 47 ]. We identified 20 patients (12 females, 8 males) with spondyloarthropathy in the Oklahoma City Area IHS database, giving an overall crude prevalence rate of 17 per 100,000 at risk. Included in this group are three patients (all female) with psoriatic arthritis (ICD-9 Code # 696.0). In the Billings Area, we identified 12 individuals (7 females and 5 males) with ICD-9 codes used to identify patients with spondyloarthopathy, yielding a prevalence rate of 42 per 100,000 at risk. These included one patient (a 12 year old male) with ankylosing spondylitis (#ICD-9 # 720.0), one (a 5 year old male) with psoriatic arthritis (#ICD-9 #696.0), and 10 patients (6 females and 4 males) with nonspecific sacroiliitis (#ICD-9 # 720.2). The Oklahoma and Billings Area estimates were within the range previously reported for childhood spondyloarthropathies (e.g. ankylosing spondylitis) in the United States and United Kingdom (12 to 33 per 100,000) and Mexico (13 to 65 per 100,000) [ 48 ], but slightly lower than previously estimated prevalence rates of 29 per 100,000 (all spondyloarthropathies) for First Nations children in western Canada [ 22 ]. The female-to-male preponderance in both Areas was unusual and has not been, to our knowledge, reported with any previous population. Discussion Although rare individually, the rheumatic diseases, taken together, are among the most common chronic health conditions affecting children [ 1 ]. Exact prevalence rates among children living in the United States are difficult to obtain, owing, in large part, to the de-centralized delivery of health care in this country. The IHS represents an exception to that decentralization and is, arguably, the closest representation to a nationalized health care delivery system currently functioning within the United States. Thus, records and data available through the IHS represent a unique opportunity to assess population-wide health needs not otherwise available to child health researchers in this country. This report provides a first-ever population-wide estimate of the prevalence of chronic arthritis in American Indian children living within the United States. While the prevalence rate for the Oklahoma City Area was within the same order of magnitude as the most recent reports from Europe [ 35 ], the prevalence rate in the Billings Area was nearly 10 times this recent European estimate. These findings are consistent with earlier studies of rheumatoid disease in American Indian adults, where prevalence rates 10 times higher than the general population were reported [ 49 , 50 ]. It should be pointed out, however, that prevalence estimates of JRA vary widely, ranging between 16 to 113 per 100,000 [ 30 , 31 , 51 - 55 ]. The reasons for the discrepancy in prevalence estimates between the Oklahoma City and Billings Areas are not clear. One possibility is that the northern plains tribes are particularly susceptible to rheumatoid disease in ways that other groups (e.g., Eastern Woodlands or Southwestern tribes) are not. It is also possible that the difference in ethnic composition of the two populations accounts for this difference. While many Oklahoma tribes require at least a 25% blood quantum of tribal ancestry (e.g., the Kiowa tribe [ 56 ]), other tribes require only proof of descent from an individual on the original Dawes rolls of 1893 [ 57 ]. Thus, the Oklahoma City Area includes many individuals whose degree of American Indian ancestry is 1/4 or less and may include individuals with less than 1/64 American Indian ancestry. In contrast, there has been less intermingling between Caucasian and American Indian populations on the northern plains, and a larger percentage of the Billings Area population includes individuals with full-blooded American Indian ancestry. Our study once again points out the rarity of the pauciarticular form of JRA in non-European populations. In studies of European and European-descended populations, pauciarticular JRA is the most common form of chronic childhood arthritis [ 3 , 8 ]. The Oklahoma City Area database listed a single child with ICD-9 codes #714.32 (pauciarticular JRA) or 714.33 (monoarticular arthritis), the codes used to identify such children. These findings are consistent with reports from the Indian subcontinent,[ 7 ] Kuwait [ 5 ], Turkey [ 12 ], Thailand [ 13 ], Japan [ 14 ], South Africa [ 15 ], and with our experience with African American children in Detroit [ 16 ]. The slight female-to-male preponderance for spondyloarthopathy is also worth noting. High prevalence rates for spondyloarthopathies have been noted in both Northwestern and Southwestern tribes [ 24 , 41 - 44 ]. However, in these studies, a strong male preponderance was noted. Whether the findings from Oklahoma City and Billings represent a novel finding or inaccuracies in the ICD-9 coding await confirmatory studies, as we discuss below. An important limitation to this study is the fact that we did not have the means to verify every individual case listed in Oklahoma City database and were unable to confirm any diagnosis in the Billings Area database. However, our limited test of the accuracy of the Oklahoma City data provided surprising confirmation of the accuracy of coding for known cases. While we could not confirm any of the Billings cases, our search of the database of a single IHS facility in the neighboring Aberdeen Area corroborated the prevalence statistics we derived from the Oklahoma City and Aberdeen databases. Indeed, our experience suggests that a search strategy like the one we used is likely to under-estimate rather than over-estimate the prevalence of rheumatic disease in the IHS user population. We are aware that there are many factors that might overestimate disease prevalence using this type of database search. The first is the possibility that a given ICD-9 code might have been used to designate a "working" diagnosis that was never established by the patient's clinical course. The second opportunity for overestimation of prevalence would occur if children were systematically misdiagnosed. This could occur easily if physicians use serologic data as the sole criterion for diagnosis. For example, many physicians routinely screen children with musculoskeletal complaints using antinuclear antibody (ANA) tests. However, the prevalence of low-titer positive ANA tests is extraordinarily high in the pediatric population [ 58 ]. Thus, if ANA-positive children with musculoskeletal pain [ 59 ] are listed as having "JRA," then there would be a gross overestimation of the actual prevalence. Similarly, there are factors that might have led to underestimation of JRA prevalence by relying solely on a three-year database search. Children or adolescents with well-controlled JRA may not have seen an IHS physician during the relevant time period, and thus would have been excluded. Similarly, physicians who rely on rheumatoid factor tests as a diagnostic criterion for JRA might fail to diagnose the disease in a child, since only a small number of children with JRA have detectable IgM rheumatoid factor [ 60 ]. The ideal method for obtaining true disease prevalence rates would include rigorous, pro-active case finding in a known population at risk. This approach was taken by Manners and Diepeveen in a study of school children in Australia [61]. Using such an approach, these authors reported a prevalence rate of 4 per 1,000 for JRA, significantly higher than any previous estimates. We are now preparing a similar project involving American Indian communities on the northern plains. Conclusion We conclude that the rheumatic diseases of childhood may represent a significant burden of morbidity in these two IHS user populations. More detailed studies with rigorous case ascertainment are required to follow up these preliminary data. Competing Interests None declared. Authors' Contributions Dan Cameron provided data from the IHS database in Oklahoma City, and Diane Jeannotte provided the Billings Area data. Joyce Mauldin and Glenn Solomon performed the database searches. Dr. Jarvis directed this study and assisted in data analysis and interpretation. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517923.xml
535566	Molecular and epidemiologic analysis of a county-wide outbreak caused by Salmonella enterica subsp. enterica serovar Enteritidis traced to a bakery	Background An increase in the number of attendees due to acute gastroenteritis and fever was noted at one hospital emergency room in Taiwan over a seven-day period from July to August, 2001. Molecular and epidemiological surveys were performed to trace the possible source of infection. Methods An epidemiological investigation was undertaken to determine the cause of the outbreak. Stool and blood samples were collected according to standard protocols per Center for Disease Control, Taiwan. Typing of the Salmonella isolates from stool, blood, and food samples was performed with serotyping, antibiotypes, and pulsed field gel electrophoresis (PFGE) following XbaI restriction enzyme digestion. Results Comparison of the number of patients with and without acute gastroenteritis (506 and 4467, respectively) during the six weeks before the outbreak week revealed a significant increase in the number of patients during the outbreak week (162 and 942, respectively) (relative risk (RR): 1.44, 95% confidence interval (CI): 1.22–1.70, P value < 0.001). During the week of the outbreak, 34 of 162 patients with gastroenteritis were positive for Salmonella, and 28 of these 34 cases reported eating the same kind of bread. In total, 28 of 34 patients who ate this bread were positive for salmonella compared to only 6 of 128 people who did not eat this bread (RR: 17.6, 95%CI 7.9–39.0, P < 0.001). These breads were produced by the same bakery and were distributed to six different traditional Chinese markets., Salmonella enterica subsp. enterica serovar Enteritidis ( S . Enteritidis) was isolated from the stool samples of 28 of 32 individuals and from a recalled bread sample. All S . Enteritidis isolates were of the same antibiogram. PFGE typing revealed that all except two of the clinical isolates and the bread isolates were of the same DNA macrorestriction pattern. Conclusions The egg-covered bread contaminated with S . Enteritidis was confirmed as the vehicle of infection. Alertness in the emergency room, surveillance by the microbiology laboratory, prompt and thorough investigation to trace the source of outbreaks, and institution of appropriate control measures provide effective control of community outbreaks.	Background Salmonellosis, resulting from the ingestion of contaminated poultry, beef, pork, eggs, and milk [ 1 ], is an important public health problem worldwide [ 2 ]. Although the high temperature of the baking process would suggest that baked goods provide a relatively inhospitable environment for colonization with infectious pathogens, there have been numerous reports of food poisoning outbreaks associated with consumption of baked goods [ 3 - 8 ], and such outbreaks can be a major public health concern [ 4 ]. Food-borne disease outbreaks due to Salmonella species are relatively uncommon in Taiwan [ 9 ] compared to those in the United States [ 21 ] and Japan [ 22 ]. Thirty-one outbreaks were reported to the Department of Health, Taiwan from 1986 to 1995, which accounted for 5.6% of all outbreaks [ 9 ]. Serovar Typhimurium was the leading serovar for Salmonella food-borne disease outbreaks but serovar Enteritidis has emerged as a new serovar in Taiwan [ 23 ], consistent with similar findings of a worldwide increase in Salmonella enterica subsp. enterica serovar Enteritidis ( S . Enteritidis) infections [ 12 , 24 , 25 ]. We report the findings of the investigation of an outbreak of S . Enteritidis that was performed after notification of the rapid increase of the number of patients with febrile gastroenteritis in an emergency room and an unusually high percentage of Group D Salmonella isolated since it comprised only 7.4% of the various Salmonella groups in the hospital one year before the outbreak. Methods Epidemiological investigation A rapid increase in the number of attendees due to acute gastroenteritis and fever was noted to have begun on July 28, 2001 at the emergency room (ER) of Kaohsiung Municipal Hsiao-Kang Hospital, and this increase continued for the following six days. Clinical and demographic features and the food reported to have been consumed three days before development of gastrointestinal symptoms for all patients with acute gastroenteritis at the ER from July 28 to August 3, 2001 were reviewed by ER personnel and the infection control team. The infection control team conducted patient interviews and reviewed charts using a standardized case record form. Stool and blood cultures were performed once specimens were available. These specimens were further analyzed by the hospital's clinical microbiology laboratory and the laboratory of the Center for Disease Control, Taiwan (CDC, Taiwan) in Kaohsiung, Taiwan. Charts of all patients with and without acute gastroenteritis who visited the ER from six weeks before until two weeks after July 28, 2001 were reviewed by infection control nurses for the baseline data of the ER. An outbreak-associated case was defined as a patient visiting the ER with the diagnosis of acute gastroenteritis and having Salmonella infection. A cohort study of all patients attending the ER with acute gastroenteritis between July 28 and August 3, 2001 was performed in order to test the hypothesis that illness was caused by a specific food. Laboratory investigation Surveillance culture Methods for sample collection, cultivation and isolation were conducted according to the standard protocol of the CDC, Taiwan for food-borne disease outbreaks as described previously [ 9 ]. Serogrouping and serotyping of Salmonella Salmonella serotypes were determined with the use of antiserum (Difco, Detroit, MI, USA) according to the manufacturer's instructions. Serogrouping and serotyping were performed by the slide agglutination method and tube agglutination method to identify the somatic O antigen and flagellar H antigen, respectively [ 10 ]. Testing for antimicrobial susceptibility Antimicrobial susceptibility was determined by the disk diffusion method according to the National Committee for Clinical Laboratory Standards [ 11 ]. The antimicrobial agents tested included ampicillin, amoxicillin/clavulanate, gentamycin, cefazolin, cefmetazone, cefperazone, imipenem, ofloxacin, and trimethoprim/sulfamethoxazole. Escherichia coli ATCC 25922 was used as the quality control organism. Genomic fingerprinting by pulsed field gel electrophoresis (PFGE) Total DNA was prepared and PFGE was performed as described previously [ 12 , 13 ]. The restriction enzyme Xba I (New England Biolabs, Beverly, MA) was used at the manufacturer's suggested temperature. Restriction fragments were separated by PFGE in 1% agarose gel (Bio-Rad, Hercules, CA.) in 0.5X TBE buffer (45 mM Tris, 45 mM boric acid, 1.0 mM EDTA, pH 8.0) for 25 h at 200 V at a temperature of 14°C, with ramped times of 2 to 40 s using the Bio-Rad CHEF-DRII apparatus (Bio-Rad Laboratories, Richmond, CA). Gels were then stained with ethidium bromide and photographed under ultraviolet light. The resulting genomic-DNA profiles, or "fingerprints," were interpreted according to established guidelines [ 14 ]. All experiments above were performed in duplicate. Environmental investigation Food items having a significant relationship with gastroenteritis cases with Salmonella infection were suspected as the vector of infection. A sample of the implicated food was collected from a patient's home within six hours of the patient having symptoms of febrile gastroenteritis and it was stored at 4°C. Further tracing the source of the suspected food and the subsequent environmental investigation were undertaken by the Bureau of Health, Kaohsiung County and CDC, Taiwan. Data collection and statistics Differences between groups were compared using Chi-Square analysis. Relative risk and 95% confidence intervals were also calculated when Chi-Square analysis was used. Continuous variables were analyzed with Student's t -test. Significance was considered at a P value < 0.05 (Epi Info. Version 3.2.2 Centers for Disease Control and Prevention (CDC), USA). Results Patients An increase in the number of patients admitted due to acute gastroenteritis and fever was noticed beginning from the early morning of July 28, 2001 with a return to normal levels one week later. Reviewing daily case-visits to the ER due to all causes and due to gastroenteritis six weeks before this outbreak period revealed significantly more cases due to all causes visiting the ER on Saturday and Sunday than on weekdays (From Monday to Friday) (mean ± standard deviation (S.D.): 140.3 ± 38.3 compared with 109.6 ± 16.7, P = 0.004). The number of patients visiting due to gastroenteritis on Saturday and Sunday were also higher than those visiting on weekdays (mean ± S.D.: 15.8 ± 7.3 compared with 10.5 ± 3.6, P = 0.032) (Fig. 1 ). A total of 162 (17.2 %) of 942 patients visiting the ER during the outbreak period had acute gastroenteritis. Compared with 85 (9.7%) of 872 patients in the week before the outbreak, significantly higher percentages of patients with gastroenteritis were observed in the outbreak week (Relative risk (RR): 1.92, 95% confidence interval (CI): 1.38 – 2.26, P < 0.001). A comparison of the number of patients with and without acute gastroenteritis (506 and 4467, respectively) during the six weeks before the outbreak week and during it also revealed a significant increase in the number of patients of gastroenteritis during the outbreak week (RR: 1.44, 95% CI: 1.22–1.70, P value < 0.001). Investigation of food consumed by gastroenteritis patients during the outbreak period revealed 34 (21%) had consumed the same kind of egg covered bread decorated with mayonnaise and fried seasoned pork fiber from six traditional Chinese markets (three located in Kaohsiung City and three in Kaohsiung County). There were 2, 2, 2, 3, 3, and 22 gastroenteritis patients that consumed the bread purchased from 6 markets, respectively. In total, 28 of 34 patients who ate this bread were positive for Salmonella compared to only 6 of 128 people who did not eat this bread (RR: 17.6, 95%CI 7.9–39.0, P < 0.001). The association of consuming the kind of bread and having Salmonella infection was significant (RR: 17.6, 95% CI: 7.9–39.0, P value < 0.001). No other identified food or restaurant exposure was significantly associated with the outbreak. Regarding the 28 Salmonella cases that consumed the implicated bread, 12 were male. Their age ranged from three to 71 years old. Eleven cases (39.3%) were under 18 years old and one case was over 65. The incubation periods ranged from 4 to 17 (median, mean ± S.D. 10, 9.4 ± 3.2) hours after consumption of the bread. Twenty-seven of the 28 patients were hospitalized. The clinical symptoms among the 28 cases included abdominal pain (100%), fever (100%), diarrhea (100%), vomiting (85.7%), chills (35.7%), and headache (7.1%). The laboratory data showed eight (28.6%) cases had white blood cell counts higher than 10000 /mm 3 . Positive occult blood reaction was found in stool specimens of 24 of 27 (88.9%) patients tested. Colitis was found in all four patients who received colonoscopy examination. The two bacteremia patients' fever subsided one day and two days after admission, respectively, without antimicrobial therapy. No mortality or sequellae occurred among these cases during hospitalization or in the three months' follow-up by infection control nurses. Further investigation by the city public health administration found the incriminated bread from the six markets was all produced by the same bakery that was prohibited from production of all kinds of bread on August 1, 2001. Because the bakery had stopped their production one day before the official prohibition, no further samples of the components of bread, such as the egg and mayonnaise, were available. Investigation of the bakery staff with stool cultures for Salmonella and surveillance culture of the workplace surfaces were not performed. Bacterial strains In the week of this outbreak, there was an extraordinarily high percentage (94.1%) of group D of Salmonella isolates. Twenty-eight Salmonella isolates were cultured from 32 available stool specimens from 34 patients who consumed the implicated food. These isolates were all of group D and were further identified as S . Enteritidis. This pathogen grew in two of the 34 patients' blood cultures. A Salmonella isolate of the same serovar was cultured from bread provided by a patient. The sample had been stored in the refrigerator because the patient had not finished eating it. A total of 30 stool specimens from 25 patients not consuming the implicated bread were collected and Salmonella was isolated from six of the 25 patients' stool specimens. For the six Salmonella isolates from cases not consuming the implicated food during the outbreak period, two were of group B and four were of group D. Further serovar analysis and PFGE analysis of the four group D isolates was not performed. Among the four group D isolates, two were resistant to ampicillin and trimethoprim/sulfamethoxazole, which was different to the antibiogram of the isolates from cases consuming the implicated bread. Antibiogram and PFGE patterns All 28 stool isolates, two blood isolates and one food isolate were susceptible to the nine antimicrobial agents tested. Only two stool isolates showed unrelated PFGE patterns with more than six band differences compared to the epidemic PFGE pattern of the other isolates. All the other stool, blood, and food isolates had the same PFGE pattern indicating a clonal relationship (Figure 2 ). The two isolates with different PFGE patterns were from two patients who had purchased and consumed the bread from two different markets. Discussion The source of the outbreak was traced to ingestion of egg-covered bread topped with mayonnaise and fried seasoned pork fiber from six traditional Chinese markets, and then back to a bakery. This outbreak had no clear association with the common sources of food poisoning outbreaks in Taiwan, such as commercial lunch boxes, or food from banquets or wedding dinners, and it was caused by an uncommon pathogen of food-borne outbreak in Taiwan [ 9 ]. Most of the patients falling ill with gastroenteritis on the 29th and 30th of July (Saturday and Sunday) had not consumed the implicated bread, (Fig. 1 ) suggesting some other source or vehicle existed. However, it could be due to the typical increase of ER visits on Saturdays and Sundays, compared to weekdays, because most clinics stop services on weekend. Compared with phage typing [ 15 ] and plasmid typing [ 12 , 16 , 17 ], pulsed-field gel electrophoresis (PFGE) is a reproducible, discriminative, and feasible typing method for S . Enteritidis [ 18 , 19 ] though it has been suggested to have limited value in epidemiological analysis because of the high genetic homogeneity among strains of S . Enteritidis [ 20 ]. Analytic epidemiologic study in addition to molecular typing is essential in thoroughly investigating the source of Salmonella isolates. Outbreaks can be missed easily if the increase in case number is not noticed due to the wide distribution of outbreak sources in different areas or if a relatively small outbreak occurs among a large population and cases are exposed to the outbreak source at different times. Recognition of the outbreak reported in this study was aided by the routine surveillance of Salmonella groups and knowledge that the percentage of group D Salmonella patients was low before. Thus, the unusually high percentage of group D Salmonella seen within the week of the outbreak led to the investigation. The implicated vehicle of infection in the outbreak was an egg-covered bun topped with mayonnaise and fried seasoned pork fiber which had been distributed to different markets after production at the same bakery. Epidemiological study, antibiograms, and serologic and molecular typing patterns revealed that almost all cases of S . Enteritidis infection during the period were the same as that of the bread isolate. Only the implicated baked good from the same bakery caused Salmonella infections while other items from the bakery were not found to be epidemiologically related with the outbreak. This finding suggests contamination of the pathogen did not occur during the common routes of production, transportation, or selling of goods from the bakery. Although the bread sold in the traditional Chinese markets frequently does not conform to sanitary requirements, and although the staff involved in distribution and selling are not subject to the routine hygiene inspections, in food stores, the possibility of contamination of a specific bakery product simultaneously at six markets in the same time period is very low, suggesting that the contamination occurred during the production process. Food contamination in the outbreak was traced to the same bakery but an investigation of infection among the bakery staff and sources of the contents of the bread was not conducted. Although the baking process involves high temperatures sufficient to kill pathogens, the manual addition of toppings or flavors, such as mayonnaise, eggs, and meat products provide possibilities for contamination with food-borne pathogens. In addition, insufficient baking may be a risk factor for human health because it may not destroy microbial contamination. Different Salmonella serovars have been related specifically with some foods. S . Enteritidis is particularly related to eggs [ 26 - 28 ]. In this outbreak and the other S . Enteritidis outbreaks related to baked goods[ 4 , 5 , 7 , 8 ], all of the food products contained eggs and the egg material had not been cooked sufficiently. Whether the topping of this bread with lightly cooked eggs or the under-cooking played a role in the contamination could not be confirmed due to lack of culture of separate parts of the bread. The finding that Salmonella isolated from the bread had the same serovar, antibiogram, and PFGE pattern as isolates from patients and the significant relationship between consumption of this baked good and the isolation of Salmonella implicated this food product as the vehicle of contamination. This outbreak had a relatively short incubation compared to the 24–72 hour range reported for other Salmonella outbreaks [ 29 , 30 ]. Whether a highly virulent strain or a high inoculum of bacteria [ 30 ] during production or rapid growth of bacteria due to hot summer weather, contributed to the short incubation time was not investigated. For many food-borne outbreaks, the pathogens and transmission vehicles are often not identified, usually because of delayed collection of epidemiologic and microbiologic information [ 28 ]. Initiating an outbreak investigation based on surveillance of emergency room admissions would provide useful information which may lead to early recognition of the pathogen and vehicle [ 31 ]. The alertness of our emergency room staff resulted in recognition of the unusual increase in cases of gastroenteritis, although the patients came from two districts and had no obvious relationship to common food sources. This led to prompt investigation and containment of a potential source of further infection. Cooperation of the emergency room, microbiology laboratory, the infection control team staff at hospitals and public health administration staff combined with the application of epidemiological and bacterial typing methods is crucial to the success of the source identification to prevent further dissemination during Salmonella Enteritidis outbreaks. Competing interest The author(s) declare that they have no competing interest. Authors' contributions Po-Liang Lu and L. K. Siu were in charge of the investigation, data handling and writing of the manuscript. Po-Liang Lu and Shang- Jyh Hwang participated in the design of the study. Po-Liang Lu, In-Jane Hwang, Ya-Lina Tung and Shang-Jyh Hwang were in the investigation team for data collecting and data analysis. Po-Liang Lu performed the statistical analysis. Chun-Lu Lin and L. K. Siu carried out microbiological assays. All authors have read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC535566.xml
534107	Psychiatric diagnoses in 3275 suicides: a meta-analysis	Background It is well known that most suicide cases meet criteria for a psychiatric disorder. However, rates of specific disorders vary considerably between studies and little information is known about gender and geographic differences. This study provides overall rates of total and specific psychiatric disorders in suicide completers and presents evidence supporting gender and geographic differences in their relative proportion. Methods We carried out a review of studies in which psychological autopsy studies of suicide completers were performed. Studies were identified by means of MEDLINE database searches and by scanning the reference list of relevant publications. Twenty-three variables were defined, 16 of which evaluating psychiatric disorders. Mantel-Haenszel Weighted Odds Ratios were estimated for these 16 outcome variables. Results Twenty-seven studies comprising 3275 suicides were included, of which, 87.3% (SD 10.0%) had been diagnosed with a mental disorder prior to their death. There were major gender differences. Diagnoses of substance-related problems (OR = 3.58; 95% CI: 2.78–4.61), personality disorders (OR = 2.01; 95% CI: 1.38–2.95) and childhood disorders (OR = 4.95; 95% CI: 2.69–9.31) were more common among male suicides, whereas affective disorders (OR = 0.66; 95% CI: 0.53–0.83), including depressive disorders (OR = 0.53; 95% CI: 0.42–0.68) were less common among males. Geographical differences are also likely to be present in the relative proportion of psychiatric diagnoses among suicides. Conclusions Although psychopathology clearly mediates suicide risk, gender and geographical differences seem to exist in the relative proportion of the specific psychiatric disorders found among suicide completers.	Background Suicide is an important public health problem that is among the leading causes of death in Western countries [ 1 ]. Over the last years, the relationship between suicide and mental disorders has been the focus of several studies and has generated important debate [ 2 ]. This relationship has been investigated by different strategies, but particularly by the psychological autopsy method [ 3 ], which is generally considered the method of choice to retrieve postmortem information on psychopathology. The psychological autopsy procedure entails the retrospective psychiatric assessment of the deceased by variable methodologies, but generally by means of proxy-based interviews. This procedure is also frequently completed by having access to medical and other relevant dossiers from the subject on whom the psychological autopsy is carried out [ 4 , 5 ]. It is well established that psychopathology is an important predictor of suicide completion [ 6 ], but there is considerable variability between studies in rates of total and specific psychiatric disorders [ 7 ]. One of the most consistent findings in suicidology is the excess of male suicides observed in most countries [ 8 ], with a few notable and important exceptions, such as China [ 1 , 9 ]. Geographic origin is another important source of variation [ 1 ]. However, the possibility that clinical and other behavioural factors could at least partly mediate gender and geographic differences in suicide rates has been little explored. The aim of this study was to carry out quantitative syntheses of overall and specific psychiatric diagnoses found in suicide studies and to explore possible gender and geographical differences in the distribution of psychiatric disorders among suicide completers. Methods Study identification To identify studies for this review, the National Library of Medicine (NLM) PubMed database was searched up to December 2002 using English language and human study limits. The Medical Subject Heading (MeSH) terms "suicide AND psychological autopsy", "suicide AND psychopathology", "suicide AND (postmortem diagnoses OR postmortem diagnosis)", and "(mental disorders/epidemiology) AND prevalence AND ((suicide/statistics & numerical data) NOT suicide attempts)" were used. Finally, in order to find other articles not obtained through electronic searches, reference lists from original studies as well as from not independent studies were screened. Study selection The inclusion criteria for considering articles for this review were as follow. Studies had to: 1) be original, 2) be published in English, 3) contain information on diagnostic distribution, 4) include suicide completers unselected according to specific mental disorders, 5) use of a psychological autopsy method, which for the purpose of this review was considered as the process of reconstructing psychiatric diagnoses based either on interviews with informants (regardless of the specific diagnostic instrument methodology) or on review of multiple official records that contained interviews with informants such as general practitioners, other professionals and relatives or friends, 6) use of standard diagnostic criteria (any versions of the Diagnostic and Statistical Manual of Mental Disorders, the International Classification of Diseases or the Research Diagnostic Criteria). Studies were excluded if: 1) their sample was not independent from that investigated in another study (see below for criteria on which one was included), 2) they were reports on suicide in one specific diagnostic category and 3) if diagnoses were simply extracted from medical records without review of multiple sources of information. A single reviewer (G.A.L.) made a prior screening to identify and select articles. When titles and abstracts were deemed adequate or when they remained too obscure to reach a verdict, full texts were retrieved for further evaluation in conformity with the inclusion and exclusion criteria. Study assessment A total of 23 variables were defined, three of which relate to demographic information, four other concern the method of diagnosis, and 16 evaluate the presence of psychiatric diagnoses. To obtain the latter 16 variables (shown in table 1; see additional file ), every diagnostic term used in the original studies was categorised into one of the 16 pre-defined groups. So diagnoses such as "intermittent depressive disorder" or "neurotic depression" reported in some studies were coded under "depressive disorders' variable and diagnoses such as "alcohol use", "alcohol misuse" and "alcohol abuse" were coded as "alcohol problems". All substances noted as other than alcohol were coded under "other substances problems". These two variables were then recoded as "any substance problems". The same was achieved with the "depressive disorders" and "bipolar disorders" which were recoded as "any affective disorders". Disorders labelled as "other" or as a subset of various disorders without further specification were left aside. For all studies the most specific diagnosis was considered. That is, when the authors broke down general diagnosis such as "affective disorder" into "depressive disorders" and "bipolar disorders", only these more specific diagnoses were noted and accounted for in our study. When two studies or more were carried on the same population, the study with the largest sample and the most informative report was consistently selected. When multiple diagnoses and principal diagnoses (those deemed by the investigators as more related to the suicide) were reported, preference was given to the former. In four cases, secondary diagnoses were added to principal diagnoses to obtain multiple diagnoses [ 10 - 13 ]. Studies for which controls were selected among psychiatric in-patients or matched to suicides by mental diagnosis, only suicide cases were included in our analysis [ 12 , 14 ]. In the study by Graham and Burvill [ 15 ], controls were older suicide completers, and so they were included in our suicide group. In the study by Hawton et al. [ 10 ], only diagnoses for suicides obtained by means of an interview were included. In three case-control studies [ 16 - 18 ], not all suicide cases were matched to a control. In these cases, we considered the full suicide sample in the descriptive analyses, but only the control-matched suicides in the quantitative analyses. Statistical analysis Descriptive analyses and homogeneity tests were carried out before pooling the data. In order to determine the risks of having had a disorder, suicides and controls were recorded in 2 × 2 tables. These data were then stratified by the 16 outcome variables and Mantel-Haenszel Weighted Odds Ratios (OR) and 95% confidence intervals (95% CI) were estimated. Gender differences were also explored by means of Odds Ratios. Major disorders were then compared between the different demographic areas by means of χ 2 to assess variations in the diagnostic distribution across these demographic areas. All statistical analyses were carried out using Epi Info 6, version 6.04d (C.D.C., U.S.A.; W.H.O., Geneva, Switzerland). Results A total of 152 studies were initially identified. After selection according to inclusion/exclusion criteria, 27 studies were included in this review. The most common reasons for exclusion were that a) no diagnostic distribution was provided (n = 46) [ 6 , 19 - 63 ], b) samples were pre-selected according to a psychiatric disorder (n = 30) [ 64 - 93 ], c) there was another report on the same sample that either included more subjects or was more informative (n = 29) [ 3 , 94 - 121 ]. Four other studies were about non-completers [ 122 - 125 ]. Another was not in English [ 126 ], and others reported only on one type of disorder [ 127 , 128 ], and therefore, they were all excluded. Additional 14 studies [ 7 , 129 - 141 ] were excluded because the diagnostic criteria were either unspecified or not standard. The studies by Rich et al. [ 99 ] and by Foster et al [ 142 ] were not independent from, respectively, Rich et al. [ 143 ] and Foster et al. [ 144 ]. Although non-independent, these studies provided information of different quality, and thus, were included in our review. Accordingly, Rich et al. [ 99 ] and Foster et al. [ 142 ] were considered, respectively in the gender difference analysis and the case-control comparisons, whereas the study by Rich et al. [ 143 ] and Foster et al. [ 144 ] were considered for the descriptive analysis. Methodological assessment Among the 27 studies that were retained, 52% (14/27) were case-control studies. Eighty-one percent (22/27) of the studies were published after 1990. Sixty-seven percent of the studies (18/27) used DSM diagnostic criteria, whereas only 22% (6/27) and 11% (3/27) used the ICD and RDC diagnostic criteria respectively. Multiple diagnoses were investigated in 63% (17/27) of the studies, whereas principal diagnoses only were given for the other 10 studies. A description of the demographic and methodological features of these 27 studies is shown in table 2 . Table 2 Description of the 27 studies included in this meta-analysis Study Year Origin Diagnostic criteria Methods Number of diagnoses n Suicide With a Dx (%) n Control with a Dx (%) Matched Appleby et al.[151]* 1999 England ICD-10 Official records and interviews Multiple 84 76 (90%) 64 17 (27%) Living ± 5 year and sex Apter et al.[145]* 1993 Israel DSM-III Official records and interviews Principal 43 35 (81%) Asgard U.[147]* 1990 Sweden RDC Official records and interviews Principal 104 99 (95%) Cavanagh et al.[14] 1999 Scotland DSM-III Official records and interviews Principal 45 44 (98%) Cheng et al.[16]* 1995 Taiwan DSM-III-R Official records and interviews Multiple 116 114 (98%) 226 130 (58%) Living ± 5 years, sex, area of residence Conwell et al.[156]* 1996 USA DSM-III-R Official records and interviews Multiple 141 127 (90%) Foster et al.[142,144] 1997/1999 Ireland DSM-III-R Official records and interviews Multiple 118 106 (90%) 117 30 (26%) List of deceased's GP Age, gender, marital status Harwood et al.[17]* 2001 England ICD-10 Official records and interviews Multiple 100 93 (93%) 54 N/A Natural deaths Age and sex Hawton et al.[10] 2002 England ICD-10 Official records and interviews Multiple 42 38 (90%) 84 6 (7%) Living nurses ± 10 years, specialty and seniority Henriksson et al.[11] 1993 Finland DSM-III-R Official records and interviews Multiple 229 225 (98%) Houston et al.[12] 2001 England ICD-10 Official records and interviews Multiple 47 40 (85%) Lesage et al.[150] 1994 Canada DSM-III-R Official records and interviews Multiple 75 69 (92%) 75 N/A Living Neighbourhood, age, marital status and occupation Phillips et al.[9]* 2002 China DSM-IV Interviews with informants Principal 519 325 (63%) 536 93 (17%) Accidental deaths Geographical areas Rich et al.[143] 1986 USA DSM-III Official records and interviews Multiple 283 258 (91%) Runeson B.[153] 1989 Sweden DSM-III-R Official records and interviews Principal 58 57 (98%) Shaffer et al.[18]* 1996 USA DSM-III Official records and interviews Multiple 119 108 (91%) Shaffi et al.[13]* 1988 USA DSM-III Official records and interviews Multiple 21 20 (95%) 21 11 (52%) Living friends Sex, age, race, education, religion, income, and father's education Vijayakumar et al.[159]* 1999 India DSM-III-R Official records and interviews Principal 100 88 (88%) 100 14 (14%) Living SES, sex and ± 2 years Waern et al.[154]* 2002 Sweden DSM-IV Official records and interviews Multiple 85 82 (96%) 153 28 (18%) Living Sex, ± 2 years Boardman et al.[152] 1999 England ICD-10 Multiple official records Multiple 212 151 (71%) 212 40 (19%) Unnatural deaths ± 5 years and sex Cantor et al.[157] 1989 Australia DSM-III-R Multiple official records Principal 47 41 (87%) Groholt et al.[149]* 1997 Norway DSM-III-R Multiple official records Multiple 121 90 (74%) Thacore et al.[158] 2000 Australia ICD-9 Multiple official records Principal 75 46 (65%) Graham et al.[15] 1992 Australia DSM-III Multiple official records Multiple 136 120 (88%) Brent et al.[148] 1999 USA DSM-III Interviews with informants Multiple 140 115 (82%) 131 32 (24%) Living Age, race, gender, country and SES Cerel et al.[155] 2000 USA RDC Interviews with informants Multiple 15 13 (87%) 201 70 (35%) Non-suicide bereaved family Arato et al.[146]* 1987 Hungary RDC Interviews with informants Principal 200 162 (81%) * Based on axis I disorders only. N/A – information not available or not clear Demographic features A total of 3275 suicides were included in our study with a mean number of 121 (standard deviation (SD) 103) suicides per study. There were 11 studies where diagnoses were given by gender for a subtotal of 933 males and 462 females [ 10 , 11 , 18 , 99 , 144 - 150 ] There were 14 studies [ 10 - 12 , 14 , 17 , 142 , 145 - 147 , 149 , 151 - 154 ] carried out in Europe, including one in Israel [ 145 ]. These 14 European studies comprised a total of 1488 suicides. Seven studies were from North America [ 13 , 18 , 143 , 148 , 150 , 155 , 156 ] with 794 suicides, three others were from Australia [ 15 , 157 , 158 ] with 258 suicides and, finally, three were from Asia [ 9 , 16 , 159 ]. with 735 suicides. Diagnostic distribution The mean percentage of suicides with a psychiatric diagnosis was 87.3 % (SD 10.0 %). However, only 14 of the 27 studies reported both axes I and II disorders (see table 2 ). The remaining 13 studies only assessed axis I diagnoses. The mean percentage of controls with a diagnosis was, as expected, lower (34.9 % SD 25.1 %). As a comparison, among studies not included because the diagnostic criteria were not specified or not standard, the mean percentage of suicides with a diagnosis was not statistically different from that of the studies included in this review (78.7% SD 21.0%, χ 2 : 2.27, p = 0.13). On average, 43.2% (SD 18.5%) of suicide cases were diagnosed with any affective disorders (including depressive and bipolar disorders) and 25.7% (SD 14.8%) with other substance problems. In these groups, respectively, depressive disorders and alcohol problems were the most frequent. Finally, personality disorders represented 16.2% (SD 8.6%) of the suicide diagnoses and psychotic disorders, including schizophrenia accounted for 9.2% (SD 10.2%). The samples from the 14 case-control studies were found homogeneous for the 16 outcome variables according to a homogeneity test (results not shown), allowing us to pool the individual studies and determine overall risks. Table 1 (see additional file ) shows that, with the exception of organic disorders and adjustment disorders, suicide cases had a higher risk of being diagnosed than controls with each of the diagnoses considered. Of these diagnoses, the risks for psychotic disorders were the highest (OR = 15.38; 95% CI: 3.53–97.82) followed by the variable "at least one psychiatric disorder" (OR = 10.50; 95% CI: 9.60–13.56). The risk for schizophrenia was also particularly high (OR = 5.56; 95% CI: 3.12–10.24). This is due to the fact that there were only 15 control subjects altogether diagnosed with schizophrenia and two with psychotic disorders. Statistically significant differences were found when male and female suicide cases were compared (see table 3 ). However, gender-based comparisons should be considered cautiously as, when available, demographic information indicated that female suicides included in the studies reviewed tended to be older than males (table 5 ). Nevertheless, even considering this potential limitation, the results are interesting. The risks for alcohol (OR = 2.19; 95% CI: 1.63–2.95), other substance problems (OR = 2.02; 95% CI: 1.32–3.10), and any substance problems (OR = 3.58; 95% CI: 2.78–4.61), personality disorders (OR = 2.01; 95% CI: 1.38–2.95) or childhood disorders (OR = 4.95; 95% CI: 2.69–9.31) were greater in male as opposed to female suicides. On the other hand, the risks of having depressive disorders (OR = 0.53; 95% CI: 0.42–0.68) or any affective disorders (OR = 0.66; 95% CI: 0.53–0.83) were lower in males. Table 3 Odds Ratios for major outcome variables across sexes Disorders n for females n for males OR (95% CI) χ 2 p-value Any psychiatric disorders 398 801 0.98 (0.70–1.36) 0.02 0.881 Schizophrenia 17 44 1.30 (0.71–2.39) 0.79 0.373 Other psychotic disorders or psychosis NOS 15 40 1.33 (0.71–2.56) 0.88 0.347 Somatoform, anxiety and neurotic disorders 33 83 1.27 (0.85–1.97) 1.24 0.265 Bipolar disorders 26 43 0.81 (0.48–1.38) 0.68 0.409 Organic disorders 6 15 1.24 (0.45–3.60) 0.20 0.656 Adjustment disorders 31 64 1.02 (0.64–1.64) 0.01 0.917 Disorders more likely if male Alcohol problems 73 272 2.19 (1.63–2.95) 29.57 0.000 Other substances problems 32 122 2.02 (1.32–3.10) 11.89 0.001 Any substances problems 110 436 3.58 (2.78–4.61) 110.18 0.000 Personality disorders 41 153 2.01 (1.38–2.95) 14.60 0.000 Childhood disorders 13 117 4.95 (2.69–9.31) 34.57 0.000 Disorders more likely if female Depressive disorders 199 268 0.53 (0.42–0.68) 28.56 0.000 Any affective disorders 272 454 0.66 (0.53–0.83) 12.91 0.000 Other disorders 16 12 0.36 (0.16–0.82) 7.44 0.006 Table 5 Descriptive analysis of the age and sex of subjects Age (mean ± SD) n [Studies] All regions ♂ 28.5 ± 12.8 880 [11,18,143,145,148-150,158] ♀ 34.5 ± 17.8 333 [11,18,143,147-149,158] Both sexes* 41.6 ± 17.8 794 [11,14,17,149,151,154,157,158] American Studies ♂ 26.0 ± 12.3 491 [18,143,148,150] ♀ 27.3 ± 18.9 127 [18,143,148] Both sexes* N/A N/A European Studies ♂ 27.2 ± 15.4 314 [11,145,149] ♀ 37.9 ± 18.9 191 [11,147,149] Both sexes* 42.3 ± 20.8 672 [11,14,15,17,151,154] Australian Studies ♂ 42.5 491 [158] ♀ 45.7 15 [158] Both sexes* 39.5 ± 5.2 122 [157,158] Asian Studies ♂ N/A N/A ♀ N/A N/A Both sexes* N/A N/A N/A – information not available * Both sexes refers to studies in which information on age by sex was not provided, and thus, only mean age for the whole sample was available. Analysing the data according to geographic areas, the diagnostic distribution of the key diagnoses found in suicides differed significantly between world regions (see table 4 ), but as mentioned above, potential age-related biases may apply (table 5 ). The American suicides were more often diagnosed with a psychiatric disorder than suicides in the other regions of the world; 89.7 % (SD 4.2 %) of the American suicides had at least one diagnosis, whereas 88.8 % (SD 8.9 %) of the European suicides, 83.0 % (SD 18.4 %) of the Asian suicides and 78.9 % (SD 15.3 %) of the Australian suicides had at least one psychiatric diagnosis. Table 4 Diagnostic distribution across different regions of the world European (%) North American (%) Australian (%) Asian (%) χ 2 Affective disorders 753 (48.5) 390 (33.6) 71 (32.7) 335 (51.3) 11.3* Substances-related disorders 390 (18.6) 573 (40.1) 106 (24.1) 135 (26.7) 12.1* Schizophrenia and other psychotic disorders or psychosis NOS 125 (7.5) 42 (4.2) 29 (24.3) 53 (8.4) 24.1* Personality disorders 197 (16.8) 75 (13.4) 75 (17.7) 20 (17.7) 1.2 n.s. At least one Diagnosis 1298 (88.8) 710 (89.7) 207 (78.9) 527 (83.0) 6.4 n.s. * Significant at p ≤ 0.01 n.s. Non significant Discussion Total psychopathology Since the first psychological autopsy studies by Robins et al. [ 139 ] in North America and by Barraclough et al. [ 7 ] in Europe, a relatively small number of studies have been carried out. These original studies were descriptive in nature, and only more recently case-control studies have been performed. The data from these studies have consistently suggested a clear relationship between mental disorders and suicide. Here we systematically reviewed these studies and pooled their results whenever possible. Our results show that, on average, 87.3 % of the subjects who committed suicide had a mental disorder. On the other hand, an average of 14.0 % of these subjects was not diagnosed with a psychiatric disorder. A possible explanation is that a diagnosis failed to be detected due to various methodological shortcomings. This possibility is concrete, as psychological autopsy studies rely on informants and/or available medical information to generate diagnostic data. In some cases, the informant has little information on the last weeks or months of life of the subject. Therefore, it is possible that the overall rate of psychopathology observed is still underestimated. This is consistent with findings from recent studies by our group focusing on suicides without an axis I diagnosis [ 160 ]. Specific diagnoses This review confirms the overall impression from individual studies that affective, substance-related, personality and psychotic disorders account for most of the diagnoses among suicides. The two single most common diagnostic categories among suicide completers were any affective disorders (diagnosed in 43.2 % of suicide cases), and any substance disorders (present in 25.7 % of suicide cases). Recent studies on comorbidity indicate that suicide completers are more likely to have more than one psychiatric diagnosis [ 142 , 161 ]. In a comparison with matched community controls, Foster et al. [ 142 ] found a significant increase in suicide risk in the presence of Axis I-Axis II comorbidity (OR = 346.0, p < 0.0001). Our group [ 161 ], investigating male completers and controls from the general population, found that suicide cases had an average of 2.36 diagnoses and that comorbidity in completers tended to be of three different patterns, according to mean number of diagnoses (range 1.19 – 4.05) and presence of impulsive-aggressive behaviours. Thus, it would have been interesting to assess overall levels of comorbidity in this review, as well as to investigate what is the amount of overlap between the different diagnoses investigated. However, very little, if any, information about comorbidity was present in the original studies reviewed and this information was impossible to retrieve from the published data. Gender differences The investigation of gender differences in rates of psychopathology associated to suicide should be regarded in light of the methodological limitations of this review, which are primarily related to difficulties in comparing studies carried out using different methodological procedures, diagnostic instruments and criteria, in addition to potential differences in sample characteristics, including age distribution. However, given the important effect that gender seems to have as a suicide risk moderator and the relative lack of appropriate investigation focusing on gender differences in suicide completion, the observed differences in rates of psychopathology in male and female suicides are interesting and should be considered for validation in future studies. Our results indicate that the risk of substance-related disorders, personality disorders and childhood disorders are significantly higher in male suicides, whereas, the risk of affective disorders, specifically, depressive disorders, are greater in female suicides. On average, any substance problems represented 41.8 % (SD 21.1 %) of the male diagnoses and 24.0 % (SD 16.5 %) of the female diagnoses (χ 2 7.29 p = 0.007), whereas affective disorders represented 59.4 % (SD 13.9 %) of the female diagnoses and 47.4 % (SD 12.7 %) of the male diagnoses (χ 2 2.88 p = 0.089). Although there has been much discussion on possible factors that could help explain gender differences in suicide rates, most of the studies have primarily focused on psychosocial and demographic risk factors. There is very little data on the possible role of psychiatric and/or behavioural characteristics, which may also mediate gender differences in suicide risk. This study suggests that the underlying psychiatric morbidity may be different in male and female suicide completers. An important question that follows is whether or not the differences found in this study between male and female suicides are the consequence of gender differences in the prevalence of psychiatric disorders in the general population. Although possible, it is unlikely that differences in population rates of psychiatric disorders could explain the different distribution of psychiatric disorders observed in this study, as the gender-specific risks found were not consistently reflecting gender-differences observed in prevalence rates (for instance, schizophrenia and psychotic disorders) and they were not always in the same direction (for instance, personality disorders). An interesting finding of this study was precisely the absence of gender differences in schizophrenia. This is not necessarily inconsistent with suggestions that most of the suicide cases in schizophrenia are males [ 162 - 164 ], as our findings basically indicate that there are no relative differences between genders in the proportion of suicide cases that are diagnosed with schizophrenia. However, our findings are inconsistent with the common generalization that schizophrenics tend to commit suicide early in the course of the disease because, given gender-differences in the age at onset in schizophrenia [ 165 ], with males more likely to have the onset at younger ages, one would expect a considerably higher proportion of schizophrenia among male completers, even if the age distribution in our sample suggests that women in general seemed older than men. In summary, despite the potential methodological limitations discussed above, our results in gender differences in clinical correlates of suicide are interesting and should be further investigated. Geographic differences We also found differences in rates of psychiatric disorders in studies from different geographic origins. This finding may indicate social and cultural factors influencing how one views and interprets suicide and cultural biases towards or against specific diagnoses. Alternatively, as discussed for gender-based comparisons, demographic (age, rural vs. urban samples, socioeconomic and educational level, etc.) differences between the samples could explain some of these results. In view of that, similar limitations, as those for the analysis of gender differences, apply to the analysis of geographical differences in rates of psychopathology associated to suicide (see table 5 ). American women seem younger than in any other region, Australian women and men appear older than those in the other regions, and no Asian studies provide age means for their sample. In spite of these limitations, our review suggests that, although psychopathology mediates suicide worldwide, there seem to be differences across different parts of the world in the relative proportion of the specific psychiatric disorders found among suicide completers. As mentioned above, these differences may be attributed to variance in psychological autopsy methodologies between countries, or yet, to important differences in the prevalence of psychiatric disorders. Although it is possible that methodological differences between studies play a certain role explaining some of the differences found, it is unlikely that they accounted for all differences found as the studies included in these regional comparisons used similar methods and diagnostic criteria, whereas the differences found were substantial. It is not likely either that diversity between countries in prevalence of psychiatric disorders account for all the observed regional differences, as for some of these disorders, such as schizophrenia, it is thought that there is little variation in prevalence rates between different populations [ 166 ]. Thus, the geographical differences observed in the relative proportion of psychiatric disorders among suicide completers is an interesting issue that should be further explored. Most limitations of this study are common to all quantitative systematic reviews. In particular to this study, one should take into account that the quantitative review was carried out with studies that, although published in a relatively short period of time (from 1986 to 2002), have variation in diagnostic criteria used and have different methodological rigor. Moreover, it is possible that between-study variation in the distribution of a series of demographic variables could have had an impact on our findings. We chose not to control for these methodological differences as given the diverse sources of possible variation, doing so would have considerably limited the number of studies included in the review. Therefore, we opted to be more inclusive and consider the results of this review as preliminary and providing information to be further investigated. Over the course of this study, a report on another meta-analysis of psychological autopsy studies was published. This study, by Cavanagh et al. [ 167 ], reviewed the literature on psychological autopsies and yielded similar overall results. However, there are differences between the study by Cavanagh et al [ 167 ] and ours, both in methodology and major aims. While they identified studies through a larger number of library databases, they included only studies up to June 2000. Moreover, they did not investigate risks attributed to specific diagnostic categories, but rather risks attributed to mental health disorders, presence of an affective disorder and comorbidity. They also investigated the role of a few social variables and did not carry out analyses exploring a possible gender and geographic difference in relative rates of psychopathology. Conclusions Our study carried out a systematic review of psychological autopsy studies of suicide and indicates that overall, 87.3% of suicide cases have a history of psychiatric disorders. We also found that male suicides have a different psychiatric profile than female suicide cases and that the relative proportion of psychiatric disorders in suicide completers tends to vary according to geographical region. Competing interests The author(s) declare that they have no competing interests. Authors' contributions GAL carried out the search, extraction of data, analysis and drafted the manuscript. CK helped with the design of the review, and the statistical analysis. GT conceived the study and participated in the design and coordination. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 Table 1 – Mantel-Haenszel Weighed Odds Ratio. This table gives Mantel-Haenszel Weighed Odds Ratio for the 14 case-control studies included in this meta-analysis for the 16 variables of psychiatric disorders. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC534107.xml
546211	Assessment of dizziness among older patients at a family practice clinic: a chart audit study	Background Dizziness is a common complaint among the elderly with a prevalence of over 30% in people over the age of 65. Although it is a common problem the assessment and management of dizziness in the elderly is challenging for family physicians. There is little published research which assesses the quality of dizziness assessment and management by family physicians. Methods We conducted a retrospective, chart audit study of patients with dizziness attending the Sunnybrook Family Practice Center of Sunnybrook and Women's College Health Sciences Center (SWCHSC) in Toronto. We audited a random sample of 50 charts of patients from 310 eligible charts. Quality indicators across all dizziness subtypes were assessed. These quality indicators included: onset and course of symptoms; symptoms in patients' own words; number of medications used; postural blood pressure changes; symptoms of depression or anxiety; falls; syncope; diagnosis; outcome; specialty referrals. Quality indicators specific to each dizziness subtype were also audited. Results 310 charts satisfied inclusion criteria with 20 charts excluded and 50 charts were randomly generated. Documentation of key quality indicators in the management of dizziness was sub-optimal. Charts documenting patients' dizziness symptoms in their own words were more likely to have a clinical diagnosis compared to charts without (P = 0.002). Conclusions Documentation of selected key quality indicators could be improved, especially that of patients' symptoms in their own words.	Background Dizziness is a common complaint among the elderly, with a prevalence of more than 30% in people over age 65 [ 1 ] and it accounts for 2% of consultations in the primary care setting [ 2 ]. Drachman and Hart [ 3 ] described four subtypes of dizziness: vertigo, lightheadedness, dysequilibrium, and others. Several recent community-based studies of dizziness shows that, among the 4 dizziness subtypes, the proportion of vertigo was more uniform, ranging from 28 to 32% [ 1 , 4 - 6 ]. Reported frequencies of specific diagnoses for dizziness varies widely however, depending on: 1) clinical setting (primary care setting, referral center or emergency department); 2) patient age or patient populations examined; and 3) investigator bias. These methodological problems limit the generalizability of the etiological studies [ 3 ]. Kroenke et al [ 7 ] found in their systematic review that dizziness was attributed to peripheral vestibulopathy in 44%, central vestibulopathy in 11%, psychiatric causes in 16%, other conditions in 26%, and an unknown cause in 13% of cases. Life-threatening illness is rare in patients with dizziness (with cerebrovascular disease accounting for 6%, cardiac arrythmia for 1.5% and brain tumor for <1%) [ 7 ]. However, many do have serious functional impairment, such as increased risks for falls and increased incidence of symptom-related fears, anxiety or depression [ 8 - 10 ]. Many patients with chronic dizziness, particularly the elderly, are under-referred for specialist consultation and thus are not receiving timely treatment [ 5 ]. When assessing dizziness, what concerns a family physician most are: 1) how to distinguish serious causes of dizziness from less urgent ones; 2) how to manage patients with chronic but yet debilitating dizziness; and 3) how to decide on the right timing and the appropriate specialty for referral. However, many family doctors describe dizziness as "confusing" and "discouraging" problem [ 8 ] and expensive investigations like electro-nystagmography and MRI are rarely helpful [ 4 ]. In fact, a diagnosis cannot be ascertained in many patients with dizziness and many patients may have more than one diagnosis [ 11 ], making management difficult. To date, there are no evidence-based guidelines in the management of dizziness among elderly patients in a primary care setting because most past studies on dizziness have been retrospective or in referral settings. Traditionally, the approach to dizziness is "disease-oriented", in which the clinician aims, at a minimum, to exclude potentially fatal causes and possibly to diagnose a specific cause for treatment. On the other hand, some authors like Tinetti et al [ 12 ] and Kao et al [ 13 ] regard dizziness in the elderly as a "geriatric syndrome", because it represents dysfunction in more than one body system and has multiple predisposing risk factors. This function-oriented approach focuses on impairment reduction to reduce morbidity associated with dizziness, regardless of etiology. Tinetti's epidemiological population-based study [ 12 ] found that seven characteristics were associated with dizziness in the elderly: anxiety, depression, using five or more medications, impaired balance, past myocardial infarction, postural hypotension, and impaired hearing. Despite differences in the above two approaches, both share in common certain key quality indicators as reflected in recent studies and reviews [ 4 , 7 , 11 - 15 ]. The purpose of this chart audit study was to assess the extent to which family physicians included these key quality indicators when assessing and managing the dizzy elderly patient. Methods A retrospective chart audit was conducted at the Family Practice Center (Sunnybrook Campus) of Sunnybrook and Women's College Hospital Health Sciences Center. Inclusion criteria for the chart audit were: 1) Patients with a International Classification of Disease (ICD-9) diagnostic billing code of "780" (dizziness); 2) Patients seen between Feb 1 st 2001 and Jan 31 st 2003; 3) Patients 65 years of age or older when seen. Exclusion criteria were: 1) Patients who are discharged from service or died; 2) Patients whose presenting symptoms were not dizziness or any of its subtypes. A chart audit intake form was designed [ Additional File 1 ], which included quality indicators important in the diagnosis and management of dizziness, based on recommendations from several recent review articles and peer-reviewed studies [ 4 , 7 , 11 - 15 ]. A random sample from the eligible charts was then audited and the data analyzed for descriptive statistics using SPSS. The general outcome measures/quality indicators across all dizziness subtypes include the documentation of : 1) onset and course of symptoms; 2) symptoms in the patients' own words; 3) number of medications used; 4) postural blood pressure changes; 5) symptoms of depression or anxiety; 6) falls; 7) syncope; 8) diagnosis; 9) outcome of dizziness; 10) specialty referrals. The quality indicators specific to vertigo include the documentation of 1) episode duration; 2) relationship to head turning; 3) tinnitus and hearing loss; 4) ear examination; 5) neurological examination; 6) spontaneous nystagmus; 7) positional nystagmus (Hallpike manoevre) (16). These patients were also audited for whether audiometry was ordered and whether Epley's manoevre [ 17 ] was offered if BPV was diagnosed. The quality indicators specific to lightheadedness include the documentation of 1) relationship to postural change; 2) cardiac symptoms; 3) syncope; 4) orthostatic blood pressure changes. These patients were also audited for whether ECG or Holter monitoring were ordered. The quality indicators specfic to disequilibrium include the documentation of 1) falls; 2) neurological exam; 3) cerebellar signs; 4) Gait examination; 5) Romberg's sign; 6) visual acuity. The quality indicators specifically to other non-classifiable dizziness include the documentation of symptoms of depression and anxiety. Results 310 charts satisfied the inclusion criteria with 20 charts excluded. A random sample of 50 charts were generated for the audit. The demographics of the sample, including age, gender living situation, are described in Table 1 . Of note is that 62% of the patients are 80 years of age or older and 28% of patients are living alone. Table 1 Demographics and living situation of patients (n = 50) Total % Male % Female Age > or = 65 100% (n = 50) 42% (n = 21) 58% (n = 29) Age < 80 38% (n = 19) 16% (n = 8) 22% (n = 11) Age > or = 80 62% (n = 31) 26% (n = 13) 36% (n = 18) Age range 65–91 66 to 89 65 to 91 Living Alone 28% (n = 14) 2% (n = 1) 26% (n = 13) Living with Spouse 18% (n = 9) 12% (n = 6) 6% (n = 3) Living with Family 8% (n = 4) 2% (n = 1) 6% (n = 3) Living situation Not Documented 46% (n = 23) 26% (n = 13) 20% (n = 10) The distribution of different subtypes of dizziness is described in Table 2 , with more patients presenting with lightheadedness (40%) and dysequilibrium (38%) than vertigo (28%). There are more females than males among the patients with lightheadedness (30% vs. 10%) and vertigo (16% vs. 10%) whereas the ratio of females to males is roughly the same among those with dysequilibrium (20% vs. 18%). 30% (n = 15) of patients presented with more than one subtype of dizziness. Table 2 Symptomatology distribution of patients Dizziness Subtype Yes No Not Documented Total Female Male Total Total Vertigo 26% (n = 13) 16% 10% 26% (n = 13) 48% (n = 24) Lightheadedness 40% (n = 20) 30% 10% 10% (n = 5) 50% (n = 25) Disequilibrium 38% (n = 19) 20% 18% 6% (n = 3) 56% (n = 28) Others 26% (n = 13) 12% 14% 6% (n = 3) 68% (n = 34) The onset and diagnoses of dizziness are described in Table 3 . 70% of patients have a precipitating factor, the commonest ones being postural change, movement and head turning. 46% of patients have no diagnosis while 10% of patients have more than one diagnoses. Among patients with an ascertained diagnosis, the most common ones are BPV (12%), labyrinthitis (10%) and TIA/Stroke (8%). Significantly, patients were more likely to be diagnosed if their symptoms were documented in their own words compared to those without such documentation (see Table 4 ). Table 3 Onset and diagnoses of dizziness Percentage of Patients Onset Spontaeous 4% (n = 2) Precipitating Factors present 70% (n = 35) Postural change 38% (n = 19) Head turning 12% (n= 6) Any movement 18% (n= 9) Walking 10% (n = 5) Anxiety 2% (n = 1) Other factors 26% (n = 13) Not Documented 26% (n = 13) Diagnosis No 46% (n = 23) Yes 54% (n = 27) BPV 12% (n = 6) Labyrinthitis 10% (n = 5) TIA/Stroke 8% (n = 4) Hypertension 6% (n = 3) Depression/Anxiety 6% (n = 3) Arrhythmia 4% (n = 2) Alcohol 4% (n = 2) Dehydration 4% (n = 2) Others 4% (n = 2) More than One Diagnoses 10% (n = 5) Table 4 Documentation of patients' symptoms in their own words Documentation With Diagnosis Without Diagnosis Yes (n = 22) 82%* (n = 18) 18% (n = 4) No (n = 28) 32% (n = 9) 68% (n = 19) *statistically significant (P = 0.002) The documentation of general and dizziness subtype-specific quality indicators in history and physical examination are described in Table 5 . It also was observed that: 1) 60% of all patients were taking at least 5 medications; 2) three vertiginous patients with associated ear symptoms were not offered audiometry; 3) none of the four vertiginous patients with an abnormal Hallpike test were documented to be treated by Epley's manoevre; 4) in the lightheadedness subgroup, ECG was ordered in only 40% and Holter monitoring in only 30% of patients. Table 5 Documentation of general and dizziness subtype-specific quality indicators in history and physical examination for Patients with Vertigo, Lightheadedness, Disequilibrium and Others Yes No Not Documented General (n = 50) Postural BP changes 6% 44% 50% Associated Depression 6% 8% 86% Associated Anxiety 14% 0% 86% Falls 6% 30% 64% Syncope 2% 36% 62% Vertigo (n = 13) Episode Duration 38% N/A 62% Relationship to Head Turning 38% 8% 54% Tinnitus 23% 54% 23% Hearing Loss 0% 0% 100% Ear Examination 39% (Normal:31%) (Abnormal:8%) N/A 61% Neurological Exam 92% N/A 8% Spontaneous Nystagmus 46% (Normal 38%) (Abnormal 8%) N/A 54% Hallpike Manoevre 39% (Normal 8%) (Abnormal 31%) N/A 61% Lightheadedness (n = 20) Relationship to Postural change 55% 15% 30% Chest Pain 5% 60% 35% Palpitation 5% 55% 40% Syncope 5% 50% 45% Orthostatic drop in BP 10% 55% 35% Orthostatic rise in pulse 0% 15% 85% Disequilibrium (n = 19) Falls 11% 32% 57% Gait examination 58% (Normal: 26%) (Abnormal:32%) N/A 42% Neurological exam 68% N/A 32% Romberg's sign 42% (Normal: 37%) (Abnormal: 5%) N/A 58% Cerebellar signs 47% (Normal: 42%) (Abnormal: 5%) N/A 53% Visual acuity exam 10% (Normal: 5%) (Abnormal: 5%) N/A 90% Others (n = 12) Depression 0% 25% 75% Anxiety 25% 0% 75% N/A: Not Applicable As for the course of dizziness, only 2 patients have worsening symptoms (4%) and 60% of patients are referred to specialty services, the commonest ones of which are ENT (12%), neurology (8%) and cardiology (6%). Discussion A striking finding from this study was that 46% of the patients did not have any diagnosis and 10% of them had more than one diagnosis. This finding is in accordance with the data from the review by Sloan et al [ 11 ] and illustrates the difficulty of diagnosing dizziness in a primary care setting. This is also reflected by a 40% referral rate to specialty services, which is higher compared to the 16% referral rate shown in a recent study [ 5 ]. On the other hand, the dizziness symptom worsened with time in only 4% of patients in this study, which is consistent with previous work [ 7 ] showing the generally "benign" course of this condition. The distribution of etiological causes of dizziness in this sample is also consistent with those of previous studies [ 7 ] with peripheral vestibular disorders (BPV and labyrinthitis) being the most common and accounting for 22% of diagnoses. Effective history taking and communication between family physicians and patients is of crucial importance in the diagnosis of dizziness. The present chart audit study showed that family physicians were more likely to reach a diagnosis when patients' symptoms were documented in their own words, compared to those without such documentation. Overall, the documentation rate of key quality indicators important to all dizziness subtypes were low, such as falls, syncope, symptoms of depression and anxiety, and orthostatic blood pressure changes. A history of falls is associated with increased morbidity but this was documented in only 36% of patients. This is especially worrying given that 28% of patients in our sample were known to be living alone. The finding that 60% of patients with dizziness are using 5 or more medications is consistent with Tinetti's population-based cross-sectional study [ 12 ]. Among vertiginous patients, the documentation rates for episode duration, relationship to head turning, hearing loss, and Hallpike maneuver were far from satisfactory. Among lightheaded patients, the documentation rates for symptom relationship to postural change, chest pain, palpitation, syncope, and orthostatic blood pressure changes are better but there is still room for improvement. Among patients with disequilibrium, the documentation rates for falls, gait examination, cerebellar signs, Romberg's sign and visual acuity examination were again sub-optimal. Among patients with non-specific dizziness, symptoms of depression and anxiety were also sub-optimally documented. This study has several limitations. First, being a retrospective study, its strength is limited because there is no standardized strategy or protocol for data collection among different family physicians. Second, a chart audit is prone to documentation bias and incompleteness, depending on individual family physicians. In addition, the sample size of 50 is relatively small given the complexity of the clinical problem. Moreover, only one diagnostic billing code was used. Although this would catch the presenting symptom at its undifferentiated stage, the drawback is that we may underestimate the actual scope of the problem by ignoring patients who were coded more specifically by their dizziness subtypes. In addition, only one hospital site, namely Sunnybrook hospital, was selected for chart audit. Being an academic teaching center with predominantly older patients, the patient data from this hospital alone may not be generalizable to those of a community clinic setting. Future directions for study would include the conduction of more prospective cohort studies on primary care patients using a standardized protocol for data collection This would assure uniform and consistent evaluation with the least amount of selection bias. Ideally, the use of inception cohorts would allow for better definition of the causes and natural history of dizziness in persons having their first episode. More prospective studies on dizziness outcomes with extended follow-up periods, as well as studies on different management strategies and specialty referral patterns for dizziness in the primary care setting would be helpful to family physicians. The ultimate goal is to identify the clinical and demographic or situational characteristics in a primary care setting that could help predict management decisions, such as diagnosing the most likely attributable cause of dizziness, delivering the most effective treatment for the symptom, and making the most timely and appropriate referral. Conclusions Regardless of whether one regards dizziness as a "geriatric syndrome" or as a discrete "disease" for which a clear assessment algorithm could be established, there are certain quality indicators that are helpful for either approach but their documentation were found to be sub-optimal in this chart audit study. Allowing patients to describe their symptoms in their own words may help to improve diagnosis and management. Competing interests The author(s) declare that they have no competing interests. Authors' contributions EK contributed to the design and conception of the study, carried out the chart audit, analyzed the data and wrote the first draft of the manuscript. NP contributed to the design and conception of the study, supervised the conduct of the chart audit and data analysis and wrote and revised subsequent versions of the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 1. Chart Audit Form MS Word document with the chart audit form used in the study Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546211.xml
545081	An assessment of factors contributing to treatment adherence and knowledge of TB transmission among patients on TB treatment	Background The treatment guidelines for tuberculosis treatment under Directly Observed Treatment, Short-course (DOTS) have been a common strategy for TB treatment in Zambia. The study was carried out in Ndola, Zambia, to investigate factors contributing to treatment non-adherence and knowledge of TB transmission among patients on TB treatment, in order to design a community-based intervention, that would promote compliance. Methods A household-based survey was conducted in six randomly selected catchment areas of Ndola, where 400 out of 736 patients receiving TB treatment within the six months period, were recruited through the District's Health Management Board (DHMB) clinics. All patients were interviewed using a pre-tested structured questionnaire, consisting of i. Socio-demographic characteristics ii. Socio-economic factors iii. Knowledge about TB transmission and prevention iv. Patterns in health seeking behaviour and v. TB treatment practices at household level. Results Most male TB patient respondents tended to be older and more educated than the female TB patient respondents. Overall, 29.8% of the patients stopped taking their medication. There were 39.1% of the females and 33.9% of the males, who reported that TB patients stopped taking their medication within the first 2 months of commencing treatment. Age, marital status and educational levels were not significantly associated with compliance. The major factors leading to non-compliance included patients beginning to feel better (45.1% and 38.6%), lack of knowledge on the benefits of completing a course (25.7%), running out of drugs at home (25.4%) and TB drugs too strong (20.1% and 20.2%). There was a significant difference [OR = 1.66, 95% CI 1.23, 2.26] in TB knowledge, with more males than females reporting sharing of cups as a means for TB transmission, after adjusting for age, marital status and educational levels. Significantly [p = 0.016] more patients who had resided in the study for less than two years (59%) were more likely to report mother to child transmission of TB, compared to 41.2% of those who had been in the area for more than 2 years. Conclusion This study established that 29.8% of TB patients failed to comply with TB drug taking regimen once they started feeling better.	Background Tuberculosis is one of the major public health risk factors of high mortality among patients in Zambia. The prevalence of tuberculosis among adults has more than doubled since the onset of AIDS epidemic in Zambia [ 1 ], with most of the cases often related to HIV/AIDS conditions. Because of the advent of AIDS epidemic in the country, tuberculosis cases have increased to over 500/100,000 population [ 2 ]. Due to AIDS, it has been projected that TB cases would reach 38,000 by the year 2004 in Zambia [ 3 ]. A multisite study conducted in Ndola in 1997 [ 4 ] revealed that HIV prevalence was as high as 42%, and in a sub-sample of women attending antenatal clinics (ANC), HIV prevalence rate was 28.4% [ 5 ]. Data from Ndola and other Zambian communities revealed that the number of TB cases had risen five-fold since 1996 [ 6 ]. Because of the communities' health seeking behaviour and their perceived causes of TB among patients, those suffering from TB would transmit the disease to others and merging drug-resistance strains would make it even more difficult to limit the spread of the infection [ 1 , 3 ]. Although tuberculosis treatment guidelines had been circulated for the management of the disease at health facility down to household level, DOTS was the only best means of increasing compliance among TB drug-taking patients in Zambia [ 1 , 7 ]. As early as 1980s, the African countries had embraced and implemented the WHO recommended DOTS strategy, which was meant to standardize methods for case detection, case management and monitoring [ 8 ]. DOTS as a strategy, entails that medication is taken in while the care provider is observing the patient swallowing the drug [ 9 ]. Diagnosing TB patients at a health facility and sending them to continue taking their treatment at home, poses serious challenges for other members of the households who assist the patients with treatment. Work carried out in Haiti [ 7 ], Zambia [ 10 ] and most other rural communities [ 11 ] present the challenges facing compliance with TB treatment. Tuberculosis like HIV/AIDS is often associated with stigmatization and thus may create resistance among patients to treatment. A study carried out in Nigeria [ 12 ], raised an important point of delays in care seeking behaviour due to stigma experienced by TB patients. Studies [ 13 ] have shown that stigmatization creates a lot of self-denial among those with diseases like TB and Sexually Transmitted Infections (STIs); hence most of them fail to comply with the treatment regime. Issues of diverse nature to caring of patients at home are rampant in communities that have low literacy levels [ 13 , 14 ]. The possibility of being exposed to tuberculosis by merely associating with patients may create some resentment by those in the household to providing proper care of the patient and encourage non-compliance [ 15 , 16 ]. In their studies [ 4 , 15 , 18 ], researchers found that fear of catching the disease was a factor in household members' negative reactions to care of the TB patient. The high incidence of chronic TB-related problems among the patients increases the probability that care of TB patients becomes costly at home, in terms of food provision and other essentials [ 18 , 19 ]. While TB treatment lasts between 6 and 8 months, the effective treatment of multidrug resistant may be more pronounced in low-income communities and can even be more complex if not properly supervised. Discrimination on the basis of disease at health facilities sometimes exacerbates problems with compliance with tuberculosis drug taking behaviour [ 19 ]. Unless there is privacy during drug collection schedules, patients may resist going to collect their drug supplies at the clinic because of discriminatory behavior by health care providers. Some studies that have been carried out in Africa and other parts of the world [ 15 , 19 ] found that health care providers discriminated against AIDS patients. The type of language used at both the health facility and the homes has a strong bearing on the reactions patients may have and their compliance with treatment. Where TB is perceived to be AIDS, a disease that is frightening and a lurking cause of premature death, those associated with it may sometimes withdraw themselves from the rest of the community because, they believe they have reached the end of their lives. More often than not, when patients are diagnosed as having TB, communities immediately construct them as social and sexual misfits in the society, which is often followed by exclusion from social interactions and relationships [ 20 ]. In homes where care is given, a patient may experience feelings intense loneliness and abandonment. A study among the Nigerian medical students [ 21 ] revealed that the majority of the students considered AID a divine punishment for sexual excess. In order to try and improve community participation, organisations such as WHO have developed strategies to enhance TB/HIV/AIDS control [ 22 ]. This paper presents factors contributing to treatment adherence and knowledge of TB transmission among patients on TB treatment. Methods Study design and population A cross sectional study was conducted among tuberculosis patients who were getting their drugs from the clinics operational in Ndola, Zambia by the Ndola District Health Management Board. In this study, compliance refers to patients who took their TB drugs daily for 8 months. On the other hand, any patient who stopped taking TB drugs during the treatment period was regarded as non-compliant patient. This study took into account an analytical approach of comparing two groups (the compliant against the non-compliant group). The study was limited to these patients because of the research focus on compliance of patients over time to prescribed TB drug regimens identified from medical records. In addition to medical records, TB patients were also directly asked if they had ever stopped taking the drugs since starting the treatment. Sample Using the medical records from the six sampled Ndola District Health Management Board Clinics, a list of all TB patients seeking treatment within the six months period was drawn. A total of 736 patients were listed from the six study sites. Using this figure, and a prevalence of 50% +_5% at 95% confidence level, the minimum sample size was determined to be 248. Although 400 TB patients were recruited in the study, 18 were excluded from the analysis because only demographic data was obtained. Reasons for incompleting questionnaires included some patients becoming restless and rushed to the hospital, while others were too sick to continue with the interview and had died by the re-appointment date. TB patients were classified as follows: those who never attended the TB-clinic after diagnosis, those who attended 1–2 times after diagnosis and commencement on treatment, and those who attended regularly (i.e. 3 or more times) after diagnosis and continuation of treatment. Patients who attended the clinic regularly and continued taking the treatment were classified as controls, and the rest of the patients were classified as cases. Out of 18 clinics that were run by the District Health Management Board in Ndola, six were randomly selected from the list of clinics so as to enable the team obtain a representative sample of the clinics in the district. In the sampled clinic catchment areas, a household-based survey was conducted and all patients that met the criteria were invited to take part in the study. In all, 114 non-compliant and 268 treatment-compliant TB patients participated in the study. Ethical considerations TB is strongly associated with HIV/AIDS because of the perceived mode of transmission, which stigmatises and discriminates patients. These are common in most African communities [ 24 ] and Zambia in particular [ 25 ]. TB like HIV/AIDS becomes difficult to discuss in public. Looking at the sensitive nature of the study, all patients were assured of confidentiality and anonymity. Information on patients, residential addresses, and health facilities to which patients were affiliated were collected for the purpose of follow-up. Respondents were informed that this information would not be made available to persons outside the study team. Respondents were further assured that no person-identifiers would be used for publication. Data collection, management and analysis Those who had agreed to participate were interviewed using a pre-tested structured survey instrument. Data collected were individual socio-demographic characteristics and economic characteristics, knowledge on TB transmission, prevention and treatment practices, patterns in health seeking behaviour, compliance and non-compliance to TB drug taking, and perceived community support during the drug taking period. The data was entered in EPI6. It was edited using consistency and range checks after data entry. However, data analysis was done in the statistical package, SPSS. The Chi-squared Yates corrected test for 2 by 2 tables and Pearson's uncorrected Chi-squared test for higher contingency tables were used to determine associations between qualitative factors. A multivariate logistic regression model was used to adjust for differences in the distributions of age, marital status and education between males and females in the relationships between knowledge items and gender. The cut off point for statistical significance was set at the 5% level. Results The demographic characteristics of household-based patient population that participated in the survey are presented in Table 1 . Information on gender was not recorded for two respondents. Most of the male patient respondents (44.0%) were in the age group 30–39 years, while the majority of female patient respondents (38.2%) were of age 20–29 years (p = 0.001). Among women, TB patients were 2.04 times more likely to be in the age group 10–19 years, compared to males in the same age group. Significantly more females than males were either divorced/separated or widowed (p < 0.001). In terms of educational attainment, males tended to be more educated than females (p < 0.001). Table 1 Socio-demographic characteristics of Patients by Gender Total Male Female Demographic characteristics n % N % n % AGE [p = 0.001] (years) Total = 379 Total = 193 Total = 186 < 10 22 5.8 10 5.2 12 6.5 10–19 20 5.3 5 2.5 15 8.1 20–29 116 30.6 45 23.2 71 38.2 30–39 145 38.3 85 44.0 60 32.3 40–49 47 12.4 31 16.1 16 8.6 50+ 29 7.7 17 8.8 12 6.5 MARITAL STATUS [p < 0.001] Total = 379 Total = 194 Total = 185 Married 168 44.3 108 55.7 60 32.4 Single 77 20.3 43 22.2 34 18.4 Divorce/Separated 69 18.2 25 12.9 44 23.8 Widowed 65 17.2 18 9.3 47 25.4 EDUCATION [p < 0.001] (Years in school) Total = 342 Total = 178 Total = 164 0–4 52 15.2 21 11.8 31 18.9 5–7 146 42.7 60 33.7 86 52.4 8–9 73 21.3 42 23.6 31 18.9 10+ 71 20.8 55 30.9 16 9.8 Table 2 shows only one significant difference in knowledge of TB transmission by gender. Significantly more males than females stated that sharing of cups as a means of transmitting TB (27.3% of males and 17.2% of females; p = 0.025). Males were still more likely to hold this belief even after adjusting for age, marital status and education. Table 2 Knowledge on TB Transmission by Gender Male Female Total = 194 Total = 186 Sources of transmission n % n % p value OR (95% CI)* Through sexual intercourse 24 12.4 25 13.4 0.875 0.91 (0.62, 1.31) From mother to child 71 36.6 77 41.4 0.393 0.93 (0.73, 1.19) Sleeping in same room with TB patient 10 5.2 7 3.8 0.684 1.03 (0.58, 1.82) Sharing cups 53 27.3 32 17.2 0.025 1.66 (1.23, 2.26) Patient coughing directly at others 15 7.7 21 11.3 0.313 0.65 (0.42, 1.02) * OR (95%CI) [Odds Ratio and 95% confidence interval] adjusted for age, marital status and education No significant differences by duration of residence in the community were observed in knowledge of TB transmission through sexual intercourse, sleeping in the same room with TB patient and patient coughing directly at others (Table 3 ). Significantly (p = 0.016) more patients with duration of stay of less than two years in the community (59.3%) reported that TB could be transmitted from mother to child compared to 41.2% of patients who had stayed for two or more years in the compounds. Furthermore, significantly (p = 0.019) more patients who had stayed for at least two years in the compounds (27.4%) stated that TB could be transmitted through sharing of cups compared to 11.9% of patients who had stayed in the compounds for less than two years. Table 3 Associations between knowledge of TB transmission and duration of stay in the compound Duration of stay in compound Knowledge of TB transmission <2 years 2+ years p value Total = 59 Total = 277 n % n % Through sexual intercourse 9 15.3 40 14.4 0.966 From mother to child 35 59.3 114 41.2 0.016 Sleeping in the same room with TB patient 3 5.1 13 4.7 1.000 Sharing cups 7 11.9 76 27.4 0.019 Patient coughing directly at others 5 8.5 30 10.8 0.762 A total of 114 (29.8%) out of 382 patients stopped taking TB drugs at some point during the treatment regimen. Analyses for the factors associated with TB drug compliance revealed that sex, education, marital status, sharing a room with any one else, relationship with head of household, anyone suffered from TB in the house, close relative or friend that has suffered from TB, number of times suffered from TB, and use of traditional healers for treatment of TB were not significantly associated with compliance. The common reason given for stopping treatment by both the compliant and the non-compliant patients were that they could not continue with the medication when they started feeling well (45.1% and 38.6%), respectively. Meanwhile, other reasons given by complaint patients were lack of knowledge on the benefits of completing TB course (25.7%), TB drugs too strong (20.1%) and lack of food in the home (11.4%). Similarly, the non-compliant patients mentioned running out of drugs at home (25.4%), TB drugs too strong (20.2%) and loss of hope to live (16.7%) as reasons for stopping. Common reasons given for stopping treatment for male patients were that they ran out of drugs at home (29.1%) or had no food (23.6%). Meanwhile, most females reported that forgetting to take the medicine (32.0%), reaction to drugs (20.0%), and running out of drugs at home (20.0%) as the reasons for stopping taking drugs (Table 4 ). Table 4 Perceived Reasons Given by Compliant and Non-compliant Patients Leading to Stoppage of TB Drug Taking Reason Compliant Non-compliant n = 268 % n = 114 % Once they start feeling better 121 45.1 44 38.6 Lack of knowledge on the benefits of completing a course 69 25.7 14 12.3 Running out of drugs at home 15 5.6 29 25.4 TB drugs too strong to continue 54 20.1 23 20.2 Lack of food 40 14.9 13 11.4 Loss of hope to live 30 11.2 19 16.7 Lack of drugs at the clinic 5 1.9 5 4.4 Denial of suffering from TB 3 1.1 6 5.3 Doctors advice 0 0.0 2 2.0 Figures cannot add up because the question allowed for multiple responses Results from the qualitative study reveal that 10 out of 17 non-compliant patients stressed running out of drugs as a reason for stopping. A young man who was a widower aged 32 years reported: Okay, the drugs got finished but like I am not feeling well, that is why I don't go to collect, because here, there are no children you send. My nephew also has started what? The one who used to get it for me [referring to the nephew who had started work] In another in-depth interview with a single non-compliant female patient aged 31 years, emphasised on running out of drugs at home and reported thus: That one we drink two tablets, I..., we went on Monday, they said the Doctor was not there [referring to the clinical officer or nurse responsible for running the TB clinic]... he comes on Wednesday, so I have missed also. The key determinant for stopping medication in both qualitative and quantitative was running out of drugs. No significant gender differences existed at the stage at which most TB patients stopped taking drugs (p = 0.743). About a third of the respondents (33.9% males and 39.1% females) indicated that they stopped taking drugs within the first two months of starting treatment. However, by the 5 th month of consistently taking medication, more men tended to stop taking their medicine (Figure 1 ). Most of the patients received the drugs from health facilities which were normally very close to their residences. Easy access to drugs created one more problem; where individuals failed to follow the drug schedules; they moved to the next health facility in the other neighbourhood and started their medication. In the new place, they gave false addresses as well as different names from those they were known in their neighbourhood. These are interesting findings that highlight why compliance continues to be difficult in most of the communities. Figure 1 Months since starting treatment at which most TB patients stopped taking drugs Male Female Discussion The Ndola study showed that 29.8% of all TB patients on treatment did not comply. The majority of patients who participated in the study were between 20 and 39 years. This population is in agreement with other studies that had been conducted in Ndola [ 4 , 5 , 24 ]. It was striking to note that females in the age group 10–19 years were more likely to suffer from TB than their male counterparts in the same age category. Interestingly, this result is consistent with findings from some HIV/AIDS studies conducted in Zambia, which showed higher prevalence of HIV among females in this age group [ 1 , 3 ]. In terms of marital status, more males were identified as divorced/separated or widowed. The educational profile in this study was identical to that of the general population in Zambia [ 17 , 21 , 27 ]. Most patients had attended primary level of education, with more male patients in the majority. However, there was no relationship between the individuals' level of education and TB infection rates. TB is a serious public health issue that needs continued attention. It has been demonstrated in other studies [ 22 , 26 ] that the TB epidemic may have changed the population's economic needs, resulting in forcing more young men and women to be exposed to situations that place them at higher risk for TB transmission, for instance, sharing overcrowded rooms even when the situation does exacerbate the risk of TB transmission. Knowledge about TB transmission No relationship has been found between marital status of an individual and the knowledge of TB. Lack of knowledge makes TB a serious public health problem, which needed an urgent attention. TB patients in Ndola could have continued sharing a room with an infected person, because of lack of accomodation, high rental costs and inadequate knowledge on TB transmission. Over 70% of TB patients easily mentioned the symptoms which included sweating at night, loss of weight, loss of appetite and prolonged coughing, which at times was accompanied by sign of blood in the sputum. Knowledge about the symptoms of the disease did not vary according to levels of education or age. Knowledge modes of transmission and methods of prevention required revisiting, as community members seemed to have had knowledge that did not relate the disease to the environment (that TB could be transmitted through the air). They believed in physical contacts with objects, for instance, sharing of cups, having sexual intercourse with TB patient and from mother to child. This result is consistent with findings from a study in Botswana [ 19 ], which reported that TB was a result of pollution from breaking taboos. It is not surprising therefore, to find no gender differences in terms of TB knowledge transmission among the age groups. The low knowledge in preventive measures in this group could be explained through our study sample itself. Although these age groups were of patients who could have received the information from the clinic during the time they were collecting drugs and their food supplies, which were often provided to patients, prevention did not seem to have been stressed by health care providers. The other minor reason could be that our questions were prompted; this could also explain why 13.3% reported that to prevent TB one should always keep to one partner. The care seeking behaviour remained the same as in other studies [ 25 , 26 ], where the majority of patients went to specialised health facilities once they were sick. Duration of residence The patients' duration of residence in the city of Ndola compared very well with another study [ 26 ], which found the median period for most women to be 10 years. It has been shown in this study that duration of residence was associated with the knowledge of tuberculosis transmission. However, the study had identified some misconceptions on TB transmission, about a third of the respondents reported having got the diseases through sexual intercourse. This study suggested a relationship between community constructions of TB and HIV/AIDS campaigns. Patients could have related TB with HIV and selected sexual intercourse as a source of transmission for the disease, and therefore considered condom use as a means for preventing it. One explanation for this misconception could be attributed to the overstressing that TB is an opportunistic disease for HIV/AIDS. The study in Haiti [ 9 ], clearly showed that patients who were diagnosed as TB patients ended up being HIV positive. This misconception was critical in changing people's behaviour and therefore required an urgent intervention. Non-compliance Non-compliance to TB drug taking is reasonably lower than we originally thought. There were 29.8% of the patients who stopped taking their medicine at two months after commencing treatment. The defaulting progressed in the second month (16.8%), and then stabilized thereafter. Factors associated with defaulting in the current study, included patients beginning to feel better, lack of knowledge on the benefits of completing a course, running out of drugs at home and TB drugs being too strong to continue. However, the major and stricking determinant of non-complaince was the patient beginning to feel better. While lack of food and other provisions were identified in other studies [ 17 ], as factors associated with defaulting, in this study lack of food did not come out as an issue. This could be due to the fact that the DHMB through the health facilities distributes food portions (herps, mealie meal, beans and cooking oil) as complimentary food availability at household level for TB patients. In the study at Lusaka University Teaching Hospital [ 28 ], it was found that over 45% of the TB patients could not comply with drug treatemnt instructions for various reasons, including poor transportation to the hospital and lack of family support. Over 79% of the patients, had suffered from TB only once, while the rest had suffered more than twice. Results present a wide range in the percentage of opposing views for TB as an air borne disease. Within age groups, there was usually no relationship between the documented preventive measures against TB and air borne germs. The reasons were attributed to the fact that these were patients obtaining their drugs from a health facility, hence, received some of the information on disease transmission. The relationship between the awareness 'coughing blood' and 'TB' was explored during the health talks. Conclusion The findings in the Ndola study showed that approximately 39.8% of all TB patients on treatment did not adhere to their treatment schedules, when they started feeling better. Competing interests The author(s) declare that they have no competing interests. Authors contributions FADK was the principal investigator of the study and was responsible for the design, implementation and supervised data entry and cleaning. He worked closely with the biostatistician during data analysis. He is the principal author of this paper. MT was the co-investigator of the study. She contributed to the design of the study, coordinated data collection, entry and cleaning. She was part of the data analysis team. SS is a co-author of this puplication who carried out data analysis as a biostatistician. LS co-authored this publication and was responsible for data entry and cleaning. He was part of the data analysis team. All authors have read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545081.xml
514543	Assessment of genetic diversity in Trigonella foenum-graecum and Trigonella caerulea using ISSR and RAPD markers	Background Various species of genus Trigonella are important from medical and culinary aspect. Among these, Trigonella foenum-graecum is commonly grown as a vegetable. This anti-diabetic herb can lower blood glucose and cholesterol levels. Another species, Trigonella caerulea is used as food in the form of young seedlings. This herb is also used in cheese making. However, little is known about the genetic variation present in these species. In this report we describe the use of ISSR and RAPD markers to study genetic diversity in both, Trigonella foenum-graecum and Trigonella caerulea . Results Seventeen accessions of Trigonella foenum-graecum and nine accessions of Trigonella caerulea representing various countries were analyzed using ISSR and RAPD markers. Genetic diversity parameters (average number of alleles per polymorphic locus, percent polymorphism, average heterozygosity and marker index) were calculated for ISSR, RAPD and ISSR+RAPD approaches in both the species. Dendrograms were constructed using UPGMA algorithm based on the similarity index values for both Trigonella foenum-graecum and Trigonella caerulea. The UPGMA analysis showed that plants from different geographical regions were distributed in different groups in both the species. In Trigonella foenum-graecum accessions from Pakistan and Afghanistan were grouped together in one cluster but accessions from India and Nepal were grouped together in another cluster. However, in both the species accessions from Turkey did not group together and fell in different clusters. Conclusions Based on genetic similarity indices, higher diversity was observed in Trigonella caerulea as compared to Trigonella foenum-graecum . The genetic similarity matrices generated by ISSR and RAPD markers in both species were highly correlated (r = 0.78 at p = 0.001 for Trigonella foenum-graecum and r = 0.98 at p = 0.001 for Trigonella caerulea ) indicating congruence between these two systems. Implications of these observations in the analysis of genetic diversity and in supporting the possible Center of Origin and/or Diversity for Trigonella are discussed.	Background The family Fabaceae includes many crops useful for food, forage, fiber, wood and ornamental purposes. In this family, a few legumes such as chickpea, soybean, fababean, fenugreek, lentil, pea etc. are consumed as grain legumes. The grain legumes are plants used as food in the form of unripe pods, mature seeds or immature dry seeds, directly or indirectly [ 1 ]. The grain legumes not only provide variety to human diet but they also supply dietary proteins for vegetarian populations that lack animal and fish protein in their diet. Considering the growing problem of malnutrition, use of legume species as high-protein food is very important. Moreover, legumes are also capable of symbiotic nitrogen fixation enriching the soil condition suitable for a crop following the legume crop [ 2 ]. The genus Trigonella is one of the largest genera of the tribe Trifoliatae in the family Fabaceae and sub-family Papilionaceae [ 3 ]. Among Trigonella species, Trigonella foenum-graecum (commonly known as fenugreek) is a flowering annual, with autogamous white flowers occasionally visited by insects. Indigenous to countries on the eastern shores of Mediterranean, fenugreek is widely cultivated in India, Egypt, Ethiopia, Morocco and occasionally in England [ 4 ]. Trigonella foenum-graecum is extensively grown in the tropical and subtropical regions of India. Different parts of the plant such as leaves and seeds are consumed in India. It is also used for medicinal purpose. According to ancient medicinal system, the Ayurveda, it is a herbal drugs that is bitter or pungent in taste. It is effective against anorexia and is a gastric stimulant [ 5 ]. Fenugreek is becoming popular around the world with its extract used to flavor cheese in Switzerland, artificial maple syrup and bitter-run in Germany, roasted seeds as coffee-substitute in Africa, seed powder mixed with flour as fortification to make flat-bread in Egypt, as an anti-diabetic herb in Israel, whole seed and dried plant used as insect and pest repellent in grain storage, and oil used in perfumery in France [ 6 ]. Research reports in recent years have indicated that fenugreek can be a remedy to diabetes by lowering blood sugar and cholesterol levels [ 7 ]. T. caerulae , from the same family and commonly known as the Blue fenugreek, on the other hand, is a less commonly grown herb. This flowering annual with autogamous blue flower is found in Alps and in the mountains of eastern and south eastern Europe. Terminal leaves are mainly used for cooking while young seedlings are eaten with oil and salt. Dried and powdered leaves as well as flowers are used for flavoring and coloring bread, cheese, etc in China and Germany. They are also used as a condiment in soups and potato dishes and a decoction of leaves is used as aromatic tea [ 8 ]. Grain legumes like T. foenum-graecum and T. caerulea although important in food and medicine are rarely grown outside their native habitat. Across the world only known and well-defined cultivars are grown in specific areas. Gene banks also harbor scanty germplasm collection of Trigonella species [ 9 ]. The neglected and the under-use status of these locally important crops indicates a risk of disappearance of important plant material developed over thousands of years of cultivation. One of the important factors restricting their large-scale production and development of better varieties is that very little information is available about their genetic diversity, inter and intraspecific variability and genetic relationship among these species. Therefore, attempts to analyze possible untapped genetic diversity become extremely essential for breeding and crop improvement. The present study was undertaken with the objective of analyzing genetic diversity in various accessions of T. foenum-graecum and T. caerulea representing various countries where they are grown using molecular marker technology. Results Assessment of genetic diversity in T. foenum-graecum and T. caerulea using ISSR and RAPD markers A set of 100 ISSR primers was used for initial screening of 7 accessions of T. foenum-graecum of which 40 gave amplification. However, only 14 ISSR primers detected intraspecific variation in I7 accessions of T. foenum-graecum generating clear reproducible patterns and revealing 100 bands in the range of 500 bp to 2 kb. Among these seventy-two bands were polymorphic amounting to 72% polymorphism [Table 3 ]. Furthermore, during ISSR analysis 11 unique bands were obtained, where 6 were contributed by accession TMP = 8714 from Yemen, 4 by accession TMP = 8691 from Turkey, and 1 by accession TMP = 8685 from Iran. Similarly, in T. caerulea , of the 100 ISSR primers used for initial screening, 47 gave amplification. Of these, 18 primers detected intraspecific variation in 9 accessions of T. caerulea showing 93.64% polymorphism [Table 4 ]. With these 18 primers, 16 unique bands were produced, where 9 and 7 bands were contributed by accessions 206901 and 206486, respectively, from Turkey alone. In case of RAPD analysis, 100 RAPD primers were used for initial screening in T. foenum-graecum of which 22 primers generated polymorphic patterns revealing 70.12% polymorphism [Table 3 ]. Eight unique bands were produced by these primers of which 3 were contributed by accession TMP = 8714 from Yemen, 2 by accession TMP = 8698 from Egypt, 1 by accession TMP = 8707 from Afghanistan and 1 each by accession TMP = 8691 and TMP = 8690 from Turkey. Similarly in T. caerulea of the 40 primers used for initial screening, 10 primers produced polymorphic pattern giving 95.83% polymorphism [Table 4 ]. Eight unique bands were produced with these primers wherein; the maximum numbers of unique bands (4 each) were again produced by the same accessions 206901 and 206486 from Turkey. Dendrogram analysis for T. foenum-graecum and T. caerulea Genetic similarity was calculated from the Nei's similarity index value for all the 17 accessions of T. foenum-graecum considering ISSR and RAPD approaches individually as well as together. Based on ISSR markers alone, the similarity index values ranged from 0.69 to 0.92. These values were used to construct a dendrogram using Unweighted Pair Group Method with Arithmetic averages (UPGMA). In the ISSR based dendrogram T. foenum-graecum genotypes formed 4 clusters [Figure not shown]. The first cluster grouped together accessions from Afghanistan (8707), Canada (1065), Pakistan (8717,8718), Iran (8675), and Turkey (8690). The second cluster contained accessions from India (8686,8689,8675). Remaining one accession from India (8687) and one from Nepal (8706) formed the third cluster. The fourth cluster contained accessions from Egypt (8698,8679) and Turkey (8692). Accessions from Ethiopia (8696), Turkey (8691) and Yemen (8714) out grouped from the main clusters. Three accessions from Turkey analyzed in the present study fell into different clusters. Based on RAPD markers alone, the similarity index values ranged from 0.71 to 0.91. In the RAPD based dendrogram, T. foenum-graecum genotypes formed 2 main clusters [Figure not shown]. The first cluster had two subgroups, the first subgroup contained accessions from Afghanistan (8707), India (8686,8687,8675), Turkey (8690) and Egypt (8698) while the second subgroup contained accessions from Pakistan (8717,8718), Turkey (8692), and Egypt (8679). The accession from Canada associated with these clusters. The second cluster contained accession from Iran (8685) and Nepal (8706) and the accession from India associated with this cluster. Accessions from Ethiopia (8696), Turkey (8691) and Yemen (8714) out grouped from these two clusters. In the RAPD based dendrogram also accessions from Turkey fell into different clusters. Based on both the marker systems together the similarity index values ranged from 0.65 to 0.89 [Fig. 1 ]. Here the T. foenum-graecum accessions from Egypt (8698,8679) were grouped together. Accessions from Pakistan (8717, 8718) and Afghanistan (8707) were grouped together in one cluster. Accessions from India (8686,8689,8675,8687), Nepal (8706) and Iran (8685) were grouped together. However, all the three accessions from Turkey fell in different clusters and did not group among themselves. Bootstrapping was done using the WinBoot program to estimate the relative support at clades. The robustness of the cluster was not very strong in T. foenum-graecum (50–70%). In T. caerulea , genetic similarity was calculated from the Nei's similarity index value considering ISSR and RAPD approaches individually as well as together. Based on ISSR marker system, the similarity index values ranged from 0.41 to 0.92 while that on the basis of RAPD markers ranged from 0.34 to 0.93. Similarity indices values based on both the marker systems together ranged from 0.38 to 0.92 indicating more diversity in case of T. caerulea . The dendrograms based on ISSR and RAPD markers showed similar clustering pattern when used individually [Figures not shown] as well as together [Fig. 2 ]. Here the dendrogram showed two main clusters. The first cluster grouped together accessions form Turkey (TMP = 8704) and Australia (PI = 186283). The second cluster contained 3 accessions of unknown geographic origin and 1 accession each from U.S.A (PI = 345743) and Spain (PI = 244288). The remaining 2 accessions i.e. accessions PI = 206486 and PI = 206901 from Turkey, out grouped from these two main clusters. Thus the accessions from Turkey did not cluster together as observed in T. foenum-graecum . The bootstrap values obtained using the WinBoot program in T. caerulea were very high (90–100%). Heterozygosity and marker index Heterozygosity was calculated using ISSR and RAPD marker systems individually as well as together as detailed in Table 3 for T. foenum-graecum and in Table 4 for T. caerulea . In T. foenum-graecum the H av values calculated for ISSR, RAPD and ISSR+RAPD were 0.214, 0.203 and 0.211, respectively. In case of T. caerulea the H av values for the same were 0.330, 0.346 and 0.338, respectively. For both, T. foenum-graecum and T. caerulea , the H av values for ISSR and RAPD systems did not differ significantly. The marker index (MI) values calculated for ISSR and RAPD for T. foenum-graecum were 1.53 and 1.51, respectively while in T. caerulea they were 3.17 and 3.35, respectively. Thus MI values in both the species did not differ significantly for ISSR and RAPD systems. Correlation between measures of similarity In T. foenum-graecum , when the similarity matrices generated using ISSR and RAPD markers were compared, a value of r = 0.78, at P = 0.001 indicated a good correlation between data generated by both the systems [Fig. 3 ]. Similarly when the similarity matrices generated using ISSR and RAPD systems were compared in case of T. caerulea , a value of r = 0.98 indicated a very good correlation between the two marker systems [Fig. 4 ]. Discussion The two marker systems, ISSR and RAPD used in the present study have also been used as effective tools to evaluate genetic diversity and to throw light on the phylogenetic relationships in Brassica napi [rapeseed, [ 10 ]], Allium sect. Sacculiferum [Alliaceae, [ 11 ]] and Asimina triloba [pawpaw, [ 12 ] and [ 13 ]]. Genetic diversity analysis using ISSR and RAPD markers in legumes like Cicer [[ 14 ] and [ 15 ]] and Cajanus [[ 17 ] and [ 18 ]] has been carried out in our own laboratory. These studies have given important clues in understanding species relationship, which may further assist in developing and planning breeding strategies. However, no such reports on genetic diversity using molecular markers were available in the genus Trigonella . In the present study, an attempt has been made to examine the level of genetic variation within T. foenum-graecum and T. caerulea accessions obtained from germplasm collection centers at Saskatoon (Plant Gene Resources of Canada) and Pullman (USDA-ARS Plant Introduction Station, Washington). The T. foenum-graecum accessions were selected in order to represent most of the countries where it is grown. In case of T. caerulea all the nine accessions available at Saskatoon and Pullman were used in present study. Analysis of polymorphism detected in T. foenum-graecum and T. caerulea Polymorphism in a given population is often due to existence of genetic variants represented by the number of alleles at a locus and their frequency of distribution in a population. Heterozygosity corresponds to a probability that two alleles taken at random from a population can be distinguished using the marker in question. Thus a convenient quantitative estimate of marker utility and the polymorphism detected can be given in terms of the mean heterozygosity and the marker index [ 18 ]. In T. foenum-graecum as well as in T. caerulea the H av and the marker index (MI) values for ISSR and RAPD markers [Table 3 and 4 ], respectively, did not differ significantly indicating that similar levels of polymorphism were detected by both the marker systems in the given germplasm pools. This was also confirmed by the high correlation co-efficent for ISSR and RAPD marker systems obtained for T. foenum-graecum and T. caerulea [Fig. 3 and 4 ]. Genetic diversity as measured by the heterozygosity was higher in T. caerulea (0.33) as compared to T. foenum-graecum (0.21). Based on allozyme diversity, the estimated mean heterozygosity values have been reported for self-pollinating species, Vigna unguiculata , H av = 0.027 [ 19 ] and Vicia tetrasperma , H av = 0.342 [ 20 ]. The heterozygosity value for Vigna unguiculata was lower while that for Vicia tetrasperma was higher as compared to T. foenum-graecum and T. caerulea . Based on ISSR markers, the estimates of genetic diversity, H av = 0.358 reported in cultivated pawpaw ( Asimia triloba ), was higher as compared to T. foenum-graecum and T. caerulea [ 13 ]. Estimation of genetic relatedness in T. foenum-graecum and T. caerulea Data collected with ISSR and RAPD marker systems were used to compare genetic similarity between various accessions of T. foenum-graecum and T. caerulea . The accessions could be a mixture of different genotypes. Therefore, to have a complete representation of a specific accession, DNAs from fifteen plants were mixed in equal proportion. Thus within accession diversity was eliminated and a complete banding profile of the accession was used for the analysis. In T. foenum-graecum , ISSR and RAPD could detect almost similar level of polymorphism (72% with ISSR and 70.12% with RAPD). In the UPGMA analysis, T. foenum-graecum accessions from one country and the nearby region grouped together in some cases while they were placed in different clusters in certain cases. Accessions from Pakistan and Afghanistan grouped together in one cluster while accessions from India and Nepal grouped in another cluster. Moreover, the three accessions from Turkey fell in different clusters inspite of being geographically very close to each other. Thus, there was no clear clustering pattern of geographically closer accessions in the present study indicating that the association between genetic similarity and geographical distance was less significant. However, it is necessary to use more number of accessions from each geographical location to confirm the available pattern. In T. caerulea also, ISSR (93.64%) and RAPDs (94.83%) detected almost equal level of polymorphism. T. caerulea showed more polymorphism as compared to T. foenum-graecum . In case of T. caerulea also, the UPGMA analysis showed that plants from different geographical regions were distributed in different groups. Here again, the accessions from Turkey were not grouped together. Two accessions from Turkey out grouped from the main cluster while one was grouped in the first cluster with Australia. In T. caerulea we could obtain three accessions from Turkey and only one accession from other countries. Therefore, it would be premature to comment precisely about the correlation of geographic distance and genetic diversity in this case. To confirm the available pattern it is necessary to use more number of accessions from each geographical location. Center of Origin and /or Center of Diversity for Trigonella The place of origin of a species as explained by Vavilov is an area that contains a large amount of genetic variability of that species. According to him variation is a function of time, hence the region containing the greatest variation in a species would have supported and sustained that species for a longer time than the other regions suggesting that region to be the Center of Origin and/or Diversity. He set up eight geographic centers, two of which namely the Near Eastern and Mediterranean Centers extent into Turkey [ 21 ]. It has been postulated by Vavilov that the Near East region extending from Israel through Syria, Southern Turkey into Iran and Iraq and the Mediterranean Center including Spain, Morocco and Turkey is the Center of Origin of Trigonella , Trifolium and Medicago species [ 22 ]. In the present study both, T. foenum-graecum and T. caerulea accessions from Turkey exhibited more diversity. These results support Vavilov's hypothesis. The Indian accessions of T. foenum-graecum i.e. accession number 8686 (Khandwa), 8685(Mumbai), and 8687 (Patiala) separated by an aerial distance of 52 km, 104 km, and 135 km, respectively from each other, were genetically more similar (similarity index 0.893) and clustered together [Fig. 1 ]. However, the accessions of T. foenum-graecum from Turkey i.e. accession number 8692 (Malatya) and 8691 (Elbistan) separated by a distance of 100 km (similarity index value 0.875-0.745) were out grouped and were genetically more distant from each other although morphologically they were similar to each other as well to the accessions from other countries. Turkey is one of the significant and unique countries in the world from the point of view of plant genetic resources and plant diversity. The country has more than 3,000 endemic plants and immense diversity has been reported in many legumes such as Vicia , Medicago , Trifolium , Lathyrus , Onbrychis , Trigonella , Pisum , Cicer , Lens , etc [ 23 ]. Many genera of cultivated plants like Cicer , Lens , Pisum , Amygladus , Prunus , Triticum , etc have their Center of Origin and/or Diversity in this country [ 22 ]. Vavilov designated southeastern Turkey and the adjoining Syria as the primary Center of Origin (now the center of diversity) for chickpea [ 24 ]. Similar to chickpea (and other grain legumes also), in T. foenum-graecum the large seeded cultivars are abounded around the Mediterranean region whereas the small seeded cultivars are predominated eastwards. Thus, Turkey may also be the primary Center of Origin of T. foenum-graecum and T. caerulea . However, this hypothesis needs to be confirmed by considering more accessions distributed over a wide geographic range especially from the Near East and the Central Mediterranean region. Conclusions In conclusion, molecular markers allowed us to estimate the overall genetic diversity in T. foenum-graecum and T. caerulea and simultaneously revealed molecular based genetic relationship. In the UPGMA analysis, no significant correlation was observed between geographic distance and genetic diversity. Our data further supported the hypothesis of the Near East and the Central Mediterranean to be the Center of Origin and/or Diversity for Trigonella as put forth by Vavilov. Methods Plant material and DNA extraction Seeds of T. foenum-graecum accessions were obtained from Plant Gene Resources of Canada (PGRC), Saskatoon, Canada. These accessions along with the TMP numbers and the country from where they have been collected are outlined in Table 1 . Seeds for various T. caerulea accessions were obtained from PGRC, Saskatoon, Canada and USDA-ARS Plant Introduction Station at Pullman, Washington (W-6), and are detailed in Table 2 . Fifteen plants of each accession were grown in pots for DNA isolation. Two gram of young leaf tissue was harvested for each plant and frozen in liquid nitrogen for DNA extraction. Plant DNA was extracted by the Doyle and Doyle method [ 25 ] and equal amount of DNA from each of the fifteen plants was pooled together for each accession. PCR amplification ISSR A set of 100 anchored micro satellite primers was procured from University of British Columbia, Canada. PCR amplification of 20 ng of DNA was performed in 10 mM Tris-HCI pH 7.5, 50 mM KCI, 1.5 mM MgCl2, 0.5 mM spermidine, 2% formamide, 0.1 mM dNTPs, 0.3 uM primer and 0.8 U of Taq DNA polymerase (Ampli-Taq DNA polymerase, Perkin Elemer, USA) in a 25 ul reaction using Perkin Elmer 9700 thermocycler. After initial denaturation at 94°C for 5 minutes, each cycle consisted of 30 seconds denaturation at 94°C, 45 seconds of annealing at 50°C, 2 minutes extension at 72°C along with 5 minutes extension at 72°C at the end of 45 cycles. RAPD RAPD analysis was performed using arbitary decamer primers procured from University of British Columbia, Canada. The reaction mixture (25 ul) contained 10 mM Tris-HCI pH 7.5, 50 mM KCI, 1.5 mM MgCl2, 0.5 mM spermidine, 0.1 mM dNTPs, 15 pmoles of primer, 20 ng genomic DNA, and 0.8 U of Taq DNA polymerase (Ampli-Taq DNA polymerase, Perkin Elmer, USA). Amplification was carried out using Perkin Elmer 9700 thermocycle for 40 cycles, each consisting of a denaturing step of 1 minute at 94°C, followed by annealing step of 1 minute at 36°C and an extension step of 2 minutes at 72°C. The last cycle was followed by 5 minutes of extension at 72°C. Amplified products were electrophoresed on 2% agarose gel using 0.5× TAE buffer (10 mM Tris HCl and 1 mM EDTA pH. 8.0) and visualized by ethidium bromide staining. The patterns were photographed and stored as digital pictures in gel documentation system. The reproducibility of the amplification was confirmed by repeating each experiment three times. Agarose gel electrophoresis Amplified products were electrophoresed on 2% agarose gel using 0.5x TAE buffer (10mM Tris HCl and 1mM EDTA pH. 8.0) and visualized by ethidium bromide staining. The patterns were photographed and stored as digital pictures in gel documentation system. The reproducibility of the amplification was confirmed by repeating each experiment three times. Data analysis Unequivocally reproducible bands were scored and entered into a binary character matrix (1 for presence and 0 for absence). The genetic similarity among accessions was determined by Nei’s genetic distance [26]. A dendrogram was constructed based on the matrix of distance using Unweighted Pair Group Method with Arithmetic averages. Scores entered in matrix were analyzed using TAXAN version 4.0 software based on the degree of bands sharing. Similarity matrix was generated using Dice coefficient as SI = 2Nab/Na+Nb where Na = total number of bands present in lane a, Nb = total of bands in lane b, Nab number of bands common to lanes a and b [27]. The dice values were then used to perform the UPGMA analysis. To evaluate the robustness of the grouping formed the binary data matrix was subjected to bootstrapping using WinBoot program [28]. The phenogram was reconstructed 1000 times by repeating sampling with replacement and the frequency with which the groups were formed was used to indicate the strength of the group. Correlation co-efficient for the similarity matrices generated by ISSR and RAPD data in both, T.foenum-graecum and T.caerulea, were calculated by method of Smouse et al [29]. The expected heterozygosity, Hn for a genetic marker was calculated as: Hn = 1 - pi2, where pi is the allele frequency of the ith allele [26]. The arithmetic mean heterozygosity Hav for each marker class was calculated as Hav = Hn/n, where n = number of markers or loci analyzed [18]. The average heterozygosity for polymorphic marker (Hav)p was further derived as; (Hav)p = Hn/np; where np is the no. of polymorphic markers or loci [18]. Marker index (Ml) for each marker system was also calculated as, MI = E (Hav)p; where E = effective multiplex ratio [E = nβ where β is the fraction of polymorphic marker or loci,18]. Authors' contributions RD carried out the PCR and molecular genetic study, participated in the design of the study and performed statistical analysis. ML participated in procuring the seeds from different sources and monitoring experiments and result. LC a taxonomist participated in procuring, growing and analyzing the seeds obtained from different sources. PR and VG conceived of the study, participated in its design and co-ordination and monitored the experiments and the results. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514543.xml
545056	A rare case of ascending colon actinomycosis mimicking cancer	Background Actinomycosis is a rare inflammatory disease caused by an anaerobic bacterium that can rarely affect the large intestine. Case presentation We present a rare case of a cecum and ascending colon actinomycosis in a 72 years old woman, mimicking clinically a malignant inflammatory tumor of the right colon. The patient complained of right lower quadrant pain. Although our first thought was a peri-appendiceal abscess, CT scan suggested a right colon tumor. The patient underwent a right colectomy and the histological examination of the specimen revealed colon actinomycosis. Conclusions Preoperative diagnosis in colon actinomycosis is difficult to achieve. Treatment of choice is antibiotics administration. A review of the possible pathogenesis and therapeutic modalities is also presented.	Background Actinomycosis is an uncommon inflammatory entity caused by the universally distributed anaerobic bacterium, Actinomyces Israeli. This Actinomyces is a gram-positive, rather microaerophilic bacterium, which consists a component of the normal human flora. Actinomyces requires the presence of many other bacteria, which destroy the over-vascularized regions and convert aerobic microenvironment to an anaerobic one. Then it's easy for Actinomyces to migrate, infect and proliferate in already injured tissue. Primary bowel involvement is rare, although it has been increased in frequency over the last years. The most common sites of the disease are the transverse colon and the cecum with the appendix [ 1 , 2 ]. Actinomycosis can mimic other abdominal diseases as diverticulitis, abscesses, inflammatory bowel disease and malignant tumors, presenting a diagnostic challenge, and identified post-operatively in most of the cases [ 3 ]. The treatment of choice is antibiotic administration, whenever it is possible due to diagnostic difficulties, although in most cases surgical intervention is performed Diagnosis can be achieved with endoscopy and imaging techniques, as computed tomography (CT scan) and magnetic resonance imaging (MRI). We present a rare case of right colon actinomycosis mimicking malignant tumor, causing bowel obstruction. The diagnosis was achieved only postoperatively. Case presentation A 72 years old woman proceeded to the Emergency Department complaining of acute lower right abdominal pain, mild fever, mild weight loss and constipation. The past medical history of the patient was free. The patient was presented with severe lower right abdominal pain, with signs of local peritonitis and a palpable mass in the same region. The body temperature was 37.2–37.4°C, while arterial blood pressure and cardiac rate 150/90 mmHg and 90/min respectively. The signs of local peritonitis combined with the palpable mass of the lower right abdominal area suggested perforation of the appendix and abscess formation. The laboratory examinations of the patient showed leucocytosis with white blood cell count (WBC) 19000 (with macrophage prevalence: 89%). Colonoscopy revealed obstruction of the right colon. Biopsies were not acquired because the colon lumen was obstructed and the endoscope could not approach the lesion. Computerized tomography (CT) scans of the abdomen (Figure 1 ) revealed a soft lobular mass, measuring 5 × 5 cm, attached to the ascending colon and cecum, compatible with a tumor. The most possible diagnosis was that of a perforated colonic tumor. The patient underwent explorative laparotomy and right hemicolectomy with an end-to-end ileocolonic anastomosis. Thorough exploration of the abdominal cavity revealed no other pathologic findings. The surgical specimen consisted of a 10 cm length of the terminal ileum and the whole of the right colon. The serosa was very heavily covered by suppurative exudates and fibrotic tissue. Microscopic examination of the surgical specimen revealed thickening of the ascending colon wall with neutrophilic infiltration. Numerous polymorphonuclear leukocytes within the muscularis and a fibro-purulent reaction over the serosa with actinomycotic "sulfur granules" in it, were found in high power magnification (Figure 2 ). Upon receiving the pathology report, systemic intravenous penicillin treatment was initiated. Therapy continued for 10 days and then followed by oral penicillin for 6 months. No postoperative complications were observed and the patient was discharged the 14 th day. Discussion Actinomyces Israeli, a filamentous, gram-positive bacillus, is a constant part of the micro flora in the human oral cavity [ 4 ]. Actinomycosis presents a worldwide distribution and no sex predilection is obvious although most of the reported cases refer to males. Abdominal involvement occurs in only 20 percent of all cases of actinomycosis and can mimic malignancy, tuberculosis and inflammatory bowel disease [ 5 ]. Actinomyces is not always pathogenic, and normally exists in stagnated cecum or sigmoid colon. Predisposing factors include previous abdominal surgical operations, intestinal necrosis, foreign bodies, appendicitis and perforation. Some authors suggest that inflammatory or neoplastic processes may contribute to actinomycosis development [ 6 , 7 ]. Bowel obstruction and perforation due to actinomycosis without predisposing factors is very rare and only few cases have been described in the literature. Most commonly actinomycosis occurs in terminal ileus and appendix and rarely in the ascending colon, which is difficult to get obstructed. In our case actinomycosis affected cecum and ascending colon, with a dramatic clinical presentation. We searched the literature in Medline from 1997 to 2004 and we found that only a few cases with clinical presentation similar to our case have been reported. Preoperative diagnosis is difficult although in some cases colonoscopy and histological examination of endoscopically acquired specimen can set the diagnosis. In our case the colon lumen was obstructed and no biopsies were taken. The CT findings suggested perforated colon tumor and an oncologic right hemicolectomy was performed. Actimomycosis usually mimics subacute infections or malignant tumors and the radiologic diagnosis of this entity may be difficult. Some authors suggest that abdominal CT scan with contrast enhancement may reveal a solid mass (intraluminal or extraluminal) with focal areas of attenuation invading the adjacent tissues and suggesting the diagnosis [ 8 , 9 ]. Most common findings in CT scan and/or barium study include mural invasion with stricture formation, mass effect with tapered narrowing of the lumen, and thickened mucosal folds. In many cases the radiologic findings are similar to those of Crohn's disease, intestinal tuberculosis, and excavated malignant tumors [ 10 , 11 ]. The most important CT feature for the correct diagnosis is a large mass adjacent to the involved bowel, which is also a very common finding in patients with colon actinomycosis. In rectosigmoid, colon cystic masses are more common, whereas in transverse or ascending colon purely solid masses are the predominant finding [ 12 , 13 ]. Goldwag et al suggest that CT guided fine needle aspiration can be both diagnostic and therapeutic. Microbiological analysis of material acquired by FNA may reveal sulfur granules, which are suggesting actinomycosis and nocardiosis. In most of the cases the sample receive is difficult especially when intestinal and colon are involved. We believe that in cases where the CT findings are non-specific, surgical exploration is necessary not only for diagnostic but also for therapeutic reasons [ 14 ]. High dose intravenous penicillin injection followed by orally administered penicillin for at least 6–12 months is the treatment of choice. Penicillin administration decreases morbidity and the patient may avoid an unnecessary operation [ 15 , 16 ]. Correct diagnosis is difficult and can be achieved preoperatively in only 10% of the cases, but it is of great importance because the appropriate treatment includes primarily penicillin administration. Surgical intervention is indicated only in cases with obscure diagnosis and for necrotic debridement removal. Although diagnosis only with imaging techniques and laboratory tests is difficult, abdominal actinomycosis should always be included in the differential diagnosis in patients with abdominal masses [ 1 ]. Conclusions In conclusion, colon actinomycosis should always be included in the differential diagnosis of abdominal masses with tumor of inflammatory characteristics. It should be especially suspected when the appearance on CT scan is of a solid mass with focal areas of attenuation or a cystic mass with a thickened wall showing inhomogeneous contrast enhancement that tends to invade adjacent tissues or structures. Immediate and accurate diagnosis, usually by FNA and cytology examination can prevent unnecessary surgical treatment. Competing interests The author(s) declare that they have no competing interests. Authors' contributions All authors contributed equally to this work. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC545056.xml
544360	Practice activity trends among oral and maxillofacial surgeons in Australia	Background The aim of this study was to describe practice activity trends among oral and maxillofacial surgeons in Australia over time. Methods All registered oral and maxillofacial surgeons in Australia were surveyed in 1990 and 2000 using mailed self-complete questionnaires. Results Data were available from 79 surgeons from 1990 (response rate = 73.8%) and 116 surgeons from 2000 (response rate = 65.1%). The rate of provision of services per visit changed over time with increased rates observed overall (from 1.43 ± 0.05 services per visit in 1990 to 1.66 ± 0.06 services per visit in 2000), reflecting increases in pathology and reconstructive surgery. No change over time was observed in the provision of services per year (4,521 ± 286 services per year in 1990 and 4,503 ± 367 services per year in 2000). Time devoted to work showed no significant change over time (1,682 ± 75 hours per year in 1990 and 1,681 ± 94 hours per year in 2000), while the number of visits per week declined (70 ± 4 visits per week in 1990 to 58 ± 4 visits per week in 2000). Conclusions The apparent stability in the volume of services provided per year reflected a counterbalancing of increased services provided per visit and a decrease in the number of visits supplied.	Background In Australia the majority of dentists work in private general practice [ 1 ]. Relatively few are specialists (10.8%), of which 16.8% are oral and maxillofacial surgeons accounting for 1.9% of all practising dentists. The major trends in oral health in Australia over recent decades indicate improved oral health among the population. For example, there has been a dramatic decline in the percentage of edentulous adults [ 2 , 3 ], and caries experience among children has declined since the 1970s [ 4 ], although in the later half of the 1990s improvements in child oral health had ceased [ 5 ]. Service trends in private general practice have reflected the trends towards improved oral health with a shift towards higher provision of services such as diagnostic, preventive and endodontic consistent with the retention and maintenance of a natural dentition [ 6 ]. Practice activity patterns among private general practitioners have shown declining levels of visits supplied but stable levels of time devoted to work [ 7 ]. Identifying trends over time in the practice activity of oral and maxillofacial surgeons is a key element in planning by informing debate on issues relevant to the speciality such as the future supply of services and training needs. Previous Australian data have shown the distribution of oral and maxillofacial surgery was dominated by dentoalveolar services [ 8 ]. However, the dominance of dentoalveolar surgery might be reduced if the practice patterns of surgeons in relation to third molar surgery were influenced by debate on the development of standards and criteria of care [ 9 ], and ongoing assessment of the risks and benefits of removal of third molars [ 10 ]. While the core of oral and maxillofacial surgery is dentoalveolar, the knowledge of the orofacial region forms the basis for the wider scope of the modern specialty [ 11 ]. Since 1996 it has been mandatory that all trainees in Australia enter dual degree programs and then exit with the College fellowship, as a result the percentage of medically qualified surgeons has increased from 2.5% to 17.1% between 1990 and 2000 [ 12 ]. It has been reported that oral and maxillofacial surgeons with medical qualifications, while maintaining a broad scope, tended to have a greater range of procedures within the major groupings [ 13 ]. Considering the trend towards oral and maxillofacial surgeons gaining medical qualifications and the potential impact that this may have on practice activity, the aims of this study are to describe practice activity trends among oral and maxillofacial surgeons in Australia in 1990 and 2000 in terms of time worked, visits supplied and services provided. Methods Study design/sample All registered oral and maxillofacial surgeons in Australia were surveyed in both 1990 and 2000. A surgeon must be registered with a dental board in the state/territory in which they practice. For the purposes of this analysis trainees were excluded. Although some of the same surgeons may have responded to the survey at both points in time for this study the design and analysis is treated as two sequential cross-sectional surveys. Data collection and analyses Data were collected using mailed self-complete questionnaires with a primary approach letter sent initially to each surgeon, followed a week later by the survey materials, with a reminder card two weeks later, and up to four follow-up mailings of survey materials to surgeons who had not yet responded in order to ensure higher response rates [ 14 ]. Surgeon background characteristics and practice factors were described using percentages and compared using chi-square tests for 1990 and 2000. Service rates, time devoted to work and number of visits supplied by surgeons were described using means and compared between 1990 and 2000 using general linear models. All analyses were performed using SAS software [ 15 ]. Study variables The questionnaire was designed to provide comprehensive coverage of a range of workforce issues and the analysis presented here is limited to a subset of the total number of variables that were collected. The questionnaire included surgeon demographic and background variables (e.g., year of birth, sex, place of birth, qualifications,), practice details (e.g., level of activity in private and public practice, type of practice, level of practice activity), and a log of services provided in a typical week. Surgeons recorded details of the patients they treated over a one-week period. Main areas of service were classified as dentoalveolar, trauma, pathology, orthognathic, reconstructive surgery and other/major medical compromise. An outline of the key variables collected and how measures of time worked, services provided and visits supplied were derived is presented in Figure 1 . The time measures of hours per day, days per week and weeks per year worked were used to calculate hours per week and hours per year worked, and were used along with visits per week and services per week to calculate visits per year and both services per visit and services per year. Results Response and background characteristics by year of study Data were available from 79 surgeons from 1990 (response rate = 73.8%) and 116 from 2000 (response rate = 65.1%). Service provision data were available for 4,847 patients from 1990 and 3,292 patients from 2000. Table 1 shows the majority of surgeons in both 1990 and 2000 were in the age group 40–49 years, were male and born in Australia. Similarly, in both 1990 and 2000 the majority of surgeons worked 80+% in private practice. The only significant difference between 1990 and 2000 in Table 1 was the higher percentage of surgeons with dual qualifications, having a medical degree and College fellowship FRACDS (OMS) in addition to a dental degree. Service rates The rate of provision of services is presented in Table 2 . The distribution of services per visit was dominated by dentoalveolar services in both 1990 and 2000. The overall rate of services per visit increased between 1990 and 2000, reflecting increased rates of pathology and reconstructive surgery. The distribution of services per year reflected the pattern for services per visit with dentoalveolar services dominating. However, there were no significant differences over time in the rate of provision of services per year. Practice activity The number of hours per year devoted to work by surgeons did not change significantly between 1990 and 2000, reflecting stable levels of hours per day, days per week and weeks per year worked. However, the number of visits per week supplied by surgeons decreased between 1990 and 2000. The number of visits per year supplied by surgeons also showed a trend towards a decreased number of visits over time, but the change (P = 0.0508) was not significant at the P < 0.05 level. The relationship between the service and visit measures is presented in Figure 2 , which shows that the stability in the number of services provided per year involved a counterbalancing of increased rates of service per visit and decreased numbers of visits supplied. Discussion While the findings of this study are based on a small sample size this primarily reflects the size of the population studied. Oral and maxillofacial surgeons comprise a small percentage of the dental workforce [ 1 ], hence all registered surgeons were included in order to maximise the sample size available for analysis. While smaller samples can reduce statistical power it is still possible to detect significant differences when the effect size is sufficiently large, and a number of statistically significant differences were observed. While small sample sizes can sometimes result in bias, the achievement of adequate response rates in this study did not suggest response bias issues were likely [ 16 ]. The use of a sequential cross-sectional design while unable to address change over time at an individual level, as in a longitudinal design that traces the same subjects, has the advantage of being representative at both points by the inclusion of new subjects that have entered the study population (assuming no bias has been introduced through other means). The practice patterns of oral and maxillofacial surgeons have been reported to be largely stable, showing no change between 1990 and 2000 in their age and sex distribution, place of birth, practice activity level, referral sources, mix of cases and perceptions of work [ 12 ]. The distribution of main areas of service has also remained relatively stable over time [ 17 ]. However, there were signs of change in the increased percentage of dual qualified surgeons, and changes in rates of some non-dentoalveolar surgical procedures over the course of the study. While still a minority of surgeons, those with dual qualifications had a different service profile with higher rates of orthognathic surgery, dental implants, and bone graft procedures. While the mix of cases was dominated by dentoalveolar rather than major maxillofacial surgery in both 1990 and 2000, there appears to be a beginning of an expansion of some selected non-dentoalveolar surgical procedures. However, the stability in orthognathic surgery rates per year observed here indicates that the higher odds of orthognathic surgery among dual qualified surgeons [ 18 ] has not increased the total provision of this type of surgery. The decline in patient visits supplied by oral and maxillofacial surgeons, while statistically significant for visits supplied per week, was not statistically significant at the conventional level of P < 0.05 for visits supplied per year. This partly reflects the stability in weeks per year worked, one of the component measures used to derive visits per year. However, the number of visits per year was at the borderline of statistical significance (P = 0.0508) and considerations other than a reliance on P values is recommended in the literature [ 19 ], with more emphasis on estimation through the use of confidence intervals to accompany point estimates [ 20 ]. A key guide is the measure of effect size, or the size of the difference being reported, which can be factored into considerations of clinical significance at the individual level and public health importance when aggregated across the individuals making up a population. The trends observed in practice activity among oral and maxillofacial surgeons show parallels with private general practice dentists in Australia. Trends in private general practice have also shown a counterbalancing effect of declining numbers of visits supplied with increased rates of services per visit resulting in a stable level of provision of services per year. Medical general practitioners have shown a trend towards longer consultations [ 21 ], consistent with the trend observed for the dental labour force. Despite the divergence in length and scope of training that oral and maxillofacial surgeons are required to fulfil, the convergence in practice activity trends may indicate specific health labour force or even broader labour force issues influencing both groups similarly. The Australian health and community services labour force in general has shown a trend towards working less hours per year with an increase in the percentage working part-time between 1996 and 2002 [ 22 ], which is different to the stable number of hours per year worked by both dentists and oral and maxillofacial surgeons. However, the reported decline in hours worked per week for medical specialists brings their average clinical hours (41.5 hours per week) close to 39.5 hours per week reported for oral and maxillofacial surgeons [ 23 ]. Possible explanations for the lower levels of patient visits per hour among private general practice dentists include increased numbers of older patients [ 24 ], who may have complex treatment needs which require more services or take longer to complete. Changes have been observed in the distribution of the characteristics of patients treated and visits supplied by oral and maxillofacial surgeons over the study period [ 17 ], with the most marked difference being a shift to an older age distribution of patients consistent with demographic trends projecting an increase in middle-aged and older adults in the Australian population. Data from New Zealand have shown increases in the rate and number of injuries among older people and a general increase in the contribution of falls to the occurrence of trauma [ 25 ]. It may also be that as Australians retain more teeth into older age trauma services will expand in these age groups of patients. However, most trauma treated by oral and maxillofacial surgeons is related to the jaw rather than teeth and despite the increased percentage of patients aged 45 years and older, the majority of patients treated by oral and maxillofacial surgeons remained in the age groups 18–24 and 25–44 years. Historical records have indicated that average length of dental appointments changed little over the period 1960–61 to 1974–75, but there was an increase since 1974–75 that was quite marked across the 1977–78 to 1982–83 period [ 26 - 29 ], and continued to increase through to 2001 [ 30 ]. Cross-infection control procedures may be another possible source of influence on productivity associated with either increased appointment or change-over times. The operation of such effects on productivity has implications for planning the delivery of services. Conclusions Estimates of the capacity to supply services and projections of labour force requirements need to consider that while the rate of services per visit has increased this has been counterbalanced by decreases in the number of visits supplied. This has resulted in a stable volume of services provided by oral and maxillofacial surgeons per year over the study period. Competing interests The author(s) declare that they have no competing interests. Authors' contributions DSB performed data analyses and drafting of the manuscript. AJS provided overall supervision of the project. KAS performed data processing and preliminary analyses. DNT was involved in data collection. ANG provided specialist advice on oral and maxillofacial surgery. All authors contributed to and approved the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544360.xml
544348	Gastric emptying is slow in chronic fatigue syndrome	Background Gastrointestinal symptoms are common in patients with Chronic Fatigue Syndrome (CFS). The objective of this study was to determine the frequency of these symptoms and explore their relationship with objective (radionuclide) studies of upper GI function. Methods Thirty-two (32) patients with CFS and 45 control subjects completed a questionnaire on upper GI symptoms, and the 32 patients underwent oesophageal clearance, and simultaneous liquid and solid gastric emptying studies using radionuclide techniques compared with historical controls. Results The questionnaires showed a significant difference in gastric (p > 0.01) symptoms and swallowing difficulty. Nocturnal diarrhoea was a significant symptom not previously reported. 5/32 CFS subjects showed slightly delayed oesophageal clearance, but overall there was no significant difference from the control subjects, nor correlation of oesophageal clearance with symptoms. 23/32 patients showed a delay in liquid gastric emptying, and 12/32 a delay in solid gastric emptying with the delay significantly correlated with the mean symptom score (for each p ≪ 0.001). Conclusions GI symptoms in patients with chronic fatigue syndrome are associated with objective changes of upper GI motility.	Background Chronic Fatigue Syndrome (CFS) is a descriptive term used to define a classifiable pattern of symptoms that cannot be attributed to any alternative condition [ 1 ]. It can be associated with immunological alterations, neuro-endocrine changes [ 2 ], sleep disturbance and disturbed neurocognitive performance with abnormal cerebral perfusion [ 3 ], but the pathophysiological significance of these is uncertain. Skeletal neuromuscular function is usually normal in CFS sufferers [ 4 ]. Many with CFS have gastro-intestinal (GI) symptoms, which are often unrecognised as being part of CFS. The commonest of the upper GI symptoms include fullness and bloating after a small meal, abdominal distension, nausea, and loss of appetite. Lower GI tract symptoms have considerable overlap with irritable bowel syndrome [ 5 ]. The hypothesis explored in this paper is that symptoms of possible upper gastrointestinal origin are more common in patients with CFS and are related to upper gastrointestinal motility as assessed by radionuclide methods. Methods Subjects Consecutive patients with CFS who met the Fukuda criteria [ 6 ] for CFS were all seen by a single physician (RB). Patients with any medical condition which could account for chronic fatigue, a BMI > 30, previous GI surgery or medication affecting the rate of gastric emptying were excluded. Overt psychiatric disease was excluded at the interview. The patients were asked to self assess their percentage reduction in activity from prior to the onset of CFS as a marker of severity. Gastro-Intestinal symptoms were evaluated in patients and controls by a standard questionnaire prior to the gastric emptying studies [ 7 ]. Symptoms were divided into " oesophageal " (dysphagia, heart burn, acid regurgitation), " gastric ": (anorexia, nausea, early satiety, bloating, abdominal distension, intermittent abdominal pain), " other " frequency of bowel actions, consistency of stools, presence or absence of diarrhoea, urgency and timing. Symptoms were scored. 0, none, 1, mild (symptom could be ignored), 2, moderate (symptom could not be ignored, but did not influence daily activities), 3, severe, (symptom influenced daily activities). A mean symptom score (maximum score 3) for the 6 gastric symptoms, and 3 oesophageal symptoms was obtained. The volunteer control subjects who completed the questionnaire were in regular full time employment, with no history of excessive fatigue, on no GI medication, and had no previous GI surgery. Radionuclide measurement of upper GI motility Details and normal ranges of this double isotope test have been previously published [ 8 ]. The solid meal consisted of 100 g of cooked ground beef containing 40MBq in-vivo labelled 99m Tc-sulfur colloid-chicken liver, and the liquid meal consisted of 150 ml of 10% dextrose in water containing with 20 MBq of 67 Ga-ethylenediaminetetraacetic acid (EDTA). All medication (except oral contraceptives) was discontinued for 24 hours prior to each study. The test was performed at 10 am (after an overnight fast) and monitored for at least two hours with the subject in the sitting position with the scintillation camera behind. The study commenced with a standardised oesophageal clearance study (solid bolus) followed by eating the solid meal and then immediately drinking the glucose solution. Each study was continued for at least 2 hours. Oesophageal clearance was expressed as time to 95% clearance (ref range < 93 sec) [ 9 ], Liquid gastric emptying as half-clearance time (ref 4–31 minutes) and solid emptying as amount remaining at 100 min (ref 4–61%). The GI questionnaires were compared between CFS and control by Chi 2 , and Gastric emptying indices compared with historical normal range (t test comparison of means), and correlated with the mean symptom score (± SD). The Study was approved by The Human Research Ethics Committee of the Royal Adelaide Hospital and informed consent given by the subjects. Results Thirty-two (32) CFS patients (22F), with a mean age of 38.5 years had gastric emptying studies. Forty-five (45) control subjects undertook the questionnaire. The demographic details of the controls vs. patients are shown in table 1 Gastro-intestinal symptoms were more common in the CFS group (mean symptom score {MSS] 1.01 ± 0.87) than controls (MSS 0.24 ± 0.34) (table 2 ). Table 1 Characteristics of CFS subjects vs controls. (SD) CFS CONTROLS Number 32 45 Sex F 22, M10 F 37, M 8 Age (yr) 38.5 ± 13.9 34.4 ± 8.5 Weight Kg 68.1 ± 12.1 71.8 ± 11.6 Duration CFS (yr) 9.5 ± 6.8 Severity, % reduction activity 65 Smoke % 18 17.8 Table 2 Percentage frequency of any gastrointestinal symptoms. CFS % (n = 32) CONTROLS % (n = 45) GASTRIC Abdominal discomfort 39 22. Fullness after small meal 70 31* Nausea 67 15* Abdominal pain 76 27* Loss of appetite 42 12* Vomiting 22 2* OESOPHAGEAL Acid regurgitation 30 28 Heart burn 48 27 Swallowing difficulty 45 9* OTHER Bowel movements/ day (mean) 1.6 1.2 Constipation % 26 30 Consistency Formed % 67 80 Loose/Watery % 33 20 Nocturnal diarrhoea % 21 0* Faecal Urgency % 51 16* * indicates symptoms more frequent in CFS group p <.05, Chi 2 The overall, grouped gastric emptying studies of CFS subjects showed no significant slowing of oesophageal clearance p = 0.45 from the control population, and no significant correlation between emptying and oesophageal symptom score (r = 0.15) although 5 of the symptomatic and 2 of the asymptomatic subjects 7/32 (22%), were slower than the 95% confidence limits, (fig 1 ), this did not reach statistical significance. The major abnormality shown is a delay in the emptying of the liquid phase in 23/32 72% of the patients, whereas 12/32 (38%) of solid emptying was delayed compared with the historic controls (t comparison of means, figs 2 and 3 , group p ≪ 0.005). When the gastric emptying results were compared to the mean symptom score there was a highly significant correlation of solid (r = 0.81) and liquid (r = 0.65) delay which increased with the symptom score (p ≪ 0.001). Figure 1 Oesophageal clearance time compared with oesophageal symptoms in Chronic Fatigue Syndrome. (Shaded area represents 95% confidence limit of normal reference range) Figure 2 Per-cent gastric retention of solid food compared with mean symptom score in chronic fatigue syndrome. (Shaded area represents 95% confidence limit of normal reference range) Figure 3 Time to 50% gastric emptying of liquid compared with mean symptom score in chronic fatigue syndrome. (Shaded area represents 95% confidence limit of normal reference range) Discussion G-I symptoms are common in patients with CFS. Abdominal pain is distressing [ 10 ], often requiring analgesia for relief. A previously unrecorded symptom in CFS patients is nocturnal diarrhoea, which disrupts an already disturbed sleep pattern. The most common upper GI symptom is fullness and bloating after a small meal. The usual medical explanation for the gut symptoms is 'irritable bowel'. Unless specific G-I questions are put to the CFS patient they will not spontaneously discuss these symptoms. An abnormality in solid or liquid emptying or combinations of these study parameters was more common in the more symptomatic patients, and liquid was more frequently affected. This is the opposite of the abnormality seen in diabetic subjects [ 8 ], where the major abnormality, delay in the solid phase of gastric emptying has been ascribed to autonomic dysfunction or hyperglycaemia. A group of elderly subjects with a number of neurological defects showed a delay in the liquid rather than the solid emptying [ 11 ]. Symptoms and delayed gastric emptying in diabetic gastroparesis studies have not correlated well. In this study there is a good correlation with symptoms. The commonest of these was early satiety, fullness and bloating after eating. There was though a poor correlation with oesophageal symptoms and a disorder of oesophageal emptying. GI motility is complex, with central, local neuromuscular and humoral influences. Non-specific endocrine disturbances have been demonstrated in CFS, but the relevance of these is unknown with regard to GI disturbances. Skeletal muscle fatigue appears to be of central rather than peripheral origin, but again it is not known whether this may be extrapolated to visceral muscle. Inconclusive central changes have been documented. The impact of disturbed sensory function is unknown, and this could also involve peripheral nerves or the central processing of sensory information. Diagnostically, there is overlap between CFS, functional dyspepsia and fibromyalgia and all may be related to abnormal sensory processing [ 10 , 12 , 13 ]. Altered gastric emptying has been shown in association with irritable bowel syndrome [ 14 ]. Conclusions These observations indicate that there is measurable disturbance in upper gut motility corresponding with symptoms in CFS. Although the cause for these findings is not apparent in this study, the more prominent delay in liquid rather than solid emptying may point to a central rather than a peripheral aetiology. The gastro-intestinal tract and function should be properly investigated and the symptoms not necessarily be ascribed to irritable bowel syndrome. Competing interests The author(s) declare that they have no competing interests. Authors contribution RB examined the patients and analysed the clinical data, BC performed the Nuclear Medicine studies, drafted the manuscript and performed the statistics. Both read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544348.xml
526283	Erytrocyte membrane anionic charge in type 2 diabetic patients with retinopathy	Background The Steno hypothesis states that changes in basement membrane anionic charge leads to diabetic microvascular complications. In diabetic nephropathy, loss of basement membrane glycosaminoglycans and the association between glomerular basement membrane heparan sulphate and proteinuria has been documented. A correlation between erythrocyte surface and the glomerular capillary wall charges has also been observed. The aim of this study is to evaluate the relationship between retinopathy and erythrocyte anionic charge and urinary glycosaminoglycan excretion in type 2 diabetic patients. Methods 49 subjects (58 ± 7 yrs, M/F 27/22) with type 2 diabetes with proliferative retinopathy (n = 13), nonproliferative retinopathy (n = 13) and without retinopathy (n = 23) were included in the study. 38 healthy subjects were selected as control group (57 ± 5 yrs, M/F 19/19). Erythrocyte anionic charge (EAC) was determined by the binding of the cationic dye, alcian blue. Urinary glycosaminoglycan and microalbumin excretion were measured. Results EAC was significantly decreased in diabetic patients with retinopathy (255 ± 30 ng alcian blue/10 6 RBC, 312 ± 30 ng alcian blue/10 6 RBC for diabetic and control groups respectively, p < 0.001). We did not observe an association between urinary GAG and microalbumin excretion and diabetic retinopathy. EAC is found to be negatively corralated with microalbuminuria in all groups. Conclusions We conclude that type 2 diabetic patients with low erythrocyte anionic charge are associated with diabetic retinopathy. Reduction of negative charge of basement membranes may indicate general changes in microvasculature rather than retinopathy. More prospective and large studies needs to clarify the role of glycosaminoglycans on progression of retinopathy in type 2 diabetic patients.	Background Diabetic retinopathy is the leading cause of blindness in diabetic adults [ 1 , 2 ]. During the first two decades of the disease, nearly all patients with type 1 diabetes and > 60 % of type 2 diabetes have retinopathy. The duration of the diabetes is probably the strongest predictor for development and progression of retinopathy [ 3 ]. The incidence of retinopathy positively correlates with HbA 1c [ 4 ]. Diabetic patients with evidence of nephropathy are characterized by a 5 to 10 times higher incidence of proliferative retinopathy [ 5 ]. Albuminuria not only is associated with kidney disease but, is also a strong predictor of cardiovascular disease and proliferative retinopathy, suggesting it reflects a generalized vascular disease [ 6 ]. Thus, the coincidence of generalized vascular dysfunction, albuminuria, mesangial expansion, proliferative retinopathy and accelerated development of atherosclerosis suggests a common cause of abnormalities in susceptible diabetic patients. The Steno hypothesis proposed that increased loss of anionic charge from basement membranes leads to diabetic microangiopathy [ 7 ]. This hypothesis has been proven for diabetic nephropathy but it's not clear for diabetic retinopathy. Anionic content of basement membranes predominantly produced by glycosaminoglycan molecules (GAG). They constitute the charge selective barrier of the glomerul basement membrane [ 8 ]. Several clinical studies indicate that loss of the GAG is associated with diabetic albuminuria in diabetic and experimental models [ 9 , 10 ]. Although abnormal GAG metabolism is a well known phenomen in diabetic nephropathy, it is still debated whether it has a pathogenic role in diabetic retinopathy. The aim of this study was to evaluate the relationship between diabetic retinopathy and erythrocyte anionic charge as well as the urinary glycosaminoglycan excretion in type 2 diabetic patients. Methods 49 outpatients (27 male, 22 female) with type 2 diabetes mellitus diagnosed after the age of 30, were included and divided into 3 subgroups according to severety of retinopathy. Diabetes was diagnosed according to American Diabetes Association criteria. The onset of diabetes was after the age of 30 in these patients. The study was approved by the institutional ethics committee and patients gave written informed consent. Exclusion criteria included hypertension, myocardial infarction, cerebrovascular disease, heart failure, treatment with antiaggregants, steroids or other drugs that effect blood pressure and glucose, serum creatinine > 200 μmol/l, connective tissue disorders and other systemic disease. Renal and bladder infection diseases of patients and controls were excluded biochemically and microbiologically. The ocular fundus was examined by an ophthalmologist after dilation of the pupils and classified. The patients were divided into three groups according to the severity of retinopathy. Group 1 consisted of 23 patients without retinopathy (R 0 ). Group 2 consisted of 13 patients with nonproliferative retinopathy (R 1 -hemorrhages, microaneurysms, cotton-wool exudates). The 13 patients in group 3 (R 2 ) had proliferative retinopathy (R 2 -neovascularisations, vitreous hemorrhages, ablatio of the retina). Age and sex matched 38 healthy volunteers were included as healthy controls. Demographic characteristics of the groups are shown in Table 1 . Venous samples were collected after an overnight fast and 24 hour urine samples collected. The charge on erythrocytes (RBC) was measured by means of cationic dye alcian blue 8GX (Sigma catalogue no: A 5268) according to the method of Levin with minor modifications as follows: [ 11 ]. From citrated venous blood samples, platelets and leukocytes were removed using the method of Beutler et al. [ 12 ]. After removal of platelets and leucocytes erythrocytes were washed 3 times in phosphate buffered saline (PBS), and subsequently a fraction of RBC was resuspended in PBS containing alcian blue at a final concentration of 250 mg/l. After a 30 min incubation at 37°C, the RBC suspension was centrifuged, and the remaining alcian blue concentration was measured in the supernatant with a Shimadzu UV 1201 V spectrophotometer (Shimadzu, Japan) at a wavelength of 650 nm. Each determination represents the mean of 2 assays. The quantity of alcian blue bound to RBCs was expressed as nanograms of alcian blue per 10 6 RBCs. In our experimental conditions the intraassay and interassay coefficients of variation were 5.8 and 7.6 % for 100 nanograms of alcian blue per 10 6 RBCs. Urinary glycosaminoglycan (U GAG ) excretion was determined in 24 hr urine samples spectrophotometrically at 520 nm by the addition of dimethylmethylene blue (Aldrich Chem Co., USA) and bovine kidney heparan sulfate as standard (Sigma catalogue no H 7640) [ 13 ]. In our experimental conditions the intraassay and interassay coefficients of variation were 1.5 and 2.4 % for 10 mg/l glycosaminoglycan, respectively. Serum sialic acid was measured in serum with a colorimetric method described by Svennerholm with a Shimadzu UV-1201 spectrophotometer (Shimadzu, Japan) with a wavelength of 525 nm. Using N-acetyl neuraminic acid (Sigma catalogue no: A 3007) as standard [ 14 ]. Intra assay and interassay coefficients of variations were 6.6 and 9.2 % for 1 mmol/l sialic acid respectively. Urinary albumin excretion was measured by an immunoturbidemetric assay (Roche diagnostics, USA) on automated clinical chemistry analyzer (Hitachi 902). In all patients, the annual level was determined as the mean of urinary albumin excretions in three 24 h urine collections taken at home during normal physical activity. The intraassay and interassay coefficients of variation were 1.3 and 4.3 % for a mean value of 25 mg/l albumin concentration. Serum glucose, cholesterol, triglyceride and HDL (Dade Behring, USA) were evaluated with enzymatic methods (Dade Behring, USA) on an automated clinical chemistry analyzer (Dimension RxL). Creatinin was assayed by the Jaffe reaction on an automated clinical chemistry analyzer (Dimension RxL). HbA 1C (reference range 4.4–6.0 %) was measured by a turbidimetric inhibition immunoassay technique (Roche Diagnostics, USA) on automated clinical chemistry analyzer (Hitachi 902). The intraassay and interassay coefficients of variation were 1.8 and 3.0 % for a mean value of 10.5 % HbA 1C concentration. The analysis of the data was performed with a PC compatible Instat-II programme. Paired t test and ANOVA were used where appropriate. Correlation analysis was performed with Spearman rank test. The differences were considered significant when the probability was p < 0.05. The results were given as mean ± SD. Results The clinical characteristics of the study groups are shown in Table 1 . There were no difference in age and BMI between the groups. Duration of diabetes mellitus was slightly longer, although not significantly different than that of the patients with proliferative retinopathy. Results of laboratory parameters are shown in Table 2 . Patients with proliferative retinopathy had higher concentrations of triglyceride and lower concentrations of HDL cholesterol as compared to diabetics without evidence of retinopathy and control group. Serum cholesterol level was the same in all study groups. Blood glucose and HbA 1C levels were not different in subgroups of our diabetic population. Distribution of erythrocyte anionic charges of all study groups are demonstrated in figure 1 . EAC was 312 ± 30 ng alcian blue/10 6 RBC in healthy controls. Diabetic patients both with nonproliferative and proliferative retinopathy had a significantly lower alcian blue binding to RBCs compared with diabetic patients without retinopathy (p < 0.001). A marked (p < 0.05) difference was found between type 2 diabetic patients (8.3 ± 4.1, 8.7 ± 4.8, 8.4 ± 3.3 mg/24 h, Table 2 ) and controls (5.0 ± 2.4 mg/24 h) regarding urinary GAG excretion. We did not observe a correlation between urinary GAG excretion and diabetic retinopathy. A significant difference (p < 0.05) was observed in serum sialic acid values in diabetic patients compared with healthy controls. Serum sialic acid level was higher in the nonproliferative retinopathy group, but the difference was statistically insignificant (p > 0.05). Microalbumin level was significantly p < 0.05 higher in diabetic patients (73 ± 40 [35–250], 75 ± 21 [50–202], 85 ± 45 [56–224] mg/24 h, Table 2 ) compared to healthy controls (10 ± 6 mg/24 h). But there wasn't a correlation between the severity of retinopathy and microalbuminuria (p > 0.05). Results of the Spearman rank correlation test are shown on table 3 . EAC is found to be negatively corralated with microalbuminuria in all groups. Discussion Our study demonstrates a statistically significant quantitative reduction of alcian blue binding to RBCs in diabetic patients than healthy subjects, which is in accordance with literature [ 15 , 16 ]. Previously we have shown a statistically significant quantitative reduction of alcian blue binding to RBCs in streptozosin diabetic rats [ 17 ]. Alcian blue is a complex amphoteric molecule that binds to acidic glycoproteins which represent a large class of cell surface anionic molecules [ 18 ]. As alcian blue binding is an expression of the anionic charge on the cell surface, this result indicates that RBC anionic charge is reduced in diabetic patients. Erythrocyte anionic charge by itself is unlikely to be important in the pathogenesis of diabetic retinopathy but reduced RBC alcian blue binding was found to be associated with the loss of glomerular basement membrane anionic charges in diabetic rats and patients [ 19 ]. Changes in the composition of basement membranes are likely to be responsible for functional disturbances and hence for the development of capillary disease. Chakrabarti et al. reported a decreased density of anionic sites in the retinal basement membrane of BB rats [ 20 ]. A similar decrease in anionic density was also demonstrated in the Bruch's membrane of BB and streptozosin diabetic rats [ 21 ]. Several authors have questioned the validity of the alcian blue binding assay. The major objections concern the incomplete alcian blue dissolution in PBS buffer and its tendency to precipitate with time [ 22 , 23 ]. We prepared a fresh alcian blue solution for each experiment. Each experiment was considered valid when no change in the optical density of the blank alcian blue solution occurred over the duration of the experiment. In addition, the experiments were done in batches with cells from a healthy person. Therefore any flaw in the methodology would not affect only one group and thus distort the results. Diabetic retinopathy is characterized by gradually progressive alterations in the retinal microvasculature leading to vascular hyperpermeability, capillary occlusion and ultimately neovascularization. Thickening of the retinal microvascular basement membrane and loss of anionic content are well documented morphological features of diabetic retinopathy [ 20 , 21 , 24 ]. These changes causes breakdown of the blood-retinal barrier resulting in capillary hyperpermeability and leakage of proteins into the deep and superficial layers of the retina [ 1 , 2 ]. Glycosaminoglycan primarily heparan sulfate has been implicated in permeability properties of basement membrane of glomerular basement membrane as well as the retinal basement membrane [ 25 ]. Kahaly et al. reported that urinary GAG excretion is increased in diabetic patients with microangiopathy [ 26 ]. Williamson et al and Ginn et al. have shown that albumin permeation is increased into the diabetic compared to normal retina [ 27 , 28 ]. Loss of heparan sulfate from the retinal basement membrane in addition to its thickening may cause increased vascular permeation of albumin and other substances into the retina and possibly into the optic nerve. Systemic treatment of IDDM patients with proteinuria with danaparoid sodium, a glycosaminoglycan, not only reduced proteinuria, but also decreased the number of hard exudates in the retina [ 29 ]. Studies in diabetic rats have shown reduced activity of the key enzyme in the biosynthesis of heparan sulfate, N-deacetylase, which results in impaired heparan sulfate biosynthesis in experimental diabetes [ 30 , 31 ]. In this study we found that urinary GAG excretion is increased in diabetic patients than healthy controls, but there was no difference in GAG excretion between nonproliferative and proliferative retinopathy patients. We couldn't find a correlation between GAG excretion and EAC but we found a correlation between albumin excretion and EAC. Although our working hypothesis is that the reduction in anionic charge in basement membranes may be due to an abnormal GAG metabolism, we couldn't confirm it with this study. Our study did not directly address the pathophysiology of diabetic retinopathy but a reduction in erythrocyte charge perhaps connected with the properties of retinal basement membrane. Several reports suggest that elevated serum sialic acid levels in type 1 and type 2 diabetic patients indicate microvascular damage [ 32 , 33 ]. In our study, diabetic patients were found to have elevated serum sialic acid levels in comparison with healthy controls. Therefore, the increase in serum sialic acid level may indicate a coincidence between the structural alterations in the retina and general vascular damage. Conclusions We conclude that type 2 diabetic patients with low erythrocyte anionic charge are associated with diabetic retinopathy. Reduction of negative charge of basement membranes may indicate general changes in microvasculature rather than retinopathy. More prospective and large studies needs to clarify the role of glycosaminoglycans on progression of retinopathy in type 2 diabetic patients. List of abbrevations BMI: Body mass index EAC: Erythrocyte anionic charge GAG: Glycosaminoglycan PBS: Phosphate buffered saline RBC: Erythrocytes U GAG : Urinary glycosaminoglycan Competing interests The authors declare that they have no competing interests. Authors' contributions YB : Evaluated laboratory parameters. Designed the study. Wrote the paper. HD: Diagnosed outpatients according to American Diabetes Association criteria. MA: Examined ocular fundus after dilation of the pupils and classified according to the severity of retinopathy. DY: Performed the analysis of the data with a PC compatible Instat-II programme. Improved the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC526283.xml
524494	Efficacy of two once-daily methylphenidate formulations compared across dose levels at different times of the day: Preliminary indications from a secondary analysis of the COMACS study data	Background Methylphenidate (MPH) is commonly prescribed in the treatment of Attention-Deficit/Hyperactivity Disorder or ADHD. Concerta and Metadate CD are once-daily formulations of MPH using different delivery mechanisms resulting in different pharmacokinetic profiles. A recent study (COMACS) showed that for near-milligram (mg) equivalent daily doses, Metadate CD provides greater symptom control in the morning (1.5 through 4.5 hours post-dose), while Concerta provides greater control in the early evening (12 hours post-dose). Non-inferential comparison of effects for different dose levels of the two formulations suggested that equivalent levels of morning symptom control could be obtained with lower daily doses of Metadate CD than Concerta; the situation being reversed in the evening. The current paper presents a secondary analysis that provides a statistical test of these observations. Method The COMACS study was a multi-center, double-blind crossover study of Metadate CD, Concerta and placebo with each treatment administered for 1 week. Children were assigned on the basis of their pre-trial dosage to either high (Metadate CD 60 mg; Concerta 54 mg), medium (Metadate CD 40 mg; Concerta 36 mg) or low doses (Metadate CD 20 mg; Concerta 18 mg) of MPH, and attended a laboratory school on the 7th day for assessment at 7 sessions across the day. For the post-hoc comparisons across dose levels presented here, total SKAMP scores with the active treatments (adjusted for placebo response) were analyzed using an analysis of covariance, with a combined measure modeling placebo response across all time period as the covariate. Results Symptom control from 1.5 through 6.0 hours post-dose was as good with lower doses of Metadate CD (20 and 40 mg) as with higher doses of Concerta (36 and 54 mg, respectively). Lower daily doses of Concerta (18 and 36 mg) and higher doses of Metadate CD (40 and 60 mg, respectively) gave equivalent control at 7.5 and 12 hours with Metadate CD giving better control from1.5 through 6.0 hours post-dose. Conclusions Different delivery profiles of Metadate CD and Concerta can be exploited to limit total daily exposure to MPH while at the same targeting a specific, especially clinically significant, period of the day. These results need to be confirmed in a study in which children are randomly allocated to different dose levels of the two formulations and plasma MPH concentrations are assessed simultaneously.	Background Attention Deficit /Hyperactivity Disorder (ADHD) is a relatively common early onset developmental condition characterised by a pervasive and persistent pattern of age inappropriate and debilitating inattention, impulsiveness and overactivity. It is reported to affect between 3 and 6 percent of the childhood population and, if untreated, to be associated with a poor prognosis in adolescence and adulthood [ 1 , 2 ]. Methylphenidate (MPH) remains a pharmacological treatment of first choice for children with ADHD [ 3 ]. Historically, effective 'all-day' management of symptoms has relied on the use of multiple doses (typically two or three) of immediate release (IR) MPH spread out across the day (early morning, midday and evening)[ 4 ]. The use of IR formulations in this way combines all-day coverage with the opportunity to tailor doses at different times of the day to meet the specific needs of children. However, there is evidence that multiple dosing leads to problems with adherence especially during the school day where children receiving medication may feel stigmatised by their classmates [ 5 ]. Once-a-day sustained release (SR) formulations have been licensed in the US for some time but the early formulations were not widely used because of the perceived lack of efficacy especially with regard to speed of onset [ 6 ]. In the last few years a second generation of more effective formulations (referred to here as extended release formulations – ER) have been licensed. These exploit a range of different delivery technologies and offer smooth patterns of symptom control across the day [ 7 - 9 ]. These new formulations represent a major advance in the clinical management of the condition and are popular with both patients and clinicians. Various ER formulations have been designed each with a different pharmacokinetic (PK) and pharmacodynamic (PD) profile that results in differing patterns of duration and timing of effect. Thus they have the potential to provide clinicians with the opportunity to simplify the dosing regime without loosing the ability to tailor treatment to the clinical profile of an individual patient. In order to exploit this opportunity clinicians need to be able to make informed decisions about the comparative benefits of differing doses of different formulations with different PK/PD profiles. Unfortunately, to date, there have been few head-to-head trials of these new ER formulations that provide the information required for this. We recently reported the results of a randomised, placebo-controlled, head-to-head comparison of the pharmacodynamic (PD) properties of near mg-equivalent daily doses of two safe and effective [ 10 , 11 ] ER formulations of MPH in children (the COMACS study; [ 12 ]). Concerta (CON) was designed to replace three-times-a-day (TID) IR MPH and provide twelve-hour symptom coverage. It consists of an insoluble OROS™ tablet with an IR drug over coat. Twenty two percent of the dose is in the IR overcoat and 78% in the ER core, which is released, by an osmotic pump process [ 7 , 10 ]. Metadate CD (MCD) has a profile more in keeping with a two-times-a-day (BID) regime of IR MPH. It consists of a soluble capsule containing a mixture of IR MPH beads (30% of the total dose) and ER MPH beads coated with a controlled-release polymer to deliver MPH gradually over the extended period (70% of the total dose) [ 8 , 11 ]. Low (20 mg MCD and 18 mg CON), medium (40 mg MCD and 36 mg CON), and high (60 mg MCD and 54 mg CON) doses are available for each formulation and have been demonstrated to deliver equivalent levels of total exposure to MPH in adults [ 13 ] (as shown by comparable AUCs and C max ). Recently a 10 and 30 mg dose of MCD and a 27 mg dose of CON were licensed, but these were not available during the COMACS study. As a result of the design differences, MCD releases 50% more IR MPH in the initial bolus delivery process than CON for a near mg-equivalent total daily doses. Furthermore although the amount of ER MPH is the same for near mg-equivalent doses the pattern of delivery differs with MPH release front-loaded with MCD (starting at 1 hour) and more back-loaded with CON. The different IR/ER ratios and time-course for dissolution of the two extended-release portions of the formulations result in distinctly different plasma concentration vs. time profiles in adults when compared at near-mg equivalent total doses [ 13 ]. Concentrations of MPH are significantly higher for MCD than for CON for up to 6 hours after dosing, and, by contrast, concentrations of MPH are significantly higher for CON from 8 through 12 hrs after dosing. Differences in plasma concentration vs. time profiles are expected to occur in children as the PK of MPH in adults and children are qualitatively similar (i.e. there are no reported age-related differences in absorption, distribution, metabolism and excretion of MPH between these populations [ 14 - 16 ]). In agreement, the plasma concentration vs. time profile for MCD in children is consistent with that observed in adults [ 8 ]. Results of the COMACS study demonstrated that the PD patterns of the two formulations in children mirrored their expected differences in plasma concentration vs. time profiles. This meant that at each of the three near-mg equivalent daily dose levels, MCD produced a greater reduction in symptoms during the morning (up to six hours from drug administration) while CON produced superior control in the early evening (i.e., at 12 hours post-dose). However, the two formulations were equivalent in their effects during the afternoon between 6 and 7.5 hours post dose [ 12 ]. The publication of the main analysis of the COMACS study data has led to a number of suggestions for alternative analyses that would usefully address questions of the clinical utility of the two formulations. Although from a scientific point of view the aim of the COMACS study – to compare across-the-day PD profiles of the two formulations – was best served by a comparison of bio-equivalent (i.e. AUC-equivalent) total daily doses, other strategies for selecting comparator doses could have been legitimately followed. For instance, it has been suggested that comparators could have been matched on the basis of the size of the IR component dose rather than the total daily dose (see Table 1 for IR and ER components of the MCD and CON). In this regard it is interesting that an informal comparison of the relative efficacy of MCD and CON at each dose level suggested that equivalent morning symptom control would possibly be obtained at doses with a near mg-equivalent IR component but with a lower total daily dose of MCD than CON, while the situation was reversed in the evening (that is, lower daily doses of CON could have equivalent efficacy to higher doses of MCD). In each case, the COMACS study data suggested that this targeted control would be gained at the cost of efficacy at other point(s) during the day. These across-the-day-changes in patterns of the relative efficacy of the two formulations are consistent with what would be expected on the basis of their IR/ER ratios, their drug delivery mechanisms, and their resulting predicted plasma concentration vs. time profiles. Table 1 Amounts of Immediate-Release (IR) and Extended-Release (ER) Methylphenidate in Available Dosage forms of Metadate CD and Concerta a IR MPH ER MPH 18 mg CON 4 mg 14 mg 20 mg MCD 6 mg 14 mg 36 mg CON 8 mg 28 mg 40 mg MCD (2 × 20 mg capsules) 12 mg 28 mg 54 mg CON 12 mg 42 mg 60 mg MCD (3 × 20 mg capsules) 18 mg 42 mg a Recently, a 10- and 30-mg dose strength of Metadate CD and a 27-mg dose strength of Concerta have become available. These were not available at the time of the COMACS study. These observations could be clinically important because it is expected that patients and their families may be willing to trade some degree of symptom control at certain times of the day if that allows them to reduce the overall dose of MPH. In such cases, clinicians will seek a dosing profile that allows them to target a specific period of the day that is especially important for a particular patient. By doing this they can retain the effectiveness of a higher dose component during these selected periods of the day while limiting total daily MPH exposure. For instance, in some cases clinicians, patients and parents may seek symptom control in the evening while being less concerned about the daytime. In other cases they may wish to focus on morning and afternoon symptom control. Because the main aim of the COMACS study design was to compare the PD profiles of MCD and CON across the day, initial analyses were limited to within-subject comparisons between equivalent total daily doses (low, medium and high) of the formulations. Thus this primary analysis of the data did not include a cross-dose analysis, which would allow the relative efficacy of different daily doses of MCD and CON to be directly tested at different times of the day. The present paper presents a statistical cross-dose analysis of the efficacy of MCD and CON at different times of the day in order to test these observations directly. The ideal way to test across-dose comparisons of different medications is to randomise patients to different dosing strata. In the COMACS study children were assigned to dosing strata on the basis of pre-trial dose levels. However, because the COMACS study was designed to have similar and substantial numbers of children in each dosing strata the data set still offers the opportunity for a preliminary post-hoc exploration of the PD profiles of different doses of MCD and CON. Furthermore, the inclusion of the placebo arm in the COMACS design enabled us to directly address one of the possible major confounds associated with across dose comparisons in this sort of stratified design. An inspection of the published COMACS data revealed that the placebo scores measured at different sessions across the day were higher in the high dose group, than the medium or low dose groups. This suggests that children on higher doses may have had a more severe expression of the disorder making cross-group comparisons complicated. In the analysis reported in the current paper these placebo scores were used to adjust the treatment outcome scores to take account of this. A number of specific across-dose comparisons were made. First, the lower daily doses (20 mg and 40 mg) of MCD were compared to the higher daily doses of CON (36 mg and 54 mg respectively). The comparison between MCD 20 mg and CON 36 mg is of particular interest to clinicians as these are the suggested dose substitutions for IR MPH 10 mg BID and 10 mg TID, respectively. On the basis of the initial observations of the COMACS study it was predicted that MCD 20 and CON 36 mg, and MCD 40 and CON 54 mg would produce equivalent control in the morning (from 1.5 through 4.5 hours post-dose). In contrast, at these dose levels it was predicted that CON would produce significantly greater effect as the day wore on (from 6 hours onwards). Second, lower daily doses (18 mg and 36 mg) of CON were compared with higher doses of MCD (40 mg and 60 mg respectively). We tested the prediction that at these dosing levels MCD would demonstrate greater efficacy only in the morning and early afternoon (from 1.5 through 6 hours); so that despite the lower total daily dosing levels, CON, designed to be effective over a 12 hour period, would still be more effective than MCD, designed to be effective over an 8 hour period, at reducing symptoms in the early evening (i.e., at 12 hours). Methods Clinical materials Metadate ® CD (methylphenidate HCl, USP) Extended-Release Capsules (MCD) were obtained from Eurand Americas, Inc (Vandalia, OH), while Concerta ® (methylphenidate HCl) Extended-release Tablets (CON) were obtained from Alza Corporation (Mountain View, CA). For a detailed description of the preparation of clinical materials see Swanson et al., 2004 [ 12 ]. Patients Six to 12 year old children, with interview-confirmed-diagnoses of ADHD who were being treated with MPH in doses of between 10 to 60 mg/day (5 mg to 20 mg per administration, one to three times a day) were recruited for the trial. Children were deemed otherwise healthy on the basis of an extensive medical history and physical examination. Children were excluded if they had an IQ below 80 or the inability to follow or understand study instructions. Other standard exclusion criteria for MPH drug trials applied [ 12 ]. Children provided signed assent, and their legal guardians signed an IRB-approved consent form. A total of 214 patients were screened for participation into the study and 184 patients (74 percent of which were male) were stratified across the three dose levels based on their previously established clinical doses of MPH. Eighty-two percent of the patients met the criteria for ADHD-Combined Type with a further 15 percent meeting the criteria for Inattentive Type. Approximately 25% of children had a co-morbid condition (e.g., anxiety and oppositional defiant disorder). At prescreening, approximately 91% of the patients were on once-a-day dosing regimens. Of the remainder 7.6% were taking BID and 1.6% TID IR MPH. Of the 184 subjects entering the study, 157 received all three levels of treatment and participated in all seven classroom sessions. The demographic characteristics of the sample of patients that completed all three treatments (n = 157) were not different than those reported for the full sample. Design This was a ten-site, double-blind, placebo-controlled crossover study comparing three treatment conditions: MCD, CON, and placebo (PLA). The study was conducted in accordance with the principle of the Declaration of Helsinki and its amendments and the International Committee on Harmonization Guidelines on Good Clinical Practice. Dose-level assignment was made according to pre-existing daily dosing requirement for MPH and children remained at the dose-level assigned for the entire study duration. Children treated with low doses (≤ 15 mg/day IR or 20 mg/day ER) of MPH were randomized to receive a daily dose of MCD 20 mg, CON 18 mg or PLA; those treated with medium doses (>10 to ≤ 30 mg/day IR or >20 to ≤ 40 mg/day ER) of MPH were randomized to receive MCD 40 mg, CON 36 mg or PLA; and children treated with high doses (> 30 mg/day IR or >40 mg/day ER) of MPH were randomized to receive MCD 60 mg, CON 54 mg, or PLA. Each of the three treatments was administered for 7 days (in randomized order) without an intervening washout period. Assessments took place in the laboratory school on Days 7, 14, and 21 (for a detailed description of the laboratory classroom day see Swanson et al., 2004 [ 12 ]). Two trained observers assessed patients during each classroom session on the Swanson, Kotkin, Atkins, M/Flynn, Pelham Scale (SKAMP; [ 17 , 18 ]) on the basis of a 1.5-hour cycle of activities with separate assessments of Attention and Deportment being made at 0, 1.5, 3.0, 4.5, 6.0, 7.5 and then 12 hours after drug administration. Because Attention and Deportment scores were highly correlated in the original COMACS analysis, these subscales were combined in the current analysis for ease of presentation. Statistical analyses A preliminary factorial Analysis of Variance (ANOVA) using the SPSS General Linear Model was conducted in order to confirm that the pattern of effects of Treatment (CON, MCD, PLA), Dose (low, med, high) and Session (0, 1.5, 3.0, 4.5, 6, 7.5 and 12 hours) found in the Swanson et al. (2004) paper [ 12 ] held when the individual SKAMP Attention and Deportment scales were combined to provide a composite score. In order to test the specific across-dose predictions Analysis of Covariance (ANCOVA) was used to make the four comparisons outlined above. In each case, Treatment/Dose (analysis 1-MCD 20 mg vs. CON 36 mg; analysis 2-MCD 40 mg vs. CON 54 mg; analysis 3-MCD 40 mg vs. CON 18 mg; analysis 4-MCD 60 mg vs. CON 36 mg) was the between-subjects independent factor. Session was the within- subject factor, and the dependent variable was total SKAMP score. Children's active drug SKAMP scores were adjusted to take account of their behaviour on PLA using a weighted combined SKAMP score for all observation points. Weighting was determined by principle components analysis and was similar for each observation point. Other between-subjects factors included in the initial COMACS study data analysis (including Site and Sequence of Drugs) were excluded from the current analysis. The GLM option that utilizes data from just those subjects with complete data (i.e., those cases without missing data) was selected for each separate analysis. Results Preliminary analysis Table 2 shows the total SKAMP scores for each observation sessions at each dose level of CON, MCD and placebo. There were significant main effects of treatment, F (2, 306) = 92.06; p < 0.001, and session, F (6, 918) = 34.70; p < 0.001, and an interaction between treatment and session, F (12, 1836) = 45.21; p < 0.001. Planned comparisons demonstrated that the relative efficacy of the two formulations in relation to PLA was as described by Swanson et al. [ 12 ] for separate SKAMP deportment and attention scales: CON = MCD < PLA at the time of dose delivery, MCD > CON > PLA for 1.5, 3 and 4.5 hours, MCD = CON > PLA for session 6 and 7.5 hours and MCD = PLA < CON at 12 hours. For a discussion on the superiority of placebo immediately after dosing, the reader is referred to Swanson et al [ 12 ]. Table 2 Mean (± SD) SKAMP Total Scores at Each Observation Session for Each Treatment at Each Dose Level a MCD CON PLA Observation Session (hrs) Low Med High Low Med High Low Med High 0 18.48 (11.82) 20.88 (12.95) 19.91 (13.15) 18.04 (10.13) 19.14 (12.14) 21.47 (13.06) 13.58 (9.72) 16.02 (11.84) 13.96 (11.14) 1.5 11.44 (7.99) 10.98 (8.62) 6.55 (5.85) 14.04 (9.85) 14.86 (12.01) 11.34 (9.71) 19.10 (12.83) 19.47 (12.56) 18.88 (13.48) 3.0 12.57 (9.92) 11.03 (9.66) 7.31 (6.10) 16.44 (12.43) 15.29 (12.72) 12.62 (11.00) 21.47 (14.61) 20.98 (14.11) 22.11 (14.10) 4.5 13.46 (11.53) 12.39 (10.32) 9.15 (8.62) 17.55 (13.37) 15.09 (12.60) 13.55 (11.91) 20.23 (11.92) 22.09 (15.46) 23.44 (12.55) 6.0 16.08 (13.27) 14.47 (11.53) 10.30 (9.71) 17.00 (12.12) 14.28 (11.73) 12.04 (11.62) 22.98 (12.79) 22.15 (13.91) 26.02 (14.56) 7.5 15.85 (11.21) 17.26 (13.63) 14.29 (12.55) 18.62 (12.66) 15.19 (13.47) 13.47 (12.97) 23.54 (12.96) 23.13 (14.72) 24.48 (14.68) 12.0 20.44 (13.75) 20.28 (15.02 19.85 (14.41) 16.90 (13.36) 17.81 (13.84) 16.74 (14.98) 19.45 (13.46) 20.73 (13.46) 22.02 (15.17) a Lower SKAMP scores indicate greater efficacy CON = Concerta; MCD = Metadate CD; PLA = placebo. Low = Low Dose (CON 18 mg; MCD 20 mg), Med = Medium Dose (CON 36 mg; MCD 40 mg), Hi = High Dose (CON 54 mg ; MCD 60 mg). Comparison across dose levels The PLA adjusted scores for the total SKAMP score relating to the specific cross-dose comparisons are presented in Figures 1a through 1d. The number of patients included in specific analysis were as follows: MCD 20 vs. CON 36 – 58 vs. 53; MCD 40 vs. CON 54 – 55 vs. 47; MCD 40 vs. CON 18 – 55 vs. 57; and MCD 60 vs. CON 36 – 49 vs. 53. Figure 1 Mean (± SE) of placebo-adjusted total SKAMP scores for each comparison . Lower SKAMP scores indicate greater efficacy. Asterisks indicate MCD was significantly better than CON (p < 0.05) while crosses indicate CON was significantly better than MCD (p < 0.05). Lower daily doses of MCD than CON The comparison of MCD 20 mg and CON 36 mg revealed no overall difference between treatments, F (1,108) = 0.001; ns. There was an effect of session, F (6,648) = 15.08; p < 0.001 and an interaction between treatment and session F (6,648) = 3.98; p < 0.01. The two treatments did not differ at the time of dose administration or 3, 4.5, 6 and 7.5 hours after it (Fs < 2.5). MCD was superior to CON at 1.5 hours, F (2, 110) = 4.48; p < 0.05, while CON was superior to MCD at 12 hours, F (2,109) = 3.94; p < 0.05. The comparison of MCD 40 mg and CON 54 mg revealed no overall difference between treatments, F (1,99) = 1.05; ns. There was an effect of session F (6,594) = 36.20; p < 0.001 and an interaction between treatment and session, F (6,594) = 3.72; p < 0.01. There was no difference between treatments at the time of administration or at 1.5, 3, 4.5 and 6 hours (Fs < 3.58; ns). However, CON was superior to MCD at 7.5, F (1,100) = 6.26; p < 0.05, and 12 hours, F (1,99) = 4.55; p < 0.05. Lower daily dose of CON than MCD For the comparison of MCD 40 mg and CON 18 mg there was a main effect of treatment with MCD being associated with lower (i.e., better) SKAMP scores than CON, F(1,109) = 6.50, p < 0.05. There was also an effect of session, F(6,654) = 18.61, p < 0.001 and an interaction between treatment and session, F(6,654) = 8.32; p < 0.001. MCD gave lower (better) scores at 1.5 (F(1,111) = 10.02; p < 0.005), 3 (F(1,110) = 18.36; p < 0.001), 4.5 (F(1,111) = 16.16; p < 0.001), and 6 (F(2,111) = 5.94; p < 0.05) hours. However, there was no difference between the two formulations at 7.5 and 12 hours, Fs < 2.75. MCD 60 mg also gave significantly lower scores than CON 36, F (1,99) = 14.03; p < 0.001. Once again there was an effect of session, F (6,594) = 35.38; p < 0.001, and an interaction between treatment and session, F (6,594) = 10.55; p < 0.001. MCD gave significantly lower scores at 1.5 (F (1,100) = 35.01; p < 0.001), 3 (F (1,100) = 30.99; p < 0.001), 4.5 (F (1,99) = 18.23; p < 0.001) and 6 (F (1,100) = 9.66; p < 0.005) hours but not at 7.5 and 12 hours, Fs < 0.55. Discussion CON and MCD are once-daily formulations of MPH using different drug delivery mechanisms resulting in different plasma concentration vs. time profiles in both adults [ 13 ] and children [ 8 ]. In the COMACS study we compared the PD profiles of equivalent daily doses of the two formulations in children with ADHD [ 12 ]. The primary analysis was restricted to a comparison of the relative efficacy of the two formulations within subjects and dosing strata (i.e., across AUC-equivalent total daily doses). As predicted by the PK-PD model proposed by Swanson [ 19 ], MCD provided greater symptom control in the morning (from 1.5 through 4.5 hours post-dose), while CON gave greater control in the early evening (at 12 hours post-dose). However, on the basis of the IR/ER ratios of the two formulations, the expected plasma concentration vs. time profiles, and the informal observation of the relative efficacy of the different formulations made between subjects across dosing levels in the COMACS study, it was predicted that a lower total daily dose of MCD (i.e., 20 and 40 mg) would give equal levels of symptom control in the morning when compared to CON at the next highest dosing strata (i.e., 36 and 54 mg); the situation being reversed for CON in the later part of the afternoon and in the evening. Specifically, it was predicted that MCD 20 mg and 40 mg would give similar levels of symptom control to that provided by CON 36 mg and 54 mg, respectively, from 1.5 through 4.5 hours post dose. This would reflect the fifty percent higher relative proportion of IR delivery in MCD compared to CON. In contrast, the higher ER doses of CON would give significantly better control between 6 and 12 hours. There was only partial support for these predictions. Not only was there no significant difference between placebo adjusted total SKAMP scores for MCD 40 mg and CON 54 mg from 1.5 through 4.5 hours, but also at 6.0 hours post-dose. In addition, comparison at the lower dose levels (MCD 20 mg and CON 36 mg) gave a surprising result: MCD appeared to be associated with a stronger effect and a more rapid onset of action than CON during the very early period post-dose. This was indicated both by the absolute values (i.e., the intercept) for the two formulations at 1.5 hours post-dose and the steeper slopes for MCD than CON between 0 and 1.5 hours. This was a particularly unexpected finding because at the dose levels being compared, MPH available from the IR component of MCD (6 mg) was less than that available for CON (8 mg). Given that the active drug is the same in the two formulations and assuming that clinical efficacy reflects MPH serum concentrations as has been proposed by Swanson [ 19 ] this would suggest that non-drug related factors may be involved. Non-drug factors may include those associated with the speed of dissolution and absorption associated with the different delivery mechanisms of the two formulations. For instance, it may be that the IR beads of MCD are associated with a higher rate of dissolution than the CON MPH overcoat, although there is no reason to expect this. Alternatively, the rate of dissolution of the IR components of the two formulations may be equivalent but the distinction between the IR and ER components may be less clear-cut in MCD than CON. This would produce early exposure to MPH from some ER beads in addition to IR beads during the early part of the day. Dissolution studies support this by suggesting that while the ER components of MCD and CON both start to release MPH at approximately 1 hour post-dose, the MCD ER MPH component seems to be more front-end loaded, while CON's ER MPH component seems to be more back-end loaded. MCD drug delivery technology, may therefore, be more efficient at delivering IR components at low doses, and this requires further investigation. The predictions relating to the efficacy of these different dose comparisons during the latter part of the day were confirmed only at the 12 hour testing session for the lower doses (CON 36 mg vs. MCD 20 mg) and the 7.5 and 12 hours sessions for the higher doses (CON 54 mg vs. MCD 40 mg). Given their expected plasma concentration vs. time profiles in children, one would expect greater efficacy from CON than MCD whatever the dose comparison being made. The 7.5 hour post-dose period is less clear-cut in terms of the relative benefit of the two formulations. The second set of predictions related to the value of lower daily doses of CON (18 mg and 36 mg) to provide equivalent symptom coverage compared to higher doses of MCD (40 mg and 60 mg, respectively) in the late afternoon and superior coverage in the evening. The predictions for late afternoon were supported at both the higher and lower cross-dose comparisons with lower doses of CON giving equivalent control at 7.5 hours. However, interestingly, at 12 hours the higher doses of MCD remained equivalent to CON. This result is in keeping with the idea, supported by the expected plasma concentration and known dissolution data that MCD continues to release MPH up to 12 hours post-dose. As expected, the higher doses of MCD had a greater effect between dosing through 6 hours post dose. Given design limitations it is possible that these effects are not 'real' effects related to dose and treatment type but are related to differences between the types of children assigned to different dose levels. The way in which the children were assigned to dose level meant that across-dose comparisons of different treatments could be subject to a number of confounds. First, it could be that children were placed on higher doses because they had a more severe form of ADHD. In the present study we attempted to deal with the possible confounds that such an approach to dose assignment might bring by using the placebo SKAMP score (across all time periods) as a covariate. This score was included as a proxy for severity of ADHD in order to control for the possibility that children assigned to the higher-dose level had a more severe form of the condition. A second possible confound, not corrected for by controlling for placebo scores, relates to the child's sensitivity to MPH: Children may have been prescribed higher pre-study doses of MPH because they were less sensitive to MPH and did not respond to the lower dose. Such variations in sensitivity could be independent of overall severity of the disorder and therefore constitute a second possible confound. If this were the case then an across-dose comparison would be between more- and less-sensitive children. There are two pieces of evidence that argue against this as an explanation for the current results. First, Swanson et al. [ 12 ] reported that the between-subject factor of dose was significant on SKAMP Attention scores in the COMACS study. If dose levels were determined by MPH sensitivity such an effect would not have been expected. Second, there is no reason to believe that sensitivity to MPH should favor one formulation at one time (i.e., MCD in the morning) and the second formulation at another time (i.e., CON in the evening). Differential MPH sensitivity, therefore, seems an unlikely explanation for the current pattern of results. It is also possible that basing the study dose level on pre-study total daily dose may have had differential effects for children receiving TID and BID doses of IR preparations prior to the study. Children on, for instance, two tablets of 10 mg IR given early in the morning and then in the afternoon would be assigned to the low dose strata of 18 mg of CON thought to be equivalent to 5 mg IR tablets taken TID. In this case, children, although receiving AUC-equivalent daily doses, would be receiving a lower morning (5 mg rather 10 mg) and afternoon dose (5 mg rather 10 mg) and a new evening dose (5 mg as opposed to no dose) during the study relative to their pre-study MPH daily treatment. However, in the current study, 91% of children were on once-a-day dosing prior at prescreening, and it is therefore difficult to estimate the effects of this factor on the current across-dose comparison. Taken together, the existence of these confounding factors and complications means that while these data provide an initial indication of the relative efficacy of different doses of CON and MCD at different points across the day they should be treated with a certain degree of caution, and the results should be confirmed in a study where subjects are randomized to dose level rather than being assigned to it on the basis of their pre-study MPH daily dose and plasma levels of MPH are assessed simultaneously. Conclusions From the point of view of the practicing clinician the results of this study further highlight the value of having available different ER MPH formulations with different expected plasma concentration vs. time profiles offering different patterns of efficacy over different periods of the day. Given the established dose-response relationship between MPH and side effects, clinicians and parents may wish to limit total daily MPH intake as far as possible while maintaining a tolerable level of efficacy over the day as a whole and/or targeting a particularly important period of the day for a particular patient. This head-to-head study makes a first step towards providing the systematic evidence base on which to make such decisions. Competing interests EJB and DC are consultant for and/or have received support from Celltech, McNeil, Jannsen Cilag and Eli Lilly. JMS is a consultant for Celltech, McNeil and Eli Lilly. SJH and HHD are employees of Celltech Americas, Inc. Authors' contributions JMS, DC, SJH and EJB participated in the design of this post-hoc analysis while EJB performed the statistical analysis. HHD and EJB prepared the manuscript. All authors contributed discussion and read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC524494.xml
547899	Hinderin, a five-domains protein including coiled-coil motifs that binds to SMC3	Background The structural maintenance of chromosome proteins SMC1 and SMC3 play an important role in the maintenance of chromosomal integrity by preventing the premature separation of the sister chromatids at the onset of anaphase. The two proteins are constitutive components of the multimeric complex cohesin and form dimers by interacting at their central globular regions. Results In order to identify proteins that by binding to SMC3 may interfere with the protein dimerization process, a human cDNA library was screened by the yeast two-hybrid system by using the hinge region of SMC3 as bait. This has lead to the identification of Hinderin, a novel five domains protein including two coiled-coil motifs and sharing a strikingly structural similarity to the SMC family of proteins. Hinderin is ubiquitously expressed in human tissues. Orthologue forms of the protein are present in other vertebrates but not in lower organisms. A mapping of the interaction sites revealed that the N- and C-terminal globular domains mediate the binding of Hinderin to SMC3. Hinderin/SMC3 complexes could be recovered by immunoprecipitation from cell lysates using an anti-SMC3 antibody, thus demonstrating that the two proteins interact in vivo. On the contrary, Hinderin did not interact with SMC1. In vivo the rate of SMC1/SMC3 interaction was decreased by the ectopic expression of Hinderin. Conclusions Hinderin is a novel binding partner of SMC3. Based on its ability to modulate SMC1/SMC3 interaction we postulate that Hinderin affects the availability of SMC3 to engage in the formation of multimeric protein complexes.	Background The structural maintenance of chromosome (SMC) proteins are involved in several aspects of chromosomal dynamic, in DNA recombination and in DNA repairs [ 1 - 3 ]. Two SMC proteins named SMC1 and SMC3 bind to and prevent the premature separation of sister chromatids at the end of mitosis [ 4 , 5 ]. SMC1 and SMC3 directly interact through their central globular binding domains by forming an heterodimer [ 6 , 7 ]. The protein complex encircles the sister chromatids and is stabilized though the interaction with two other cohesin proteins known as Scc1 and Scc3 in s. cereviasie [ 7 , 8 ]. At anaphase, the ring-shaped complex is broken down when separase, a cysteine protease, cleaves Scc1, thus freeing the sister chromatids to move in opposite directions [ 6 , 9 ]. Somatic cells with deranged separase activity or lacking Scc1 develop aneuploidy at increased rate. This suggests that the cohesin complex plays a major role in the maintenance of chromosomal stability [ 10 - 13 ]. The mechanism regulating the interaction between SMC1 and SMC3 is still poorly understood. It has however been established that a single point mutation of the central globular domain (known as hinge) of either one of these proteins strongly affects the dimerization rate and prevents the attachment of the cohesin complex to chromatid DNA [ 7 ]. Conceivably, proteins that interact with the hinge domain of SMC1 or SMC3 can act as modulator of the cohesin complex formation and may thus affect chromosomal stability. In this paper we report the identification of a new SMC3-interacting protein that specifically binds to SMC3's central globular domain. The sequence of the identified gene product matched that of a previously discovered hypothetical new protein with no known function. We have named the protein Hinderin. The gene is expressed in all the human tissues analyzed thus far. Orthologue forms of this protein are expressed in vertebrates but not in lower organisms. Hinderin is a five-domain proteins and its structure resembles that of SMC proteins with N- and C-terminal globular domains that are joined by a coiled coil region interrupted at the center by a third globular domain. However, unlike the canonical SMC proteins, Hinderin does not harbor ABC-like ATPase sequences. We have found that the protein interacts with the hinge region of SMC3 but not with SMC1. Hinderin acts as a binding competitor of SMC1 and, as such, qualifies as a putative modulator of the SMC3 function. Results Identification of Hinderin, an SMC3-interacting protein with five domain structure including coiled-coil motifs The region of SMC3 encompassing the protein hinge domain (Gln 465 to Gln 807 ) was used as bait in a yeast two-hybrid system to identify interacting proteins expressed by a human fetal brain Matchmaker two-hybrid cDNA library (Clontech). About 3 × 10 6 library clones were screened. Forty blue colonies reaching 2 mm in size after one week were collected and 21 of the isolated plasmids with inserts greater than 500 bp were sequenced. Three of the sequences matched the same region of the published cDNA Genbank clones AB037749 (coding for the hypothetical protein KIAA1328) and AL832625 (corresponding to the hypothetical protein DKFZp451C1618). The inserts of ~2 kb included part of the gene 3'-UTR. However the 5'-end of the coding region was not present in the retrieved clones. The issue was complicated by the fact that the sequences of AB037749 and AL832625 diverged at their 5'-end. 5'-RACE was thus employed to identify the transcriptional start site of the interacting gene by using mRNA derived from fetal kidney 293, hepatoma HepG2, and cervical HeLa human cells. All the cloned sequence coincided with that of the AL832625 clone and matched in full the putative coding sequence obtained by automated computational analysis of the human genome (Genbank XM029429). The conceptually translated sequence coded for a protein of 578 amino acids. The exam of the secondary structure of the protein revealed a remarkable similarity with the structural organization of SMC proteins, particular with regard to the five-domain structure typical of the SMC protein family. The protein features C- and N-terminal globular domains, joined by a coiled-coil sequence interrupted in the middle by a third globular domain. The size of the central globular domain is similar to that of SMC3 and SMC1 but the remaining domains are smaller in size. Based on its ability to interact with the hinge domain of SMC3 and thus to potentially affect SMC3/SMC1 dimerization, the new protein was named Hinderin. The Hinderin gene spans ~400 kb of the human genome and is located on chromosome 18p11-2. The coding sequence is organized in 10 exons (fig. 1A ). Exon 10 is the largest and contains an extended 3'-UTR. Northern blot hybridization analysis performed on mRNA from HeLa and colon carcinoma HCT116 human cell lines, showed a single major transcriptional product of ~4.5 kb (fig. 1B ). A correlation between the exonic organization and the protein structural motifs was apparent (fig. 1C ). The N-terminal globular domain is encoded by exons 1–3. The first coiled-coil domain is encoded by exons 4–6. Exon 7 harbors the entire central globular domain, the central coiled-coil region, and part of the C-terminal globular domain. The polypeptide encompassing exons 3 to 6 displays a 36% sequence homology to the consensus sequence of SMC family of proteins. A bipartite nuclear localization signal (R 441 KERK) is located in exon 7 within a coiled-coil region. The Hinderin expression pattern in a panel of 16 human tissues was analyzed by semiquantitative RT-PCR (fig. 1D ). The results indicate that the gene is expressed ubiquitously with the highest expression level detected in the lung, liver, placenta, kidney, and pancreas. The lowest expression level was detected in leukocytes and the prostate. Figure 1 Hinderin: Genomic organization, structural domains, and expression pattern in human tissues. A) The sequence of the Hinderin ORF was used to BLAST the human genome database. The numeration of the exon boundaries is relative to the transcriptional start site. The size of the intronic sequences is based on the numeration of the human contiguous sequences. B) Northern blot hybridization of Hinderin in HeLa and HCT116 cells. A single main transcript of ~ 4.5 kb is identifiable. C) Prediction of coiled-coil domains in Hinderin. The numbers on the abscissa corresponds to the amino acid residues. The probability of the polypeptide to assume a coiled coil conformation is plotted on the ordinate axis. The contribution of the different exons to the five-domain structural organization of Hinderin is also illustrated. The sequence encoded by exons 3 through 6, bears a 36% homology to the SMC protein consensus sequence. A bipartite nuclear localization signal is harbored in exon 7. D) Expression of Hinderin in different human tissues. PCR amplification was stopped after 30 cycles and the product analyzed on 2% agarose staining with ethidium bromide. The G3PDH transcript was amplified in 20 cycles and used to show uniformity of the source cDNA in the different specimens. Mapping of the Hinderin and SMC3 interaction domains In order to map the region of Hinderin interacting with SMC3, AH109 yeast was cotransformed with the SMC3-465/807 bait and with different Hinderin constructs (fig. 2 ). Yeast cotransformants of SMC3-465/807 and the Hinderin plasmid retrieved from the two-hybrid library (H-47/578) grew rapidly. This result was confirmed by switching the bait and prey vectors. However, the H-47/578 construct did not interact with the N- and C-terminal domains of SMC3 (SMC3-1/186 and SMC3-976/1217 respectively) or with the hinge region of SMC1 (SMC1-485/670). A Hinderin construct harboring only the protein central globular domain (H-177/360) furthermore did not interact with the SMC3 bait, thus pointing to a substantial difference in the binding properties of the hinge domain of the SMC proteins and the corresponding domain of Hinderin. On the contrary, both the N- and C-terminal domains of Hinderin (H-1/85 and H-360/578) interacted with the SMC3 hinge domain. The assay of the β-galactosidase activity expressed by the yeast as a result of the protein-protein interaction provided a quantitative measure of the rate of the process. The strongest interaction was detected between SMC3-465/807 or SMC3-465/716 and H-47/578. The Hinderin constructs H-1/85 and H-360/578 displayed a lower interaction activity suggesting that the protein N- and C-terminal domains synergically bind to SMC3. An Hinderin construct (H-64/360), consisting of the central globular domain in addition to part of the protein N-terminal domain, displayed binding activity similar to that of the N-terminal globular domain alone (H-1/85). The protein central globular domain therefore does not affect the binding rate of Hinderin to SMC3. Truncation mutants of SMC3-465/807 were generated to map the region of SMC3 responsible for the interaction with Hinderin. When tested only SMC3-465/716 gave origin to colonies growing at the same rate as observed with the SMC3 bait (fig. 2 ). The remaining constructs, in which the SMC3 hinge region had been partially or completely deleted, produced no colonies. The intact SMC3 hinge domain is thus required for the interaction with Hinderin. Similarly the interaction between SMC3-465/807 and the hinge region of SMC1 (SMC1-485/670) gave strong reactivity whereas truncation mutants of SMC3 lacking part of the cohesin protein central globular domain (SMC3-552/807 and SMC3-465/643) did not interact. This finding is consistent with previous reports showing that SMC3 and SMC1 dimerization occurs through interaction of the terminal regions of the hinge domains of the two proteins [ 7 , 14 ]. Figure 2 Yeast two-hybrid assay of the interaction of different regions of SMC3 and SMC1 with Hinderin. A) SMC3 and Hinderin constructs in pGBKT7 are schematized on the left-hand side. All numerals refer to the amino acid sequence. The SMC3-465/807 construct harbors the protein hinge domain and was utilized as bait for the screening of the Clontech Matchmaker human cDNA library. SMC1, SMC3 and Hinderin constructs cloned in pACT2 or pGADT7 are illustrated on the right-hand side of the panel. H-47/577 represents the pACT2 clone retrieved from the screening. We scored the strength of interaction based on the rate of appearance of Blue colonies and the intensity of the color developed. The null score (-) was assigned when no blue colonies were visible after ten days. To obtain a quantitative measure of the rate of protein interaction, yeast colonies were grown in selective media and after lysis the β-galactosidase activity released in the supernatant assayed using a chromogenic substrate as detailed in the Methods. The results shown are the mean ± SD of three independent determinations. Hinderin interacts with SMC3 in vivo and is a binding competitor for SMC1 In order to investigate whether Hinderin associates with SMC3 in vivo, 293 cells were transiently transfected with 1 μg/ml of Hinderin-V5 expression vector and incubated 48 h. The transfected cells displayed three-fold elevation of the Hinderin transcript level but SMC1 and SMC3 expression was not affected (fig. 3A , V5-IP input lanes and fig 3B , RT-PCR results). After addition of anti-SMC3 or alternatively anti-SMC1 antibody to the cell lysate, the immunocomplexes were analyzed on SDS-PAGE followed by Western-immunoblotting using anti-V5 antibody (fig. 3A ). This revealed the presence of Hinderin-V5 in the SMC3 immunoprecipitate. On the contrary, Hinderin-V5 was not detected in the SMC1 immunoprecipitate, thus corroborating the results of the yeast two-hybrid system interaction experiments. Furthermore, SMC3 but not SMC1 could be recovered in Hinderin-V5 immunocomplexes. In order to examine whether the interaction between SMC1 and SMC3 is affected by Hinderin concentration in vivo, experiments were conducted in 293 cells expressing increasing levels of Hinderin and by assessing the rate of SMC1-SMC3 interaction using a mammalian two-hybrid system (Promega Checkmate). In this assay the interaction between the GAL4:SMC3 activating domain fusion protein and the VP16:SMC1-binding domain fusion protein activates the expression of the firefly luciferase encoded by a reporter vector (pGL5). The rate of interaction between SMC3 and SMC1 is thus directly proportional to the level of luciferase that is expressed. We found that transfection of 293 cells with increasing amount of the Hinderin-V5 expression vector caused a dose-dependent decrease of luciferase activity. In cells overexpressing Hinderin, the interaction between the cohesin proteins was gradually reduced down to ~34% (n = 3, p < 0.05) (fig. 3B ). Ectopic expression of Hinderin had no effect on the expression of the bait and prey fusion proteins as determined by RT-PCR, using primers encompassing the coding sequence of fusion proteins. The results are consistent with the notion that Hinderin competes with SMC1 for binding to SMC3, thereby negatively affecting the interaction between the two cohesin proteins. Figure 3 Protein coimmunoprecipitation and competitive binding of Hinderin to SMC3. A) To monitor the Hinderin interaction, 293 cells were transfected with 1 μg/ml of Hinderin-V5 expression vector and incubated for 48 h. Mock transfected cells served as control. For SMC1 and SMC3 coimmunoprecipitation experiments, cell lysates (500 μl) were incubated with either 25 μg of anti-SMC1 or anti-SMC3 antibody for 2 h at RT. The immunocomplexes were then captured on agarose-protein G and analyzed by electrophoresis on 12% SDS-PAGE. After transblotting to a nitrocellulose membrane, Hinderin-V5 was identified using an anti-V5 monoclonal antibody and HRP-conjugated secondary antibody. The migration position of the Hinderin-V5 fusion protein present in the input cell lysate (200 μl) and in the immunoprecipitate is indicated by an arrow. Goat immunoglobulins are identified by an asterix. The presence of SMC1 and SMC3 in the Hinderin-V5 immunoprecipitate was assessed by incubation of cell lysates (500 μl) with 10 μg murine anti-V5 antibody. The immunocomplexes absorbed on agarose-protein G were analyzed on 8% SDS-PAGE and immunoblotted with either anti-SMC1 or anti-SMC3 antibody. SMC1 and SMC3 immunoblots of the input material (200 μl) are also illustrated. B) The effect of Hinderin on the interaction between the SMC3 and SMC1 hinge domains was investigated in 293 cells expressing different level of Hinderin by using a mammalian two-hybrid assay system. SMC3-474/702 and the SMC1-474/663 hinge domains in pBIND and pACT vectors respectively, were cotransfected in 293 cells together with Hinderin-V5 expression vector (0.3 or 1 μg/ml) or alternatively 1 μg/ml of the empty pcDNA3.1 vector (mock-transfected) and the reporter pG5/luc vector using Lipofectamine. Cells transfected with pBIND-Id and pACT-MyoD fusion protein expression vectors were used as control to assess the specificity of the Hinderin effect on protein-protein interaction. Luciferase activity was assayed after 48 h. The bars represent the mean ± SD of the values (n = 3). Semiquantitative RT-PCR was employed to assess the transcript level of GAL4:SMC3 and VP16:SMC1 fusion proteins and of G3PDH in cells ectopically expressing different levels of Hinderin. Discussion In eukaryotic cells, the disjunction of the sister chromatids at the onset of anaphase is required to maintain chromosomal stability and to prevent aneuploidy [ 12 , 15 ]. Human somatic cells with unrestrained separase activity display retarded sister separation and increased rate of chromosomal breakage [ 10 ]. On the other hand, the overexpression of Scc1 in mouse fibroblasts inhibits the cells proliferation, implicating this protein and its complex with the other cohesin components in the control of mitotic cell cycle progression [ 16 ]. We have previously reported that overexpression of SMC3 has cell transforming potential [ 17 ] and we have determined that SMC3 level is controlled in intestinal epithelial cells through the APC/β-catenin/TCF4 transactivation pathway [ 18 ], a signaling system that is almost invariably altered in colon carcinomas. These findings support the idea that alteration of the level of the components of the cohesin complex has important consequences that may trigger a tumorigenic cascade. A key event in the formation of the cohesin multimeric complex is the dimerization of SMC1 with SMC3 which occurs through the interaction between the two proteins central globular domains. The hindrance of this process is likely to have a significant effect on the formation and the function of the cohesin complex and on the maintenance of a stable chromosome population. Given this postulate, we have screened a human cDNA library to identify proteins interacting with the hinge region of SMC3. Clones coding for Hinderin accounted for 15% of those identified in this screening. Interaction between Hinderin and SMC3 has been further confirmed by mapping the site of binding, in co-immunoprecipitation experiments using cell lysates, and in a two-hybrid mammalian system. Furthermore we have determined that SMC1 and SMC3 interaction rate is inversely related to the level of Hinderin expressed by mammalian cells. Hinderin displays the same five-domain structural organization of the SMC family [ 19 ]. However, the central globular domain of the protein does not appear to be involved in the binding with SMC3. Protein-protein interaction studies with a set of truncation constructs are rather consistent with the conclusion that SMC3 specifically interacts with the Hinderin terminal globular domains. An interaction model that is plausible with the protein-protein interaction results is illustrated in fig 4 . The model predicts the occupancy of the hinge region of SMC3 by the two terminal globular domains of Hinderin that come in close proximity by virtue of the flexibility of the protein central globular domain. In SMC1 and SMC3 the coiled-coil domains are in antiparallel orientation and by interacting they allow the N- and C-globular domains to join and form a functional ATPase head that interacts with Scc1 [ 19 ]. Given the similarity with the structural organization of the SMC proteins and the interaction modality with SMC3, the Hinderin coiled coil domains might contribute to the binding of the N- and C-terminal globular domains to SMC3. Figure 4 Proposed model of interaction of Hinderin with SMC3. A) The five-domain structure of Hinderin is compared to that of SMC3. The different structural domains are drawn in scale to allow a direct comparison of the size of the terminal globular regions, the two coiled-coil domains and the central globular domain in the two proteins. B) Mode of interaction of the SMC1-SMC3 dimer. The juxtaposition of the cohesins hinge domains interacting through sites located at the globular-coiled coil domain boundaries (ref. 6) is illustrated. C) Postulated mechanism for the competitive binding of Hinderin to the hinge domain of SMC3. The N- and C-terminal globular domains of Hinderin are shown to interact with binding sites located on the SMC3 hinge. The Hinderin central globular domain is not involved in the binding to SMC3 but may play a role by orienting the N- and C-terminal globular domains toward their targets. Conclusions In summary, we have identified a novel interacting partner of SMC3. The protein, named Hinderin, specifically interacts with the hinge domain of SMC3. The protein is ubiquitously expressed in human tissues. We speculate that when in a certain context, SMC3 association with Hinderin becomes favored compared to the association to SMC1, the availability of SMC3 to engage in the cohesin complex formation is reduced. The binding of SMC3 to proteins affecting its association to functional partners represents a new modality of regulation of SMC3 activity. Methods Screening of a human cDNA library by yeast two-hybrid system In order to identify proteins encoded by the human fetal brain cDNA library and interacting with the hinge domain of SMC3, a bait plasmid was generated by subcloning the SMC3-465/807 polypeptide coding region into the yeast two-hybrid system bait vector pGBKT7 (Clontech) (see fig. 2 for a diagram of the constructs used and their designation). The insert was generated by PCR using Pfu DNA polymerase and primers terminated with restriction sites that allowed the directional cloning of the products into the accepting vector (see Additional file 1 for a listing of the primers used in this study). Mouse full-lenght SMC3 cDNA was used as template. To identify the gene encoded by the interacting plasmid, the prey insert was sequenced by priming at the Gal4AD site (5'-AATACCACTACAATGGA-3'). The sequences were BLASTed against the nr and human dEST databases. DNA restriction digestions provided information on the size of the clones retrieved. 5' RACE and cloning of the complete Hinderin coding sequence The total RNA was isolated using TRI-reagent from 293, HepG2 and HeLa cells. The 5' RACE assay was performed using a RLM-RACE Ambion kit. The generated cDNA was used as template for nested PCR. The inner PCR product was cloned into pCRII-Topo vector (Invitrogen) and sequenced using a T7 primer. The sequence matched that of the DKFZp451C1618 clone (AL832625) and extended 5' to the published sequence of KIAA1328 (AB037749). Based on this information, the complete Hinderin coding region was generated by RT-PCR utilizing mRNA extracted from 293 cells, and the product cloned in frame with the tag sequence in pcDNA3.1/V5-His TOPO (Invitrogen). When transfected into 293 cells, the expressed product detected with an anti-V5 monoclonal antibody had size 69 KDa, as expected for a protein encoded by the Hinderin-V5 fusion gene (fig. 3A ). Mapping of the protein interacting sites In order to map the SMC3 interacting site(s) a series of truncated constructs were produced in pGBKT7 vector. The inserts required for the SMC3-1/186, SMC3-976/1217, SMC3-552/807 and SMC3-711/807 constructs were produced by PCR. SMC3-465/550 and SMC3-465/716 were generated by introducing stop codons in the sequence of SMC3-465/807 using a QuikChange XL kit (Stratagene). The mutated duplex oligonucleotides used had the forward sequence: 5'-CTTTCTATACTTGTGT AAGTCACTGCTGGTAAC-3' and 5'-GACCAGTTGATGAACTAAATGCAGATAGAG-3', respectively. SMC3-465/643 and SMC3-643/807 were obtained by digesting SMC3-465/807 at the Pst I or alternatively the Nde I restriction sites present in the vector multiple cloning site and at the Sma I site of the insert. The resulting linear constructs were blunt-ended and religated. pGADT7-SMC3-465/807 was generated by retriving the insert from the bait vector. SMC1-485/670 encoding the entire SMC1 hinge region, was generated by RT-PCR from 293 cells mRNA and cloned in pGADT7. Hinderin deletion constructs H-64/360 and H-360/578 were generated by restriction digestion of the pACT2-Hinderin-47/578 clone identified in the yeast two-hybrid system screening with Nco I/ Bgl II and Bgl II/ Xho I respectively, followed by religation of the blunt-ended vector. Hinderin in bait pGBKT7 vector was generated by retrieving the H-47/578 insert from the prey clone. H-177/360 and H-1/85 inserts were obtained respectively by PCR and by digestion of the Hinderin pcDNA3.1 expression vector with BamH I/ EcoR I. Both were subsequently subcloned in pGADT7. To assess the strenght of interaction between different SMC3 and Hinderin domains, three colonies were randomly selected and grown overnight in 5 ml of selection media. After cell lysis by freeze thawing in 300 ml of 100 mM Na 2 HPO 4 pH 7.0, 10 mM KCl, 1 mM MgSO4, buffer, β-galactosidase activity was assessed using ortho-nitrophenyl-β-D-galactopyranoside as substrate (1 mM) in the presence of 50 mM β-mercaptoethanol. After 2 h incubation at 37°C the color intensity was read at 420 nm. Northern blot hybridization and semiquantitative PCR Total mRNA was extracted with TRI-reagent from subconfluent HeLa and HCT116 cells and separated on 1% agarose. After transfection to a nitrocellulose filter, the Hinderin transcript was identified by hybridization to a 335 bp 32 P-labeled cDNA probe annealing to the 5'-end region of the gene. In order to examine the expression of Hinderin in different human tissues, semiquantitative RT-PCR was performed by using 0.5 μg of Marathon-ready first strand cDNA (Clontech) from 16 tissues. The PCR reaction was monitored at 20 and 30 cycles and the product analyzed on 2% agarose. In order to normalize for possible differences in mRNA content, the expression of the housekeeping gene G3PDH was analyzed in each sample. Protein secondary structure prediction and identification of orthologue forms of Hinderin in other species The human polypeptide sequence was analyzed with the COILS program to predict globular and coiled-coil domains. The scanning window was set at 21. The homology with other known protein family was examined by querying the NCBI Conserved Domain protein database. We used PSORT to scan for nuclear and other localization signal consensus sequences. To identify orthologue forms of Hinderin in other species, the human protein sequence was BLASTed against the translated EST database of m. fascicularis, mouse, rat, cow, sheep, dog, zebrafish, c. elegans, drosophila, and s. cereviasie. When a homologous sequence had been identified in lower organisms, we ran the COILS program to assess whether it displayed the same secondary structure as that of the matching human sequence. Protein complex immunoprecipitation 293 cells were grown at ~80% confluence in 35 mm cm plates and transfected with 1 μg of Hinderin-V5 expression vector. After 48 h of incubation, the cells were washed in ice-cold phosphate-buffered saline and lysed in 1.2 ml of 50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 1% Nonidet P40, 0.5% Na-deoxycholate buffer containing 100 mM NaF, 2 mM Na 3 VO 4 , and a cocktail of protease inhibitors. The cell lysate was centrifuged at 12,000 g and the recovered supernatant preabsorbed on protein G-agarose. Aliquots (500 μl) of the cell lysate were then incubated overnight at 4°C with either 25 μg of goat anti-human SMC3 antibody or goat anti-human SMC1 antibody (Santa Cruz Biotech) or alternatively with 10 μg mouse anti-V5 antibody (Invitrogen). The immunocomplexes were captured on Protein G-agarose by incubating 1 h at 4°C. After washing in immunoprecipitation buffer containing 300 mM NaCl, the protein immunocomplex was analyzed by SDS-PAGE and the proteins transferred to nitrocellulose membranes by electroblotting. After saturation in 4% dry milk/0.1% Tween 20 in PBS, the filter was incubated 1 h at RT with primary antibody. The anti-SMC1 and anti-SMC3 immunoprecipitate filter blots were incubated with anti-V5 monoclonal antibody (200 ng/ml) whereas the V5 immunoblots were incubated with either anti SMC1 (100 ng/ml) or anti-SMC3 (100 ng/ml) antibodies. After incubation with species-specific anti IgG HRP-conjugated secondary antibody (1:10,000), the immunocomplexes were visualized by enhanced chemiluminescence reaction (ECL). Mammalian two-hybrid interaction assay Inserts corresponding to the SMC3-474/702 and SMC1-474/663 hinge domains were generated by PCR and ligated in the pBIND or the pACT vectors (Promega) respectively through ligation at the BamH I and Sal I sites. The resulting bait and prey DNA constructs (0.25 μg/ml each) together with the Hinderin-V5 expression vector (either 0.3 or 1 μg/ml or alternatively 1 μg/ml of empty pcDNA3.1 vector) and the reporter plasmid pG5/luc encoding the firefly luciferase (0.1 μg/ml), were co-transfected into 293 cells using Lipofectamine. Control experiments were conducted using pACT-Id and pBIND-MyoD vectors (Promega) encoding respectively GAL4:Id and VP16:MyoD fusion proteins known to interact in vivo [ 20 ]. After 48 h incubation, cells were lysed and the expressed luciferase activity quantitated using a dual luciferase reporter assay kit (Promega) and a Lumat LB 9501 luminometer. Firefly luciferase values were corrected for the transfection efficiency using the values of the Renilla luciferase activity encoded by the pBIND vector under the control of a strong constitutive promoter. In order to assess the effect of Hinderin on the bait and prey expression, total RNA was extracted from a group of transfected cells and the specifc transcript levels quantitated by RT-PCR using primers annealing to the ends of the fusion protein coding sequence. Authors' contributions CAP carried out the two-hybrid system experiments including the immunoprecipitation studies, generated the necessary DNA constructs, and performed the gene expression analysis. GG conceived and coordinated the studies and drafted the manuscript. Supplementary Material Additional File 1 Primer and adapter sequences oligonucleotide sequence Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC547899.xml
514557	Does preoperative abduction value affect functional outcome of combined muscle transfer and release procedures in obstetrical palsy patients with shoulder involvement?	Background Obstetric palsy is the injury of the brachial plexus during delivery. Although many infants with plexopathy recover with minor or no residual functional deficits, some children don't regain sufficient limb function because of functional limitations, bony deformities and joint contractures. Shoulder is the most frequently affected joint with internal rotation contracture causing limitation of abduction, external rotation. The treatment comprises muscle release procedures such as posterior subscapularis sliding or anterior subscapularis tendon lengtening and muscle transfers to restore the missing external rotation and abduction function. Methods We evaluated whether the preoperative abduction degree affects functional outcome. Between 1998 and 2002, 46 children were operated on to restore shoulder abduction and external rotation. The average age at surgery was 7.6 years and average follow up was 40.8 months. We compared the postoperative results of the patients who had preoperative abduction less than 90° (Group I: n = 37) with the patients who had preoperative abduction greater than 90° (Group II: n = 9), in terms of abduction and external rotation function with angle measurements and Mallet classification. We inquired whether patients in Group I needed another muscle transfer along with latissimus dorsi and teres major transfers. Results In Group I the average abduction improved from 62.5° to 131.4° (a 68.9° ± 22.9°gain) and the average external rotation improved from 21.4° to 82.6° (a 61.1° ± 23°gain). In Group II the average abduction improved from 99.4°to 140°(a40.5° ± 16°gain) and the average external rotation improved from 33.2°to 82.7° (a 49.5° ± 23.9° gain). Although there was a significant difference between Group I and II for preoperative abduction (p = 0.000) and abduction gain in degrees (p = 0.001), the difference between postoperative values of both groups was not significant (p = 0.268). There was also no significant difference between the two groups in the preoperative external rotation, the external rotation gain and the postoperative external rotation (p = 0.163, p = 0.181 and p = 0.803, respectively). Conclusions Obstetric palsy patients with shoulder sequela who had a preoperative abduction less than 90°hadas good functional results using latissimus dorsi, teres major muscle transfer and subscapularis muscle release as the patients who hada preoperative abduction greater than 90°.	Background Although the extent and severity of the deformity in obstetrical palsy may vary from patient to patient, shoulder is the most frequently affected joint. Many clinical and radiological classification systems had been described to address the problem and tailor the solution at the shoulder [ 1 - 5 ]. Among deformities affecting this joint, internal rotation contracture causing limitation of abduction and external rotation and dislocation of the shoulder is commonly observed on the follow up [ 4 ]. The shoulder instability was not caused directly by obstetrical trauma but related to a dynamic phenomenon of muscle imbalance. It was discovered that subscapularis muscle usually recovers more quickly than the external rotators and abductors [ 6 , 7 ]. Despite initial conservative therapy, patients with partial recovery may develop internal rotation and adduction contractures at the shoulder. If this persists long enough, flattening of the humeral head and joint incongruence will ensue. Under these conditions the main goal of the treatment is to reestablish the muscle equilibrium. The first step is to treat the muscle release such as the posterior subscapularis slide or the modified anterior subscapularis tendon lengtening procedures. Afterwards, muscle transfer is performed to restore the missing external rotation and abduction function in patients with congruent glenohumeral joint and limited glenohumeral deformity. However external rotational humeral osteotomy is preferred in patients with similar external rotation weakness, internal rotation contracture but with severe glenohumeral deformity [ 1 ]. Usually release and transfer procedures are seperated depending on the age of the patient and condition of the shoulder but in older age group (>4 years), it is preferred to combine the shoulder release and muscle transfer simultaneously [ 6 - 8 ]. Some authors perform other muscle transfers like levator scapulae and trapezius along with latissimus dorsi and teres major only for total flail shoulders [ 9 ], others claim that if the preoperative abduction degree of the shoulder is less than 90° indicating a week deltoid and external rotator muscles, a trapezius muscle transfer must be added to latissimus dorsi and teres major transfers to achieve significant improvement of both abduction and external rotation [ 10 ]. In this study, we compared the results of combined release and tendon transfer operations performed at the same stage because of the late presentation of the cases, for shoulder abduction and external rotation in two groups of patients who did not have any surgical treatment before. In one group preoperative abduction degree was less than 90°, while in the other group it was equal or more than 90°. We compared the gains of abduction and external rotation in two groups and evaluated whether the preoperative abduction degree affects the functional outcome and in patients with preoperative abduction degree <90° and also another muscle transfer is needed along with latissimus dorsi and teres major transfers. Methods General data about the patients Between 1998 and 2002, 46 children with obstetrical brachial plexus palsy who had no surgical treatment before (30 male and 16 female) were operated to restore shoulder abduction and external rotation in our clinic. The average age at surgery was 7.6 years (3–16). The patients had an average follow up of 40.8 months (range 24 to 60 months). Involvement of right side was seen in 30 patients and the left side in 16 patients. No patient had bilateral involvement. All of the patients were vaginally delivered with vertex presentations. Obstetrical history revealed that most mothers were multiparous and 8 of the patients were delivered with the help of forceps or vacuum. The mean birth weight of the patients was 4.5 kg (3–6.6 kg). The pool of the patients consisted 12 of patients with C5–C6 spinal root involvement, while 11 of the patients had additional C7 involvement. Finally 23 of the patients had total brachial plexus roots involvement. Accompanying birth complications were fracture of clavicle in one case, injury of sternocleidomastoideus muscle in one case and Horner's Syndrome in 2 cases. Table 1 [see Additional file 1 ] summarizes the specific qualifications, preoperative and postoperative evaluation values of all patients. We compared the postoperative results of the patients who had preoperative abduction less than 90° [Group I: n = 37, mean abduction 62.5° (20°–85°) and mean external rotation 21.4° (0–80°)] with the patients who had preoperative abduction equal or more than 90° [Group II: n = 9, mean abduction 99.4° (90–110) and mean external rotation 33.2° (0–65)], in terms of abduction and external rotation function with angle measurements and Mallet classification. The statistical analysis was performed with analysis of variance (ANOVA) and Student's t-test. Statistical significance was presumed at p < 0.05. Preoperative and postoperative evaluation Preoperative and postoperative active and passive range of motion degrees of abduction and external rotation were measured, videos were recorded during shoulder abduction and external rotation and Mallet scores (Figure 1 ) were noted. Abduction degree is measured at standing position and external rotation is measured in prone position with 90° shoulder abduction and at 90° elbow flexion. Radiography of the shoulder in adduction and 90° abduction and axial magnetic resonance imaging of the shoulder was performed. Shoulder deformity was classified according to Waters-Peljovich grading system, Table 2 , and the patients with type I and type II deformities were included in the series [ 1 ]. Most of the times, preoperative physical examination revealed weakness of the deltoid muscle and external rotators as well as co-contraction of pectoralis major, latissimus dorsi and teres major muscles at the anterior and posterior margin of the axillary fossa during shoulder elevation specially in Group I patients. Operative technique In our series we used a technique similar to the Hoffer technique [ 8 ]. We observed tightness and/or hypertrophy of latissimus dorsi, teres major, subscapularis and sometimes pectoralis major muscles intraoperatively. We performed subscapularis muscle release and the latissimus dorsi and teres major muscle transfer at the same session since the mean age at surgery was 7.6 years. Patients were placed in the lateral decubitus position and conjoined tendon of latissimus dorsi and teres major was explored with a posterior zigzag incision parallel to the lateral border of scapula to prevent scar contracture (Figure 2 ). Extensive dissection of the conjoined tendon and the related muscles from the surrounding structures was needed so that the conjoined tendon could reach to its new insertion easily. While preparing the muscles for transfer care must be taken not to injure the pedicles, which were located at medial side of these muscles (Figure 3 ). The next step was detachment of the conjoined tendon from its insertion on the inner side of the humerus by retracting the neurovascular structures of the arm superiorly. Afterwards a tunnel was prepared between long head of triceps and deltoid muscle with an extensive care to prevent injury to the axillary nerve. Rotator Cuff Quick Anchor ® Plus (Johnson & Johnson) which had size 2 green ethibond polyster sutures on it, was applied at the insertion point of infraspinatus muscle on the rotator cuff (Figure 4 ). The suture material was transferred to the posterior incision from the tunnel and the conjoint tendon was interwoven with this suture material. Then conjoined tendon was transferred to the posterior deltoid incision. Finally, reinsertion of the conjoined tendon to humerus was achieved while the arm was at 90° abduction and full external rotation (Figure 5 ). Fractional tenotomy or Z plasty procedure was applied to the pectoralis major tendon, if necessary, through an anterior axillary incision or by retracting posterior incision superiorly. Subscapularis muscle was released from the anterior surface of the scapula subperiostally from a small incision at the lateral side of the scapula carefuly, in order to prevent injury to the pedicle of lattisimus dorsi and teres major muscles (Figure 6 ). A cast, stabilizing the shoulder at 90° abduction and 90° external rotation and elbow at 90° flexion was applied for five weeks and physiotherapy was started after the removal of the cast under the control of custom-made splint. Results The average abduction degree improved from 62.5° (20°–85°) to 131.4° (90°–165°) and the average external rotation degree improved from 21.4° (0°–80°) to 82.6° (30°–95°) in Group I. We obtained 68.9° ± 22.9° (109%) gain for abduction and 61.1° ± 23° (285%) gain for external rotation. The difference between preoperative and postoperative values of abduction and external rotation was significant (ANOVA, F = 265 p = 0.000, F = 201 p = 0.000, respectively). The average abduction degree improved from 99.4° (90°–110°) to 140° (110°–170°) and the average external rotation degree improved from 33.2° (0°–65°) to 82.7° (45°–90°) in Group II. We obtained 40.5° ± 16° (40%) gain for abduction and 49.5° ± 23.9° (149%) gain for external rotation. The difference between preoperative and postoperative values of abduction and external rotation was significant (ANOVA, F = 25 p = 0.000 and F = 32.3 p = 0.000, respectively). The results were summarized in Table 3 . Although difference between preoperative abduction and abduction gain values in terms of degrees of Group I and Group II were significant (ANOVA, F= 43.1 p = 0.000 and F = 12 p = 0.001, respectively), the difference between postoperative values of both groups was insignificant (ANOVA, F = 1.257 p = 0.268). There was also no significant difference between the preoperative external rotation, the external rotation gain and the postoperative external rotation for both groups (ANOVA, F = 2.017 p = 0.163, F = 1.848 p = 0.181 and F = 0.063 p = 0.803, respectively). The results of both groups were shown in graphics in Figure 7 . Mean Mallet scores increased from 2.8 to 3.9 for global abduction, from 2.5 to 3.9 for global external rotation, from 2.1 to 3.6 for hand to head and from 2.5 to 3.5 for hand to mouth for Group I. Hand to back Mallet score decreased from 2.5 to 2.2. All the differences between preoperative and postoperative values were found significant according to the Student's t-test (t = -17.14 p = 0.000 for abduction score, t = -12.04 p = 0.000 for external rotation score, t = -13.06 p = 0.000 for hand to head score, t = 2.372 p = 0.023 for hand to back score and t = -7.361 p = 0.000 for hand to mouth score). Mean Mallet scores increased from 3.5 to 4 for global abduction, from 2.8 to 4 for global external rotation, from 2.7 to 4 for hand to head and from 3.2 to 3.5 for hand to mouth for Group II. Hand to back Mallet score decreased from 2.8 to 2.1. All the differences, apart from hand to mouth score, between preoperative and postoperative values were found significant according to the Student's t-test (t = -2.53 p = 0.035 for abduction score, t = -3.592 p = 0.007 for external rotation score, t = -4.4 p = 0.002 for hand to head score and t = 2.8 p = 0.023 for hand to back score). The difference between preoperative and postoperative hand to mouth scores was found insignificant (t = -2, p = 0.081). While the difference between preoperative Mallet scores of Group I and Group II concerning abduction, hand to head and hand to mouth was significant (ANOVA, F = 23.211 p = 0.000, F = 11.407 p = 0.002 and F = 7.692 p = 0.008, respectively), the difference was insignificant for external rotation and hand to back scores (ANOVA, F = 1.393 p = 0.244 and F = 2.475 p = 0.123, respectively). The difference between postoperative values of both groups was insignificant (ANOVA, F = 0.239 p = 0.627 for abduction, F = 0.76 p = 0.388 for external rotation, F = 2.764 p = 0.106 for hand to head, F = 0.354 p = 0.555 for hand to back and F = 0.363 p = 0.555 for hand to mouth). The results were summarized in Table 4 . Example cases from each group can be seen in Figure 8 and 9 . Discussion Internal rotation contracture is the most frequent and important secondary deformity of the shoulder in birth palsy. The problem is sometimes addressed by muscle release procedures such as the posterior subscapular slide or an anterior subscapularis tendon lenghtening operations. Once passive external rotation is improved, the child is later assessed for muscle transfers to reconstruct active external rotation if necessary [ 4 ]. According to Chang et al [ 5 ] there are two types of residual muscle impairment after recovery in the late obstetric brachial plexus palsy: motor recovery with cross-innervation and paralysis or paresis. Contractures of the pectoralis major, teres major, brachialis and biceps muscles, which are most frequently observed, cause the deformity of the shoulder and elbow. The reconstructive strategy include releasing of the antagonistic muscles (elongation of the pectoralis major and latissimus dorsi muscles) and augmentation of the paretic muscles (teres major transfer to the infraspinatus muscle for augmentation of shoulder external rotation and abduction and reinsertions of both ends of the clavicular part of the pectoralis major laterally for deltoid augmentation). However, there are still many controversies concerning donor muscle choice for transfer, timing and operative tecniques of palliative surgical theraphy for the shoulder deformity. The infraspinatus muscle works to center the humeral head in the glenoid throughout elevation. External rotation of the shoulder allows greater arm elevation by clearing the greater tuberosity from impingement by the coracoacromial arch. External rotation of the humerus also positions the long head biceps centrally to aid in its function as humeral stabilizer and loosens the inferior glenohumeral ligaments, thereby allowing greater arm elevation. Hence the infraspinatus muscle plays a key role in shoulder elevation as a humeral head stabilizer, an active abductor, and an external rotator of the shoulder [ 6 ]. The importance of transferring the teres major and latissimus dorsi as one conjoined tendon and anchoring into the posterior aspect of the greater tuberosity at the insertion of the infraspinatus similar to Hoffer method is augmentation of the weakened infraspinatus. Transfer with this technique instead of rerouting around humeral neck enables a stronger external rotator power because of the increased mechanical advantage at its insertion in the humeral head as opposed to the humeral shaft. The reason for the dramatic improvement of shoulder abduction after latissimus muscle transfer is probably because the transfer enhances the stabilizing effect of the rotator cuff which enables the deltoid to act more effectively, this phenomenon was called "force couple" effect by Phipps and Hoffer [ 11 ]. In many centers, muscle release procedures are performed before the age of two years, however for older children tendon transfer to restore abduction and external rotation is added [ 12 ]. It is accepted that the corrective procedures to rebuild the muscle equilibrium are best undertaken before permanent bony deformity occurs at 3 to 4 years of age [ 13 ]. Gilbert [ 14 ] suggested that release of the subscapularis is indicated if the external rotation does not improve more than 20°. Based on his 5 years of follow up, he reported excellent results after subscapularis release especially in patients before the age of 2 years. Raimondi also waits for the active external rotation due to the reinforcement of weak external rotator muscles after subscapular muscle release procedure in early ages but since recovery of the external rotators cannot be expected, he preferres the tendon transfer and muscle release operations at the same time in children older than 4 years of age [ 9 ]. Muhlig et al [ 12 ] described a common policy accepted by most of the centers. According to this; if passive external rotation of the shoulder stays < 30°, surgical treatment is indicated. If there is no posterior displacement of the humeral head than a subscapular slide will be used. However, if there is posterior displacement of the humeral head than subscapular lengthening by an anterior approach will be preferred. Indications for tendon transfer for improving external rotation and abduction are determined as well. If infraspinatus muscle does not show signs of reinnervation by the age of 2 years, a muscle transfer should be added to the subscapularis lengthening to avoid recurrence. If there is a fixed medial rotation contracture and posterior luxation of the humeral head with deformities of the glenoid than derotational osteotomy of the humerus should be added to the subscapularis lengthening. As all of our patients were older than 2 years of age, we performed latissimus dorsi and teres major transfer at the same session with subscapularis and pectoralis major release. In total flail shoulders, despite a certain degree of innervation, the functional results of shoulder corresponds to zero with the absence of a strong latissimus dorsi. In that condition, the levator scapulae muscle is utilised as an intrinsic stabilizer of glenohumeral joint and trapezius muscle is used as a prime mover for shoulder abduction with or without latissimus dorsi and teres major transfers [ 9 ]. Gilbert in his series of 44 patients with transfer of latissimus dorsi, the improvement of abduction was satisfactory in the shoulders which preoperatively coded as Grade III (Shoulder abduction is between 90°–120°, external rotation is between 0°–30°) or more, but not in those coded as Grade II (Shoulder abduction is between 45°–90°, external rotation is to neutral) or less. Hence he thought it may be necessary to add a concomitant transfer of the trapezius to the patients whose abduction of the shoulder was weak or absent [ 15 ]. Chen et al [ 10 ] asserted the need for an additional trapezius muscle transfer for shoulder of the patients who had less than 90° abduction to increase the success of the classic latissimus dorsi + teres major transfer. They transferred latissimus dorsi by fixing its tendon to the insertion of the infraspinatus and tenotomized the teres major and then attached to the belly of the latissimus dorsi and found out in their early stage of treatment that, 10 of 18 cases with abduction less than 90°, with transfer of the latissimus dorsi and the teres major, patients gained no improvement of abduction but some recovery of external rotation, while five of seven patients with abduction equal or more than 90°, made significant progress in both abduction and external rotation. Al-Qattan [ 16 ] performed latissimus dorsi transfer on 12 children with variable preoperative shoulder abduction (range 60–150°, mean 100°) and postoperatively ten children achieved a modified Mallet score of 4 and were able to reach the occiput easily and they had mean 140° active shoulder abduction (range 90–170°). Hence the author also did not find any difference in patients with weak or strong preoperative abduction. It is our opinion that in our Group I patients, the co-contraction between shoulder abductors (supraspinatus, infraspinatus and deltoid) and adductors (mainly, pectoralis major, teres major and latissimus dorsi) and also subscapularis muscle tightness cause limitation of shoulder elevation. If antagonistic muscles (teres major and latissimus dorsi) are transferred for the paretic muscles (infraspinatus) and the pectoralis major and subscapularis muscles are released with preserving their shoulder stability function, these children can have as succesfull postoperative shoulder abduction and external rotation as the children in Group II, who has less cross innervation hence better preoperative abduction value. Group I and II patient had almost similar postoperative mean abduction (131.4° & 140°, respectively) and external rotation values (82.6° & 82.7°, respectively). Extensive dissection of latissimus dorsi and teres major muscles from the surrounding structures gave us the opportunity to utilize both muscles for transfer, without any difficulty during the passage of the conjoined tendon through the tunnel which was prepared between long head of triceps and deltoid muscle, and also during reinsertion to the humerus. Several authors reported recurrences of the deformity in terms of reduction of external rotation and abduction gain. Two of the 12 children in Al-Qattan series [ 16 ] and three of 35 cases in Phipps and Hoffer series [ 11 ] had recurrence of the deformity. Al-Qattan [ 16 ] classified the possible cause of this late complication as recurrence of the internal rotation contracture (mainly in the subscapularis), gradual contracture of the teres major as part of the inferior glenohumeral angle contracture and co-contraction of the muscles. We did not have any recurrence of the deformity during the follow-up period which may be related to the use of rigid fixation with bone anchors for reinsertion of the conjoint tendon. We believe that in cases who has congruent glenohumeral joint (Type I-III Waters-Peljovich grading system), and deltoid muscle strength of M3–M4 (British Medical Research Council evaluation) but weak or absent external rotation, if the latissimus dorsi and teres major muscles have sufficient strength (M3 or more), the ideal procedure is transfer of latissimus dorsi and /or teres major onto the posterior aspect of the greater tuberosity of humerus, at the insertion of the infraspinatus. So we are not totally convinced about adding trapezius muscle transfer concomitantly with the latissimus dorsi + teres major transfer session. We rather preserve this muscle for the patients that could not achieve enough shoulder abduction after the first operation. Conclusions Since our study was not randomised to treatment, the groups did not comprise equal number of patients, the assessments were not performed by independent observers but by a physiotherapy group of our team at postoperative 24 – 60 months (not at the same time for everyone), we would like to interpret our conclusions cautiously. Almost near normal shoulder function can still be reached in children who could not receive primary early neural reconstruction, by combined muscle release and muscle transfer operations which are performed before the severe glenohumeral deformities occur. We found out that the patients with obstetric palsy shoulder sequela who had a preoperative abduction value less than 90° could have good functional results as the patients who had preoperative abduction values equal or more than 90°, with latissimus dorsi, teres major muscle transfer and subscapularis muscle release. Competing interests None declared. Authors' contributions All authors participated in the design of the study, operations and drafting the manuscript. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 Preoperative and postoperative range motion degrees and Mallet scores of the patients. Patients with bold numbers are in Group II with preoperative abduction values ≥ 90° and the others are in Group I with preoperative abduction values < 90° (Abd Deg: abduction degree, Ex. Rot. Deg: external rotation degree). Passive range of motion degrees are in parenthesis. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC514557.xml
546205	Expression profile of immune response genes in patients with Severe Acute Respiratory Syndrome	Background Severe acute respiratory syndrome (SARS) emerged in later February 2003, as a new epidemic form of life-threatening infection caused by a novel coronavirus. However, the immune-pathogenesis of SARS is poorly understood. To understand the host response to this pathogen, we investigated the gene expression profiles of peripheral blood mononuclear cells (PBMCs) derived from SARS patients, and compared with healthy controls. Results The number of differentially expressed genes was found to be 186 under stringent filtering criteria of microarray data analysis. Several genes were highly up-regulated in patients with SARS, such as, the genes coding for Lactoferrin, S100A9 and Lipocalin 2. The real-time PCR method verified the results of the gene array analysis and showed that those genes that were up-regulated as determined by microarray analysis were also found to be comparatively up-regulated by real-time PCR analysis. Conclusions This differential gene expression profiling of PBMCs from patients with SARS strongly suggests that the response of SARS affected patients seems to be mainly an innate inflammatory response, rather than a specific immune response against a viral infection, as we observed a complete lack of cytokine genes usually triggered during a viral infection. Our study shows for the first time how the immune system responds to the SARS infection, and opens new possibilities for designing new diagnostics and treatments for this new life-threatening disease.	Background Severe acute respiratory syndrome (SARS) emerged in 2003, as a new epidemic form of life-threatening infection [ 1 ]. As of September 2003, there were 8098 cases of SARS from 29 countries with 774 deaths (WHO). SARS is characterized by high fever, malaise, rigors, headache, dry cough, and progression to interstitial infiltration in lungs with eventual mortality of greater than 10% in many countries [ 2 ]. SARS has been shown to be caused by a novel coronavirus; SARS-CoV, with genome sequences recently published [ 3 - 7 ]. However, the pathogenesis of SARS is poorly understood. Major hematological features of this disease are lymphopenia, transient thrombocytopenia, and normal neutrophil and monocyte counts [ 8 ]. It has been shown that SARS coronavirus infects and replicates in a wide variety of host cells, including PBMCs, in susceptible animals and human beings [ 9 , 10 ]. Hence, to understand the host response to this pathogen, we profiled the gene expression patterns of peripheral blood mononuclear cells (PBMC) from SARS patients, compared to healthy controls using oligo nucleotide microarrays. We found that in the PBMC from SARS patients a number of genes were differentially expressed, as compared to healthy controls, including immune-related genes and these genes are not the typical ones expected in a viral infection. During a viral infection, most cell types in the body respond by secreting high levels of type 1 interferons (IFN-α and IFN-β) [ 11 ]. IFN-α/β can directly induce antiviral activities in neighboring cells, preventing viral spread by increasing the resistance of uninfected cells toward the virus. Moreover, these IFNs can activate Natural Killer (NK) cells mediated cytotoxity toward virus-infected cells [ 11 , 12 ], and there is accumulating evidence that IFN-α/β contribute to driving the adaptive-immune response in the T helper cell type 1 (Th1) direction, via stimulation of IFN-γ expression [ 12 ]. NK cells can produce IFN-γ [ 13 ], which activates leukocytes, such as monocytes/macrophages, that, in turn, participate in the antiviral responses by producing free radicals and proinflammatory cytokines such as TNF-α [ 13 ]. During the response to viral infections, a key role is played by the expansion and activation of CD4+ and CD8+ T cells, which are central to the antiviral immunity, including their capability to inhibit replication and clear the infection. CD8+ cells have a direct effector role through cytotoxic T lymphocyte mediated lysis, and cytokine and chemokine production [ 14 ]. The role of CD4+ T cells in antiviral immunity is highly dependent on production of cytokines, notably IFN-γ [ 15 ], and the cytolytic activity exerted by a subset of CD4+ T cells [ 16 ]. Activation, coordination, and regulation of the above-described antiviral responses are mediated by complex mechanisms, where cytokines play important roles. However, to our surprise, we found that the patients' response of SARS appears to be mainly an innate inflammatory response, rather than a specific immune response against a viral infection. There is no significant level of up-regulation of MHC-I genes, neither for major cytokines including IFNs, nor for genes involved in complement mediated cytolysis, suggesting that the immune response against SARS-CoV may be different from other viral infections. Results and discussion To study the differential immune-gene expression patterns induced by SARS coronavirus, PBMCs from patients with SARS and normal subjects were examined using microarray technology. We shooed to use PBMCs as these cells are more easily obtained from patients compared to other infected tissues, and the SARS-CoV has recently been shown to infect PBMCs [ 9 , 10 ]. The method of global gene expression analysis using oligonucleotide microarrays has proven to be a sensitive method to develop and refine the molecular determinants of several human disorders, including cancer and autoimmune diseases, and has provided us with signatures of the immune response [ 17 ]. Using this technology, complemented with powerful analytical methods, we compared the gene expression profiles of PBMC from a series of SARS patients with those of healthy donors. To ensure the reliability and reproducibility of the microarray analysis, Pearson Correlation factors using the signal from all the normal samples were calculated. All four control arrays have Pearson correlation coefficients (r) of >0.95, which suggests an excellent reproducibility among individual arrays in the same experiment and between normal control experiments. We analyzed the expression pattern of over ~8,700 genes from the PBMCs of 10 SARS patients, and compared the expression patterns with healthy control samples. Student's unpaired T-test ( P < 0.01) was performed and significant genes were selected. Cluster analysis was performed to find distinct groups of genes that were significantly changed. The criteria used for cluster analysis is different from filter criteria used for differentially expressed genes. 248 genes were selected for Hierarchical clustering (Figure 1 ) which includes all the 186 differentially expressed genes. The severity of the disease does not seem to change the genetic profile of the PBMCs: 5 of our patients were in intensive care, whereas the other 5 patients did not require intensive care, at the time the blood was taken. However, no substantial gene-expression differences were observed that could correlate to disease severity. To our surprise, the expression of genes coding for cytokines was completely absent, and the immune-related genes which were overexpressed are usually associated with innate-immune response against bacterial infection and not against a viral infection (Table 1 ). Several genes were highly up-regulated in patients with SARS, such as, genes coding for Lactoferrin, S100A9 and Lipocalin 2 among others (Table 1 ). Lactoferrin is a 78- to 80-kDa glycoprotein synthesized by glandular epithelial cells and mature neutrophils, frequently used as a marker for neutrophil degranulation at sites of inflammation, having well recognized bacteristatic and bactericidal properties [ 18 ]. It is now clear that the biological actions of lactoferrin are not restricted to its bacteristatic and bactericidal properties. Indeed, a wide array of actions has been reported for lactoferrin, including enhancing NK cell activity, as well as stimulating neutrophil aggregation and adhesion [ 18 ]. The myeloid-related proteins, S100A8, S100A9, and S100A12 have been shown to regulate lymphocyte, monocyte, and neutrophil migration and are found in the extracellular milieu during infections and inflammatory episodes [ 19 , 20 ]. The high levels of serum S100A8/A9 in chronic inflammatory pathologies such as rheumatoid arthritis and inflammatory bowel disease, as well as during bronchitis and tuberculosis [ 20 ], suggest that they might play a role in inflammatory reactions. Lipocalins are a family of small, secreted proteins, which have little amino acid sequence homology (20–30%) but share a common three-dimensional structure [ 21 , 22 ]. Tissue distribution studies have revealed that a member of this family 24p3 lipocalin (24p3) is mainly expressed in the liver during an acute phase response [ 21 ]. It has also been detected in spleen, lung and the uterus. In the latter location, its expression has been found to be coincident with parturition, a time of major tissue remodeling and inflammation [ 22 ]. 24p3 has also been detected in the conditioned media of LPS stimulated murine PU5.1.8 macrophages and therefore it has been suggested to function in defense against infectious agents [ 22 ]. However, recent evidence proposed that the lipocalins may trigger apoptosis in immune cells via an unknown cell surface receptor [ 23 ]. SARS patients with respiratory distress fulfill criteria for Acute Respiratory Distress Syndrome (ARDS) and diffuse alveolar damage is seen in the lungs on histological examination of postmortem [ 24 ]. We speculate that upregulation of lipocalins are part of the host response in limiting unwanted tissue damage and in reducing inflammation and lung fibrosis. In addition, enhanced expression of lipocalins may play a role in the potential cause of lymphopenia observed in the majority of patients. There is also up-regulation of expression of genes, such us, Bactericidal Permeability Increasing Protein (BPI) and Carcino Embryonic Antigen related Cell adhesion Molecule 8 (CEACAM 8 or CD66b). BPI is released from activated neutrophils and is an endogenous antibiotic, which rapidly kills Gram negative bacteria by high affinity binding to the LPS component of the cell wall [ 25 ]. CEACAM 8 is expressed by activated monocytes and granulocytes, and significant up-regulation of this gene indicates the involvement of innate immune cells in SARS. Other genes which encode proteins like Leukotrien-B4 receptor (LTB4R), Leukotrien-A4 hydrolase, IL-8 receptor (IL-8RA), anaphylatoxin C3a receptor-1 (C3aR1), Neutrophil Cytosolic Factor 1 (NCF 1), S100 calcium binding protein A9, Defensin, (DEFA 1/4), LPS binding protein CAP18 (CAMP), and Peptidoglycan Recognition Protein (PGLYRP) are involved in chemotaxis, inflammatory reaction and superoxide metabolism [ 26 ]. Similarly, Formyl Peptidase Receptor (FPR) genes are expressed by activated neutrophils. Regulation of expression of FPR on neutrophils plays a key role in neutrophil polarization and chemotaxis. Chemotaxis is effected by a neutrophil membrane mesh, via remodeling of the actin component of the membrane [ 27 ]. Up-regulation of CD24 and FcγR3A indicates some degree of neutrophil, B cell and NK cell activity. FcγR3a, a low affinity receptor for IgG, expressed in activated macrophages and NK cells, facilitates Antibody Dependent Cell mediated Cytotoxicity (ADCC) by NK cells [ 13 ]. There is moderate up-regulation of the kappa light chain of the Nuclear Factor (NFκB 1A) and the B-cell lymphoma 3-encoded protein (BCL3) (Table 1 ). The major form of NFκB is a heterodimer of p65/p50 subunits, which regulates expression of many proteins in activated immune cells by binding to DNA. It interacts with NFκB through its ankyrin repeats and it favors p50 dimerization by recruiting p50 monomers from the cytoplasmic pool of p105/p50 dimers, and thereby enhancing nuclear translocation and DNA binding of p50 dimers [ 28 ]. BLC3 is expressed by activated B cells and T cells on mitotic stimuli, playing an important role in transcription activation [ 29 ]. In SARS patients there is down-regulation of genes (Table 2 ) which regulate proliferation and differentiation of T-cells, such as the Lymphocyte-specific protein tyrosine kinase (LCK), which is required for phosphorylation of CD3. There is down-regulation of genes coding for the epsilon polypeptide of the TCR (CD3E), IL-2 induced T-cell kinase (ITK), the Zeta chain of the TCR (CD3Z), the Alpha 4 subunit of VLA-4 receptor (ITGA4), the Chemokine (C-C motif) Receptor 7 (CCR7), and the Interleukin 10 receptor alpha (IL10 RA). All these point to a potential state of unresponsiveness towards the SARS-CoV antigens. It has been described that T cell function in patients with peritonitis had decreased Th1 function, without increased Th2 response, and furthermore T cell proliferation and cytokine secretion correlated with mortality [ 30 ]. In our study, there is up-regulation of genes involved in homeostasis and cell growth (Table 3 ); for example, genes involved in DNA synthesis, nucleosome assembly and protein synthesis, such as, genes encoding Translation Initiation Factor 1A (EIF1 AY), and histone proteins. There is down-regulation of genes coding for negative regulators of the cell cycle, such as, Retinoblastoma-like 2 protein (RBL2), Cyclin-Dependent Kinase Inhibitor 1B (CDKN1B) and anti-apoptoic protein TNF induced protein GG2-1 etc (data not shown). These findings indicate a high degree of proliferation of immune cells; however, there is no significant level of increase in the expression of cytokines like IL-2 or IL-3 and their receptors, which are required for the proliferation and the differentiation of T cells and effector functions of B cells and NK cells. So the increased cell proliferation may be of a granulocyte lineage rather than a monocytic lineage. We confirmed our microarrays findings by Real time PCR on selected upregulated genes such as Lactoferrin Lipocalin, S100P, FCGR3A, and TLR2 (Figure 2 ). The average expression of all the five genes by real-time RT-PCR was compared to the average expression levels found by microarray analysis (Table 4 ). The real-time PCR method verified the results of the gene array analysis and showed that those genes that were up-regulated as determined by microarray analysis were found to be up-regulated by real-time PCR analysis. Moreover, since we did not find any upregulation of cytokine-genes usually associated with a viral infection, we selected a few cytokines and compared the SARS samples, with samples collected from 5 influenza virus-infected patients. The observed results, in figure 2 , showed that in the influenza patients the mRNA for type I interferons (IFNα and IFNβ), TNFα and IL-12-p40 were upregulated, whereas genes upregulated in the SARS patients (LTF, Lipocalin, S100P, FGR3A and TLR2), were not upregulated in the influenza patients. Therefore, firstly, this confirms that cytokines usually triggered during a viral infection were not detected in the SARS samples (but were triggered in the influenza patients); and secondly, genes triggered in the SARS patients were not triggered in influenza patients. In this study we have performed extensive analysis of gene expression of PBMCs of SARS patients using a microarray platform that includes more than 8000 gene sequences. However, one potential drawback in our current experimental design is that the gene expression levels were compared between patient samples and normal controls using PBMCs, rather than purified individual cell types. While it would be of interest to determine the exact proportions of monocytes, lymphocytes and contaminating granulocytes in our PBMC preparations, unfortunately, given the highly infective nature of patient samples and lack of suitable flowcytometry facility with appropriate bio-safety control, FACS analysis was, however, not possible. Our results suggest that the response of SARS affected patients seems to be mainly an innate inflammatory response, rather than a specific immune response against a viral infection. There is no significant level of up-regulation of MHC-I genes or major cytokines including IFNs (α,β, and γ), or genes involved in complement mediated cytolysis, suggesting that the immune response against the SARS-CoV may be different from other viral infections, or that the virus may be using a unusual strategy to evade the host immune system and cause the pathogenesis and mortality. Conclusions This differential gene expression profiling of PBMCs from patients with SARS, strongly suggests that the response of SARS affected patients seems to be mainly an innate inflammatory response, rather than a specific immune response against a viral infection. There is no significant level of up-regulation of MHC-I genes or major cytokines, including IFNs, or complement mediated cytolysis, suggesting that the immune response against the SARS-CoV may be different from other viral infections or that the virus may be using a different strategy to evade the host immune system and cause the pathogenesis and mortality. The severity of the disease does not seem to change the genetic profile of the PBMCs: 5 of our patients were in intensive care, whereas the other 5 patients did not require intensive care, at the time the blood was taken. However, no substantial gene-expression differences were observed that could correlate to disease severity. Moreover, we confirmed (by real-time-PCR) that cytokines usually associated with viral infections (such as interferons and other cytokines) were not detected in the SARS samples, but were triggered in another viral infection (influenza patients). We also showed that genes triggered in the SARS patients were not triggered in the influenza patients. Our study shows, for the first time, how the immune system responds to the SARS infection and opens new possibilities for designing new diagnostics and treatments for this new life-threatening disease. Methods Patients and peripheral blood RNA extraction Our study included ten adult patients (S1-S10) who were diagnosed with SARS according to the World Health Organization (WHO) SARS criteria and admitted to the Tan Tock Seng Hospital, Singapore, for treatment and four additional normal human subjects (C1-C4). Our protocol was approved by the Tan Tock Seng Hospital Research Ethics committee. Peripheral blood was collected by vein-puncture of the brachial vein; mononuclear cells (PBMC) were prepared using histopaque (Sigma- Aldrich, St. Louis, MO) according to manufacturer's recommendations. Sample preparation and processing procedures were carried out as described in the Affymetrix Gene Chip Expression Analysis Manual (Affymetrix Inc., Santa Clara, CA). Briefly, total RNA was extracted from PBMCs, using Trizole method (Invitrogen, Carlsbad, CA) and further purified using RNeasy columns according to manufacturer's instructions (Qiagen, Valencia, CA). Integrity of total RNA was confirmed by formamide gel electrophoresis and quantification was carried out by measuring the A 260 nm . Generation of cDNA and labeled cRNA 5 μg of total RNA was used to synthesize double stranded cDNA using T7-(dT24) oligonucleotide primer and Superscript reverse transcriptase (Invitrogen). The resultant cDNA was purified by phenol:chloroform extraction and ethanol precipitation in presence of 7.5 M ammonium acetate. 1 μg of purified cDNA was subsequently used to synthesize biotin labeled cRNA by in vitro transcription (IVT) using T7 RNA polymerase at 37°C for 5-6 hr as per manufacturers' instructions (ENZO labeling kit, Ambion, USA). Labeled cRNA obtained after IVT was purified using RNeasy columns (Qiagen). Purified cRNA was fragmented using fragmentation buffer (40 mmol/L Tris acetate, pH 8.1, 100 mmol/L Potassium acetate, 30 mmol/L Magnesium acetate) at 94°C for 35 min. Microarray hybridization and scanning Fragmented cRNA (10 to11μg/ probe array) was used to hybridize to human focus array (HG-Focus Array) at 45°C for 16 hr with constant rotation of 60 rpm in a Gene chip hybridization oven 640 (Affymetrix). The chips were washed and stained using Gene chip fluidics Station 400 (Affymetrix). Staining was performed using streptavidine phycoerythrin conjugate (SAPE, Molecular Probes, Eugene, OR), followed by the addition of biotinylated antibody to streptavidine (Vector Laboratories, CA), and finally with streptavidine phycoerythrin conjugate. Probe arrays were scanned using Agilent Gene Array Scanner Series US 74900593 (Agilent technologies, USA). Data filtering and analysis An absolute expression analysis was performed using Microarray Suite Software 5.0 (Affymetrix) and relative mRNA expression levels were expressed as plus or minus fold changes compared to normal controls. Each chip was scaled to an overall intensity of 500 to correct for minor differences in overall chip hybridization intensity, and to allow comparison between chips. The data from ~8700 genes was imported in to MicroDB 3.0 and Data Mining Tool 3.0 (Affymetrix) for further analysis. Pearson Correlation coefficient (r) was used to ensure the reproducibility of the data using signal from normal samples. T-test was performed to identify genes that were differentially expressed in SARS patients over normal samples. The statistical significance of the differential expression of any gene was assessed by computing P value for each gene. Any gene for which this P value was < 0.01 was considered to be differentially expressed. We selected 186 genes for further analysis that met the following criteria: (i) Changes in expression of at least 2 fold higher or lower comparing the normal (ii) Signal >500 (iii) Detection P < 0.01 and (iv) Genes which met the above criteria in at least 30% of samples. Finally, data representing 10 SARS patients fold changes over control samples were averaged. Genes were annotated according to biological process using the Gene ontology: tool for the unification of biology from the Gene Ontology Consortium [ 31 ]. The complete set of raw data was deposited into the NCBIs' Gene Expression Omnibus (GEO) and it can be accessed through the GEO accession 'GSE1739' . Hierarchical clustering Unweighted average linkage Hierarchical clustering was applied for samples using the 'Genesis' software [ 32 ]. Genes with 'Present' call (P < 0.01) and significantly changes in expression of at least 2 fold higher or lower comparing the normal were selected. Finally 248 genes which are passing this filter criteria in at least 15% of samples and above were selected for clustering. Real-time quantitative PCR Real-time PCR was performed for ten genes, namely: Lactoferrin (Primers: 5' tcg tcc tgc tgt tcc tcg ggg 3' and 5' tcc agc ggt cct gcg aag gcc 3'); Lipocalin (Primers: 5' aag ccc ctg ctc ctg gcc atc agc 3' and 5' cga cct gat gct gta tgc cac gtg 3'); S100P (Primers: 5' cat gat cat aga cgt ctt ttc 3' and 5' aca cga tga act cac tga agt 3'); TLR2 (Primers: 5' gta tct gca agg gca gct cag gat 3' and 5' ttc ctc aag gaa ggt aag tcc agc 3'); FCGR3A (Primers: 5' ctc cgg ata tct ttg gtg act 3' and 5' tgc aga gca gtg ttc ttc cag 3'); IFNA (Primers: 5'atg gcc ttg acc ttt gct tt 3' and 5'tgg aag att tcc tca tag c 3'); IFNB (Primers: 5'atg acc aac aag tgt ctc ctc caa a 3' and 5' ttc ttc cag gac tgt ctt ca 3'); IL-12 p40 (Primers: 5'atg tgt cac cag cag ttg gtc atc 3' and 5'ctg aat gtc aaa tca gta ct 3'); TNFA (Primers: 5'gag tga caa gcc tgt agc cca tgt tgt agc 3' and 5'gca atg atc cc a aag tag acc tgc cca gac 3'); GAPDH (Primers: 5'acc aca gtc cat gcc atc ac 3' and 5'tcc acc acc ctg ttg ctg ta 3'). For amplicon detection, the Light Cycler RNA Master SYBR Green Kit (Roche) was used as described by the manufacturer. PCRs were performed in a LightCycler ® instrument (Roche) as follows: reverse transcription at 61°C for 20 min, initial denaturation at 95°C for 2 min; amplification for 45–65 cycles of denaturation (95°C, 5s, ramp rate 2°C /s), annealing (optimal temperature, 5s, ramp rate 2°C /s) and extension (72°C, product length [bp]/25 s, ramp rate 2°C /s). A single online fluorescence reading for each sample was taken at the end of extension step. Quantitative results were expressed by identification of the second derivative maximum points, which marked the cycles where the second derivatives of the fluorescence signal curves are at maximum. These points were expressed as fractional cycle numbers. Then, these cycle numbers were plotted against the logarithm of the concentrations of serially 2-fold diluted standard samples to obtain a standard curve. The concentrations of unknown samples were calculated by extrapolation from this standard curve. Positive sample specificity was confirmed by determining the melting curve (95°C, 5s, ramp rate 20°C /s; 68°C, 15s, ramp rate 20°C /s; 95 °C, 0s, ramp rate 0.1°C /s, continuous measurement). Authors' contributions RR carried out the sample preparation, hybridization and drafted the manuscript. MJ carried out statistical analysis, data filtration, and data deposition and helped with manuscript preparation. LYH, HHC and DT provided the blood samples and the clinical comments for the study. BPL advised on study design and development. AJM devised and coordinated the study, revised and finalized the manuscript. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546205.xml
509248	Are p.I148T, p.R74W and p.D1270N cystic fibrosis causing mutations ?	Background To contribute further to the classification of three CFTR amino acid changes (p.I148T, p.R74W and p.D1270N) either as CF or CBAVD-causing mutations or as neutral variations. Methods The CFTR genes from individuals who carried at least one of these changes were extensively scanned by a well established DGGE assay followed by direct sequencing and familial segregation analysis of mutations and polymorphisms. Results Four CF patients (out of 1238) originally identified as carrying the p.I148T mutation in trans with a CF mutation had a second mutation (c.3199del6 or a novel mutation c.3395insA) on the p.I148T allele. We demonstrate here that the deletion c.3199del6 can also be associated with CF without p.I148T. Three CBAVD patients originally identified with the complex allele p.R74W-p.D1270N were also carrying p.V201M on this allele, by contrast with non CF or asymptomatic individuals including the mother of a CF child, who were carrying p.R74W-p.D1270N alone. Conclusion These findings question p.I148T or p.R74W-p.D1270N as causing by themselves CF or CBAVD and emphazises the necessity to perform a complete scanning of CFTR genes and to assign the parental alleles when novel missense mutations are identified.	Background Cystic fibrosis (CF) is a common, often fatal disease with a well-defined genetic cause, so that it is now recommended in many countries in Europe and the United States to offer genetic screening for CF mutations to identify carriers among adults with a positive family history of CF, partners of individuals with CF, couples planning a pregnancy, couples seeking prenatal care and, recently, neonatal screening. Because of the mutational heterogeneity and the rarity of many mutations, most clinical DNA laboratories offer tests that aim to detect 75–95 % of CF alleles depending on the ethnic and geographic backgrounds of the population, using available commercial kits motsly including 20 to 31 mutations selected on the basis of their frequency as CF-causing mutations. A few laboratories (most often national reference laboratories) have developed the scanning of coding/flanking CFTR sequences to detect unknown mutations. As of February 2004, about 1200 disease-causing mutations have been identified in the cystic fibrosis transmembrane conductance regulator (CFTR) gene Frameshift, splice-site, nonsense, and in-frame but nonfunctional deletions (such as p.F508del) are disease-causing mutations. By contrast, the status of some missense mutations is extremely difficult to assess and functional studies are not available to diagnostic laboratories. Missense mutations may represent, depending on the populations, up to 45% of mutations responsible for CF or CBAVD (congenital bilateral absence of vas deferens). Moreover, due to improved scanning strategies, a growing number of complex alleles (several sequence changes on the same gene) are thought to affect the expression of the disease phenotype by modulating the effect of a mutation [ 1 - 5 ]. The most striking exemple is the length of the intron 8 polythymidine tract (7, 9, or 5 thymidines) on exon 9 splicing as a genetic modifier of the severity of the p.R117H mutation [ 1 ]. Another exemple is the revertant mutation p.R553Q which, when carried on the same gene as p.F508del, is associated with a CF phenotype with normal chloride concentration in sweat test [ 6 ] and which, when expressed in heterologous cells, can partially correct the processing and Cl- channel gating defects caused by the p.F508del mutation [ 7 ]. With the start of population screening for CF carriers, new data on the prevalence of some missense mutations have been provided, questioning their involvement as disease-causing mutations. In North American populations, missense mutations p.I148T and p.D1270N were found >100 times and >200 times, respectively, more frequently in carrier screening than in CF patients [ 8 , 9 ]. Moreover, we and others have found that individuals affected with CF or CBAVD carry p.D1270N associated with p.R74W on the same allele [p.R74W;p.D1270N] [ 10 , 11 , 5 ]. Similarly, p.I148T has been shown to be associated with a CF phenotype only as a complex allele, i.e. when associated with mutation c.3199del6 on the same gene [ 8 ]. A completely asymptomatic male individual who is a compound heterozygote for p.D1270N and p.I148T has been recently identified [ 9 ]. These findings provided evidence that these missense changes may not be the true mutations and prompted us to reanalyze all the patients in our CF or CBAVD cohort who had been originally diagnosed as compound heterozygotes for either p.I148T or [p.R74W;p.D1270N] and another mutation on the other allele. The result of full scanning of CFTR sequences showed that a second mutation (c.3199del6 or the novel mutation c.3395insA) was associated in cis with p.I148T in all individuals with a CF phenotype, and that a third missense mutation (p.V201M) was associated in cis with complex allele [p.D1270N;p.R74W] in patients with a CBAVD phenotype in this series. Methods CFTR scanning for individuals with p.I148T or [p.D1270N;p.R74W] From 1990 to 2003, we have analysed for CFTR mutations genomic DNA from 437 families with CF and 170 with isolated azoospermia caused by CBAVD, using a combination of mutation screening for known and scanning for unknown mutations. The first step was the search of 31 CF mutations detected by the ABI oligonucleotide ligation assay and 3 common intronic mutations by using restriction analysis. The second step was the scanning of coding/flanking sequences by DGGE (Denaturing Gradient Gel Electrophoresis) using 32 GC-clamped amplimers (which in our experience detected 98% of CFTR mutants), followed by sequencing to resolve abnormal PCR products (BigDye terminator cycle sequencing on ABI 310 automate sequencer). Whenever possible, family members were assayed for the mutations and associated polymorphisms. We detected 160 different mutations in the CF group accounting for 97 % of CF alleles, and 64 different mutations in the CBAVD group accounting for 85% of CBAVD alleles, which represents one of the highest allelic heterogeneity reported so far. Usually, mutation scanning is stopped when two mutations are found to be in trans . In this study, we analyzed by DGGE the entire coding and flanking regions of the CFTR gene of individuals who had been previously found to carry p.I148T or the complex allele [p.R74W;p.D1270N] and assayed their relatives for the additional sequence changes identified. We also re-analyzed the CFTR gene of a CF patient who had been originally described with c.394delTT in trans of c.3195del6 [ 12 ], now renamed c.3199del6 (see the Results). Studies to determine the frequency of each sequence alteration described in this report were performed on 600 chromosomes from our general population (Southern France). In addition, we also reanalyzed two additional CF patients previously found to carry p.I148T in the Lyon genetic center. The study was approved by the institutional ethical committees and informed consent was obtained from families. Nomenclature Gene variants and mutants are described using DNA and protein designation: intronic changes, deletions, insertions and frameshifts are reported at the cDNA level (c.) and amino acid changes at the protein level (p.), as recommended in the Human Genome Variation Society web page . Results A CF mutation (c.3199del6 or c.3395insA) is associated in cis with p.I148T in CF patients Two out of 437 CF patients analyzed in Montpellier and two out of 801 CF patients analyzed in Lyon were found to carry p.I148T, which was initially thought to be one of the two mutations responsible for CF in these patients. However, thorough re-analysis of the entire CFTR sequence determined that a CF mutation (c.3395insA or c.3199del6) was present on the same gene in both cases (table 1 ). Table 1 CFTR haplotypes associated with mutations found in CF patients carrying p.I148T in cis with c.3395insA or c.3199del6 and in one CF patient carrying c.3199del6 alone Indiv No. Age at Diagnosis Phenotype CFTR Mutations CFTR haplotype IVS1 IVS8 IVS8 IVS8 470 IVS17B IVS17B EGH a CA CA TGm Tn TA CA CF1 7 yrs CF-PI c.394delTT 21 23 10 9 M 36 13 B c.3199del6 22 16 11 7 V 7 17 C CF2 10 yrs CF-PS [c.3395insA;p.I148T] 21 23 10 9 M 7 17 B p.R334W 22 17 11 7 V 46 13 A CF3 6 ms CF-PI [c.3199del6;p.I148T] 21 23 10 9 M 7 17 B p.F508del 21 23 10 9 M 31 13 B CF4 3 yrs CF-PI [c.3199del6;p.I148T] 22 23 nd 9 M 7 17 B p.F508del 22 17 nd 9 M 31 15 B CF5 6 ms CF-PI [c.3199del6;p.I148T] 22 23 nd 9 M 7 17 B p.F508del 22 23 nd 9 M 31 13 B a EGH, extragenic haplotype XV2c/ TaqI , KM19/ PstI ; nd, not determined Patients CF1-3 were from the cohort of Montpellier (n = 437), patients CF4-5 were from the cohort of Lyon (n= 801). Mutation c.3199del6 can also occur alone as a CF-causing allele Mutation c.3199del6 was found to be carried without p.I148T in a young CF male with 394delTT on the other allele, diagnosed at the age of 7 years on the basis of typical pulmonary disease, pancreatic insufficiency, poor growth and positive sweat test [ 12 ]. Mutation c.3199del6 was initially described by us in 1994 in this patient as c.3195del6 [ 12 ], in accordance with the first draft of mutation nomenclature [ 13 ]. However, it occurred in the same palindromic sequence in exon 17a than mutation c.3199del6 reported in 1998 by Bozon et al . [ 14 ]. Both mutations are expected to delete either amino acids Val1022 and Ile1023, or Ile1023 and Val1024 from the CFTR protein. As it is impossible to determine at the genomic level in which part of the palindrome each of them occurred, the most 3'copy of the repeat is arbitrarily assigned to have been mutated, according to the current rule [ 15 ]. Consequently, mutations c.3195del6 and c.3199del6 should be considered as identical and reported as c.3199_3204del. The familial segregation analysis of polymorphisms covering the CFTR gene showed that p.I148T, when present in individuals with a CF phenotype, occurred on a unique haplotype carrying IVS8-9T whatever the mutation in cis , c.3395insA or c.3199del6 (table 1 ). By contrast, c.3199del6 without p.I148T occurred on a different haplotype carrying IVS8-7T. Mutations p.I148T, c.3199del6 and c.3395insA were not found on 600 chromosomes from our general population. Triple-mutant allele [p.R74W;p.V201M;p.D1270N] is found in males with CBAVD whereas double-mutant allele [p.R74W;p.D1270N] is found in asymptomatic individuals Re-analysis of the CFTR gene in families carrying [p.R74W;p.D1270N] identified a third mutation (p.V201M) on the same chromosome in three unrelated individuals with CBAVD (table 2 ). Only the double-mutant p.R74W-p.D1270N was present in the two unaffected individuals who were found with these changes in our sample. The first case was a young boy who had been initially suspected of having CF at age 4 years because of allergic rhinitis but for whom the diagnosis of CF was later ruled out; no other CFTR sequence alteration could be identified and the sweat tests were negative (chloride values <40 mM). The second individual was the mother of a CF girl who was compound heterozygous for p.F508del and p.P67L. This woman, who was carrying p.P67L on one CFTR gene and [p.R74W-p.D1270N] on the other (table 1 ), was completely asymptomatic at age 45 years and displayed three negative sweat tests (chloride values <20 mM). The triple and double mutant alleles seem to have occurred on the same haplotype TG11-T7-V470. Table 2 CFTR sequence changes found in individuals carrying missense alterations p.R74W, p.D1270N, or p.V201M Mutations Haplotype IVS1 IVS8 IVS8 IVS8 470 IVS17B IVS17B CA CA TGm Tn TA CA CBAVD1 p.R1066C 22 16 11 7 V 30 13 [p.R74W;p.V201M;p.D1270N] 22 16 11 7 V 31 13 CBAVD2 p.M952I 26 17 10 7 M 7 17 [p.R74W;p.V201M;p.D1270N] 22 16 11 7 V 31 13 CBAVD3 [p.R74W;p.V201M;p.D1270N] 22 16 11 7 V 31 13 [p.R74W;p.V201M;p.D1270N] 22 16 11 7 V 31 13 Individual non affected with CF No mutation 21 nd 10 7 M 7 17 [p.R74W;p.D1270N] 22 nd 11 7 V 30 13 Asymptomatic mother of a CF affected girl p.P67L 23 16 10 7 M 7 17 [p.R74;p.D1270N] 22 16 11 7 V 31 13 Discussion p.I148T is a low penetrance CF mutation or a neutral polymorphism Since its initial description in a CF Canadian patient with pancreatic insufficiency [ 16 ], the mutation p.I148T, which changes a conserved amino acid and occurs in the first cytoplasmic loop of the CFTR protein, has been considered as a severe CF allele in many countries. It was thought to be the second most common CF mutation in the French Canadian population, accounting for 9.1% of the French Canadian chromosomes [ 17 ], whereas in France, p.I148T accounted for only 0.11 % of the CF alleles in a sample of 3,710 patients affected with the disease [ 5 ]. p.I148T can now be detected by several commercially available kits developed for routine screening of CF carriers and for CF neonatal screening, and recently it has been included in the core panel of 25 CF mutations recommended by the American College of Medical Genetics (ACMG) [ 18 ]. Thousands of individuals are being screened for this mutation worldwide and it is possible that several prenatal diagnosis have been or will be performed. However there are now several lines of evidence that question the role of p.I148T by itself in causing disease. First, compound heterozygosity for p.I148T and a severe CF mutation was recently identified in several healthy individuals [ 9 , 19 ]. When affected and unaffected individuals carrying apparently the same mutational genotype were re-analyzed for additional changes that could explain the different phenotypes, p.I148T was found to be associated in cis with another mutation, c.3199del6, in patients with a classic CF phenotype, whereas healthy adults who were compound heterozygous for p.I148T and a severe CF mutation or homozygous for p.I148T did not carry the deletion [ 8 ]. In a recent study, the p.I148T mutation has been further documented to be linked with the 3199del6 mutation in all 24 CF patients of French Canadian descent originally identified as compound heterozygous for the p.I148T mutation and a second severe CFTR mutation [ 20 ]. Second, p.I148T was found to be over 100 times more common in two independent U.S. carrier screening programmes than in CF patients: it accounted for 6.4 to 7.7% of chromosomes detected in the screened populations versus 0.06 to 0.068% of CF chromosomes in CF patients [8. 9]. This discrepancy suggests that p.I148T is either a poorly penetrant mutation or a neutral polymorphism. Third, when transiently expressed in epithelial cells, p.I148T mutant protein is normally processed and is able to mediate normal chloride transport with properties identical with those of wild-type cells [ 21 ]. As the mutant seems to suppress the ability of CFTR to support HCO3- transport, it has been hypothesized that p.I148T may contribute to disease through Cl- coupled HCO3- altered transport; however, the major CFTR functions are retained by the mutant [ 21 ]. Fourth, we show in this study for the first time that p.I148T can be associated with a frameshift mutation c.3395insA in exon 17b instead of in-frame deletion c.3199del6 in exon 17a. Insertion c.3395insA (designated as c.3395_3396insA) is a previously unreported mutation that is predictive of premature termination of translation at amino acid residue 1155. The truncated protein lacking the 325 last amino acids is believed to be not functional and be degraded rapidly, generating no detectable protein. A CFTR alteration producing a premature termination signal is a class I mutation, considered severe enough to cause CF by itself and exclude the contribution of any other sequence change on the same allele. Fifth, in contrast with other studies that stated that only the complex allele [p.I148T;9T;c.3199del6] appeared to be associated with a classic CF phenotype [ 8 ], we demonstrate that c.3199del6 is associated with a CF phenotype even if the deletion occurs on a chromosome that does not carry p.I148T, which adds further value to the consideration that p.I148T is not a true mutation but simply a polymorphism. Although no functional test was performed to prove its contribution to the severe phenotype, mutation c.3199del6 has been considered as a defective allele as it results in the loss of two amino acid residues in the TM10 domain of the CFTR protein and has not been detected in non-CF alleles. Our data fully support the recent recommendation that p.I148T should not be included in the mutation panel selected for prenatal screening strategy [ 22 ]. The complex allele [p.R74W;p.D1270N] may be not enough to cause disease We and others had initially described p.R74W [ 23 ] and p.D1270N [ 24 ] in isolation but they have since been found in association in many CBAVD or CF patients [ 10 , 11 ] and these two changes were thought to be deleterious, alone or in combination. A few complex alleles have been expressed in heterologous systems to evaluate the impact on CFTR processing and Cl- channel activity and better understand the contribution of each missense mutation on phenotype. When expressed in HeLa cells, mutant p.R74W, p.D1270N and [p.R74W;p.D1270N] did not affect CFTR processing, however a lower cAMP-responsive anion conductance was observed with the double mutant [p.R74W;p.D1270N] [ 3 ]. The assay suggested that p.R74W alone should be considered as a polymorphism, p.D1270N alone could generate a CBAVD phenotype while the complex allele could produce a more severe phenotype as p.R74W could enhance the effect of p.D1270N [ 3 ]. However these findings have not yet been confirmed by other studies. We have found here that a triple-mutant [p.R74W;p.V201M;p.D1270N] allele was carried in all three patients with CBAVD whereas only the double mutant [p.R74W;p.D1270N] allele was present in two asymptomatic individuals including an obligate carrier who was compound heterozygous for a CF mutation. Another mother carrying [p.R74W;p.D1270N] in trans of a CF mutation has been described previously; despite two positive sweat tests she was absolutely asymptomatic [ 25 ]. Missense p.V201M in exon 6a changes a valine for a methionine in the third transmembrane domain; it was initially reported alone in a French patient with CBAVD [ 26 ], then in Brazilian patients with CF [ 27 ]. Recent large scale screening for CF carrier showed that p.D1270N was present 205 times more commonly in the screened population than in the CF patients (frequency of 14% versus 0.068%); in addition, a completely asymptomatic adult compound heterozygote for p.D1270N and p.I148T has been identified [ 9 ]. Although it is not known whether these alleles are associated or not with the third change p.V201M, there are now enough evidence to question the role of the complex allele [p.R74W;p.D1270N] as being a CF or CBAVD mutation. Further experimental and genetic investigations will be necessary to demonstrate the role of p.V201M in causing disease. Conclusions This report further corroborates the recent hypothesis [ 9 ] that p.I148T and p.R74W-p.D1270N may not be true CF/CBAVD mutations. If these observations are further confirmed by a large multicentric study, they will have important implications for genetic counseling of patients and couples found to carry p.I148T or [p.R74W;p.D1270N]. They also pinpoint several important points in genetic testing for CF : first, the necessity of scanning the whole regions of the CFTR gene for diagnosis purposes, whatever the cost; second the necessity to better standardize mutation nomenclature, and third the usefulness of confirming inheritance of mutations from both parents whenever possible to avoid the risk for erroneously reporting changes in trans that are in fact complex alleles. Competing interests None declared. Authors'contributions JPA, CG, CT and FC carried out the molecular genetic studies. MDG coordonated the molecular analysis. MC conceived the study and drafted the manuscript. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC509248.xml
539348	Leptin and insulin stimulation of signalling pathways in arcuate nucleus neurones: PI3K dependent actin reorganization and KATP channel activation	Background Leptin and insulin are long-term regulators of body weight. They act in hypothalamic centres to modulate the function of specific neuronal subtypes, by altering transcriptional control of releasable peptides and by modifying neuronal electrical activity. A key cellular signalling intermediate, implicated in control of food intake by these hormones, is the enzyme phosphoinositide 3-kinase. In this study we have explored further the linkage between this enzyme and other cellular mediators of leptin and insulin action on rat arcuate nucleus neurones and the mouse hypothalamic cell line, GT1-7. Results Leptin and insulin increased the levels of various phosphorylated signalling intermediates, associated with the JAK2-STAT3, MAPK and PI3K cascades in the arcuate nucleus. Inhibitors of PI3K were shown to reduce the hormone driven phosphorylation through the PI3K and MAPK pathways. Using isolated arcuate neurones, leptin and insulin were demonstrated to increase the activity of K ATP channels in a PI3K dependent manner, and to increase levels of PtdIns(3,4,5)P 3 . K ATP activation by these hormones in arcuate neurones was also sensitive to the presence of the actin filament stabilising toxin, jasplakinolide. Using confocal imaging of fluorescently labelled actin and direct analysis of G- and F-actin concentration in GT1-7 cells, leptin was demonstrated directly to induce a re-organization of cellular actin, by increasing levels of globular actin at the expense of filamentous actin in a PI3-kinase dependent manner. Leptin stimulated PI3-kinase activity in GT1-7 cells and an increase in PtdIns(3,4,5)P 3 could be detected, which was prevented by PI3K inhibitors. Conclusions Leptin and insulin mediated phosphorylation of cellular signalling intermediates and of K ATP channel activation in arcuate neurones is sensitive to PI3K inhibition, thus strengthening further the likely importance of this enzyme in leptin and insulin mediated energy homeostasis control. The sensitivity of leptin and insulin stimulation of K ATP channel opening in arcuate neurones to jasplakinolide indicates that cytoskeletal remodelling may be an important contributor to the cellular signalling mechanisms of these hormones in hypothalamic neurones. This hypothesis is reinforced by the finding that leptin induces actin filament depolymerization, in a PI3K dependent manner in a mouse hypothalamic cell line.	Background Leptin and insulin function as peripherally-derived hormone signals involved in the long-term regulation of energy balance [ 1 - 4 ]. Their circulating levels are directly proportional to adipose mass and CNS access occurs via saturable receptor-mediated processes. The primary CNS target for these adipostats is the ARC, where leptin and insulin receptors are highly expressed, and where direct administration of either hormone has a potent effect on food intake and body weight. Two specific ARC neurone populations have been strongly implicated in sensing changes in levels of circulating leptin and insulin and transducing these signals into neuronal outputs [ 1 , 3 ]. These "first-order" neurones encompass the melanocortin precursor, POMC containing neurones and NPY and AgRP co-containing neurones, the former associated with catabolic, the latter anabolic, outputs. Leptin and insulin increase POMC mRNA levels and decrease NPY & AgRP mRNA levels respectively. However, transcriptional control is not the only effector mechanism elicited by these hormones on ARC neurones. Electrophysiological studies have shown that leptin depolarizes and increases the firing rate of ARC POMC neurones and inhibits the tone of NPY/AgRP neurones [ 5 ]. Although the electrophysiological actions of insulin have not been reported for identified POMC and NPY/AgRP neurones, both leptin and insulin have been demonstrated to inhibit, by hyperpolarization, the firing of a sub-population of ARC neurones, identified by their sensitivity to changes in extracellular glucose concentration [ 6 , 7 ]. For these latter neurones, termed glucose-responsive (GR), K ATP channels have been identified as an effector mechanism through which leptin and insulin elicit neuronal inhibition. Consequently, leptin and insulin signal the status of body energy stores by activating their receptors on ARC neurones, eliciting changes in the electrical activity and amounts of releasable peptides in specific neuronal populations, leading to compensatory effector outputs, such as changes in food intake, energy balance and glucose homeostasis [ 8 ]. Obese humans have elevated leptin and insulin levels, indicative of central resistance to these hormones [ 9 ]. The mechanisms underlying this resistance are unclear, with defective hormone passage through the BBB and flawed receptor-signal transduction in ARC neurones being the prime candidates [ 10 , 11 ]. Consequently, it is important to understand the molecular mechanisms underlying leptin and insulin receptor modulation of ARC first-order neurones. Leptin and insulin, by stimulation of their respective receptors, have been demonstrated to activate various signalling pathways in peripheral tissues [ 10 - 13 ]. However, as these hormones induce seemingly identical actions on ARC neurones, both in terms of behavioural output and effects on ARC neurone excitability, some parallelism or convergence of signalling is likely [ 12 , 13 ]. Leptin, by binding to the long form of the leptin receptor (ObRb) has been demonstrated to activate three main signalling cascades, JAK2 – STAT3, MAPK and PI3K, the latter two of which are also intermediates in insulin receptor activation [ 14 , 15 ]. However, recent studies have strongly implicated PI3K as the key signalling intermediate in leptin and insulin actions on hypothalamic neurones influencing food intake and body weight [ 16 , 17 ]. Thus, to elucidate further the pathways that contribute to convergent actions of leptin and insulin on ARC neurones, we have examined the phosphorylation status of key leptin and insulin signalling intermediates in the ARC and have explored the linkage, with a focus on PI3K mediated signal transduction pathways, between these hormones and ARC neurone K ATP channel activation. Results Leptin and insulin stimulate phosphorylation of signalling proteins in ARC Rat hypothalamic tissue sections, predominantly made up of arcuate nucleus, were treated with aCSF alone or with leptin (10 nM) or insulin (0.1 or 1 nM, which produced identical results) for 1, 5, 15 and 30 minutes. Leptin and insulin stimulation induced comparable immunoblot profiles (Figure 1 ), with increased levels of phosphorylated STAT3 (p-STAT3), phosphorylated MAPK (p-MAPK), phosphorylated PKB/Akt (p-PKB) and GSK3 (p-GSK3). The phosphorylation status of the latter two proteins, PKB and its downstream effector GSK3, were utilised as a sensitive assay for hormone activation of PI3K. Leptin (10 nM) or insulin (0.1 nM) treatment was observed to cause an increase in phosphorylation of all four proteins. This increase in phosphorylation was generally transient with the highest levels of phosphorylation at the 1 and/or 5 minute time points. Subsequent to this peak level, in the majority of experiments, the phosphorylation was not sustained over the time period examined and returned to control values within 30 minutes (Figure 1A,1B ). These data demonstrate that all 3 pathways potentially contribute to insulin and leptin signalling in ARC neurones and thus play a role in connecting leptin or insulin receptor activation to neuronal effector outputs. As both leptin and insulin signalling in the ARC require PI3K activity for reduction in food intake and body weight [ 16 , 17 ], we examined the sensitivity of the phosphorylation of PKB/GSK3 and MAPK to the presence of PI3K inhibitors. Isolated ARC sections were incubated either in control aCSF, 10 nM wortmannin or 10 μM LY294002, for 20 minutes prior to exposure to control aCSF, leptin (10 nM) or insulin (0.1 nM), in the continued presence of the appropriate inhibitor. The presence of LY294002 prevented leptin or insulin induced phosphorylation of PKB and GSK3 following 1 minute exposure to these hormones, illustrated in Figure 2A , as expected for proteins downstream of PI3K [ 18 ]. Furthermore, the presence of the PI3K inhibitor per se reduced p-PKB and p-GSK3 levels significantly, indicating that PI3K is active to a limited degree in these ARC neurones. Similar results were obtained for wortmannin (data not shown). However, surprisingly the PI3K inhibitors also reduced the leptin- and insulin-stimulated phosphorylation of MAPK (Figure 2B ). These data further establish PI3K as a key component of neuronal leptin and insulin signalling in ARC neurones and suggest a potential role for PI3K in leptin and insulin driven transcriptional activity. Because the phosphorylation status of these signalling intermediates was examined in whole ARC extracts, this supplies little information as to the mechanisms by which adiposity hormones target specific ARC neurones. Thus, we have tried to delineate the molecular events connecting leptin and insulin receptor activation, PI3K activity and effector outputs. Here we focus on the activation of K ATP channels, responsible for leptin and insulin inhibition of an electrophysiologically identified subset of ARC neurones [ 6 , 7 ]. Figure 1 Effects of leptin and insulin on phosphorylation of STAT3, MAPK, PKB and GSK3 Rat ARC wedges were incubated for 0, 1, 5, 15 or 30 minutes with 10 nM leptin (A) or 0.1 – 1 nM insulin (B) before cells were lysed and equal amounts of lysate were subjected to SDS-PAGE and transferred to nitrocellulose membrane. The phosphorylated levels of p42/p44 MAPK, PKB, STAT3 and GSK3α/β were detected by immunoblotting with appropriate specific antibodies. The total amount of PKB is also shown. Bands were quantified using densitometry. The values are expressed as relative to the corresponding aCSF control group, and normalized for protein loading. Values represent the mean ± SEM for between 4–6 animals for each time point. * P < 0.05 and ** P < 0.01. Figure 2 Changes in phosphorylation of PKB, GSK3 and MAPK by inhibition of PI3K Rat ARC wedges were pretreated with 10 μM LY294002 or aCSF for 20 minutes and incubated for 1 minute with 10 nM leptin or 1 nM insulin or aCSF. Equal amounts of protein lysate were subjected to SDS-PAGE and transferred to nitrocellulose membrane. The phosphorylated levels of PKB, and GSK3αβ (A) and p42/p44 MAPK (B) were detected by immunoblotting with appropriate specific antibodies. The total amount of PKB is also shown. Bands were quantified using a densitometer. The values are expressed as relative to the corresponding aCSF control group, and normalized for protein loading. Values represent the mean ± SEM for between 4–5 and 3–5 animals for each time point with leptin and insulin respectively. * P < 0.05 and P < 0.01. Leptin and insulin activate K ATP channels in acutely isolated ARC neurones Cell-attached recordings from rat isolated ARC neurones were used to confirm that leptin and insulin activate the large conductance K ATP channel as previously described [ 6 , 7 ]. Leptin, present in the recording electrode during cell-attached recordings, increased mean K + channel activity in 45% of unidentified neurones (n = 25/55). Mean channel activity (N f .P o ), 1–2 minutes following cell attached formation was 0.08 ± 0.02 and increased to 0.38 ± 0.03 (n = 10, P < 0.01) after peak activation had occurred (10.6 ± 1.0 minutes after patch formation). In control cell-attached recordings of between 10–25 minutes, with 10 nM leptin in the pipette solution, there was no effect on K + channel currents observed in other ARC neurones (n = 30). During peak K ATP channel activation by leptin, bath application of the K ATP channel inhibitor tolbutamide (200 μM), reduced mean channel activity by 56 ± 12% (n = 4; P < 0.05), an effect reversible on washout of drug (Figure 3A ). Following leptin-induced increase in channel activity, patch excision into the inside-out configuration allowed channel sensitivity to ATP to be assessed. At a patch potential of 0 mV and in asymmetric cation gradients, mean channel activity was 0.40 ± 0.09 (n = 3) and bath application of 3 mM MgATP (Figure 3B ) reversibly reduced channel activity by 67.1 ± 9.7% (P < 0.05). Current-voltage relations under symmetrical K + conditions were linear, with a mean single channel conductance of 156 ± 15 pS (n = 3). The sensitivity to tolbutamide, ATP and single channel characteristics are consistent with leptin activation of the large conductance K ATP channel of GR neurones [ 6 , 19 ]. In a separate series of cell-attached recordings from isolated ARC neurones, bath application of insulin (0.1 – 10 nM) also increased K ATP channel activity in 45% of unidentified neurones (n = 14/31). N f .P o increased from 0.14 ± 0.03 under control conditions to 0.49 ± 0.08 (n = 7; P < 0.01) after approximately 10 – 20 minutes exposure to insulin (Figure 3C ). Insulin had no effect on other K + channel currents in cell-attached recordings from the remaining neurones (n = 17). Consequently, these data are in agreement with previous studies on rat ARC GR neurones [ 6 , 7 ]. Figure 3 Leptin and insulin activate large conductance K ATP in ARC neurones A , representative cell-attached recording from an acutely dissociated ARC neurone. Leptin (10 nM), present in the electrode solution, increased the activity of a K + channel, which was inhibited reversibly by bath application of 200 μM tolbutamide. Upward deflections in this and subsequent cell-attached recordings are extracellularly recorded action current activity. B , representative recording from an inside-out patch under asymmetrical K + conditions and held at 0 mV obtained from an acutely dispersed ARC neurone, following cell-attached leptin-induced increase in channel activity. Note that bath application of 3 mM MgATP reversibly inhibited K + channel activity. Inserts are expanded regions of traces showing channel activity in more detail. C , representative cell-attached recording from an acutely dissociated ARC neurone. Under control conditions, few channel openings are observed. Subsequent to bath application of 0.1 nM insulin, there is a marked increase in K ATP channel activity. Corresponding diary plot of channel activity (N f .P o ) with time displays the insulin induced increase in activity. Leptin and insulin activation of ARC neurone K ATP is PI3K dependent As leptin and insulin activation of K ATP channels in ARC neurones is rapid (<5–10 minutes) and the leptin increase in K ATP activity demonstrated in isolated patches [ 6 ], this action is unlikely to be mediated by changes in transcription. Furthermore, in cell-attached recordings, following leptin (10 nM) stimulated K ATP channel activity, application of the MAPK pathway inhibitor, PD98059 (10 μM; n = 4) had no effect on N f .P o (Figure 4A ). A previous study has demonstrated that insulin activated ARC neurone K ATP channels are similarly insensitive to this MAPK pathway inhibitor [ 7 ]. However, inhibition of PI3K does reverse both leptin (Figure 4B ) and insulin-induced activation of ARC neurone K ATP channels. Leptin (10 nM) increased K ATP mean N f .P o from 0.21 ± 0.10 to 0.68 ± 0.28 (n = 3, P < 0.02), and subsequent bath application of 10 nM wortmannin reduced K ATP activity to a mean value of 0.33 ± 0.13 (n = 3, P < 0.02) over a period of 15–20 minutes, an N f .P o indistinguishable from control (P > 0.4). Similarly, in a separate series, leptin increased N f .P o from 0.15 ± 0.04 to 0.35 ± 0.05 (n = 4, P < 0.05) and subsequent application of 10 μM LY294002 reduced N f .P o to 0.19 ± 0.04 (n = 4, P < 0.01) within 15–20 minutes. Essentially identical data have been reported previously for the effects of these PI3K inhibitors on insulin-activated ARC K ATP channel activity [ 7 ]. Thus these results demonstrate that leptin and insulin signalling pathways converge on PI3K to elicit GR neurone hyperpolarization, and confirm that PI3K is a key enzyme in individual ARC neurone responsiveness to both leptin and insulin. Figure 4 PI3K activity mediates leptin activation of K ATP Representative cell-attached recordings with leptin (10 nM), present in the recording electrode. A , traces illustrate that leptin-induced increase in K ATP activity is not reversed on bath application of the MEK inhibitor, PD 98059 (50 μM). The corresponding diary plot for part of the recording, initiated 20 mins after recording began, and following attainment of maximal leptin-induced K ATP channel activity is shown. B , traces show that leptin-induced K ATP channel activity is inhibited by subsequent application of the PI3K inhibitors, wortmannin (10 nM) or LY294002 (10 μM). Corresponding diary plots for N f .P o from a single experiment for each PI3K inhibitor are shown below the relevant traces. Such a central role for PI3K suggests that its main lipid product, PtdIns(3,4,5)P 3 may serve as an important second messenger for downstream effectors such as the K ATP channel. The mechanism by which PtdIns(3,4,5)P 3 recognises downstream target proteins is by binding to specialised phosphatidylinositol recognition sites, such as the pleckstrin homology (PH) domain [ 20 ]. Thus, to demonstrate that PtdIns(3,4,5)P 3 production is elevated in ARC neurones following exposure to leptin and insulin, we used the PH domain of GRP-1, which selectively binds PtdIns(3,4,5)P 3 [ 21 ] coupled to GFP (PH-GRP1-GFP) in an overlay assay on fixed freshly isolated ARC neurones. In non-stimulated ARC neurones there is significant labelling of all neurones with PH-GRP1-GFP (Figure 5 ). This is likely due to inherent PI3K activity of the neurones, rather than non-specific binding, as a PH-GRP1-GFP fusion protein with a single point mutation (K273A), which does not bind PtdIns(3,4,5)P 3 [ 21 , 22 ], displays very little reactivity with non-stimulated ARC neurones (n = 7). Stimulation of isolated ARC neurones with leptin (10 nM) for 10 minutes resulted in a proportion (38 ± 8%; n = 5) of dispersed neurones displaying increased fluorescence after treatment with wild type fusion protein (Figure 5A ). In addition, exposure of dispersed neurones to insulin (1 nM) for 5–10 minutes induced increased binding of PH-GRP1-GFP fusion protein in a similar proportion of neurones (43 ± 15%; n = 6), although insulin appeared to induce greater levels of binding/fluorescence than leptin (Figure 5B ). However, leptin and insulin driven phosphorylation of PKB and GSK3 along with induction of elevated PtdIns(3,4,5)P 3 levels in ARC neurones, are only indicative of increased PI3K activity. Thus we examined IRS-2 associated PI3K activity [ 16 ] in isolated ARC wedges exposed for one or two minutes to leptin (up to 50 nM) or insulin (up to 100 nM). Although we observed an increase in activity in 4/8 and 4/7 experiments for leptin and insulin respectively, there was no overall significant increase observed (data not shown). This may be due to the relative paucity of leptin and /or insulin sensitive neurones in the overall cellular population. Figure 5 Leptin and insulin increase PtdIns(3,4,5)P 3 in isolated neurones Acutely isolated ARC neurones were incubated in the absence and presence of 10 nM leptin (A) or 1 nM insulin (B) for 10 minutes. Cells were fixed and permeabilized, as described in Methods, prior to incubation with wild type (wt) or K273A mutant (mt) PH-GRP1-GFP fusion protein for 1 hour. Cells were subsequently processed for visualising GFP by confocal microscopy. Note that leptin and insulin increased the binding of wild type PH-GRP1-GFP in ARC neurones, and this is shown as both the fluorescence image (upper panels in A, B) and as a false colour image (lower panels in A, B), where blue represents low or non-detectable fluorescence and red the highest fluorescence intensity. Leptin and insulin activation of GR neurone K ATP requires actin filament re-organisation Previous studies have demonstrated that the phosphatidylinositol lipid second messenger, PtdIns(3,4,5)P 3 activates K ATP channels in an insulin-secreting cell line when applied directly to the internal aspect of isolated patches [ 23 ]. However, activation of K ATP is probably not due to direct binding of the lipid to channel subunits, as the effect of PtdIns(3,4,5)P 3 on K ATP was prevented by the presence of the actin stabilizing agent, phalloidin. Additionally, leptin-induced opening of this insulin-secreting cell K ATP channel was occluded when phalloidin was present in the cell interior [ 23 ]. Thus we examined whether adiposity hormone signalling in ARC GR neurones also requires actin remodelling in order to manifest a specific effector output, the activation of hypothalamic neurone K ATP channels. As our assessment of hormone activation of GR neurone K ATP channels uses cell-attached recordings, we used the membrane permeable actin stabilizing toxin, jasplakinolide to induce actin polymerization [ 24 ]. In preliminary experiments, jasplakinolide (100 nM) was demonstrated to have no effect when applied directly to isolated inside-out patches obtained from ARC neurones, under asymmetrical recording conditions, containing spontaneously active K ATP channels (n = 4; P > 0.5; data not shown). Cell-attached recordings with leptin (10 nM) present in the pipette solution, increased mean K ATP channel activity from 0.07 ± 0.02 to 0.44 ± 0.07 (n = 4; P < 0.01). Subsequent bath application of jasplakinolide (50 – 100 nM) reversed the leptin-induced K ATP activation (Figure 6A ), with channel activity returning to 0.10 ± 0.02 (n = 4; P < 0.01) within 5–10 minutes, a level indistinguishable from pre-leptin controls (P > 0.5). In a second series of cell-attached recordings, bath application of insulin (0.1 – 10 nM) increased mean K ATP activity from 0.06 ± 0.01 to 0.64 ± 0.16 (n = 4; P < 0.01) and subsequent bath application of jasplakinolide (100 nM), in the continuous presence of insulin, reduced channel activity to 0.17 ± 0.06 (n = 4; P < 0.01) within 5–10 minutes (Figure 6B ). The inhibition of leptin and insulin stimulated K ATP channel activity by jasplakinolide was reversible on washout of the toxin in 2/4 and 3/4 patches for leptin and insulin respectively. Figure 6 Actin dynamics mediate leptin and insulin activation of K ATP A , representative cell-attached recording from an acutely isolated ARC neurone with leptin (10 nM) in the electrode solution. Following attainment of increased K ATP activity, bath application of jasplakinolide (100 nM) reversibly reduced channel activity. The corresponding diary plot for this experiment is shown. B , representative cell-attached recording from an ARC neurone. Bath application of insulin (10 nM) increased K ATP channel activity and subsequent bath addition of 100 nM jasplakinolide, concomitant with 10 nM insulin, reversibly inhibited the insulin stimulated K ATP activity. The corresponding diary plot is shown. PI3K mediates leptin-induced actin filament reorganisation in GT1-7 cells Preliminary experiments labelling rat ARC slices with rhodamine-conjugated phalloidin to stain for F-actin were unsatisfactory due to the overall high levels of staining in the slices and inability to distinguish clearly individual neurones and their responses to hormone stimulation. Similarly, use of freshly isolated ARC neurones was precluded as the phalloidin staining was inconsistent among individual neurones within a single preparation and between neuronal preparations. Thus we have used the mouse hypothalamic cell line, GT1-7, to demonstrate that leptin utilises a PI3K-dependent signalling cascade to modify cytoskeletal dynamics. A previous study indicated that GT1-7 cells express ObRb [ 25 ], although others do not concur with this conclusion [ 26 ]. Using RT-PCR, we detected the presence of leptin receptor mRNA in GT1-7 cells by amplification of a common extracellular domain of the mouse receptor. Further analysis using primers specific to ObRb, which contains a long cytosolic domain with the intracellular protein motifs required for signalling [ 10 ], demonstrates the presence of this receptor isoform in GT1-7 cells (Figure 7A ). We have also used this hypothalamic cell line to examine whether leptin is capable of increasing PI3K activity directly. In response to 50 nM leptin, IRS-2-associated PI3K activity was modestly, but significantly, increased (Figure 7B ). Thus, using native GT1-7 cells leptin (1–10 nM) induced a decrease in cortical F-actin as visualised by alexa 488 conjugated phalloidin staining, which was prevented or reversed by the presence of 100 nM jasplakinolide (n = 8; Figure 8 upper panels). As cellular cortical actin structure is determined by the dynamic equilibrium between F- and G-actin, a reduction in F-actin at the plasma membrane should be accompanied by a corresponding increase in the concentration of free G-actin in the cells [ 27 ]. Figure 8 (middle panel) demonstrates, using alexa 594 conjugated DNase I staining of the same cells, that leptin does indeed increase the concentration of G-actin and that this effect is also sensitive to the presence of jasplakinolide (n = 8). Indeed, dual staining of the GT1-7 cells demonstrates (Figure 8 , lower panel) that leptin alteration of the cortical cytoskeleton is due to a concomitant increase in the content of G-actin at the expense of F-actin, and that this action is completely inhibited in the presence of jasplakinolide. The alteration in cytoskeletal dynamics by leptin is also PI3K dependent as shown in Figure 9A , where the presence of either 10 nM wortmannin (n = 13) or 10 μM LY294002 (n = 13) substantially reduced the ability of leptin to decrease the levels of F-actin and increase G-actin as assessed by phalloidin and DNase I staining respectively. This cell staining method of assessing leptin stimulated changes in actin dynamics was compared to direct quantitative analysis of actin. Live cells were treated with leptin (10 nM) ± jasplakinolide (100 nM) or LY294002 (10 μM) or wortmannin (10 nM) for 20 minutes, Triton-X-100 soluble (G) and insoluble (F) actin fractions separated and run on a gel [ 28 ]. Exposure of cells to leptin did not alter total cellular actin, whereas G-actin levels increased by 2 fold, at the expense of F-actin, the levels of which declined by 65% (n = 4; Figure 9C,9D ). Thus, leptin induced a change in the G/F actin ratio from a control value of 0.54 to 3.17. These data correlate well with the change in fluorescence intensity observed in leptin-treated fixed cells (n = 8), where G-actin levels were also increased by 2 fold and F-actin decreased by 70% (Figure 9B ). Exposure of cells to the F-actin stabilizing agent, jasplakinolide or the PI3K inhibitors, LY294002 or wortmannin prevented leptin from inducing F-actin disassembly as observed by either assay (Figure 9A,9B,9C,9D ). In addition, protein overlay experiments using wild type PH-GRP1-GFP fusion protein binding to assess PtdIns(3,4,5)P 3 levels in GT1-7 cells demonstrate that leptin increases PI3K activity concurrently with the re-organization of cortical actin in this cell line, with leptin stimulation inducing little change in the K273A mutant PH-GRP1-GFP binding to these cells (Figure 10 ). Figure 7 Leptin stimulates PI3K activity in GT1-7 cells A , expression of the leptin receptor mRNA in mouse GT1-7 cells. Lanes 1–3, RT-PCR detection of the common ObR isoform in hypothalamus (lane 2) and GT1-7 cells (lane 3), together with a negative control (lane 1). Lanes 4–6, RT-PCR detection of the ObRb isoform in the hypothalamus (lane 5) and GT1-7 cells (lane 6) together with a negative control (lane 4). Note the presence of PCR products of the appropriate sizes in GT1-7 and hypothalamus (465 bp ObR and 647 bp ObRb). B , PI3K activity associated with IRS-2 in GT1-7 cells stimulated with 50 nM leptin. PI3K activity was measured in immunoprecipitates and was quantitated using a Phosphoimager. Data are mean ± SEM for 4 experiments. P < 0.01. Figure 8 Leptin disrupts cortical actin filaments in GT1-7 cells Cultured GT1-7 cells were incubated in the absence and presence of leptin (10 nM) ± jasplakinolide (100 nM) for 30 minutes (jasplakinolide added 10 minutes prior to leptin). Following treatment cells were fixed and permeabilized, as described in the Methods, incubated with Alexa 488 conjugated phalloidin and Alexa 594 conjugated DNase I and subsequently processed for visualising F- and G-actin respectively by confocal microscopy. Figure 9 PI3K mediates F-actin disruption in GT1-7 cells A , Cultured GT1-7 cells were incubated in the absence and presence of leptin (10 nM) ± wortmannin (10 nM) or LY 294002 (10 μM) for 30 minutes. Following treatment cells were fixed and permeabilized, as described in the Methods, incubated with Alexa 488 conjugated phalloidin and Alexa 594 conjugated DNase I and subsequently processed for visualising F- and G-actin respectively by confocal microscopy. B , Plot of average alexa 488-phalloidin fluorescence intensity (green) and alexa 594-DNase 1 (red) in fixed cells treated with 10 nM leptin (L) alone or cells treated with leptin and jasplakinolide (100 nM; J + L), LY294002 (10 μM; L + L) or wortmannin (10 nM; W + L) relative to cells untreated (C) with drug (n = 8 separate experiments, with 8 cells measured under each condition for each experiment). C , GT1-7 cells were incubated with PBS only (C), 10 nM leptin (L) or leptin and jasplakinolide (100 nM; J + L), LY294002 (10 μM; L + L) or wortmannin (10 nM; W + L). Cells were treated to extract actin pools as described in Methods and equal amounts of pool lysate were subjected to SDS-PAGE and transferred to nitrocellulose membrane. The levels of actin were detected by immunoblotting with an actin monoclonal antibody. D , Plot of average Triton-X-100 soluble (G, red), Triton-X-100 insoluble (F, green) and total actin (gray) concentration from live cells, relative to control untreated cells (n = 4 separate experiments), for data as shown in C . Error bars indicate s.e.m. and * significance of P < 0.01. Figure 10 Re-organisation of F-actin is associated with raised PtdIns(3,4,5)P 3 levels in GT1-7 cells Cultured GT1-7 cells were incubated in the absence and presence of 10 nM leptin for 20 minutes. Cells were fixed and permeabilized, as described in Methods, prior to incubation with rhodamine-conjugated phalloidin and wild type (wt) or K273A mutant (mt) PH-GRP1-GFP fusion protein for 60 minutes. Cells were subsequently processed for visualising GFP and rhodamine by confocal microscopy. Discussion PI3K – a pivotal enzyme in ARC signalling Previous studies have demonstrated that leptin applied in vivo stimulates hypothalamic ObRb to increase phosphorylation of the signalling protein intermediates STAT3 and MAPK and that both leptin and insulin increase hypothalamic PI3K activity [ 12 , 29 ]. Here we have applied hormones directly to ARC wedges isolated from hypothalamic slices to enable improved signal detection (with respect to amplitude and temporal resolution), localisation of signalling to the arcuate nucleus and to fix external conditions so that potential compensatory changes associated with in vivo studies are obviated. Exposure of ARC wedges to leptin or insulin induced rapid (≤1 minute) phosphorylation of MAPK (ERK1 & 2 subfamilies), STAT3 and the PI3K activity indicators, PKB and its downstream target GSK3. These hormone-induced increases in phosphorylation were transient in the majority of experiments, usually lasting 1–5 minutes at ~34°C with return to control levels of phosphorylation within 30 minutes. Such rapid recovery has also been noted in other studies [ 13 , 30 ] and may be due to activation of endogenous phosphatases such as PTP1B curtailing this acute signalling process [ 14 , 31 ]. The phosphorylation of MAPK is quite modest and at present there are few data which link this pathway directly with the actions of either insulin [ 32 ] or leptin [ 16 ] on energy homeostasis, although recently it has been shown that centrally driven insulin-mediated sympathoactivation of brown adipose tissue is MAPK-dependent [ 33 ]. As expected, exposure of ARC wedges to leptin induced an increase in tyrosine phosphorylated STAT3 [ 11 , 13 , 29 , 34 ]. However, unexpectedly insulin also induced an increase in tyrosine phosphorylation of STAT3 in ARC neurones. In a previous study [ 30 ] in vivo application of insulin (icv) demonstrated no such change, unless leptin was co-applied. The data reported here indicate that insulin per se is capable of increasing STAT3 phosphorylation, as no exogenous leptin was present or endogenous leptin likely to remain in the ARC sections following the extensive washes and incubations prior to stimulation. This difference may be due to an increased signal to noise delivered using ARC tissue over whole hypothalamus and that rapid transient signals are more readily detectable by this method. Both leptin and insulin rapidly increased the phosphorylation of PKB and its downstream effector GSK3 in a wortmannin and LY294002 sensitive manner, indicative of increased PI3K activity in ARC neurones, in agreement with previous in vivo studies [ 17 , 35 ]. However, our results did not demonstrate that either leptin or insulin induced a significant increase in IRS-2-associated PI3K activity measured directly in ARC tissue. This may be due to a low signal to noise ratio, as only a (unknown) proportion of cells would be expected to respond to the hormones in the ARC tissue block, and/or that hormone mediated increases in PI3K activity are limited to plasma membrane microdomains. This is supported by the very modest increase in PI3K activity detected in GT1-7 cells when stimulated by leptin. Although hypothalamic activation of PKB by insulin has been reported previously [ 17 ], these are the first reports that leptin increases PKB activity and that both hormones increase the phosphorylation of GSK3 in the ARC. The presence of the PI3K inhibitors, wortmannin or LY294002, also reduced the leptin and insulin driven increase in MAPK phosphorylation. The mechanism by which leptin and insulin cause phosphorylation of this protein is most likely through the Ras pathway, as this protein has been demonstrated to interact directly with the catalytic subunit of PI3K [ 36 ] and inhibitors of PI3K have been reported to inhibit insulin induced increased MAPK activity, for example in rat adipocytes [ 37 ]. The insulin mediated enhanced STAT3 tyrosine phosphorylation in an interesting observation that requires further examination. Although phosphorylation of tyrosine-705 on STAT3 is a prerequisite for dimerisation and translocation of STAT3 to the nucleus [ 38 ], phosphorylation of serine-727 may also be required for maximal activation of STAT3 DNA binding [ 39 ]. Interestingly one pathway candidate for phosphorylating serine-727 is the Ras/Raf/MEK signalling cascade, and indeed a recent study has demonstrated that leptin can induce S727 phosphorylation of STAT3 in a PD98059 dependent manner in macrophages, and this is required to produce full stimulation of STAT3 [ 40 ]. Insulin mediated serine phosphorylation of STAT3 has also been reported, using transfected Chinese hamster ovary cells, to be mediated by a MEK-dependent pathway [ 41 ]. A similar mechanism in hypothalamic neurones would indicate an inter-connection between the three identified signalling pathways activated by these hormones and an important effector molecule, STAT3. Studies are underway to examine this proposal. The importance of STAT3 signalling to the central mechanisms that control energy homeostasis has recently been directly demonstrated by transgenic mouse studies. Using a 'knock-in' strategy to induce defective STAT3 binding to ObRb [ 42 ] or a 'knock-out' strategy to ablate STAT3 from some hypothalamic neurones [ 43 ], loss or reduction in hypothalamic STAT3 signalling initiates hyperphagia, increased body weight and adiposity with alterations in glucose homeostasis. Indeed, the JAK2-STAT3 and IRS2-PI3K signalling pathways are purported to underpin the genomic and acute or membrane functions of these signalling pathways respectively [ 12 ]. Clearly, further work is required to determine the exact signalling mechanisms controlling insulin stimulated STAT3 phosphorylation in hypothalamic neurones. Leptin and insulin signalling to K ATP channels Leptin and insulin cause inhibition, by hyperpolarization through activation of a sulphonylurea-sensitive K + conductance, of a subset of hypothalamic neurones, defined by their acute sensitivity to changes in external glucose concentration, termed GR neurones [ 6 , 7 , 19 ]. Single channel recordings from acutely isolated ARC neurones demonstrate that both hormones activate the same K + channel, the sulphonylurea-sensitive large conductance K ATP channel. This action is rapid and independent of transcriptional events, so most likely is mediated by MAPK or PI3K signalling. Pharmacological inhibition of the MAPK pathway with PD98059 did not reverse leptin (as shown above) or insulin [ 7 ] stimulated K ATP activity, abrogating this pathway from causing the hyperpolarising response. In contrast, inhibition of PI3K with either wortmannin or LY294002, reversed both leptin (as shown above) and insulin [ 7 ] raised K ATP activity. Furthermore, use of the fusion protein PH-GRP1-GFP as a specific detector of PtdIns(3,4,5)P 3 in isolated neurones also demonstrated that both hormones rapidly increase the cellular content of this PI3K lipid product in a sub-population of neurones. These results are consistent with class 1 PI3K [ 44 ] acting as a point of convergence for leptin and insulin signal transduction pathways to K ATP channels in GR neurones. The functional significance of PI3K in the control of energy balance has been demonstrated by in vivo studies, which show that leptin [ 16 ] and insulin [ 17 ] stimulate IRS2-associated PI3K activity in the hypothalamus and pharmacological inhibition, using wortmannin and LY294002, of hypothalamic PI3K activity prevents the anorectic actions of icv leptin or insulin, whereas the MAPK inhibitor PD98059 had no effect on leptin driven attenuation of food intake [ 16 ]. Remodelling of cortical actin filaments as a leptin and insulin signalling event Leptin and insulin stimulated K ATP activity in isolated ARC neurones was also reversed, within 5–10 minutes, by the marine sponge toxin, jasplakinolide. This toxin binds to F-actin with high affinity, resulting in its stabilization and prevention of depolymerization to its monomer G-actin [ 24 ]. These data indicate that the adiposity hormones require actin filament depolymerization for K ATP activation to occur. Such a mechanism is supported by reports that agents, which promote actin depolymerization, activate K ATP channels in cardiac myocytes [ 45 , 46 ] and the insulin-secreting cell line, CRI-G1 [ 47 ]. Furthermore, in this latter study leptin stimulated K ATP channel activity was also shown to depend on actin filament depolymerization. Insulin is also well documented to cause actin filament re-organization in peripheral cells associated with various functional outputs, which depend on PI3K activity, including metabolic and mitogenic effects [ 48 ]. The reversal of hormone-stimulated K ATP activity by jasplakinolide was faster (5–10 minutes) than for the PI3K inhibitors (15–20 minutes). This temporal difference suggests that the site of jasplakinolide action is downstream from the PI3K signal transduction pathway to K ATP channels. However, alteration of the cellular cortical actin structure is inferred through the use of natural agents like jasplakinolide. In order to verify directly that hormone-driven structural re-arrangements did occur we decided to use the hypothalamic cell line, GT1-7, as preliminary experiments using freshly isolated neurones did not produce reliable and reproducible data due to the presence of dead and dying cells showing as false positives for hormone induced actin depolymerization. Use of this cell line also obviated any problems with identification of ObRb containing neurones and neuronal subtypes in slices. RT-PCR analysis indicates that this cell line does express the main signalling form of the leptin receptor and analysis of PI3K activity shows functional coupling of this receptor to this signalling pathway. We have shown, by cell staining of fixed cells and, independently by analysis of cellular G- and F-actin concentration from live cells, that leptin disrupts cortical actin structure by disturbing the processes that maintain the equilibrium between F-actin and G-actin, in the direction of depolymerization to G-actin. This effect of leptin was completely inhibited by the presence of either jasplakinolide or the PI3K inhibitors. In addition, there is a good temporal and spatial association between PtdIns(3,4,5)P 3 production, as determined by PH-GRP1-GFP binding, and actin filament depolymerization. Thus, leptin and insulin signalling in, at least some sub-groups of hypothalamic neurones maintains a close parallel with leptin signalling in insulin-secreting cells, where it has been reported that leptin increases K ATP activity by a PI3K-dependent cortical actin re-arrangement [ 47 ]. Conclusions The effect of leptin and insulin on the phosphorylation status of various cellular signalling intermediates and on K ATP channel activation in arcuate neurones indicates that both hormones activate the same signalling cascades, and can produce common outputs. The sensitivity of both K ATP opening and the phosphorylation of certain intermediates to PI3K inhibition is significant as this enzyme has been previously demonstrated to play an important role in leptin and insulin mediated energy homeostasis control. Furthermore it is interesting that leptin and insulin induce rapid phosphorylation of MAPK and STAT3 as these data support the view that these hormones may influence genomic and membrane neuronal outputs by common mechanisms. The inhibition of leptin and insulin stimulation of K ATP channel opening of arcuate neurones by jasplakinolide suggests a role for cytoskeletal dynamics in modulation of membrane events such as neuronal hyperpolarization. This hypothesis is further strengthened by the finding that leptin induces actin filament depolymerization in a mouse hypothalamic cell line, which is PI3K dependent, demonstrating that this cell line may be a useful model for further analysis of leptin signalling mechanisms in hypothalamic neurones. Methods Preparation of hypothalamic lysates and immunoblots Male Sprague-Dawley rats (50–100 g) were killed by cervical dislocation in accordance with Schedule 1 of the UK Government Animals (Scientific Procedures) Act (1986). The brain was rapidly transferred to ice-cold aCSF solution, containing (in mM): 128 NaCl, 5 KCl, 1.2 NaH 2 PO 4 , 26 NaHCO 3 , 1.2 CaCl 2 , 2.4 MgSO 4 , and 10 glucose, equilibrated with 95% O 2 , 5% CO 2 to give a pH of 7.4. The tissue was maintained in ice-cold aCSF whilst horizontal 400 μm coronal brain slices were prepared using a Vibratome (Intracel, Royston, Herts. UK). Slices containing the ARC were incubated in aCSF at room temperature for 20 minutes, and then at 33–35°C for 1 hour. Hypothalamic wedges, predominantly containing the ARC were cut, and these were incubated in aCSF ± hormones and/or kinase inhibitors (10 mls) for the required time. The reaction was stopped by the addition of 2 ml of cold lysis buffer containing (in mM) 100 NaCl, 10 NaF, 25 Tris HCl, 10 NaPP i , 5 EGTA, 1 EDTA, 1 Na 3 VO 4 , 1 Benzamidine, 0.1 PMSF, 0.1% (v/v) mercaptoethanol, 1% Tritron X-100 (v/v) and 92 mg ml -1 sucrose. The tissue was homogenised on ice, the lysate sonicated for two 10 s periods and then centrifuged for 10 minutes at 12000 rpm at 4°C. The supernatant was retained and the pellet discarded. The protein content of the clarified lysate was determined by the method of Bradford [ 49 ]. Proteins (10 μg) were separated by SDS-PAGE, and subsequently transferred to nitrocellulose membranes. Membranes were incubated in blocking buffer (10% non-fat dried milk in TBST (20 mM Tris HCl, 150 mM NaCl, 0.5% Tween, pH 7.4)) for 1 hour at room temperature following which phospho-specific p44/p42 MAPK (Thr202/Tyr204), phospho-specific STAT3 (Tyr705), phospho-specific GSK-3α/β(Ser21/9), phospho-specific PKB (Thr308) and PKB (all polyclonal and used at 1:1000) antibodies were applied overnight at 4°C with gentle shaking. All antibodies were obtained from Cell Signalling Technology Inc. The membranes were washed four times with TBST and incubated for 1 hour at room temperature with horseradish peroxidase conjugated Goat anti-Rabbit IgG (1:5000). After further washing with TBST, total amount of specific protein was visualised by enhanced chemiluminescence detection as described by the manufacturer (NEN Life Science Products). Immunoreactive bands were scanned and quantified using AIDA software. As an internal control, the membranes were immunoblotted with a monoclonal anti β-actin antibody (Sigma: used at 1:5000) or with the PKB antibody. The values for proteins were normalized with respect to the internal control to account for variations in gel loading. Determination of PI 3-kinase activity Cell and tissue lysates were made as described. The immunoprecipitation and PI3K activity assay were carried out as previously described [ 50 ]. Briefly, frozen samples were thawed before centrifugation to remove precipitated material. 10 μl Protein-G-Sepharose beads pre-coupled to 5 μg anti-IRS2 antibody (Upstate Biotechnology) was used to immunoprecipitate PI3K activity from ~0.5 mg cell lysate. The immunoprecipitated material was washed once with ice cold lysis buffer and three times with ice cold assay buffer, both of which were freshly supplemented with protease inhibitors, reducing agent and sodium vanadate as described [ 51 ]. Washed beads were re-suspended in 40 μl assay buffer supplemented with 1 μM unlabelled ATP, 25 μCi/assay radiolabelled ATP and phosphatidylinositol/phosphatidylethanolamine vesicles (final concentration of each lipid 100 μM). Samples were incubated at 37°C for 30 mins and the reaction was stopped by addition of 0.6 ml methanol/chloroform/12 M HCl (80:40:1, v/v), 0.2 ml chloroform and 0.32 ml 0.1 M HCl. Samples were processed and PtdIns(3)P separated from contaminating materials by thin layer chromatography (TLC) as previously described [ 51 ]. Bands corresponding to [ 32 P]PtdIns(3)P were located using a phosphorimager (Fuji FLA 5000) and analyzed with AIDA software. Preparation of acutely isolated ARC neurones and electrophysiology Coronal slices containing the medial hypothalamus were obtained (as described above) and sections containing the ARC were removed. The sections were transferred to 5 ml aCSF containing 1 mg ml -1 protease XIV (Sigma-Aldrich, Dorset, U.K) and incubated for 1 hour at room temperature. The aCSF was continuously gassed with 95% O 2 : 5% CO 2 for the entire incubation period. Sections were removed and washed in 50 ml aCSF five times prior to re-suspension in 5 ml normal saline containing (in mM): 135 NaCl, 5 KCl, 1 MgCl 2 , 1 CaCl 2 , 10 HEPES, 3 glucose, pH 7.4. Sections were sequentially triturated with fire polished Pasteur pipettes with decreasing tip size. The cell suspension was evenly distributed onto concanavalin A (Sigma-Aldrich) pre-treated 35 mm diameter culture dishes. The culture dishes were left for 15–20 minutes allowing cell adhesion prior to use. Cell-attached single channel currents were recorded from single neurones at room temperature, using an Axopatch 200B amplifier (Axon Instruments, Foster City, CA USA). Patch pipettes were prepared from thick walled borosilicate glass and had open tip resistances of 8 – 15 MΩ when filled with high K + solution containing (in mM) 140 KCl, 1 MgCl 2 , 1 CaCl 2 , 10 HEPES, pH 7.2. This solution was used in order to allow easy identification of K + currents in the cell-attached configuration [ 19 ]. All cell-attached recordings were made in the presence of normal saline, with no applied pipette potential, thus utilizing the cell membrane potential to drive current flow (with inward current shown as downward deflections). Single channel recordings from inside-out patches isolated from ARC neurones were made either under asymmetrical conditions, in the presence of normal saline, or under symmetrical K + conditions with the intracellular aspect of the membrane exposed to a bathing solution containing (in mM): 140 KCl, 1 MgCl 2 , 2.7 CaCl 2 , EGTA 10 (free Ca 2+ of 100 nM), HEPES 10, pH 7.2. Data were recorded onto digital audio-tape using a Biologic DTR 1200 recorder and analysed off-line. Pre-recorded data were transferred via a Digidata 1200 interface into a PC, digitised at 10 kHz and measured using the PCLAMP6 software, Fetchan 6. The mean current (I) and single channel amplitude (i) were determined for recordings ranging in duration from 30 s to 120 s and channel activity (N f .P o ) determined as described previously [ 52 ], where N f is the number of functional channels and P o is the open probability. Drug effects were measured by comparison of N f .P o from individual patches in the presence and absence of the drug. Data for a given set of experiments were normalised and statistical significance determined by employing the Students t -test for unpaired data. Results are presented as mean ± SEM and the number of experiments denoted by 'n.' Leptin receptor mRNA expression Reverse transcription was performed in a 20 μl reaction containing 1 × First Strand Buffer, 1 mM DTT, 0.5 mM of dNTP, 0.5 μg anchored oligo(dT)18, 4 μg RNA and 1 μl (200 U) M-MLV Reverse Transcriptase (Gibco), at 25°C for 5 minutes, 42°C for 60 minutes, 70°C for 15 minutes and stored at -20°C. After RT, a 2 μl aliquot of the reaction was added to 48 μl of PCR mix. The mix containing 1 × PCR buffer, 2.5 mM MgCl 2 , 0.5 mM PCR nucleotide mix, 1 μM each of the gene specific primers (mObRcom F:ggaatgagcaaggtcaaaa; mObRcom R:gtgacttccatatgcaaacc; mObRb F:tcttctggagcctgaacccatttc; mObRb R:ttctcaccagaggtccctaaact; ref [ 53 ]) and 5 units of Taq DNA polymerase (Promega). PCR was performed using the following profile: 94°C for 5 minutes, 25 cycles at 94°C for 30 seconds, 55°C for 30 seconds, 72°C for 30 seconds, with a final extension at 72°C for 7 minutes. GT1-7 cell culture, staining and actin analysis The mouse hypothalamic cell line GT1-7 [ 54 ] was grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (Sigma), 1 mM L-glutamine and 1% penicillin-streptomycin at 37°C in a humidified atmosphere of 95% air and 5% CO 2 . Cells were passaged every 3–4 days, plated on poly-L-lysine (Sigma) coated glass coverslips in 3.5 cm Petri dishes and used 1–2 days after plating. Cells were treated with 10 μM LY294002 or 10 nM wortmannin or 100 nM jasplakinolide (all Sigma) in normal saline for 10 minutes, prior to a challenge with 10 nM leptin (or saline), in the continuous presence of inhibitor, for 20 or 60 minutes before fixing. GT1-7 cells were fixed in 4% methanol-free formaldehyde in cytoskeletal buffer (10 mM MES, 3 mM MgCl2, 138 mM KCl, 2 mM EGTA, pH 6.1) with 0.32 M sucrose for 30 minutes [ 28 ]. They were then washed in phosphate-buffered saline (PBS), permeabilised in PBS/0.5% Triton X-100 for 10 minutes, rinsed in PBS, blocked with 20% goat serum (Sigma) for 30 minutes, rinsed in PBS and incubated with rhodamine conjugated phalloidin, or 2 μg ml -1 Alexa 594-DNase I and 2 U ml -1 Alexa 488-phalloidin (all Molecular Probes) for 90–120 minutes, rinsed in PBS, and mounted on coverslips. Cells were observed with a 63X oil objective and images acquired using a laser-scanning confocal microscope (Zeiss LSM 510), under identical conditions with randomly selected regions of each coverslip. For quantitative analysis of G- and F-actin cellular pools, we used a direct method to partition the actin pools from live cells [ 28 ]. In brief, equal cell numbers were added to 3.5 cm culture dishes and cells grown to 80% confluence. The Triton-X-100 soluble (G-actin) pool was isolated first, by incubating cells for 5 minutes at room temperature with 1 ml PBS containing 1% Triton-X-100, protease inhibitors and 1 μg ml -1 phalloidin (to prevent filament dissociation). Cells were then washed with PBS, and the Triton-X-100 insoluble pool (F-actin) prepared by addition of 1 ml of PBS lysis buffer, containing 1% Triton-X-100, protease inhibitors, 2% SDS and 1 μg ml -1 phalloidin for 5 minutes prior to harvesting cells from dishes. For determination of total actin, cells were exposed to the second step only. Each cellular pool was passed through a 25 gauge needle and total protein concentration determined, before equal amounts of protein were loaded onto SDS-PAGE gels, and actin detected using an actin monoclonal antibody (Chemicon). Quantitative measurements of G- and F-actin in fixed cells were made using Velocity software (Improvision), where individual cell total fluorescence, normalized to cell area, was determined and background fluorescence subtracted. Average fluorescence intensity was calculated for 8 cells in each experiment, and expressed relative to control (non-drug exposed cells). Actin bands on gels were quantified by densitometry, where total density was determined with respect to constant area, background subtracted and average relative band density calculated. PH-GRP1-GFP fusion protein overlays Following stimulation with hormone for 10–20 minutes, acutely isolated neurones (room temperature) or GT1-7 cells (37°C) were fixed at room temperature with 2–4% paraformaldehyde for 15 and 30 mins, respectively. Cells were permeabilized by washing with 0.05% PBS-Tween 20 (PBS-T; x2 for 10 mins). Non-specific binding was minimised by blocking with 3% BSA for 1 hour at room temperature. Cells were subsequently washed with 0.05% PBS-T prior to incubation with wild type PH-GRP1-GFP or K273A mutant PH-GRP1-GFP (50 μg ml -1 ) fusion protein for 1 hour at room temperature, and images acquired by confocal microscopy. List of abbreviations ACSF, artificial cerebrospinal fluid; AgRP, agouti-related protein; ARC, arcuate nucleus; NPY, neuropeptide Y; POMC, proopiomelanocortin; K ATP , ATP-sensitive potassium channel; BBB, blood-brain-barrier; DNase I, deoxyribonuclease I; F-actin, filamentous actin; G-actin, globular actin; GFP, green fluorescent protein; GR neurone, glucose-responsive neurone; GRP1, general receptor for phosphoinositides-1; GSK3, glycogen synthase kinase 3; IRS2, insulin receptor substrate 2; JAK2, janus kinase 2; MAPK, mitogen-activated protein kinase; MEK, MAPK kinase; ObR, leptin receptor; PH domain, pleckstrin homology domain; PI3K, phosphoinositide 3-kinase; PKB, protein kinase B; PtdIns(3,4,5)P 3 , phosphatidylinositol 3,4,5-trisphosphate; STAT3, signal transducer and activator of transcription 3; Authors' contributions SM carried out the electrophysiology studies, participated in the western blot and protein overlay experiments. HL carried out the arcuate western blot experiments and PI3K activity measurements. KN carried out the actin imaging and actin quantitative analysis experiments. EA contributed to the actin imaging experiments, western blots and participated in the protein overlay experiments. LB performed all tissue culture and participated in the western blot experiments. AG made the fusion proteins and participated in the design of the overlay experiments. CS participated in the design and implementation of the western blot experiments. MA conceived of the study, participated in its design and co-ordination and drafted the manuscript. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC539348.xml
368162	The DNA Story, Part III	Maurice Wilkins's autobiography provides an engaging perspective of the events leading to the discovery of the structure of DNA	As the year-long celebration of the 50 th anniversary of the discovery of the structure of DNA came to an end, the engaging autobiography of one of the participants further enlivened the drama of this event. Maurice Wilkins, now 87, postpones the account of his involvement in the DNA affair until the second half of the book. Recounting his background and interesting life before DNA (34 years) in plain but telling sentences brings to life a character that is almost as much out of the ordinary as those of the more flamboyant James Watson and Francis Crick. Wilkins' first six years in New Zealand (a Garden of Eden) were followed by a long, vividly described trip to England, where the family eventually settled in Birmingham. His boyhood was marked by immersion in astronomy and telescope-making, but saddened by the painful illness of his sister. Success in school physics was the key for getting into Cambridge, where he reveled in the world of Ernest Rutherford, Mark Oliphant, and John Bernal. Given his leftist leanings, it was inevitable that Wilkins would become involved in the pacifist movement in Cambridge, with its close connection to the Communist Party. Perhaps too much involvement led to a low degree grade in 1938 and no hope of remaining at Cambridge. Instead, he returned to Birmingham and joined the Luminescence Lab being established by John Randall, a man with whom he would be closely connected for many decades. The work there contributed to Randall's scheme for making radar practical in air defense—the cavity magnetron that may have turned the course of World War II. Early in 1944, Oliphant, then at Birmingham, left to work on the atomic bomb at Berkeley and took Wilkins along. Life in Berkeley was exciting, but beneath the excitement of bomb work and mixed feelings upon its success at Hiroshima, Wilkins read Erwin Schrodinger's What Is Life? Along with others who were to unravel the secrets of DNA, this planted the seed. When, after three transitional years, Randall became head of Kings College London's physics department and director of a biophysics research unit sponsored by the Medical Research Council (MRC), Wilkins was his deputy. The attack on DNA structure soon began. That X-ray diffraction might play a major role in this search rested on two pillars unique to England. One was the British lead in using X-ray diffraction to determine molecular structures—a crown jewel built on the work of the Braggs (father and son), Bernal, and Dorothy Hodgkin. The other was the pre-World War II work of William Astbury in showing that DNA fibers displayed some crystallinity that, if developed, might be the basis of helping to determine the structure. Wilkins confides that in 1950 he knew little of how such X-ray analysis might be done. But in that year he was presented with an opportunity in the form of samples of carefully prepared calf DNA, given to him by a Swiss chemist, Rudolf Signer. With this DNA, much better fibers could be obtained and much sharper diffraction diagrams emerged. The exploitation of this advance, however, became mired in a colossal error in Randall's management of the group. Without telling Wilkins, he wrote to Rosalind Franklin, who was on her way to join the DNA effort, that Wilkins was withdrawing from DNA work and that she would take over. Unaware of this, Wilkins and Alec Stokes continued their work and reported at a meeting in Cambridge in July 1951 that DNA chains were probably in a helical conformation with a diameter of 20 Å. At the close of the meeting, Franklin assailed Wilkins, saying that he should stop his DNA work (as Randall had written would be the case). Understandably, but regrettably, the two groups continued working in isolation from each other. Matters worsened. In October, Watson arrived at Cambridge and set up DNA structure studies with Crick. They quickly arrived at a three-stranded helical structure. But Franklin and Wilkins soon demolished it. Likewise, a three-stranded model at Kings College had a very short life. As if to trump these failures, Bragg at Cambridge and Randall at Kings agreed that DNA studies at Cambridge should stop and that the work should continue only at Kings. Mismanagement and noncooperation were taking their toll. Franklin was moving toward a two-stranded structure, but away from helices. Indeed, in mid-1952 she initiated a discussion with an announcement about the death of the helix. Mysteriously, she put aside a striking photo of the diffraction pattern of B-DNA (one of the two major structural forms of DNA) that emerged in early 1953 as a perfect signature of the helical form. But 1952 continued downhill. Even Wilkins stopped DNA work that November. Suddenly, in the new year, life returned to the DNA effort. Linus Pauling had just published a structure (three-stranded) that did not long survive, but the entrance of the world's leading structural chemist into the race reawakened everyone to the centrality of DNA structure. In January, Raymond Gosling gave to Wilkins the very well-oriented diffraction photo of B-DNA that he and Franklin had taken in July 1952. Wilkins assumed that it was given to him to do as he wished; a few days later, he showed it to Watson. Though hardly an expert in X-ray diffraction, Watson sensed that it was strong evidence for helices and sketched it for Crick on his return to Cambridge. Later that January, Franklin announced she would be moving from Kings College to Birkbeck College to join Bernal's group. In giving her final seminar, she switched from her earlier insistence that B-DNA was nonhelical, but did not show the photo that gave the strongest evidence for helicity. This shift put Franklin in a position to move forward on the structure of DNA, but without others' resorting to model building, the goal would have remained elusive. Finally, in mid-February, Max Perutz, who was a member of the MRC committee overseeing the Biophysics Unit at Kings College, passed on to Crick his copy of a report from that unit. This report contained Franklin's results that the phosphates were on the outside and that the A-form of DNA had a special crystalline arrangement called the monoclinic C2 space group. From his work with proteins, Crick saw immediately that the chains in the helical structure must be antiparallel and that there were probably two chains entwined. Watson used other data in the report to deduce that there were indeed two chains, not three or four. Erwin Chargaff had recently shown that in the base composition of all DNAs examined, adenines and thymines as well as guanines and cytosines are equal, i.e., A = T and G = C. Now released from the ban on DNA studies, Watson and Crick engaged in a frantic search using model building. They found a unique way to fit the bases in the structure by pairing, and by March 7 they had the double-helix model constructed: it obeyed the Chargaff ratios, it fit the X-ray data for B-DNA, and it provided a rational way to encode and transmit genetic information to subsequent generations. Wilkins was invited to view the model in Cambridge. He found it stunning. Watson asked him to be a coauthor of the paper. Wilkins, true to his character, declined, as he had not been involved in the final monumental stage. Back in London, Franklin had already moved to Birkbeck. She received the news of the discovery with equanimity. But a later examination of her notebooks showed that she had moved to favor helices and a two-chain (or possibly a one-chain) model. With the rather complicated story of the greatest discovery in biology in the century now reasonably complete, what is one to make of it? There are many answers. I will mention only three. The first is the key role played by model building. In fiber diffraction there is not enough information, by orders of magnitude, to locate every atom, as would be possible in diffraction by perfect crystals that give thousands of sharp reflections. Instead, the fiber diagram can only provide cues and some specifics, such as the repeat distance. Model building is a way of bringing into the picture previously determined bond distances and bond angles of components such as the purine and pyrimidine bases and the sugars that are unavailable from the fiber diagram. That this was not seen at Kings College left the researchers there well behind in a field that they had pioneered. A second lesson is the importance of bringing the full knowledge of single crystal analysis to fiber diagram interpretation. That Franklin and Wilkins missed noting that the monoclinic C2 space group meant that the chains in the fiber had to be antiparallel robbed them of an important clue to the structure. And third, the management of the Biophysics Unit at Kings College was a recipe for failure. Riddled by secrecy, diffuse lines of authority, the absence of strategies, and a lack of open congeniality, all so well described by Wilkins, who refers to it as Randall's Circus, this unit is a model of how not to succeed in group research. DNA research continued at Kings College in a gradually improving environment: important details were worked out. But there was no real renewal, such as aiming at how DNA is configured to accommodate proteins in the nucleus. Wilkins enjoyed being included in the subsequent awards—the Lasker and the Nobel prizes. With Crick, he was annoyed by Watson's rendering of events in The Double Helix . The final chapter of his own autobiography addresses the criticism that some have leveled against his cold relation with Franklin, but also his happiness in newfound family life. Research gradually gave way to the pursuit of pacifist goals in a number of organizations and to the popularization of science. His has been a useful life, a part of which contributed to the great revolution in biology. It is good to have the insight that this book presents in a candid and personal way.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC368162.xml
548957	Correction: Components of Coated Vesicles and Nuclear Pore Complexes Share a Common Molecular Architecture	null	In PLoS Biology , volume 2, issue 12. 10.1371/journal.pbio.0020380 In Materials and Methods, the sentence “The magnetic beads were added to the extract to a ratio of about 8 ×10 9 beads per g of cells” contains an error. The correct amount of beads is 8 × 10 8 , i.e., ten times less. Published February 15, 2005	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC548957.xml
368176	Tracking Blood-Forming Stem Cells through Development	null	Of the 200-plus different types of cells that form the mammalian body, most have a finite life span. Like nearly everything in biology, there are exceptions—neurons and muscle cells, for example, can last a lifetime—but the vast majority of cells eventually wear out and must be replaced. Among the most short-lived cells, blood cells are generated continuously, mainly in the bone marrow of an adult, recharging the bloodstream as their depleted predecessors are efficiently dispatched and removed from circulation every 120 days. Some 2.5 million new red blood cells are generated every second from a small pool of stem cells. Fluorescence activated cell sorting is used to study the early development of hematopoietic stem cells Blood cell development, called hematopoiesis, passes through discrete stages in specific tissues in the developing embryo before converging in the bone marrow, where it continues throughout adulthood. Some researchers have proposed that hematopoietic stem cells (HSC) flood the bloodstream during short, precise intervals to build the developing hematopoietic system (which includes the liver, bone marrow, spleen, tonsils, and lymph nodes). Presenting an alternative model for HSC migration, Julie Christensen and her colleagues in Irving Weissman's lab at Stanford University report that HSC in mice gradually leave the fetal liver to colonize the developing spleen and bone marrow as the organs acquire the means to support them. In mouse embryos, HSC precursors develop first in the yolk sac and a region called the aorta-gonad-mesonephros (AGM), then they migrate to the liver, and later to the spleen, before finally settling into the bone marrow just before birth. It was thought that this migration occurs in distinct waves of HSC production because HSC numbers decrease in one region just before increasing in newly forming hematopoietic sites. Analyzing the concentration and activity of HSC, Christensen et al. found the cells in the blood at low but fairly constant levels during much of late fetal development, when they migrate from the liver to the spleen and bone marrow. Although the HSC population decreases in the liver at 15.5 days after conception, the authors propose that this drop occurs primarily because the HSC have differentiated into mature blood cells, not because they've exited the liver en masse to help build the spleen and bone marrow. On the other hand, the slight decrease in circulating HSC, which also occurs around this time, may be attributed to their recruitment from the bloodstream to these developing tissues. Christensen et al. also examined the impact of intercellular signaling proteins called chemokines, which help regulate fundamental developmental processes, on HSC migration. To effectively “seed” developing tissues, HSC must first be recruited from the blood, guided to the appropriate nascent tissue, then corralled and sustained. The chemokine SDF-1 attracts and retains HSC in the bone marrow but was thought to have a lesser effect on fetal liver HSC. Christensen et al. demonstrate not only that liver HSC migrate in response to this chemokine, but that their migratory response increases dramatically when both SDF-1 and a signaling protein called steel factor (SLF) are present. While adult marrow HSC respond to SDF-1, they do not respond to SLF alone and do not show improved migration in the presence of both SLF and SDF-1. Bone marrow transplants have become increasingly common for a number of hematological disorders, including leukemia and aplastic anemia. Since hematopoiesis occurs primarily in the bone marrow in both mice and humans after birth, these findings offer valuable insights into the migratory behavior of these stem cells and suggest how HSC migration might be applied to bone marrow transplants and other clinical therapies.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC368176.xml
544189	GenomeViz: visualizing microbial genomes	Background An increasing number of microbial genomes are being sequenced and deposited in public databases. In addition, several closely related strains are also being sequenced in order to understand the genetic basis of diversity and mechanisms that lead to the acquisition of new genetic traits. These exercises have necessitated the requirement for visualizing microbial genomes and performing genome comparisons on a finer scale. We have developed GenomeViz to enable rapid visualization and subsequent comparisons of several microbial genomes in an interactive environment. Results Here we describe a program that allows visualization of both qualitative and quantitative information from complete and partially sequenced microbial genomes. Using GenomeViz, data deriving from studies on genomic islands, gene/protein classifications, GC content, GC skew, whole genome alignments, microarrays and proteomics may be plotted. Several genomes can be visualized interactively at the same time from a comparative genomic perspective and publication quality circular genome plots can be created. Conclusions GenomeViz should allow researchers to perform visualization and comparative analysis of up to eight different microbial genomes simultaneously.	Background Current efforts in genome sequencing have led to a rapid increase in the number of microbial genome sequences. A total of 522 ongoing microbial genome projects are listed in the GOLD database [ 1 ] and while 167 microbes have been completely sequenced. These sequencing projects now include several bacterial pathogens and different isolates of the same bacterial species that differ with respect to virulence and physiology. Thus, complete and partial genome sequences of a number of closely related species/strains from various genera such as Escherichia , Bacillus , Helicobacter , Mycobacterium , Streptococcus , Staphylococcus and Listeria are available. Genomic data too, is diverse, ranging from COG functional classification data [ 2 ], genomic islands [ 3 , 4 ], expression data from microarrays and proteomics, GC skew, AT skew, GC%, to whole genome alignments. Such rapid increase in genomic information necessitates the development of tools that offer rapid and convenient visualization capabilities. Furthermore, it is important to contrast and compare data deriving from several different sources (computational, genomic, proteomic) to have a better understanding of genome function. Several genome visualization tools have been developed in the last few years. The Microbial Genomes Viewer [ 5 ] offers a good online solution to genomic visualization, allowing flexibility in using one's own data. However, the plot is not very interactive and provides no undo facility as once a mistake is made one has to recreate the entire plot. GenoMap [ 6 ] can be used to create circular genome plots. Although the visualization is helpful, only limited interaction is possible with the resulting plot. GenomeAtlas [ 7 ] provides picture-based structural DNA analysis for a large number of genomes via a web-interface. GenomePlot [ 8 ] also provides a method to render chromosome wheel plots using tab-delimited input files, although it lacks the interactivity with the pictures and requires a rather specific input file format that may have to be customized for each genome. BugView [ 9 ] is another application that allows comparative analysis of microbial genomes, however it allows only two genomes to be viewed and compared simultaneously. The linear plots are useful and offer much flexibility but the circular plots are static. Genome2D [ 10 ] offers useful visualization options for data visualization and integration of several algorithms for a single genome at a time. Artemis [ 11 ] and ACT [ 12 ] are convenient programs for visualizing single genomes or comparing multiple genomes on linear scales. Implementation GenomeViz has been programmed in ActiveTcl. ActiveTcl [ 13 ] is available freely from ActiveState. PERL [ 14 ] is needed to run the scripts available with GenomeViz. This is usually installed on Linux and Solaris systems but can also be freely downloaded. GenomeViz works only on Unix-based platforms and has been tested on Linux and Solaris operating systems. Currently, it does not work on Windows because of a bug in the Tcl library on Windows which causes narrow arcs on a canvas to be drawn incorrectly. We recommend a minimum of 512 MB RAM to run the program. Results and discussion GenomeViz uses the concept of "tags" which may be applied to groups of genes for classification-type data. A tag file is tab-delimited text file of three columns. It has the "tags", their colors, and their brief descriptions. A pre-prepared tag file for the COG functional categories is available for immediate use. The map file has all the information required to create the plot (gene name, strand, start and end in genome, annotation, and the tag for the gene). It is also a tab-delimited text file. Both file formats (tag and map) are easy to manipulate in a spreadsheet application like Microsoft Excel. However, care must be taken while manipulating data in such applications since errors may creep in the data as demonstrated by Zeeberg et al. [ 15 ]. The map file alone is sufficient for plotting numerical data, but both the map and tag files are needed to plot classification-type data. Data type (qualitative or quantitative) is automatically detected from the map file. A PERL script "tagit" is also available for "tagging" a particular set of genes with user-defined tags. Another script, "avid2viz" is also available which reformats whole genome alignments created by the AVID program [ 16 ] to a map file format that can be visualized in GenomeViz (Figure 1 ). In order to minimize initial difficulty that users may encounter in creating their own map files, we provide pre-prepared map files for over a hundred genomes. Of these nearly seventy genomes are loaded with the COG classification scheme and may be used immediately. The program also performs checks on the input map file for possible formatting errors and attempts to indicate location of errors (if any) before creating the plot. Several types of plots may be created; on either single or double strands and color gradients and line-graphs are available for numerical data. Once the plots are done, mouse-over on any gene immediately displays associated information from the map file in a display area. Using GenomeViz, it is also possible to search, highlight and retrieve genes of interest. Each loaded genome may be queried separately. Regular expression searches are fully supported and results are highlighted in the genome. For instance, the simple expression ribosomal\|ribosome will mark in color all ribosomal proteins in any genome and retrieve all the information for these genes from the map file. The "\|" operator is the standard OR operator in Tcl expressions. Genes involved in iron metabolism/regulation which are usually annotated with keywords like ferrous, ferric or iron may be retrieved with the expression "ferric\|ferrous\|iron". The results can be saved as a text file. Users can also use their own annotations to visualize and query their genome of interest provided these annotations are available in the map file format. It is also possible to display, in different colors, the results of different queries on the same genome by changing the search color before performing a search. This will enable visualization of, for instance, the distribution of genes/operons involved in iron and zinc metabolism/regulation separately. A 'Select COGs' option enables one to retrieve all genes from a particular COG category, e.g. "Cell division and Chromosome partitioning" or "Transcription". Each loaded genome can thus be queried separately. Usually, this is more useful when using a special tag file (CogsGrayScale.tag) that colors all genes as "grey", so that a neutral background is available for highlighting the distribution of genes of a particular COG category over the entire genome. Categories of interest can be highlighted in different colors simply by changing the selection color before selecting the category. Results of each query are also displayed in a text box from where they may be saved as a text file. Thus, GenomeViz allows a rapid overview of the similarities in distribution of various functional categories in closely related genomes (Figure 2 ). It is also possible to visualize differences/similarities in data derived from various different sources e.g. horizontally transferred genes (Figure 3 ). Several options are available for printing the circular plot. The graphics can be directly sent to the printer or saved to a PostScript file and read by standard graphics programs. A number of page size options are available and extra large plots spanning many pages may also be printed. A detailed program manual is available with notes on installation, usage and examples. Conclusions We describe a rapid and convenient application GenomeViz for simultaneous visualization and comparison of varied genomic data from several microbial genomes. Future updates for software and data will be available from the project home page. Availability and requirements • Project name: GenomeViz • Project home page: • Operating system(s): Linux, Solaris, Unix • Programming language: Tcl/Tk • Other requirements: ActiveTcl, PERL • License: Free for academic use • Any restrictions on use by non-academics: Contact corresponding author for a license. List of abbreviations used COG: Clusters of Orthologous Groups SIGI: Score-based Identification of Genomic Islands Authors' contributions RG conceived the program, wrote and tested it, prepared the manuals and the website. TC oversaw the entire development process. TH offered suggestions on program features. RG and TC prepared the manuscript. All authors read and approved of the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544189.xml
544837	Simultaneous two organ metastases of the giant basal cell carcinoma of the skin	Background Basal Cell Carcinoma (BCC) is the most common carcinoma in humans. It accounts for 20% of carcinomas in men and 10–15% of carcinomas in women. Despite its high incidence, metastatic events are exceedingly rare. The reported frequency of metastatic dissemination is estimated at 0.0028–0.5 percent. Once metastasis is detected, there is a high mortality rate of 50% within 8 months. Methods In this study, we present a case of simultaneous lung and parotid metastases of giant BCC primary located on the right medial canthus of a 62 year old female. Results Examination of the tumor located on the medial canthus obtained showed "adenoid BCC". Computed tomography (CT) was performed to evaluate parotid region for evaluation of parotid gland and chest. Parotid and lung metastasis were detected in CT. Routine labarotory tests and radiological investigations were done. There was no abnormal finding. We also investigated this patient with a bone scan (normal), abdominal and cranial CT scans (also normal). Conclusion Although metastasis of BCC is a very rare condition, this study reports a case of simultaneous parotid gland and lung metastasis originating from a giant BCC primary that was located on the right inner canthus of a 62 year old female.	Background Basal cell carcinoma (BCC) is the most common carcinoma in humans and accounts for 20% of carcinomas in men and 10–15% of carcinomas in women. Approximately 75–86% of primary BCCs are found on the head or neck. The most common location on the head is the nose, specifically the nasal tip and alae. It constitutes 90% of periorbital malignancies [ 1 , 2 ]. Sun exposure is the primary etiologic agent for the development of BCC. The tumors are more frequent in individuals with fair complexions. BCC arising on the medial canthus tends to be deep and invasive and may result in perineural extension and loss of optic nerve function. Pieh et al reported that the highest recurrence rates of BCC following attempted excision, (approximately 60%), was seen with lesions arising from the medial canthus since these lesions tend to be more invasive and difficult to manage [ 3 ]. Reclusive patients or patients who neglect the lesions for long periods of time are more likely to have giant, invasive tumors [ 4 ]. Giant BCC, defined as lesions more than 5 cm at its largest diameter, are rare forms of BCC [ 4 ]. Giant BCCs more commonly appear on the trunk and display a more aggressive behavior, resulting in local invasion and metastasis. The reported incidence of metastatic BCC ranges from 0.03 % to 0.55 [ 5 ]. We report a case of simultaneous lung and parotid gland metastases of giant BCC located on medial canthus. Case report A 62-year-old woman was referred to the Plastic and Reconstructive Surgery Department for treatment of a bleeding exophytic tumor located on the right inner canthus. She had had the lesion for approximately 11 years. Initially, the patient was treated with excision and primary closure ten years ago. At this time the tumor had a diameter of 5 cm. The tumor was diagnosed as adenoid BCC microscopically and surgical margins were tumor-positive. The patient was operated on two years later when the diameter of the recurrent tumor was 15 mm. Histological examination of this second specimen revealed an "adenoid BCC" with clear surgical margins. Although the tumor recurred again after the second excision, the patient neglected medical advice and did not undertake any treatment (Figure 1 ). More recently, however, the tumor began growing rapidly and became hemorrhagic. On examination the lesion was located on the right inner canthus and involved 1/3 of the eyelid. The size of tumor was approximately 55 mm × 45 mm. Visual functions of the patient were normal. However, a fixed mass developed in the patient's periauricular area six months ago (Figure 2 ), although there were no palpable cervical nodes. We therefore investigated this region with computed tomography (CT), which revealed a tumor involving the right orbital structures extending to the ethmoidal cells. The tumor also involved the right parotid gland and multiple cervical lymph nodes. Figure 1 Giant BCC located on the inner canthus Figure 2 Involvement of the parotid gland of the patient We also investigated this patient with a bone scan (normal), abdominal and cranial CT scans (also normal) and a thoracic CT. Multiple metastatic lesions were seen in the chest CT (Figure 3 ). Figure 3 CT scan of the chest of the patient. Multiple metastatic lesions were seen. Examinations of the cardiovascular, gastrointestinal, neurological, urogenital and hematological systems and other parts of the skin were performed by physical and routine laboratory and radiological techniques. There were no abnormal findings. Biopsy was performed from the tumor located on the inner canthus and revealed "adenoid basal cell carcinoma" (Figure 4 ). Also, Fine needle biopsies were performed on the parotid and pulmonary lesions which confirmed the presence of adenoid BCC in these regions. Figure 4 Histological view of the specimen showed "adenoid BCC". (Hematoxylene eosin staining, × 50 magnification) The patient did not accept the offer of surgical treatment for the tumor on the inner canthus. She was referred to the Oncology Department and treated with radiotherapy and chemotherapy. The patient received approximately 6000 cGy of external beam radiation over 3 weeks totally. Also, chemotherapy was initiated with cisplatin and 5-fluoruracil. She was followed up with physical examination and CT scans for six months and there were no metastases to other organs. She is still being followed. Discussion Spates et al. noted that metastatic BCC was first reported in 1894 [ 6 ]. As outlined by Lattes and Kessler, metastatic basal cell carcinoma is defined by the following criteria: 1) the primary tumor must occur in skin containing hair follicles and not the mucous membranes; 2) metastasis cannot be by simple extension, but occurring at a site distant from the primary tumor; 3) the primary tumor and the metastasis must have similar histologic appearances of basal cell carcinoma; and 4) squamous cell features must not be present in the lesions [ 7 , 8 ]. The case presented here satisfied these three criteria. More than 300 cases of metastatic BCC have been reported in the literature. Two-thirds of metastatic BCCs arise from primary tumors on the face, with the ear being the most common location. Higher rates of metastasis also occur from primary lesions on the scalp and genitalia [ 9 - 11 ]. Primary BCC metastasizes through hematogenous and lymphatic routes. As was the case with our patient, metastasis to the lymph nodes has been estimated to occur in 70% of cases. The most common organs involved in hematogenous spread are lungs, bone, and skin [ 12 , 13 ]. While the usual BCC that gives rise to metastases is a large, ulcerated, locally invasive BCC of the head and neck that recurs despite repeated surgical procedures or radiotherapy, these features are not absolute prerequisites for metastasis [ 14 ]. Immunosupression may be a factor in the pathogenesis of the metastasis of BCC, but there was no finding suggestive of immune system abnormality in the case presented here [ 14 ]. Some authors have suggested that immunosuppression or impaired cell-mediated immunity (including AIDS) may predispose to BCC and BCC metastasis [ 15 - 17 ]. BCC usually has multiple skin recurrences before metastases become evident as in the case presented here [ 18 ]. BCCs with any of the following: long duration, location in the mid face or ear, a diameter larger than 2 cm, with aggressive histological subtype, previous treatment, neglected, or a history of radiation exposure, should be considered "high risk [ 19 ]. Although giant BCC is commonly located on the trunk, Takemato et al reported a case of with giant basal cell carcinoma which invaded the orbital tissue and anterior skull base [ 4 ]. They concluded that giant basal cell carcinomas have aggressive character to destroy tissue and more metastatic potential. Many investigators have reported that radical treatment with a wide excision of the tumor at an early stage is essential to treat a potentially aggressive BCC. Takemato et al used a free rectus abdominis musculocutaneous flap in the treatment of giant BCC which invaded the orbital tissue [ 4 ]. Tumors greater than 3 cm in diameter have a 2 % incidence of metastasis and/or death. This increases to 25% in those lesions more than 5 cm in diameter and to 50% in lesions more than 10 cm in diameter [ 20 ]. The prognosis of metastatic BCC is extremely poor. Once metastasis is detected, there is a high mortality rate of 50% within 8 months [ 20 ]. This poor outcome has led to the use of systemic chemotherapy in a number of individual cases. Several chemotherapeutic agents that have been used in metastatic BCC, including fluorouracil and combination of vincristine, bleomycine and prednisone [ 18 ].Kaufman suggested that cisplatin, alone or in combination is probably the most active in metastatic BCC [ 21 ]. Recently, Jefford et al presented their experience in the treatment of metastatic BCC [ 22 ]. According to their study, systemic chemotherapy with cisplatin and paclitaxel provided palliative benefit to their patient with acceptable toxicity and conclude that the regimen is a reasonable choice for the rare patient presenting with metastatic BCC. About half of metastatic BCCs have metastasis to lymph nodes as the first site, but hematogenous route to lung and bone is also equally represented [ 23 ]. Robinson and Dahiya reported a case of BCC with pulmonary and lymph node metastasis causing death [ 14 ]. BCC located on the nose, eyebrow, ear, nose, and temple frequently metastasizes to the parotid and lymph nodes of neck [ 24 ]. To the best of our knowledge this report is the first to present simultaneous lung and parotid gland metastases of giant BCC. Competing interests The author(s) declare that they have no competing interests. Authors' contributions EC, conceived the study and coordinated the write-up and submission. AA participated in the writing of the manuscript. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544837.xml
538254	Atrial fibrillation and survival in colorectal cancer	Background Survival in colorectal cancer may correlate with the degree of systemic inflammatory response to the tumour. Atrial fibrillation may be regarded as an inflammatory complication. We aimed to determine if atrial fibrillation is a prognostic factor in colorectal cancer. Patients and methods A prospective colorectal cancer patient database was cross-referenced with the hospital clinical-coding database to identify patients who had underwent colorectal cancer surgery and were in atrial fibrillation pre- or postoperatively. Results A total of 175 patients underwent surgery for colorectal cancer over a two-year period. Of these, 13 patients had atrial fibrillation pre- or postoperatively. Atrial fibrillation correlated with worse two-year survival (p = 0.04; log-rank test). However, in a Cox regression analysis, atrial fibrillation was not significantly associated with survival. Conclusion The presence or development of atrial fibrillation in patients undergoing surgery for colorectal cancer is associated with worse overall survival, however it was not found to be an independent factor in multivariate analysis.	Background In general, colorectal cancer patients are three times more likely to be in atrial fibrillation preoperatively than matched controls undergoing non-colorectal cancer surgery [ 1 , 2 ]. It may also occur postoperatively. Recent data suggest that atrial fibrillation may be an inflammatory complication, resulting from the initiation of an inflammatory response to surgery or infection [ 3 - 6 ]. Colorectal cancer patients may have elevated C-reactive protein (CRP) levels [ 7 ] indicating a systemic inflammatory response. Elevated CRP levels may be associated with a worse prognosis in colorectal cancer patients [ 8 ]. Postoperative dysrhythmias may [ 9 ] or may not [ 10 ] be associated with worse survival following surgery for lung cancer. We hypothesised that atrial fibrillation (AF) is an adverse prognostic indicator in patients undergoing surgery for colorectal cancer. Patients and methods Patients who underwent a resection for colorectal cancer between 1 st January 2000 and 31 st December 2001 in a 600-bed district general hospital in the United Kingdom National Health Service were identified. The hospital serves a population of approximately 230,000. About 90 elective and emergency laparotomies are performed each year for colorectal cancer. Patients were identified from the prospectively maintained colorectal cancer database maintained by the colorectal surgical department. Patients with radiological, endoscopic or clinical examinations suspicious of colorectal cancer are referred to the weekly colorectal multi-disciplinary team (MDT) meeting. In the case of suspicious radiology, the referral to the MDT is made automatically by the radiology department. This avoids the possibility of the responsible clinical firm failing to refer a patient for consideration. Similarly, the pathology department automatically refers any patient in whom histology shows colorectal malignancy. In addition, patients who undergo surgery where a suspicion of colorectal cancer is raised are referred for consideration. The colorectal meeting is attended by the colorectal surgeons, radiologists, pathologists, palliative care physicians and nursing staff. Patients determined to have colorectal cancer by the MDT are entered into the database. The colorectal department periodically compares the database to clinical coding data for patients with colorectal cancer in order to ensure complete data capture. All patients are followed-up regularly by a team of colorectal nurse specialists in a dedicated clinic. Age, sex, mode of presentation (emergency or elective), Dukes stage, postoperative anastomotic leakage and adjuvant therapy were recorded for all patients. The colorectal cancer database was cross-referenced with the hospital clinical-coding database to identify those patients who were in atrial fibrillation at any time before or after their surgery. Patients with colorectal cancer who did not undergo surgery or who only underwent palliative stoma formation were excluded. All patients were followed up for at least two years postoperatively. Overall survival and recurrence-free survival were recorded. Recurrence-free survival was defined as the time interval between operation and first diagnosis of local or distant recurrence. Patients with no recurrence were censored at the time of death from any cause other than cancer or at the time they were last seen by the colorectal team. Characteristics between those with and without AF were compared using the Student t-test and Fisher Exact test for continuous and categorical data respectively. Potential prognostic factors were compared by the log-rank test. Significant prognostic factors identified from the univariate analysis were entered into a multivariate Cox regression model of survival to test for independence. The 5% level was considered significant in the multivariate analysis. Statistical analysis was performed using Statsdirect ® version 2 (Statsdirect Ltd., UK). Results One hundred and seventy-five patients (M:F = 111:64) who underwent bowel resection for colorectal cancer were identified from the database. Their median age was 74 years (interquartile range 66 to 80 years). Tumour site, Dukes stage and mode of operation (emergency or elective) are summarised in Table 1 . Anastomotic leaks occurred in three patients while another three patients received preoperative radiotherapy. Median follow-up was 2.38 years. There were 60 deaths (42 cancer-specific deaths) during the follow-up period. The remaining 18 patients died from conditions such as pneumonia, pulmonary embolus or myocardial infarction. Cause of death was recorded for all patients in the database. Twenty-eight patients (16%) developed recurrence during postoperative surveillance. The remaining 14 patients who died were noted to have incurable disease at the time of surgery. Table 1 Baseline characteristics of study group No. of patients (%) Age ≥72 years 98 (56%) < 72 years 77 (44%) Gender Male 111 (64%) Female 64 (36%) Site Colon 119 (68%) Rectum 56 (32%) Dukes' Stage A 26 (15%) B 75 (43%) C 57 (33%) D 17 (9%) Presentation Elective 147 (84%) Emergency 28 (16%) Cross-referencing with the clinical coding database identified thirteen patients with a history of atrial fibrillation. Five patients were in AF preoperatively. The remaining eight patients developed postoperative AF. A comparison of baseline characteristics between patients with and without AF is shown in table 2 . The AF was paroxysmal in three patients, persistent but eventually resolved in five and permanent in the remaining five patients. There were seven deaths among the patients with atrial fibrillation: two from recurrent colorectal cancer, four from other causes (pneumonia in two patients, respiratory failure secondary to pulmonary fibrosis in one and left ventricular failure in one) with recurrent cancer present and one non-cancer related death (peri-operative myocardial infarction). Table 2 Comparison of characteristics between patients with AF and those without. Characteristic Sinus rhythm Atrial fibrillation p Male gender 102 9 0.77 Mean age 72 73 0.64 Rectal cancer 52 4 0.99 Elective surgery 139 8 0.05 Pre-operative radiotherapy 2 1 0.22 Anastomotic leak 3 0 0.79 Dukes Stage A 24 2 0.73* B 69 6 C 53 4 D 16 1 *Fisher-Freeman-Halton Exact Test Survival analysis There was no correlation between overall survival and the following variables in the univariate survival analysis: gender, postoperative anastomotic leak, site of tumour (rectal versus colonic) or preoperative radiotherapy. Mode of surgery (emergency or elective), age (<72 years or ≥ 72 years) and Dukes' stage had a significant effect on survival. When patients with atrial fibrillation were compared to those without (Figure 1 ), atrial fibrillation correlated with worse overall survival. Dukes' stage, mode of surgery, age and atrial fibrillation were entered into a Cox regression model overall survival (Table 2 ). Mode of surgery, age and Dukes' stage retained significance but atrial fibrillation did not (Model chi-square 49.6; 7 degrees of freedom; p < 0.0001). There was no significant correlation between atrial fibrillation and recurrence-free survival (p = 0.74). Figure 1 Kaplan-Meier survival curves for patients with a history of atrial fibrillation (1) versus those without (0). P = 0.04 (Log-rank test). Discussion Tumours stimulate an inflammatory response when they invade or metastasise [ 11 ]. This inflammation may cause a rise in serum levels of inflammatory markers such as CRP. Serum CRP levels correlate with survival in colorectal cancer patients [ 8 ]. Atrial fibrillation may be precipitated or maintained by an inflammatory mechanism [ 3 ]. Thus, we hypothesised that the presence of AF, an inflammatory complication, would be associated with poorer survival in colorectal cancer as it is a manifestation of more advanced disease. Our data demonstrate a significant correlation between atrial fibrillation and survival following surgery for colorectal cancer in univariate analysis. Previous investigators have found that atrial arrhythmias were significantly more common in those patients with a history of cancer. In addition, CRP levels were higher in those patients with a history of cancer. However, there was no independent link between cancer and arrhythmias. This is consistent with inflammation being the causal link between cancer and arrhythmias [ 12 ]. In our cohort, atrial fibrillation was not an independent predictor of survival. Dukes' stage was the strongest predictor of survival. Previous data from colorectal cancer patients suggest the degree of inflammatory response may reflect the degree of disease progression [ 13 ]. The incidence of liver metastases, peritoneal tumour deposits, lymph node metastases and intravascular invasion are higher in patients with elevated CRP levels [ 14 ]. Thus, atrial fibrillation may be a manifestation of systemic inflammation due to more advanced disease and would not be independent of Dukes' stage. Cox regression analysis of our study cohort (Table 3 ) appears to show poor survival in Dukes B patients (Hazard ratio 9.13) compared to Dukes C patients (Hazard ratio 0.15). There was a trend for Dukes B patients to be older (mean age 73.79 years) than Dukes C patients (70.99 years; p = 0.1, student t-test, β = 0.67) and this may account for the reduced survival. There were no differences in the Dukes' stage distribution between patients with and without AF (Table 2 ). However, the small numbers in our series (only thirteen patients with AF) render it underpowered to detect a correlation between the presence of AF and colorectal cancer stage. Table 3 Overall univariate and multivariate survival analysis Variable Univariate p Multivariate p Hazard Ratio 95% Confidence interval of hazard ratio Age (< 72 years versus ≥ 72 years) 0.03 0.003 2.38 1.339 to 4.224 Pre-operative radiotherapy 0.26 - - - Anastomotic leak 0.23 - - - Gender (Male versus female) 0.33 - - - Site of tumour (Rectal versus colonic) 0.87 - - - Emergency surgery <0.0001 0.004 2.35 1.312 to 4.205 Atrial fibrillation 0.04 0.06 2.21 0.979 to 4.985 Dukes' stage (A, B, C, D) <0.0001 <0.0001 Dukes A 0.09 0.032 to 0.267 Dukes B 9.13 4.343 to 19.206 Dukes C 0.15 0.073 to 0.312 Dukes D 1 Reference There are several other limitations to our study. The quality of the clinical coding has not been assessed. Thus, the sensitivity and specificity of our clinical coding department in coding AF is unknown. Only 13 patients (7.4%) developed AF. Previous work suggests that 13% of elective colorectal patients develop postoperative AF [ 15 ]. Thus, it is possible that some patients with unrecorded AF have been incorporated into the control arm of our series. The study cohort has been followed for a relatively short time, only two years. However, most trials of follow-up following surgery for colorectal cancer report a mean time to relapse of approximately 24 months [ 16 ]. We were unable to demonstrate a significant relationship between atrial fibrillation and recurrence-free survival. This may be due to an insufficient sample size. However, atrial fibrillation does correlate with overall survival. The presence of atrial fibrillation may be a clinical marker of poor overall survival in patients undergoing surgery for colorectal cancer. Competing interest The author(s) declare that they have no competing interests. Authors' contribution SRW conceived the study, performed the statistical analysis and drafted the paper. KMG and NJW performed the literature search and collected the data. TAJ and NJK co-wrote the paper with SRW . All authors approved the manuscript. Funding support None declared	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC538254.xml
528735	Reuse of imputed data in microarray analysis increases imputation efficiency	Background The imputation of missing values is necessary for the efficient use of DNA microarray data, because many clustering algorithms and some statistical analysis require a complete data set. A few imputation methods for DNA microarray data have been introduced, but the efficiency of the methods was low and the validity of imputed values in these methods had not been fully checked. Results We developed a new cluster-based imputation method called sequential K-nearest neighbor (SKNN) method. This imputes the missing values sequentially from the gene having least missing values, and uses the imputed values for the later imputation. Although it uses the imputed values, the efficiency of this new method is greatly improved in its accuracy and computational complexity over the conventional KNN-based method and other methods based on maximum likelihood estimation. The performance of SKNN was in particular higher than other imputation methods for the data with high missing rates and large number of experiments. Application of Expectation Maximization (EM) to the SKNN method improved the accuracy, but increased computational time proportional to the number of iterations. The Multiple Imputation (MI) method, which is well known but not applied previously to microarray data, showed a similarly high accuracy as the SKNN method, with slightly higher dependency on the types of data sets. Conclusions Sequential reuse of imputed data in KNN-based imputation greatly increases the efficiency of imputation. The SKNN method should be practically useful to save the data of some microarray experiments which have high amounts of missing entries. The SKNN method generates reliable imputed values which can be used for further cluster-based analysis of microarray data.	Background DNA microarray is a popular high-throughput technology for the monitoring of thousands of gene expression levels simultaneously under different conditions [ 1 ]. The typical purposes of microarray studies are to identify similarly expressed genes under various cell conditions and associate the genes with cellular functions[ 2 , 3 ]. The analysis performed to meet the purposes of microarray studies mentioned above usually involves clustering genes according to their pattern of expression levels in various experimental conditions. In fact, cluster analysis means grouping samples (or genes) by similarity in expression patterns. To measure the similarity in cluster analysis, correlation distance and Euclidean distance are widely used[ 4 ]. Principal component analysis (PCA) is also a powerful technique when used with the clustering method to specify the number of clusters[ 5 ]. However, these widely-used methods in microarray data analysis can be both seriously biased and misled by missing values in the dataset[ 6 - 8 ]. Missing values of microarray data commonly occur during data preparation mainly due to imperfections in the various steps in DNA microarray experiments. One of the yeast microarray data sets shows that the number of genes having at least one missing value was 2419 of 6198 rows (genes) (in other words, 39 %)[ 9 ] and 566 of 918 rows (72.5%) [ 10 ]; and 1741 of 2364 rows (73.6%) [ 11 ] had missing values in other reports. As mentioned previously, some statistical analyses require complete data sets and one should discard the entire data in a row, usually all the values for one gene, that have a single missing value. The rows with missing values can be utilized for further analyses after the imputation of the missing values in many cases. Imputation has been used in many fields to fill the missing values in incomplete data using observed values. There are many different algorithms for imputation: hot deck imputation and mean imputation [ 7 ], regression imputation [ 12 , 13 ], cluster-based imputation [ 14 ], and tree-based imputation [ 15 , 16 ], maximum likelihood estimation (MLE)[ 17 ], and multiple imputations (MI)[ 17 , 18 ]. Proper selection of an algorithm for a given data set is important to achieve maximum accuracy of imputation. Recently, several methods have been applied to the imputation of microarray data, including row average [ 7 ], singular value decomposition (SVD) [ 19 ] and KNN imputation [ 20 ] methods. In general, it seems the recently developed KNN-based method is most efficient. KNN imputation method is an improved hot deck imputation method [ 21 ] that uses the mean values of most similar genes for estimating missing values. The KNN imputation method can be considered a cluster-based method since missing values are imputed using selected similar genes. In the previously developed method, the efficiency of imputation was limited both in accuracy and computational complexity in that it did not efficiently use the information of the gene having missing values. The existence of missing values in a gene limits the use of other observed values of that gene in the conventional imputation method. In our work, this problem could be improved by using the imputed values sequentially for the later nearest neighbor calculation and imputation. We suggest a sequential KNN (SKNN) imputation method that boasts improved accuracy in estimation of missing values in a wide range of missing rates with high computational speed. We also suggest an EM-style sequential KNN (EM-SKNN) method that uses a sequential KNN method repeatedly to improve accuracy. We evaluated the efficiency of the SKNN imputation method through comparison with the known KNN-based method and other well known imputation methods such as maximum likelihood estimation (MLE) and multiple imputations (MI). Results We evaluated the efficiency of our new SKNN method and the EM-SKNN method with three other imputation methods: KNN-based imputation, the MLE method, the MI method, by applying them to three different types of microarray data sets with different missing rates. The appropriate number (k) of nearest neighbors was dependent on the data types and missing rates. The RMS errors were minimal when k was 10 for time-series data and mixed type data regardless of the missing rate, and the RMS errors of the non-time series data showed similarly low values when k values were between 10 and 20. For comparison of different imputation methods, we used 10 for k which showed a minimal RMS error in every data type with different missing rates. The performance of the KNN-based imputation method depends on the similarity of k -nearest neighbors to be used for imputation. The overall similarity of the entire data set can affect, on average, the similarity of all possible k -nearest neighbors. The time-series data set, which has the narrowest distribution of Euclidean distances among genes, shows the least RMS error after imputation as we can see in Figure 1 and 2 . Figure 1 also shows that the performances of the SKNN and EM-SKNN methods are better than that of the conventional KNN method over a whole range of tested missing rates. The range of RMS error by the new SKNN method, for example, was 0.194 to 0.269 in comparison with 0.194 to 0.324 of the KNN method in time-series data. The accuracies of our new methods are especially superior when the missing rate is over 30%. The RMS errors for time-series and non-time series data nearly approached their maximum values at missing rates of 50% and 60%, respectively, in the KNN method. The RMS error of the mixed data set is stable over a wide range of missing rates, but becomes unstable and increases dramatically after a 40% missing rate. The slight difference of KNN algorithm could lead to a large improvement in the accuracy of imputation at a high missing rate because the SKNN method is able to select more similar k -neighbors than the conventional KNN method as the missing rate grows. In the conventional KNN method, the selection pool and the dimension (or number of existing values for a gene) of the distance measurement of neighbor genes are reduced according to the increase of missing rate. In this situation, the method inevitably selects less related (or less similar) neighbors for imputation. In addition, the size of data set can limit the maximum missing rate for stable imputation. In our data sets, the size of a mixed data set is about 20% larger than other data sets, which may affect the mixed data set in terms of having stable RMS error in a relatively larger (10%) range of missing rate. We tested the performance of other well known non-KNN-based methods such as maximum likelihood estimation (MLE) and multiple imputations (MI) methods. These methods are well known imputation methods but there has been no report on their application in microarray data analysis. The efficiency of the MLE method was much worse than the SKNN method for all tested data sets. The RMS errors in the MLE method were 0.11 to 0.33 in time-series data, 0.30 to 0.38 in mixed data, and 0.58 to 0.69 in non-time series data. The efficiency of the MI method was generally similar to SKNN but the former is more dependent on data types. The efficiency of the MI method was better at a lower missing rate, but slightly worse at a higher missing rate for the time-series data set. The MI method was worse than SKNN in terms of overall range of missing rate of non-time series data. However, the best imputation method for mixed data set proved to be the MI method. We can conclude that the MI method is as efficient as the SKNN method for the imputation of microarray data, even though the efficiency of the MI method experienced more fluctuations than the SKNN method depending on the data type. The result is similar in a comparison of overall RMS error after imputation of a data set having unequally distributed missing entries over the columns. We show a comparison of one of the data sets (time series data set) in Table 1 . As expected, the efficiency of the SKNN method is higher, especially for the data sets having a higher missing rate. For more careful estimation of imputation efficiency, we examined the structure of data after imputation. We calculated the Pearson correlation coefficients for each column (experiment) between original data and imputed data. The larger the correlation coefficient is, the better the relationship between original complete data and imputed data is preserved in a column. Figure 3 shows that the SKNN method preserves the structure of the original data set better than the conventional KNN method and MI method for all columns of the time-series data set. The situation was the same for the other data sets (data not shown). Interestingly, the MI method was much worse than the SKNN method, differing from RMS error analysis. This column-wise comparison gives us more specific information on the efficiency of imputation method. In Figure 3-b , we can see that the performance of SKNN is relatively better for the column with highly missing entries (column 17 and 18) than for other columns. Through measuring the means and standard deviations for each column of data sets, we discovered that the dispersion of values in a column does not affect the accuracy of KNN-based imputation. The SKNN algorithm improves execution time for imputation. The computational complexities are approximately O ( m 2 n 2 ) in the conventional KNN method and O ( mn log m ) in the SKNN method for a matrix with m rows (genes) and n columns (experiments). This is because the sequential KNN algorithm imputes all missing values in a gene simultaneously with given nearest neighbors, while the conventional KNN method must calculate neighbors for each missing entry. The application of Expectation Maximization (EM) to the sequential KNN method marginally improved the accuracy in compensation for the increase of computational time proportional to the number of iterations. For MI methods, the execution time increased as M times of single imputation method when MI used M multiple imputation. Using the SKNN imputation method, it took 28.3 seconds on a Pentium IV 2.4 GHz computer to estimate missing values for a data set with 4489 genes, 18 experiments and a 40% missing rate. The processing time using the EM-SKNN method was proportional to the number of iterations. Discussion The SKNN method offers better performance than the previously developed KNN method for both time series and non-time series microarray data sets and for data sets having different missing patterns. As the missing rate increased, sequential reuse of imputed data did not propagate errors of imputation as in the conventional KNN method. It showed the best improvement of accuracy for the data set with a high missing rate. Notably, the SKNN method is also robust on the imputation of a data set with unequally distributed missing entries. A real microarray data set usually has non-random distribution of missing data. Furthermore, some systematic errors during the experiment can generate an abnormal increase in distribution of missing entries for the corresponding column of microarray data set. In this type of data, the SKNN method, which is especially efficient on the data set having heavy missing entries, can exert relatively more accurate imputation than other imputation methods as shown in our model data set. The MI method has not been well introduced in the field of microarray analysis, although it is a well known imputation method in other fields [ 18 ]. In comparison with the SKNN method, we discovered the potential of the MI method for microarray analysis. The MI method did not preserve original data structure as well as the SKNN method, but the overall RMS error was close to the SKNN method. The MI method is executed under the assumption of multiple normality of all dimensions of data. This assumption may not be satisfied in real-world data. Nevertheless, the performance of the MI method was much higher than the simple KNN method, which suggests that the MI method is practically applicable for the imputation of microarray data. The computational complexity is reduced in the SKNN method for the dimension of both the number of genes and the experiments compared with the simple KNN method. Particularly, computation time can be saved substantially for microarray data with a large number of experiments. The SKNN method works efficiently in a wider range of missing rate with high speed. We want to emphasize that our results showed that the method using estimated values achieved even better accuracy than the method using only observed values in the case of the KNN-based imputation method. We suppose that this result could be applicable to other cluster-based analysis. It would be hardly acceptable for the experimentalist to use imputed data for further analysis. However, analysis could become more errorneous without imputation due to loss of information caused by missing values. The use of imputed data should definitely depend on the type of later process. If the next process is a cluster-based analysis, the genes with imputed values could be efficiently used, as we had good results for KNN-based imputation with the reuse of imputed values. For future works, it may be possible to integrate the imputation and gene clustering of microarray data for classification of genes with proper evaluation steps. This may offer more and better information sources of microarray data for the final decision of gene classification. All the procedures used in this paper are done by R-code and C++ and the programs are available upon request. Conclusions The SKNN method is an especially efficient imputation method on data having high missing entries. It can be practically useful in saving data of some accidental microarray experiments having high missing entries. Our results also suggest that the imputed values generated by the SKNN method can be used reliably for further cluster-based analysis of microarray data. Methods We developed and implemented SKNN and EM-SKNN methods for the imputation of microarray data, and we compared their accuracies with the previously developed KNN imputation method. Data sets used in this work were selected from publically available microarray data. The data sets were from a study of gene expression in yeast Saccharomyces cerevisiae cell-cycle regulation [ 22 ], calcineurin/crz1p signaling pathway [ 23 ], and certain environmental changes[ 9 ]. These data sets can be classified into time series data set [ 22 ], mixed (time-series and non-time series) data set[ 23 ] and non-time series data set[ 9 ]. The efficiencies of imputation methods were assessed by Root Mean Squared (RMS) error and correlation coefficients using three different data types as described later. KNN Imputation method To assess the relative efficiency of the imputation methods, we implemented known KNN imputation method developed by Troyanskaya et al. (2001)[ 20 ]. The source code of the KNN imputation was available from the Helix group at Stanford University [ 24 ]. The matrix form of microarray data is composed of rows and columns that represent genes and experimental conditions respectively. Before any further process, the rows of original data sets containing missing values are removed to make complete matrices and test data sets for imputation methods were generated by random deletion of values in the complete matrices. The sizes of test data sets were 4489×18 for time series data set, 4380×24 for mixed data set, and 3779×22 for non-time series data set. The missing rates generated randomly in the test data sets were between 1% and 70% (1, 3, 5, 10, 20, 30, 40, 50, 60, 70). The occurrence of missing values can depend on the specific experiment in real miroarray data. Considering this case, we also generated test data set having missing values non-randomly along the columns which represent each experiment. The overall missing rate of data set was fixed to one of the value ranging from 10% to 60%. In a data set, the missing rates for two experiments (columns) were set to 80% and remaining columns have randomly generated missing entries. In KNN method, k -nearest neighbor genes are taken from the whole matrix of the test data set except any genes that has missing value at the same position with the gene to be imputed. Euclidean distance is used as the metric to estimate the similarity of neighboring genes. To compare the similarity by this metric, each gene should have the same dimension and missing positions of values inside. Missing value is imputed with weighted average of the corresponding column of the k -nearest genes. The weight of i th gene is calculated as equation (1), where k is the number of selected genes and D i is the distance between i th gene and a gene to be imputed. For the performance comparison of the imputation methods, we selected appropriate k -values for each data set and each method with different missing rates. Different k -values ranging from 1 to 500 were tested and we selected the k -values with the least error between imputed values and real values. Sequential KNN Imputation method SKNN method that we suggest in this report is distinguished in two main points from previously implemented KNN method described above. In SKNN method, genes are ordered by its missing rate and the imputation was executed sequentially from the gene that had least missing rate. In addition, these sequentially imputed genes are used for the later imputation of the other genes. The test data set was separated into incomplete and complete set that has or has not missing values respectively. The genes in incomplete set were imputed by the order of missing rate. Missing value was filled by the weighted mean value of corresponding column of the nearest neighbor genes in complete set. Once all missing values in a gene are imputed, the imputed gene was moved into the complete set and used for the imputation of the rest of genes in incomplete set. In this process, all missing values in one gene can be imputed simultaneously from the selected neighbor genes in complete set. This reduced execution time from previously developed KNN method that should select nearest neighbors for each imputation. EM-style Sequential KNN Imputation method EM-style imputation algorithm was originally suggested by Rich Caruana (2001)[ 25 ]. EM-style imputation is executed by two steps. It estimates missing values from observed values and improves accuracy of fill-in values through recursive process. We integrated EM-style imputation algorithm and SKNN method to increase the accuracy of imputation. All of missing values were estimated by SKNN imputation method at the first step. The estimated missing values were re-estimated by SKNN method again. In this second step, we could use newly imputed values to select k -nearest neighbors for the estimation of missing values. EM-style method executes this process repeatedly until the differences between newly updated values and previous values converge. Because all the imputed values were converged within less than 10 iterations, we did 10 iterations for the comparison of accuracy with the other methods. Maximum Likelihood Estimation (MLE) and Multiple Imputation (MI) In MLE method, data set with missing values are centered, scaled, and sorted by the patterns of missing through the preliminary manipulations. Missing entries in a data matrix are estimated under the multivariate normal model with user-supplied parameters and observed data (non-missing entries in the data set). The parameters are estimated using imputed and observed data. A vector of parameters representing the MLE, means and variance-covariance matrix are returned by using EM algorithm. Missing values are estimated through this iterative process until estimated parameters converge. We executed MLE method to estimate missing values by using 'norm' library of R [ 26 ] based on the description of Rubin[ 17 ] for this work. The whole MI procedure is made of three steps. They are imputation, analysis, and pooling processes. We applied only the first step, imputation process, of the three steps because our interest is to fill in missing values with estimated values. We used Predictive Mean Matching (PMM) as a method for missing values estimation. It uses a linear regression on observed variables to impute missing values. The estimated coefficients provide the mean vector and the variance matrix to generate multiple sets of coefficients that leads M imputed sets. M plausible values for missing observations were created by above MI algorithm and then the mean of M imputed values was filled in the missing value. We implemented Multiple imputation method using 'mice' library of R [ 26 ] based on the description by Rubin [ 17 ]. Evaluation of imputation methods The accuracy of imputation method was evaluated by calculating error between actual values and imputed values after missing values were estimated. The metric used to assess the accuracy of estimation was RMS error. RMS error was calculated as follows, where R i is the real value, I i is the imputed value, and N is the number of missing values. Besides RMS error, Pearson correlation coefficients were used to evaluate the sequential KNN method. Correlation coefficients were calculated between imputed data and complete data for each column. From this evaluation, we could find how the data structure of each column was preserved after imputation with different imputation methods. List of abbreviations KNN: K Nearest Neighbor; SKNN: Sequential KNN; EM: Expectation Maximization; RMS: Root Mean Squared; PCA: Principal Component Analysis; MLE: Maximum Likelihood Estimation; MI: Multiple Imputation Authors' contributions KK participated in the design of algorithms, performed statistical analysis and drafted the manuscript. BK participated in the design of algorithms and carried out C++ programming. GY conceived of the study, participated in its design and coordination, and finalized manuscript. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC528735.xml
517710	Exonuclease activity and P nucleotide addition in the generation of the expressed immunoglobulin repertoire	Background Immunoglobulin rearrangement involves random and imprecise processes that act to both create and constrain diversity. Two such processes are the loss of nucleotides through the action of unknown exonuclease(s) and the addition of P nucleotides. The study of such processes has been compromised by difficulties in reliably aligning immunoglobulin genes and in the partitioning of nucleotides between segment ends, and between N and P nucleotides. Results A dataset of 294 human IgM sequences was created and partitioned with the aid of a probabilistic model. Non-random removal of nucleotides is seen between the three IGH gene types with the IGHV gene averaging removals of 1.2 nucleotides compared to 4.7 for the other gene ends (p < 0.001). Individual IGHV, IGHD and IGHJ gene subgroups also display statistical differences in the level of nucleotide loss. For example, within the IGHJ group, IGHJ3 has average removals of 1.3 nucleotides compared to 6.4 nucleotides for IGHJ6 genes (p < 0.002). Analysis of putative P nucleotides within the IgM and pooled datasets revealed only a single putative P nucleotide motif (GTT at the 3' D-REGION end) to occur at a frequency significantly higher then would be expected from random N nucleotide addition. Conclusions The loss of nucleotides due to the action of exonucleases is not random, but is influenced by the nucleotide composition of the genes. P nucleotides do not make a significant contribution to diversity of immunoglobulin sequences. Although palindromic sequences are present in 10% of immunologlobulin rearrangements, most of the 'palindromic' nucleotides are likely to have been inserted into the junction during the process of N nucleotide addition. P nucleotides can only be stated with confidence to contribute to diversity of less than 1% of sequences. Any attempt to identify P nucleotides in immunoglobulins is therefore likely to introduce errors into the partitioning of such sequences.	Background The variable domain of the immunoglobulin heavy chain (IGH) is encoded by the IGHV (variable), the IGHD (diversity) and the IGHJ (joining) genes. In developing B cells these genes are brought together via a process of recombination involving the selection of one of each gene type from sets of genes present within the genome [ 1 ]. The bringing together of the selected IGHV, IGHD and IGHJ genes generates combinatorial diversity [ 1 ]. The first genes to join are the IGHD and IGHJ genes, followed by the bringing together of the IGHV gene with D-J. Further junctional diversity is generated at the points between the joining genes [ 2 , 3 ]. Junctional diversity results from the loss of nucleotides through the action of unknown exonuclease(s) and from the addition of N [ 3 ] and P nucleotides [ 2 ]. The final IGH V-D-J rearrangement in mature B cells is finally subject to the process of somatic hypermutation in secondary lymphoid organs which involves the targeted introduction and accumulation of point mutations [ 4 ]. The addition of N nucleotides is performed by the enzyme terminal dideoxynucleotidyl transferase (TdT), and in the IGH locus this addition can occur at both the D to J and the V to D-J joins [ 5 ]. The regions of N addition are denoted as N regions, and nucleotides that fall between the V and D genes are denoted as N1 regions, while those that lie between the D and J genes are denoted as N2 regions. P nucleotides are derived from the asymmetric opening of hairpin loops that form at gene ends as part of the rearrangement process [ 6 ]. The opening of the hairpin loops produces short, self-complementary single stranded extensions that can be incorporated into junctions, or may alternatively be removed via exonuclease activity [ 6 ]. It is the self-complementarity of P nucleotides that leads to their palindromic appearance and thus to their name. Hairpin opening is said to produce inserts of 0–4 nucleotides [ 2 ]. P nucleotides have been associated with the IGHV and IGHJ genes, as well as with each end of the IGHD gene [ 7 ] and estimates of the frequency of P nucleotide addition suggest a presence in about ten percent of sequences [ 7 - 10 ]. The mechanism of immunoglobulin gene rearrangement was first proposed by Tonegawa in the late 1970's [ 1 ]. Since that time, much has been learnt about the processes involved. Some areas, however, remain relatively uninvestigated, including the nature of exonuclease removal and the contribution of P nucleotide addition to junctional diversity. The lack of research in these fields may reflect the inherent difficulties in studying the relevant gene sequences, because IGH V-D-J junctions are the result of random and imprecise processes. It can therefore be difficult to distinguish between gene ends and N or P additions. The very few reports of exonuclease removal in the literature mainly describe analysis of murine sequences [ 11 - 14 ]. These investigations revealed nucleotide loss to be significantly different for murine IGHJ and IGHD genes. Differences were seen in the average exonuclease removal from IGHJ and IGHD gene subgroups, with individual gene subgroups possessing significantly different average levels of nucleotide removal. Influences upon gene processing that have been proposed to explain these observations include the presence of TG motifs [ 15 ], the relative location of stretches of 3 or more W (A or T) nucleotides and their positional relationship with respect to 2 or more S (G or C) nucleotides [ 12 ], and the presence of TAT motifs [ 13 ]. Recent advances in data standardisation in immunogenetics has allowed for improved statistical analysis. The standardisation emanates from IMGT-ONTOLOGY [ 16 , 17 ] upon which one of the most widely used immunogenetics tools, IMGT/V-QUEST, is based [ 18 ]. IMGT, the IMmunoGeneTics Information System R , also offers standardised nomenclature [ 19 ] and standardised numbering of positions within immunoglobulin sequences [ 20 ]. Despite the importance of IMGT within the field, the tools offered still suffer from shortcomings especially in the analysis of IGH V-D-J junctions. Alternative means of analysing junctions are therefore still sought by researchers [ 21 , 22 ]. The development of a statistically based algorithm for the partitioning of immunoglobulin sequences [ 21 ] as an alternative to IMGT/Junction Analysis [ 18 ], combined with the large amount of sequence data available through public nucleotide databases, has allowed us to investigate the nature of nucleotide removal from human immunoglobulin heavy chain genes in the expressed repertoire. Improved means of identification of gene ends facilitates the development of datasets with more certain partitioning. This study reports the extent of P nucleotide addition and the nature of exonuclease removal in the expressed human repertoire. Analysis of nucleotide loss and addition within the dataset reveals that different gene subgroups undergo distinct processing by exonuclease(s) and shows that there is no significant contribution by P nucleotides to the diversity of the expressed repertoire. Results Dataset creation The collection of human IgM sequences from public databases resulted in a dataset of approximately 1500 sequences. The exclusion of fetal, moderately and highly mutated (>5 mutations) and disease associated sequences reduced the dataset to 306 sequences. Further exclusions were made of those sequences that showed signs of IGHV gene replacement or the utilization of multiple D genes. Five sequences utilized two D gene segments, as identified using strict criteria as previously described [ 21 ] (EMBL:U97246, L12190, L29154, AJ519292, AJ245025). Evidence of IGHV gene replacement, in the form of V gene 'footprints', was seen in 7 sequences (EMBL:L29154, AJ245008, AJ245280, AJ519296, AY003831, X54445 Kabat:AL311). The footprints were unique sequences of 6 or more nucleotides derived from V gene ends containing a cryptic recombination signal sequence (cRSS) which is thought to be essential for replacement events [ 23 - 26 ]. The final dataset contained 294 sequences. Within the final dataset of 294 sequences there were 245 sequences for which IGHV, IGHD, IGHJ, N1 and N2 regions could be defined. A further 49 sequences, lacking determinable D, N1 and N2 region but possessing identifiable IGHV and IGHJ genes were also included in the dataset. For these 49 sequences, it was not possible to confidently determine the utilized IGHD gene within the junctional nucleotides, however, the ends of the V and J regions could be determined accurately. All IGHV, IGHD and IGHJ gene subgroups were represented within the dataset. Details of the dataset can be seen in the Appendix [see Additional File 1 ]. Exonuclease removals from genes and gene subgroups Exonuclease removal was evident in each of the 245 IGH V-D-J rearrangements examined, with 25% of IGH V-D-J rearrangements displaying removal from all four gene ends. A further 48% of IGH V-D-J sequences had removals from three of the four gene ends. The average number of nucleotides lost from each of the gene ends is presented in Figure 1 . Figure 1 Average exonuclease removal from IGH genes. The average nucleotide removal from of the gene ends was examined for 294 IGHV and IGHJ genes and 245 IGHD genes. For the IGHD genes, removals were considered from each end of the gene; 5' (V-D side) and 3' (D-J side). Bars represent standard error. Examination of the 294 IGHV and IGHJ segments revealed 41% of IGHV ends lacked removals, compared to just 18% of IGHJ ends. Sixteen percent of the 245 IGHD genes lacked removals from the 5' end of the D gene, and 17% had no removals from the 3' end of the D region. Exonuclease removals ranged from 0 to 13 nucleotides at the 3' V-REGION and 0 to 14 nucleotides at the 5' D-REGION end. At the D-J junction, 3' D-REGION and 5' J-REGION removals both ranged from 0 to 20 nucleotides. A significant difference in the extent of exonuclease removals was observed between the IGHV, IGHD and IGHJ gene ends (p < 0.0001, Kruskal-Wallis Test). Average removals from IGHV region ends were significantly lower than removals from IGHD and IGHJ region ends (p < 0.001, Dunn's Multiple Comparison Test). On average, only 1.2 nucleotides were lost from IGHV region ends while average removals of 4.7 nucleotides were evident from each of the IGHD region ends as well as from the IGHJ region ends. The average number of nucleotides removed from each gene subgroup within the three genes was calculated to identify differences in processing of sequences at the gene subgroup level (Figure 2 ). It was necessary to exclude IGHV7 from the analysis as only a single sequence from this subgroup was in the dataset. Average removals differed significantly among the six IGHV subgroups analyzed (p = 0.03, Kruskal-Wallis Test) (Figure 2A ). Comparison testing was, however, unable to identify the source of the difference. Figure 2 Average exonuclease removal from gene subgroups. The average exonuclease removals from gene ends was investigated for each IGHV, IGHD and IGHJ subgroup. Significant differences were seen among the 294 IGHV genes (A). No significant difference between the 245 IGHD genes were seen at the 5' end (B). The 3' IGHD end does show significant differences for the IGHD subgroups (C) as do the six IGHJ subgroups (D). Bars represent standard error. Removals from D gene subgroups were examined at each of the region ends. The removals from the 5' D end did not reveal significant differences, but significant differences were seen between the subgroups at the 3' end (p < 0.0001, one-way ANOVA) (Figure 2B , Figure 2C ). More extensive removals, of 6.0 and 7.5 nucleotides respectively, were observed from IGHD2 and IGHD3 subgroup members (p < 0.0001, Tukey's Multiple Comparison Test). The remaining 5 IGHD subgroups experienced average deletions of 2.1 nucleotides at their 3' ends. Comparison of average nucleotide loss for each of the six IGHJ subgroups revealed significant differences between the average removals (p < 0.0001, one-way ANOVA). The low level of removals from IGHJ3 was notable. On average just 1.3 nucleotides were removed from IGHJ3 sequences, while an average of 6.4 nucleotides were removed from IGHJ6 sequences (p < 0.002, Tukey's Multiple Comparison Test) (Figure 2D ). Influence of W and S motifs A more detailed examination of exonuclease removals from the IGHJ genes was undertaken, to investigate the influence of W and S motifs. The presence of these motifs in the first 15 5' nucleotides of the IGHJ ends was considered. IGHJ ends containing 5' S motifs showed significantly lower average removals than those lacking a 5' S motif (p < 0.0001, Kruskal-Wallis Test). The IGHJ genes whose sequences did not possess an S motif within the first 15 5' nucleotides had, on average, three more nucleotides removed (Figure 3 ). Figure 3 Influence of W and S motifs on nucleotide loss. IGHJ genes were grouped by the presence of W and S motifs within the first 15 nucleotides of the IGHJ subgroup sequence. Average exonuclease loss was examined for the three sets; 5' W only, 5' S only and S then W. Contribution of P nucleotides to diversity Putative P nucleotides were identified among those gene ends that remained untrimmed by exonuclease activity during the process of IGH V-D-J rearrangement. Examples of nucleotides that satisfied the P nucleotide criteria were observed at 3' V, 5' D, 3' D and 5' J gene ends. The identified nucleotides ranged in length from 1 to 4 nucleotides. The observed P nucleotides fell into twenty-three sets based upon unique sequences, and the gene end at which they were observed (Table 1 ). Each of the 23 sets was analyzed to determine the likelihood that apparent P nucleotides were actually the result of N additions. The p-values for each of the P nucleotide sequences are shown in Table 1 . Correction of the significance level for the comparison of the 23 sets using the Bonferroni adjustment resulted in a required alpha value of 0.003. Table 1 Putative P Nucleotides in a dataset of 294 human IgM sequences Gene Segment Putative P Sequence Observed Total Junctions p-value 1 IGHV C 10 111 0.99 T 25 111 0.023 CC 2 105 0.99 TC 10 105 0.049 TG 2 105 0.98 TCT 3 100 0.045 TGT 1 100 0.55 TCTC 1 86 0.21 IGHD 5' A 1 37 0.99 C 5 37 0.99 AC 1 36 0.86 CA 1 36 0.86 CC 2 36 0.96 CCC 1 33 0.77 IGHD 3' G 11 40 0.88 GT 2 36 0.57 TC 1 36 0.86 GTT 3 33 0.0022 IGHJ T 11 42 0.042 GT 3 34 0.26 AGC 1 32 0.45 ACT 2 32 0.026 1 α equal to 0.003 A single case of significance was observed among the putative P nucleotide sequences. This was for a sequence of 3 nucleotides (GTT) which was associated with the 3' end of the IGHD region. The occurrence of 3 'GTT' sequences in the dataset remains the only significant putative P nucleotides even if the alpha value is increased to 0.01. Using a 0.05 significance level, 6 sets out of 23 appear significant, however this conclusion carries a 69% chance of being incorrect and that the results occurred by chance. Although putative P nucleotide sequences are present in 10% of sequences most of these are likely to have arisen as the result of N nucleotide addition. P nucleotides can only be confidently attributed to less than 1% of sequences with three sequences from the IgM dataset contained statistically significant P nucleotides out of the 245 IGH V-D-J rearrangements examined. The overall contribution of P nucleotides to junctional nucleotides was 9 nucleotides out of 2899 junctional nucleotides, or 0.3% of junctional nucleotides, within the IgM dataset. The probability of 'GTT' occurring within an N region is 0.007875, therefore, the sequence could be expected to occur twice at the observed position in the 245 junctions examined, by chance alone. Of the three identified P nucleotides it is therefore possible that only one is a true P nucleotide. The contribution of P nucleotides to junctional nucleotides could therefore be as low as 0.1%, with P nucleotide inclusions occurring in less than 0.5% of sequences. Discussion Investigation of the role played by nucleotide loss and addition in the generation of immunoglobulin diversity has been limited by the ability of researchers to accurately determine gene ends. The development of a statistically based partitioning method has allowed this study to gain insights into the nature of nucleotide loss and addition in the expressed human immunoglobulin repertoire. Analysis of 294 human IgM sequences revealed significant differences between average nucleotide losses from different heavy chain genes segments. IGHV genes suffer less removal in comparison to other genes, suggesting that a process or processes may act to prevent the removal of critical components or to select against sequences in which such removals have occurred. Critical components may include the conserved TGT that defines the start of the CDR3 [ 27 ] and the internal heptamer site utilized in V H gene replacement [ 23 - 25 ]. The extension of the analysis to the gene subgroup level showed significant differences among removals from the IGHV gene subgroups, among the IGHD subgroups at the 3' end of the D gene and between the six IGHJ gene subgroups. The most striking contrast was observed for the IGHJ gene subgroups, specifically for IGHJ3 and IGHJ6. Removals from IGHJ3 averaged only a single nucleotide, while IGHJ6 on average lost in excess of 6 nucleotides. The differences observed as part of this study suggest that the loss of nucleotides during the creation of human heavy chain immunoglobulin sequences is not random. Unique 'patterns' of exonuclease removal between gene subgroups have previously been reported in murine immunoglobulins [ 12 , 28 ], however we are not aware of any such reports from studies of the human repertoire. Murine IGHJ4 genes have been reported to undergo an average removal of 2 nucleotides more than any other murine IGHJ subgroup [ 28 ]. Comparison of the murine IGHJ4 sequence to that of human IGHJ6 shows these two sequences to be identical for the first 7 nucleotides. The common sequence, ATTACTA, is unique to these IGHJ subgroups. This suggests that the common sequence may be linked to the high levels of nucleotide loss experienced by these gene subgroups, relative to the other IGHJ genes. The nucleotide composition of gene has previously been stated to influence exonuclease processing in the murine system [ 12 ]. The presence of two 'motifs' was thought to be the determining factor in the outcome of exonuclease processing. One motif involves stretches of two or more G or C nucleotides and is referred to here as the S motif. The other motif is composed of stretches of three or more consecutive A or T nucleotides and is referred to here as the W motif. S motifs in murine sequences were associated with low average removals from gene region ends, while the presence of W motifs correlated with high average removals [ 12 ]. Similar results were seen for human immunoglobulins in this study, with average removals from IGHJ genes containing 5' S motifs being significantly lower than from those containing 5' W motifs. Interestingly, the average position of the first S motif within the IGHJ genes coincided with the average level of removal from IGHJ genes. The correlation between S motif position and average exonuclease removal suggests that the S motif may act to block continued exonuclease removal from the gene region end. This may explain the high removals from the human IGHJ6 and murine IGHJ4 gene segments, as these sequences lack S motifs which may prevent such extensive exonuclease processing. Consideration of the W and S motif composition of the 3' D gene segments showed the IGHD2 and IGHD3 gene subgroups to be rich in W motifs and to lack S motifs (data not shown). These two subgroups showed higher average removals compared to other IGHD subgroups. The relationship between nucleotide composition and exonuclease activity could therefore explain the significant differences observed at the 3' end of the IGHD gene subgroups. IGHD2 and IGHD3 are both long D genes. A relationship between D gene length and exonuclease activity may have therefore been acting to influence exonuclease processing. Examination of exonuclease activity of D genes grouped by length did reveal significant differences (data not shown), however, these differences were only evident at the 3' end of the D genes and as the analysis was confined to IGHD2 and IGHD3 sequences, it is difficult to conclude whether sequence length has a role. The influence of nucleotide composition on exonuclease removals from heavy chain gene is easily examined in the IGHJ genes, due to the small number of alleles and the clear division of sequences based on the presence or absence of S and W motifs. Significant differences in the V genes were not further examined as the larger number of alleles made sample groups too small to allow for meaningful statistical analysis. The absence of significant differences at the 5' end of the IGHD genes may result from the lack of distinct differences in the nucleotide composition of these sequences. This would make any variations in exonuclease processing more subtle, and thus a larger sample size would be necessary to observe any differences. P nucleotide addition has been reported to contribute to diversity in between 10% [ 7 , 8 , 29 ] and 41% [ 9 ] of immunoglobulin sequences. Initial analysis of putative P nucleotides in the dataset of 294 human IgM sequences in this study revealed the frequency the presence of putative P nucleotides to be around ten percent of sequences. This is consistent with previous reports [ 7 , 10 , 29 ]. Statistical analysis of the putative P nucleotides sequences, however, revealed that only one 'P nucleotide sequence' was observed at a frequency that was significantly above the frequency that would be expected from N nucleotide addition alone. This suggests that the true contribution of P nucleotides to diversity in the expressed human IgM repertoire is much lower than previously reported, with P nucleotides present in less than 1% of sequences and accounting for approximately 0.3% of junctional nucleotides in the IgM dataset. It should be noted that even those P nucleotides accepted on the basis of statistical analysis carry a degree of uncertainty. Accounting for the possible misidentification of TdT additions among P nucleotides suggests that the contribution of P nucleotides to junctional nucleotides may be even lower than 0.3%. The identification of P nucleotides during partitioning of immunoglobulin sequences introduces a greater margin of error than would result from their exclusion from partitioning processes. For example, the rare nature of P nucleotide inclusion in rearranged immunoglobulins means that in a dataset of 1000 sequences less than 10 sequences may possess P nucleotides. Arbitrary identification of putative P nucleotides in the same dataset could however lead to 100 of the sequences being identified as having P addition. The error is therefore smaller if P nucleotides are not allocated as part of the partitioning process. The statistical demonstration of the phenomenon of P nucleotides by Meier and Lewis utilized altered recombination substrates, where the IGHV, IGHD and IGHJ gene regions were replaced by restriction sites on a plasmid vector [ 8 ]. The recombinant substrates were then transfected into murine cell lines and the processing of the substrates was then examined. Meier and Lewis observed that putative P nucleotides occur at a significant frequency among the processed recombinant substrates and this has been used as the basis for the allocation of P nucleotides in subsequent immunoglobulin studies [ 8 ]. Examination of the frequency of P nucleotides among the recombination substrates, however, revealed a five fold greater frequency of P nucleotide inclusions than was seen among adult murine T cell receptors and immunoglobulin sequences [ 8 ] and adult human immunoglobulins [ 7 , 29 ]. The applicability of Meier and Lewis' statistical analysis of the altered recombination substrates to demonstrate the contribution of P nucleotides must therefore be questioned, as the reporting of statistical significance is likely to be a direct result of the elevated frequency of putative P nucleotides among the recombinant substrates. Examination of putative P nucleotides at five-fold lower frequencies eliminates the significance observed in the original studies and supports the figure reported here of a contribution to junctional diversity in less than 1% of sequences (data not shown). Conclusions Substantial in vitro evidence in support of the formation of hairpin loops as part of the immunoglobulin rearrangement mechanism [ 6 , 30 , 31 ] and for the creation of P nucleotides as part of the process of hairpin loop opening exists [ 32 , 33 ]. The results reported here suggest that the P nucleotides generated by the loop opening do not, however, contribute significantly to the diversity of the final rearranged immunoglobulin. Exonuclease processing of IGH genes is not random and the nucleotide composition of the gene end appears to be influential. Further investigations into factor(s) influencing the exonuclease processing of gene ends will be required in order to elucidate the exact nature of the relationship between the gene end and exonuclease processing. Methods Dataset creation Human IgM sequences were obtained from public nucleotide databases; IMGT/LIGM-DB available through the IMGT, The IMmunoGeneTics Information System R , [ 18 ], the Entrez nucleotide database from the National Center for Biotechnology Information (NCBI) [ 34 ] and the Kabat Database of Sequences of Immunological Interest [ 35 ]. The sequences obtained were screened to exclude those sequences of fetal origin, those associated with diseases and those of a non-productive nature. Screening was necessary to avoid the introduction of any biases that may be associated with particular disease states or stages of immunological development. Sequences that contained in excess of 5 mutations within the V gene were also excluded from the final dataset. Sequences displaying higher levels of mutation were excluded from the analysis as the partitioning method utilised is most accurately applied to sequences with low levels of mutation. Partitioning of rearranged immunoglobulin sequences The determination of genes and N regions of the sequences within the dataset was performed with the aid of a statistical analysis of point mutations [ 21 ]. This method uses the number of mutations in the core region of the V genes to predict the level of mutation within other regions of the immunoglobulin sequence. The approach is based upon the mutability of trinucleotides, while also factoring in the exponential decay of somatic point mutations [ 36 ] and the effects of antigen selection [ 21 ]. The key to the analysis is the calculation of mutability scores for the various genes. Mutability scores can be used to indicate the likelihood that mutations will be distributed in a particular way between two or more parts of an immunoglobulin sequence. The focus of the study upon exonuclease removal required careful definition of rules for the identification of nucleotide losses. Preliminary alignment of sequences were performed using IMGT/V-QUEST [ 18 ]. Where identical IMGT/V-QUEST alignment scores were achieved to different alleles of a germline gene, the allele first allele reported was recorded. IMGT/V-QUEST compares input sequences to the IMGT reference sets which are the most complete sets of sequences that are available for IGHV, IGHD and IGHJ functional genes, their alleles and open reading frame genes. Alignments produced by IMGT/V-QUEST give an overall indication of similarity between a rearranged sequence and germline genes. D gene determination was performed using previously described criteria [ 21 ], where the level of required similarity to a germline sequence was dictated by the length of the junction and the likelihood of N nucleotides being misidentified as IGHD segments by chance. To aid in the allocation of D genes, a D Gene Alignment Utility was developed. This web based tool allowed alignments to be performed between a junctional sequence and all germline D genes, including inverted D gene sequences [ 37 ] obtained from the IMGT Reference Directory [ 19 ]. The program utilized an altered Smith-Waterman algorithm that did not allow for gaps [ 38 ]. Difficulties with immunoglobulin partitioning are often experienced, especially in the determination of gene ends. In this study, runs of consecutive nucleotide differences between a IGH V-(D)-J rearrangement and a germline sequence at a gene end were always attributed to exonuclease removal, rather than mutation of the gene end. The portion of a IGH V-(D)-J rearrangement that contained such differences, with respect to the germline sequence, was allocated to the N region, and the gene was considered to have undergone exonuclease processing at the gene end. Situations where IGHV or IGHJ ends revealed a series of differences and similarities to the germline sequence required the consideration of each possible combination of exonuclease removal and point mutation that could have led to the creation of the observed region end. Probabilities were calculated for each 'path' to the observed segment end. Mutability scores for the region end were used to determine the likelihood of point mutations contributing to the region end. TdT addition probabilities were used to calculate the likelihood of N additions generating particular nucleotide sequences. The TdT probabilities used were p(G) = 0.35, p(C) = 0.35, p(A) = 0.15, p(T) = 0.15 [ 5 ]. The path that displayed the greatest likelihood was used to allocate nucleotides to either an N region or a region end. Sequences that lacked exonuclease removals from gene region end(s) were examined for the presence of P nucleotides [ 2 ]. Self-complementary repeats of the gene end located in the neighboring N region were designated as putative P nucleotides. For example, if the V gene ended with the nucleotides GA, then CT was sought at the start of the N1 region. All possible lengths of P insertions were considered as part of the identification process. Examination of P nucleotides The contribution of P nucleotides to immunoglobulin diversity was investigated through data generated by the investigations reported here. Unique putative P nucleotides were identified and their appearance was tallied within the dataset. The number of junctions that displayed a lack of exonuclease activity at one or more gene ends was also calculated. The cumulative binomial probability of observing a given number of P nucleotides or greater was then calculated for each putative P nucleotide sequence in an attempt to estimate the contribution of P nucleotides to immunoglobulin diversity. Role of W and S Motifs in exonuclease processing The effect of W and S motifs upon exonuclease activity was investigated within the IGHJ genes. The IGHJ genes were grouped based upon the relative location of such motifs. W motifs were defined as sequences of 3 or more consecutive A or T nucleotides. S motifs were defined as sequences of 2 or more consecutive G or C nucleotides. The presence of these motifs was considered within the first 15 5' nucleotides of the IGHJ genes. Three sets were established, based upon the observed configurations of the motifs in the IGHJ subgroups; S motif followed by W motif, S motif only and W motif only (Table 2 ). Analysis of exonuclease removals for each of these groups was then performed by calculating the average number of nucleotides removed from the IGHJ end for each set. Table 2 Grouping of IGHJ genes by relative location of W and S motifs J Gene Sequence 1 Group IGHJ101 GC TGAATACTT CC AG 5' S only IGHJ201 TGCTACT GG TACTTG 5' S only IGHJ301 TGAT GC TTTT GATGT 5' S then W IGHJ302 AT GC TTTT GA TAT CT 5' S then W IGHJ401 ACTAC TTT GACTACT 5' W only IGHJ402 ACTAC TTT GACTACT 5' W only IGHJ403 GC TAC TTT GACTACT 5' S then W IGHJ501 ACAACT GG TT CG ACT 5' S only IGHJ502 ACAACT GG TT CG ACC 5' S only IGHJ601 ATTA CTACTACTACT 5' W only IGHJ602 ATTA CTACTACTACT 5' W only IGHJ603 ATTA CTACTACTACT 5' W only 1 W motifs shown underlined , S motifs shown in italics Statistical analysis The extent of nucleotide deletion was calculated as the average number of nucleotides removed for a given dataset. Significant differences between average removals were determined using one-way ANOVA for normally distributed datasets, and Kruskal-Wallis Test for other datasets. Where significant differences were found, multiple comparison testing was carried out using Tukey's Multiple Comparison Test, for normally distributed datasets, and Dunn's Multiple Comparison Test for non-normally distributed datasets. All analysis of exonuclease removals was carried out using GraphPad Prism (Version 3.00, 1999, GraphPad Software) with an alpha value of 0.05. An analysis of P nucleotides was performed by calculating the probability that putative P nucleotides may actually have resulted from N nucleotide addition by TdT. Probabilities were calculated as described by Meier and Lewis [ 8 ]. The probability of the presence of the observed or a greater number of P nucleotides was calculated as follows [ 8 ]: where, n is the observed number of P nucleotides and N is the total number of sequences containing junctional inserts equal to or greater than the length of the P nucleotide(s) being examined, and p is the expected frequency of the observed P nucleotide sequence, which was calculated using reported TdT frequencies [ 5 ]. Probabilities were calculated for each observed putative P nucleotide sequence, and the alpha value was adjusted for the number of comparisons made, using the Bonferroni correction. Authors contributions KJ carried out the dataset creation and partitioning, performed the statistical analysis and drafted the paper. BG aided in the design of the study and the development of the statistical analysis. WS advised on study design and development. AC devised the study, aided in study design and co-ordination and revised the manuscript. All authors read and approved the final manuscript. Supplementary Material Additional File 1 294 Partitioned human IgM Sequences 294 human IgM sequences were partitioned using a statistically based model and used in the examination of exonuclease activity and P nucleotide addition in the expressed human repertoire. The file is in Microsoft Excel format. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517710.xml
517704	In vivo clearance of surfactant lipids during acute pulmonary inflammation.	Background A decrease in pulmonary surfactant has been suggested to contribute to the lung dysfunction associated with pulmonary inflammation. A number of studies have implicated surfactant clearance as a possible mechanism for altered pool sizes. The objective of the current study was to specifically investigate the mechanisms of surfactant clearance in a rodent model of acute pulmonary inflammation. Methods Inflammation was induced by intrapulmonary instillation of lipopolysaccharide (LPS: 100 μg/kg). Lipid clearance was assessed at 18 and 72 hours post-LPS instillation by intratracheal administration of radiolabel surfactant-like liposomes 2 hours prior to isolation and analysis of inflammatory cells and type II cells. Results At both 18 and 72 hours after LPS instillation there was significantly less radioactivity recovered in the lavage fluid compared to respective control groups (p < 0.05). At both time points, the number of cells recovered by lavage and their associated radioactivity was greater compared to control groups (p < 0.01). There was no difference in recovery of radioactivity by isolated type II cells or other cells obtained from enzymatic digestion of lung tissue. Conclusion These results show that increased clearance of surfactant lipids in our model of acute pulmonary inflammation is primarily due to the inflammatory cells recruited to the airspace and not increased uptake by alveolar type II cells.	Background Pulmonary surfactant is a phospholipid-protein complex that lines the inner surface of the lung and is essential for normal pulmonary function. Surfactant acts to promote lung stability by reducing surface tension within the lung, while also protecting against inhaled pathogens. Surfactant is composed of approximately 90% lipids and 10% proteins by weight. The lipid component is primarily phospholipids with phosphatidylcholine (PC) being the most abundant, and the protein component comprised of four surfactant-associated proteins designated SP-A, SP-B, SP-C and SP-D [ 1 ]. The reduction of surface tension within the lung is a result of the interaction between surfactant phospholipids and the two hydrophobic surfactant proteins, SP-B and SP-C [ 1 ], while the two hydrophilic proteins, SP-A and SP-D, are members of a family of innate immune molecules called collectins [ 2 ]. Collectins opsonize bacteria and viruses and enhance their phagocytosis by macrophages and neutrophils [ 2 ]. Alterations of the pulmonary surfactant system, including decreased total surfactant levels, have been implicated in the pathophysiology of acute lung injury. Multiple studies of patients with a variety of lung diseases have shown that surfactant levels are decreased in the inflamed, injured, or infected lung [ 3 - 5 ]. In agreement, decreases in alveolar surfactant lipid pools have also been observed in several animal models of lung inflammation induced by both direct insults to the lung, such as bacterial infection [ 6 ], oxygen toxicity [ 7 , 8 ], endotoxin administration [ 9 - 11 ], and by indirect insults, such as N-nitroso-N-methylurethane [ 12 ] and cecal ligation and perforation [ 13 ]. Alveolar metabolism of surfactant is a complex process, primarily involving type II epithelial cells that synthesize, secrete and clear surfactant from the airspaces [ 14 ], along with phagocytic cells such as macrophages and neutrophils that participate in surfactant clearance [ 15 , 16 ]. In a situation of pulmonary inflammation, altered type II cell metabolism has been thought to play a role in the alterations of surfactant lipid levels. Viviano et al. observed a decrease in alveolar surfactant levels and a corresponding increase in intracellular surfactant after lipopolysaccharide (LPS) administration [ 10 ]. Additionally, exposure of rat lungs ventilated ex vivo to LPS resulted in the presence of giant lamellar bodies within the type II cells [ 17 , 18 ]. Collectively these studies suggest that type II cell metabolism is altered after LPS administration and that a possible explanation for decreased surfactant pool size may be increased clearance of surfactant lipids by the type II cells. Additional experimental evidence has also implicated recruited inflammatory cells as having an impact on surfactant pool sizes. Both neutrophils and macrophages recovered from LPS exposed lungs had a greater capacity to internalize surfactant-like lipids compared to control cells in vitro [ 16 ]. Therefore these recruited inflammatory cells may also have a significant impact on surfactant pool size by increasing the overall surfactant lipid clearance. The objective of the current study was specifically to investigate the mechanisms of in vivo surfactant clearance in a rodent model of acute pulmonary inflammation induced by intrapulmonary instillation of LPS. We hypothesized that during pulmonary inflammation there is altered clearance of alveolar phospholipid by both alveolar type II cells and inflammatory cells within the airspaces. We analyzed in vivo clearance of surfactant lipids by a variety of pulmonary cells at 18 and 72 hours after LPS instillation. Results indicated that increased clearance of surfactant lipids in our model of acute pulmonary inflammation is primarily due to enhanced lipid uptake by the inflammatory cells recruited to the airspace, whereas uptake by type II cells was unaltered. However the in vivo surfactant uptake by inflammatory cells in the current study was significantly less than that predicted by previous in vitro studies [ 16 ]. Methods Materials Dipalmitoylphosphatidylcholine (DPPC), egg phosphatidylcholine (PC), dipalmitoylphosphatidylglycerol (DPPG) and cholesterol were purchased from Avanti Polar Lipids (Birmingham, AL). L-α-dipalmitoyl [2-palmitoyl-9,10- 3 H(N)]PC was obtained from DuPont New England Nuclear (Boston, MA). Elastase for type II cell isolations was purchased from Worthington Biochemicals (Freehold, NJ). Dulbecco's PBS, DMEM, and fetal bovine serum (FBS) were obtained from Life Technologies (Gaithersburg, MD). Low-endotoxin BSA, O26:B6 Escherichia coli LPS, Rat IgG and all other chemicals were purchased from Sigma (St. Louis, MO). Chloroform and methanol were obtained from EM Science (Gibbstown, NJ). Preparation of Liposomes Small unilamellar liposomes were prepared with a lipid composition similar to pulmonary surfactant: 52% DPPC, 26% egg PC, 15% DPPG and 7% cholesterol by weight with trace amounts of 3 H-DPPC (12 μCi/mg phospholipid). The lipids were dried under nitrogen, reconstituted in 0.15 M saline and extruded from a French Press cell under 900 psi. This resulted in small unilamellar liposomes at a concentration of approximately 1 mg lipid/ml. Animal Model Male pathogen-free Sprague Dawley rats (150–200 g; Taconic Farms, Germantown, NY) were used for the current study. For LPS animals, endotoxin (0.1 mg/kg of O26:B6 E. coli LPS) was suspended in 300 μl of sterile 0.15 M saline. Control animals received an equal volume (300 μl) of sterile 0.15 M saline. For instillation, animals were anesthetized with halothane such that they remained unconscious throughout the entire instillation procedure and had no cough reflex upon intubation. Animals were placed on a board at a 45° angle, intubated with an 18-gauge blunt ended catheter and either sterile saline or LPS suspension was instilled followed by five 1 ml boluses of air to facilitate the distribution of the instilled fluid. Two hours prior to killing, liposomes (100 μg lipid) were intratracheally instilled in all groups following the identical instillation procedure. Separate groups of both control and LPS animals were sacrificed at 18 or 72 hours after saline or LPS instillation. Whole Lung Studies One control animal and one LPS animal were investigated simultaneously, thus all procedures were done in parallel. At 18 or 72 hours after saline or LPS instillation, animals were anesthetized with 0.3 mg of sodium pentobarbital and 700 U of heparin. After loss of toe pinch reflex, the trachea was cannulated and the rat was exsanguinated via transection of the descending aorta. The chest cavity was subsequently opened and the lungs perfused through the pulmonary artery with 40–50 ml of a calcium buffer (140 mM NaCl, 5 mM KCl, 2.5 mM Na 2 HPO 4 , 10 mM HEPES, 2.0 mM CaCl 2 , and 1.3 mM MgSO 4 at 37°C). The lung from the first animal was carefully removed from the thoracic cavity and placed between saline soaked gauze pads while the second lung was perfused and removed in identical fashion. The time for lung perfusion and removal was approximately 10 minutes. After lungs were isolated from both a control and LPS animal, they were lavaged simultaneously with eight 10-ml volumes of EGTA buffer (140 mM NaCl, 5 mM KCl, 2.5 mM Na 2 HPO 4 , 10 mM HEPES, and 0.2 mM EGTA at 37°C). For each animal the individual lavages were collected and combined to make up the bronchoalveolar lavage fluid (BALF), which was immediately stored on ice. After the lavage procedure, individual lobes were dissected away from the major airways. The lung tissue was then cut into 5 mm pieces in 5 ml calcium buffer and subsequently homogenized with a Polytron PT-MR-2100. After complete homogenization, total lung tissue volume was diluted to 40 ml with calcium buffer and stored on ice. A small aliquot (2 ml) of the total BALF was removed and the remainder was centrifuged at 250 g for 10 minutes at 4°C to generate a pellet that was primarily BALF cells. The cell pellet was subsequently suspended in 10 ml of PBS. Lavage cell numbers were determined by a hemocytometer, viability determined by trypan blue exclusion and cell differential determined by Hemacolor staining of cytospins that were prepared using a Shandon Cytospin 2 centrifuge. Liposome-association in total BALF, cell-free BALF, BALF cells and homogenized lung tissue was determined by scintillation counting. Surfactant Phospholipid Measurement Alveolar phospholipid levels were measured in the cell-free BALF by phospholipid-phosphorous measurement. Lipids were extracted using the method of Bligh & Dyer [ 19 ] and phospholipid levels were determined using a modification of the Duck-Chong phosphorous assay [ 20 ]. Briefly, 100 μl of 10% magnesium nitrate in methanol was added to the extracted lipids. After drying, the samples were ashed in a fume hood on an electric rack for approximately 1 min. After 1 ml of 1 M HCl was added, the samples were reheated on a heating block while covered for 15 min at 95°C. After cooling, a 66 μl aliquot of each sample was added to individual wells of a 96 well plate along with 134 μl of a dye consisting of 4.2% ammonium molybdate in 4.5 M HCl with 0.3% malachite green (1:3 vol/vol). The absorbency of each sample was read at 650 nm using a Biorad 550 microplate reader and compared to reference standards on the same plate. Surfactant Protein A (SP-A) Analysis Relative quantities of alveolar SP-A from 18-hour control animals, 18-hour LPS animals and 72-hour LPS animals were determined by loading equal volumes of BALF on 15% SDS-PAGE gels under reducing conditions. Total BALF recovery was not different among the individual animals. Proteins were then transferred to nitrocellulose where SP-A was probed with a well characterized polyclonal rabbit anti-rat SP-A [ 21 ]. The nitrocellulose was subsequently developed using the ECL system (Amersham Pharmacia Biotech, Piscataway, NJ). Type II Cell Isolation A separate cohort of animals was required for type II cell isolation studies. As with the whole lung studies, type II cell isolation was completed for one control and one LPS rat in parallel for both 18- and 72-hour time points after instillation of saline or LPS. The type II cell isolation procedure was described previously with minor modifications of the original protocol [ 22 , 23 ]. The modifications were designed for isolation of type II cells from inflamed lungs, and included increasing the elastase to 3,000 orcein units/lung, and increasing the surface area for IgG panning by two-fold. Briefly the isolation procedure involved killing the rats, removing the lungs and lavaging the lungs eight times as stated in Whole Lung Studies. Subsequently the lungs were lavaged twice with calcium buffer (37°C) and once with elastase solution (3000 orcein/40 ml calcium buffer; 37°C). The lungs were then filled with the elastase solution and suspended in warmed saline (37°C) for 20 minutes. Lung tissue was removed from the major airways, cut into 5-mm pieces and chopped 200 times with sharp scissors in 5 ml of the calcium solution along with 2 mg of DNase. The tissue suspension was added to a flask with 4 ml of FBS and 35 ml of calcium buffer and shaken vigorously for 2 min in a 37°C water bath. The resultant tissue suspension was strained through gauze, and then decreasing sizes of nylon mesh (150-, 15-, and 8-μm). The cell suspension was centrifuged at 250 g for 10 min at 4°C, the resulting cellular pellet was suspended in 20 ml of warm DMEM (37°C) and incubated on two IgG coated Petri dishes (150 × 15 mm) for 30 minutes at 37°C. After the incubation period, nonadherent cells were removed from the plate, washed once with PBS, suspended in calcium buffer and used for determination of radiolabel recovery associated with Type II cells. The remaining adherent cells were gently scraped off the plate in 5 ml of DMEM and transferred into a polypropylene tube; this population of cells is referred to as "plate cells" and consisted primarily of macrophages and/or neutrophils as described in detail below. The cells were washed once and suspended in calcium buffer for determination of associated radioactivity. A small aliquot from the isolated type II cells and plate cells was saved for differential cell count and viability. Cell purity was determined by counting a minimum of 250 cells from random fields after staining by the Papanicolaou method [ 22 ]. Statistics All data reported are means ± standard error (SE). Repetitions (n) used to calculate means ± SE were from independent experiments, not from replicates within an experiment. An analysis of variance (ANOVA) was used to determine differences between all experimental groups at a specific time point, followed by a Tukey post-hoc test for multiple comparisons. Significance was accepted when p < 0.05. Results LPS-induced alterations: BALF cells and alveolar surfactant Table 1 reveals total alveolar cell numbers and differentials from the BALF of control and LPS groups killed at the 18- and 72-hour time points. Eighteen hours after instillation of LPS there were significantly more cells recovered in the BALF from the LPS treated animals compared to the saline control group (p < 0.01). Cell differentials from the two 18-hour groups revealed that the LPS group had primarily neutrophils in the BALF, whereas the cells recovered from the control group were predominantly macrophages. Seventy-two hours after LPS instillation there were significantly greater numbers of cells in the BALF of the LPS group compared to the 72-hour control group (p < 0.01), but significantly fewer numbers of cells compared to the LPS 18-hour group (p < 0.01). Cell differentials from the 72-hour groups revealed that the cells were predominately macrophages in both the LPS and control groups. Of note, there were no differences in cell numbers or differentials between the two control groups at the different time points. Table 1 Bronchoalveolar lavage fluid characteristics from animals killed 18 and 72 hours after instillation of saline (Control) or lipopolysaccharide (LPS). Control 18 (n = 6) LPS 18 (n = 6) Control 72 (n = 6) LPS 72 (n = 6) Cell Numbers (10 6 ) 12.3 ± 1.1 102.8 ± 8.1* 11.3 ± 0.9 57.3 ± 6.6* # % Macrophages 98.2 ± 0.3 7.3 ± 0.8* 97.7 ± 0.4 90 ± 1.5* % Neutrophils 1.8 ± 0.3 92.7 ± 0.8* 2.3 ± 0.4 10 ± 1.5* Alveolar Phospholipid (mg PL/kg BW) 14.7 ± 0.7 12.8 ± 0.9 14.0 ± 0.4 12.9 ± 0.5 Abbreviations: PL = phospholipid; BW = body weight. Values are mean ± SEM. * = p < 0.01 vs. appropriate control; # = p < 0.01 vs. LPS 18. Table 1 also displays the total phospholipid levels measured in the BALF of the four experimental groups. There was an approximate 15% decrease in total phospholipid recovered by lung lavage from both LPS groups compared to their respective saline control groups, however this did not reach statistical significance. Of note, there was no difference in mean body weights among the four experimental groups (Con 18: 177 ± 7 g; LPS 18: 180 ± 10 g; Con 72: 186 ± 9 g; LPS 72: 192 ± 7 g). Figure 1 shows Western blot analysis of SP-A measured in the BALF recovered from animals killed 18 hours post saline instillation (control), and 18 hours and 72 hours after LPS instillation. There was relatively greater quantity of SP-A in the BALF of animals killed at both 18 and 72 hours after LPS instillation compared to the control animals, consistent with previous reports using similar models of intrapulmonary LPS administration [ 9 - 11 ]. Figure 1 Alveolar surfactant protein A (SP-A) levels. Relative quantities of alveolar SP-A were measured by Western Blot. Equal volumes of bronchoalveolar lavage fluid (BALF) from the individual animals were utilized and total BALF recovery was not different among these animals. Lanes 1 and 2 represent individual animals that were lavaged 18 hours after saline instillation. Lanes 3 and 4 represent individual animals that were lavaged 18 hours after instillation of 100 μg/kg O26:B6 lipopolysaccharide (LPS). Lanes 5 and 6 represent animals lavaged 72 hours after instillation of 100 μg/kg 026:B6 LPS. Whole Lung Studies: Clearance of Radiolabel Liposomes Figure 2 shows distribution of the recovered radioactivity associated with the cell free BALF, the isolated BALF cells, and whole lung tissue for the two groups killed 18 hours after instillation of saline or LPS. There was no difference in total lung radioactivity recovered compared to the total radiolabel instilled between the Control 18 and LPS 18 groups (46.0 ± 4.6% and 48.4 ± 4.0%; respectively). There was significantly less radiolabel recovered in the BALF (p < 0.05) and significantly more radiolabel associated with the BALF cells (p < 0.01) in the LPS group compared to the control group. There was no difference in radiolabel association with lung tissue between the two groups. Figure 2 Distribution of radiolabel liposomes 18 hours after instillation of LPS. Whole lung distribution of the total recovered 3 H-liposomes instilled 2 hours prior to killing. Animals were killed 18 hours after instillation of saline (Con 18) or 100 μg/kg O26:B6 lipopolysaccharide (LPS 18) where the lungs were lavaged, lavage cells isolated and whole lung tissue homogenized. Radioactivity was subsequently measured in cell free bronchoalveolar lavage fluid (BALF), isolated BALF cells and whole lung tissue homogenate (Tissue). Data are means ± SEM and expressed as a percentage of total recovered radiolabel; n = 6 animals/group. Statistical significance * = p < 0.05 vs. respective control group. Figure 3 displays total recovered radioactivity associated with the cell free BALF, the isolated BALF cells, and whole lung tissue for the two groups killed 72 hours after instillation of saline or LPS. There was no difference in the total lung radiolabel recovered compared to the total radiolabel instilled between the Control 72 and LPS 72 groups (47.3 ± 4.0% and 44.8 ± 3.8%; respectively). There was significantly less radiolabel recovered in the BALF (p < 0.05) and significantly more radiolabel liposomes associated with the BALF cells (p < 0.01) in the LPS group compared to the control group. There was no difference in radiolabel associated with lung tissue between the two groups. Figure 3 Distribution of radiolabel liposomes 72 hours after instillation of LPS. Whole lung distribution of the total recovered 3 H-liposomes instilled 2 hours prior to killing. Animals were killed 72 hours after instillation of saline (Con 72) or 100 μg/kg O26:B6 lipopolysaccharide (LPS 72) where the lungs were lavaged, lavage cells isolated and whole lung tissue homogenized. Radioactivity was subsequently measured in cell free bronchoalveolar lavage fluid (BALF), isolated BALF cells and whole lung tissue homogenate (Tissue). Data are means ± SEM and expressed as a percentage of total recovered radiolabel; n = 6 animals/group. Statistical significance * = p < 0.05 vs. respective control group. Additional controls were performed in which animals (n = 2) received radiolabel liposomes followed by whole lung lavage and lung tissue homogenization 5 minutes after the instillation procedure. Mean total lung radiolabel recovery for this control group was 77% suggesting that ~25% of the radiolabel liposomes were lost in the instillation procedure (i.e., syringe, catheter), or possibly were adhered to the airway epithelium. The distribution of the recovered liposomal radioactivity within the 5-minute control lungs was 91% associated with the BALF, 2% associated with BALF cells, and 7% associated with lung tissue. Type II Cell Isolation: Cell Recoveries, Purities and Radioactivity per Cell Table 2 demonstrates the cell recovery and cell differential after type II cell isolation for the two experimental groups at the 18-hour time point. Type II cell recovery for the control group ranged from 10.4 × 10 6 to 21.6 × 10 6 cells with a mean of 17.1 × 10 6 . The purity of this cell fraction averaged greater than 80% type II cells and viability was greater than 95%. Plate cell recovery from the control group ranged from 1.8 × 10 6 to 4.7 × 10 6 cells with a mean value of 3.2 × 10 6 . This cell fraction was predominately macrophages and the viability was greater than 90%. For the LPS group at the 18-hour time point, type II cell recovery ranged from 10.6 × 10 6 to 20.5 × 10 6 cells and a mean of 16.3 × 10 6 cells, with a purity greater than 85% type II cells and viability greater than 95%. Plate cells from the LPS 18 group ranged from 2.7 × 10 6 to 13 × 10 6 cells with a mean of 6.6 × 10 6 cells. This cell population consisted primarily of neutrophils and macrophages and viability was greater than 90%. There were no significant differences in cell numbers obtained from control and LPS animals 18 hours after saline or LPS instillation. Table 2 Cell recovery and differential after lung digestion and IgG panning for animals killed 18 hours after instillation of saline (Control) or lipopolysaccharide (LPS). Control 18 (n = 7) LPS 18 (n = 7) Type II Cells Plate Cells Type II Cells Plate Cells Cell Numbers (10 6 ) 17.1 ± 1.6 3.2 ± 0.5 16.3 ± 1.8 6.6 ± 1.7 % Type II 81 ± 2 15 ± 7 86 ± 2 14 ± 2 % Macrophages 9 ± 1 61 ± 5 3 ± 1* 33 ± 2* % Neutrophils 1 ± 1 17 ± 10 4 ± 1 53 ± 3* % Other Cells 9 ± 2 7 ± 2 7 ± 2 - Values are mean ± SEM. * = p < 0.01 vs. Control. Figure 4 reveals the radioactivity per million tissue cells for the isolated type II cells and the other tissue associated cells (plate cells) for the control and LPS 18-hour groups. There were no significant differences in the radioactivity per type II cell or plate cell between the control group and the LPS group at the 18-hour time point after instillation. Figure 4 Liposome uptake by lung tissue cells 18 hours after instillation of LPS. Association of 3 H-liposomes instilled 2 hours prior to killing in cells isolated from lung tissue. Animals were killed 18 hours after instillation of saline (Con 18) or 100 μg/kg O26:B6 LPS (LPS 18). Lungs were lavaged and type II cells and other tissue-associated cells (plate cells) were isolated. Data are means ± SEM and expressed as DPM/10 6 cells, n = 7 animals/group. Table 3 shows the cell recovery and cell differential after type II cell isolation for the two experimental groups at the 72-hour time point. Type II cell recovery for the control group ranged from 21.2 × 10 6 to 40 × 10 6 cells with a mean of 28.4 × 10 6 . The purity of this cell fraction averaged 75% type II cells and viability was greater than 95%. The number of cells obtained from the plate group ranged from 1.8 × 10 6 to 15.9 × 10 6 cells with a mean value of 6.2 × 10 6 with the primary cell type being macrophages with a small percentage of neutrophils and type II cells. Viability for the plate cells were greater than 90%. For the LPS group at the 72-hour time point, type II cell recovery ranged from 19.2 × 10 6 to 64.6 × 10 6 cells and a mean of 33.2 × 10 6 cells, purity greater than 80% type II cells and viability greater than 95%. Plate cells from the LPS 18-hour group ranged from 3.3 × 10 6 to 16.2 × 10 6 cells with a mean of 7.8 × 10 6 cells. This cell population consisted primarily of macrophages and viability was greater than 90%. There were no significant differences in cell recoveries between the control and LPS groups at the 72-hour time point. Of note, there were a greater number of cells recovered from both 72-hour groups compared to both 18-hour groups. Table 3 Cell recovery and differential after lung digestion and IgG panning for animals killed 72 hours after instillation of saline (Control) or lipopolysaccharide (LPS). Control 72 (n = 6) LPS 72 (n = 7) Type II Cells Plate Cells Type II Cells Plate Cells Cell Numbers (10 6 ) 28.4 ± 3.9 6.2 ± 2.2 33.2 ± 5.6 7.8 ± 2.2 % Type II Cells 75 ± 2 22 ± 3 83 ± 1 19 ± 3 % Macrophages 11 ± 2 56 ± 5 12 ± 2 59 ± 4 % Neutrophils - 22 ± 7 2 ± 1 22 ± 3 % Other Cells 14 ± 1 - 3 ± 3* - Values are mean ± SEM. * = p < 0.01 vs. Control. Figure 5 reveals the radioactivity per million tissue cells for the isolated type II cells and the other tissue-associated cells (plate cells) for the control and LPS 72-hour groups. There were no significant differences in the radioactivity per type II cell or plate cell between the LPS group and the control group 72 hours after the appropriate instillation. Of note there was no significant difference in radioactivity per cell for both the type II cells and plate cells between the two 72-hour groups and the two 18-hour groups. Figure 5 Liposome uptake by lung tissue cells 72 hours after instillation of LPS. Association of 3 H-liposomes instilled 2 hours prior to killing in cells isolated from lung tissue. Animals were killed 72 hours after instillation of saline (Con 72) or 100 μg/kg O26:B6 LPS (LPS 72). Lungs were lavaged and type II cells and other tissue-associated cells (plate cells) were isolated. Data are means ± SEM and expressed as DPM/10 6 cells; Con 72 = 6 animals, LPS 72 = 7 animals. Discussion In the present study, we evaluated the clearance of surfactant-like liposomes at 18 and 72 hours following the intrapulmonary instillation of LPS. At both time points, radiolabel liposomes were instilled two hours prior to killing and the distribution within the lung was determined. Radioactivity was measured in cell-free BALF, isolated BALF cells, whole lung tissue, isolated type II cells and remaining tissue-associated cells. At both time points there was a significant increase in clearance of exogenous liposomes from the airspace and a small decrease in alveolar surfactant phospholipid levels in the LPS groups compared to control groups that received vehicle only. This corresponded with increased radiolabel associated with isolated BALF cells and no difference associated with lung tissue at either time point after LPS instillation compared to controls. There was no difference in liposomal radioactivity associated with isolated type II cells or other tissue cells between both LPS groups and their respective control groups, which was in agreement with whole lung measurements. These data suggest that cells recruited into the airspace as a consequence of the LPS-induced inflammatory response can significantly contribute to increased clearance of surfactant-like liposomes from the airspace (Figs. 2 and 3 ). Eighteen hours after LPS instillation there was an infiltration of neutrophils into the airspace which resulted in a 10-fold increase in BALF cells. At this time point there was a doubling of total liposomal radioactivity associated with these inflammatory cells compared to alveolar cells recovered from control animals. Neutrophils recruited into alveolar spaces during infection are essential for host defense by phagocytosis and killing of bacteria and other infectious agents. Previous studies have documented that surfactant can in fact modulate neutrophil functions, as surfactant proteins A and D can enhance neutrophil uptake of bacteria [ 24 ]. Although these phagocytic cells are extremely important in host defense, they have also been implicated as a contributing factor to the lung injury in a variety of inflammatory disorders. In a recent study by Quintero et al., it was demonstrated by confocal microscopy that neutrophils isolated from inflamed lungs were able to significantly internalize and degrade surfactant lipids in vitro [ 16 ]. Based on in vitro measurements of lipid uptake, they estimated that neutrophils could account for up to 48% of the observed clearance and could significantly impact surfactant homeostasis. Data from the current study supports this idea by demonstrating that in vivo , neutrophils can indeed contribute to the clearance of surfactant lipids during pulmonary inflammation. At the 72-hour time point after LPS instillation there was an accumulation of macrophages in the alveolar space that resulted in a 5-fold increase in BALF cells compared to control animals. There was a doubling of radiolabeled phospholipids associated with these alveolar cells and no increase in the amount of radiolabeled phospholipids associated with whole lung tissue. This resulted in a significant increased in clearance of total liposomal radioactivity from the BALF, 72 hours after LPS instillation. Alveolar macrophages have been shown in vitro to take up and degrade surfactant lipids and have also been shown in vivo to account for approximately 20% of surfactant lipid clearance in normal lungs [ 16 , 25 ]. Macrophages isolated from LPS-exposed lungs were shown to have a greater capacity to take up surfactant lipids in vitro than macrophages isolated from normal lungs [ 16 ], suggesting that in an inflamed lung, the recruited macrophages may have a significant role in surfactant metabolism. Indeed, the in vivo data presented in the current study support the theory that activated alveolar macrophages can impact clearance of surfactant lipids. Of note, previous in vitro data had predicted that inflammatory cells (neutrophils and macrophages) could account for a 6 to 13 fold increase in lipid clearance [ 16 ]. In the current study there was a 5–10 fold increase in alveolar inflammatory cells but only a doubling of cell associated radioactivity and a change in alveolar phospholipid pool size of approximately 15%. Importantly, the current study instilled a similar dose of LPS, had similar end points and the same number of inflammatory cells recovered from the BALF as the aforementioned in vitro study. Although the present in vivo data demonstrated that the inflammatory cells can indeed contribute to increased lipid clearance as suggested previously, the absolute value was considerably less than that predicted from the in vitro studies. Although we do not have an explanation for this quantitative difference between the in vivo and in vitro observations, it is possible that the process of cell isolation in the in vitro study resulted in their activation for subsequent lipid uptake. Alternatively, in vivo , multiple factors can impact total surfactant pool size in addition to clearance by inflammatory cells. Type II cells are the predominant cell type that regulates surfactant metabolism, being involved in synthesis, secretion and clearance. Secretion of surfactant phospholipids is solely a property of alveolar type II cells, whereas clearance in a healthy lung is regulated primarily by both type II cells and alveolar macrophages. In one study utilizing uninjured rabbits, it was determined that type II cells accounted for approximately 65% of the clearance of alveolar phospholipid [ 25 ]. In a situation of pulmonary inflammation induced by LPS administration, there have been documented decreases in alveolar surfactant phospholipids [ 9 - 11 ], increases in intracellular surfactant phospholipids [ 10 ] and the appearance of giant lamellar bodies within type II cells [ 17 , 18 , 26 ]. From these observations, LPS induced inflammation could either inhibit surfactant secretion or enhance surfactant clearance by type II cells, resulting in diminished alveolar surfactant pool and increased intracellular pool. We originally hypothesized that after LPS administration there was increased clearance of surfactant phospholipids by type II cells. However, data from this study demonstrated that in vivo , surfactant clearance by type II cells was similar at both 18 and 72 hours after intrapulmonary LPS instillation compared to control groups killed at the same time points (Figs 4 and 5 ). Data from the current study cannot lead to any conclusions pertaining to alterations in type II cell phospholipid synthesis or secretion. As mentioned, we observed a small but not statistically significant decrease in alveolar phospholipid pools after LPS administration, which is in contrast to previous studies by Viviano et al. and MacIntosh et al. that used 1 mg/kg and 0.5 mg/kg of LPS respectively [ 9 , 10 ]. Nevertheless, we did observe a similar increase in alveolar levels of SP-A after intrapulmonary LPS (Fig 1 ), which is a well documented characteristic of lung inflammation after LPS exposure [ 9 - 11 ]. This increase has been primarily attributed to increased SP-A synthesis by type II cells [ 11 ] suggesting that indeed type II cell metabolism was altered in our model. Additionally, these data provide further experimental evidence that surfactant phospholipid and SP-A pools can be independently regulated in response to inflammatory agents. We also investigated additional tissue-associated cells that were isolated after lung digestion (primarily macrophages and neutrophils). Similar to type II cells, surfactant clearance from these tissue-associated cells isolated from the two LPS groups were similar to their respective control groups. In addition, the clearance attributed to type II cells and tissue-associated cells was similar at both time points for the LPS and control groups. This observation is in agreement with a study by Gurel et al., that documented that alveolar type II cells and tissue macrophages contributed equally to the alveolar clearance of phospholipid in adult mice [ 27 ]. Conclusions In conclusion, we demonstrated that after intrapulmonary LPS administration in adult rats, the in vivo clearance of surfactant lipids by type II cells is unaltered and that the increased clearance of surfactant lipids is primarily due to either neutrophils or macrophages that are recruited to the alveolar space. Authors' contributions JLM conceived, designed and performed all aspects of the study, and was the primary participant in its writing. JRW aided in conception and design of study, and participated in its writing.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517704.xml
550665	A novel surgical procedure for bridging of massive bone defects	Background Bony defects arising from tumor resection or debridement after infection, non-union or trauma present a challenging problem to orthopedic surgeons, as well as patients due to compliance issues. Current treatment options are time intensive, require more than one operation and are associated with high rate of complications. For this reason, we developed a new surgical procedure to bridge a massive long bone defect. Methods To bridge the gap, an in situ periosteal sleeve is elevated circumferentially off of healthy diaphyseal bone adjacent to the bone defect. Then, the adjacent bone is osteotomized and the transport segment is moved along an intramedullary nail, out of the periosteal sleeve and into the original diaphyseal defect, where it is docked. Vascularity is maintained through retention of the soft tissue attachments to the in situ periosteal sleeve. In addition, periosteal osteogenesis can be augmented through utilization of cancellous bone graft or in situ cortical bone adherent to the periosteal sleeve. Results The proposed procedure is novel in that it exploits the osteogenic potential of the periosteum by replacing the defect arising from resection of tissue out of a pathological area with a defect in a healthy area of tissue, through transport of the adjacent bone segment. Furthermore, the proposed procedure has several advantages over the current standard of care including ease of implementation, rapid patient mobilization, and no need for specialized implants (intramedullary nails are standard inventory for surgical oncology and trauma departments) or costly orthobiologics. Conclusions The proposed procedure offers a viable and potentially preferable alternative to the current standard treatment modalities, particularly in areas of the world where few surgeons are trained for procedures such as distraction osteogenesis ( e.g. the Ilizarov procedure) as well as areas of the world where surgeons have little access to expensive, complex devices and orthobiologics.	Background Replacement of bone where there is none is one of the most challenging problems facing orthopedic surgeons today. In the case of tumor resection or trauma, massive bone defects must be filled with regenerate bone as quickly as possible in order to restore function. Current standards for bridging of massive bone defects in long bones generally follow a theme of i ) filling the defect with bone autograft or allograft (including cancellous bone graft or bone transplantation via vascularized or non-vascularized fibula transfer) and ii ) accelerating functional remodeling and integration through addition of physical and/or chemical stimuli such as tension (e.g. Ilizarov technique, which is a standard surgical treatment modality for bone transport whereby an osteotomy is performed far from the defect site and the transport segment thus created is moved, approximately one millimeter per day, under the constant tension by wires attached to a cumbersome external fixator, until the defect is bridged and the segment can be docked onto the other side of the defect), and orthobiologics ( e.g. bone graft or bone graft replacement, BMP's). Surgical treatment modalities involving auto/allografting and bone regeneration via distraction osteogenesis are complex, time intensive procedures of inherently high risk due to vagaries of organ donation, in the case of allografts, and the complexity of soft and hard tissue salvage during the process of distraction osteogenesis. In addition, orthobiologics are costly and their dosage regimes as well as efficacy are currently the subject of much research [ 1 - 3 ]. Both bone grafting and bone transport procedures are complex for the surgeon as well as for the patient. Furthermore, they are susceptible to complications such as delayed union, extensive treatment time periods, infections, and insufficient mechanical function outcomes that can result in fractures. The high complication rates of these procedures exacerbate the previously mentioned difficulties associated with these treatment modalities, from the perspective of the surgeon as well as that of the patient. In short, the inherent risk of complications increases the need for patient compliance and clinical follow-up. Despite the effort associated with these procedures, their results are often less than satisfactory. Hence, the complexity and shortcomings of current state-of-the-art surgical procedures have provided impetus to develop a new treatment modality that provides a relatively straightforward, single step procedure with a high probability of success for the bridging of massive bone defects in long bones. The procedure is straightforward and can be implemented in operating rooms across the world without the need for high-tech equipment or expensive orthobiologics. The purpose of this manuscript is to describe the novel procedure. Technical innovation – methodology and proof of feasibility The proposed procedure is applicable for clinical scenarios including tumor resection as well as debridement after an infection or non-union. In the case of reconstruction after tumor resection (Fig. 1A ), a transport-segment of the diaphysis adjacent to the defect is pealed out of the surrounding periosteum (Fig 1B,C ), an osteotomy is performed and the transport-segment is moved out of the periosteal sleeve and docked to the other side of the defect (Fig. 1D ). The periosteal sleeve is then closed like a tube surrounding the newly created defect. Either an internal fixation device such as an intramedullary nail, a plate or an internal fixator or an external fixator can be used to provide stabilization through the healing and regeneration phase. Figure 1 Schematic diagram showing the concept for the new surgical procedure. The proposed procedure depends to a large degree on bone's inherent healing strategies. Bone is a remarkably resilient tissue capable of adaptation to the most extreme biological and mechanical environments; this capacity for self-regeneration without scarring is based on bone's endogenous healing strategies. First, bone remodels itself through osteoclastic resorption and osteoblastic matrix apposition; by constantly reweaving itself, the structure is dynamic and optimal for prevailing mechanical function. Furthermore, the natural healing cascade of bone after trauma recapitulates embryonic endochondral ossification. Hence, modeling, growth and remodeling confer a means to regenerate functional tissue at any time in the life cycle of a bone. The "raw materials" necessary to replace bone are located in the environment or produced by the cells that do the work of regeneration, i.e. osteoclasts and osteoblasts. In the case of regeneration of bone in defects, further potentially key constituents to the formation of a functional regenerate in situ include a patent blood supply, chemical gradients of morphogens and/or cytokines, a template onto which the cells can anchor themselves during the rebuilding process (e.g. graft or a scaffold), and biophysical stimuli such as fluid flow and/or cell level strains. The proposed procedure essentially replaces the defect site in a pathological zone with a defect site in a healthy bed of tissue and provides for progenitor cells through the surrounding, healthy periosteum as well as many of the other key constituents for successful tissue regeneration, as defined above. A clinical case described below demonstrates the osteoinductive potential of the periosteum and serves is a proof of feasibility for the proposed procedure. An 11-year old male presented with a low grade surface osteosarcoma of the tibia. After resection of the tumor, the fibula was resected for transfer and the surrounding periosteum was left behind to serve as an osteo-inductive and -conductive sleeve (Figure 2A ). Already 3 weeks after the procedure, bone regenerate is visible within this sleeve (Figure 2A ). Impressive remodeling of the fibula is also evident in follow up radiographs and includes extensive remodeling of the intramedullary canal by three months post procedure (Figure 2B and 2C ). Based on this clinical case as well as one author's previous experience with an in vivo segmental defect in an ovine model [ 4 , 5 ], the osteogenic potential of the periosteum as a source of progenitor cells and as a "membrane" or boundary template for guided bone generation is demonstrated. Taking this one step further, the corresponding author conceived of the idea to exploit the potential of the healthy periosteum by moving the defect site from a pathological zone to a healthy one and then providing sufficient mechanical stability to let bone's endogenous healing capacity regenerate functional tissue within the new defect zone. Figure 2 11 year male with malignant tumor. Fibula-pro-tibia following local resection. Cortical regeneration from periosteum. Performed at Mt. Sinai Medical Center, NYC, 2000. A : 3 weeks post-operative radiograph, B : 6 weeks post-operative, C : 3 months post-operative, D : approximately 6 months post-operative. Discussion The proposed procedure is novel in that it introduces for the first time the possibility to bridge a massive defect in a long bone using a single stage procedure. Furthermore, the proposed procedure has several advantages over the current standard of care including ease of implementation, lack of requirement for specialized implants (intramedullary nails are standard inventory for surgical oncology and trauma departments) or costly orthobiologics, and rapid patient mobilization. This makes the proposed procedure a viable and potentially preferable alternative to the current standard treatment modalities, particularly areas of the world where few surgeons are trained for procedures such as distraction osteogenesis ( e.g. the Ilizarov procedure) as well as where surgeons have less access to expensive, complex devices and orthobiologics. Conclusion In summary, the authors propose a new procedure which obviates the need for several surgical procedures, reduces the risk for complications, reduces the time frame for the treatment and is much more comfortable for and requires less compliance of the patient. This novel, one stage procedure exploits the osteogenetic potential of the periosteum for bone formation to bridge the defect with concomitant bone transport and does not require the use of expensive hardware or orthobiologics. Competing interests The author(s) declare that they have no competing interests. Authors' contributions UK conceived of the technical innovation, organized its design and coordination, and drafted the manuscript. DSS was the senior surgeon in the clinical case showing feasibility. Both authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC550665.xml
520816	Characterization of a novel large deletion and single point mutations in the BRCA1 gene in a Greek cohort of families with suspected hereditary breast cancer	Background Germline mutations in BRCA1 and BRCA2 predispose to breast and ovarian cancer. A multitude of mutations have been described and are found to be scattered throughout these two large genes. We describe analysis of BRCA1 in 25 individuals from 18 families from a Greek cohort. Methods The approach used is based on dHPLC mutation screening of the BRCA1 gene, followed by sequencing of fragments suspected to carry a mutation including intron – exon boundaries. In patients with a strong family history but for whom no mutations were detected, analysis was extended to exons 10 and 11 of the BRCA2 gene, followed by MLPA analysis for screening for large genomic rearrangements. Results A pathogenic mutation in BRCA1 was identified in 5/18 (27.7 %) families, where four distinct mutations have been observed. Single base putative pathogenic mutations were identified by dHPLC and confirmed by sequence analysis in 4 families: 5382insC (in two families), G1738R, and 5586G > A (in one family each). In addition, 18 unclassified variants and silent polymorphisms were detected including a novel silent polymorphism in exon 11 of the BRCA1 gene. Finally, MLPA revealed deletion of exon 20 of the BRCA1 gene in one family, a deletion that encompasses 3.2 kb of the gene starting 21 bases into exon 20 and extending 3.2 kb into intron 20 and leads to skipping of the entire exon 20. The 3' breakpoint lies within an Alu Sp repeat but there are no recognizable repeat motifs at the 5' breakpoint implicating a mechanism different to Alu -mediated recombination, responsible for the majority of rearrangements in the BRCA1 gene. Conclusions We conclude that a combination of techniques capable of detecting both single base mutations and small insertions / deletions and large genomic rearrangements is necessary in order to accurately analyze the BRCA1 gene in patients at high risk of carrying a germline mutation as determined by their family history. Furthermore, our results suggest that in those families with strong evidence of linkage to the BRCA1 locus in whom no point mutation has been identified re-examination should be carried out searching specifically for genomic rearrangements.	Background Germ line mutations in the BRCA1 and BRCA2 genes predispose individuals to breast and ovarian cancer. The lifetime risk of breast cancer in female carriers of a BRCA1 mutation is 60–80% while that of ovarian cancer is 20–40%. The median age of diagnosis of breast cancer is 42 years, i.e. 20 years earlier than the median of unselected women in the U.S.A. and Western Europe [ 1 ]. BRCA1 is a large gene with 22 coding exons encoding a 220 kD protein [ 2 ] that functions in maintaining genomic integrity and in transcriptional regulation [ 3 , 4 ]. A multitude of mutations scattered throughout the 5592 bp coding sequence have been described. In particular, germ line mutations in BRCA1 have been identified in 15–20% of women with a family history of breast cancer and 60–80% of women with a family history of breast and ovarian cancer [ 5 , 6 ]. The percentage of mutations identified is strongly dependent on the population studied, with strong founder effects evident in some populations [ 7 - 9 ]. The vast majority of mutations described to date are point mutations and small insertions and deletions . Such mutations are detected by PCR-based screening methods such as the Protein Truncation Test (PTT), Single Strand Conformational Polymorphism (SSCP), Denaturing Gradient Gel Electrophoresis (DGGE), heteroduplex analysis (HA) and more recently denaturing high performance liquid chromatography (dHPLC) with varying degrees of sensitivity for each method. Direct DNA sequencing is used in order to confirm and characterize mutations detected by any of these approaches [ 10 , 11 ]. While in the early studies for mutation detection only single point or small insertion/deletion mutations were screened for, recent studies have shown that genomic rearrangements are also a common type of mutation in the two genes accounting for 10–30 % of all mutations identified in some populations [ 12 - 16 ]. In this study we document and extend previous work suggesting the necessity of screening for large genomic rearrangements in a complete program for mutation detection of the BRCA1 gene by characterizing a deletion encompassing 3.2 kb of the gene including exon 20. In addition, we add more evidence supporting pathogenicity of a previously described variant, G1738R, which seems to be specific to the Greek population [ 17 ]. Methods Patients BRCA1 patients and their families were referred through the Oncology Departments of Hygeia Hospital and other hospitals throughout Greece. Patients were included on the basis of affected family members, types of cancer present in the family and the age at diagnosis of breast cancer in the proband. The families were subdivided into high risk when multiple cases of breast and ovarian cancer were diagnosed, medium risk if there were only 2–3 cases of breast cancer, and low risk in isolated cases of breast cancer with diagnosis before the age of 40 years. The study population consisted therefore of 12 high risk families, 2 with medium risk, and 4 families at low risk as determined by their cancer history. Ethical approval was obtained from the hospital's advisory committee and all patients signed informed consent. Screening has been completed in 25 individuals from 18 families. Testing was initially carried out on DNA from an affected family member and upon detection of an inactivating mutation the rest of the family members were directly tested for this mutation. DNA and RNA isolation Genomic DNA and RNA were purified from peripheral blood leukocytes using standard extraction protocols. PCR amplification The complete coding sequence of BRCA1 including splice junctions was amplified by PCR. Similarly, exons 10 and 11 of the BRCA2 gene were amplified in 3 of the patients. Primers used have been chosen from the BIC database . In addition primers: mBRCAF: GAG TTT GTG TGT GAA CGG ACA CTG and mBRCAR: GTG CCA AGG GTG AAT GAT GAA AGC, designed during the course of this study, were used for amplification of cDNA from the patients found to carry a deletion of exon 20 of the BRCA1 gene. Reactions of 50 μl were heated on a PTC-200 MJ Research Thermocycler (MJ Research Inc., USA) at 95°C for 5 min then cycled 35 times of denaturation at 95°C for 40 sec, annealing at the appropriate temperature for 30 sec and extension at 72°C for 30 sec, followed by a final extension step at 72°C for 6 min. Reaction mixture was 20 mM TrisHCl (pH 8.4), 50 mM KCl, 1.5 mM MgCl 2 , 200 μM each dNTP, 1.5 U Taq DNA polymerase (Invitrogen, Netherlands) or 2.5 U Optimase polymerase (Transgenomic, Inc., USA) and 12.5 pmol of each primer. dHPLC analysis The WAVE DNA Fragment Analysis System (Transgenomic, Inc., USA) and associated WAVE-Maker™ software were used as previously described [ 18 ]. Sequence analysis Purification of the PCR products was performed using the Concert Rapid PCR purification or gel extraction system kits (Marligen Biosciences INC, U.S.A.). Automated cycle sequencing for both strands was performed with the ABI Prism ® 310 Genetic Analyzer using the Big-Dye Terminator Cycle Sequencing Kit. Sequences obtained were aligned, using Sequencher ® PC software, with normal sequences from Genbank ( BRCA1 : L78833, BRCA2: U43746) and examined for the presence of mutations. All nucleotide numbers refer to the wild-type cDNA sequence of BRCA1 as reported in GenBank (accession number U14680). Multiplex Ligation – dependent PCR Amplification (MLPA) MLPA was carried out using the P002_BRCA1 kit (MRC-Holland, Netherlands) as instructed by the manufacturer. Fragment analysis was carried out on ABI Prism ® 310 Genetic Analyzer (Applied Biosystems, USA) using TAMRA-500 (Applied Biosystems, USA) as size standard. A peak pattern of 34 peaks ranging in size from 127 to 454 nt is obtained [ 19 ]. Long PCR The deletion in BRCA1 exon 20 was confirmed by long PCR using the GeneAmp XL PCR System (Applied Biosystems, U.S.A.) according to the manufacturer's instructions. PCR was carried out in 50 μl reactions consisting of 1 × buffer II (supplied with the enzyme), 12 pmole each primer, 0.2 mM each dNTP and 2.4 U rTth DNA polymerase, XL (Applied Biosystems, U.S.A.). After 2.5 min denaturation at 95°C PCR was carried out for 19 cycles of 95°C for 30 sec, 58°C for 30 sec and 68°C for 8 min followed by 15 cycles of 95°C for 30 sec, 58°C for 30 sec and 68°C for 4 min with a time increment of 10 sec per cycle. A final extension step was carried out at 72°C for 10 min. PCR products were separated by agarose gel electrophoresis and visualized by EtBr staining. RT-PCR Total RNA was extracted from whole blood of patients from family D using Trizol (Life Technologies, USA) according to the manufacturer's instructions. First strand synthesis was performed by denaturing approximately 500 – 1000 ng total RNA, random hexamers (5 μM final concentration) for 4 min at 70°C, followed by snap freezing on ice and addition of dNTPs (0.5 mM final concentration), 1 U/μl recombinant RNase inhibitor (Invitrogen, Netherlands) and 200 U MMLV reverse transcriptase (Invitrogen, Netherlands). The mixture was incubated at 37°C for 1 hour followed by denaturation of the enzymes at 95°C for 5 min. 4 μl of cDNA were used for subsequent PCR amplification. Results A total of 18 families, 12 of which were at high risk of having hereditary breast cancer, have been examined for mutations at the BRCA1 locus. A pathogenic mutation was identified in five families, where four distinct mutations have been observed. In addition, 18 polymorphisms, including a novel silent polymorphism in exon 11 of the BRCA1 gene, have been detected (Table 1 ). Furthermore, in 3 of the families exons 10 and 11 of BRCA2 were analyzed. Table 1 summarizes the results of single nucleotide variants detected in this study. All of the variants have been identified by dHPLC and characterized by sequencing. In family A there were four cases of breast cancer affecting four successive generations. The proband was diagnosed with breast cancer at the age of 45. The most frequently occurring mutation, 5382insC in exon 20, was identified. Analysis of the proband's daughter in whom cancer developed at the age of 38 revealed that she also carried the mutation. The same mutation was also identified in family E where breast cancer was diagnosed in 3 members in three successive generations at 42, 37 and 41 years (data not shown). In family B there were seven cases of breast cancer and one case of colorectal cancer at age 50. Mutation analysis at the BRCA1 locus revealed a missense mutation 5331 G > A. The mutation results in substitution of a Glycine by an Arginine at codon 1738. The mutation was not detected in the patient's unaffected sister. Unfortunately, no other family members were available for analysis. In family C there were 4 cases of breast cancer in addition to cancer of the larynx, ovaries, lung and the genitals. A single base substitution, G > A at nucleotide 5586 was identified. This mutation causes a splicing defect resulting in a protein lacking exon 23. Unfortunately, no DNA was available from other family members in order to test the correlation of this mutation with the other tumor types. In Family D (Figure 1a ) there were 5 cases of breast cancer with age at onset ranging from 29 to 50 years. In addition there was one individual who had CRC. Sequencing of the complete coding region of BRCA1, failed to reveal a mutation. This prompted the analysis of other genes, namely BRCA2 and p53 in order to try and characterize the phenotype in this family, but no mutation was identified. For this reason, we decided to use the MLPA technique [ 19 ] to screen the proband for genomic rearrangements which have been shown to be responsible for a large proportion of BRCA1 mutations. MLPA analysis of the proband (IV:10 in figure 1a ) revealed deletion of exon 20 of the BRCA1 gene. This was confirmed by Long PCR (Figure 1b ) and RT – PCR. Sequencing of PCR products generated by RT-PCR confirmed absence of the entire exon 20 in the mRNA of the patient. Subsequently, a number of restriction endonucleases were employed in order to narrow down the deletion breakpoints and facilitate their characterization. A Sma I-generated irregular fragment was finally isolated from agarose gel and sequenced revealing a deletion of 3.2 kb starting 21 bases into exon 20 and extending 3.2 kb into intron 20 (Figure 1c ). The 3' breakpoint lies within an Alu Sp repeat but there are no recognizable repeat motifs at the 5' breakpoint. The deletion was also found in 4 relatives of the proband (Figure 1a ) who were tested, two of whom had breast cancer. The other two relatives had not yet developed cancer at the ages of 36 and 55 at the time of testing and are therefore assumed to be pre – symptomatic carriers. Finally, the proband's sister, who at the time of testing had not developed cancer at the age of 33 was found not to carry the mutation. Discussion A total of 18 breast cancer families have been examined for mutations at the BRCA1 locus. In three of the families analysis was extended to exons 10 and 11 of the BRCA2 gene. A pathogenic mutation in BRCA1 was identified in 5/18 (27.7 %) families, where four distinct mutations have been observed. In addition, 18 polymorphisms have been found to be present in more than one family. In family B a missense mutation, 5331 G > A, was identified. The exact effect of this single amino acid change, G1738R, on the protein function is unclear. The altered glycine is located on the surface of the coil structure of the BRCT linker region and the mutation therefore may disrupt the interface or affect protein interaction [ 20 ]. This mutation has been previously described in 4 unrelated Greek patients [ 21 ]. An alternative mutation affecting the same amino acid, G1738E, has been shown to result in loss of function of the protein in vitro [ 22 , 23 ]. Based on this information, the family history of the individual described here and absence of the mutation in the proband's unaffected sister we hypothesize that the mutation is pathogenic although additional data are needed. Our results indicate, as has been documented by others that family history is the major determinant of the risk of breast cancer. As can be seen in Table 1 in all families where a pathogenic mutation was identified there were at least 3 cases of breast cancer. In addition, the extremely high risk suggested by the family history in some cases prompted a more intense analysis of further genes and approaches in an attempt to characterize the underling reason for such a family history. This was the case in family D where sequencing of the complete coding region of BRCA1, failed to reveal a mutation. This prompted the analysis of other genes, namely BRCA2 and p53 , but without finding a mutation. For this reason, we decided to use the MLPA technique [ 19 ] to screen the proband for genomic rearrangements which have been shown to be responsible for a large proportion of BRCA1 mutations [ 12 - 16 ]. This led to the identification of deletion of an entire exon of the BRCA1 gene, namely exon 20. Deletion of exon 20 has been previously described in an Italian family [ 12 ]. However, the authors of that report have localized the breakpoints of the deletion in the two flanking introns [ 12 ]. In the deletion described here the 5' breakpoint is localized inside exon 20 (21 bp upstream of the splice donor site) while the 3' breakpoint is located 3.2 kb into intron 20 within an Alu Sp repeat. The deletion described here, therefore, is different to the majority of rearrangements described so far for the BRCA1 gene, since they have been shown to result from homologous recombination of Alu repeats [ 24 , 25 ]. In this case, the 5' breakpoint does not correlate with any recognizable repeat motifs suggesting a repeat – independent recombination mechanism at play. In 2/5 pedigrees in whom a pathogenic mutation was identified there was one case each of colorectal cancer (CRC). In particular, in family D (Figure 1a ) the disease seems to have originated from a CRC patient. Data on the correlation between BRCA1 mutations and the risk of CRC are not conclusive. Two recent studies carried out on Ashkenazi Jewish patients suggest that there is no correlation [ 26 , 27 ]. Other reports carried out on more diverse populations suggest 2- to 3-fold elevated risks of CRC among first-degree relatives of BRCA1 mutation carriers [ 28 , 29 ]. Unfortunately, there was no DNA available for analysis from the two CRC patients in the two families. A further interesting point observed in this study is the identification of two BRCA1 unaffected mutation carriers at the ages of 39 and 55. The penetrance and age of onset of disease in BRCA1 mutation carriers is variable. Various reports have suggested the existence of modifying genetic and environmental factors on the penetrance of BRCA1 and BRCA2 mutation carriers [ 30 , 31 ]. In this respect it is interesting to examine this large pedigree for such modifying factors. Conclusions To our knowledge this is the first report of a genomic rearrangement identified as the underlying mutation of BRCA1 in a Greek family. Although our sample group is quite small identification of 1 genomic rearrangement in 5 mutations detected, suggests that in yet another population this type of mutations contribute to the BRCA1 mutation spectrum. This therefore warrants use of a combination of techniques capable of identifying both single base mutations in addition to large genomic rearrangements. In this respect, we have found that use of dHPLC for single base mutations and MLPA for large genomic rearrangements is a reliable combination for use as an initial screening step followed by sequencing for characterization of the mutations identified. Furthermore, our results suggest that in those families with strong evidence of linkage to the BRCA1 locus in whom no point mutation has been identified re-examination should be carried out searching specifically for genomic rearrangements. List of abbreviations dHPLC denaturing High Performance Liquid Chromatography MLPA Multiplex Ligation – dependent PCR Amplification PCR Polymerase Chain Reaction RT-PCR Reverse Transcription Polymerase Chain Reaction CRC Colorectal Cancer Competing interests None declared. Authors contributions IB carried out mutation detection by dHPLC and sequencing for the patients analyzed. AA carried out MLPA, the molecular characterization of the deletion, and drafted the manuscript. MM contributed in the molecular studies and preparation of the manuscript and figures, ER, SL, AP, VG, AK NP, PK and DY provided the patient material, diagnosis and management,, KK and AH carried out molecular genetic analysis of the p53 gene, GN conceived of the study, and participated in its design and coordination. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC520816.xml
555537	Use of biological based therapy in patients with cardiovascular diseases in a university-hospital in New York City	Background The use of complementary and alternative products including Biological Based Therapy (BBT) has increased among patients with various medical illnesses and conditions. The studies assessing the prevalence of BBT use among patients with cardiovascular diseases are limited. Therefore, an evaluation of BBT in this patient population would be beneficial. This was a survey designed to determine the effects of demographics on the use of Biological Based Therapy (BBT) in patients with cardiovascular diseases. The objective of this study was to determine the effect of the education level on the use of BBT in cardiovascular patients. This survey also assessed the perceptions of users regarding the safety/efficacy of BBT, types of BBT used and potential BBT-drug interactions. Method The survey instrument was designed to assess the findings. Patients were interviewed from February 2001 to December 2002. 198 inpatients with cardiovascular diseases (94 BBT users and 104 non-users) in a university hospital were included in the study. Results Users had a significantly higher level of education than non-users (college graduate: 28 [30%] versus 12 [12%], p = 0.003). Top 10 BBT products used were vitamin E [41(43.6%)], vitamin C [30(31.9%)], multivitamins [24(25.5%)], calcium [19(20.2%)], vitamin B complex [17(18.1%)], fish oil [12(12.8%)], coenzyme Q10 [11(11.7%)], glucosamine [10(10.6%)], magnesium [8(8.5%)] and vitamin D [6(6.4%)]. Sixty percent of users' physicians knew of the BBT use. Compared to non-users, users believed BBT to be safer (p < 0.001) and more effective (p < 0.001) than prescription drugs. Forty-two potential drug-BBT interactions were identified. Conclusion Incidence of use of BBT in cardiovascular patients is high (47.5%), as is the risk of potential drug interaction. Health care providers need to monitor BBT use in patients with cardiovascular diseases.	Background The use of complementary and alternative medicine (CAM), defined by the United States National Center for CAM as a group of diverse medical and health care systems, practices, and products that are not presently considered to be a part of conventional medicine, has grown tremendously in the United States [ 1 - 4 ]. A recent national survey reported that four out of every ten Americans used at least one form of CAM, and one out of five used prescription medications together with CAM [ 3 ]. The prevalence is even higher in patients with chronic medical problems (for example, 28 to 90% in patients with arthritis, 11–56% in those with cancer, 60% in patients with asthma and 67.8% in patients with human immunodeficiency virus) [ 5 - 11 ]. Biological based therapies (BBT) is an important type of CAM and is defined by the National Center for CAM as use of substances found in nature, such as herbs, foods, and vitamins. BBT is the second most commonly utilized CAM, with the first being prayer therapy [ 2 ]. Similar to other chronic medical conditions, patients suffering from a variety of cardiovascular diseases including coronary heart disease, congestive heart failure, stroke, arrhythmia and congenital cardiovascular defects, may also be looking to CAM to prevent or treat their illnesses. This is particularly likely since a number of BBT products including vitamin E, vitamin C, beta-carotene, fish oils, and coenzyme Q10 have been evaluated for prevention and/or treatment of cardiovascular diseases [ 12 - 25 ]. Despite a wide array of available BBT for cardiovascular conditions, studies evaluating the prevalence of usage of these agents are limited [ 21 - 25 ]. As CAM, in general, has become widely accessible to the public, and BBT may be purchased in pharmacies, health food stores and supermarkets, it is difficult to control patient usage of these products. In addition, the likelihood for adverse effects and interactions between conventional therapies and BBT places patients using such products at an increased risk of adverse drug events. It is, therefore, important to examine patient usage so as to advise and monitor them properly. Among the studies conducted to-date that included patients with a broad spectrum of cardiovascular diseases, few focused only on BBT. The studies examined different factors that may determine BBT use, but none of them examined the potential for side effects and drug interactions with other prescription and non-prescription medications that the patients were utilizing. Since patients with cardiovascular conditions consume many prescription medications with narrow therapeutic indexes and extensive drug interaction profiles [ 26 , 27 ], it is important to look at the prevalence of use and potential drug interactions in a cohort of cardiovascular patients. The primary objective of this study was to determine the effect of the education level on the use of BBT in cardiovascular patients. The study also investigated the attitudes and beliefs towards BBT by patients with cardiovascular diseases. In addition, patient perceptions regarding the safety and efficacy of BBT, common BBT used and a list of potential BBT-drug interactions were reviewed. Methods This is a cross-sectional, descriptive study utilizing structured interviews to assess the level of education, usage, beliefs and attitudes towards BBT among inpatients with cardiovascular diseases. The study was conducted in the Cardiac Care Center at the Mount Sinai Hospital, New York, USA from February 2001 to December 2002. Participants provided informed consent and were interviewed by one of the investigators. To maintain consistency of the interview and to prevent interviewer bias, a scripted letter was drafted for the investigators to invite the patients to participate in the study and to explain the process of the study. After a patient was enrolled, an investigator read the survey questions verbatim to the patient and tried not to elaborate whenever possible. Patients were included in the study if they had at least one of the following diagnoses: cardiovascular disease(s) including congestive heart failure, coronary heart disease, thromboembolic diseases, valvular heart disease, arrhythmia, vascular aneurysm, peripheral vascular disease, pulmonary hypertension, congenital heart disease and post heart transplant. Additional inclusion criteria included being 18 years of age or older, English speaking, no documented cognitive deficits precluding the patient from understanding the interviewer, and willingness to provide an informed consent. Prior to patient contact, the attending physicians of the eligible patients were contacted and informed about the study. If the attending physicians chose not to have their patients participate in the study, the patients were not included. Upon agreement of the physicians, subjects were invited to participate and were asked to sign an informed consent at their convenience prior to being interviewed. This study was approved by the Institutional Review Boards of Mount Sinai Hospital and Long Island University. Biological based therapy survey Utilizing a structured instrument (see Additional file 1 ), eligible subjects were interviewed by one of the investigators during their stay at the hospital. The BBT survey was modified and adapted from a previously published survey [ 6 ]. The participants could choose to answer or not to answer any question at their discretion, and could discontinue their participation in the study at any time during the interview. During the interview, demographic data, including age, gender, race, marital status, level of education, annual income, and working status, were collected. Additionally, history of cardiovascular and other medical conditions, and medications utilized were recorded. The definition of BBT in this study was similar to that defined by the United States National Center for CAM, which included all herbal supplements, vitamins and mineral supplements. The patients' attitudes and beliefs towards BBT were assessed by asking them about their perceived safety and efficacy of BBT. The side effects and potential drug/food interactions listed by the patients were compared against those listed in the MicroMedex ® Database [ 28 ]. The patients' assessments of benefits of BBT as compared with conventional medicine were recorded. Additionally, the participants were asked whether they reported the use of BBT to their physicians, pharmacists, or other healthcare professionals. A review of patients' medical records was performed to collect data about patients' cardiovascular diseases and to confirm medications used. Although, the identity of the participants in this research study was kept confidential, patients were notified in the informed consent process that if potential BBT-prescription medication interactions were identified, their physicians would be notified. Statistical analysis For this study to have an 80 percent power to detect a 20 percent clinically significant difference in determining factors of BBT use such as education level between the users and non-users of BBT, and establishing a p value of < 0.05 as the level of statistical significance, approximately 200 patients (100 patients in each group) needed to be enrolled. For demographic parameters, continuous variables were compared between the two groups using Students' t-test and categorical variables were compared using chi-square. Attitudes and beliefs regarding the safety and efficacy of BBT were compared between the two groups using a chi-square test for categorical data and Students' t test for Likert-type scale questions. The BBT products used by cardiovascular patients were recorded and the likelihood of potential drug interactions between BBT and other medications the patients were taking were described. Statistical analyses were conducted using Statistical Product and Service Solutions program (SPSS ® for Windows, Rel. 10.01 1999). Results A total of 200 patients who were admitted to the Cardiac Care Center at Mount Sinai Hospital were enrolled into the study. All the physicians approached agreed to have their patients participate in the survey. Two of the patients were, eventually, excluded from data analysis due to incomplete survey data. Of the remaining 198 patients, 94 (47.5%) reported BBT use at some point in their lifetime and 84 (42%) reported using such products during the immediate 12 months prior to the survey. Of the 94 patients who used BBT, 32% reported using the products all the time. Demographic characteristics The demographic and socioeconomic characteristics of the total sample are presented in Table 1 . The mean age of surveyed participants was 60.0 ± 16.5 years, and their ages were normally distributed ranging from 19 to 102 years of age. The overall sample consisted of 118 (59.6%) men and 80 (40.4%) women. Table 1 Demographic Characteristics of Study Subjects Users (%) Nonusers (%) p value Total 1 N = 198 94 (47.5) 104(52.5) Age 2 61.4 ± 16.7 58.7 ± 16.3 NSS Gender NSS Male 51(54) 67(64) Female 43(46) 37(36) Race NSS White 53(56) 46(44) Black 19(20) 27(26) Hispanic 12(13) 18(17) Other 3 10(11) 13(7) Education 0.0006 < High school 12(13) 28(27) High school 23(24) 42(40) Some college 18(19) 12(12) College Graduate 28(30) 12(12) Graduate School 13(14) 8(8) Annual Household Income NSS < $10,000 17(27) 19(30) $10,000–$30,000 12(19) 15(23) $30,000–$50,000 8(13) 14(22) $50,000–$75,000 15(23) 7(11) $75,000–$100,000 4(6) 4(6) > $100,000 5(13) 5(8) 1 lifetime use of BBT; 2 mean age ± standard deviation (range); 3 Asian/Pacific Islander/Indian Overall, education level significantly influenced the use of BBT, p = 0.0006. Among users, more patients had college degrees (28 [30%]) as compared with nonusers, (12 [12%], p = 0.003). In contrast, 42 (40%) nonusers finished high school versus 23 (24%) of users, p = 0.023. There were no significant differences between users and nonusers in other demographic variables. Cardiovascular diseases, other medical conditions and medications used The distribution of cardiovascular and noncardiovascular diseases reported by the sample is presented in Table 2 . The mean number of disease diagnoses carried by users and nonusers were 4 (range 1 – 10) and 3 (range 1 – 8), respectively, p = 0.723, while the mean number of cardiovascular diagnoses for users versus nonusers were 3 (range 1 – 6) and 3 (range 1 – 5), respectively, p = 0.134. The most common cardiovascular conditions diagnosed in users and nonusers were hypertension, 56 (24%) and 65 (27%), followed closely by coronary heart disease, 45 (19%) and 54 (18%), respectively. Distribution of other cardiovascular diseases between users and nonusers was similar as well. Similarly, the mean number of noncardiovascular diagnoses reported by users and nonusers of BBT were not significantly different: 2 (range 1 – 5) and 1 (range 1 – 5), respectively, p = 0.288. Table 2 Distribution of Cardiovascular and Noncardiovascular Diseases Users Nonusers p value Average number of diseases 4 3 NSS Range 1 – 10 1 – 8 Average number of cardiovascular diseases 3 3 NSS Arrhythmia 25 32 Congestive Heart Failure 38 52 Coronary Heart Disease 45 54 Hyperlipidemia 31 43 Hypertension 56 65 Other 1 12 14 Post-heart Transplant Recipient 1 2 Thromboembolic Disease 13 16 Valvular Heart Disease 11 14 Average number of noncardiovascular diseases 2 1 NSS Arthritis 25 17 Cancer 3 4 Diabetes Mellitis 33 33 Gastrointestinal Disease 19 13 Hypothyroidism 5 11 Nervous System Disorder 8 9 Ocular Disease 6 6 Other 2 30 13 Pulmonary Disease 12 13 Renal Disease 7 12 1 aneurysm, peripheral vascular disease, infective endocarditis, pulmonary hypertension, congenital heart disease; 2 allergic rhinitis, scleroderma, systemic lupus erythematosis, Raynaud's disease, benign prostatic disease, osteoporosis The mean number of traditional prescription and nonprescription medications taken by the users of BBT (7 [range 1–15]) and nonusers (7 [range 0–18]) was not significantly different, p = 0.445. The most common cardiovascular medications prescribed for both users and nonusers were diuretics (71%, 73%), followed by aspirin (56%, 43%) and beta-blockers (49%, 50%). The cardiovascular and noncardiovascular medications taken by both groups were not different. Types and patterns of BBT use The BBT products utilized by cardiovascular patients are presented in Table 3 . The mean number of BBT products utilized by cardiovascular patients was two. Vitamin E (41, [44%]) was the most commonly utilized BBT, followed by vitamin C, (30 [32%]). The prevalence of the remaining BBT is summarized in Table 3 . Table 3 All Biological Based Therapies Utilized by Cardiovascular Patients (at any time) Product Number Reasons for Use (per patient) AA#5 (anti-arthritis) 1 treat arthritis Acidophillus 1 health Aloe vera 4 headache, stomach gas, skin pigmentation Atomic Drops 1 treat headache Bee Pollen 1 prevent cold Beta carotene 5 maintain good health, energy, improve heart contraction Bioflavinoid 1 bone health instructed by chiropractor Calcium 19 supplement, prevent osteoporosis, improve heart function Chamomile 2 stomach ache, improve heart condition Chromium picolante 2 supplement for heart condition, muscle strength Coenzyme Q 10 11 improve heart contraction, supplement to diet Dexatrim 1 weight lost DHEA 1 1 supplement Echinacea 5 prevent or treat cold, flu, stay healthy, boost immune system Fish Oil 12 decrease cholesterol, maintain circulation and good health, scleroderma Folic Acid 3 supplement, help with heart condition, sickle cell anemia Garlic 5 decrease cholesterol, help maintain good health Ginkgo biloba 5 antioxidant, enhance memory, energy Ginseng 4 increase energy, stamina, virility Glucosamine/chondroitin 10 treat arthritis, decrease joint pain Golden seal 1 body cleaner Grapeseed oil 1 preserve health Green tea 3 decrease cholesterol, improve circulation Mixed herbal tea 1 Sooth stomach upset, anxiety Insulin leaf tea 1 diabetes Iron supplement 4 anemia, increase energy Lecithin 2 improve heart condition and circulation, decrease cholesterol Alpha-Linolenic acid 1 improve heart condition Magnesium 8 supplement, antioxidant, improve metabolism, improve heart function Multivitamins 24 supplement, energy, sickle cell anemia Primrose oil 1 scleroderma Saint John's Wart 1 depression Saw palmetto 1 for prostate health Selenium 4 antioxidant, supplement, improve heart condition Strong bark 1 help with heart condition Valerian 4 decrease anxiety, improve sleep, to decrease blood pressure Vitamin B complex 17 supplement, improve heart condition, decrease leg cramps, energy Vitamin B12 2 supplement, anemia Vitamin C 30 antioxidant, supplement, help with heart condition, improve circulation, strengthen immune system, prevent or treat cold Vitamin E 41 antioxidant, supplement, help with heart condition, increase energy, decrease cholesterol, improve circulation, thin blood, treat hypertension Yohimbine 1 increase energy, stamina, virility Zinc 4 supplement 1 dehydroepiandrosterone The patients reported that their physicians were aware of their BBT use in 60% of the instances and pharmacists were aware in 32% of the cases. Only 33% of users reported that they were asked about BBT use during a history/physical examination by a health care professional. The patients were not surveyed about their pharmacists' assessments of BBT use. Perceived benefits and attitudes towards BBT A greater percentage of users (74.5%) of BBT reported that they believed these products to be safe substances as compared with nonusers (26.2%), p < 0.001. Likewise, significantly more users believed that BBT was effective (70.2%) compared to nonusers (30.1%), p < 0.001. More nonusers (72%) than users (45%) did not know whether BBT products work better, as well as, or worse than traditional medications, p < 0.001. More users than nonusers believed that BBT works as well as traditional medications (30.9% versus 9.7%, p < 0.001). Concerning adverse effects, more users (44.7%) of BBT than nonusers (17.5%) believed that BBT causes fewer side effects than traditional medications, p < 0.001. At the same time, 32.2% of users and 62.1% of nonusers did not know whether BBT causes more or fewer side effects than traditional prescription medications, p < 0.001. Potential drug-BBT interactions Examination of patients' prescription and nonprescription medication profiles and BBT utilized, revealed 42 potential drug-BBT interactions. The onset of the interaction and the degree of severity were classified according to that used by the MicroMedex HealthCare Series Integraded Index [ 28 ] and the published literature [ 29 - 43 ]. Suspected or potential interactions were communicated to the patients and their primary physicians. The most common interaction identified was coadministration of aspirin and vitamin E (16 cases) that could potentially result in an increased risk of bleeding due to additive inhibition of platelet aggregation [ 29 - 31 ]. Similar effects may result from coadministration of clopidogrel and vitamin E; this potential interaction was recognized in five cases [ 31 ]. Five instances of potential interaction between warfarin and vitamin E were identified and close monitoring of the International Normalized Ratio (INR) was recommended [ 32 - 34 ]. A list of other potential drug-BBT interactions that were identified is presented in Table 4 . Table 4 Potential Drug-Biological Based Therapies Interactions & Management Medication Alternative Pharmacotherapy Product Potential Interaction and Management Number of Patients Onset 28 * Level of Severity 28 Warfarin Green tea Green tea may antagonize warfarin effects. Monitor INR. 1 Delayed Moderate Warfarin Coenzyme Q 10 Coenzyme Q 10 may have procoagulant effects and may decrease response to warfarin. Monitor INR. 2 Delayed Moderate Warfarin Vitamin E Vitamin E may potentiate warfarin effects. Monitor INR and signs and symptoms of bleeding. 5 Delayed Moderate Warfarin Garlic Garlic has antiplatelet effects, and risk of bleeding may be increased. Monitor INR and signs and symptoms of bleeding. 1 Delayed Major Warfarin Ginkgo biloba Ginkgo inhibits platelet aggregation, risk of bleeding may be increased. Monitor INR and signs and symptoms of bleeding 2 Delayed Major Warfarin Fish oils Concomitant use of warfarin and fish oils may increase risk of bleeding. Monitor INR and signs and symptoms of bleeding. 1 Not specified Not specified Aspirin Vitamin E Coadministration of vitamin E and aspirin may increase the risk of bleeding. Monitor for signs and symptoms of bleeding 16 Not Specified Not specified Aspirin Garlic Additive antiplatelet effects may occur with coadministration of garlic and aspirin, risk of bleeding may be increased. Monitor for signs and symptoms of bleeding 1 Delayed Moderate Aspirin Fish oils Concomitant use of aspirin and fish oils may increase risk of bleeding. Monitor for signs and symptoms of bleeding 3 Not specified Not specified Aspirin Ginko biloba Ginkgo inhibits platelet aggregation; risk of bleeding may be increased. Monitor for signs and symptoms of bleeding. 1 Delayed Major Clopidogrel Vitamin E Concomitant use of vitamin E and clopidogrel may increase the risk of bleeding. Monitor for signs and symptoms of bleeding 5 Not specified Not specified Clopidogrel Garlic Additive antiplatelet effects may occur, risk of bleeding may be increased. Monitor for signs and symptoms of bleeding 1 Delayed Major Clopidogrel Coenzyme Q Coenzyme Q 10 may have procoagulant effects, may partially antagonize antiplatelet effect of clopidogrel. 1 Not specified Not specified Insulin Insulin leaf Enhanced hypoglycemic effects. Patient was advised to discontinue insulin leaf while receiving insulin for glycemic control. 1 Delayed Moderate Paroxetine Saint John's Wort Saint John's Wort induces cytochrome P450 3A4 enzyme and has mechanism of action similar to serotonin reuptake inhibitors. Administering it concomitantly with paroxetine will enhance the toxicity of paroxetine. Patient was advised to discontinue the use of Saint John's Wort 1 Rapid Severe INR = International Normalized Ratio * Delayed onset of interactions occurs after multiple doses of both agents. Discussion Despite a wide array of available BBT for cardiovascular conditions, studies evaluating the prevalence of their usage are limited. A review of the literature at the time of study initiation (January 2001) and more recently (June 2004) identified five studies in this area, and only a few included a broad spectrum of cardiovascular disease patients [ 21 - 25 ]. Wood et al. conducted a telephone survey in 107 patients randomly selected from the "Improving Cardiovascular Outcomes in Nova Scotia" database [ 24 ]. A majority of patients (64%) in this study utilized CAM. Ackman and colleagues evaluated the patterns of BBT use by patients with congestive heart failure (CHF) [ 21 ]. Out of 180 CHF patients, 59% used vitamins and minerals and 38% used herbal or health food products. Liu and associates evaluated the prevalence of CAM used in 263 patients undergoing cardiovascular surgery [ 22 ]. Seventy-five percent of respondents utilized CAM including vitamins (53.6%), nutritional therapy (17.1%) and herbs (9.9%). Compared with non-CAM users, users were older (p = 0.027), belonged to the Caucasian racial group, (p = 0.001) and had a higher level of education (p = 0.017). Additionally, the evaluation of attitudes towards the effectiveness of CAM revealed that users were more likely to believe that CAM would work in a complimentary manner with conventional medical treatments, p < 0.05. Furthermore, more users than nonusers believed that CAM would promote general health and wellness, p < 0.05. Of the patients surveyed, only 17% reported discussing CAM with their medical doctors. Another recent study evaluated the use of CAM among 246 patients attending a cardiac clinic prior to cardiac surgery [ 23 ]. A total of 182 (80.9%) patients used CAM, and 12.9% utilized megavitamins. Another study, conducted in Canada, focused on the use of over-the-counter medications and herbal products among patients with cardiac diseases, the majority of whom were diagnosed with coronary artery disease (74%) [ 25 ]. The authors reported that 23% of the patients used multivitamin or multivitamin/mineral products. Overall, the results of these five studies confirm a high prevalence of all kinds of CAM used, including BBT among patients with cardiovascular diseases. Biological based therapy used by cardiovascular patients In our study, the lifetime prevalence of BBT use in the sample of 198 cardiovascular patients was 47.5%, which is very similar to the results reported by Eisenberg and colleagues in the general population (42%) [ 3 ]. Comparisons with other investigations of alternative medicine use in cardiovascular patients are difficult to make due to the various definitions of "alternative medicine" [ 21 - 25 ]. Some studies included items such as alternative procedures including acupuncture, while others included "pharmacotherapy" only (herbs, BBT, CAM and nutritional supplements). In addition, certain studies looked at only one cardiovascular disease while others looked at a broad range of patients with multiple diseases. Furthermore, different studies also used different measurements of incidence and prevalence; i.e., some investigators focused on lifetime use, others reviewed the previous 12-month history only, yet others limited their findings to BBT use in the 14 days prior to the investigation. Regardless, results indicated that a high percentage of patients with cardiovascular diseases are taking some kind of BBT (almost one out of every two patients). Health care professionals need to be aware of these findings and routinely inquire about BBT use by patients when taking a medication history. Similar to Ackman's study [ 21 ], the most frequently utilized BBT in the present study of cardiovascular patients was vitamin E, (41 [43.6%]). The popularity of vitamin E among cardiovascular patients is not surprising due to the alleged benefits of this vitamin in heart disease from early literature [ 12 - 14 ], and its relative availability in pharmacies, health food stores and supermarkets. As was the case with vitamin E, the use of vitamin C in the current investigation, (30 [31.9%]) was similar to previously reported results, being the second most common product utilized by cardiovascular patients [ 21 ]. Similar to other studies [ 1 , 3 , 6 , 21 ], the current study confirmed that based on the patients' reports, a high percentage of physicians (40%) and pharmacists (68%) were not aware of BBT used by cardiovascular patients. Use of BBT by the patients in this study was not routinely discussed with health care providers during medical history evaluation (33%). This is an alarming but not an isolated finding. On average, cardiovascular patients consumed 7 (range, 1 – 18) prescribed medications, and 2 (range, 1 – 12) BBT products. Considering the complex medical treatment received by cardiovascular patients, addition of unmonitored BBT to the patients' regimens may place the users of these therapies at a greater risk for the development of adverse events and interactions with prescribed medications. As addressed by this analysis, 42 potential drug-BBT interactions were identified. Demographics impact on biological based therapy use in cardiovascular patients Among the demographic variables collected in this analysis, no parameter other than the level of education had a significant impact on whether patients used BBT. The present investigation revealed that the education level among users (63% received some college or college and graduate school education) of BBT was higher than that of the nonusers (32%), p < 0.001. This finding has also been reported in other studies [ 3 , 6 ]. Higher level of education has been shown to significantly influence the use of alternative products and services [ 3 ]. Education level is sometimes directly related to economic status, thus the patients may have more resources to spend on BBT. Similarly, better-educated consumers may be more likely to be exposed to various less conventional forms of healthcare reading about their illnesses and treatment options. Educated patients also might be less inclined to accept their physicians' knowledge and expertise, and may seek other treatment options. Perceived safety and efficacy of BBT The majority of BBT users (as compared to nonusers) in the current investigation believed that BBT products are safe, effective and cause fewer side effects as compared to prescription medications. These results are consistent with the findings by other investigators [ 6 , 21 ], supporting a logical conclusion that patients who believe that BBT are safe and effective, are more likely to use them. Study limitations Several limitations of the study need to be addressed. Those that are intrinsically related to survey data collection in general include potential bias of responders and nonresponders. Since participation in this study was voluntary, patients who chose to participate in the study may be more motivated and knowledgeable about the subject of BBT as compared to those who refused to participate. Therefore, the results of this study may not be applicable to the entire cardiovascular patient population. Also, a cross-sectional nature of this study precludes drawing any definitive conclusions regarding cause and effect relationships. For example, one cannot definitively conclude that more education will significantly effect one's decision to take BBT because other factors such as exposure to these types of products may have influenced more educated consumers to take these products. As with other surveys, recall bias cannot be eliminated as data collection relied on patients' self-report, rather than objectively documenting BBT use. Additionally, with surveys there is always a possibility of misunderstanding the questions and responses, miscommunication between the interviewer and the patients, and inaccurate recording of the information. The majority of the survey collected factual information (i.e., demographics and BBT usage history). However, the questions that evaluated patients' perception on the safety and efficacy of conventional medicine and BBT were not validated. This study was conducted in a large urban inner city hospital; the results may not be extrapolated to other cities or clinical settings. Another limitation may be the exclusion of non-English speaking patients, which may be particularly important since there are certain ethnic groups who are documented to use more BBT than are other ethnic groups (e.g., persons of or from Chinese descent utilize traditional Chinese herbal medications). Finally, terminology varies greatly in the published literature. Some authors use the term CAM in all instances, others differentiate CAM from BBT, etc. As a result, it is very difficult to compare published reports with certainty. Conclusion In recent years, the interest of using BBT in disease management has increased dramatically in the medical and layman communities. The amount of valid scientific research in this area of therapy continues to increase. Yet, there are still many unknowns concerning BBT, especially in the area of adverse effects and drug interactions. The finding of a high prevalence of BBT (47.5%) use among cardiovascular patients and the lack of communication between patients and their physicians/pharmacists should be addressed by the health care community. Higher education level, as shown in the present study and other previous investigations [ 1 , 3 , 6 , 22 ], is associated with an increased use of BBT, but it does not necessarily mean that these patients are aware of the potential detrimental effects of BBT, as demonstrated in the current study. In cardiovascular patients, the perceived effectiveness and safety of BBT, and assumed lack of side effects of these products as opposed to traditional medications, highlights an area for further education. A high incidence of potential drug-BBT interactions was also identified in this study (42 interactions in 94 users). Given that the use of BBT can have a direct effect on patient care, and users of these therapies do not always voluntarily report their use of these products to their providers, health care professionals need to inquire about BBT use routinely. Collecting complete patient histories and educating patients about potential dangers and possibilities of adverse effects and interactions between prescription medications and BBT (or other CAM) will lead to better overall patient care. Competing interests The author(s) declare that they have no competing interests. Authors' contributions LC participated in literature review, protocol development, collected and analyzed data, and submitted the manuscript for publication. DB collected the data. JWMC participated in study design, data analysis and preparation of manuscript. VR participated in design development and data analysis. HLK proposed the project and reviewed the manuscript. GCC, JM and BM participated in protocol implementation and manuscript review. All authors read and approved the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 Appendix A – Biological based therapy survey, It's a survey tool utilized in the study to collect patient data. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC555537.xml
538268	CpG methylation of the FHIT, FANCF, cyclin-D2, BRCA2 and RUNX3 genes in Granulosa cell tumors (GCTs) of ovarian origin	Background Granulosa cell tumors (GCTs) are relatively rare and are subtypes of the sex-cord stromal neoplasms. Methylation induced silencing in the promoters of genes such as tumor suppressor genes, DNA repair genes and pro-apoptotic genes is recognised as a critical factor in cancer development. Methods We examined the role of promoter hypermethylation, an epigenetic alteration that is associated with the silencing tumor suppressor genes in human cancer, by studying 5 gene promoters in 25 GCTs cases by methylation specific PCR and RT-PCR. In addition, the compatible tissues (normal tissues distant from lesion) from three non-astrocytoma patients were also included as the control. Results Frequencies of methylation in GCTs were 7/25 (28 % for FHIT ), 6/25 (24% for FNACF ), 3/25 (12% for Cyclin D2 ), 1/25 (4% for BRCA2) and 14/25 (56%) in RUNX3 genes. Correlation of promoter methylation with clinical characteristics and other genetic changes revealed that overall promoter methylation was higher in more advanced stage of the disease. Promoter methylation was associated with gene silencing in GCT cell lines. Treatment with methylation or histone deacetylation-inhibiting agents resulted in profound reactivation of gene expression. Conclusions These results may have implications in better understanding the underlying epigenetic mechanisms in GCT development, provide prognostic indicators, and identify important gene targets for treatment.	Background Ovarian cancer is one of the most common cancers in women of all age groups. Among ovarian neoplasms, granulosa cell tumors (GCTs) are relatively rare, accounting for approximately 3% of all ovarian cancers. One common DNA modification is promoter hypermethylation associated with loss of expression of a tumor suppressor gene. During cancer development, there is a shift in methylation patterns and some promoter region CpG islands become methylated leading to silencing of the adjacent genes [ 1 ] and this process is considered to be a critical step in cancer development. The causes of methylation change in normal cells remain unknown. It has been hypothesised that fixation of methylation occurs when random seeding of methylation occurs in the promoter regions of silent genes [ 2 ]. Methylation seems to provide an ideal set of cancer specific markers for early detection of or for monitoring response to treatment. However, the use of methylation at a given site as a marker to detect low numbers of tumor cells relies on the background levels of methylation in normal tissues to be nearly zero. The human FHIT gene is a member of the histidine triad gene family [ 3 ], the function of which remains unknown. FHIT gene has been shown to be hypermethylated in oesophageal, lung, breast, prostate, bladder, cervical, and oral cancers [ 4 - 9 ]. Recent studies indicate that FANC proteins interact with both BRCA1 and BRCA2 genes through a common pathway [ 10 ]. Methylation changes may disrupt the FANC-BRCA pathway and hence may be a marker change in the cancer development in granulosa cells of the ovary. Aberrant expression of cyclin D2 was also demonstrated in human ovarian granulosa cell tumors and testicular germ cell tumor cell lines [ 11 ]. In breast cancer, repression of expression was attributed to methylation of the cyclin D2 gene promoter region. Several studies have indicated that methylation of cyclin D2 and its mRNA and protein were absent in most breast cancer cell lines examined and in primary breast cancers although normal breast epithelial cells had abundant expression [ 12 - 14 ]. BRCA2 may play role in regulation of the cell cycle during proliferation and differentiation. To date, only one study has shown the absence of methylation in the promoter region of BRCA2 in breast cancers cell lines and other normal human breast, bladder, colon, and liver tissues [ 15 ]. RUNX3 is one of the genes with RUNT domain, which has been identified to have a tumor suppressor role that frequently shows loss of expression due to hemizygous deletion and hypermethylation in gastric cancers [ 16 , 17 ]. The role of epigenetic gene inactivation in GCTs of ovarian origin is yet not fully understood. Previously published reports on GCTs and its precursor lesions showed varying degree of promoter methylation of many tumour suppressor genes [ 18 - 20 ]. To investigate the role of promoter methylation in detail in ovarian tumorigenesis, we further evaluated CpG methylation of 5 more tumour suppressor genes in 25 GCTs and cell lines. We found 68% of GCTs patients exhibiting promoter methylation in at least one gene. The FHIT , FANCF and RUNX3 gene promoters were frequently methylated. Methylation status was correlated with histologic characteristics. We also found evidence that promoter methylation inactivates gene expression in GCTs and exposure to methylation and/or histone deacetylase (HDAC)-inhibiting agents reactivate the gene expression. Results We examined the hypermethylation status of a panel of 5 normally unmethylated tumor suppressor or cancer genes: FHIT , FANCF , Cyclin-D2 , BRCA2 and RUNX3 in 25 ovarian GCTs and ovarian cell line DNAs using the MSP assay (Fig. 1 ). The frequency of promoter hypermethylation of the tumor suppressor gene loci included in the panel was FHIT 2 8%, FANCF 24 %, Cyclin - D2 12%, BRCA2 4%, and RUNX3 56% of the 25 tumours (Table 1 ). DNA methylation and mRNA expression results in 5 ovarian cell lines are shown in Table 2 . Fig. 2a and 2b shows representative examples of MSP of each gene. Figure 1 Methylated profile of GCTs of ovarian origin in 25 patients for FHIT, FANCF, cyclin-D2 and BRCa2 genes. Results were scored as methylated (dark boxes) or unmethylated (light grey boxes). Table 1 DNA methylation and mRNA expression in ovarian cancer cell lines Cell line MSP mRNA expression FHIT FANCF Cyclin-D2 BRCA2 RUNX3 FHIT FANCF Cyclin-D2 RUNX3 TOV-21G M M/U M U U + - + + C13* M/U M M U M - + - - OAW42 U M/U U U U + + + + OAW28 U U U U U + + + + OV-90 M/U U M U U + + - + KGN M U U U M + + + - Table 2 Association of grade classification (IA, IB and IC) and promoter methylation of different genes in granulosa cell tumours of ovarian origin Genes Methylation Status Grade classification p-value * p-value † p-value ** p-value ‡ p-value # IA IB IC FHIT Methylated 3 1 3 1 0.0526 0.5856 0.2352 0.041 # Unmethylated 8 9 1 FANCF Methylated 2 1 3 0.6609 0.0312 † 1 0.0769 0.041 # Unmethylated 9 9 1 Cyclin-D2 Methylated 0 0 3 0.23 0.0017 † - 0.0088 ‡ 0.011 # Unmethylated 11 10 1 BRCA2 Methylated 0 0 1 1 0.16 - 0.2667 0.2857 Unmethylated 11 10 3 RUNX3 Methylated 7 3 4 0.6887 0.1052 0.1984 0.5165 0.0699 Unmethylated 4 7 0 Fisher's exact test. IA type vs. IB and IC type; †Fisher's exact test. IA and IB type vs. IC type GCTs of ovarian origin; Fisher's exact test. IA type vs. IB; ‡Fisher's exact test. IA type vs. IC type; #Fisher's exact test. IB type vs. IC type GCTs of ovarian origin Figure 2 (a) Representative examples of MSP of FHIT , FANCF , Cyclin - D2 and BRCA2 genes in cell lines and tumor samples; -ve, negative control (water blank PCR mixture without template); +ve, positive control (normal lymphocyte DNA treated with Sss I methyl transferase). The unmethylated form of p16 was amplified as a control to check DNA integrity. (b) Representative examples of MSP with RUNX3 gene and its expression before and after treatment with 5'-AZA-dC in different ovarian tissue (both normal and cancerous) using RT-PCR. UT: untreated; T: treated with 5'-AZA-dC; T2–T25: tumour samples. Hypermethylation was observed in all of the histological stages of cancer examined and in patients of all ages except BRCA2 which is seen in only one patient with stage IC. Eighteen tumors showed methylation of at least 1 gene, and 7 tumors showed no methylation of any of the 5 genes. A total of 28% of these tumors had one gene, 16% two genes, 8% three genes and 8% four genes hypermethylated (Fig 1 ). No methylation was observed in 15 normal ovarian tissue DNAs and 50 lymphocyte DNAs from females. Using statistical analysis, we examined methylation with regard to these cancer patient clinicopathological parameters of age and stage. None of these patients have any smoking history. FHIT methylation was only found in all three stages of the tumors but was significantly more pronounced in IC (P = 0.041 vs IB). Hypermethylation of FANCF was significantly more frequent in stage IC (P = 0.0312 when compared with IA and IB collectively; P = 0.041 vs IB). Cyclin - D2 methylation was found only in patients with IC type of cancer ( P = 0.0017 vs. IA+ IB; P = 0.0088 vs. IA and P = 0.011 vs. IB; Table 1 ). BRCA2 methylation was found only in one patient with stage IC only cancer. RUNX3 methylation was found in all the three stages of these tumours. However, all the patients with stage IC showed methylation to associate with high stage but not at a statistically significant level. FANCF and Cyclin-D2 methylation was found to be more pronounced in older than younger patients. However, we are unable to amplify the modified DNA for either of the PCR products. Methylation is known to inactivate the tumour suppressor genes. To test the hypothesis, we examined the expression of all these genes except BRCA2 by RT-PCR after cellular exposure to 5-Aza-2'-deoxycytidine (DAC). No FANCF mRNA in TOV-21G, Cyclin-D2 in C13 and OV-90, and RUNX3 in C13* and KGN was detected in untreated cell lines (Table 2 ). However, DAC treatment for 120 hours induced an increase in the detectable level of mRNA expression in these cell lines and the tumors samples that initially lacked the expression. We were unable to detect expression of RUNX3 gene in patient 11 that also shows no PCR products with either methylated or unmethylated PCR primers. Further studies involving this patient with markers specific for RUNX3 revealed the loss of heterozygosity (LOH) which may be responsible for the loss of expression (Data not given). Discussion Cancer cells especially of ovarian origin keep on accumulating genetic changes that allow them to evade various chemotherapeutic drugs and hence become increasingly dangerous. We try to answer some of the questions by exploring the role of methylation mediated gene silencing in five tumour suppressor gene in the present study. We asked a question whether hypermethylation of FHIT , FANCF , cyclin-D2 , BRCA2 and RUNX3 resulted in the loss of gene expression. We performed reverse transcription-PCR to test the expression of all these genes on RNA extracted from tumour samples. With few exceptions, expression of mRNA correlated well with promoter hypermethylation of these genes. This could be partly due to complex process that controls gene expression in which the chromatin conformation, cofactors availability, repressor process and enhancer molecules all play a part. The present study shows that methylation is one of the important determinants, because in majority of the cases, expression of these genes in these tumours correlates with hypermethylation of the promoter sequences. We found a link between aberrant methylation of genes investigated and the clinicopathological features of stage IC, even though the sample size was small. One has to bear in mind that GCTs of ovarian origin constitute less than 5% of the total ovarian cancers. Similar reports have been reported for the FHIT methylation in highly malignant osteosarcomatous [ 21 ]. Taken together, our results demonstrate that promoter aberrant methylation of FHIT is an important mechanism for inactivation of this tumor suppressor gene in many malignancies. Cytogenetic investigation in these tumours revealed that chromosome 3 and 11 are found to carry deletions and reciprocal translocations along with the trisomy 14 and monosomy 22. It could be important because some of the tumour suppressor genes ( FHIT , RASSF1A , RAR-β and FANCF ) studied extensively in these tumours are in fact localised on these chromosomes. FANC genes are essential in DNA repair pathways and recently, it has been shown that promoter hypermethylation of FANCF gene disrupts the FA-BRCA pathway, resulting in cisplastin resistance. We found FANCF promoter hypermethylation in 24% of the tumours. This is in line with the previous findings in squamous cell carcinomas of lung and oral cavity and, in cervical and ovarian cancer [ 22 - 25 ]. To test whether other epigenetic mechanisms such as partial methylation and histone deacetylation play a role, we examined the expression of all these genes except BRCA2 after treatment with DAC. DNA hypermethylation-mediated gene silencing is closely associated with histone modifications such as methyl-H3-K9. In this regard, DNA-demethylating agents 5'-aza-2'deoxycytidine (DAC) and trichostatin (TSA) reactivates expression of epigenetically silenced genes [ 26 ]. Although DNA hypermethylation is essential to maintain repressive state of histone code, histone modifications precede DNA hypermethylation in silencing specific genes [ 27 , 28 ]. In the present study, reactivation and/or increased expression of FHIT , cyclin-D2 and RUNX3 in C13, FANCF in TOV-21G and cyclin-D2 in OV-90 cell line after exposure to DAC in the absence of promoter methylation suggests that key histone modifications, either by direct or indirect involvement of promoter methylation, also play a role in down-regulating FANCF gene expression in this cancer type. It is therefore assumed that FANCF silencing might be considered a candidate-mechanism underlying the state of genomic instability may be considered to be a rate-limiting in the origin of GCTs of ovarian origin. FANCF gene silencing has been shown to revert in vitro as a result of demethylation in some ovarian cell lines as seen in the present study also [ 23 ]. These reversible changes in the methylation status further complicate the tumour assessment that shows the contribution of FANCF methylation. The present finding along with the results from other cancer types suggests the presence of genomic instability due to the methylation mediated FANCF gene silencing. Cyclin D2 gene promoter is hypermethylated in 12% of GCTs, a number that is consistent with previous findings in other cancers [ 14 , 29 , 30 ]. We noted a trend that Cyclin D2 methylation is found only in IC type tumors. The present study showed that hypermethylation of the RUNX3 gene promoter frequently occurred in ovarian GCTs. In human gastric cancers it has been shown that promoter hypermethylation and hemizygous deletion of the RUNX3 gene correlated with a significant reduction in expression, and the tumorigenicity of cell lines in nude mice was inversely related to their level of RUNX3 expression, indicating that RUNX3 is a tumor suppressor involved in the development of gastric cancers [ 17 ]. The presence of RUNX3 CpG island hypermethylation in GCTs, but not in normal ovarian tissue or peripheral blood suggests that RUNX3 hypermethylation might be associated with the genesis of this cancer type and this frequent methylation might serve as a biological marker. Loss of expression in RUNX3 gene in patient 11 which also lacks any PCR product with either of the unmethylated and methylated primers indicates that it could be due to the large deletion in the gene. However, studies involving the markers D1S234 and D1S199 showed the loss of heterozygosity which could to some extent explain the loss of expression in this patient. We could not confirm this due to the lack of further sample. However, the sample quality was good as indicated by the presence of band in GAPDH1 which was used as an internal control. The histological examination of granulosa cells revealed that these are small, usually round to polygonal, but may be spindle-shaped with scanty amphophilic cytoplasm, containing only occasional small lipid droplets, and having indistinct cell borders. Granulosa cells regularly express inhibin and contain vimentin and smooth muscle actin intermediate filaments and, less commonly, cytokeratins. However, during their malignant transformation, theses cells exhibit one or more of so-called 'microfollicular', 'macrofollicular', 'trabecular' or 'insular' patterns. The microfollicular variant is characterized by multiple small rounded spaces formed by cystic degeneration in small aggregates of granulosa cells and often fragments of nuclear debris or pyknotic nuclei. These spaces, known as Call-Exner bodies, are found in only 30–50% of tumors. The granulosa cell nuclei are oriented somewhat radially around these structures. The nuclei are typically bland and often grooved in granulosa cell tumors and have few mitoses, but show variation in size and shape, and marked hyperchromatism. There exist two different pathways that can contribute to the development of cancers; genome-wide hypomethylation may lead to the loss of chromosomes (as seen in these tumours) leading to chromosomal instability, whereas promoter methylation in tumour suppressor genes, which are responsible for gene silencing, can lead to the development of cancers in somatic cells. Therefore, the balance in DNA methylation is very important, and alteration in these may be protective in one pathway but deleterious in the other. Conclusions In view of the high mortality rates associated with ovarian cancer, a better understanding of the molecular mechanisms underlying tumor progression in the disease could reveal novel pathways of high clinical relevance. It can therefore be concluded that promoter hypermethylation of these genes, and loss of expression in ovarian cancers (GCTs) is relatively common and this may also be useful as a tumour marker for early diagnosis and subsequent disease monitoring. Hence, these epigenetic signatures could play a decisive role in designing treatment options for this category of ovarian cancer. These results may also have implications in better understanding the underlying epigenetic mechanisms in GCT development thus can provide prognostic indicators, and identify important gene targets for treatment of this type of cancer in females. Methods Study Population The subjects were 25 patients affected with GCTs of ovarian origin without a positive family history. The study was approved by the Health Research Ethics Board of the Faculty of Natural Sciences. All of them were untreated at the time of study. Fifty normal individuals ranging in age from 20–60 years were studied simultaneously under similar experimental conditions. The tumor grading is as follows: 11 FIGO stage IA, 10 FIGO stage IB and 4 FIGO stage IC. All these diagnoses were reviewed by a gynaecologic pathologist and the tumor were assessed using standard criteria. An informed consent was taken from all the subjects prior the study. Fresh cancer tissue specimens were received from all the patient and these were cultured as per standard protocol. A part of the fresh tissue was used to isolate DNA and RNA for further analysis where as the cultured cells were used for the 5'-Aza-2'-deoxycytidine experiments. Microscopically, GCTs are composed of granulosa cells, theca cells, and fibroblasts in varying amounts and combinations. The term granulosa-theca cell tumor had been applied to all tumors in which both cell types were identified, regardless of the amounts present. Cell Lines and DNA Isolation Five human ovarian cancer cell lines (TOV21G, C13, OAW28, OAW42, OV90 and KGN) from the American Type Culture Collection and European Collection of Cell Culture were maintained in RPMI 1640 supplemented with 10% fetal bovine serum (Hyclone, Logan, UT) and were grown at 37°C in 5% CO 2 [ 31 ]. DNA was isolated from cultured cells using QIAamp DNA Mini kit (Qiagen Inc.) and quantified. LOH studies were performed using markers D1S199 and D1S234 specific for locus 1p36.11. Methylation Status by Methylation-Specific PCR (MS-PCR) DNA methylation patterns in CpG islands of tumor suppressor genes FHIT, FNACF and BRCA2 were determined by chemical modification with sodium bisulphite as described previously [ 32 ]. Briefly, 1 μg DNA l was denatured by NaOH (50 μl, final concentration, 0.2 M) for 10 min at 37°C. 1 μg of salmon sperm DNA (Sigma) was added as carrier before modification. Freshly prepared 30 μl of hydroquinone (10 mM, Sigma) and 520 μl of sodium bisulfite (3 M, pH 5.0, Sigma) were mixed and samples were incubated under mineral oil at 55°C for 16 hr. The DNA samples were desalted through Wizard columns (Promega, Madison, WI), desulfonated by NaOH (final concentration, 0.3 M) for 5 min at room temperature, followed by ethanol precipitation. DNA was resuspended in water and used immediately or stored at -20°C. 50 μl of bisulphite modified DNA was used for each MSP. Following primer pairs for FHIT gene, methylated CpG site, forward 5'-ttggggcgcgggtttgggtttttacgc-3' and reverse 5'-cgtaaacgacgccgaccccacta-3', unmethylated CpG site, forward 5'-ttggggtgtgggtttgggtttttatg-3', and reverse 5'-cataaacaacaccaaccccacta-3'; 189–262 bp relative to transcription start site, FANCF gene, methylated CpG site, forward 5'-tttttgcgtttgttggagaatcgggttttc-3' and reverse 5'-atacaccgcaaaccgccgacgaacaaaacg-3', unmethylated CpG site, forward 5'-tttttgtgtttgttggagaattgggttttt-3' and reverse 5'-atacaccacaaaccaccaacaaacaaaaca-3'; the primers corresponds to the position +280 to +432, BRCA2 gene, methylated CpG site, forward 5'-gacggttgggatgtttgataagg-3' and reverse 5'-aatctatcccctcacgcttctcc-3', unmethylated CpG site, forward 5'-agggtggtttgggatttttaagg-3' and reverse 5'-tcacacttctcccaacaacaacc-3'; these primers are 135 and 211 bp upstream of transcription start site and cyclin-D2 gene, methylated, forward 5'-tacgtgttagggtcgatcg-3' (-1427 to -1409) and reverse 5'-cgaaatatctacgctaaacg-3' (-1152 to -1171) and unmethylated, forward 5'-gttatgttatgtttgttgtatg-3' (-1616 to -1594) and reverse 5'-taaaatccaccaacacaatca-3' (-1394 to -1414). Each PCR reaction generated 74 bp products both with methylated and unmethylated primers for FHIT , 153 bp for FANCF both with methylated and unmethylated primers, 337 and 250 bp product with BRCA2 primers specific for methylated and unmethylated primers and 276 and 222 bp product for cyclin-D2 primers specific for methylated and unmethylated PCR reactions respectively. For the PCR reaction of 25 μl, 50 ng sodium bisulfite treated DNA was added to reaction buffer containing 0.2 mM dNTP, 16.6 mM (NH4) 2 SO4, 67 mM Tris pH 8.8, 10 mM β-mercaptoethanol, 1.5 mM MgCl2, 10 pmol of forward and reverse primers specific to the methylated and methylated DNA sequences and 1.25 units of AmpliTaq Gold (PE Biosystems, Foster City, CA, USA). The PCR reactions were cycled in a GeneAmp 9600 thermal cycler (Applied Biosystems) under the following conditions: preheat at 94°C for 3 min. followed by 40 cycles (94°C for 40 sec, 65°C for 40 sec, for FHIT , 4 cycles of 65°C and 36 cycles of 55°C for 60 seconds for FANCF and 62°C or 56°C for 40 seconds in methylated or unmethylated BRCA2 gene, 72°C for 45 sec and a final extension at 72°C for 7 min. The PCR conditions for cyclin-D2 are as follows: 1 cycle of 95°C for 5 min and 35 cycles at 95°C for 30 s, 55°C for 30 s, and 72°C for 45 s; and 1 cycle of 72°C for 5 min. In addition to these genes we also investigated the RUNX3 gene (Accession no AL023096) methylation status using the primers: forward, 5'-ataatagcggtcgttagggcgtcg-3', and reverse, 5'-gcttctactttcccgcttctcgcg-3' (64917–65031; 115 bp), for methylated DNA of RUNX3 ; and forward, 5'-ataatagtggttgttagggtgttg-3', and reverse, 5'-acttctactttcccacttctcaca-3' (64917–65031; 115 bp), for unmethylated DNA of RUNX3 with annealing temperature of 55°C for 20 seconds. For each PCR set, DNA isolated from normal peripheral lymphocytes of healthy individuals served as a negative methylation control. Human placental DNA was treated in vitro with SssI methyltransferase (NEB, Beverly, USA) to create completely methylated DNA at all CpG-rich regions and served as positive methylation control. Methylation-specific PCR products were analysed on 3% agarose gel electrophoresis with ethidium bromide staining. A positive control and a negative control were included in each amplification reaction. FHIT , FANCF , Cyclin-D2 and RUNX3 Expression by RT-PCR FHIT , FNACF and cyclin-D2 mRNA levels in cancer patients were compared to their expression in normal cells using semi-quantitative reverse transcriptase polymerase chain reaction. In brief, cDNA was synthesized using AMV reverse transcriptase (Promega, Mannheim, Germany) and amplified in duplex reactions in a total volume of 25 μl containing 150 μM of each dNTP, 1.5 mM MgCl2, 10 pmol of each primer pair and 1.0 units of Taq polymerase. After initial denaturation at 95°C (94°C) for 5 min, 30 cycles of 30 sec (45 sec) at 95°C (94°C), 30 sec (50 sec) at 58°C (55°C) and 45 sec (50 sec) at 72°C were performed followed by a final 10 min elongation step at 72°C. The values in the brackets correspond to FANCF . The following primers were used to amplify FHIT : forward 5'-gctcttgtgaataggaaacc-3' and reverse 5'-tcactggttgaagaatacagg-3' which yields 532 bp product spanning within exon 5 to exon 10, FANCF : forward 5'-ttcggaagtctttgctgcct-3' and reverse 5'-agtaataacacacgattgcc-3' which yields 413 bp product spanning from +733 to +1144. RT-PCR was performed for Cyclin D2 using the primers 5'-catggagctgctgtgccacg-3' (forward) and 5'-ccgacctacctccagcatcc-3' (reverse) with PCR conditions being: 1 cycle of 94°C for 3 min and 35 cycles at 94°C for 20 s, 55°C for 30 s, 72°C for 45 s followed by 72°C for 5 min. RT-PCR was also carried out for RUNX3 gene using the primers: forward 5'-aggcattgcgcagctcagcggagta-3' and reverse 5'-tctgctccgtgctgccctcgcactg-3' (152 bp). GAPDH1 was used as internal control. Each reaction was performed in triplicate. PCR products were electrophoresed on 2% agarose gels and quantified using densitometer (Molecular Dynamics). Fold increases in expression in cancer cells were calculated with respect to the levels of the transcripts in normal cells. 5-Aza-2'-deoxycytidine and n-butyrate treatment Cell lines (TOV21G, C13*, OAW28, OAW42 and OV90), normal ovarian cells and tumour cells of ovarian GCTs were treated with demethylating agent 5-Aza-2' deoxycytidine (Sigma) for five days at a concentration of 2.5 μM, HDAC-inhibiting agent trichostatin (TSA) at a final concentration of 5 μM for the last 24 hours or a combination of both. Total RNA isolated from treated, untreated cell lines and cell of the cancerous tissue was reverse transcribed using random primers and the Pro-STAR first strand RT-PCR kit (Stratagene, La Jolla, CA). A semi-quantitative analysis of gene expression was performed in replicate experiments using 30 cycles of RT-PCR as described above. Competing Interests The authors declare that they have no competing interests. Authors' Contributions VSD for executing the MSP and RT-PCR experiments; completing manuscript; MS, for carrying out cell culturing; and SAH conceived and coordinated the study. All authors read and approved the final manuscript. Figure 3 (a). Expression of FHIT, FANCF and Cyclin D2 genes before and after treatment with 5'-AZA-dC in different cell lines and ovarian tissue (both normal and cancerous) using RT-PCR. NT: no treatment; T: treatment with 5'-AZA-dC; T1–T3: tumour samples; NOV: normal ovarian cells. Co-amplified product of GAPDH served as an internal control. (b) Results obtained with another cell line KGN and tumours before and after treatment with 5'-AZA-dC.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC538268.xml
523853	GeneXplorer: an interactive web application for microarray data visualization and analysis	Background When publishing large-scale microarray datasets, it is of great value to create supplemental websites where either the full data, or selected subsets corresponding to figures within the paper, can be browsed. We set out to create a CGI application containing many of the features of some of the existing standalone software for the visualization of clustered microarray data. Results We present GeneXplorer, a web application for interactive microarray data visualization and analysis in a web environment. GeneXplorer allows users to browse a microarray dataset in an intuitive fashion. It provides simple access to microarray data over the Internet and uses only HTML and JavaScript to display graphic and annotation information. It provides radar and zoom views of the data, allows display of the nearest neighbors to a gene expression vector based on their Pearson correlations and provides the ability to search gene annotation fields. Conclusions The software is released under the permissive MIT Open Source license, and the complete documentation and the entire source code are freely available for download from CPAN .	Background Microarray experiments produce vast amounts of data. The resulting datasets are highly complex and contain large matrices of expression measurements as well as sequence and experiment annotations that provide biological context to the data. To organize these different types of data in a way that allows intuitive exploration of the data, and provides the ability to gain important insights into relationships within a given dataset requires sophisticated visualization tools. Such visualization tools are of benefit not only to researchers analyzing and presenting or publishing their own data, but also to Model Organism Databases (MODs) for compiling and displaying microarray data for a given model organism. There are several excellent free tools available that allow an individual user to analyze their own data. These tools are either accessible on the web, or can be downloaded and used on a desktop machine. Examples include the EPCLUST [ 1 ], GEPAS [ 2 , 3 ] and FGDP [ 4 , 5 ] web-based tools and the TMEV [ 6 , 7 ] desktop tool from TIGR. However, once these tools have been used, and a cluster or other group of genes has been selected, this resulting dataset needs to be made available to other people for browsing and exploration. There are a few visualization tools that allow display of such a static dataset that are available as free software tools, e.g. Michael Eisen's TreeView [ 8 , 9 ], JavaTreeView [ 10 ], or the more recent MapleTree [ 8 ]. All of these tools are, however, desktop tools that themselves have to be downloaded and work on locally stored datasets. The impetus for the development of GeneXplorer was the desire to provide access to datasets via the Internet, without the requirement to download and install additional software. We developed GeneXplorer for use in web supplements of microarray publications whose raw data are housed within the Stanford Microarray Database (SMD) [ 11 , 12 ] and for use as a tool to allow SMD users to browse their own data within SMD before publication. Using GeneXplorer, hierarchically clustered gene expression data can be interactively viewed using a web browser on any computer platform. GeneXplorer uses the widely accepted CDT file format [ 13 ] produced by several freely available clustering programs (e.g. [ 9 , 14 ]), which between them have been downloaded several thousand times. Thus GeneXplorer should be widely usable my SMD and non-SMD users alike. Implementation The application was written using object oriented Perl following the Model-View-Controller (MVC) design paradigm [ 15 ]. GeneXplorer consists of two classes, the data model class Microarray::CdtDataset (M), and the presentation logic class Microarray::Explorer (V). The controller, named gx, is a Perl CGI script that dispatches CGI requests to the viewer. The MVC paradigm was used because it dissociates how data are represented internally (the Model) from how they are displayed (the View), from how they are interacted with (the Controller) (see Figure 1 .). The goal of such a separation is that by keeping consistent APIs for the components to interact with each other, each component may be modified extensively internally, with little or no effect on the other parts of the application, thus making code maintenance easier. The Microarray::CdtDataset class provides an application programming interface (API) that allows details of a particular expression cluster to be queried. In turn, instances of the Microarray::Explorer class use this API to retrieve and then display information about the dataset. The controller is a relatively simple CGI Perl script that is responsible for capturing CGI parameters and using them to first create a dataset Microarray::CdtDataset object, which is subsequently used in the instantiation of a Microarray::Explorer object. The controller then invokes the appropriate Microarray::Explorer methods, depending on where, and in which frame the user clicked. The Microarray::CdtDataset has two essential functions: during dataset creation (see below) it decomposes the data file into its constituent data parts and creates the files needed during data viewing (see below). During data viewing it provides the API for the viewer, and allows searching and retrieval of the data. Under the current model the dataset object itself is immutable. Microarray::CdtDataset was implemented as a client of the Microarray::DataMatrix module, which provides an API for accessing matrices of expression data. In the design of the classes certain compromises had to be made to accommodate the stateless client server environment in which the program operates. Specifically, to allow rapid responses, pre-generated images and correlation data are cached in a compact format on the web-server. There are two stages required to publish a microarray dataset on the web using GeneXplorer. The first stage (executed only once per dataset) involves creation of all the necessary files for GeneXplorer to use. The second stage uses these files to produce the display using the GeneXplorer web front-end. Dataset creation Dataset creation requires a file in the Clustered Data Table (CDT) format: a simple tab delimited text file format (see [ 13 ] for file format details). This format was introduced with the 'Cluster' and 'TreeView' applications [ 8 ] and is widely used for microarray data. A perl script (makeMicroarrayDataset.pl) uses Microarray::CdtDataset to create the various required data files. Correlations between expression-vectors within the dataset are calculated for each pair-wise combination of vectors using the C program 'correlations'. Correlations for each vector above the default cutoff value of 0.5 (which is configurable) are saved in a binary format that facilitates rapid searching. Depending on the version of the Perl GD module [ 16 ] installed on the system, either png or gif formatted images representing the cluster will be created. These images include both a 2-color image representation of the data matrix and an image representation of the experiment names. The program that creates these files is configurable, such that these images can be created using either a red/green or a yellow/blue color scheme, and in addition, the contrast of the images can be customized and set in steps of log(2) scale. The name and path of a dataset can be defined hierarchically within the file system (see Figure 2 ) allowing the creation of many datasets within the same project. Dataset viewing GeneXplorer is a Perl application that produces a set of html frames that can be used for viewing the expression data (Figure 3 .). The three frames that it produces are: 1) A radar frame. This frame displays an image map of the data matrix and gives an overall view of the clustered data. The rows correspond to the features or genes (also referred to as reporters), and the columns correspond to the experiments within the dataset. When the image is clicked the next 100 expression patterns starting at the position of the click are displayed in the zoom frame. The position of a bracket on the right side of the radar window indicates the section of the whole radar image that is displayed in the zoom frame. 2) A toolbar frame. Actions in the toolbar may affect either the radar or the zoom frame. There is a tool to set the scaling of the radar image, while the search box allows searching of gene annotations and the expression patterns of the resulting hits are displayed in the zoom frame. In addition the toolbar frame also contains a JavaScript enabled text box that gives feedback depending on the user's mouse position, to provide additional information about the genes and experiments within the cluster. 3) A zoom frame. This frame displays a zoomed view of selected expression patterns, such that the user can see both the individual patterns and the associated annotations. The source of the selected patterns can be either a section of the radar image, the result of a search the user performed in the toolbar, or the result of a nearest neighbor search initiated in the zoom frame itself. The expression profiles themselves in the zoom frame are clickable and the resulting search will display the expression pattern for the most similarly expressed genes to the gene that was clicked on, and provide visual feedback as to the level of similarity in their expression profiles. In addition, when the user moves the mouse over parts of the zoom window, additional information is directed back to the textbox in the toolbar. The experiment name, correlation value (Figure 3d ) and gene annotation is displayed when the mouse is over the experiment image map, the correlation bar, and the expression pattern, respectively. Full text searching The search box in the toolbar enables a string search of either all, or specific gene annotation fields. The string may contain more than one term, where each term in the search string should be at least 2 characters long. Spaces between the terms are interpreted as term separators and the terms are combined using the logical 'AND' operator. Wildcard searches are allowed using the '*' character, such that at least one character should precede the wildcard character. The hits resulting from the search are displayed in the zoom frame, as expression patterns. The number of hits displayed in the zoom window is limited to 200 hits. Display configuration GeneXplorer allows configurable linking out of the gene annotations to external databases. The number of these links per a gene is not limited, making it easy to be able to look at the information for a gene in several different databases. A configuration file in the dataset directory is used to control where the various gene identifiers are linked. Templates are available for various organisms, and the existing files can be edited manually if a link to a new database is desired. Because of the current limitations of the input cdt file format, setting up the external database links might require manual editing at the time of dataset creation. This is fully described in the README document that is part of the distribution. The external database annotations are not currently updatable in any automated fashion; this will be addressed as part of our plans to make GeneXplorer able to read MAGE-ML (see future plans) that would allow us to do the updates via web services. Installation and use The GeneXplorer package is provided as a typical Perl distribution on the Comprehensive Perl Archive Network (CPAN), and adheres to the usual installation mantra of perl modules. After unpacking the software, a user with administrative privileges merely needs to type: perl Makefile.PL make make test make install This will install the libraries and the executable files that are needed for dataset creation by GeneXplorer into the regular system locations, unless otherwise specified during the first step above. The example in Figure 2 shows the file structure if the library and bin directories under the web server's root had been specified for installation of the libraries and executables respectively. To actually use the gx script, it must be copied into a cgi-bin directory, and the various html files must be copied to the appropriate location under the web server's root (see Figure 2 ). Results and discussion In addition to its use within SMD, GeneXplorer has been used by many publications to provide access to microarray datasets through their web supplements, that can be accessed through SMD's publication page [ 17 ], and was used as the basis for visualization of fuzzy k-means cluster data [ 18 ]. We demonstrate on an example dataset how GeneXplorer works [ 19 , 20 ]. Figure 3a shows a display of this dataset in the browser window. The whole dataset is displayed in the radar frame, and the zoom window shows the section of this image that was selected, with the gene annotations at a readable size. Clicking on any of the hyperlinks in the zoom frame brings up a new window displaying the biological information for the selected gene that is found in SOURCE (Figure 3b .) [ 21 ]. Searching the dataset for all the genes whose name field contains the keyword 'kinase' results in the zoom window shown in Figure 3c . This type of search allows comparison of the expression patterns of a subset of the genes based on some functional category – e.g. GO process-terms, if the annotation fields contain these terms. Clicking on one of the expression profiles (the one belonging to 'Estrogen Receptor 1', in this case) leads to the display in Figure 3d . In the zoom frame it shows the expression profile of the selected gene as the top row, and all the other expression profiles below with Pearson correlation above 0.5. The length of the small orange bar on the right side of the expression profiles gives a graphical representation of these correlation values, while the actual value is displayed in the info box in the toolbar when the mouse is over the orange bar. Future developments We are planning to further develop GeneXplorer to enable it to handle other data formats. Specifically, we would like it to be able to accept data files in MAGE-ML format [ 22 ], which is becoming a standard file format for communicating gene expression data. In addition, we would like it to be able to display tree views of the clustered data and allow zooming on specific nodes of the cluster. Conclusions We have developed a web-application, GeneXplorer, which allows the visualization of microarray datasets over the Internet using only a web browser. This application has been extremely useful in our experience, where it serves both SMD users during analysis of their data and the public while browsing published datasets. Availability and requirements GeneXplorer is available at [ 23 ] under the MIT Open Source license. It should work on any UNIX-type system capable of running Perl and a Web server, though we ourselves have deployed it on Sun Solaris. Additional information on installation and usage is provided in the installation instructions and documentation that is part of the distribution. List of abbreviations used SMD: Stanford Microarray Database. Authors' contributions CAR designed and wrote the initial version of the GeneXplorer. This was extensively re-factored and modularized by JCM (library modules for dataset) and JD (for explorer). DB was involved in the guidance of the early stages of this project. GS wrote the correlations software and the DataMatrix classes, and guided the development of this project. All authors read and approved the final version of the manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC523853.xml
523847	PLoS Medicine— A Medical Journal for the Internet Age	A message from the founders of the Public Library of Science	The Internet is awash with medical information. Eight hundred million people have direct access to the Internet [ 1 ], and in the United States over 60% have searched for health or medical information on the Web [ 2 ]. Go to any search engine and type in the name of a disease or drug, and you will be directed to hundreds of sites, ranging from the sound and useful to the quackish and dangerous. Google “medical” and you get 85 million pages, “drug,” 40 million, and “health,” 230 million. But something is conspicuously missing. The most reliable medical information on the Internet—the contents of peer-reviewed medical journals—is hidden from the public and most of the world's physicians. Although most medical journals are available online, their publishers limit access to those who choose, and can afford, to pay for access. This should not, and need not, be so. In the 19th and early 20th centuries independent physicians and small medical societies, interested in making the best new medical knowledge available to doctors, students, and the public, began to publish general medical journals containing case reports, ideas for new treatments, and the results of medical experiments. These pioneers took advantage of the best available technology for disseminating information, printing titles like The Lancet, The New England Journal of Medicine, and The Journal of the American Medical Association on cheap paper and selling them to subscribers at a few pennies a copy. For more than a century, printed journals like these were the dominant means of conveying medical knowledge around the world. But technology has changed. The Internet is now the most economical and efficient conduit for the delivery of information to most places. Publishers of medical journals realize this—when the Internet took off, they took their journals online. But while they adapted their means of distribution to the 21st century, they left their business model in the 19th century, continuing to charge readers for access just as they had done for their printed journals. This has been good for business—medical publishing has never been more profitable—but it comes at a huge cost. The established medical publishers have turned their back on the opportunity to make the latest and best medical information available to anyone with an Internet connection. With the launch of PLoS Medicine, we are embracing this opportunity. Everything we publish is immediately, freely available online throughout the world, with no restrictions on distribution, copying, printing, or legitimate use. Everything published in PLoS Medicine is immediately freely available online throughout the world, with no restrictions on distribution, copying, printing, or legitimate use. Of course, it costs us money to publish this journal, and we must cover our expenses. But the fee-for-access business model that made perfect sense for the printed journal is no longer consistent with the mission of medical publishing because it needlessly limits the reach of the medical literature. And so we have adopted a new model. Instead of charging readers for access to our journal, we ask the authors of accepted research articles to pay a publication fee to cover the costs of peer review, editorial oversight, and production. This “open access” business model ensures our financial health as a publisher while allowing us to convey everything we publish to the widest possible audience. Of course, we do not expect authors to cover publication costs personally—rather, we expect the government agencies, companies, foundations, research institutions, hospitals, or universities that sponsor the research to pay the fee. These organizations have always considered the wide dissemination of the results of the research they support to be an integral part of their mission. Virtually every leading sponsor of medical research has announced its willingness to pay for open-access publication, the costs of which average less than one percent of the cost of the research itself—a small price to pay to ensure that everyone who could benefit from their research can benefit from it. We realize that not everyone with something important to convey in a medical journal has access to such funds. To ensure that we don't replace a barrier to access with barriers to publication, we've raised money to cover the publication costs of articles whose authors are unable to pay them. And, for every PLoS journal, an author's ability to pay will never be a consideration in our decision to publish an article. Despite its obvious benefits, open-access publication has met with fierce opposition. Established medical publishers—now businesses more than forces for change—see open-access not as an opportunity to fulfill a mission of public service but as a threat to their lucrative businesses. They contend that their journals still serve the community well, and object that open access threatens their very existence. This is nonsense! It is our responsibility as publishers and members of the medical community not only to give patients access to the medical literature, but to provide them with tools to use it wisely. The Wellcome Trust, the world's largest charitable sponsor of biomedical research, seeking to ensure that the results of the science it funds are “disseminated widely and freely available to all,” recently commissioned a thorough analysis of the scientific and medical publishing industry [ 3 ]. It concluded that the current market “does not operate in the long-term interest of the research community,” and issued a strong statement in support of open access [ 4 ]. Responding to concerns about journal finances, the trust commissioned a detailed economic analysis of open-access publishing [ 5 ], based on which it concluded that “the open access model of scientific publishing—where the author of a research paper pays for peer reviewed research to be made available on the web free to all who wish to use it—is economically viable, guarantees high quality research and is a sustainable option which could revolutionise the world of traditional scientific publishing” [ 6 ]. (This report, freely available online, is an excellent resource for anyone with questions about the economics of open-access publishing). We know firsthand that the Wellcome Trust is right. In October 2003, we launched our first journal, PLoS Biology, and it is thriving—not only as a destination for the best research in all areas of biology, but also as a resource for students, teachers, and members of the public who have never before had direct access to the product of scientific inquiry (see for yourself at www.plosbiology.org ). We are now bringing this success and this spirit to medicine. The world of medical journals needs a fresh infusion of idealism. All of today's leading medical journals are more than 70 years old, and PLoS Medicine is here to challenge the status quo. We are first and foremost an open-access publisher working to ensure that everyone has access to the latest medical research and expertise. But we aim to be more than just an open-access alternative to established general medical journals. We are determined to make PLoS Medicine the best medical journal in the world by providing outstanding original research and new ideas; thought-provoking, educational, and imaginative features for readers; and the fastest, fairest, and most rigorous peer review for authors. As an open-access journal, we see our audience differently than do the conventional medical journals: our audience is composed of medical researchers, physicians, and other health-care providers, patients and their advocates, students, and the public around the world. It will be a great challenge to create a journal that will serve such a diverse audience—we welcome this challenge. We will make it possible for the results of advanced research on infectious diseases to guide treatment in remote clinics thousands of miles away. We will make the results of a clinical trial of a new drug accessible and understandable both to doctors who might prescribe it and to people who might start taking it. We will make research on rare diseases accessible to general practitioners and patients so that they can work together to recognize and treat them. Whereas some would argue that medical journals should not be accessible to patients because patients are unable to use the information effectively, we believe it is our responsibility as publishers and members of the medical community not only to give patients access, but to provide them with tools to use the medical literature wisely. Medical research is a partnership between medical scientists and millions of voluntary human participants, conducted largely with public funds. What better way to acknowledge the public's contribution and ensure their willingness to sponsor and participate in future research than to openly share the product of this research with them? We hope that you will enjoy reading PLoS Medicine and find it useful and provocative. Please share the journal with your colleagues, patients, and friends. Tell us what you want to see, what you like, and what we could do better. Give us your ideas for changes that will make PLoS Medicine a better journal for you and the community. Join us in reinventing the medical journal.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC523847.xml
521077	Suboptimal clinical response to ciprofloxacin in patients with enteric fever due to Salmonella spp. with reduced fluoroquinolone susceptibility: a case series	Background Salmonella spp. with reduced susceptibility to fluoroquinolones have higher than usual MICs to these agents but are still considered "susceptible" by NCCLS criteria. Delayed treatment response to fluoroquinolones has been noted, especially in cases of enteric fever due to such strains. We reviewed the ciprofloxacin susceptibility and clinical outcome of our recent enteric fever cases. Methods Salmonella enterica Serotype Typhi ( S . Typhi) and Serotype Paratyphi ( S . Paratyphi) blood culture isolates (1998–2002) were tested against nalidixic acid by disk diffusion (DD) and agar dilution (AD) and to ciprofloxacin by AD using NCCLS methods and interpretive criteria. Reduced fluoroquinolone susceptibility was defined as a ciprofloxacin MIC of 0.125–1.0 mg/L. The clinical records of patients treated with ciprofloxacin for isolates with reduced fluoroquinolone susceptibility were reviewed. Results Seven of 21 (33%) S . Typhi and S . Paratyphi isolates had reduced susceptibility to fluoroquinolones (MIC range 0.125–0.5 mg/L). All 7 were nalidixic acid resistant by DD (no zone) and by AD (MIC 128- >512 mg/L). The other 14 isolates were nalidixic acid susceptible and fully susceptible to ciprofloxacin (MIC range 0.015–0.03 mg/L). Five of the 7 cases were treated initially with oral ciprofloxacin. One patient remained febrile on IV ciprofloxacin until cefotaxime was added, with fever recurrence when cefotaxime was discontinued. Two continued on oral or IV ciprofloxacin alone but had prolonged fevers of 9–10 days duration, one was switched to IV beta-lactam therapy after remaining febrile for 3 days on oral/IV ciprofloxacin and one was treated successfully with oral ciprofloxacin. Four of the 5 required hospitalization. Conclusions Our cases provide further evidence that reduced fluoroquinolone susceptibility of S . Typhi and S . Paratyphi is clinically significant. Laboratories should test extra-intestinal Salmonella spp. for reduced fluoroquinolone susceptibility.	Background Salmonella enterica Serotype Typhi ( S . Typhi) and Serotype Paratyphi ( S . Paratyphi) with reduced susceptibility to fluoroquinolones are common in India and Southeast Asia [ 1 ]. Such isolates have elevated minimum inhibitory concentrations (MICs) to ciprofloxacin and other fluoroquinolones, although they are still considered "susceptible" using current National Committee for Clinical Laboratory Standards (NCCLS) interpretive criteria [ 2 ]. Reduced susceptibility to fluoroquinolones most often arises from point mutations in the quinolone resistance determining region (QRDR) of the gyrA gene which encodes the A subunit of DNA gyrase. These mutations lead to resistance to nalidixic acid, a quinolone, and resistance to this agent can thus be used as an indicator of reduced fluoroquinolone susceptibility [ 3 ]. Although the mechanisms of resistance have been defined, there is still little information on the clinical significance of reduced susceptibility to fluoroquinolones in Salmonella . Limited published data suggest that treatment of these infections with a fluoroquinolone may result in a delay in clinical response and possibly treatment failure [ 2 ]. As a result of this clinical data, NCCLS currently recommends testing of extra-intestinal Salmonella isolates for nalidixic acid resistance as a marker for reduced fluoroquinolone susceptibility [ 4 ]. Others have suggested a re-evaluation of the current fluoroquinolone MIC breakpoints for Salmonella [ 2 ] while some recommend differentiating between "low level resistance" (MICs of 0.125 – 1.0 mg/L) and "high level resistance" (MICs >1.0 mg/L) [ 5 ]. A poor response to ciprofloxacin therapy in a patient with typhoid fever at our hospital prompted us to review retrospectively the ciprofloxacin MICs and clinical outcomes of other enteric fever cases that were treated with a fluoroquinolone. Methods We tested 21 S . Typhi and S . Paratyphi blood culture isolates recovered from patients at The Ottawa Hospital and The Children's Hospital of Eastern Ontario between 1998–2002. Susceptibility testing was performed using NCCLS methods and interpretive criteria [ 4 , 6 , 7 ]. Disk Diffusion testing was performed using nalidixic acid (30 μg), ampicillin (10 μg), trimethoprim-sulfamethoxazole (1.25 μg/23.75 μg), and chloramphenicol (30 μg) disks (Oxoid, U.K.) on Mueller-Hinton agar plates (PML Microbiologicals, Mississauga, ON). MICs to nalidixic acid (Sigma-Aldrich, St-Louis, MO) and ciprofloxacin (Bayer Inc.) were determined by agar dilution using Mueller-Hinton agar (Becton Dickinson and Company, Sparks, MD). Reduced susceptibility to fluoroquinolones was defined as a ciprofloxacin MIC of 0.125–1.0 mg/L [ 3 ]. We sequenced the QRDR region of the gyrA gene for all isolates with reduced susceptibility to fluoroquinolones to detect nucleotide mutations. PCR amplification and DNA sequencing was performed using primers [ 8 ] and conditions previously described [ 9 ]. Sequencing was performed by the Ottawa Genome Centre DNA Sequencing Institute using the Big Dye Terminator v 3.1 method. Sequences were compared to the gyrA sequence of the fully susceptible S. Typhimurium NCTC 74 (accession number X78977, EMBL GenBank database) [ 8 ] and to 3 fully susceptible S . Typhi isolates. Pulsed field gel electrophoresis (PFGE) was performed on isolates with reduced susceptibility to fluoroquinolone using XbaI . Medical records of patients who were treated with ciprofloxacin for infections due to isolates with reduced fluoroquinolone susceptibility were reviewed to determine clinical outcomes. Results Laboratory Seven of 21 isolates or 33% (4 of 12 S . Typhi and 3 of 9 S. Paratyphi) were nalidixic acid resistant by disk diffusion, all with no zones of inhibition. These isolates were all nalidixic acid resistant with an MIC range of 128->512 mg/L, and demonstrated reduced susceptibility to fluoroquinolones (ciprofloxacin MIC range 0.125–0.5 mg/L). The 14 nalidixic acid susceptible isolates by disk diffusion (zone size range 20–25 mm) were also susceptible by agar dilution (MIC range 2–4 mg/L), and none had reduced susceptibility to fluoroquinolones (ciprofloxacin MIC range 0.015–0.03 mg/L). Multidrug resistance was identified in 5 of 21 isolates (24%) with resistance to ampicillin, trimethoprim-sulfamethoxazole, and chloramphenicol. All 7 isolates with reduced susceptibility to fluoroquinolones had nucleotide point mutations in the gyrA gene that resulted in amino acid substitutions: 5 had a Ser83 to Phe change, one had a Ser83 to Tyr change, and one had an Asp87 to Asn change. PFGE of the 7 resistant isolates showed 3 of 4 S . Typhi to have identical patterns. These 3 strains were also multidrug resistant to ampicillin, trimethoprim-sulfamethoxazole and chloramphenicol. The fourth strain, which was susceptible to ampicillin, trimethoprim-sulfamethoxazole and chloramphenicol, was a closely related strain (2 band difference). Two of 3 S. Paratyphi had identical PFGE patterns, with the third being possibly related (4 band difference) [ 10 ]. All 3 isolates of S . Paratyphi were susceptible to the other antibiotics tested. Clinical (Table 1- see Additional file 1 ) Infections with the 7 isolates with reduced susceptibility to fluoroquinolones were all acquired in the Indian subcontinent. Five of the 7 patients with infections due to these isolates were treated initially with ciprofloxacin. The clinical and microbiological findings for these 5 are summarized in Table 1 (see Additional file 1 ). S. Typhi isolates from cases 1, 3 and 4 were also resistant to ampicillin, trimethoprim-sulfamethoxazole and chloramphenicol. The Salmonella isolates from the other 2 cases were susceptible to these antibiotics. Of note, 3 of 4 patients with S . Typhi had fever durations well in excess of the expected 95% upper confidence limit of 3.9 days found in a pooled analysis of multiple studies. These studies involved patients with predominantly nalidixic acid susceptible S . Typhi who were treated with fluoroquinolones [ 1 ]. The patient infected with S . Paratyphi with reduced susceptibility to fluoroquinolone was judged to be failing ciprofloxacin, since there was no clinical improvement after 3 days of treatment. Although fever of this duration is common, patients with enteric fever given fluoroquinolones may show some clinical improvement by this time. Three days with no amelioration of symptoms has been used as indicator of a poor response to therapy in some studies [ 11 , 12 ]. Thus, this patient possibly may have had a poor response to therapy due to the elevated ciprofloxacin MIC. None of the 5 had a recurrence of bacteremia after antibiotics were discontinued, and none died. Discussion Fluoroquinolones have been considered the treatment of choice for enteric fever. As noted, a mean fever clearance time of only 3.9 days (for predominantly nalidixic acid-susceptible strains) has been reported, which is shorter than for other agents [ 1 ]. Calculated clinical and microbiologic failure rates of 2.1% and 0.4%, respectively, are also low in comparison to other antibiotics [ 1 ]. Relapse rates (1.2%) and fecal carriage rates (1.5%) have also been lower than for the traditional drugs, trimethoprim-sulfamethoxazole and chloramphenicol. However, reduced susceptibility to fluoroquinolones jeopardises the usefulness of these agents. It is now seen frequently in S . Typhi and S . Paratyphi isolates acquired in the Indian subcontinent and Southeast Asia [ 1 ], where common clones appear to be circulating. In fact, the common PFGE profile of 3 of our S . Typhi isolates with reduced susceptibility to fluoroquinolones appears to be the same as that described for S . Typhi strains from India, Pakistan, Bangladesh, and Tajikistan [ 13 ]. The gyrA point mutations in our isolates have also been previously described in isolates from these countries [ 1 , 3 ]. Several lines of evidence support the clinical significance of elevated fluoroquinolone MICs in Salmonella spp., including case reports, animal experiments, and pharmacodynamic modelling [ 2 ]. In a report of patients treated with short courses of oral ofloxacin [ 11 ], the median time to fever clearance was greater for patients infected with nalidixic acid resistant S . Typhi than for those with nalidixic acid susceptible strains. In addition, one third of nalidixic acid resistant infections required re-treatment vs. 0.8% of infections due to susceptible strains. The authors concluded that short courses (less than 5 days) of oral fluoroquinolone therapy should not be used for treating nalidixic acid resistant isolates. Our small series adds to the evidence that there is a delayed clinical response to fluoroquinolone therapy in this setting, and demonstrates that fever may be prolonged even with long courses of fluoroquinolone therapy, and even when given parenterally in some cases. Until results from randomized controlled trials of enteric fever due to strains with reduced susceptibility to fluoroquinolones are available, the best treatment regimen is uncertain. Current options include higher dose and longer duration of fluoroquinolones, 3rd generation cephalosporins and azithromycin, either alone or in combination [ 1 ]. If susceptible, agents such as ampicillin, trimethoprim-sulfamethoxazole, and chloramphenicol may also be considered, but rates of resistance to these agents are generally too high to recommend as first-line empiric therapies [ 1 ]. In our series, resistance to all these agents was high among the isolates with reduced fluoroquinolone susceptibility (3 of 7 or 43%). When fluoroquinolones are given, it is important to bear in mind that two pharmacokinetic/pharmacodynamic ratios are predictive of successful treatment with these agents [ 2 ]. These are the ratio of peak serum antimicrobial level to the MIC, and the ratio of the 24-hour area under the serum concentration-versus-time curve to the MIC (AUC/MIC). Higher dosing will maximize these ratios, and thus when oral ciprofloxacin is given for suspected enteric fever, use of 750 mg rather than 500 mg bid seems logical. It is important that microbiology laboratories have procedures to detect Salmonella strains that have reduced susceptibility to fluoroquinolones. The current NCCLS guidelines (2004) recommend that all extra-intestinal isolates of Salmonella be tested for resistance to nalidixic acid in order to detect reduced fluoroquinolone susceptibility [ 4 ]. In addition, it is recommended that physicians be informed that isolates that test fluoroquinolone "susceptible" but nalidixic acid resistant may not be eradicated with fluoroquinolone therapy. There have also been suggestions to change the NCCLS breakpoints to reflect the risk of treatment failure of Salmonella spp. with reduced susceptibility to fluoroquinolones [ 1 ]. This is clearly a difficult decision, since fluoroquinolones have previously been the most active antibiotic class for treatment of enteric fever. Some patients with these isolates may do poorly on short courses and lower doses of fluoroquinolones but potentially could respond adequately to higher doses and longer durations of treatment. Rather than reclassifying all isolates with reduced susceptibility to fluoroquinolones as being resistant to ciprofloxacin, others have recommended differentiating between low-level (MIC of 0.125 – 1.0 mg/L) and high-level (MIC >1.0 mg/L) ciprofloxacin resistance based on the MIC range [ 5 ]. It is not completely clear at present which of these is the optimal approach for laboratory detection of Salmonella spp. with reduced fluoroquinolone susceptibility. Not all laboratories routinely perform MICs and many rely on commercially-available automated susceptibility methods that cannot currently detect reduced fluoroquinolone susceptibility. The current NCCLS recommendations to test for nalidixic resistance are easy to implement for routine testing but may have reduced specificity [ 3 ]. Laboratories need to decide which approach is best for their workflow and available resources. Conclusions Given the accumulating evidence, including our own clinical experience, enteric fever due to Salmonella spp. with reduced susceptibility to fluoroquinolones is clinically significant. It is important that laboratories test S. Typhi and S . Paratyphi, as well as other extra-intestinal Salmonella isolates, for reduced susceptibility to fluoroquinolones. Competing interests None declared. Authors' contributions RS: participated in the design of study and authored first draft; MD: conceived study, coordinated susceptibility testing, and revised manuscript; AM: conceived study and reviewed clinical data; KR: participated in study design and coordinated pulsed field gel analysis; PJ: reviewed clinical data and helped coordinate susceptibilty testing; CB: carried out molecular genetic studies and participated in sequence alignment; BT: conceived study and supervised study, revised manuscript All authors read and approved of the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 Clinical and microbiological information for patients treated with ciprofloxacin for S. Typhi or S. Paratyphi isolates with reduced fluoroquinolone susceptibility provides information regarding clinical course for patients with isolates with reduced fluoroquinolone susceptibility as well as antibiotic susceptibility and gene mutation results for these isolates Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC521077.xml
550671	Propagated repolarization of simulated action potentials in cardiac muscle and smooth muscle	Background Propagation of repolarization is a phenomenon that occurs in cardiac muscle. We wanted to test whether this phenomenon would also occur in our model of simulated action potentials (APs) of cardiac muscle (CM) and smooth muscle (SM) generated with the PSpice program. Methods A linear chain of 5 cells was used, with intracellular stimulation of cell #1 for the antegrade propagation and of cell #5 for the retrograde propagation. The hyperpolarizing stimulus parameters applied for termination of the AP in cell #5 were varied over a wide range in order to generate strength / duration (S/D) curves. Because it was not possible to insert a second "black box" (voltage-controlled current source) into the basic units representing segments of excitable membrane that would allow the cells to respond to small hyperpolarizing voltages, gap-junction (g.j.) channels had to be inserted between the cells, represented by inserting a resistor (R gj ) across the four cell junctions. Results Application of sufficient hyperpolarizing current to cell #5 to bring its membrane potential (V m ) to within the range of the sigmoidal curve of the Na + conductance (CM) or Ca ++ conductance (SM) terminated the AP in cell #5 in an all-or-none fashion. If there were no g.j. channels (R gj = ∞), then only cell #5 repolarized to its stable resting potential (RP; -80 mV for CM and -55 mV for SM). The positive junctional cleft potential (V JC ) produced only a small hyperpolarization of cell #4. However, if many g.j. channels were inserted, more hyperpolarizing current was required (for a constant duration) to repolarize cell #5, but repolarization then propagated into cells 4, 3, 2, and 1. When duration of the pulses was varied, a typical S/D curve, characteristic of excitable membranes, was produced. The chronaxie measured from the S/D curve was about 1.0 ms, similar to that obtained for muscle membranes. Conclusions These experiments demonstrate that normal antegrade propagation of excitation can occur in the complete absence of g.j. channels, and therefore no low-resistance pathways between cells, by the electric field (negative V JC ) developed in the narrow junctional clefts. Because it was not possible to insert a second black-box into the basic units that would allow the cells to respond to small hyperpolarizing voltages, only cell #5 (the cell injected with hyperpolarizing pulses) repolarized in an all-or-none manner. But addition of many g.j. channels allowed repolarization to propagate in a retrograde direction over all 5 cells.	Introduction There are no low-resistance connections between the cells in several different cardiac muscle and smooth muscle preparations [reviewed in refs. [ 1 ] and [ 2 ]]. In a computer simulation study of propagation in cardiac muscle, it was shown that the electric field (EF) that is generated in the narrow junctional clefts, when the prejunctional membrane fires an action potential (AP), depolarizes the postjunctional membrane to its threshold [ 3 - 5 ]. Propagation by mechanisms not requiring low-resistance connections have also been proposed by others [ 6 - 9 ]. This results in excitation of the postjunctional cell, after a brief junctional delay. The total propagation time consists primarily of the summed junctional delays. This results in a staircase-shaped propagation, the surface sarcolemma of each cell firing almost simultaneously [ 4 ]. Propagation has been demonstrated to be discontinuous (or saltatory) in cardiac muscle [ 10 - 13 ]. Fast Na + channels are localized in the junctional membranes of the intercalated disks of cardiac muscle [ 5 , 14 , 15 ], a requirement for the EF mechanism to work [ 1 - 5 ]. We recently modeled propagation of APs of cardiac muscle and smooth muscle using the PSpice program for circuit design and analysis [ 16 - 18 ]. Like the mathematical simulation published in 1977 [ 3 ] and 1991 [ 4 ], the EF developed in the junctional clefts (negative V JC ) was large and sufficient to allow transfer of excitation to the contiguous cell, without the requirement of gap-junction (g.j.) channels. Propagation of excitation can occur by the EF mechanism alone, even when the excitability of the cells was made low. In connexin-43 (heterozygous) and Cx40 knockout mice, propagation in the heart still occurs, but it is slowed [ 19 - 22 ] as predicted by our PSpice simulation study [ 18 ]. The present experiments were carried out to study propagated repolarization in this model of simulated action potentials (APs). Propagation of repolarization is a phenomenon that occurs in cardiac muscle [ 23 ]. It has been shown that propagation of vasodilation occurs in the microvasculature [ 24 ], and that the endothelial cells are involved in the conduction of hyperpolarization and vasodilation in an artery [ 25 ]. Therefore, our hypothesis was that propagated repolarization would also occur in our PSpice model. Methods The methods used and PSpice program (Cadence Co, Portland) have been described in detail previously, including the circuit [ 17 , 18 ]. In brief, each cell was represented by four basic excitable units, two for the long surface membrane of the cell (one upward-facing and one downward-facing) and one basic unit for each of the two junctional membranes (left end of cell and right end) (Fig 1 ). The radial (shunt) resistance of the junctional cleft (R JC ) was placed in the junctions between adjoining cells. The basic units were connected internally by the intracellular longitudinal resistance (r i ). The basic units were connected externally with the extracellular resistance (R O ), broken down into a longitudinal component (R ol ) and a transverse (radial) component (R or ). R O was connected to ground as depicted in Figure 1 . The circuit used for each unit was kept as simple as possible, using only those ion channels that set the resting potential (RP) and predominate during the rising phase and plateau phase of the AP. Figure 1 Cardiac Muscle and Smooth Muscle. Circuit diagram used for study of propagated repolarization in cardiac muscle and smooth muscle. A: 5-cell chain. A depolarizing stimulating pulse (I S1 ; 0.50 ms, 0.25 nA) was applied to the inside of the first cell (A1; left side). A hyperpolarizing pulse (I S2 ; variable intensity and duration) was applied to the inside of the fifth cell (A5; right side) a few milliseconds later when the action potentials (APs) initiated by I S1 were in their plateau phase (peak overshoot). B: Enlarged diagram to show a portion of the circuit for details of the basic units. The myocardial cell was assumed to be a cylinder 150 μm long and 16 μm in diameter, and the smooth muscle cell a cylinder 200 μm long and 5 μm diameter. Since in vascular smooth muscle (VSM), the muscle fibers run in a circular direction, if transverse velocity is calculated, the fiber diameter should be used. The values of the capacitive and the resistive elements in each basic unit were set to reflect the input resistance (ca 20 MΩ) and input capacitance (ca 100 pF) of the individual cells, and the junctional units were prorated, with respect to the surface units, based on relative areas represented. At rest, the resistance of K + compared to Na + (cardiac muscle) or Ca ++ (smooth muscle) were set to give resting potentials (RPs) of -80 mV for cardiac muscle and -55 mV for smooth muscle. During excitation, the action potentials (APs) overshot to +32 mV and +11 mV, respectively. Electrical stimulations (I S1 ) were always applied internally to the first cell of the chain (cell A1). Rectangular depolarizing current pulses of 0.25 nA amplitude and 0.50 ms duration were applied. The delay time before the I S1 pulse was applied was usually set to 1.0 ms in SM. A second stimulus (I S2 ) that was hyperpolarizing was applied to the inside of the last cell (A5) of the chain when the APs of all 5 cells were in their plateau phase. The intensity and duration of the I S2 pulses were varied over a wide range in order to generate strength / duration (S/D) curves. Because the PSpice program does not have a voltage-dependent resistance (to generate the increase in Na + or Ca ++ conductance during excitation), this function had to be done with a V-controlled current source (our "black-box"). The sigmoidal relationship between conductance and membrane potential (V M ), over a relatively narrow V M range, was mimicked by the black-box. The Na + or Ca ++ current required for excitation had to be calculated for several V M values and inserted into the GTABLE function. Experiments were done with a single chain of 5 cells or 2 cells. There were no gap junctions between the cells of the chain under initial conditions. The presence of gap junction connexons (tunnels) was represented by adding a variable shunt resistance (R gj ) across each cell-to-cell junction. This resistor connected the inner surface of the prejunctional membrane with the inner surface of the postjunctional membrane. This R gj shunt resistance was varied between 10,000 MΩ (1 tunnel), 1000 MΩ (10 tunnels in parallel), 100 MΩ (100 tunnels), 10 MΩ (1,000 tunnels), and 1.0 MΩ (10,000 tunnels). Each tunnel was assumed to have a conductance of 100 pS. Results A. All-or- None Repolarization of Stimulated Cell A5 There was a sharp (all-or-none) repolarization of the stimulated cell (A5) of the 5-cell chain in both cardiac muscle (Fig. 2AB ) and smooth muscle (Fig. 2CD ). As shown, stimulation of cell A1 with a depolarizing current pulse (I S1 ) produced propagation of APs down the chain. At the plateau (peak) of the APs, a repolarizing pulse applied intracelluarly to cell A5, if of sufficient intensity (duration constant), produced a sudden repolarization of only cell A5 (Fig. 2B for cardiac muscle and D for smooth muscle). A slightly lower current intensity failed to produce a stable repolarization of cell A5 (Fig. 2A for cardiac muscle and 2C for smooth muscle). Note that the potential change (repolarizing) produced in neighboring call A4 was very small (< 1 mV). This emphasizes that there are indeed no low-resistance connections between the modeled cells under standard conditions. The hyperpolarizing pulse had to bring the V m of cell A5 into the region of the GTABLE's sigmoidal curve. The transient repolarization is in agreement with the biological case [ 23 - 25 ]. Figure 2 Sharp Repolarization of Stimulated Cell (A5). Sharp repolarization of only the last cell (5 th ) of the 5-cell chain when a repolarizing I S2 pulse was applied in cardiac muscle ( A-B ) and in smooth muscle ( C-D ). Panels A and C illustrate the records obtained when the applied I S2 pulse was just not quite strong enough to produce a permanent repolarization of cell #5. In panels B and D , the I S2 intensity was slightly increased to produce an all-or-none repolarization. The membrane potential of adjacent cell #4 (A4) was only slightly changed when cell A5 underwent a very large change. The velocity of antegrade propagation (θ a ) was about 54 cm/sec in CM and 8.9 cm/sec in SM under that conditions. B. Propagation of Repolarization As indicated in the Methods section, it was not possible to insert a second black-box in the K + leg of the basic circuit, because the PSpice program became erratic. Therefore, in order to achieve propagation of the repolarization of cell A5 in the retrograde direction, it was necessary to insert gap-junction channels between the cells of the chain (1, 10, 100, 1000, 10000 channels). This corresponded to adding resistive shunts between the cells across the junctions (R gj ) of 10000, 1000, 100, 10, and 1.0 MΩ (assuming each channel has a conductance of 100 pS). The results of doing such an experiment are shown in Fig 3 for cardiac muscle (A – C) and for smooth muscle (D – F). When there were many channels (e.g. 10,000 in Fig. 3A and 3D or 1000 in Fig. 3B and 3E ), the rising phase of the APs of all 5 cells were superimposed. This means that all 5 cells fired nearly simultaneously, as expected because of the high degree of low-resistance coupling. However, when a repolarizing current pulse was applied to cell A5, its repolarization spread to the neighboring cells. But the other cells did not repolarize simultaneously, as can be seen. Instead, there was a propagation of the repolarization at a certain velocity. This repolarization velocity became slower and slower as the number of channels was decreased. For example, with 100 channels (Fig. 3C and 3F ), the propagated repolarization velocity was slower than with 1000 channels (Fig. 3B and 3E ) or 10,000 channels (3A and 3D). With only 10 channels, the repolarization did not persist in either cardiac muscle or smooth muscle (not illustrated). Figure 3 Insertion of gap junction channels. Propagation of the repolarization of cell A5 was produced when sufficient gap-junction (g.j) channels were inserted between the cardiac muscle cells and smooth muscle cells. A: Record obtained when 10,000 gj-channels were inserted (equivalent to a g.j. resistance (R gj ) of 1.0 MΩ). Note that the rising phase of the APs from all 5 cells were superimposed, indicating that they all fired simultaneously. Also note that repolarization propagated in a retrograde direction down the 5-cell chain. B: 1,000 gj-channels inserted (R gj of 10 MΩ). Again, the rising phase of the APs of the 5 cells were nearly superimposed. C: 100 gj-channels (R gj of 100 MΩ). With less coupling, the rising phase of the APs of the 5 cells were separated in time. The velocity of propagated repolarization (θ r ) was further slowed. D: R gj = 1.0 MΩ(10,000 channels). The rising phase of the APs from all 5 cells were superimposed. Retrograde propagation of repolarization was very fast. E: R gj = 10 MΩ(1000 channels). The rising phase of the 5 APs were still superimposed, but now the retrograde propagation velocity was slowed. F: R gj j = 100 MΩ(100 channels). The rising phase of the 5 APs are now separated, indicating velocity of antegrade propagation (θ a ) was slowed. Velocity of retrograde propagation (θ r ) was slow. C. Strength/Duration Curves The intensity (strength) and duration of the rectangular hyperpolarizing current pulses (I S2 ) applied to cell A5 were varied over a wide range in order to generate strength / duration curves. This was done when R gj was infinite (i.e., 0 channels) and when R gj was 10 MΩ (1000 channels) for strong coupling. The pulse duration was initially constant at 1.0 ms (near the chronaxie value) and then lowered to 0.5 ms and to 0.25 ms. The current intensity was varied until the sharp endpoint occurred, namely the stable repolarization of all cells in the chain. These results are plotted in Figure 4 for cardiac muscle and smooth muscle. Panel A is the strength / duration (S / D) curve for when R gj was infinite (0 channels), and Panel B is the S/D curve for when R gj was 10 MΩ(1000 channels). Note that the I S2 intensity was about 8–10-fold greater when the cells were well-coupled, because the applied hyperpolarizing current had to spread to all 5 cells of the chain. Regardless, the chronaxie values were about the same (ca. 1.0 ms). Figure 4 Strength/Duration Curves. Strength / duration (S/D) curves for cardiac muscle cells (filled circles) and smooth muscle cells (unfilled circles) (5-cell chains) when R gj was ∞ (0 channels) ( A ) or when R gj was 10 MΩ(1000 channels) ( B ). The S / D curves are rectangular hyperbolas. The time (pulse duration) it takes for a current intensity of twice the rheobasic intensity to produce the all-or-none repolarization is the chronaxie (σ) The rheobase is the asymptote of the data points extrapolated back to the ordinate, as shown. The chronaxie was about 1.0 ms, in both panels A and B . But the absolute current intensity required was about 8–10-fold greater in panel B compared to panel A . The membrane time constant (τ m ) is related to the chronaxie (σ) by the equation shown in panel A . Discussion In principle, the addition of a second black-box into the K + leg of the basic circuit would allow the cell to repolarize in an all-or-none fashion to small repolarizing currents. When this was attempted, the program behaved erratically. So in the absence of g.j. channels, only the cell (A5) injected with repolarizing current (I S2 ) was able to repolarize in an all-or-none manner. The neighboring cell (A4) exhibited only a slight repolarization of <1 mV when cell A5 had repolarized completely back to the RP (-80 mV for CM and -55 mV for SM). This fact emphasized that there were no low-resistance connections between the cells under our initial conditions. However, addition of 10,000, 1,000, or 100 g.j channels (corresponding to R gj values of 1.0, 10, and 100 MΩ) did allow propagation of repolarization to occur. The borderline value was 10 g.j. channels (1000 MΩ R gj ), e.g., repolarization propagated part-way down the chain in SM and almost succeeded in CM. Of course, inserting the g.j. channels required that the I S2 repolarizing current applied be much greater. This is because the I S2 current had to spread down the entire chain, with the threshold current required to cause all cells to repolarize being determined by sufficient current entering distal cell A1 to repolarize it to the GTABLE sigmoidal region. Thus the proximal cells, like A5 and A4, became hyperpolarized beyond the level required for their repolarization. The repolarizing I S2 current intensity required for the all-or-none repolarization was lower when the rectangular pulse duration was increased. This was true for both when only the injected cell A5 was repolarized (R gj = ∞) and when all 5 cells repolarized (R gj of 1.0, 10, and 100 MΩ). Plots of current intensity (ordinate) versus current duration (abscissa) gave the typical hyperbolic strength/ duration curve for excitable membranes. The chronaxie values were about 1.0 ms, for which a time constant τ m of about 1.44 ms could be calculated. The S/D curves for the two conditions (R gj = ∞ and R gj = 10 MΩ) show that the current intensity required was about 8–10-fold greater when there were many gj-channels, in both CM and SM. The calculated velocity for propagated repolarization (θ r ) varied with the number of gj-channels (Table 1 ), as expected. The more channels, the faster the velocity. For cardiac muscle, the θ r was about 200 cm/s in the very well coupled case (10,000 channels) and about 50 cm/s in the less coupled case (100 channels) (Table 1 ). In all cases, the velocity for propagated repolarization (θ r ) was much lower than the velocity for antegrade propagation (θ a .). In the 2-cell chain, the calculated velocities of propagated repolarization were similar to those for the 5-cell chain (Table 1 ). Table 1 Calculated velocity of retrograde propagation (θ r ) as compared to that for antegrade propagation (θ a ) for cardiac muscle (CM) and smooth muscle (SM). 2-Cell Chain (Cardiac) 5-Cell Chain (cm/sec) No. of GJ-Channels Rgj (MΩ) Threshold* (nA) θ r (cm/s) CM SM θ r θ a θ r θ a 0 ∞ # # ## 32 ## 3.7 1 10,000 # # ## 38 ## 6.8 10 1,000 95.4 7.1 ## 55 ## 13 100 100 15.8 30 50 115 73 42 1000 10 5.4 75 86 550 114 820 10,000 1 5.6 750 200 3000 400 3600 * Pulse duration was held constant at 1.0 ms. # Second cell (A1) failed to repolarize. # # Some cells failed to repolarize. Velocities of antegrade propagation (θ a ) were taken from previously published data [17,18]. The present study provides some new and important information about the PSpice simulations. First, it verifies that propagation (orthodromic) can occur in the complete absence of gap-junction channels, as previously reported [ 3 , 4 , 16 - 18 ]. Second, it demonstrates for the first time that activation of Na+ (in CM) or Ca++ (in SM) channels is reversible, by bringing V m back to the level of the sigmoidal activation curve (GTABLE). Third, it shows for the first time that, in the PSpice model, the membranes exhibit the characteristic strength/duration curves. Fourth, it shows that the PSpice program has some serious limitations. In summary, because of technical difficulties with the PSpice program, it was necessary to insert gj-channels in order to produce propagation of repolarization. Otherwise, only the modeled cell injected (A5) with the repolarizing I S2 current was able to repolarize. Since the potential change in the neighboring cell was only about 1 mV or less, this emphasizes that there were no low-resistance connections between the simulated cells under initial conditions. Propagation in the orthodromic direction occurs by the electric field (EF) discussed in previous papers (1–4, 16–18). The repolarizing I S2 current gave S/D plots that were typical rectangular hyperbolic curves for excitable membranes, with chronaxie values of about 1.0 ms, both for CM and SM. The calculated velocity for propagated repolarization was greater when the number of gj-channels was increased. The antidromic (retrograde) propagation velocity was usually considerably slower than the orthodromic (antegrade) propagation velocity for depolarization. The present findings do not necessarily imply that, in biological tissue, gap junctions are required for propagated repolarization to occur.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC550671.xml
546007	Nonsteroidal anti-inflammatory drug use and Alzheimer's disease risk: the MIRAGE Study	Background Nonsteroidal anti-inflammatory drugs (NSAID) use may protect against Alzheimer's disease (AD) risk. We sought examine the association between NSAID use and risk of AD, and potential effect modification by APOE-ε4 carrier status and ethnicity. Methods The MIRAGE Study is a multi-center family study of genetic and environmental risk factors for AD. Subjects comprised 691 AD patients (probands) and 973 family members enrolled at 15 research centers between 1996 and 2002. The primary independent and dependent variables were prior NSAID use and AD case status, respectively. We stratified the dataset in order to evaluate whether the association between NSAID use and AD was similar in APOE-ε4 carriers and non-carriers. Ethnicity was similarly examined as an effect modifier. Results NSAID use was less frequent in cases compared to controls in the overall sample (adjusted OR = 0.64; 95% CI = 0.38–1.05). The benefit of NSAID use appeared more pronounced among APOE-ε4 carriers (adjusted OR = 0.49; 95% CI = 0.24–0.98) compared to non-carriers, although this association was not statistically significant. The pattern of association was similar in Caucasian and African Americans. Conclusions NSAID use is inversely associated with AD and may be modified by APOE genotype. Prospective studies and clinical trials of sufficient power to detect effect modification by APOE-ε4 carrier status are needed.	Background Several cross-sectional [ 1 , 2 ], case-control [ 3 - 6 ], and prospective studies [ 7 - 9 ] have reported an inverse association between nonsteroidal anti-inflammatory drug (NSAID) use and the risk of Alzheimer's disease (AD), whereas others [ 10 - 12 ] have not. In this report, we present results of analyses of data from the Multi-Institutional Research in Alzheimer's Genetic Epidemiology (MIRAGE) Study in which we examined potential effect modification by APOE-ε4 carrier status and ethnicity on this association. Methods Subjects and data collection The MIRAGE Study is a multi-center family study of genetic and environmental risk factors for AD sponsored by the National Institute on Aging since 1991. The details of MIRAGE Study data collection procedures, protocols for obtaining family histories, and reports of validity studies of the MIRAGE questionnaires have been published elsewhere. [ 13 - 15 ] Briefly, families were recruited through probands meeting NINCDS-ADRDA criteria [ 16 ] for probable or definite AD who were ascertained through research registries and memory clinics. After obtaining informed consent from non-demented family members, and a combination of consent or assent – along with informed consent by proxy – on living demented subjects, questions eliciting demographic data and information about presumptive risk factors for AD were obtained using standardized MIRAGE questionnaires. Questions pertaining to NSAID use were added to the questionnaire in 1996, and the data presented in this report were collected from May, 1996 through May, 2002. Questions about the proband were answered by a surrogate source within the family, typically the spouse or adult offspring. The same information was sought on non-demented first-degree family members of these probands over 50 years of age, usually a sibling or spouse (less commonly parents or children). 1020 family members in this analysis claimed to be cognitively normal, or were reported by family informants to be dementia-free. Of these, 982 were evaluated using the modified Telephone Interview of Cognitive Status (mTICS) [ 17 , 18 ], and normal cognitive status was confirmed in 973 (99.1%). Information on both patients and first-degree family members was supplemented where available by multiple informants, and medical and nursing home records. To elicit information on prior NSAID use, the following question was asked: "Have you ever taken a nonsteroidal anti-inflammatory medication (e.g. Advil, Motrin, etc.) on a daily basis for more than 6 months ?" No distinction was made between aspirin and other classes of NSAIDs. For proxy reporting about a relative with AD, the question substituted "your relative" for "you". For any affirmative answer, a follow-up question asked for the dates at which the medications were first used and the names of all NSAIDs that had been used. A discrete "index date" was established within each family corresponding to the earliest date that the family or medical records reported AD symptoms to have begun in the proband. Subjects from each family (whether AD cases or non-demented family members) were considered to have been exposed to NSAIDs only if the starting date for NSAID use preceded this index date by at least one year. Age represented the age of cases and of non-demented relatives at the index date, and was treated as a continuous variable. As shown in Figure 1 , there were 756 probands and 1020 relatives over the age of 50 with APOE genotype who were queried about prior NSAID use. After exclusions for those subjects who had missing or unsure responses for the name of their medication, did not include a medication start date, or had missing data for the variables age, sex, education or ethnicity, there remained for analysis 682 probands and 982 relatives. Of the 982 relatives, nine were reported to be demented with the onset of their dementia prior to the index date for that family, and their diagnoses were verified by review of medical records as having probable or definite AD by research criteria, so these were classified with the probands as having AD. Figure 1 MIRAGE subjects ≥ 50 years (adjusted for family index date) with APOE genotype who completed the personal history questionnaire. Statistical analysis Analyses were performed using SAS version 8.2. The primary independent and dependent variables were prior NSAID use and AD case status, respectively. Crude odds ratios were computed in the first instance, followed by adjusted estimates using generalized estimating equation (GEE) models [ 19 ] to account for the possibility that variables of interest (e.g., medication use, APOE status) could be correlated among individuals within families. Adjustments were made for the following covariates: age, sex, ethnicity (categorized as White, African-American, or other), education (less than versus equal to or greater than high school level), and APOE-ε4 carrier status (one or two ε4 alleles vs. none). We stratified the dataset in order to evaluate whether the association between NSAID use and AD was similar in APOE-ε4 carriers and non-carriers. In addition, we formally evaluated these associations by adding an interaction term (ε4 * NSAID use) to the GEE model. Ethnicity was similarly examined as an effect modifier. Results Characteristics of the 1664 subjects are listed in Table 1 . AD patients were more likely to be older and to be APOE-ε4 carriers compared to controls. The distributions of sex and education were not different between cases and controls. Table 1 Characteristics of AD patients and non-demented family CHARACTERISTIC AD (N = 691) NON DEMENTED (N = 973) AGE ADJUSTED PERCENT AD AGE ADJUSTED PERCENT NON-DEMENTED P-VALUE* Mean Age (SD) 70.0 (8.2) 65.0 (8.8) <0.0001 Sex (%male) 242 (35.0) 381 (39.2) 36.5 39.9 0.39 Greater than HS Ed (%) 403 (58.3) 643 (66.1) 59.3 65.1 0.10 African American (%) 215 (31.1) 204 (21.0) 28.4 21.3 0.01 Use of NSAIDs† (%) 24 (3.5) 66 (6.8) 3.5 6.7 0.08 APOE-ε4 carrier (%) 448 (64.8) 370 (38.0) 65.3 38.0 <0.0001 * Reported p-values use General Estimating Equations to account for correlation among observations. †Exposure to non-steroidal anti-inflammatory drugs had to precede index date by one year or more. Sixty-six out of 973 non-demented relatives (6.8%) and 24 of 691 cases (3.5%) reported previous NSAID use (odds ratio = 0.49; 95% CI = 0.31–0.80). After adjustment for age, sex, educational level, and ethnicity, the odds ratio (OR) of NSAID use among AD cases compared to non-users was 0.57 (95% CI = 0.35–0.93); it was 0.64 (95% CI = 0.38–1.05) when APOE carrier status was added to the GEE model (see Table 2 ). Table 2 Risk of AD with and without prior use of NSAIDs, stratified by APOE-ε4 carrier status EXPOSURE AD (N = 691) NON-DEMENTED FAMILY MEMBERS (N = 973) CRUDE ODDS RATIO (95% CI) AGE-ADJUSTED ODDS RATIO (95% CI) ADJUSTED ODDS RATIO (95% CI) Overall No NSAIDs 667 907 1.0 1.0 1.0 Use of NSAIDs 24 66 0.49 (0.31, 0.80) 0.55 (0.34, 0.88) 0.64 (0.38, 1.05)* Having no ε4 alleles No NSAIDs 231 556 1.0 1.0 1.0 Use of NSAIDs 12 47 0.61 (0.32, 1.18) 0.75 (0.38, 1.46) 0.78 (0.39, 1.52) Having at least one ε4 allele No NSAIDs 436 351 1.0 1.0 1.0 Use of NSAIDs 12 19 0.51 (0.24, 1.06) 0.47 (0.24, 0.96) 0.49 (0.24, 0.98) Adjusted for age, sex, ethnicity, education, and APOE-ε4 status. Adjusted for age, sex, ethnicity, and education. *NSAIDAPOE-ε4 interaction p-value = 0.04. The magnitude of inverse association was greater among APOE ε4 carriers (OR = 0.49; 95% CI = 0.24–0.98) than non-carriers (OR = 0.78; 95% CI = 0.39–1.52). However, formal evaluation of the interaction between NSAID use and APOE-ε4 carrier status did not reveal a significant difference (p = 0.40). The association between NSAID use and AD risk was similar among Caucasian and African Americans (data not shown). Discussion This study supports the findings of previous reports [ 1 - 6 ] suggesting that use of NSAIDs for at least six months is associated with a reduced risk of AD. This association appears to be more robust among APOE-ε4 carriers than non-carriers, although the difference in associations between these two groups was not statistically significant. The MIRAGE Study includes the largest number of well-characterized AD cases and family controls to date, and this large sample size permits adjustment for important potential confounders, as well as the power to examine effect modification by APOE genotype and ethnicity. The subjects without dementia were first-degree family members of AD cases, providing some degree of informal matching on age, socioeconomic status, and health-seeking behavior. However, these results must be interpreted in light of some methodological limitations. Data on NSAID use was collected with a single retrospective question that did not distinguish between aspirin use and non-aspirin NSAID use. Moreover, while non-demented participants reported on themselves, a proxy historian reported on most of the demented individuals. Differential reporting is a potential source of bias in a study that uses self-report on most of the non-demented subjects, yet relies on surrogate respondents for all of the subjects with AD. Asymmetric data collection is difficult to avoid when cases are cognitively impaired, but may be more accurate than expected in AD patients where the surrogate historian has a long association with the subject. We addressed this potential bias by performing an independent validation study to determine the accuracy of surrogate information on a number of questions, including the same questions used in this report about NSAID use. [ 15 ] This study found substantial reliability on the NSAID item (kappa = 0.70). While a validation study comparing proxy historians for non-demented persons does not perfectly mirror the situation in which proxy historians report on demented individuals, our study revealed excellent concordance for surrogate responses from most categories of relatives. This result is consistent with those of prior studies which found a similar association despite differences in study design (cross-sectional [ 1 , 2 ] vs. case-control [ 3 - 6 , 10 , 11 ] vs. prospective [ 7 - 9 , 12 ]), sampling frame (family members [ 10 ] vs. registry-based [ 11 ] vs. general population [ 1 - 9 , 12 ]), ascertainment of exposure, type of medication considered (aspirin [ 2 - 4 , 6 , 8 - 10 , 12 ] vs. non-aspirin NSAIDs [ 1 - 11 ] vs. 'any' NSAID [ 4 , 8 ]), duration of exposure (current [ 1 - 3 ] vs. any history of use, duration ranging from a week to at least six months [ 3 , 5 - 12 ]), and degree of matching or adjustment (usually adjusted for age, sex, and education, less frequently APOE genotype [ 7 - 9 , 12 ]). While many studies have examined the association between NSAID use and risk of AD, few have examined the impact of APOE genotype on this association. The Cache County Study [ 9 , 20 ], the Canadian Study for Health and Aging (CSHA) [ 8 ], and the Rotterdam Study [ 7 ] adjusted for APOE and tested for effect modification and found none. But they had fewer AD cases, and, in the CSHA had a smaller proportion of genotyped subjects. The Rotterdam Study [ 7 ] reported separate odds ratios for APOE-ε4 carriers and non-carriers, but this sample did not have any subjects who were both APOE-ε4 carriers and who reported long-term use of NSAIDs. They found no difference in risk between those with at least one ε4 allele compared to ε4 non-carriers among subjects who used NSAIDs between one month and two years. Our data suggest an enhanced protective benefit of NSAID use among those with ε4. A smaller protective effect was also evident among those lacking ε4. In our sample there were relatively few AD cases who were not ε4 carriers, thus the appearance of different patterns of association between NSAID use and AD risk among APOE genotype subgroups may be spurious. This difference could also have arisen as a result of bias and confounding. The genotype-specific association could be explained by differential inclusion of subjects into the study on the basis of APOE-ε4 carrier status and NSAID use. This might occur if there were differential mortality, according to APOE genotype, among those with AD who had a history of NSAID use; or if for any reason among NSAID users APOE-ε4 carriers were less likely than non-carriers to be diagnosed with AD (or conversely, if among non-users of NSAIDs AD was more likely diagnosed in APOE-ε4 carriers compared to non-carriers). Differential recall could also give rise to this observation. However, these explanations are unlikely because subjects were not selected on the basis of APOE genotype. It is also possible that APOE-ε4 carrier status is a proxy for differentially distributed unmeasured confounders related to NSAID use such as inflammatory disease processes. Alternatively, our results imply that NSAID use affects AD risk differently between APOE-ε4 carriers and non-carriers. For example, because ε4 carriers are inherently more vulnerable to AD, there is a greater opportunity for attributable risk reduction. This explanation does not imply biological interaction between NSAIDs and ε4. On the other hand, the ε4 isoform may have greater pro-inflammatory properties [ 21 ] and ε4 individuals may be more responsive to the benefits of NSAID use than those lacking ε4. Examination of this finding in prospective studies and clinical trials of sufficient power (such as the ADAPT Study [ 22 ], a prospective trial of anti-inflammatory use in the prevention of AD) to detect effect modification by APOE-ε4 carrier status is needed. Such confirmation would provide critical insights into the mechanisms by which APOE isoforms modulate AD risk and into novel therapeutic strategies. Conclusions NSAID use is inversely associated with AD and may be modified by APOE genotype. Prospective studies and clinical trials of sufficient power to detect effect modification by APOE-ε4 carrier status are needed. Competing interests The author(s) declare that they have no competing interests. Authors' contributions Study concept and design (LAF, RCG, LAC); acquisition of data (LAF, RCG, MIRAGE investigators); analysis and interpretation of data (AGY, MH); drafting of the manuscript (AGY, RCG, LAF); critical revision of the manuscript for important intellectual content (AGY, LAF, RCG, LAC); statistical expertise (LAC); obtained funding (LAF, RCG, LAC, MIRAGE investigators). All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546007.xml
550659	Comparison of seven methods for producing Affymetrix expression scores based on False Discovery Rates in disease profiling data	Background A critical step in processing oligonucleotide microarray data is combining the information in multiple probes to produce a single number that best captures the expression level of a RNA transcript. Several systematic studies comparing multiple methods for array processing have used tightly controlled calibration data sets as the basis for comparison. Here we compare performances for seven processing methods using two data sets originally collected for disease profiling studies. An emphasis is placed on understanding sensitivity for detecting differentially expressed genes in terms of two key statistical determinants: test statistic variability for non-differentially expressed genes, and test statistic size for truly differentially expressed genes. Results In the two data sets considered here, up to seven-fold variation across the processing methods was found in the number of genes detected at a given false discovery rate (FDR). The best performing methods called up to 90% of the same genes differentially expressed, had less variable test statistics under randomization, and had a greater number of large test statistics in the experimental data. Poor performance of one method was directly tied to a tendency to produce highly variable test statistic values under randomization. Based on an overall measure of performance, two of the seven methods (Dchip and a trimmed mean approach) are superior in the two data sets considered here. Two other methods (MAS5 and GCRMA-EB) are inferior, while results for the other three methods are mixed. Conclusions Choice of processing method has a major impact on differential expression analysis of microarray data. Previously reported performance analyses using tightly controlled calibration data sets are not highly consistent with results reported here using data from human tissue samples. Performance of array processing methods in disease profiling and other realistic biological studies should be given greater consideration when comparing Affymetrix processing methods.	Background Affymetrix microarrays are high throughput assays for measuring the expression levels of thousands of gene transcripts simultaneously. This type of microarray measures the expression of each transcript multiple times through a set of "probe pairs". Since the advent of the Affymetrix microarray, numerous methods have been proposed for producing numerical expression summaries for each transcript based on the probe pair data. Several systematic studies have appeared comparing a number of methods on a common basis (e.g. [ 1 - 5 ]). These studies rely heavily on calibration data sets derived from spike-in, dilution series, and mixture experiments for comparing methods. Our goal here was to carry out a comparative study of Affymetrix array processing methods using data sets from typical biological experiments seeking differentially expressed genes in human tissue samples. The following seven methods are considered here: Dchip [ 10 ], GCRMA-EB and GCRMA-MLE [ 11 ], MAS5 [ 12 ], PDNN [ 13 ], RMA [ 2 , 3 ], and TM [[ 6 , 7 ], and ]. While not every popular method is included in our study, several highly distinctive and original approaches are studied. For example, Dchip was one of the first approaches to attempt to learn probe weights directly from the probe data, and RMA pioneered the approach of disregarding the control mismatch probes. PDNN uses physical modeling to determine probe weights, while the two GCRMA methods use GC content of the probe sequences to reduce variance in the mismatch (control) probe levels. The MAS5 method is the current default method provided by Affymetrix. In addition to the six methods cited previously, we also include a method designated TM (trimmed mean). This is a simple method that has been used in a number of published investigations (e.g. [ 6 , 7 ]), but has not been considered in any previous systematic comparison of Affymetrix processing methods. To produce the probe-set summary score, the PM-MM differences are rank ordered, and the brightest 20% and dimmest 20% of values are deleted. The mean of the remaining values is used as the summary score. The scores for all probe-sets are then quantile normalized to a reference array using a piecewise linear spline with 100 knots. An important feature of this study is the use of False Discovery Rate (FDR) to quantify the sensitivity of a processing method in terms of its ability to distinguish differentially expressed genes from genes having invariant expression. This is a highly relevant property, as differential expression analysis is the most common application of microarray data. A key advantage of using FDR to compare processing methods is that FDR values can be calculated accurately using real disease profiling data where the identities of differentially expressed genes are uncertain. In contrast, most previous systematic comparisons of array processing methods have focused on calibration data sets in which concentrations of certain genes were experimentally manipulated. When it is highly likely that at least one gene is differentially expressed, false discovery rate may be defined as the expected ratio of the number of false positive calls to the total number of positive calls in a differential expression analysis between two groups of samples [ 8 ]. If the groups are biologically distinct, a sensitive processing method should result in many genes with low FDR. Thus to compare the performances of different array processing methods, we looked at two datasets in which a verified biological characteristic divided the samples into two classes, and compared the methods based on the number of genes having FDR smaller than various thresholds. For this to be a valid basis for comparison, the FDR values must be estimated with reasonable accuracy. Following other recent work (e.g. [ 9 ]), we used a permutation approach for this estimation, arguing that there is no reason that this approach favors or disfavors any particular array processing method. A small FDR is due either to a small numerator, a large denominator, or both. The denominator of the FDR depends on the actual data distribution, so variation in this value may be due to factors such as accuracy in modeling the physical and chemical nature of probe binding. Variation in the FDR numerator, however, depends only on the distribution of values produced for randomized data, a purely statistical quantity reflecting the tendency of the method to incorrectly produce test statistic outliers. Our results suggest that both factors are important in determining sensitivity. The best methods produce many large test statistic values in the actual data, and also produce consistently small test statistic values for randomized data. Poor performance of one method can be directly explained by the tendency of the method to produce outlier expression values, leading to greater numbers of incorrectly large test statistics. For overall comparison, we evaluated every pair of methods on the basis of whether the first method is expected to call at least one truly differentially expressed gene that is not also called by the second method. If this is not expected to occur, the second method is said to strongly outperform the first. Based on this comparison, two of the methods considered are clearly favored, two are inferior, and results for the other three methods are mixed. Results Sensitivity differences Our primary basis for comparison is sensitivity – the number of genes detected at a given FDR 0 level, where FDR 0 is a rescaled FDR (see methods). Figure 1 shows the key sensitivity results, using both the t-test statistic and the rank-sum statistic to assess differential expression. Setting aside at first differences between the seven processing methods, we note two findings. First, in the colon data, analysis using the rank-sum statistic is substantially more sensitive than analysis using the t-test statistic. For the ovary data, where the sample sizes would not naturally suggest a robust analysis, there is no harm to sensitivity in using the rank-sum statistic. Second, the ovary curves are substantially higher overall than the colon curves. This may be due to a greater number of true positives in the ovary data, or it may be that the small sample size for the MSI group makes it difficult to attain high evidence levels for differential expression in the colon data. In any case, both data sets have many genes with small FDR values, supporting the biological relevance of the tumor groupings for both colon and ovary samples. The more challenging colon set distinguishes the seven processing methods to a greater extent than the ovary set. Using FDR 0 = 0.1 as a reference point, there is roughly 7-fold variation across the seven methods in the number of detected genes in colon data using t-test statistics, while for rank-sum statistics the range is roughly 2-fold. In the ovary data, the range is around 1.25-fold for both statistics. Also notable is that variation in sensitivity due to the choice of test statistic (t-test or rank-sum) is smaller than variation in sensitivity due to the seven processing methods. No single method stands out as having the best or worst performance in every case. However some methods generally perform better than others. The Dchip and TM methods perform consistently well, while the GCRMA-EB and MAS5 methods consistently perform poorly. PDNN performs well on the ovary data, but poorly on the colon data, and results for the other methods are mixed. Level of agreement between methods Identities of probe sets falling below a given FDR 0 threshold vary across the methods. Figure 2 summarizes this variation. The ratio of the number of probe sets falling below various FDR 0 thresholds in k or more of the seven methods to the number of probe sets falling below the threshold for at least one method is plotted against the FDR 0 threshold, for k = 3, 4, 5, 6, 7. In the ovary data there is a very high level of agreement in this measure. For the rank-sum analysis, almost 90% of called genes are called by at least four methods, and more than 70% of called genes are called by all seven methods. For the t-test analysis, the agreement is slightly higher yet. For the colon data, the methods are much more inconsistent. For the rank-sum analysis, three of the methods agree on up to 90% of genes, but all seven methods only agree on around 30% of genes. The t-test analysis is even worse, with only around 10% of genes common to all seven methods. Turning to pairwise agreement, Table 1 shows the percentage of genes called by both members of a pair of methods out of the genes called by at least one of the two methods. In the ovary data, MAS5 shares the fewest calls with the other methods for both t-test and rank-sum analysis, while GCRMA-EB has relatively weak agreement for the t-test analysis. In the colon data, the GCRMA-EB method is highly inconsistent, with less than a quarter of calls in common with four of the six other methods for t-tests. A notable similarity is that the DChip and TM methods have at least 90% agreement in all analyses. Complementing comparison of the statistical tests, we also compared the expression levels produced by the seven processing methods. For each pair of methods, and for each pair of samples within one of the two data sets, we calculated Pearson correlation coefficients of expression levels over all genes. These values were summarized by taking the median over all pairs of samples within a data set, shown in Table 2 . Interestingly, methods calling similar genes as differentially expressed do not exhibit particularly strong correlation in expression levels. For example, TM and DChip perform very similarly in terms of which genes are identified as significant, but the pairwise correlation between expression levels for these two methods is less than the average. On the other hand, the TM and MAS5 methods are generally at the extreme high and low ends of the sensitivity scale respectively, but their expression levels are the most strongly correlated of any pair of methods. Calibration Variation in FDR across the seven methods is due to two factors – variation in the number of transcripts with large test statistics, and variation in the expected number of transcripts with large test statistics when there is no real differential expression. Here we investigate the second factor, which is driven by the tendency of each method to produce outlier expression values. The numerator of the FDR aims to correct for variation in the number of false positives, so that a method claiming large numbers of differentially expressed genes is not considered superior unless it also produces relatively small numbers of false positives. This can be viewed as a calibration, in which for each method, the test statistic must reach a certain threshold in order that the proportion of false positives is no greater than a specified value. Calibration results are summarized in Figure 3 . For each method, the threshold test statistic value required to obtain FDR 0 less than f was calculated, and plotted against f . For example, to achieve any FDR 0 value between 0.05 and 0.1 in the colon rank-sum data, GCRMA-MLE requires the lowest test statistics, RMA requires a rank-sum statistic 0.15 units larger than that of GCRMA-MLE, and MAS5 requires a rank-sum statistic 0.3 units larger than that of GCRMA-MLE. Figure 3 indicates that the methods differ substantially in terms of calibration. Notably, the ordering of the seven methods in Figures 1 and 3 are quite similar, suggesting that calibration plays a major role in determining sensitivity. Variation in thresholds among the seven processing methods is greater in the colon than the ovary data, particularly for the t-test analysis. Since calibration depends only on randomized data, it should be possible to trace variation in thresholds across the processing methods to statistical properties of the expression levels. For example, if one method produces expression levels with heavier tails, it is easier to get a large t-test statistic value by chance, particularly for the colon data with small sample sizes. This would necessitate a higher threshold. To quantify this, let denote the log 2 expression level of transcript i in sample j for method k , where k = 1, ..., 7 denotes the seven processing methods, and let where is the p th quantile of , and med is the median value. This is an affine-invariant measure of the size of the right tail of the expression values. Values of B k for the seven methods and two data sets are shown in Table 3 . For reference, a Gaussian distribution has a B value of 3.74 when the sample size is as in the ovary data, and 3.56 when the sample size is as in the colon data. The GCRMA-EB method is seen to have a much greater propensity for producing extreme expression values, explaining its low sensitivity, poor agreement with other methods, and conservative calibration. Variation in observed test statistics In addition to calibration differences, FDR variation is also influenced by the observed test statistic values. This is summarized in Figure 4 . For each method, and for a range of test statistic values t , the number of probe sets for which the observed test statistic value exceeds t was calculated and plotted against t . For example, in the colon rank-sum data, PDNN had the smallest test statistics, with MAS5 having around 500 more probe sets meeting a log test statistic threshold of 5 compared to PDNN. The Dchip and TM methods have over a thousand more probe sets meeting this threshold. Variation in test statistic values across the methods is greater in colon than in ovary data, and generally tracks with sensitivity. However note that in the colon rank-sum data, Dchip has substantially larger test statistics than GCRMA-MLE, even while GCRMA-MLE has better sensitivity (Figure 1 ), due to its less stringent calibration (Figure 3 ). Identification of genes with large fold changes An interesting possibility that can not always be excluded is that the intergroup differences are so vast that nearly every gene is affected to a small degree. If this were the case, the FDR values for the t-test and Wilcoxon statistics would converge to zero for every gene as the number of samples grows, making FDR values difficult to interpret. To further investigate this issue, we repeated the analysis using t-statistics truncated to zero when the fold change is less than 1.5 as test statistics for FDR analysis. The corresponding FDR values remain bounded away from zero for genes having true fold change smaller than 1.5, while genes with true fold change exceeding 1.5 have FDR values converging to 0. Thus the statistic identifies a meaningful subset of genes even when all genes are differentially expressed to some degree. Results for this analysis are shown in Figure 5 . In the ovary data, the GCRMA-EB method performs best, with GCRMA-MLE, MAS5, and TM slightly inferior. Several of the methods, specifically PDNN, DChip, and RMA exhibit flat curves indicating that only a limited number of genes meet the 50% change criterion. In the colon data, GCRMA-MLE and TM are nearly tied as the best performers. Overall, variation in sensitivity across the methods exists at a similar level to that found in the t-test and Wilcoxon analyses. Only the GCRMA-MLE and TM methods give consistently good performances in the two data sets for this analysis. Strong outperformance Thus far we have focused on sensitivity as a criterion for comparing methods. However even if one method is less sensitive than another, if the overlap in the called gene sets is not too great then the less sensitive method may still contribute to our understanding of which genes are differentially expressed. Suppose two methods denoted 1 and 2 give N 1 and N 2 genes respectively at a given FDR level. Then n k = (1 - p 0 FDR 0 )· N k estimates the expected number of truly differentially expressed genes called by method k . Now suppose that I is the number of genes called by both methods. Then n k - I is an estimated lower bound for the expected number of genes correctly called by method k but not by the other method. We will say that method 1 strongly outperforms method 2 if n 1 - I ≥ 0 but n 2 - I < 0. This means that in terms of differential expression, method 2 is not expected to contribute any true positives that were not called by method 1. Table 4 summarizes the results of this analysis using p 0 = 1 and FDR 0 = 0.05, showing the number of times that each method was strongly outperformed by other methods in our study. This analysis clearly favors the TM and Dchip methods, while the MAS5 and GCRMA-EB methods are nearly always found to be strongly outperformed by the other 5 methods. These results are not sensitive to choices of p 0 between 0.5 and 1 (more than half of values are constant within this range and non-constant values do not vary by more than 1). Discussion Impact of processing method choice The choice of processing method for Affymetrix array data evidently has a major impact on the ability to confidently report the results of differential expression analysis. The effect is greater, for example, than the choice of using a robust or a non-robust analysis, even in the colon data where robust analysis results in substantial improvements. Differences among processing methods are much greater in the more challenging colon data set compared to the ovary data, yet it should be noted that the sample sizes in the colon data are not atypical in real investigations. While results from two data sets can never conclusively determine the optimal method, it is notable that across both data sets, using both t-statistic and rank-sum analyses, there is a high degree of similarity in the rank ordering of the methods from the best to the worst performer. The trimmed mean (TM) and Dchip methods consistently perform as well or better than any of the other methods. A possible explanation for this is that the weights used by the Dchip may tend to downweight the least and greatest PM-MM differences, just as the TM method excludes these differences. Interpretation of FDR comparisons When comparing array processing methods using experimental data in which the identities of differentially expressed genes are unknown, great care must be taken to ensure that apparent differences in sensitivity are not due to other factors. One critical point is that the null distribution providing the expected number of false positives at a given test statistic threshold (the numerator of the FDR) must fairly reflect the statistical behavior of null genes. Permutation approaches have been extensively used to produce empirical p-values (e.g. [ 14 ]) and were used by Efron et al. [ 9 ] to estimate FDR values. Although permutation approaches are known to be slightly biased for estimating the FDR, the size of the bias (e.g. as shown in figure 5 of Efron et al. [ 9 ]) can not explain the magnitude of differences found here. In addition, for a comparative analysis, as carried out here, it is more crucial that the biases be relatively constant across the methods. However, since permutation approaches may not be highly accurate when the sample size is small, it is important to check performance on multiple data sets before conclusions about performance are drawn. While we have focused on FDR as the basis of comparison, the pursuit of small FDR values is not the only desirable operating characteristic of an array processing method, and other reports have also emphasized the accuracy of estimating the precise size of concentration differences. However to the extent that most actual studies seek to find differential expression between groups, the use of small FDR values seems more instrumental as the basis for judging methods. Variation due to choice of test statistic Although our primary aim was to investigate variation in sensitivity due to the seven processing methods, all analysis was carried out independently for two test statistics. The t-statistic is widely used in practice, but is well-known to be sensitive to outliers, particularly when the sample size is small. We found that certain processing methods, particularly EB-GCRMA, had a tendency to produce outlier expression values in the colon data set. Thus the combination of using the EB-GCRMA method with t-statistics in the colon data led to particularly poor performance. Variation due to log transform and array normalization In practice, the approach used for array normalization and for forming log-transformed expression values may be equally or more influential than the method used for producing probe set summaries [ 15 ]. In this study, we used implementations of the seven processing methods as prepared by their developers, and thus array normalization and and log-transforms were applied in a method-specific fashion. This provides a comparative analysis of the various methods as they are used in practice, which is most directly relevant since few investigators will override the default normalization and log-transform methods provided by the developers of each method. Nevertheless it remains of interest whether these routine processing steps are the determining factor of performance. In a future study it will be important to investigate this question further by modifying the implementations of the processing methods so that uniform log transforms and array normalizations are applied. Comparison of methods using data from disease profiling data sets A key point that we advocate in this work is that false discovery rates in actual disease profiling data constitute a valuable complement to benchmarking results obtained from spike-in, dilution series, and mixture experiments (e.g. [ 4 , 5 ]). The primary obstacle that must be overcome is that proper null sampling distributions are essential to ensure that the methods are compared on a common basis. Since numerous data sets covering a wide range of Affymetrix platforms are available, to the extent that multiple data sets are in agreement about relative performances it is unlikely that the randomization procedure used to calculate FDR values is systematically biased against a particular method. In spite of the statistical challenges in using disease profiling data for benchmarking, we argue that these data sets also offer some unique advantages. Calibration data sets are relatively few in number and are not available for all platforms. Newer platforms in particular are under-represented. Therefore overtraining to the available calibration data through manipulation of the many tuning parameters in the more complicated processing methods is an unavoidable concern. In addition, the calibration data sets likely do not represent the same degree of challenge as disease profiling data in that reproducibility of fold changes for affected and unaffected genes is quite high compared to data from, say, human tissues where a large number of uncontrolled sources of variability are present. Conclusions Performances of multiple array processing methods on disease profiling data sets vary widely across the seven methods studied here, but results are generally consistent between the two data sets studied. Results of our analysis generally do not parallel results obtained using calibration data sets [ 4 , 5 ], suggesting that such comparisons may not completely capture the most relevant aspects of performance. A major determinant of sensitivity is test statistic variability for randomized data. Such variability will affect false discovery rates as well as empirical p-values, which are an often-used alternative approach for identifying differentially expressed genes (e.g. [ 14 ]). Therefore it will be important in future work to seek a better understanding of statistical sampling properties of array processing methods. A particular focus should be the way that sampling variance in probe masking and probe weighting is controlled. Methods seeking to incorporate mechanistic information about the dynamics of probe binding, such as the two GCRMA methods and PDNN, should in principal outperform more generic approaches such as the TM method. Our results, particularly in the colon data, suggest that in medium-sized data sets this potential is not yet reached. In this comparative analysis we did not seek to draw definitive conclusions about the "best" or "worst" methods. Such conclusions may be made after investigating a greater number of data sets, including disease profiling data, data from controlled experiments, and calibration data. Moreover, it may be that the correct choice of method may depend on the scientific question being asked. The key message of this work is that the wide range of data sets collected in actual scientific investigations may be used for comparison of processing methods, and that in at least the two data sets considered here, similar results were obtained in the rank ordering of the methods. Methods Data sets We used two data sets – one consisting of 79 ovary tumors and the other consisting of 47 colon tumors. Both sets were generated at the University of Michigan using Affymetrix HG_U133A arrays, which consist of 22283 probe-sets, each of which is designed to assay a RNA transcript. Each probe-set consists of a set of (typically 11) probe-pairs, with each probe-pair comprising a "perfect match" (PM) probe which is a 25-base oligonucleotide complementary to the transcript, and a "mismatch" (MM) probe that is identical to the PM sequence except for alteration of the central base. The MM probe is intended as a control for nonspecific hybridization, so that the difference PM-MM measures only specific binding. However not all processing methods use the MM data in this way. For differential expression analysis, the 79 ovary samples were partitioned according to histological class into 38 endometrioid and 41 serous samples. The 47 colon samples were partitioned into 40 microsatellite stable (MSS) samples and 7 microsatellite instable samples (MSI). In both data sets, the partition is based on an independently measured biological characteristic, so there almost certainly are differentially expressed genes to be found. However in neither case are the two classes highly distinct, and numerous other sources of biological variation are undoubtedly present in the data. Normalization across arrays Array normalization refers to an adjustment of data distributions within each array in order to make the arrays more comparable. Each array processing method has been coupled with a normalization procedure by its developers (see references). We followed these method-specific normalization practices in our analysis. All methods other than MAS5 use some form of quantile normalization. Log transform and truncation All analysis was based on log-transformed data. Log-transformed values, including truncations where needed, were calculated in the manner recommended by the developers of each method (see references). Methodology of comparison We compared the seven methods based on their sensitivity in detecting differential expression at a fixed false discovery rate (FDR). For each method, two different two-sample test statistics were calculated for each gene – the standard two-sample t-statistic, and the Wilcoxon rank-sum statistic (equivalent to the Mann-Whitney statistic). The t-test statistic T is always analyzed as \| T \|, and the rank-sum statistic R is standardized as , where m 0 , m 1 are the numbers of samples in the two classes, and m = m 0 + m 1 is the total number of samples. Our FDR approach closely follows the "global estimate" of Efron et al. ([ 9 ] equation 5.9). For a given test statistic threshold t , the FDR was estimated as follows. Randomized data sets were constructed by randomly reassigning the class identifiers to the samples. The average number of transcripts with test statistic value exceeding t was calculated over 1000 randomized data sets. This number was divided by the number of actual transcripts with test statistic value exceeding t to produce a value that we denote FDR 0 . In practice the value of FDR 0 should be scaled by the proportion p 0 of non-differentially expressed genes, giving FDR = p 0 FDR 0 . Although various estimates of p 0 exist, we elected to ignore this factor since it is constant across the methods for a given data set, and any estimate of p 0 would add an additional source of uncertainty to our results. Thus it should be noted that the reported FDR 0 values, while comparable across methods, are somewhat larger than the usual estimates. Since p 0 would generally be greater than 1/2, the bias is likely less than a factor of 2. Authors' contributions KS, JMGT, RK, and DG participated in all phases of design and analysis. WC and JM performed the data analysis. KRC, TJG, SBG, ERF, and SH assisted in study design and supervised data collection. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC550659.xml
406392	Pax7 Is Necessary and Sufficient for the Myogenic Specification of CD45+:Sca1+ Stem Cells from Injured Muscle	CD45 + :Sca1 + adult stem cells isolated from uninjured muscle do not display any myogenic potential, whereas those isolated from regenerating muscle give rise to myoblasts expressing the paired-box transcription factor Pax7 and the bHLH factors Myf5 and MyoD. By contrast, CD45 + :Sca1 + isolated from injured Pax7 −/− muscle were incapable of forming myoblasts. Infection of CD45 + :Sca1 + cells from uninjured muscle with retrovirus expressing Pax7 efficiently activated the myogenic program. The resulting myoblasts expressed Myf5 and MyoD and differentiated into myotubes that expressed myogenin and myosin heavy chain. Infection of CD45 − :Sca1 − cells from Pax7 −/− muscle similarly gave rise to myoblasts. Notably, infection of Pax7 -deficient muscle with adenoviral Pax7 resulted in the de novo formation of regenerated myofibers. Taken together, these results indicate that Pax7 is necessary and sufficient to induce the myogenic specification of CD45 + stem cells resident in adult skeletal muscle. Moreover, these experiments suggest that viral transduction of Pax7 is a potential therapeutic approach for the treatment of neuromuscular degenerative diseases.	Introduction Skeletal muscle regeneration has long been considered to be mediated solely by monopotential skeletal muscle stem cells known as satellite cells ( Bischoff 1994 ; Charge and Rudnicki 2004 ). However, recent studies have identified novel populations of adult stem cells in skeletal muscle. For example, “side-population” (SP) cells isolated from muscle tissue participate in the regeneration of skeletal muscle and give rise to satellite cells ( Gussoni et al. 1999 ; Asakura et al. 2002 ). In vitro, muscle SP cells readily form hematopoietic colonies, but do not spontaneously differentiate into muscle cells unless cocultured with satellite-cell-derived myoblasts ( Asakura et al. 2002 ). Various cell surface markers have been employed to purify adult stem cell populations from skeletal muscle, including c-kit, Sca1, CD34, and CD45 (reviewed by Charge and Rudnicki 2004 ). Almost all muscle-derived hematopoietic progenitor and blood reconstitution activity is derived from CD45 + cells ( Asakura et al. 2002 ; McKinney-Freeman et al. 2002 ). Muscle-derived CD45 + cells purified from uninjured muscle are uniformly nonmyogenic in vitro and do not form muscle in vivo ( Asakura et al. 2002 ; McKinney-Freeman et al. 2002 ). However, coculture and in vivo injection experiments indicate that CD45 + SP, as well as CD45 − SP, cells possess myogenic potential ( Asakura et al. 2002 ; McKinney-Freeman et al. 2002 ). Recent experiments have established that CD45 + adult stem cells have a normal physiological role in tissue regeneration ( Polesskaya et al. 2003 ). CD45 + :Sca1 + cells display a 30-fold expansion in number following cardiotoxin-induced (ctx-induced) injury. Importantly, a large proportion of CD45 + :Sca1 + cells isolated from regenerating muscle acquire myogenic potential and appear to represent a significant source of myogenic progenitors during regenerative myogenesis ( Polesskaya et al. 2003 ). Moreover, the myogenic specification of these adult stem cells during regeneration occurs by a Wnt-signaling-dependent mechanism ( Polesskaya et al. 2003 ). The paired-box transcription factor Pax7 is specifically expressed in satellite cells and is required for the specification of the satellite cell lineage ( Seale et al. 2000 ). Following Wnt treatment of isolated CD45 + adult stem cells, Pax7 is rapidly induced as an early marker of satellite cell myogenic specification ( Polesskaya et al. 2003 ). Together, these data suggest the hypothesis that Pax7 represents the target of Wnt signaling that directs the myogenic specification of adult stem cells resident in muscle. To investigate this hypothesis, we examined the myogenic potential of adult stem cells from Pax7 −/− muscle, and employed viral vectors to transduce Pax7 into cells in vivo and in vitro. Our experiments demonstrate that Pax7 induces the myogenic program in specific populations of adult stem cells within muscle tissue and support the conclusion that Pax7 regulates myogenic determination during regenerative myogenesis. Results Pax7 Is Required for the Myogenic Commitment of CD45 + :Sca1 + Cells To determine whether Pax7 is required for myogenesis in muscle-derived CD45 + cells, we analyzed the myogenic differentiation capacity of CD45 + :Sca1 + cells from Pax7 −/− muscle undergoing ctx-induced regeneration. Flow cytometry analysis revealed a higher average proportion of CD45-expressing cells in Pax7 −/− muscle relative to wild-type ( Figure 1 A). Specifically, in muscle suspensions from Pax7 −/− and wild-type littermates, 39% ± 4% versus 31% ± 9% of cells were CD45 + :Sca1 − and 9% ± 2% versus 5% ± 2% of cells were CD45 + :Sca1 + , respectively ( n ≥ 6). Four days following ctx injury, a significantly higher proportion of CD45 + :Sca1 + cells (26% ± 3% compared to 19% ± 4% for Pax7 −/− and wild-type, respectively, p < 0.05) and a reduced proportion of CD45 − :Sca1 + cells were observed in Pax7 −/− muscle (19% ± 4% compared to 25% ± 6% for Pax7 −/− and wild-type, respectively, p = 0.07, n = 3) ( Figure 1 A). Figure 1 Pax7 Is Required for the Myogenic Specification of CD45 + :Sca1 + Cells (A) Flow cytometric analysis of cell suspensions derived from uninjured and regenerating wild-type and Pax7 −/− muscle (4 d after ctx injection) showed an increased proportion of CD45 + cells in Pax7 −/− samples. (B and C) Pax7 protein was expressed in approximately 6%–10% of CD45 + :Sca1 + cells purified from regenerating Pax7 +/− muscle. (D–K) MyoD (D and E) and Desmin (F and G) were induced in CD45 + :Sca1 + cells from regenerating Pax7 +/− but were not expressed in CD45 + :Sca1 + cells from regenerating Pax7 −/− muscle (H–K). By immunohistochemical analysis, Pax7 protein was upregulated in 6%–10% of CD45 + :Sca1 + cells from wild-type muscle 4 d after ctx injury ( Figure 1 B and 1 C). Importantly, Pax7 expression was not detected in CD45 + :Sca1 + cells purified from uninjured muscles ( Polesskaya et al. 2003 ). Furthermore, MyoD− ( Figure 1 D and 1 E) and Desmin− immunoreactive cells ( Figure 1 F and 1 G) were readily detected in cultured (18 h in growth medium) CD45 + :Sca1 + cells purified from regenerating Pax7 +/ − muscle (4 d post-ctx). By contrast, Pax7 −/− CD45 + :Sca1 + cells from regenerating muscle did not give rise to any MyoD-expressing ( Figure 1 H and 1 I) or Desmin-expressing ( Figure 1 J and 1 K) myogenic cells ( n = 3 independent isolations with three mice per experiment). Taken together, these results support a central role for Pax7 in the myogenic specification of CD45 + :Sca1 + cells in response to acute muscle damage. Pax7 Is Sufficient to Induce Myogenesis in CD45 + :Sca1 + Cells Adenoviral and retroviral expression systems were developed to ectopically introduce the Pax7 gene into putative adult stem cell populations. Pax7 was efficiently expressed from retrovirus (HAN-Pax7) in C3H10T1/2 fibroblasts and other cell cultures ( Figure 2 A). Stable expression of Pax7 did not induce MyoD (data not shown) or Myogenin protein expression ( Figure 2 B and 2 C) in C3H10T1/2 cells. MyoD, as expected, readily converted C3H10T1/2 cells into skeletal myocytes ( Figure 2 D and 2 E). These results show that Pax7 is not sufficient to induce myogenic determination in an established multipotent mesenchymal cell. Figure 2 Pax7 Induces Myogenic Commitment in CD45 + :Sca1 + Cells (A) Western blot analysis with anti-Pax7 antibody confirmed high levels of ectopic Pax7 in C3H10T1/2 cells infected with retrovirus-Pax7 (HAN-Pax7) but not with control virus expressing a puromycin-resistance marker (HAN-puro). (B and C) HAN-Pax7 did not induce expression of myogenin in C3H10T1/2 cells. (D and E) By contrast, MyoD virus (HAN-MyoD) efficiently converted C3H10T1/2 cells to myogenin-expressing myocytes (green) . (F–I) HAN-Pax7 (F and G) but not HAN-puro (H and I) activated expression of MyoD (red) in CD45 + :Sca1 + cells from uninjured muscle. (J–M) HAN-Pax7 (J) but not HAN-puro (K) also induced Myf5nLacZ expression in CD45 + :Sca1 + cells. Furthermore, HAN-Pax7-infected CD45 + :Sca1 + cultures differentiated into MyHC-expressing myocytes (green) under differentiation conditions (L), whereas HAN-puro-infected cells did not undergo myogenic differentiation (M). DAPI staining (blue) was used to visualize all nuclei. To determine whether Pax7 expression was sufficient to activate myogenesis in adult CD45 + progenitors, cells were fractionated from uninjured muscle and infected with Pax7-expressing retrovirus. Strikingly, CD45 + :Sca1 + cells expressed Myf5 (data not shown) and MyoD ( Figure 2 F– 2 I) protein after infection only with Pax7 (HAN-Pax7), and not with puromycin-alone control virus (HAN-puro), indicating that these cells undergo myogenesis in response to Pax7 Infection of CD45 + :Sca1 + cells from Myf5nLacZ reporter mice with HAN-Pax7 retrovirus specifically induced Myf5nLacZ expression and myogenesis in about 50% of infected cells ( Figure 2 J). The Myf5nLacZ allele faithfully recapitulates the expression pattern of the endogenous Myf5 gene and is rapidly induced following myogenic commitment ( Tajbakhsh et al. 1996 ; Tajbakhsh et al. 1997 ). Importantly, infection of CD45 + :Sca1 + cells with control retrovirus expressing the puromycin-resistance gene (HAN-puro) did not activate Myf5nLacZ expression ( Figure 2 K). Moreover, exposure of these cultures to differentiation conditions caused Pax7-expressing cells to differentiate into myotubes expressing myosin heavy chain (MyHC) ( Figure 2 L and 2 M). Ectopic expression of Pax7 in CD45 − :Sca1 + or CD45 + :Sca1 − cells did not result in the generation of myogenic cells. Taken together, these results demonstrate that Pax7 induces the myogenic program selectively in CD45 + :Sca1 + adult stem cells from skeletal muscle. Expression of Pax7 Converted CD45 + :Sca1 + Cells into Myogenic Progenitors CD45 + :Sca1 + cells expressing retroviral Pax7 were stably selected using puromycin, hereafter called CDSC-Pax7 cells ( n = 4 independent isolates analyzed). CDSC-Pax7 cells displayed a stellate, fibroblastic morphology that was distinct from the round, refractile appearance of primary satellite-cell-derived myoblasts. Proliferating CDSC-Pax7 cells expressed the myogenic determination bHLH factors, Myf5 ( Figure 3 A– 3 C), and MyoD ( Figure 3 D– 3 F). CDSC-Pax7 cells cycled approximately three times faster than satellite-cell-derived myoblasts isolated simultaneously (data not shown) and maintained their myogenic identity as primary cultures in excess of three months. CDSC-Pax7 cultures also differentiated efficiently into multinucleated myotubes expressing the terminal differentiation markers MyHC ( Figure 3 G– 3 I) and myogenin ( Figure 3 J– 3 L). These results demonstrate that the constitutive expression of Pax7 ( Figure 3 M– 3 O), which is normally downregulated during differentiation ( Seale et al. 2000 ), did not interfere with cell-cycle arrest and normal myotube formation. By contrast, overexpression of Pax7 in C2C12 myoblasts prevented their differentiation into MyHC-positive myocytes (data not shown). These experiments therefore demonstrate that myoblasts derived from Pax7-infected CD45 + :Sca1 + stem cells are amenable to ex vivo expansion and subsequent terminal muscle differentiation. Figure 3 CDSC-Pax7 Cells Become Myogenic Progenitors Myf5 (A–C) and MyoD (D–F) protein (green) are expressed in proliferating CDSC-Pax7 cells. Exposure of CD45 + :Sca1 + cultures to low mitogen medium induced the formation of multinucleated myotubes and expression of myogenic differentiation markers including MyHC (red) (G–I) and myogenin (red) (J–L). Sustained expression of Pax7 (red) (M–O) in these cultures did not interfere with their differentiation. DAPI staining (blue) was used to visualize all nuclei. CDSC-Pax7 Cells Express High Levels of Myf5 and Sca1 The expression pattern of myogenic factors in proliferating and differentiating CDSC-Pax7 cell lines was analyzed by Western blot ( n = 2). These experiments indicated that Myf5 was expressed at high levels in proliferating CDSC-Pax7 cells ( Figure 4 A; day 0). Moreover, CDSC-Pax7 cells continued to express Myf5 protein during their differentiation. CDSC-Pax7 cells also expressed MyoD but at low levels relative to primary myoblasts. MyoD was transiently upregulated in CDSC-Pax7 cells as they entered their differentiation program ( Figure 4 A; days 1 and 2). Figure 4 CDSC-Pax7 Cells Express High Levels of Myf5 and Sca1 (A) Western blot analysis of CDSC-Pax7 cells in proliferation conditions (day 0) and during differentiation (days 1–4) revealed high levels of Myf5 expression and low levels of MyoD expression. By contrast, satellite-cell-derived myoblasts (Wt-Mb) displayed the opposite profile of Myf5 and MyoD expression. Myogenin (Myg) was upregulated during the differentiation of CDSC-Pax7 and satellite-cell-derived myoblasts (Wt-diff). Note the sustained expression of Pax7 during the differentiation of CDSC-Pax7 cells. C3H10T1/2 (10T) lysate was used as a negative control. (B) RT-PCR analysis indicated that CDSC-Pax7 cells (two different lines) upregulated the endogenous Pax7 mRNA. Satellite-cell-derived myoblasts (Wt-Mb) and Jurkat cells were used as positive and negative controls, respectively. (C) Flow cytometry indicated that CDSC-Pax7 cells lost expression of CD45 but retained high levels of Sca1. About 24% of satellite-cell-derived myoblasts (wt-myoblasts) expressed low levels of Sca1. (Black graph depicts staining with IgG-PE control antibody; red graph shows target staining using Sca1-PE or CD45-PE.) The primary myogenic regulatory factor (MRF) expression profile in CDSC-Pax7 cells contrasted with the pattern observed in satellite-cell-derived primary myoblasts ( Figure 4 ; Wt-Mb). Primary myoblasts expressed higher levels of MyoD and lower levels of Myf5 and downregulated Myf5 immediately upon differentiation (Wt-diff). Myogenin (Myg) was upregulated during the differentiation of CDSC-Pax7 cells, albeit at lower levels compared with differentiating satellite-cell-derived myoblasts (Wt-diff). Interestingly, CDSC-Pax7 cells also expressed endogenous Pax7 mRNA as demonstrated by reverse transcriptase PCR (RT-PCR) using primers that amplify a sequence within the Pax7 3′ UTR that is not present in the viral-Pax7 vector ( Figure 4 B). This result suggests that autoregulatory mechanisms may control Pax7 gene expression. Taken together, these analyses demonstrate that CDSC-Pax7 cells and primary satellite-cell-derived myoblasts express different levels of MyoD and Myf5 but are similar in their ability to undergo terminal differentiation. CDSC-Pax7 cells were originally derived from cells expressing cell surface CD45 and Sca1 proteins. Flow cytometry was employed to determine whether expression of these markers was maintained in vitro. Notably, CDSC-Pax7 cells continued to express high levels of Sca1 (approximately 90% of cells showed intense staining), but CD45 expression was extinguished ( Figure 4 C). Interestingly, approximately 24% of primary satellite-cell-derived myoblasts displayed low levels of Sca1 staining. Sca1 levels were not increased in satellite-cell-derived myoblasts overexpressing Pax7, demonstrating that CDSC-Pax7 cells did not arise from a small number of committed myoblasts fractionated with CD45 + :Sca1 + cells (data not shown). CDSC-Pax7 Cells Differentiate In Vivo To establish whether CDSC-Pax7 cells were capable of integrating and differentiating as myofibers in vivo, intramuscular transplantation studies were performed in dystrophic ( dystrophin -deficient) muscle. Specifically, 1 × 10 5 CDSC-Pax7 cells were injected into the tibialis anterior (TA) muscle of 4- to 6-week-old mdx:nude mice. Mdx mice carry a point mutation in the dystophin gene and are a mouse model of Duchenne muscular dystrophy ( Bulfield et al. 1984 ; Sicinski et al. 1989 ; Blaveri et al. 1999 ). As expected, dystrophin was localized at the myofiber sarcolemma in wild-type muscle ( Figure 5 A) and was absent in mdx:nude skeletal muscle ( Figure 5 B). Two months after transplantation, TA muscles were processed for immunohistochemical detection of dystrophin and Pax7. These experiments revealed that CDSC-Pax7 cells differentiated in vivo, readily forming dystrophin-expressing myofibers in the dystrophin -deficient recipient muscle ( Figure 5 C and 5 D). Endogenous Pax7 protein expression was not observed in central nuclei within differentiated wild-type myofibers (data not shown). Therefore, the expression of Pax7 protein (from retrovirus) in central nuclei within dystrophin + fibers established the contribution of CDSC-Pax7 donor cells to recipient muscles ( Figure 5 E and 5 F). These results thus document the capacity for CDSC-Pax7 cells to differentiate in vivo and contribute to the repair of dystrophic muscle. Figure 5 CDSC-Pax7 Cells Efficiently Contribute to the Repair of Dystrophic Muscle (A) Wild-type muscle expressed dystrophin at the plasmalemma of all myofibers. (B) Dystrophin protein was not detected in muscle sections from dystrophin -deficient mdx:nude mice (mdx:nu). (C–F) CDSC-Pax7 cells differentiated in vivo after transplantation, readily forming large numbers of dystrophin-expressing myofibers (green) in mdx:nude muscle (C and D). Serial cross sections showing the viral expression of Pax7 protein in central nuclei of regenerated fibers (red staining in [E]) confirmed the donor origin of dystrophin-positive myofibers (red staining in [F]). Pax7 Does Not Induce Myogenesis in CD45 + :Sca1 + Cells from Pax7 −/− Muscle The myogenic differentiation of wild-type CD45 + :Sca1 + muscle cells suggested that ectopic Pax7 would induce myogenesis in this cell population from Pax7 −/− muscle. Infection of Pax7 −/− CD45 + :Sca1 + cells with Pax7 retrovirus resulted in high levels of retroviral Pax7 transcript but no expression of Myf5 mRNA by Northern blot hybridization ( Figure 6 A) or RT-PCR (data not shown). The absence of Myf5 ( Figure 6 B– 6 D) or MyoD (data not shown) expression, determined by immunochemical staining of Pax7-transduced cells, ruled out the possibility that a minor subpopulation of CD45 + :Sca1 + cells underwent myogenesis. These experiments illustrate that Pax7 −/− CD45 + :Sca1 + cells do not enter the myogenic lineage in response to Pax7, suggesting that intrinsic differences exist between wild-type and Pax7 -deficient populations of CD45 + :Sca1 + cells. Figure 6 Pax7 Does Not Induce Myogenesis in CD45 + :Sca1 + Cells from Pax7 −/− Muscle (A) Northern analysis shows that MyoD −/− satellite-cell-derived myoblasts ( MD −/− M) and differentiating cells ( MD −/− D) express endogenous Pax7 (upper arrow, Pax7 blot) and Myf5 transcripts. Pax7 −/− CD45 + :Sca1 + cells (CDSC) transduced with HAN-Pax7 (+Pax7) or HAN-puro (+puro) did not initiate expression of Myf5 mRNA. The retroviral transcript producing Pax7 (lower arrow) is smaller than the endogenous Pax7 mRNA (e.g., lower arrow). (B–D) Ectopic expression of Pax7 (red) (B) in Pax7 −/− CDSC cells did not induce Myf5 protein expression (C). DAPI staining (blue) was used to visualize nuclei (D). Pax7 Promotes Myogenic Commitment in Pax7 -Deficient CD45 −- :Sca1 − Cells In cell suspensions from uninjured muscle, satellite cells and their daughter myogenic precursors are uniformly CD45 − and Sca1 − ( Polesskaya et al. 2003 ). In Pax7 −/− mice, the extremely rare myogenic cells in muscle tissue do not express CD45 or Sca1, and do not survive or expand in a variety of culture conditions (S.B.P. Chargé, P. Seale, and M.A. Rudnicki, unpublished data). Interestingly, ectopic expression of Pax7 in CD45 − :Sca1 − cells isolated from Pax7 −/− muscle resulted in the expression of Myf5 protein in more than 50% of infected cells ( n = 3) ( Figure 7 A– 7 C). Analysis of HAN-puro-infected control cultures did not reveal any myogenic cells ( Figure 7 D– 7 F). Importantly, all Myf5-expressing myoblasts ( Figure 7 G– 7 I) and MyHC-expressing differentiated myotubes ( Figure 7 J– 7 L) in Pax7-infected CD45 − :Sca1 − cultures expressed viral Pax7. Figure 7 Pax7 Promotes Myogenesis in CD45 − :Sca1 − Cells from Pax7 −/− Muscle (A–C) Ectopic expression of Pax7 (HAN-Pax7) induced Myf5 expression (green) and myogenic commitment in CD45 − :Sca1 − cells from Pax7 −/− muscle. (D–F) By contrast, Myf5-expressing cells were completely absent from HAN-puro-infected cultures after selection. (G–L) CD45 − :Sca1 − cells from Pax7 −/− muscle expressed Myf5 (red) (H) and MyHC (red) (K) only in cells that also coexpressed high levels of Pax7 protein (G and J). Arrowheads indicate cells coexpressing Pax7 and Myf5/MyHC. Arrow in (G) and (I) depicts a Pax7 + , Myf5 − cell. In these experiments we cannot formally exclude the possibility that Pax7 promoted the survival and proliferation of committed myoblasts. However, given the extremely low number of myogenic cells recovered in culture (less than 0.7%), the low efficiency of primary myoblast infection (approximately 5%–10%), and the absence of any Myf5- or MyoD-expressing cells in control HAN-puro cultures, our results strongly suggest that Pax7 induces myogenic specification in a nonmyoblast, CD45- and Sca1-negative cell. Adenoviral Expression of Pax7 Enhances Regeneration in Pax7 -Deficient Muscle To investigate whether Pax7 was sufficient to stimulate myogenesis in vivo, adenovirus was used to ectopically express Pax7 in damaged Pax7 −/− muscle. Adenoviral particles (1 × 10 8 ) expressing either Pax7 (Ad-Pax7) or the bacterial β-galactosidase gene (LacZ) (Ad-LacZ) were injected directly into injured TA muscles of 4- to 6-week-old Pax7 −/− animals 2 d after administration of ctx ( n = 3). Immunohistochemistry for Pax7 in adenovirus-infected muscles demonstrated widespread Pax7 expression primarily in mononuclear cells within the damaged tissue (data not shown). To assess the effect of Pax7 expression in damaged tissue, TA muscles were analyzed and scored for regeneration 12 d after infection by enumerating the number of regenerated fibers with centrally located nuclei. The newly regenerated status of centrally nucleated fibers was confirmed by Desmin and embryonic MyHC immunoreactivity. Ad-Pax7 induced a markedly enhanced regenerative response relative to Ad-LacZ in Pax7 −/− muscle as evidenced by the increased number of Desmin-positive ( Figure 8 A and 8 B) and centrally nucleated fibers ( Figure 8 C and 8 D). Figure 8 Adenovirus-Pax7 Significantly Improves Regeneration In Vivo (A and B) Infection of ctx-damaged Pax7 −/− muscles with Ad-Pax7 resulted in markedly improved muscle integrity and a significantly increased number of Desmin immunoreactive (green) regenerated fibers (B) relative to muscles treated with Ad-LacZ (A). (C and D) Hematoxylin and Eosin staining similarly showed an increased number of centrally nucleated fibers in Ad-Pax7-treated Pax7 −/− muscles. (E) In three separate experimental trials, the number of regenerated fibers was markedly increased in Ad-Pax7-treated muscles relative to Ad-LacZ; however, the response was biologically variable between groups. On average, Ad-Pax7 infection resulted in a 4.1 ± 0.72–fold increase in regenerated Pax7 −/− myofibers (F). Wild-type TA muscles typically contained in excess of 700 regenerated fibers 14 d after injury (data not shown). In three independent experiments, ctx-damaged TA muscle from Pax7 −/− mice typically contained an average of 46 surviving or regenerated fibers following regeneration ( Figure 8 E). By contrast, infection of regenerating Pax7 −/− TA with Ad-Pax7 resulted in the generation of an average of 192 myofibers ( Figure 8 E). Therefore, Pax7-infected tissue contained a 4.1 ± 0.72–fold increase in the number of regenerated fibers ( Figure 8 F). Together, these results demonstrate the ability of virally transduced Pax7 to direct the de novo generation of myogenic progenitors capable of forming new myofibers and participating in regenerative myogenesis. Discussion In this article, we demonstrate that expression of Pax7 induces the myogenic specification of CD45 + muscle-derived adult stem cells. First, CD45 + :Sca1 + cells isolated from regenerating Pax7 −/− muscle were incapable of undergoing myogenic specification (see Figure 1 ). Second, expression of Pax7 with viral vectors in CD45 + :Sca1 + cells purified from uninjured muscle promoted the formation of highly proliferative myoblasts that readily differentiated as multinucleated myotubes (see Figures 2 and 3 ). CD45 + :Sca1 + cells engineered to express Pax7 (CDSC-Pax7) also differentiated in vivo, readily contributing to the regeneration of dystrophic muscle (see Figure 5 ). Lastly, Ad-Pax7 gene delivery into chemically damaged Pax7 −/− muscle resulted in the efficient de novo generation of myofibers in the absence of endogenous satellite cells. Taken together, these data unequivocally establish that Pax7 plays a key regulatory role for directing myogenic specification in some populations of adult stem cells during regenerative myogenesis. Moreover, these results emphasize the possibility of designing strategies to upregulate or ectopically express Pax7 in stem cells for the treatment of muscle degenerative diseases. The presence of adult stem cell populations distinct from satellite cells resident in skeletal muscle tissue has been well documented ( Gussoni et al. 1999 ; Jackson et al. 1999 ; Torrente et al. 2001 ; Asakura et al. 2002 ; McKinney-Freeman et al. 2002 ; Qu-Petersen et al. 2002 ; Cao et al. 2003 ). An understanding of the developmental origin of these various cell populations and their physiological relevance in the maintenance of tissue integrity is beginning to emerge. Several lines of evidence argue that skeletal muscle regeneration is normally mediated entirely by stem cells resident in muscle tissue. First, destruction of stem cells resident in muscle with high-dose local irradiation of limbs results in a long-term deficit in muscle growth and regeneration ( Wakeford et al. 1991 ; Pagel and Partridge 1999 ; Heslop et al. 2000 ). Second, transplanted muscles do not incorporate host nuclei after injury and regeneration ( Schultz et al. 1986 ). Together, those experiments argue that CD45 + stem cells from marrow do not normally transit in significant numbers through the circulation to sites of muscle damage. Our experiments, however, suggest that a population of specialized CD45 + cells resides in muscle and can efficiently form myogenic progenitors in response to Wnt signaling ( Polesskaya et al. 2003 ). In the current work we demonstrate that induction of Pax7 is required for the myogenic specification of CD45 + stem cells and that retroviral transduction can dominantly induce the myogenic specification of these cells. These observations therefore provide compelling evidence that some adult stem cells participate in regenerative myogenesis by forming myogenic progenitors following Pax7 induction in response to Wnt signaling. These data additionally suggest the hypothesis that Pax7 is a transcriptional target of the β-Catenin complex in Wnt-stimulated adult stem cells. Interesting parallels exist between the inductive mechanisms and transcriptional networks in embryonic and regenerative myogenesis ( Parker et al. 2003 ). For example, the Pax7-dependent myogenic specification of CD45 + adult stem cells appears analogous to the Pax3-dependent induction of muscle precursors during somitogenesis. In the early embryo, Pax3 is expressed in the presomitic mesoderm and immature epithelial somites prior to the onset of muscle-specific gene expression ( Goulding et al. 1994 ; Williams and Ordahl 1994 ). Moreover, Pax3 functions upstream of MyoD in the formation of trunk and body-wall muscle ( Tajbakhsh et al. 1997 ). Consistent with a direct role for Pax3 in myogenic induction, ectopic Pax3 activates MyoD expression in embryonic tissues ( Maroto et al. 1997 ; Bendall et al. 1999 ; Heanue et al. 1999 ). However, Pax3 also regulates cell survival in the presomitic mesoderm in areas that do not express Pax7, suggesting an indirect mechanism by which Pax3 may act genetically upstream of MyoD ( Borycki et al. 1999 ). Our experiments do not rule out the possibility that Pax7 promotes the survival of CD45 + progenitors that are already competent to give rise to myogenic cells. Characterization of the downstream targets of Pax7 in CD45 + cells will be required to directly address this issue. In explanted embryonic tissues, signals from the floor plate and neural tube are required for induction of the MRFs ( Munsterberg and Lassar 1995 ; Pourquie et al. 1995 , 1996; Cossu et al. 1996 ). In particular, Wnt7a activates expression of MyoD in explanted paraxial mesoderm from 10.5-d-old mouse embryos ( Tajbakhsh et al. 1998 ). The requirement for Pax3 expression in somitic precursors prior to the onset of MyoD expression suggests that Wnt signals may activate Pax3 and indirectly promote MRF expression ( Borycki et al. 1999 ). An analogous requirement for Pax7 in the myogenic commitment of adult CD45 + progenitors suggests a conserved hierarchy whereby Wnt signaling activates Pax3 or Pax7 expression upstream of the MRFs in somitic and adult muscle stem cells, respectively. This notion is supported by the observed loss of Pax3 expression in P19 mesodermal precursors engineered to express a dominant negative form of the Wnt effector protein, β-Catenin ( Petropoulos and Skerjanc 2002 ). A confounding result from our study was the inability of Pax7 to induce myogenesis in CD45 + :Sca1 + cells recovered from Pax7 −/− muscle (see Figure 6 ). Several possible explanations may account for this observation. First, CD45 + :Sca1 + muscle cells represent a heterogeneous cell population, as evidenced by their nonuniform response to stimuli such as myoblast coculture, Wnt proteins, and ectopic expression of Pax7 (results herein and Polesskaya et al. 2003 ). Analysis of muscle suspensions from young Pax7 −/− mice revealed a significantly increased number of hematopoietic progenitors and adipogenic cells ( Seale et al. 2000 ). We also observed altered proportions of CD45- and Sca1-expressing cells in uninjured and regenerating muscle (see Figure 1 A). The putative stem cell subfraction coexpressing CD45 and Sca1 may have been exhausted prematurely during postnatal Pax7 −/− muscle development. It is also conceivable that a reduced proportion of stem cells in the Pax7 −/− CD45 + :Sca1 + muscle fractions was not detected in our assay due to a low efficiency of retroviral transduction (approximately 10% of surviving CD45 + :Sca1 + cells with GFP virus). The identification of additional markers expressed by adult muscle-derived stem cells is required to more thoroughly explore these issues. Alternatively, adult stem cells may require additional signals to undergo myogenesis in response to Pax7. The profound growth deficit in Pax7 −/− muscles is likely to invoke nonspecific and indirect changes to the muscle microenvironment ( Seale et al. 2000 ). Specific cues required to “prime” or activate adult stem cells may thus be absent or ineffective in Pax7 −/− muscle. Finally, our experiments also revealed that the endogenous Pax7 gene is upregulated during the myogenic specification of CD45 + :Sca1 + cells ( Figure 4 B). Therefore, endogenous gene activity, possibly through the regulated expression of different isoforms ( Kay et al. 1995 ; Ziman et al. 1997 ), may be essential to the stability of myogenic commitment. Future experiments addressing the functional differences between CD45 + :Sca1 + cells in wild-type and Pax7 -deficient muscle will provide a unique opportunity to gain a more complete understanding of the role of these cells during postnatal muscle development. Although CD45 + cells from Pax7 −/− muscle were apparently unable to undergo myogenesis, ectopic Pax7 induced expression of Myf5 and myogenic specification in Pax7 -deficient CD45 − :Sca1 − cells (see Figure 7 ). Moreover, Ad-Pax7 significantly increased the in vivo regenerative capacity of Pax7 −/− muscle (see Figure 8 ). Skeletal muscle in adult Pax7 −/− mice displays a profound regeneration deficit with only occasional regenerated fibers observed at the site of injury 30 d after ctx injection (S.B.P. Chargé, P. Seale, and M.A. Rudnicki, unpublished data). Taken together, these results imply the presence of Pax7 −/− muscle progenitors that require the activity of Pax7 to generate sufficient numbers of myoblasts for effective regeneration. Further studies will be required to molecularly characterize the responsive cells and their developmental relationship to other muscle stem cell populations. The dominant expression of Myf5 in Pax7-infected CD45 + :Sca1 + cells (CDSC-Pax7) (see Figure 4 A) suggests a paradigm wherein Pax7 preferentially activates Myf5 compared to MyoD. Interestingly, Pax3 has been implicated in myogenesis specifically upstream of MyoD ( Tajbakhsh et al. 1997 ). Taken together, these observations suggest the hypothesis that Pax3 and Pax7 specify distinct myogenic lineages through the preferential activation of MyoD and Myf5, respectively. Several experimental observations have noted a role for Myf5 in promoting myoblast proliferation. For example homozygous Myf5nLacZ , (e.g., Myf5 -deficient) embryos display significantly reduced numbers of LacZ-expressing myogenic progenitors ( Tajbakhsh et al. 1996 ). In avian embryos, Myf5 is preferentially expressed in proliferating myoblasts, whereas MyoD appears to be upregulated in differentiating cells ( Delfini et al. 2000 ). Furthermore, Myf5 −/− satellite-cell-derived myoblasts display a profound proliferation deficit ( Montarras et al. 2000 ). The increased growth rate of CDSC-Pax7 cells is reminiscent of MyoD −/− myoblasts that also express elevated levels of Myf5 ( Sabourin et al. 1999 ). These observations raise the possibility that Pax7 activates expression of Myf5 to promote adult myoblast expansion whereas Pax3 preferentially induces MyoD and differentiation. The requirement for Pax7 in the specification of muscle satellite cells ( Seale et al. 2000 ) and its induction during the myogenic recruitment of CD45 + adult stem cells provide further evidence for a developmental relationship between CD45 + adult muscle stem cells and satellite cells. Together, our experiments suggest the hypothesis that CD45 + :Sca1 + cells give rise to satellite cells by a Pax7-dependent mechanism in response to Wnt signals. In conclusion, our work establishes that Pax7 is necessary and sufficient for the myogenic specification of specific populations of adult stem cells resident in muscle tissue. The proliferative myogenic character of CDSC-Pax7 cells and their efficient engraftment into dystrophic muscle further argue that methods to deliver Pax7 or upregulate its expression in stem cells may be useful in treating degenerative muscle disease. Materials and Methods Mice Mice carrying a targeted null mutation in Pax7 (hereafter referred to as Pax7 −/− ) were generously provided by Drs. A. Mansouri and P. Gruss ( Mansouri et al. 1996 ) and outbred into the SV129 background to increase survival. Myf5nLacZ mice were provided by Dr. S. Tajbakhsh ( Tajbakhsh et al. 1996 ). Mdx mice were obtained from Jackson Laboratory (Bar Harbor, Maine, United States). Mdx:nu mice were provided by Dr. T.A. Partridge (see Blaveri et al. 1999 ). Cell sorting Mononuclear cells were recovered from uninjured hindlimb muscles or from ctx-damaged TA muscles of Pax7 +/+ , Pax7 +/− , and Pax7 −/− mice as described previously ( Megeney et al. 1996 ). Cells were washed twice with ice-cold DMEM supplemented with 5% FBS, passed through 30-μm filters (Miltenyi Biotec, Bergisch Gladbach, Germany) and suspended at a concentration of 2–3 × 10 6 cells/ml. Staining was performed for 30 min on ice using the antibodies CD45-APC (30-F11), CD45.2-FITC (104), Sca1-PE, or FITC (D7), all from BD Biosciences Pharmingen (San Diego, California, United States) and CD45-TC (30-F11) from Caltag Laboratories (Burlingame, California, United States). Primary antibodies were diluted in cell suspensions at 1:200. After two washes with cold PBS supplemented with 2% FBS, cells were separated on a MoFlo cytometer (DakoCytomation, Glostrup, Denmark). Sort gates were strictly defined based on isotype control stained cells and single antibody staining. Dead cells and debris were excluded by gating on forward and side scatter profiles. Sorting was performed using single cell mode to achieve the highest possible purity. The purity of sorted populations was routinely greater than 98%. Retroviral and adenoviral gene expression Retrovirus was produced according to the 3-plasmid HIT system with plasmids pHIT60, pHIT456, and pHAN-puro as described elsewhere ( Soneoka et al. 1995 ). pHIT60 encodes the MLV retroviral gag-pol, pHIT456 expresses an amphotrophic envelope protein, and pHAN-puro is an expression vector with a hybrid CMV-5′ LTR promoter driving production of the retroviral transcript. Pax7 expression vectors were generated using the mouse Pax7d isoform containing a single Ala→Thr substitution at the seventh amino acid (the Thr residue is conserved in human, chicken, and zebrafish Pax7 proteins). Pax7d or mouse MyoD are translated from the full retroviral transcript, whereas the puromycin-resistance marker is expressed following integration from a shorter transcript produced by the SV40 early promoter located 3′ to the multiple cloning site. Transient cotransfection of all three plasmids into 293FT cells (Invitrogen, Carlsbad, California, United States) by the calcium phosphate method ( Graham and van der Eb 1973 ) routinely produced viral titres between 10 6 and 10 7 cfu per ml. pHAN-puro was used to produce puromycin-resistant virus for controls. Purified CD45 + :Sca1 + or CD45 − :Sca1 − cells were spun down, counted, and 20,00–50,000 cells were then cultured overnight on collagen-coated 4-well chamber slides in HAM's F10 medium (Invitrogen) supplemented with 20% FBS, antibiotics, and 10 ng/ml Stem Cell Factor (R & D Systems, Minneapolis, Minnesota, United States). The following day, cells were incubated for 6 h with retrovirus at a 1:1 ratio (complete medium: retrovirus supernatant) with 8 μg/ml polybrene (hexadimethrine bromide; Sigma, St. Louis, Missouri, United States). After infection, cells were rinsed twice with PBS, and all cells were replated in myoblast growth medium. After 48 h, infected pools were selected in 1 μg/ml puromycin (Sigma) to establish stable CDSC-Pax7 lines. C3H10T1/2 cells were incubated overnight with MyoD, Pax7, or puro virus and 8 μg/ml polybrene. Adenovirus (type V) was prepared using the Ad-Max adenovirus creation kit (Microbix Biosystems, Toronto, Ontario, Canada). Ad-Pax7d (cDNA as described above) and Ad-LacZ were expressed from the murine CMV promoter. Adenovirus was purified in CsCl gradients by centrifugation, dialyzed against sterile PBS, and frozen down in 15% glycerol at −80 °C. Titres of purified adenovirus were determined by plaque assays on 293 cells and were always above 10 10 pfu/ml. Western blot analyses Cell cultures were lysed in RIPA extraction buffer (50mM Tris-HCl [pH 7.4], 1% Nonidet P-40, 0.5% NaDeoxycholate, 0.1% Sodium-dodecyl-sulphate, 5 mM EDTA, 150 mM NaCl, 50 mM NaF) supplemented with protease inhibitors (Complete; Roche, Basel, Switzerland). The extracts were normalized for protein content using Bio-Rad dye (Hercules, California, United States). Forty micrograms of lysate was separated by sodium-dodecyl-sulfate-polyacrylamide gel electrophoresis and transferred to PVDF filters (ImmobilonP; Millipore, Billerica, Massachusetts, United States). Filters were probed with antibodies to Pax7 (Developmental Studies Hybridoma Bank [DSHB], Iowa City, Iowa, United States); Myf5, 1:1000 (C-20, Santa Cruz Biotechnology, Santa Cruz, California, United States); MyoD, 1:1000 (C-20, Santa Cruz Biotechnology); myogenin (F5D, DSHB); and α-tubulin, 1:2000 (T 9026, Sigma). Secondary detection was performed with horseradish peroxidase-conjugated antibodies (Bio-Rad). Protein expression was visualized using the ECL Plus kit (Amersham Biosciences, Little Chalfont, United Kingdom). Ctx-induced regeneration and in vivo adenovirus infections Four- to six-week-old Pax7 −/− and wild-type littermates were anesthetized with Halothane gas. Twenty-five microliters of 10 μM ctx (Latoxan, Valence, France) was injected into the midbelly of the TA muscle, using a 29½ G insulin syringe. Mice were sacrificed at 4 d or 2 wk after ctx injection. For adenovirus infections, 25 μl of sterile PBS containing 10 8 particles of purified Ad-Pax7 or Ad-LacZ was injected 2 d after ctx injection into damaged TA muscles with a 29½ G insulin syringe. Cell transplantation Primary CDSC-Pax7 cells cultured in myoblast conditions were trypsinized, washed twice with PBS, and suspended at 5 × 10 5 cells/25-μl in sterile PBS for cell transplantation. Cells were injected directly into the TA midbelly of 4- to 6-wk-old mdx:nude mice. Mice were sacrificed 2 mo after cell injections to analyze the myogenic contribution of transplanted cells. Cell cultures Primary satellite-cell-derived myoblasts were established from purified CD45 − :Sca1 − fractions of hindlimb muscle of 4- to 6-wk-old Pax7 +/+ or Pax7 +/− mice. Myoblasts and CDSC-Pax7 cells were maintained in HAM's F-10 medium (Invitrogen) supplemented with 20% FBS and 2.5 ng/ml bFGF (Invitrogen) on collagen-coated dishes. CDSC-Pax7 cells and primary satellite-cell-derived myoblasts were differentiated for 1–3 d in DMEM supplemented with 5% horse serum. C3H10T1/2 and HEK 293 cells were obtained from the ATCC (Manassas, Virginia, United States) and maintained in DMEM supplemented with 10% FBS. Histology and immunocytochemistry For analysis of regeneration and enumeration of regenerated myofibers, TA muscles were isolated, embedded in OCT (Tissue-Tek; Sakura Finetek, Torrance, California, United States)/20% sucrose and immediately frozen in liquid nitrogen. Ten-micrometer cryosections (cross sections) from the TA midbelly at the site of ctx injection were stained with Hematoxylin and Eosin. Central myonuclei in regenerating muscles were counted on at least two independent cross sections of the entire TA muscle per mouse analyzed. Fibers were further identified by immunostaining with antibodies specific to Desmin, 1:200 (D33, DakoCytomaton); dystrophin, 1:500 (Sigma); Pax7 (DSHB); or embryonic fast MyHC (F1.652, DSHB) followed by secondary detection with anti-mouse FITC conjugated antibody, 1:200 (Chemicon, Temecula, California, United States). Sections were analyzed on a Zeiss (Oberkochen, Germany) Axioplan 2 microscope. Cultured cells were fixed with 4% paraformaldehyde, nonspecific antigens were blocked in 5% horse serum/PBS, and cells were reacted with primary antibodies as follows: Desmin, 1:200 (DakoCytomaton); MyoD, 1:200 (5.8A, BD Biosciences Pharmingen); all MyHC (MF-20, DSHB); Myf5, 1:1000 (C-20, Santa Cruz Biotechnology); Pax7 (DSHB); and myogenin (F5D, DSHB). Secondary detection was performed using fluorescein- or rhodamine-conjugated antibodies, 1:200 (Chemicon). Myf5nLacZ expression was detected by X-Gal reaction as described previously ( Polesskaya et al. 2003 ). RT-PCR and Northern blot analysis Total RNA was extracted using RNeasy kits (Qiagen, Valencia, California, United States), according to manufacturer's instructions. RT-PCR analysis for endogenous Pax7 mRNA was performed using the GeneAmp PCR Core kit (PerkinElmer, Wellesley, Massachusetts, United States). RT-PCR using 1 μg of total RNA was conducted as per manufacturer's instructions with the following modifications. cDNA synthesis was extended for 1 h at 42 °C, and 5 μl of the first-strand RT product was used for PCR amplification. PCR conditions for endogenous Pax7 were 94 °C for 5 min; 35 cycles of 94 °C for 45 s, 56 °C for 45 s, 72 °C for 45 s; and finally 72 °C for 7 min. The PCR primers span intron 8 of the Pax7 gene (Pax7-exon8-fwd 5′ gct acc agt aca gcc agt atg 3′ and Pax7-exon9-rev 5′ gtc act aag cat ggg tag atg 3′) and amplify sequence in the 3′-UTR of the gene that is not contained in the viral Pax7 expression cassette. RT-PCR products were analyzed by electrophoresis through a TAE-ethidium-agarose gel. Northern blot studies were performed according to standard techniques using random-primed 32 P-dCTP radiolabeled cDNA fragments as probes (Redi-prime, Amersham Biosciences)( Sabourin et al. 1999 ). Fifteen micrograms of total RNA from various cell cultures was electrophoresed in denaturing formaldehyde gels and transferred to Hybond-N filters (Amersham Biosciences). Supporting Information Accession Numbers The accession numbers for the proteins discussed in this paper are Desmin (LocusLink ID 13346), mouse MyoD (GenBank NM_010866), MyoD (LocusLink ID 17927), Myogenin (LocusLink ID 17928), Pax7 (LocusLink ID 18509), and Pax7d isoform (GenBank AF_254422).	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC406392.xml
517512	Predictors of vigorous exercise adoption and maintenance over four years in a community sample	Background Very little is known about the correlates of adoption and maintenance of vigorous exercise. The purpose of this study was to understand the sociodemographic correlates of exercise adoption and maintenance in a community sample. Methods 917 women and 229 men completed annual surveys as part of a community-based weight gain prevention trial over four years. Multivariate regressions evaluated predictive factors for maintenance of vigorous exercise over time in regular exercisers, and predictors of adoption of exercise in adults who were sedentary at baseline. Results Exercise maintenance at Years 2 and 3 was associated with ethnicity and exercise level at baseline, while exercise maintenance at Year 4 was associated with television watching, BMI and exercise at baseline. Exercise level at baseline was associated with exercise initiation at Year 2 and Year 3. Income level, marital status, and smoking status predicted exercise initiation at Year 4. Conclusions Predictors of vigorous exercise maintenance were more consistent than predictors of vigorous exercise initiation. Results suggest that those who adopt vigorous exercise are a heterogeneous group and intervention messages could be more broadly focused. These data also suggest that exercise maintenance interventions should continue to target low-income populations with messages regarding smoking, weight and television. Clearly further research is needed to understand the factors that contribute to exercise initiation and maintenance, and to develop effective interventions to improve levels of physical activity levels.	Background Vigorous exercise is considered one of the key-components of a healthy lifestyle and cardiovascular fitness. Higher levels of exercise are associated with lower risks of hypertension, diabetes, osteoporosis, colon cancer, coronary heart disease, depression, anxiety, and has been shown to enhance weight control [ 1 - 3 ]. While earlier public health recommendations targeted vigorous physical activity, current recommendations include both vigorous and moderate physical activity [ 4 ]. Current vigorous activity recommendations include "increasing the proportion of adults who engage in vigorous physical activity that promotes the development and maintenance of cardiorespiratory fitness three or more days per week for 20 or more minutes per occasion" [ 5 ]. Although the benefits of exercise are well documented, research shows that only 15–25% of adults are engaging in vigorous exercise three or more times a week for at least 20 minutes [ 2 , 6 ]. More than 60% of American adults are not exercising at a moderate or vigorous level frequently enough to reap the health benefits [ 2 ]. Nearly 25% of all U.S. adults report no leisure time exercise at all [ 6 ]. Numerous studies have examined predictors of exercise maintenance among participants in organized exercise programs [ 7 ]. However, data using longitudinal community based studies are sparse. Two community studies that examined the maintenance and adoption of exercise showed that maintenance of vigorous activity over one year was associated with attitudes toward exercise, exercise knowledge, female gender, and self-efficacy [ 8 ]. Over two years, vigorous physical activity maintenance was associated with self-efficacy and younger age for initially active men and with education for initially active women [ 9 ]. Adoption of moderate activity over one year was associated with health knowledge [ 8 ]. Adoption of vigorous activity over two years was associated with self-efficacy, younger age, and neighborhood environment in men, and education, self-efficacy, and friend and family support in women [ 9 ]. Although more research is being conducted on correlates of adoption and maintenance of exercise, more longitudinal research is needed among demographically diverse community samples [ 10 ]. Considering only 10–25% of community residents have been successful in adopting even short-term leisure time exercise [ 8 , 11 , 12 ] there is a need to learn more about the people in the community who adopt and continue exercising. Research on potential correlates of long-term maintenance, initiation, and change in these correlates is needed to better understand how changes in life circumstances might be associated with changes in vigorous exercise. The present study sought to cross-sectionally and prospectively examine the correlates of regular vigorous exercise in a sample of community adults over a four year period. Specifically, this study attempted to expand on the current literature by understanding the relationship between regular vigorous exercise and exercise adoption and maintenance by evaluating three comparisons in a community sample: 1) cross-sectional differences in characteristics of regular vigorous exercisers compared to non-exercisers, 2) predictive factors for maintenance of vigorous exercise over time in people who are already exercising at a high level as compared to those who do not maintain their exercise level, and 3) predictors of adoption of exercise in community adults who were sedentary at baseline as compared to those who do not adopt vigorous exercise. Methods Sample The study sample included 917 women and 229 men who completed baseline surveys as part of a 3-year community-based weight gain prevention trial (Pound of Prevention; POP). Participants were recruited using a variety of methods, including direct mail sent to university employees, and advertisements in community newspapers, health department employee newsletters and radio public service announcements. In addition, recruitment also targeted lower-income women at commercial shopping centers and at community health department clinics. Lower income women were paid $20 to enroll in the study. All participants were informed that they would be randomly assigned to either a mail-based educational program, or a no contact control group and that they would be measured once per year for a total of 4 years. This study was approved by the University of Minnesota Institutional Review Board. Measures Study participants completed questionnaires and were measured for height and weight at baseline and at three annual data collection visits (Years 2, 3 and 4) following baseline. Measures included in the present study are listed below. Vigorous exercise Planned vigorous exercise behavior was assessed using 5 questions from a self-administered version of the Physical Activity History questionnaire (PAH). The PAH is reliable and valid and has been used in several large epidemiologic studies [ 13 ]. For this study, exercise activities were limited to those that would deliver significant cardiovascular benefit. Participants rated how frequently in the past year they had engaged in one or more of the activities listed. Response choices on the PAH included; never or less than once per month; 1–3 per month; 1–2 per week; 3–4 per week; or 5 or more per week. Five questions were used in this study to represent high intensity planned exercise; 1) Vigorous jogging, running, backpacking or mountain climbing, 2) Bicycle faster than 10 MPH or exercise hard on an exercise bicycle or rowing machine, 3) Vigorous exercise class or vigorous dance, 4) Brisk walking, hiking, skating or cross-country skiing or 5) other vigorous exercise (including lap distance swimming, vigorous racket sports and other strenuous sports such as competitive basketball, football, volleyball and soccer). Hours spent watching television To evaluate competing activities, we evaluated participant's report of time spent watching television. Time spent watching television has been positively correlated with body weight, presumably in part due to its displacement of exercise behaviors [ 14 ]. Participants reported the number of hours they watch television on an average day. Social Support Social support from family and friends has been found to be related to exercise in a number of studies [ 9 ]. Participants in this study ranked both their family and friends on the extent to which they were supportive of healthy eating and exercise behaviors on a 5 point scale, with 1 representing "Not at all helpful" to 5 representing "Very helpful". Body Mass Height was measured to the nearest centimeter using a wall-mounted ruler and weight was measured to the nearest half pound using calibrated balance beam scales. Body mass index was calculated using the formula weight (kg)/height (m2). Demographic Information All demographic information was self-reported and included age, educational attainment (highest level completed), gender, employment status, income, ethnicity, smoking status and marital status. Statistical Analysis Before performing analyses evaluating the predictors of exercise maintenance and initiation, the possibility of an intervention effect was evaluated. Maintenance of high intensity exercise at Year 2 was not significantly associated with treatment, however, maintenance of high intensity exercise at Years 3 and 4 were associated with treatment status. Intervention participants were more likely to be maintaining exercise behavior at Year 3 (X 2 = 14.14, p = .042) and at Year 4 (X 2 = 6.42, p = .011). Initiation of exercise at Year 2 and Year 3 were not significantly associated with treatment status. However, initiation of exercise at Year 4 was marginally positively associated with treatment (X 2 = 3.30, p = .069). Due to these associations, treatment assignment (intervention vs control) was controlled in multiple regression analyses. Participants were classified as regular vigorous exercisers (> 3 times per week) and non-regular exercisers (< 3 times per week) based on their responses to the five planned high-intensity exercises on the PAH at each of the 4 assessment points. Using these classifications (yes/no vigorous exercise), the participants were classified as maintainers if they exercised more than three times per week at both baseline and at a subsequent annual evaluation (Year 2, 3, or 4), and non-maintainers if they exercised three times per week at baseline and less than three times per week at a subsequent annual evaluation (Year 2, 3, or 4). Similarly, participants were classified as initiators if they reported less than three exercise sessions per week at baseline and more than three exercise sessions per week at subsequent annual evaluation (Year 2, 3, or 4), and non-initiators if they exercised less than three times per week at both baseline and a subsequent annual evaluation (Year 2, 3, or 4). Descriptive analyses using t-tests and chi-square tests were utilized to assess univariate associations between the baseline predictor variables and exercise status. Maintainers were compared with non-maintainers, and consistently sedentary participants were compared with participants who adopted vigorous activity at Years 2, 3 and 4. Logistic regression was used to further assess the statistical significance of associations between the predictor variables and exercise maintenance and adoption. These univariate logistic regression models predicted exercise maintenance (yes/no) and exercise initiation (yes/no) at Years 2, 3 and 4 using the baseline demographic factors as independent variables. Furthermore, the variables that were significant in the univariate analyses were entered into a multivariate logistic regression model to assess the relative weight and statistical significance of associations between the predictor variables and the variables representing exercise maintenance and initiation. These multivariate logistic regression models predicted exercise maintenance (yes/no) and exercise initiation (yes/no) at Years 2, 3 and 4. In these multivariate regressions, we controlled for exercise level at baseline and treatment status. All statistical analyses were conducted using the SAS Version 6.12 [ 15 ]. Since other research studies have indicated gender differences in predictors of exercise maintenance and initiation [ 16 ], the gender by exercise interaction was evaluated in a regression model using the frequency of vigorous exercise per week at Year 2 as the dependent variable. The independent variables in this model were the main effects for gender, the main effect for vigorous exercise at baseline, and the gender by vigorous exercise at baseline interaction. The interaction was not significant and thus the data were not stratified for analyses. Results Cross-sectional sample description Using the definition of vigorous exercise described above, 564 participants (453 females and 111 males) vigorously exercised more than 3 times per week and 582 participants (464 females and 118 males) did not vigorously exercise more than 3 times per week at baseline. Sociodemographic characteristics of these participants are described in Table 1 . Not surprisingly, participation in vigorous exercise was associated with being employed, non-smoker, lower BMI, fewer hours spent watching television, and higher perceived social support from family and friends. None of the other demographic variables were significantly associated with vigorous exercise at baseline. Table 1 Correlates of exercise status at baseline Exercisers < 3 X/wk at baseline N = 582 Exercisers > 3 X/wk at baseline N = 564 Gender Female 464 (79.7%) 453 (80.3%) Age (years) 37.6 (SD = 6.3) 37.6 (SD = 7.2) BMI (kg/m2) 28.1 (SD = 6.3) 26.2 (SD = 5.2)* Employed Yes 474 (81.4%) 495 (87.8%) Ethnicity White 500 (85.9%) 503 (89.1%) Income group < 25 K 218 (37.5%) 190 (33.8%) Marital Status Married 288 (49.5%) 266 (47.2%) Sep/Div/Wid 96 (16.5%) 96 (17.0%) Never Married 198 (34.0%) 202 (35.8%) Education HS Degree or less 73 (12.54%) 70 (12.4%) Some college 235 (40.4%) 192 (34.0%) College degree or more 274 (47.1%) 302 (53.6%) Smoking Yes 131 (22.5%) 80 (14.2%)* TV hours/day 2.6 (SD = 2.5%) 2.1 (SD = 1.9%) Social support family 2.7 (SD = 1.3) 2.9 (SD = 1.3) Social support friend 2.6 (SD = 1.2) 3.0 (SD = 1.2)* * denotes significant difference at the p < .001 Prospective analyses evaluating predictors of exercise maintenance Prospective univariate evaluations of maintenance of exercise at the three annual evaluations Maintainers and non-maintainers were compared on baseline demographic variables, social support, hours spent watching television and smoking. Similar patterns of associations were found for annual visits at Years 2, 3 and 4. Participants who maintained their exercise level at evaluation Years 2, 3 and 4 weighed less, were employed, Caucasian, of a higher income group, more highly educated, more likely to be non-smokers and watched less television per day at baseline. At evaluation Years 3 and 4, the maintainers were older than the non-maintainers at baseline. The maintainers and non-maintainers did not report differences in marital status or social support. These results are presented in Table 2 . Table 2 Associations between exercise maintenance and baseline demographic, smoking, social support and hours watching television. Year 2 Year 3 Year 4 Maintain N = 378 Non-maintain N = 186 Maintain N = 277 Non-maintain N = 230 Maintain N = 216 Non-maintain N = 256 Gender Female 300 79.4% 153 82.3% 214 77.3% 194* 84.4% 170 78.7% 209 81.6% Age (years) 38.0 SD = 6.6 36.9 SD = 8.4 38.3 SD = 6.6 36.9 SD = 8.0 38.7 SD = 6.5 37.1 SD = 7.9 BMI (kg/m2) 25.5 SD = 4.7 27.8* SD = 6.0 25.7 SD = 4.7 26.7 SD = 5.8 25.4 SD = 4.6 27.0* SD = 5.8 Employed Yes 341 90.2% 154* 82.8% 250 90.3% 197 85.7% 200 92.6% 217* 84.8% Ethnicity White 350 92.6% 153* 82.3% 259 93.5% 192* 83.5% 203 94.0% 222* 86.7% Income Group < $25,000 111 29.4% 79* 42.5% 82 29.6% 89 38.9% 57 26.4% 108* 42.2% Marital Status Married 181 47.9% 85 45.7% 130 46.9% 108 47.0% 106 49.1% 115 44.9% Sep/Div/ Widowed 59 15.6% 37 19.9% 42 15.2% 39 17.0% 35 16.2% 49 19.1% Never Married 111 29.4% 79 42.5% 105 37.9% 83 36.1% 75 34.7% 92 35.9% Education HS or less 35 9.26% 35 18.8% 23 8.3% 40 17.4% 15 6.9% 41 16.0% HS + some college 128 33.8% 64 34.4% 93 33.6% 79 34.4% 73 33.8% 96 37.5% College or more 215 56.9% 87* 46.8% 161 58.1% 111 48.3% 128 59.3% 119* 46.5% Smoking yes 42 11.1% 38* 20.4% 29 10.5% 39 16.7% 24 11.1% 45 17.6% TV/day 1.8 SD = 1.8 2.5* SD = 2.20 1.78 SD = 1.6 2.4* SD = 2.1 1.7 SD = 1.5 2.4* SD = 2.1 Social support family 3.0 SD = 1.4 2.8 SD = 1.2 3.0 SD = 1.4 2.9 SD = 1.3 3.0 SD = 1.4 2.9 SD = 1.3 Social support friend 3.0 SD = 1.2 3.1 SD = 1.2 3.0 SD = 1.2 3.0 SD = 1.2 3.0 SD = 1.2 3.0 SD = 1.2 denotes significant difference at the p < .05 * denotes significant difference at the p < .001 Prospective multivariate predictors of exercise maintenance Multivariate logistic regression analyses were performed to examine the predictors exercise maintenance at Years 2, 3 and 4. Demographic variables at baseline (BMI, income group, employment status, education, ethnicity, smoking status, hours watched TV/day, age, gender) were entered into logistic regression model along with exercise level at baseline and treatment group. Exercise maintenance at Years 2, 3 and 4 were the dependent variables in the three models. At Year 2, ethnicity (OR = .52, CI = .27–.99), BMI (OR = .93, CI = .89–.96) and exercise at baseline (OR = 1.21, CI = 1.11–1.31) were significant predictors of exercise maintenance. Ethnicity (OR = .463, CI = 0.229–0.933) and exercise at baseline (OR = 1.21, CI = 1.12–1.31) were significant predictors of exercise maintenance at Year 2. Television hours per day (OR = 0.86, CI = .74–.99), BMI (OR = .95, CI = .91–.99) and exercise at baseline (OR = 1.22, CI = 1.13–1.31) were significant predictors of exercise maintenance at Year 4. Prospective analyses evaluating predictors of exercise initiation Univariate evaluations of initiation of exercise at the three annual evaluations Results of univariate analyses relating exercise initiation with baseline variables are shown in Table 3 . Participants who initiated a vigorous exercise program differed from consistently sedentary persons in ways that were similar to those observed for exercise maintenance. These results, however, were weaker in magnitude and not as consistent over the three time periods as the results from the exercise maintenance analyses. At baseline, initiators at Year 2 had a higher income, reported higher social support from family and friends, and had a lower BMI. The only significant difference between the initiators and the sedentary persons was hours watching television per day at Year 3. Finally, initiators at Year 4 were more likely to be married at baseline. Table 3 Associations between exercise initiation and baseline demographic, smoking, social support and hours watching television. Year 2 Year 3 Year 4 Adopt N = 135 Sedentary N = 447 Adopt N = 73 Sedentary N = 347 Adopt N = 48 Sedentary N = 326 Gender Female 108 80.0% 356 79.6% 51 69.9% 305*** 81.5% 36 75.0% 269 82.5% Age (years) 37.9 SD = 6.3 37.5 SD = 6.3 38.2 SD = 5.2 37.4 SD = 6.5 37.4 SD = .6.7 37.4 SD = 6.6 BMI (kg/m2) 27.34 SD = 6.1 28.4* SD = 6.3 27.6 SD = 5.5 28.5 SD = 6.4 28.3 SD = 6.9 28.6 SD = 6.4 Employed Yes 114 84.4% 360 80.5% 63 86.3% 297 79.4% 38 79.5% 259 79.5% Ethnicity White 118 87.4% 382 85.5% 63 86.3% 319 85.3% 42 87.5% 277 84.5% Income Group < $25,000 41 30.4% 177** 39.6% 26 35.6% 151 40.4% 20 41.7% 131 40.2% Marital Status Married 69 51.1% 219 49.0% 33 45.2% 186 49.7% 30 62.5% 156 47.9% Sep/Div/ Widowed 22 16.3% 74 16.5% 13 17.8% 61 16.3% 9 18.8% 52 16.0% Never Married 44 32.59% 154 34.5% 27 37.0% 127 34.0% 9 18.8% 18.8* 36.2% Education HS degree or less 14 10.4% 59 13.2% 7 9.6% 52 13.9% 5 10.4% 47 14.4% HS degree + some college 49 36.3% 186 41.6% 25 34.3% 161 43.1% 19 39.6% 142 43.6% College degree + more 72 53.3% 202 45.2% 41 56.2% 161 43.1% 24 50.0% 137 42.0% Smoking status yes 25 18.5% 106 23.7% 15 20.6% 91 24.3% 6 12.5% 85 26.1% TV/day 2.4 SD = 1.9 2.7 SD = 2.6 2.1 SD = 1.6 2.8* SD = 2.8 2.6 SD = 3.2 2.8 SD = 2.7 Social support family 3.0 SD = 1.3 2.7* SD = 1.3 2.6 SD = 1.4 2.7 SD = 1.3 2.9 SD = 1.2 2.6 SD = 1.3 Social support friend 2.8 SD = 1.2 2.6 SD = 1.2 2.5 SD = 1.1 2.6 SD = 1.2 2.5 SD = 1.2 2.6 SD = 1.2 * denotes significant difference at the p < .001 denotes significant difference at the p < .05 Prospective multivariate predictors of exercise initiation Logistic regression analyses were also performed to examine the predictors of exercise initiation and Years 2, 3 and 4. Similar to the evaluations for exercise maintenance, demographic variables at baseline which were significant in the univariate analyses (baseline BMI, income group, marital status, gender, smoking, hours spent watching television, family social support and friend social support) were entered into three separate models controlling for exercise level at baseline and treatment group to identify predictors exercise initiation at Years 2, 3 and 4. Four variables were significant predictors of exercise initiation at one or more time points. Exercise level at baseline was positively related to exercise initiation at Year 2 (OR = .23, CI = 1.71–2.90) and at Year 3 (OR = 1.71, CI = 1.23–2.38). Baseline income group (OR = 2.38, CI = 1.08–5.23), marital status (OR = 6.07, CI = 0.39–0.95) and smoking status (OR = 3.53, CI = 0.14–0.91) predicted exercise initiation at Year 4. Discussion This study evaluated the demographic predictors of vigorous exercise initiation and maintenance in a community sample. To our knowledge, this is the first study to look at baseline demographic factors as predictors for exercise adoption and maintenance over a four-year period. This study found that demographic predictors were more consistent for exercise maintenance than exercise initiation over the four years evaluated in this study. Results showed that compared to those who did not maintain vigorous exercise, participants who maintained exercise over a 2, 3 and 4 year period were more likely to be employed, Caucasian, have a higher income, have more education, be a nonsmoker, watch less television and have a lower BMI. One of the interesting findings was that correlates of exercise initiation were less consistent than correlates of exercise maintenance over time. There did not appear to be a consistent pattern that described the associations between the predictors included in this study and vigorous exercise initiation. Although certain predictors were associated with exercise initiation at Years 2, 3 or 4, such as gender or BMI, none of the predictors were consistent over the years measured in the study. These results suggest that people in community populations who initiate vigorous exercise are a heterogeneous group, and there may be greater difficulties predicting who will initiate exercise. It could be much simpler to identify the participants who will be more likely to continue exercise, as compared to the participants who will adopt exercise. It is possible that changes in attitudes, life circumstances (such as sickness), new relationships, or variables that were not measured in this study may be better predictors of exercise adoption or maintenance. However, since we did not measure these variables, we can only speculate whether they may disrupt exercise patterns. The processes of vigorous exercise adoption may be better represented by theoretical understandings of exercise behavior, such as the health belief model [ 17 ], transtheoretical model [ 18 , 19 ], social cognitive theory [ 20 ], or the theory of planned behavior[ 21 ]. These results may have implications for designing and implementing exercise interventions. The results support others studies that suggest targeting exercise maintenance interventions at the lower income participants. However, to achieve the Healthy People 2010 objectives by increasing participation rate in vigorous exercise, interventions need to be designed to promote exercise adoption. Unfortunately, the present study's results only suggest that we can not characterize this group well. This lack of characterization of this group may actually be a benefit for interventions. Ethnic group, gender, and socio-economic status did not predict exercise initiation in this study. This suggests that each group was as likely to begin a vigorous exercise program. This is good news as exercise initiation may not be limited to those who can afford health clubs and trainers, or those who are the member of a specific ethnic group or gender. Although previous studies have stratified participants by gender [ 16 ], no interaction was found in this study between gender and exercise status. The only associations with gender were seen during Year 3, when fewer women than men maintained exercise and fewer women than men initiated exercise. It is possible that the gender by exercise interaction was not significant due to the limitation of exercise in the vigorous activity range, versus the moderate or mild activity range. One hypothesis is that the predictors of exercise adoption that are gender specific may have to do with moderate or minimal exercise adoption. Strengths and limitations of the present study should be recognized. This is one of a few studies that evaluate predictors of exercise initiation, and the first to evaluate exercise initiation over a four-year period. In addition, the sample is large, diverse, and includes longitudinal measurements. This study did not evaluate the predictors of moderate exercise initiation or maintenance. Thus, these results can not be directly interpreted for public health interventions targeted at increasing moderate activity in the general population. In addition, this study may also be affected by a selection bias. The participants in this study volunteered for a study on weight gain prevention, for which the participants may have expected to include a message on increasing exercise. Of note, 49% of the sample reported vigorous exercise more than 3 times per week at baseline, which is much higher than the 15–25% reported in national surveillance studies [ 6 , 2 ]. Although treatment status was controlled for in the analyses, these participants were recruited for an intervention trial, rather than a cohort trial, which could also contribute to a selection bias. In addition, the measures in this study are self-report and include few theoretically based variables. It is possible that the self-report nature of exercise in this study may have allowed over-reporting of vigorous exercise. Considering these limitations, this study does add to the knowledge base about who initiates and maintains vigorous exercise. The results suggest that vigorous exercise maintenance interventions should continue to target low-income populations and that interventions could incorporate messages regarding smoking, weight control, and television. We found that those who adopt vigorous exercise are a more heterogeneous group, and that no one group is more likely to adopt exercise than the others. This suggests that vigorous exercise intervention messages could be more broadly based. This study also suggests that further research is needed to identify participants and effective interventions for those who begin exercise programs. Competing interests None declared. Authors contributions KB conceptualized the study, planned and executed the analyses, interpreted the results, and drafted the manuscript. RJ participated in conceptualization of the study, interpretation of the results and assisted in drafting the manuscript. SF participated in conceptualization of the study, interpretation of the results and assisted in drafting the manuscript. All authors read and approved the final manuscript.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517512.xml
538283	Foxo3a induces motoneuron death through the Fas pathway in cooperation with JNK	Background Programmed cell death of motoneurons in the developing spinal cord is thought to be regulated through the availability of target-derived neurotrophic factors. When deprived of trophic support, embryonic spinal motoneurons in vitro over-express FasL, a ligand activating a Fas-mediated death pathway. How trophic factors regulate the expression of FasL is presently unclear, but two regulators of FasL, FOXO3a (FKHRL1) and JNK have been described to play a role in other cell types. Thus, their potential function in motoneurons was investigated in this study. Results We show here that as a result of removal of neurotrophic factors and the consequent reduction in signalling through the PI3K/Akt pathway, Foxo3a translocates from the cytoplasm to the nucleus where it triggers cell death. Death is reduced in Fas and FasL mutant motoneurons and in the presence of JNK inhibitors indicating that a significant part of it requires activation of the Fas/FasL pathway through JNK. Conclusions Therefore, in motoneurons as in other cell types, FOXO transcriptional regulators provide an important link between other signalling pathways and the cell death machinery.	Background During development of higher vertebrates, motoneurons within the spinal cord are generated in excess, and about half the cells initially generated undergo programmed cell death (PCD) during the days following target muscle contact [ 1 ]. The most frequently proposed explanation for this death is that motoneurons compete for access to limiting quantities of neurotrophic factors produced by their target tissue, and that only those which are successful survive (reviewed in [ 2 ]). Primary motoneurons purified from embryonic spinal cords and cultured in the absence of neurotrophic support mimic this process; many of them undergo programmed cell death over a period of 2–3 days [ 3 , 4 ]. Cell death in these conditions results from lack of activation of the survival pathways which normally inhibit the PCD machinery (reviewed in [ 5 ]). Thus, it is essential to identify the precise mechanisms by which motoneurons die, and the ways in which removal of trophic factors leads to their activation. We have shown that a major driving force for the death of motoneurons deprived of neurotrophic factors in vitro is activation of the Fas/CD95 death receptor by its cognate ligand, FasL [ 6 ]. Fas and FasL are expressed by embryonic motoneurons at the stage at which naturally-occurring PCD is about to occur [ 6 ]. While levels of Fas are not affected by the presence or absence of neurotrophic factors, FasL expression is strongly upregulated in motoneurons cultured for 3 days without neurotrophic factors [ 6 ], as in cerebellar granule neurons [ 7 ]. Moreover, reagents such as Fas-Fc which prevent FasL from activating Fas save a majority of motoneurons from death in the absence of trophic support, presumably by blocking interactions in cis between FasL and Fas on individual motoneurons [ 6 ]. Understanding how expression of FasL is upregulated in motoneurons is thus an important step in linking neurotrophic signalling to cell death mechanisms. The transcription factor Foxo3a (also known as FKHRL1) was a clear candidate. In conditions in which the PI3K/Akt survival and growth pathway is activated, Foxo3a is phosphorylated by Akt and exported to the cytoplasm where it is sequestrated by the 14-3-3 protein [ 8 ]. Overexpression of a constitutively activated form of Foxo3a (mutated at the three Akt phosphorylation sites and therefore unable to be phosphorylated) leads to PCD of many cell types in culture, including primary cerebellar granule neurons [ 8 - 14 ]. The FasL promoter contains three FOXO DNA-binding sites, and Foxo3a-induced apoptosis of cerebellar neurons is decreased when Fas/FasL interaction is blocked by the decoy fusion protein Fas-Fc [ 8 ]. Thus, in these cells, Foxo3a induces apoptosis in part by its ability to induce the expression of the FasL gene. The JNK pathway has also been shown to regulate FasL expression in some neuronal cells, through its effects on the transcriptional activity of the AP-1 complex. Although this pathway can play different roles, in neurons it is involved in apoptosis in response to several stresses, including withdrawal of survival factors [ 15 ]. In cerebellar granule neurons, FasL upregulation in neurons induced to die results from JNK activation and phosphorylation of c-Jun [ 7 ]. Moreover, in cerebellar neurons derived from gld mice, which are defective for FasL, the killing effect induced by trophic deprivation is reduced compared to wt mice. We therefore wished to study the function of Foxo3a in motoneurons and its relation to JNK signalling. We show that, in the absence of survival signalling through the Akt pathway, Foxo3a is translocated to the nucleus where it triggers motoneuron death. Using appropriate mouse mutants, we show that Fas signalling is required for roughly half of the cell death induced by Foxo3a, suggesting that FasL activation both directly and through JNK is a major target of its actions. Results Control of subcellular localization of Foxo3a in motoneurons We first analysed the subcellular localization of Foxo3a in motoneurons using an antibody that recognizes Foxo3a independently of its phosphorylation state. Motoneurons were purified from mouse embryonic spinal cord at E12.5, at the beginning of the period of naturally-occurring cell death, and cultured in the presence of a cocktail of neurotrophic factors (NTFs; see Methods ) to strongly activate the PI3K/Akt pathway (Perez-Garcia et al., 2004; Dolcet et al., 2001; Dolcet et al., 1999). In these conditions, Foxo3a is predominantly detected in the cytoplasm (Fig. 1 ). To confirm that exogenous Foxo3a adopts the same subcellular localization, we electroporated purified motoneurons with a vector encoding HA-tagged wildtype Foxo3a (HA-wt-Foxo3a) together with a vector encoding GFP to identify transduced neurons. When motoneurons were grown in the presence of NTFs, only cytoplasmic staining for HA-wt-Foxo3a was detected (Fig. 2A ). We then tested the effects of inhibiting survival signalling through the PI3K/Akt pathway by treatment with the PI3K inhibitor LY294002 either in the presence (not shown) or the absence (Fig. 2A ) of neurotrophic factors. In both cases, the majority of HA-wt-Foxo3a becomes localized in the nucleus of motoneurons (Fig. 2A and not shown). This pharmacological evidence suggests that Foxo3a localization reflects its phosphorylation by Akt. To confirm this, we studied the cellular distribution of the non-phosphorylatable form of Foxo3a, triple mutant (TM-) Foxo3a, in which all three Akt phosphorylation sites are mutated (Brunet et al ref). In the absence of survival signalling, HA-TM-Foxo3a showed a very similar nuclear distribution to HA-wt-Foxo3a (Fig. 2B ). However, in contrast with the wildtype form, no redistribution of HA-TM-Foxo3a to the cytoplasm was observed in the presence of the cocktail of neurotrophic factors (Fig. 2B ). Results in all conditions were quantified by counting 2 independent experiments (Fig. 2C ). They clearly show that, in motoneurons as in other cell types, Akt-induced phosphorylation of Foxo3a is required to prevent its accumulation in the nucleus. Activated Foxo3a triggers motoneuron death To investigate the ability of Foxo3a to trigger death of cultured motoneurons, we took advantage of the electroporation technique we recently developed for high-efficacy transduction of primary neurons [ 16 ]. We first confirmed that electroporation did not intrinsically inhibit motoneuron survival, by overexpressing a constitutively active form of Akt (Akt ca) in which the PH domain is replaced by a myristyl moiety which constitutively targets Akt to the membrane. Survival of electroporated motoneurons was reproducibly increased by overexpression of Akt ca as compared to the empty vector (Fig. 3A ). Indeed, survival with Akt ca was greater than with NTFs alone. This was a result of increased Akt activity, since motoneurons electroporated with wild-type Akt showed survival values no greater than those with empty vector (not shown). We therefore used electroporation to analyse the effects of Foxo3a on motoneuron survival. Overexpression of HA-wt-Foxo3a with GFP had no significant effect as compared to GFP alone or GFP with the empty vector (GFP: 100 ± 10; GFP + HA-wt-Foxo3a: 116 ± 12). To mimic Foxo3a activation and nuclear translocation, we then overexpressed HA-TM-Foxo3a together with GFP. The triple mutant triggered significant death of motoneurons cultured with or without NTFs (Fig. 3B ). In the presence of NTFs, survival was reduced by 65%, which is similar to the proportion of motoneurons that die in the absence of trophic support (Fig. 3B ). FasL is a target of Foxo3a for motoneuron death One candidate gene downstream of Foxo3a was FasL , known to be regulated by Foxo3a and shown by us to be able to trigger motoneuron death. Motoneurons were therefore isolated from mice bearing mutations that lead to reduced signalling through the Fas pathway: gld mice (point mutation in FasL) and lpr mice (regulatory defect in Fas). As before, electroporation of HA-TM-Foxo3a in control motoneurons led to loss of 60 to 70 % of them compared to HA-wt-Foxo3a (Fig. 4 ). This figure was reduced from 70 to 50 % in gld motoneurons, and from 60 to 35% in lpr mutants. Thus, induction of Fas signalling is responsible for approximately half of the cell killing induced by Foxo3a. Foxo3a acts partly through JNK to trigger motoneuron death In other cell types, Foxo3a can upregulate FasL either by direct interaction with the Foxo3A promoter [ 8 , 17 ] or indirectly through JNK activation [ 18 ], which itself leads to upregulation of FasL [ 7 ]. Motoneurons were therefore electroporated with either wild-type or TM-Foxo3a, and treated immediately after plating with an inhibitor of JNK, L-JNKI1, which has been shown to inhibit the interaction between JNK and its substrates. In the presence of NTFs, JNKI1 had no effect on the survival of motoneurons electroporated with wt Foxo3a or the empty vector. In contrast, inhibition of JNK led to a reduction of 20% in the number of motoneurons triggered to die by TM-Foxo3a (Fig. 5 ). Since this percentage is lower than the fraction saved by reduced Fas signalling, it is likely that both direct and indirect (through JNK) mechanisms are used by Foxo3a to trigger Fas-dependent death of motoneurons. Discussion Death of motoneurons as a result of insufficient trophic support was one of the first examples of developmental PCD to be discovered, but we still only partially understand the underlying mechanisms. We show here that as a result of removal of neurotrophic factors and the consequent reduction in signalling through the PI3K/Akt pathway, Foxo3a translocates from the cytoplasm to the nucleus where it triggers cell death. A significant part of this death requires activation of the Fas/FasL pathway through JNK. Thus, in motoneurons as in other cell types, FOXO transcriptional regulators provide an important link between other signalling pathways and the cell death machinery. We overexpressed the constitutively active mutant TM-Foxo3a in the presence of neurotrophic factors to mimic the nuclear translocation of Foxo3a in their absence. We observed a 50% reduction in death induced by TM-Foxo3a when we used motoneurons that were mutant for Fas/FasL signalling. This could in theory result through several indirect mechanisms. However, the literature on direct regulation of FasL by Foxo3a [ 8 ] and our earlier demonstration of upregulation of FasL in motoneurons deprived of trophic support [ 6 ] make it likely that TM-Foxo3a is acting through FasL in these cells. This may not only reflect direct upregulation of the FasL gene. The partial protection obtained using a JNK inhibitor suggests that in some motoneurons, Foxo3a acts to upregulate FasL through a parallel pathway already described in other cell types, involving JNK and the AP-1 complex [ 7 , 18 ]. The importance of the Fas/FasL pathway in motoneuron death during development remains to be determined. However, both in vitro and in vivo , recent studies clearly demonstrate a potential role in pathological motoneuron loss. After axotomy of the facial nerve in neonates, there is a massive loss of motoneurons over the following week. The numbers of surviving motoneurons are increased 2-fold in mice deficient for the Fas/FasL pathway [ 19 ]. In vitro , Fas engagement leads to activation of a motoneuron-specific signalling pathway involving p38 kinase and neuronal nitric oxide synthase. Embryonic motoneurons purified from mouse models of familial amyotrophic lateral sclerosis (ALS) show greatly exacerbated death responses to activation of this pathway [ 16 ]. It will therefore be of interest to determine whether regulation of Foxo3a plays a role in determining the responsiveness of motoneurons to cell death activation in pathological situations as well. The killing effect of Foxo3a was not totally abolished in motoneurons from lpr or gld mice. This may in part reflect the fact that these mutants are strong hypomorphs, rather than complete nulls. For example, in lpr mice, some Fas is still synthesized following splicing-out of the inhibitory transposon from primary transcripts [ 20 ]. However, it is equally likely that Foxo3a induces death of some motoneurons through other, Fas-independent mechanisms. Foxo3a has been shown to regulate the expression of the cyclin-dependent kinase inhibitor p27 kip1 [ 11 ], the glucocorticoid-induced leucine zipper protein [ 21 ], transforming growth factor-b2, [ 22 ], the DNA damage-induced protein Gadd45a [ 23 ] and the ubiquitin ligase atrogin-1 [ 24 ]. Of particular interest is its well-characterized effect on expression of Bim (Bcl-2-interacting mediator of cell death), a proapoptotic member of the Bcl-2 family that contains only the BH3 domain, which allows it to bind anti-apoptotic Bcl-2 family members and neutralize their function. During death of sympathetic neurons induced by NGF deprivation, Foxo3a directly activates Bim expression and thereby triggers cell death [ 14 ]. Moreover, JNKs have been shown to potentiate trophic deprivation-induced apoptosis in cerebellar granule cells, through phosphorylation of Bim (Putcha et al., 2003). The following unpublished results from our laboratory make it probable that a similar mechanism occurs in purified motoneurons. Three isoforms of Bim are produced by alternative splicing in both mouse and human: Bim S , Bim L and Bim EL [ 25 ]. We detected all three isoforms by RT-PCR in motoneurons isolated from E12.5 ventral spinal cord of embryonic mice ([see Additional file 1 ], part A). Moreover, >50% of motoneurons are induced to die by overexpression of Bim L and 89% by overexpression of Bim S ([see Additional file 1 ], part B). Therefore, expression of Bim induced by Foxo3a would be likely to trigger motoneuron death and provides the most likely explanation for the Fas-independent actions of Foxo3a in these cells. Conclusions In conclusion, one of the upstream events now known to trigger death of neurons deprived of trophic support is the failure of Akt to phosphorylate Foxo3a and thereby prevent it from entering the nucleus. The fact that this mechanism is found to occur in motoneurons, a classical system for the study of neuronal cell death, opens the door to a better understanding of its role during development and in neurodegenerative pathology. Methods Animals Normal CD1 and C57BL/6 mice were obtained from Iffacredo (L'Arbresle, France). lpr/lpr and gld/gld mice were purchased from the Jackson Laboratory (Bar Harbor, ME). lpr/lpr mutants present an insertion of an early transposon into the fas gene, resulting in fas transcriptional repression [ 20 ]. gld/gld mutants show a loss-of-function mutation in the fasL gene [ 26 ]. All mutants were maintained on a C57BL/6 genetic background. Controls were done in either C57BL/6 or in CD1 mice (we previously showed that their response to Fas activation was identical). Reagents LY 294002 was purchased from Calbiochem and used at a final concentration of 100 μM. L-JNKl1 was purchased from Alexis Biochemicals and used at a final concentration of 1 μM. Rabbit polyclonal antibody to HA-tag was from Clontech. Antibodies against Foxo3A were as described previously [ 8 ] and were a generous gift from A. Brunet (Stanford). Expression constructs We are grateful to Anne Brunet for donating vectors used as the basis for our expression constructs. The vectors encoding HA-tagged Akt and Akt ca were as described previously [ 27 ]. The vectors encoding HA-tagged wt and triple-mutant forms of Foxo3A were developed by Brunet et al[ 8 ] The cDNAs were excised from the original clones and subcloned in the pCAGGS expression vector at ClaI site, since this vector gives higher expression levels in motoneurons. Motoneuron purification and culture Motoneuron cultures were prepared from E12.5 mouse spinal cords essentially as described [ 28 ], except that the magnetic column step was omitted and motoneurons in the enriched metrizamide fraction were identified by morphological criteria. Motoneurons were plated in the presence or not of a cocktail of neurotrophic factors (referred to as "NTFs": 1 ng/mL BDNF, 100 pg/mL GDNF, 10 ng/mL CNTF), added at the time of cell seeding. L-JNKl1 was added at the time of seeding, and LY294002 was added for 90 mins after 22 hours of culture. Electroporation of motoneurons Cells dissociated from E12.5 mouse ventral spinal cords were centrifuged over a 6.5% (v/v) Metrizamide cushion at 2000 rpm for 15 min. Cells at the interface were collected and washed on a BSA cushion at 1500 rpm for 5 min. Cells were resuspended in electroporation buffer at a density of 50,000 cells per 100 μL. 100 μL aliquots of the suspension were transferred to 4 mm gap cuvettes (Eppendorf) and five μg of the pCAGGS-GFP vector as well as the same molar amount of the vector of interest were added. After 15 min of incubation at room temperature, cells were electroporated using three pulses of 5 ms at 200 V with intervals of 1 s. Immediately after electroporation, 1.5 ml of culture medium was added to dilute the cells which were plated in four four-wells plates [ 16 ]. Immunocytochemistry and HA-tag localization For anti-HA staining, motoneurons electroporated with HA-tagged wt or TM Foxo3A were seeded on polyornithine-laminin-coated 12-mm diameter glass coverslips and cultured for 24 hr at 37°C in complete Neurobasal medium supplemented or not with NTFs. Motoneurons cultured without NTFs were treated with LY294002 for the last 90 min. They were then fixed in 3.6% (v/v) formaldehyde for 30 min, washed in PBS-50 mM L-Lysine, and blocked for 1 hr in 5% donkey serum, 4% BSA, 0.1% Triton X-100 in PBS-50 mM L-Lysine. Cells were incubated with the anti-HA antibody (dilution 1:500) in blocking buffer, followed by fluorochrome-conjugated anti-rabbit secondary antibody. The cells were then observed under a fluorescence microscope. Cells with typical motoneuron morphology that were both GFP- and HA-positive were analyzed. The fraction of motoneurons expressing HA-wt-Foxo3A or HA-TM-Foxo3A exclusively in the nucleus, exclusively in the cytoplasm or in both compartments was evaluated in a total of 20 cells per well. Staining for endogenous Foxo3A (dilution 1:100) was performed using the same method on non-electroporated motoneurons cultured for 24 hr with NTFs. Survival assays Motoneurons were counted 48 hr after electroporation under a fluorescence microscope. Only green cells with healthy motoneuron morphology, i.e. with large cell bodies and long non-fragmented neurites, were taken into consideration. Two different wells for each condition were counted (between 50 and 350 green motoneurons) in all the experiments. Cross percentages of test versus control (being set at 100%) were calculated yielding 4 values, used to calculate the medium and the SD for each experiment. Authors' contributions C.B. conducted most of the experiments, B.P. some. C.H and B.P. conceived the experiments and the experimental design. C.B., B.P. and C.H. wrote the paper. Supplementary Material Additional File 1 A – The 3 forms of Bim are expressed in motoneurons: RT-PCR was performed on extracts of 80,000 motoneurones cultured for 3 days in the presence of NTFs, in a 35 mm diameter dish. Primers used for Bim were the following: 5'-GTGACAGAGAAGGTGGACAAT-3' and 5'-ATACCAGACGGAAGATAAAGC-3'. The 3 products were:BimS 284 bp, BimSL 374 bp, BimL 542 bp; B – Overexpression of Bim S or Bim L kills a major proportion of motoneurons: motoneurons purified from E12.5 mice embryos have been electroporated with a vector coding GFP alone or coelectroporated with a vector coding GFP and a vector coding either BimL or BimS, obtained by direct cloning of the PCR products described above first cloned into pGEMTeasy, then subcloned into pcDNA3 into EcoRI sites. Conditions of electroporation were as described in Material and Methods. The surviving electroporated motoneurons were counted after 2 days in culture in the presence of NTFs. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC538283.xml
518972	Setting priorities in health care organizations: criteria, processes, and parameters of success	Background Hospitals and regional health authorities must set priorities in the face of resource constraints. Decision-makers seek practical ways to set priorities fairly in strategic planning, but find limited guidance from the literature. Very little has been reported from the perspective of Board members and senior managers about what criteria, processes and parameters of success they would use to set priorities fairly. Discussion We facilitated workshops for board members and senior leadership at three health care organizations to assist them in developing a strategy for fair priority setting. Workshop participants identified 8 priority setting criteria, 10 key priority setting process elements, and 6 parameters of success that they would use to set priorities in their organizations. Decision-makers in other organizations can draw lessons from these findings to enhance the fairness of their priority setting decision-making. Summary Lessons learned in three workshops fill an important gap in the literature about what criteria, processes, and parameters of success Board members and senior managers would use to set priorities fairly.	Background Hospitals and regional health authorities in Canada and elsewhere are facing significant resource allocation challenges. Priorities must be set among competing opportunities because demand for health care exceeds available resources. Board members and senior administrators are looking for practical ways to improve how they set priorities under resource constraints. The priority setting literature describes priority setting in various health care contexts [ 1 - 9 ]. It identifies a number of decision-making principles and approaches that could be used to set priorities [ 10 - 16 ]. However, very little has been reported from the perspective of Board members and senior administrators themselves about what decision-making elements (criteria and processes) they would find most useful in setting priorities or how they would evaluate the success of a priority setting exercise. Fairness is a key ethical goal of priority setting when health care resources are scarce. Experience shows that there is often disagreement on what principles should be used to make fair allocation decisions (i.e., distributive fairness) [ 8 , 17 ]. This means that decision-makers must rely instead on a fair process (i.e., procedural fairness) to establish the legitimacy of priority setting decisions [ 16 , 18 ]. Norman Daniels and James Sabin have developed a fair process model for priority setting called 'accountability for reasonableness' (A4R) [ 16 ]. Based on justice theories of democratic deliberation, A4R identifies four conditions of a fair priority setting process (Table 1 ). We, and others, have been exploring the application of A4R in various health care settings [ 19 - 23 ]. Our experience suggests that A4R can provide valuable practical guidance to improve the fairness of actual priority setting in health care organizations and to enhance public accountability for priority setting [ 9 , 23 ]. To assist decision-makers in developing fair priority setting processes, we conducted one-day workshops for Board members and senior administrators at three Canadian academic health science centres (Saskatoon Health Region, Kingston General Hospital and The Ottawa Hospital), who were seeking ethics advice on how to improve priority setting in their organisations. Each organization was faced with setting priorities among their clinical services to guide resource allocation under significant budget constraints. The goal of each workshop was to help decision-makers develop a strategy for fair priority setting based on the conditions of A4R. Using case-based plenary sessions to introduce the key concepts (e.g., a case about how one organisation developed and used criteria to set clinical service priorities illustrated the importance of priority setting criteria for operationalising the Relevance condition of A4R) and facilitating consensus through small and large group discussions, we assisted workshop participants in reaching agreement on: a) the criteria decision-makers would use to set clinical service priorities, b) the processes they would follow, and c) the parameters according to which they would evaluate the success of the priority setting exercise. We summarize key lessons learned from these workshops to help decision-makers in other health care organizations develop their own fair priority setting strategies and to improve understanding of researchers and policy makers about priority setting from the point of view of decision-makers. Discussion Presentation of lessons learned Priority setting criteria When decision-makers were asked what criteria they would use to set clinical service priorities, we found that responses clustered around eight (8) criteria (Table 2 ). As a step toward operationalising the Relevance condition of A4R, these criteria describe what decision-makers considered to be the most relevant decision factors (or 'reasons') for setting clinical service priorities in their organisations. 'Strategic fit' described the extent to which clinical services contributed to advancing the strategic directions of the organisation, i.e., "fit" with the organization's vision, mission, values, and goals. This criterion was consistent with the idea that strategy should be a key driver of operational planning as a counterpoint to planning based on historical or short-term political considerations. 'Alignment with external directives' identified existing government mandates and legislated obligations as relevant considerations for setting priorities. For example, each organisation had government directives to provide particular health services at prescribed volumes. This criterion recognised explicitly the limited degrees of freedom within which priorities could be set, but also highlighted the importance for decision-makers of participating with government in achieving regional and provincial health service objectives. 'Academic commitments' consisted of two sub-criteria reflective of each organization's close affiliation with a local university and medical school. The 'education' sub-criterion emphasised the role of clinical programs in educating future health care professionals and in facilitating the integration of these activities with health service delivery. The 'research' sub-criterion emphasised the role of academic health science centres in establishing best practice standards, in generating new medical knowledge (including practice-based and bench research), and in developing technological innovation. Workshop participants felt that this criterion affirmed the unique role of academic health science centres in advancing society's health care knowledge and capacity. 'Clinical impact' was defined primarily in terms of the service volumes necessary to ensure the clinical competence of medical staff to provide safe and effective care to patients. Other relevant factors included: evidence of effectiveness in health promotion and disease prevention, uniqueness of the health service in the local area, and quality of the service provided. Workshop participants expressed concern about their ability to measure clinical impact given the limitations of their institutional decision support capabilities (e.g., data, trained decision support staff). However, they felt that by identifying these factors, this could provide direction for the collection of appropriate data and information. 'Community need' described the health service needs of patients in the organisation's local catchment area. This included current demand for health services, which could be measured on the basis of utilisation rates and waiting list data, as well as future demand based on population data and trends (e.g., aging population). Community need was further defined in terms of the availability of other health service providers. For example, community need was seen to be greater if the organisation were the sole provider of a health service to patients in the region than if there were other local providers whom patients might access for care. 'Partnerships' highlighted existing formal agreements and commitments with other organisations in coordinating delivery of health care to defined populations (e.g., referral agreements to ensure access to speciality care, or transfer agreements to coordinate the transition of patients from a hospital to a chronic continuing care facility). Partnerships were seen as effective ways to enhance service quality and to optimise resource utilisation within the region or local catchment area. 'Interdependency' described the coordination and collaboration between clinical services within the organization to enhance service quality (e.g., through interdisciplinary models of care) or to use institutional resources more efficiently. In the two organisations that had achieving a "healthy" workplace as a strategic goal, workshop participants also related this criterion to quality of work life factors as key enablers of effective clinical coordination and collaboration. 'Resource implications' included a cluster of factors related to the mobilisation and use of human and fiscal resources. Although recognising that strategic planning should not be over-determined by operational issues, workshop participants felt that the resource context was relevant for setting clinical service priorities. For example, the implications of prioritisation depended in part on the source of funding (e.g., base hospital budget, ministry of health volume-based funding, donation), the availability of staff (e.g., nurses) and capital resources (e.g., equipment, space), the flexibility of contractual agreements (e.g., union contracts), and the model of health service delivery, which could be more or less efficient in using available resources. Priority setting processes When asked what key process elements would be needed in order for priority setting to be accountable and fair, workshop participants identified ten (10) elements (Table 3 ). Some of these process elements reflected the Publicity, Revision, and Enforcement conditions of A4R. However, decision-makers identified additional process considerations that they felt were also essential for a successful priority setting process. Workshop participants identified a number of preparatory steps that should be taken before priority setting can begin: (1) The organisation should establish, refine, or confirm its strategic plan. This is to ensure that the clinical service priorities that emerge through the priority setting process align with and advance the organisation's mission and strategic goals. In effect, workshop participants felt that they needed to know first where the organisation was going so that they could set the right priorities for getting there. (2) The programmatic architecture of the organization (i.e., what services are offered and how they are grouped administratively and programmatically) should be clarified in order to set clinical service priorities relative to current activities. This step was also felt to be important for defining precisely what order of clinical service activity was to be prioritised and for creating an accurate inventory of clinical services for prioritisation. (3) The specific responsibilities of the Board and senior management in relation to the priority setting process should be clarified explicitly and upfront. Decision-makers identified some confusion about these responsibilities given that clinical service priority setting involved an overlap of the strategic responsibility of the Board with the operational responsibility of Senior Management. During the workshop, Board members and Senior Managers drafted a memorandum of agreement delineating their respective roles and responsibilities in the priority setting process. Workshop participants also identified a number of elements that were critical to the design of the priority setting process itself: (4) The executive decision-making group should be multidisciplinary and its role should be clearly and explicitly defined in advance of priority setting. Workshop participants emphasised the importance of shared accountability for priority setting across the clinical and administrative leadership. Engaging the medical leadership in a decision-making role was identified as key to developing a successful priority setting process. The engagement of other non-medical clinical leaders (e.g., nursing leadership) was also thought to be important for ensuring the legitimacy of the priority setting process. (5) Stakeholders should be engaged in the priority setting process. Although the organisational executive would ultimately be accountable for making the priority setting decisions, workshop participants felt that stakeholders could be engaged particularly as key informants through expert and broader stakeholder consultation. This consultation should include both internal stakeholders (e.g., staff, patient advisory groups) and external stakeholders (e.g., institutional partners, community groups, government officials). (6) Priority setting criteria should be clearly defined and understood by decision-makers and stakeholders. Data/information should be collected to support their application in the priority setting process. Workshop participants felt that the criteria identified in the workshop could be further refined through stakeholder engagement and tested with decision-makers to ensure a common interpretation of each criterion and consistency in their implementation. (7) An effective communication strategy should be developed to ensure a transparent priority setting process. The purpose of the communication strategy should be to ensure that stakeholders know and understand the scope and necessity of priority setting decision-making, the degrees of freedom within which priority setting would take place (including explicit identification of any "sacred cows" that would be immune from priority setting), and the particularities of the priority setting process (who will do what, how the process will work, and why). In addition, the rationales for priority setting decisions should be communicated to stakeholders and should clearly demonstrate how these decisions are defensible in light of the priority setting criteria and available data/information. (8) Decision review processes should be developed to incorporate opportunities to revisit and review decisions. Workshop participants saw these as additional opportunities to engage stakeholders around difficult priority setting decisions, although they also expressed concern that this might invite conflict between stakeholders and decision-makers. However, it was generally felt that this could be mitigated if decision review processes were focused explicitly on providing a vehicle for new data/information to be brought forward, material errors in the original decision to be corrected based on available data/information, and procedural inconsistencies to be addressed. Workshop participants identified additional elements that were important to improve quality and strengthen capacity for fair priority setting in their organisations over time: (9) Process monitoring and formal evaluation strategies should be developed to ensure quality improvement and to realise a commitment to organizational learning. Workshop participants felt that the process should be monitored for adherence to the conditions of A4R, thus allowing for mid-course corrections to enhance fairness as the priority setting process unfolded. A formal evaluation process after priority setting would allow institutional good practices as well as opportunities for improvement to be captured so that this information could lead to improved priority setting in the future. For example, Martin & Singer have developed an ethics-based quality improvement model that focuses on evaluating and improving the fairness of priority setting processes [ 23 ]. (10) The priority setting process should be supported by leadership development and change management strategies to strengthen institutional capacity for priority setting decision-making. Capacity strengthening should focus in particular on middle managers, who may not be among the decision-making group but who would play key roles in communicating with staff and in implementing the priority setting decisions. Parameters of successful priority setting When asked how they would know that the priority setting process had been a success, workshop participants identified both outcome and process parameters (Table 4 ). In either case, key marks of its success were whether the process were perceived to be an improvement over past priority setting initiatives and whether it were implemented in subsequent iterations of priority setting. Outcome parameters focused on the effects of priority setting on organizational priorities and budget, on staff, and on the community. Effects on organizational priorities and budget concerned the extent to which the priority setting process was successful in changing organizational priorities and shifting resources, in supporting and/or enhancing the mission of the organization, in contributing to conditions for growth, and in balancing the organizational budget. Effects on staff involved an evaluation of the impact of priority setting on staff satisfaction and morale, organizational recruitment and retention initiatives, and overall understanding of new priorities across the organization. Effects on the community focused on how external stakeholders, including members of the public, regional partners, health care peers (e.g., other academic health science organisations), and affiliated academic institutions, responded to the priority setting initiative. Process parameters focused on the efficiency and fairness of the priority setting process. Efficiency of the priority setting process could be evaluated in terms of whether priority setting improved institutional capacity for allocating resources and making priority setting decisions, and whether stakeholders and decision-makers felt that the priority setting process provided a worthwhile return on the time invested to set priorities. Fairness of the priority setting process could be evaluated in terms of whether stakeholders understood and felt engaged in the priority setting process, whether priority setting decisions were justified and seen to be reasonable, and whether 'winners' and 'losers' both felt that they had been fairly treated. It was interesting to us that, although A4R was presented as an ethical framework for fair priority setting, workshop participants did not specifically identify conformity with its conditions as a parameter of success related to fairness. The importance of these conditions is clearly evident, however, among the fairness considerations they cited as well as the process elements they identified as key to setting priorities. Moreover, we had been invited to work with these executive teams precisely because they were seeking an ethical framework through which to improve how they set priorities in their organisations. This suggests to us that A4R was seen by workshop participants primarily as an ethical framework for process design rather than for the evaluation of priority setting processes ex post facto . Implications of lessons learned Our findings from these three priority setting workshops illuminate the complex challenges faced by decision-makers in managing scarce health care resources. The range of criteria identified in the workshops provides insight into the competing goals (e.g., clinical vs. academic, local vs. systemic, strategic vs. operational) and multiple stakeholder relationships that decision-makers must consider in setting clinical service priorities. This is consistent with previous findings that efficiency considerations or simple technical solutions have only limited influence on decision-making and are not sufficient alone to guide priority setting decision-making [ 8 , 17 , 24 , 25 ]. Given the range of interested stakeholders and competing values, our findings underscore the importance of procedural fairness to secure socially acceptable priority setting decisions and to ensure public accountability [ 8 , 18 , 26 ]. This suggests that a fair process model like A4R may be particularly suitable to help decision-makers set legitimate and fair clinical service priorities. Although we report only on three health care organizations, the organisations were all academic health science centres facing similar resource challenges. Consensus around priority setting criteria and processes emerged independently among workshop participants in their large and small group discussions. However, this does not mean that these findings are exhaustive of the priority setting criteria that might be relevant for setting clinical service priorities (e.g., in community hospitals without academic affiliations) or the process elements that would be necessary to ensure a legitimate and fair priority setting process. Moreover, although our approach was based on the notion that fair priority setting requires a normative grounding in procedural justice – in this case, A4R – this does not mean that these findings are normatively 'right' for clinical service priority setting in all health care organisations. An evaluation of the normative 'rightness' depends to some extent on the specific institutional circumstances under which priority setting is taking place, the stakeholders who are affected, and the strategic goals that are being pursued. Experience shows, moreover, that the conditions of A4R are sufficiently general to guide fair priority setting in various institutional settings [ 9 , 16 , 20 , 27 ]. Thus, decision-makers in other health care organisations may draw lessons from these workshops to operationalise fair priority setting processes that reflect the particularities of their institutional circumstances and ensure accountability for the reasonableness of their clinical service priorities. Our experience shows that, from the perspective of Board members and senior leaders, our practical approach using A4R offers useful guidance for developing fair and publicly accountable priority setting processes under resource constraints. However, alternative priority setting approaches may also be beneficial. For example, program budgeting and marginal analysis, an economics-based approach, has been used with senior health care administrators in Canada and elsewhere to improve how priority setting optimises health and non-health benefits within available resources [ 13 ]. A comparison of priority setting approaches has not been done, however preliminary work has begun to explore a more interdisciplinary priority setting approach (Gibson JL, Mitton C, Martin DK, Donaldson C, Singer PA, manuscript submitted) [ 21 ]. Despite these possible limitations, the lessons we report here fill an important gap in the literature about the criteria, processes, and parameters of success decision-makers would use to set priorities using an ethical framework. We expect that decision-makers in other health care organizations may find themselves in the workshop participants' experience of priority setting and may use these findings as a basis for discussing how they could enhance the fairness and public accountability of their own priority setting processes. Summary • Hospitals and regional health authorities must set priorities in the face of resource constraints. • Decision-makers seek pragmatic ways to set priorities fairly in strategic planning, but find limited guidance from the literature. • We facilitated workshops for board members and senior leadership at three organizations to assist them in developing a strategy for fair priority setting. • Workshop participants identified 8 priority setting criteria, 10 key priority setting process elements, and 6 parameters of success that they would use to set priorities in their organizations. • Decision-makers in other organizations can draw lessons from these findings to enhance the fairness of their priority setting decision-making. Competing interests The authors were compensated by the health care organizations for facilitating the priority setting workshops and continue to consult with these and other health care organizations. Authors' contributions JLG conducted the workshops on which this paper is based, collated and analysed the data, and drafted the manuscript. DKM participated in analysing the data and commented on earlier drafts of the manuscript. PAS conducted the workshops on which this paper is based, participated in analysing the data, commented on earlier drafts of the manuscript, and conceived of the paper. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC518972.xml
515305	Opposite role of Bax and BCL-2 in the anti-tumoral responses of the immune system	Background The relative role of anti apoptotic (i.e. Bcl-2) or pro-apoptotic (e.g. Bax) proteins in tumor progression is still not completely understood. Methods The rat glioma cell line A15A5 was stably transfected with human Bcl-2 and Bax transgenes and the viability of theses cell lines was analyzed in vitro and in vivo. Results In vitro , the transfected cell lines (huBax A15A5 and huBcl-2 A15A5) exhibited different sensitivities toward apoptotic stimuli. huBax A15A5 cells were more sensitive and huBcl-2 A15A5 cells more resistant to apoptosis than mock-transfected A15A5 cells (pCMV A15A5). However, in vivo, in syngenic rat BDIX, these cell lines behaved differently, as no tumor growth was observed with huBax A15A5 cells while huBcl-2 A15A5 cells formed large tumors. The immune system appeared to be involved in the rejection of huBax A15A5 cells since i) huBax A15A5 cells were tumorogenic in nude mice, ii) an accumulation of CD8+ T-lymphocytes was observed at the site of injection of huBax A15A5 cells and iii) BDIX rats, which had received huBax A15A5 cells developed an immune protection against pCMV A15A5 and huBcl-2 A15A5 cells. Conclusions We show that the expression of Bax and Bcl-2 controls the sensitivity of the cancer cells toward the immune system. This sensitization is most likely to be due to an increase in immune induced cell death and/or the amplification of an anti tumour immune response	Background Glioblastoma Multiforme (GBM) are the most common and aggressive tumors of the central nervous system (CNS) [ 1 ]. Current treatments (e.g. chemo-and radiotherapy) have been relatively unsuccessful and no significant improvement in the prognosis has been recorded over the last 20 years [ 1 ]. Cancer cells are capable of inducing a specific immune response against the tumor and this property has been used in an attempt to design new therapeutic strategies [ 2 , 3 ]. In particular, anti-tumor vaccination strategies using known tumor-associated antigens or whole tumor extracts have been developed over the last few years [ 2 , 3 ]. However, a growing body of evidence has recently shown that tumors are capable of suppressing anti-cancer immune responses by the induction of tolerance, anergy or by selective killing of immune cells, thereby preventing their destruction by the immune system [ 4 , 5 ]. Dead cells have been also widely used as a source of tumor antigen but contradictory results have been reported on their effect on tumor growth [ 6 ]. We [ 7 ] and others [ 8 , 9 ] have shown that apoptotic bodies are capable of inducing a long-lasting and efficient immune response against tumors whereas others have suggested that necrotic cells could be anergic and tolerogenic [ 10 , 11 ]. Thus, the ability of dead cells to generate an immune response against a tumor could be associated with the nature of the death inducer used and/or the modus operandi of cell death (i.e. necrosis vs. apoptosis). Cytotoxic T lymphocytes (CTLs) and natural killer cells (NK), the major actors of the immune surveillance, have the ability to induce cell death by apoptosis mainly through two mechanisms: the death receptor pathway (i.e. CD95/Fas/APO-1, TRAIL) or the cytotoxic granules (i.e. perforin/granzyme pathway) [ 4 , 12 ]. Activation of death receptors appears to be sufficient to induce the cytosolic activation of caspases, the main proteolytic enzymes of apoptosis, in some tumors (class I) while, in class II tumors, amplification of the death signal occurs through mitochondrial activation of caspases [ 13 ]. Proteins of the BCL-2 family play a major role in the control of apoptosis both in vitro and in vivo in the latter pathway [ 14 ]. These proteins can be divided into anti-apoptotic proteins such as Bcl-2 and pro-apoptotic proteins such as Bax [ 14 ]. Inhibition of apoptosis through the overexpression of Bcl-2 promotes oncogenesis as demonstrated in some follicular B-lymphomas [ 15 ] while, on the other hand, the loss of Bax function has been associated with tumor progression and a bad prognosis in colon and gastric tumors having a microsatellite mutator phenotype [ 16 ]. We have recently observed that the expression of a gain of function variant of Bax can be associated with a longer survival in GBM patients [ 17 ]. Note that the expression of the anti-apoptotic molecule Bcl-2 and that of the pro-apoptotic Bax increased in parallel in low grade to high grade tumors of glial origin, suggesting that Bax and Bcl-2 could play an antagonistic but essential role in these tumors [ 18 ]. One of the most powerful mechanisms of control of tumor growth is exercised by the immune system. The immune surveillance hypothesis suggests that potentially dangerous cells could be eliminated through induction of cell death [ 3 ]. Although the CNS is usually considered an immune privileged site [ 5 ], specific cellular immune responses against tumoral antigens have been achieved in some animal models [ 19 , 20 ]. In GBM, the absence of Bax protein is compensated by an increased expression of Bak, another multidomain pro-apoptotic protein, which also maintains the immune-induced cell death [ 21 ]. Thus, manipulation of the expression of the BCL-2 family members could also be involved in the sensitivity of glial tumors to the immune system. We have tested this hypothesis by establishing cell lines, which stably express transgenes encoding either for human Bax or Bcl-2 in a rat glioma model and analyzed the effects of these transgenes on the in vitro and in vivo growth of these cell lines. Methods Reagents Unless specified, all reagents used in this study were from Sigma (St Quentin-Fallavier, France). Monoclonal anti-human Bax antibody (clone 4F11) was from Immunotech (Villepinte, France) and monoclonal anti-human Bcl-2 antibody (M 0887) was from Dako (Trappes, France); antibodies against rat Bcl-2 or Bax were respectively from Oncogene (Ab5) (Fontenay sous Bois, France) and from Pharmingen (13456E) (Le Pont de Claix, France). The fluorogenic peptide Ac-DEVD-AMC was from Bachem (Voisins les Bretonneux, France) and the lactate dehydrogenase (LDH) activity was measured using the Cytotox 96 ® assay from Promega (Charbonnières, France) as described previously [ 17 , 18 , 21 ]. Experimental research on animals have been conducted according to recommendations of the French National Ethics committee, and are in compliance with the Helsinki Declaration . In vitro transfection, proliferation and induction of apoptosis The rat glioma cell line, A15A5, was obtained from the European Collection of Animal Cell Culture (Valbonne, France). The cell line was maintained in RPMI-1640 (Invitrogen, Cergy-Pontoise, France) supplemented with 10% heat-inactivated FCS (Eurobio, Les Ulis, France), 100 μg/ml streptomycin, 100 U/ml penicillin and 2 mM L-glutamate in a 5% CO 2 air-humidified atmosphere at 37°C. Plasmids encoding for human Bcl-2 or Bax were subcloned into pRcCMV (Invitrogen) as described by the manufacturer. A15A5 cells were transfected with either 2 μg pCMV vector, pCMV Bcl-2 or pCMV Bax. Plasmid DNA was introduced into 10 6 cells by electroporation (GenePulser, BioRad, Yvry sur Seine, France) using 200 V/cm and 250 μF. Transfected cells were selected and cloned in a medium containing neomycin (250 μg/ml) for several weeks before clonal dilution. Apoptosis was induced by a short UV-treatment. Both untreated and UV-treated cells were cultured for a further 24 h under serum-free conditions. Cell death was also induced with FasL (0.5 μg/ml; a gift of Dr P. Saas EPI 119, Besançon, France), doxorubicin (doxo; 2 μM), staurosporine (STS; 1 μM), Na-Butyrate (NaB; 10 mM) or serum deprivation (d-serum) for 3 days. A MTT assay was used according to the manufacturer's instructions (Promega) to determine in vitro cell proliferation of transfected A15A5 cells. Tumor and cell extracts and Western blots A15A5 transfected cells (10 5 cells) or tumors established in rats or mice were homogenized vol./vol. in RIPA buffer (PBS containing 1% NP-40, 0.5% Na-deoxycholate, 0.1% SDS, 10 nM PMSF, 10 nM aprotinin, 1 nM Na-orthovanadate). After several passages in a 2 ml glass Dounce homogenizer, the homogenates were centrifuged at 4°C at 13,000 g for 30 min. The resulting supernatants were assayed for protein concentration using the Bradford technique prior to analysis on 15% SDS-PAGE. Western blots were performed as described earlier [ 21 ], using primary antibodies anti-Bcl-2 (1 μg/ml), anti-Bax (2 μg/ml) and actin (0.5 μg/ml). The antibodies bound to Immobilon-P (Millipore, France) were detected by enhanced chemiluminescence (Amersham, Aylesbury, UK) using a second peroxydase-labelled antibody. The amount of immunoreactive protein was quantified using IP-Lab Gel Program (Signal Analytics, Vienna, USA) after scanning with an Imager (Q-Biogene, Strasbourg, France). Animal experiments Inbred BDIX rats and Swiss nude mice were purchased from Iffa-Credo (L'Abresle, France) and were housed under standard conditions in our laboratory. huBax and huBcl-2 as well as pCMV A15A5 cells were injected into the brain of BDIX rats weighing between 200 and 240 g. All animal procedures were performed with approved protocols and in accordance with published recommendations for the proper use and care of laboratory animals. Briefly the rats were anaesthetised with an intraperitoneal injection of pentobarbital (50 mg/kg) and positioned in a stereotactic head frame. Aseptic surgical techniques were used to open the scalp in the midline and to expose the frontal and temporalis bones. A 1.0 mm aperture for implanting tumor cells was drilled through the skull. The stereotaxic position of this injection site was 2.5 mm anterior to the bregma and 2.0 mm to the right of midline. 10 4 tumor cells were implanted stereotactically at a depth of 3.0 mm into the cerebral parenchyma using a 10 μl Hamilton syringe with a 26-gauge needle. The volume injected was 5 μl and the hole was sealed with sterile bone wax. Alternatively, rats were injected sc with 10 5 cells into the hindlimbs and tumor growth was monitored every week by measuring the volume of the growing tumors. Swiss nude mice were treated similarly except that 10 4 cells were injected subcutaneously. In order to evaluate the implication of huBax A15A5 cells preventive anti-tumoral treatment, we designed a protocol consisting of three sc injections of 3.3 × 10 4 huBax A15A5 cells 15, 10 and 5 days before injection of pCMV or huBcl-2 A15A5 cells. As a control PBS, A15A5 cell oncolysate, or apoptotic bodies derived from Na-Butyrate (NaB)-treated A15A5 cells as previously described [ 7 ]. Briefly, apoptosis was induced in vitro by a 10 mM NaB treatment in subconfluent A15A5 cultures. When signs of apoptosis were observed under a microscope (changes in cell morphology, detachment from dishes, chromatin condensation as viewed with Hoechst 33342) apoptotic bodies were collected, centrifuged and conserved at -80°C prior to use. Typically, 250 μg apoptotic bodies were mixed with 5 mg/ml BCG and injected subcutaneously three times over 15 days. As a control we used oncolysates obtained after several cycles of rapid freezing/ thawing of A15A5 cells and the lysates was injected together with 5 mg/ml BCG to the animals as above. On day 0, 5 days post-treatment, four groups of rats were sc challenged with 10 5 pCMV or huBcl-2 A15A5 cells. Tumor growth was evaluated daily over 60 days. Immunohistochemical analysis of tumor cell injection site Immunochemical analysis was performed as on 12 μm brain frozen sections. Briefly, sections were fixed with 4% paraformaldehyde in PBS for 30 min at room temperature. Endogenous peroxydase activity was inhibited by a treatment with 0,3% H 2 O 2 in methanol for 20 min. The sections were incubated overnight at 4°C with anti-rat CD8 (hybridoma supernatant, Ox 8) diluted 1 in 2 in 1% BSA in PBS, then a secondary antibody coupled to peroxydase was added. The staining was revealed with an AEC substrate. Flow cytometry determination of intratumoral immune population Tumors were resected and minced into 1–2 mm 3 pieces, which were incubated in extraction buffer (30 U/ml hyaluronidase; 500 U/ml DNase, 0.01% w/v collagenase in PBS) at room temperature for 45 min and under constant agitation. The cell suspension was filtered through a sterile grid and washed three times with RPMI and maintained overnight in RPMI before analysis. For the phenotypic analysis, monoclonal antibodies obtained from Pharmingen raised against the following molecules were used: CD3 (556970; cl. 1F4), CD4 (554835, cl ox35), CD8 (554854, cl. Ox 8), CD161 (555006, cl. 10/78), CMH I (22301 D), OX 41/ CD172 (552297) and OX 62 (555010). Monoclonal antibodies raised against granzyme B (GrB, cl. 2C5/F5) was obtained from Chemicon (France) and Fas (AF 126) from R&D Systems (Lille, France). Cells were incubated with the primary antibodies for 30 min at 4°C and washed twice in PBS + 0.1% BSA. For the intracellular detection of GrB, cells were first fixed in 4% paraformaldehyde for 10 min at room temperature, washed with PBS + 0.1% BSA, then permeabilized with 0.1% saponin then incubated with the anti-GrB antibody for 30 min. The secondary antibody was then added for 30 min at 4°C and the cells washed 3 times with PBS + 0.1% BSA. Cells were analyzed on FACScalibur (Becton Dickinson, Le Pont de Claix France) using Cell Quest Pro software. A total of 5000 cells were counted in each experiment. Results Charaterization of rat glioma cells transfected with human Bax or Bcl-2 transgenes A15A5 cells transfected with the pCMV vector, human Bax or Bcl-2 transgenes, were obtained as described in materials and methods. Two different clones were used in each experiment. The expression of the transgenes was monitored by immunoblot analysis using antibodies specific for human Bcl-2 or Bax and, as shown in figure 1A , transfections of the rat glioma cell line with human transgenes were efficiently achieved. Note that the overexpression of Bax did not induce apoptosis in the A15A5 cells, suggesting that the different clones selected expressed sublethal amounts of Bax. To examine the effect of the different transgene expression to cell death, we studied their sensitivity toward different inducers of apoptosis using both drugs such as doxo (20 ng/ml) and STS (20 μM) or treatments such as d-serum, NaB (10 mM) or a short (1 min) UV irradiation. Cell death was monitored and quantified by measuring the activity of caspase 3 (namely the cleavage of the peptide Ac-DEVD-AMC) and the activity of the cytosolic enzyme LDH released into the culture medium (see materials and methods). As shown in figure 1B , when compared to pCMV A15A5 cells, huBax A15A5 cells were more sensitive to cell death in all cases. Conversely and as expected, huBcl-2 A15A5 cells were more resistant to all death inducers. Note that the resistance of pCMV A15A5 cells to apoptosis was already high, suggesting that these cells were naturally resistant to apoptosis. Nevertheless, these results suggest that human Bax and Bcl-2 transgenes were functional in the rat glioma A15A5 cells. The immune system exerts its anti-tumoral surveillance mainly through cell death induced by CTLs and NK [ 12 ]. These cells use different effectors to mediate apoptosis in target cells: the death receptor mechanism such as the FasL/Fas receptor system or the perforin/GrB cytotoxic pathway [ 12 ]. The ligation of the death ligands to their receptors initiates cell death by the activation of the intracellular initiator caspase 8, which in turn can induce apoptosis either through the direct activation of caspase 3 in type I cells or by using mitochondria as an obligatory amplifier of the death signal in type II cells [ 4 ]. Cytolytic granules function through the serine protease GrB, which activates apoptosis mainly through the mitochondrial pathway [ 12 ]. We investigated the in vitro response to externally added FasL or by transient transfection of pCMV GrB, huBax A15A5 and huBcl-2 A15A5 cells as described previously for human glioma cells [ 21 ]. As shown in figure 1C , the expression of Bax sensitized the A15A5 cells to apoptosis induced by both FasL and GrB while apoptosis was inhibited by the presence of Bcl-2 under the same conditions. Incidentally, the fact that huBax or huBcl-2 transfection modulated apoptotic sensitivity toward FasL suggests that the A15A5 cells belong to the type II group. Bcl-2 expression has been associated with retardation in cell cycle entry [ 22 ] and as such we examined the proliferation rate of the different transfected cells using a MTT assay as described in material and methods. All transfected cells appeared to have a similar doubling time, which means that the expression of human Bax or Bcl-2 transgenes did not affect their proliferative capacity in vitro (figure 2A ). Similarly, no significant differences in the effect of the transfection of pCMV, huBcl-2 or huBax A15A5 in A15A5 cells on the clonogenicity of the glioma cells were observed as shown in figure 2B . Tumorigenicity of Bax and Bcl-2 transfected cells To determine the influence of Bax and Bcl-2 on tumoral growth in vivo , 10 4 pCMV, huBcl-2 or huBax A15A5 cells were delivered intracranially (ic) into syngenic BDIX rats as described in materials and methods. The survival rates of the different groups of rats were different as the median survival for the group of rats which had received huBcl-2 A15A5 and pCMV A15A5 cells were respectively 10.2 and 15.6 days, a difference, which was highly significant ( P = 0.0086) (figure 3A ). On the other hand, 80% of rats, which were injected with huBax A15A5 cells, had a disease-free survival of at least 30 days (figure 3A ). It should be noted that 20% of the death, which occurred after injection of huBax, huBcl-2 or pCMV-A15A5 cells appeared to be due to operation-induced traumatisms since sham-operated animals gave similar results ( data not shown ). Brain sections from rats injected with huBcl-2 or A15A5 cells were histologically examined at the time of their death or after 30 days for rats injected with huBax A15A5 cells. As illustrated in figure 3B , brain sections from rats injected with huBax A15A5 cells showed little or no tumoral growth contrary to that observed in rats injected with pCMV or huBcl-2 A15A5 cells. This result suggested that tumoral growth was severely impaired in huBax A15A5 tumors but stimulated in huBcl-2 A15A5 tumors when compared to pCMV A15A5 tumors. To gain information about the kinetics of tumoral growth, cells were subcutaneously (s.c.) injected into the hindlimbs of syngenic BDIX rats and the volume of the tumors evaluated every 10 days. As shown in figure 3C , results similar to that obtained with i.c. experiments were observed as no or little growth was observed with huBax A15A5 tumors whereas huBcl-2 A15A5 tumors exhibited a faster growth than pCMV A15A5 tumors. To test the involvement of the innate immunity in the control of proliferation, the different types of transfected cells (10 5 ) were inoculated s.c. into Swiss nude mice (see materials and methods). Tumor growth in the mice was measured for a period of 45 days (figure 4 ), tumors developed rapidly with similar kinetics for all the A15A5 transfected cells including huBax A15A5 cells. Immunoblot analysis of Bax and Bcl-2 in tumors did not reveal any changes in the expression of Bcl-2 between the established tumors and the huBcl-2 A15A5 cells ( data not shown ). This result showed that the growth of the huBax A15A5 was not due to the induction of rat Bcl-2 expression during tumoral growth in the Swiss nude mice. Accumulation of CD8+ T cells in huBax A15A5 tumors The latter result suggested that immune-induced apoptosis could be involved in the control of tumoral progression of A15A5 cells in BDIX rats, a feature partially lost in athymic mice. To assess local anti-tumoral response, we first investigated the presence of the CD8 marker in the different rat brain sections (figure 3B ). The immunochemical analysis revealed a significant number of infiltrating CD8+ cells in huBax A15A5 tumors whereas huBcl-2 A15A5 or pCMV tumors showed very few CD8+ cells (figure 5A ). Moreover, we noticed that necrotic tissue was absent, thus excluding a non-specific recruitment of lymphocytes due to an inflammatory process. Next, we also quantified the accumulation of CD8+ cells by flow cytometry and examined the phenotypes of the infiltrating immune cells in ic tumors for the presence of intra-tumoral lymphocytes, monocytes and dendritic cells or the loss of the major histocompatibility (MHC) class I molecules. Tumoral cells were dissociated and enzymatically treated as described in materials and methods. We observed a significant increase in double positive cells CD3/CD8 in huBax A15A5 tumors (~17%) compared to huBcl-2 (~4%) and pCMV (~6%) tumors (figure 5B ). An in-depth analysis of these CD3/CD8 cells (figure 5C ) showed that only the lymphocyte T markers and the NK marker CD161 were significantly increased in huBax A15A5 tumors. On the other hand, no differences were found in the expression of class I MHC among the different tumors. The latter result suggested that T-cell mediated immunity could be involved in the rejection of the huBax A15A5 cells in syngenic rats. However, NK cells could play an auxiliary role in this process as suggested by the results obtained in nude mice. huBax A15A5 cells and A15A5 apoptotic bodies confer a protection against pCMV A15A5 cells and to a lesser extent against huBcl-2 A15A5 cells The A15A5 rat glioma cells are highly immunogenic in the syngenic host [ 23 ], this immune response could eradicate tumors prone to apoptosis such as huBax A15A5 cells without affecting the viability of the cells resistant to apoptosis such as the pCMV or the huBcl-2 A15A5 cells. We have previously shown that apoptotic bodies derived from cells obtained from a rat colon carcinoma were a source of anti-tumoral antigens [ 7 , 24 ]. It is conceivable that the enhanced rate of cell death observed in huBax A15A5 tumors in vivo could generate a stronger immune response to pCMV or huBcl-2 A15A5 tumors. To address this question, we compared the efficacy of an anti-tumoral immune response induced by huBax A15A5 cells to that observed with apoptotic bodies or oncolysates obtained from A15A5 cells (see protocol in figure 6A ). Apoptotic bodies were generated from A15A5 cells treated with 10 mM NaB as described earlier [ 7 ] and their molecular characterization will be published elsewhere (Bougras et al. in preparation). BDIX rats were vaccinated with these apoptotic bodies using as a control A15A5 cell oncolysates or PBS (cf. materials and methods). The rats were then challenged either with pCMV A15A5 cells (figure 6B ) or huBcl-2 A15A5 cells (figure 6C ). As shown in figure 6B , the tumoral growth of pCMV A15A5 cells was reduced after the apoptotic body treatment although the effect was limited in amplitude and in time. On the other hand, no effect on tumor growth was observed after a similar treatment with A15A5 oncolysates (compared figure 6B with figure 3A ). Interestingly, the treatment with apoptotic bodies triggered a weaker response to huBcl-2 A15A5 cells, which grew rapidly in animals treated with apoptotic bodies or oncolysate (figure 6C ). As shown in figure 6B , rats that had received huBax A15A5 cells, then challenged with pCMV A15A5 cells developed small palpable tumors, which did not progress after 35 days post-challenging (i.e. pCMV A15A5 cell proliferation was abolished after 35 days by a huBax A15A5 cell pretreatment). Note that huBcl-2 A15A5 cell proliferation was reduced but not abolished in these rats (figure 6C ). These results suggested that rats that had received huBax A15A5 cells were specifically protected against tumor growth and that the overexpression of Bcl-2 could not completely overcome this protection. Discussion Tumor recurrence after surgical resection is often observed in GBM patients and additional chemo-or radiotherapy has not been shown to substantially improve survival in these patients [ 1 ]. In animal models, cancer vaccines have been shown to procure an appropriate immune response, to be highly specific and to favor tumor rejection [ 25 ]. In the case of CNS tumors, intensive research has been performed in the field of immunotherapy since the discovery that the CNS could not be regarded any longer as a completely immunologically privileged site [ 25 ]. Phase I studies have demonstrated the feasibility and the safety of this approach in human gliomas (see for example [ 26 ]). However, although some promising results have been obtained in preclinical studies, so far most clinical attempts have been disappointing. This setback in the application of immunotherapy is, however, not restricted to CNS tumors and new strategies are now being elaborated to enhance the efficiency of this approach. Others and we have observed that apoptotic bodies, the entities derived from apoptotic cells, could be a source of « new » anti-tumoral antigens and as such could be a source of potent tumor vaccines [ 7 - 9 ]. However, the nature of the cell death program, which gives an appropriate anti-tumoral immune response remains controversial [ 6 ]. Apoptosis is thought to be critical for the development and the progression of cancer and it appears to be involved in numerous steps in tumor progression [ 27 ]. Impairment or dysregulation of apoptosis clearly provides a selective advantage to tumoral cells. This resistance could allow the neoplastic cells to evade immunosurveillance as well as environmental changes inherent to tumoral transformation. Indeed, this could explain why, at diagnosis, most tumors have already acquired a certain resistance to apoptosis [ 28 ]. This positive selection for apoptosis-resistant tumor cells is accentuated by current therapies (chemo-and radio-therapies), which also use the apoptotic program to kill cancer cells, and thus tumors resistant to these treatments are often highly resistant to apoptosis [ 29 ]. Thus, all therapeutic approaches should take into account this innate or acquired resistance in the design of new anti-cancer strategies. We have addressed the question of the role of apoptosis in tumor progression by using human Bax or Bcl-2 transfected cells and then analyzing tumoral growth in rats. Several clones of the rat glioma A15A5 cells stably transfected with human Bcl-2 or Bax were used (figure 1A ). In vitro experiments suggest that the expression of the trangenes confer the expected different sensitivities toward apoptosis (figure 1B ). We also show that neither the in vitro proliferation nor clonogenicity of these cells was affected by the expression of Bcl-2 or Bax (figure 2 ). Quite remarkably, the expression of Bax suppresses the growth of these tumors in syngenic rats while that of Bcl-2 seems to stimulate the growth (figure 3 ). The control of tumor growth appeared to be under the control of a specific immune response against tumors. i)Since all tumors proliferated at the same rate in nude mice (figure 4 ). ii) In the rejected tumors (figure 5 ), a specific increase in CD8+ cytotoxic lymphocytes, which have the potential to recognize and attack the major histocompatibility complex (MHC) class I-expressing brain cells including tumoral cells [ 30 , 31 ] were detected. iii) Syngenic rats that have received huBax A15A5 cells develop an anti-tumoral reaction against A15A5 cells (figure 6 ). However, this response was partially occluded by the presence of Bcl-2, a result consistent with the fact that its expression rendered the cells more resistant to apoptosis including that induced by the immune system (figure 1C ). However, the accumulation of CD8+ CTL at the site of injection of huBax A15A5 cells could also suggest that induction of cell death in tumors facilitated or triggered a greater specific response. This could explain why an immune response to tumors was specifically observed in animals treated with huBax A15A5 cells (figure 6 ). On the other hand, the transfection with huBax could modify the phenotype of the rat cells as suggested by previous results [ 32 ]. However, MicroArray analysis of huBax A15A5 versus huBcl-2 A15A5 cells did not reveal any changes in the transcriptome of the cells (Cartron and Jézéquel, unpublished observation). Of note, in human glioblastomas expressing Bax ψ, a highly apoptogenic variant of Bax α [ 17 ], we observed an accumulation of intra tumoral CD8+ cells when compared to Bax α tumors (Bougras et al., unpublished results). The apoptosis index has been found to be of significant prognostic significance in patients with high-grade astrocytomas [ 33 ]. Our study provides three new findings, which should be considered in immunotherapy: i) the expression of pro-or anti-apoptotic molecules can control the response of the tumor to the immune system during the course of tumor progression (at least in class II). Thus, we suggest that the over-expression of Bcl-2, which often occurs during tumorigenesis could account for the escape from the immune surveillance. ii) Anti-apoptotic mechanisms that often impede the success of treatments could also be an obstacle to immunotherapy. In addition other anti-apoptotic processes such as the existence of soluble decoy receptor that impairs FasL induced-apoptosis in malignant gliomas [ 34 ] or the existence of an inactive granzyme [ 35 ] could also be involved in the resistance to the immune system. iii) The type of cell death (e.g. apoptosis vs. necrosis) that induces the best stimulation of the immune system is still a matter of controversy but our results suggest that cells sensitized to apoptosis are capable of providing a long lasting and efficient protection against tumoral growth. Interestingly, a rat model of colon carcinoma has been described in which some clones gave rise to tumors that constantly expanded in the animal to eventually formed metastasis (Pro) while others progress for several days before complete regression (Reg) [ 36 ]. The disappearance of the latter clone has been shown to be controlled and to be triggered by the immune system and the transfection of Reg cells by Bcl-2 has been shown to prevent apoptosis and to restore its tumorigenicity [ 36 ]. However, the effect of the ectopic expression of Bax in the progressive counterpart cell line Pro was not investigated in this work [ 36 ]. Our results show thus for the first time, in the same type of cells, the adverse effects of Bax and Bcl-2 in the antitumoral role of the immune system. Conclusion Taken together, our data provide evidence that BCL-2 family members control tumor growth through their sensitivity to immune-induced cell death and enhancement of the immunogenicity of tumor cells. Competing interests The authors declared that they have no competing interests. Abbreviations Bax: Baxα ; CNS: central nervous system; CTL: cytotoxic T lymphocytes; GBM: Glioblastoma Multiforme; ic : intra-cerebral; sc : subcutaneous. Author's contributions Gwenola Bougras carried out the phenotyping characterization of the cell lines and in vivo experiments with tumors, Pierre Francois Cartron the molecular genetic studies, Fabien Gautier carried out the experimental studies with apoptotic bodies and Stephane Martin the intracerebral implantation of the rat glioma cell lines. Marité LeCabellec participated in immunohistochemical characterization of the tumors. Marc Grégoire and Khaled Meflah participated in the coordination of the study. Francois M. Vallette conceived of the study, and participated in its design and coordination and drafted the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC515305.xml
517506	Improving prescribing of antihypertensive and cholesterol-lowering drugs: a method for identifying and addressing barriers to change	Background We describe a simple approach we used to identify barriers and tailor an intervention to improve pharmacological management of hypertension and hypercholesterolaemia. We also report the results of a post hoc exercise and survey we carried out to evaluate our approach for identifying barriers and tailoring interventions. Methods We used structured reflection, searched for other relevant trials, surveyed general practitioners and talked with physicians during pilot testing of the intervention. The post hoc exercise was carried out as focus groups of international researchers in the field of quality improvement in health care. The post hoc survey was done by telephone interviews with physicians allocated to the experimental group of a randomised trial of our multifaceted intervention. Results A wide range of barriers was identified and several interventions were suggested through structured reflection. The survey led to some adjustments. Studying other trials and pilot testing did not lead to changes in the design of the intervention. Neither the post hoc focus groups nor the post hoc survey revealed important barriers or interventions that we had not considered or included in our tailored intervention. Conclusions A simple approach to identifying barriers to change appears to have been adequate and efficient. However, we do not know for certain what we would have gained by using more comprehensive methods and we do not know whether the resulting intervention would have been more effective if we had used other methods. The effectiveness of our multifaceted intervention is under evaluation in a randomised controlled trial.	Background Much research has been carried out with the aim of influencing the performance of clinicians. The results have varied [ 1 , 2 ]. As with any human behaviour, clinical practice is difficult to change. Some strategies that have been evaluated, like passive dissemination of clinical practice guidelines, have had little or no effect on practice [ 3 ]. Others, like educational outreach visits ("academic detailing") and multifaceted interventions, may be more effective than passive interventions [ 1 ]. The reasons why clinical practice sometimes is not consistent with current best evidence varies across clinical problems and from one clinician to another. A logical consequence of this is to tailor quality improvement strategies to address specific barriers [ 4 ]. Several trials of tailored interventions have been conducted. The methods used for identifying barriers to change have varied and there is limited evidence of the relative usefulness of different approaches. However, the choice of method for identifying barriers has implications, particularly with regards to resources, since some methods are time consuming and demand the involvement of many individuals. This represents a practical and financial constraint. On the other hand, if such approaches lead to the identification of important barriers that otherwise would have been overlooked, they may be worth the effort. In this article we describe a simple approach we have used to identify barriers to changing professional practice. This was done as the first step in a process of developing an intervention to improve the pharmacological management of hypertension and hypercholesterolaemia [ 5 ]. The intervention focused on three specific recommendations in clinical practice guidelines for hypertension and hypercholesterolaemia [ 6 - 8 ] based on evidence of a gap between the recommendations and current practice in Norway: • Contrary to recommendations, physicians seem to rarely estimate the risk of cardiovascular disease before initiating treatment [ 9 ] • Sales of thiazides are low, despite these drugs being recommended as first-line medication [ 10 ] • Relatively few patients reach recommended treatment goals [ 11 , 12 ] We also report the results of a post hoc exercise and a survey we carried out to evaluate our approach to identifying barriers and interventions. Methods We developed the intervention through a process of identifying barriers to implementation of recommendations and measures specifically addressing these barriers ("tailoring"). The methods we used were structured reflection, searching for other relevant trials targeted at improving the management of hypertension or hypercholesterolaemia, conducting a survey among general practitioners and discussion with physicians during pilot testing of the intervention. Structured reflection The three authors reflected over possible barriers based in part on our own experience as physicians working in primary care in Norway. We used a worksheet to structure our reflection (see Additional file 1). The worksheet included factors that might act as barriers in the practice environment, the professional environment, and related to physicians' knowledge, skills and attitudes. One worksheet was completed for each targeted behaviour: increasing the use of cardiovascular risk assessment before initiating treatment for hypertension or hypercholesterolaemia, increasing the prescribing of thiazides for the treatment of uncomplicated hypertension, and increasing the proportion of patients on medication for hypertension and hypercholesterolaemia that reach recommended treatment goals. The worksheet was used to facilitate our group discussion of possible interventions to address the identified barriers. Our research group had recently completed a trial of a strategy for guidelines implementation when we were planning this study [ 13 ]. In that study the multifaceted intervention consisted of several passive components. Information and materials were distributed by mail and to a large degree we relied on the physicians themselves to make an effort at changing their practice. The observed changes in practice were small. In another trial we had found that the use of active sick leave for back patients was significantly increased through a proactive intervention compared to a passive one [ 14 ]. Based on these experiences our research group decided to test an active strategy in this study. Therefore we decided to use outreach visits ("academic detailing") prior to considering specific barriers. We considered systematic reviews of interventions to improve professional practice when we designed our strategy [ 1 ]. We searched the Cochrane Group of Effective Care and Organisation of Care trial register for trials of interventions targeted specifically at the management of hypertension or elevated cholesterol in general practice. Questionnaire to physicians We surveyed general practitioners about some of the interventions about which we were uncertain after our structured reflection. The details of the survey have been described elsewhere [ 9 ]. Briefly, 265 physicians who had participated in an earlier trial conducted by our research group [ 13 ] were asked to complete a questionnaire as part of the study-evaluation. We used that opportunity to seek answers to the following questions: 1. Do physicians assess cardiovascular risk before prescribing antihypertensive or cholesterol-lowering drugs? 2. If not, would physicians be more likely to do so it they received a fee for this? 3. Do physicians comply with current regulations limiting the reimbursement of cholesterol-lowering drugs? The last question was asked for two reasons. Firstly, we were considering making risk assessment a condition for reimbursement of the drugs. Secondly, the existing regulations were a possible barrier to adhering to our recommendations because they conflicted with these. Pilot testing During pilot testing of the intervention at two practices, which were selected for convenience, comments from physicians relevant to possible barriers were noted. We also informally evaluated each component of the intervention. Post hoc focus groups and structured reflection exercise After we had finished designing the intervention we had the opportunity of testing our method of structured reflection at a gathering of international researchers in the Research Based Education and Quality Improvement group (ReBEQI) , December 2003. Each participant was asked to complete a worksheet to identify barriers and possible interventions related to the low use of thiazides among general practitioners. They were randomly allocated to four different groups where they collaborated on completing the worksheet. They were also asked to grade the importance of each barrier or intervention as minor, moderate or major. We disregarded those rated as minor. We compared the results from the four groups with the barriers and interventions we had identified. Post hoc survey of physicians exposed to the intervention While conducting the randomised trial to test the effectiveness of our multifaceted intervention we carried out telephone interviews with physicians allocated to the experimental group. They were asked if they adhered to our recommendations and, if not, why. The responses where noted down during the interviews. Results Barriers and interventions Figure 1 illustrates the timeframe for the methods used to identify barriers and interventions. Tables 1 , 2 , 3 give an overview of the barriers and interventions that we identified for each clinical problem. Many of the barriers were related to a lack of knowledge and could be addressed through educational interventions. The use of educational outreach visits was logical since we had planned to use an active intervention, based on our previous experience, and since this type of intervention has consistently lead to improved professional behaviour in randomised trials [ 1 ]. Similarly, based on previous experience and the capabilities of the software we hade developed [ 13 ], we planned on using an electronic risk calculator, electronic prescriptions, patient information materials, and computerised reminders. The search (July 2001) of the EPOC trial register for randomised trials with the word "hypertension" in any field yielded 58 references. Most were excluded after reading the abstracts, leaving eight, for which the full text was reviewed [ 15 - 22 ]. This did not lead to any changes in our intervention strategy. A search for randomised trials with the word "cholesterol" yielded 13 references. The full text was reviewed for only one of these [ 23 ]. This also provided little further guidance for designing our intervention. The nine trials that we reviewed are summarised in table 4 . The survey results did not indicate that a fee for estimating cardiovascular risk before initiating drug therapy would affect practice [ 9 ]. The survey results also indicated that physicians are largely not affected by conditions for drug reimbursement [ 9 ]. Moreover, there were no mechanisms in place to enforce such regulations. We did not identify additional barriers during pilot testing of the intervention with five physicians in two practices, but several of those already identified were confirmed, particularly barriers to prescribing thiazides. Based on our findings and an assessment of the feasibility and evidence of effectiveness for various interventions, we designed a multifaceted intervention. The elements of the intervention are described in table 5 . We also considered a number of interventions that we excluded. For example: • We considered placing computers in waiting rooms so that patients could assess their cardiovascular risk before seeing the physician, but concluded this would be costly and difficult to implement. • We considered providing pre-printed prescriptions, but found this would not to be relevant because most physicians use computerised systems for prescribing. • We considered exposing conflicts of interest among clinical specialists who advocated using other first line drugs than thiazides, but elected not to do so. • We considered exposing techniques used in pharmaceutical advertisements, such as using relative risk reductions rather than absolute risk reductions [ 24 ], but concluded this would have at best a limited impact. Post hoc focus groups and structured reflection exercise Nineteen researchers were divided into four groups. All groups considered advocacy by drug companies to be a major barrier to change. Routines or habits were also included as an important barrier by all the groups, as well as lack of knowledge concerning the effectiveness of thiazides, their favourable adverse effects profile, and their low cost. All the groups also mentioned competing guidelines or diverging opinions as part of the problem. Three of the groups considered local or national opinion leaders as potential barriers to change. Patients' expectations or perceived expectations were also mentioned by three of the groups. The interventions recommended by the groups to address the identified barriers are presented in table 6 . All the groups suggested the use of computerised reminders to address physicians' lack of knowledge or their habits and routines. All the groups also suggested some form of interactive education, mainly as a counter force to promotional activities by the pharmaceutical industry, and patient information was suggested by three of the groups. Two suggested training physicians to address patient expectations. Two groups suggested developing clinical guidelines and two suggested audit and feedback, but one group considered this to be of minor importance. Post hoc survey of physicians exposed to the intervention Among the 195 physicians exposed to the intervention, 149 (76%) were contacted during the trial period and agreed to answer our questions. No major additional barriers were identified. However, some physicians questioned whether adhering to the recommendations would represent a good use of resources, specifically the recommended treatment goals. Discussion Addressing barriers to change with tailored interventions makes sense and there is some empirical support for this [ 1 ]. It is unclear, however, what methods are the most useful for identifying barriers and interventions. Several qualitative methods can be used to identify barriers, such as interviews, focus groups and observation. These methods may be valuable, but they are relatively labour-intensive. We used a simpler approach to identifying barriers to change. Would the use of other methods have provided us with important additional information? Pilot testing and discussions with five physicians in two practices and interviews with 140 participating physicians did not indicate additional barriers. The post-hoc focus groups with international experts did not add much with regards to barriers and interventions. Several of these groups included "routines and habits" as a potential barrier, which was not explicitly mentioned among the barriers identified by the investigators. However, all interventions that were mentioned by more than one of the groups in the post-hoc focus group exercise were included in our multifaceted intervention. Our use of computerised reminders was based on the assumption that this would help to establish new routines, although we did not record routines and habits as a barrier when we developed the intervention. There are inherent weaknesses in our approach. One is that the investigators undertaking the structured reflection were few and we were prejudiced by our own experiences. The lack of patient involvement is another limitation, which possibly lead to an under-emphasis of patient-mediated interventions. A weakness with the group of international researchers who participated in the post-hoc focus groups is their lack of familiarity with the Norwegian context. A number of trials of tailored interventions have been conducted. The methods used to identify barriers to change have varied. Some investigators have simply hosted a meeting [ 25 , 26 ], others have used questionnaires [ 23 ], conducted focus-groups [ 27 - 30 ] or interviews [ 31 - 33 ], or both [ 34 ]. Others have used a combination of several qualitative methods [ 35 - 37 ]. Some investigators have used identification of barriers as an intervention in itself [ 19 , 38 , 39 ]. The methods that were used have been poorly described in most of these studies. Conclusions Our simple approach to identifying barriers to improving practice appears to have been effective in identifying all of the important barriers, and it was efficient. However, we do not know for certain what barriers other methods would have identified or whether the intervention could have been more effective, if we had used other methods. Further work to address these questions is planned, including direct comparison of alternative methods and evaluations of theory-based approaches . The effectiveness of our multifaceted intervention is under evaluation in a randomised controlled trial. Competing interests None declared. Author contributions All the authors participated in the process of structured reflection and in conducting the survey of physicians. AF was responsible for reviewing the results from previous research and pilot testing of the intervention. AF drafted the article while SF and ADO contributed to critical revisions of the manuscript. All authors read and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517506.xml
517499	Antroduodenal motility in neurologically handicapped children with feeding intolerance	Background Dysphagia and feeding intolerance are common in neurologically handicapped children. The aim is to determine the etiologies of feeding intolerance in neurologically handicapped children who are intolerant of tube feedings. Methods Eighteen neurologically handicapped children, followed in the Tube Feeding Clinic at the Children's Hospital of Wisconsin who were intolerant of gastrostomy feedings. The charts of these 18 patients were reviewed. Past medical history, diagnoses, history of fundoplication and results of various tests of gastrointestinal function including barium contrast radiography, endoscopy and antroduodenal manometry were documented. Results Five of 11 children had abnormal barium upper gastrointestinal series. Seven of 14 had abnormal liquid phase gastric emptying tests. Two of 16 had esophagitis on endoscopy. All 18 children had abnormal antroduodenal motility. Conclusions In neurologically handicapped children foregut dysmotility may be more common than is generally recognized and can explain many of the upper gastrointestinal symptoms in neurologically handicapped children.	Backround Oral pharyngeal dysphagia due to disordered swallowing has become increasingly recognized in children with cerebral palsy and other neurodevelopmental disorders. This has led to the increasing use of enteral tube feedings either for full or supplemental nutritional support. Symptoms of foregut dysmotility, such as vomiting, retching gagging and bloating, are often associated with tube feeding in neurologically handicapped children [ 1 - 4 ]. Previous studies have demonstrated that recurrent vomiting, aspiration and/or failure to thrive may be present in as many as 10–15% of institutionalized patients with psychomotor retardation. Antroduodenal motor function has been little studied in such children [ 5 , 6 ]. In order to elucidate the mechanisms behind these symptoms we reviewed the charts of a group of children followed in the Tube Feeding Clinic at the Children's Hospital of Wisconsin with neurological dysfunction, who were intolerant of tube feedings and who had undergone antroduodenal motility studies as part of their evaluations. Methods The charts of 18 neurologically handicapped children (mean age 4 years, range 1–10 years; 10 males) with dysphagia and symptoms of foregut motility were reviewed. The symptoms, underlying disorders, feeding route, the presence of fundoplication, and the use of prokinetic agents and H 2 b1receptor antagonists are summarized in Table 1 . All except one patient were completely or partially fed enterally. One patient ate orally but required frequent venting of his gastrostomy. Table 1 Patient population Patient Number Age (y) Diagnosis Symptoms Feeding Route UGI Gastric Emptying EGD PEG Fundoplication Other Medications H2RA Cisapride 1 9 CP GER vomiting retching oral paraesophageal hernia normal normal + gastric bezoar 2 4 feeding aversion vomiting retching gastrostomy esophageal dysmotility ND esophagitis + + + 3 3 Down's syndrome gagging retching gastrostomy normal delayed normal + 4 2 cerebral dysgenesis seizures Irritability jejunostomy GER delayed esophagitis + omeprazole 5 5 chromosome 19 deletion, GER subglotic stenosis retching gastrostomy normal normal ND + 6 4 hydrocephalus retching gastrostomy normal normal normal + + + 7 1 CP irritability gastrostomy paraesophageal hernia normal normal + + + 8 2 1/2 cerebral atrophy recurrent aspiration pneumonia vomiting gastrostomy GER normal normal + + + 9 2 diphragmatic hernia GER vomiting jejunostomy GER delayed ND + surgical jejunostomy + + 10 5 CP spina bifida retching bloating constipation jejunostomy normal normal normal + surgical jejunostomy + + 11 9 Floating Harbour syndrome retained food vomiting oral ND rapid retained food + + + + 12 17 mitochondrial disease vomiting diarrhea retching gagging bloating gastrostomy esophageal dysmotility delayed normal + + + 13 3 CP GER retching gagging gastrostomy ND normal normal + + 14 6 charge syndrome chromosome 13 Deletion retching gagging gastrostomy normal normal esophagitis + + + 15 2 CP hepatoblasoma vomiting gastrostomy paraesophageal hernia delayed normal + Nissen breakdown liver resection pyloroplasty + + 16 1 CP vomiting gastrostomy GER delayed esophagitis surgical gastrostomy - pyloroplasty + + 17 2 feeding aversion vomiting gastrostomy normal + normal + + + 18 10 CP seizures hydrocephalus vomiting retching gagging bloating TPN + + + + + GER: gastroesophageal reflux CP: cerebral palsy: NE: not done TPN: total parenteral nutiriton Following an overnight fast, antroduodenal motility studies were performed using a multilumen catheter with 8 recording ports spaced 2.5–5 cm apart, passed through the gastrostomy either under fluoroscopic guidance or endoscopically, connected to a low compliance, pneumohydraulic capillary infusion system (Arndorfer Medical Specialties, Greendale, WI) and a computerized motility system (Redtech, Calabasas, CA). Fasting activity was recorded for 3–4 hours. Erythromycin (1 mg/kg) was given intravenously over 10 minutes and the recording continued for another hour. Octreotide (0.5 mcg/kg) was then given intravenously over 5 minutes. 45 minutes later a liquid meal was given followed by an additional 2–3 hour recording period. The meal varied and consisted of the usual formula and volume given at home. Patients receiving jejunal feedings or TPN were given a bolus gastrostomy feeding. In the one patient receiving TPN, the TPN was discontinued during the study. Prokinetics were stopped at least 48 hours prior to study. Phase 1 of the MMC was defined as motor quiescence. Phase 2 was defined as the time between Phases 1 and 3 and is characterized by random contractions of varied amplitude and frequency. Phase 3 of the MMC is characterized by an aborally propagating cluster of repetitive contractions with a frequency of 11–13/minute in the duodenum and 3/minute in the antrum with a duration of 3–10 minutes. The tracings were analyzed by visual inspection. This study was approved by the Research and Publications Committee/Human Rights Review Board of the Children's Hospital of Wisconsin and the Institutional Review Board of the Medical College of Wisconsin. Results Eleven children had had recent upper GI series. Of these 5 were normal, 3 had gastroesophageal reflux and 3 had paraesophageal hernias (all following fundoplications), 1 had a bezoar and 1 had esophageal dysmotility. Fourteen patients had liquid phase gastric emptying studies. Of these 7 were normal, 6 had delayed emptying and 1 had rapid emptying. Two of 16 patients who had had recent endoscopies had esophagitis, 14 were normal. The diagnoses and clinical histories of the patients are summarized in Table 1 . Twelve of the 18 patients had had a fundoplication and 9 of the 12 had had a pyloroplasty. Indications for fundoplication were frequently poorly described in the medical records but typically included vomiting and feeding intolerance. The incidence of symptoms such as retching and sweating could not be determined No patients had a normal antroduodenal motility study (Figures 1 , 2 , 3 ). In the fasting state 12/18 failed to have phase 3 of the MMC. Eight had predominantly phase 2 activity and did not demonstrate normal fasting phase 1. Nine had non-propagating clusters. One had MMCs that propagated in a retrograde fashion. Figure 1 An 8-year-old boy who was TPN dependent demonstrates reverse peristalsis. Note lack of antral contractions. Channel 1–2 antrum; 3–6 duodenum. Figure 2 Lack of fasting phase 3 activity during a 3 hour monitoring period in an 8-year-old boy. He also had no phase 3 like activity following erythromycin. Channel 1 antrum; 2–6 duodenum. Figure 3 Lack of responsiveness to erythromycin in a 3-year-old boy with feeding aversion. This patient also had postprandial hypomotility. Channels 1 stomach; 2–3 antrum; 4–7 duodenum. Following erythromycin eight had a normal response consisting of antral contractions with a frequency of 3/min followed by phase 3 like activity in the duodenum. Three patients had no response and seen had abnormal responses consisting of abnormal clusters in three, no antral response in three and no duodenal response in three. Fourteen patients had a normal response to octreotide consisting of cessation of antral activity and the development of phase 3 activity in the duodenum. Two patients had continued antral contractions and two did not develop phase 3 duodenal activity. These patients had non-propagating duodenal clusters. Fourteen patients had normal postprandial phase 2-like activity. Eight developed premature phase 3 activity within 30 minutes following the meal. Seven patients failed to develop phase 3 activity during the 2–3 hour postprandial monitoring period and seven had no antral contractions in the postprandial monitoring period. One patient had a retrograde MMC. Three patients developed severe pain or irritability associated with antral or duodenal contractions following erythromycin [ 2 ] or octreotide [ 1 ]. There was no correlation between any constellation of symptoms and manometric abnormalities. Conclusions Foregut dysmotility is common in children with neurodevelopmental disorders such as cerebral palsy [ 1 - 4 ]. Up to 75% of institutionalized children with psychomotor retardation have GER [ 7 - 11 ]. A number of investigators have reported that neurologically handicapped children have abnormalities of lower esophageal function [ 12 , 13 ]. Delayed gastric emptying is common in such patients [ 1 , 2 ]. Many of these patients undergo fundoplication. Continued symptoms of GER following fundoplication or the development of new symptoms such as retching and gagging suggests that a more generalized foregut motility disorder is present in many of these patients [ 5 ]. The rates of complications of surgical treatment of GER that might relate to foregut dysmotility include breakdown of the wrap (0.9–13%) and the gas bloat syndrome (1.9–8%) [ 14 ]. Other complications not reported in enough detail to estimate complication rates include dumping, and gastroparesis. Ravelli and Milla have shown that gastric electrical activity as measured by the electrogastrogram (EGG) was abnormal in 31/50 neurologically handicapped children [ 2 ]. Eleven of 18 patients who were symptomatic after fundoplication had gastric dysrhythmias. Richards et al showed that neurologically impaired children with pallor, sweating, retching or forceful vomiting preoperatively were at high risk for postoperative retching and vomiting. They hypothesized that these symptoms were indicative of activation of the emetic reflex and that children with these symptoms had a more generalized disorder than those children without such symptoms [ 15 ]. In a previous study we compared gastric electrical activity as measured by EGG in a group of neurologically handicapped children who were tolerant of their tube feedings to a group that were intolerant or symptomatic during tube feedings [ 16 ]. The percentage of children in each group who had undergone fundoplication was the same. We found that although the percentage of time that normogastria, bradygastria and tachygastria were present was not different in the 2 groups, there was a significant difference in the postprandial power between the groups. This finding suggests that symptoms present in these patients such as vomiting, retching and gagging might be due to an underlying foregut motor disorder. There have been few reports of antroduodenal motility, which have focused on neurologically handicapped children with feeding intolerance. DiLorenzo and colleagues reported that 25/28 children who remained symptomatic following fundoplication had abnormal antroduodenal motility [ 5 ]. Similar to our patients a wide variety of abnormalities were found. These authors did not report how many of their patients that had neurological handicaps. Miki et al found that fasting antroduodenal motility was abnormal in 11 neurologically impaired children with symptoms of gastroesophageal reflux [ 6 ]. Due to the nature of our patients there are some limitations in the design and interpretation of our study. Because many of our patients had had multiple formula changes, we decided to use the formula, which the child was receiving at the time of the study, so that any symptoms occurring during the studies could not be attributed to a formula change. Bolus feeds were given to all patients during the study even those who had been receiving drip feeds. The fasting state could not be recorded for 4 hours in all patients, thus it is possible that absence of phase 3 of the MMC in these patients might not be abnormal. Since normal manometry data have not been published for children, non propagating clusters may or may not be normal in our patients. Since only 12 of our 18 patients had undergone fundoplication, we agree with DiLorenzo et al [ 5 ] that these motor abnormalities were not caused by surgery, rather we believe that the underlying motility disorder was more generalized than had been recognized at the time of fundoplication. In this study we have confirmed that the incidence of foregut dysmotility is very high in neurologically handicapped children with feeding intolerance. Prokinetics and acid suppression did not resolve the symptoms in our patients. Twelve of our 18 patients had had fundoplications and two had undergone two fundoplications in unsuccessful attempts to control what had been thought to be reflux symptoms. While there is no way to know how much abnormal antroduodenal motility contributed to our patient's feeding disorders, following antroduodenal manometry a number of our patients were treated successfully with jejunal feeding, suggesting that while they had foregut dysmotility, midgut motility is normal. In neurologically handicapped children foregut dysmotility may be more common than is generally recognized and can explain many of the upper gastrointestinal symptoms in neurologically handicapped children. Thus in this patient population generalized foregut dysmotility may mimic reflux and the decision to perform a fundoplication should be made very cautiously and only after a complete evaluation of foregut motility particularly in children with gagging retching and forceful vomiting. Competing Interests None declared. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC517499.xml
546013	Obsessive-compulsive disorder and trichotillomania: a phenomenological comparison	Background Similarities between obsessive-compulsive disorder (OCD) and trichotillomania (TTM) have been widely recognized. Nevertheless, there is evidence of important differences between these two disorders. Some authors have conceptualized the disorders as lying on an OCD spectrum of conditions. Methods Two hundred and seventy eight OCD patients (n = 278: 148 male; 130 female) and 54 TTM patients (n = 54; 5 male; 49 female) of all ages were interviewed. Female patients were compared on select demographic and clinical variables, including comorbid axis I and II disorders, and temperament/character profiles. Results OCD patients reported significantly more lifetime disability, but fewer TTM patients reported response to treatment. OCD patients reported higher comorbidity, more harm avoidance and less novelty seeking, more maladaptive beliefs, and more sexual abuse. OCD and TTM symptoms were equally likely to worsen during menstruation, but OCD onset or worsening was more likely associated with pregnancy/puerperium. Conclusions These findings support previous work demonstrating significant differences between OCD and TTM. The classification of TTM as an impulse control disorder is also problematic, and TTM may have more in common with conditions characterized by stereotypical self-injurious symptoms, such as skin-picking. Differences between OCD and TTM may reflect differences in underlying psychobiology, and may necessitate contrasting treatment approaches.	Background Trichotillomania (TTM) is characterized by repetitive stereotypical hair-pulling from different sites resulting in noticeable hair loss [ 1 ]. Phenomenological observations have suggested that symptoms of repetitive hair-pulling are reminiscent of the compulsions seen in obsessive-compulsive disorder (OCD) [ 2 , 3 ]. For example, both TTM and OCD patients describe compulsive urges and ritualistic behaviours [ 2 , 4 ]. Comorbidity data also suggest some overlap between TTM and OCD [ 2 ]. Thus, a number of authors have suggested that TTM might be classified with OCD in a spectrum of disorders having similar phenomenology [ 4 - 8 ]. However, in addition to overlapping phenomenology between OCD and TTM, there are also significant differences. For example, in contrast to compulsions in OCD, hair-pulling in TTM is not in response to obsessive thoughts (such as worry about harm to self or others) but rather because of an irresistible urge and the promise of gratification when pulling out hair [ 2 , 6 ]. Also, unlike patients with OCD whose symptoms change over time in terms of focus and severity (e.g. from washing of hands to checking locks, stoves, appliances, etc) [ 9 ], TTM patients usually only present with hair-pulling without evolution to non-self-injurious compulsive rituals. Examination of demographic variables in OCD and TTM supports the argument that these are two distinctive disorders. TTM is much more prevalent in females (10:1 female to male ratio) [ 10 ] whereas OCD is equally common in males and females [ 11 ]. Age of onset also differs somewhat: TTM typically presents in early adolescence, with the mean age of onset of hair-pulling in males later than that in females [ 10 , 12 , 13 ] whereas OCD has its onset from childhood through to early adulthood [ 14 ], but with males reporting an earlier onset compared to females [ 15 ]. Additional clinical observations further support a distinction between OCD and TTM. Patients with TTM tend to have fewer comorbid obsessive-compulsive symptoms, as well as less depression and anxiety compared to OCD patients [ 16 ]. Response prevention in OCD patients eventually leads to anxiety reduction, whereas in people with TTM it may lead to an increase in anxiety [ 17 ]. Although a selective response to serotonergic reuptake inhibitors (SRI's) has been suggested to characterize both OCD and TTM, there is good evidence that response to SRI's is sustained in OCD, whereas the evidence-base for the efficacy of these agents in TTM is much more mixed. Relatively few empirical studies have, however, documented the phenomenological similarities and differences between OCD and TTM [ 3 , 18 , 19 ]. A large clinical database comprised of patients with OCD and TTM provided us an opportunity to investigate the relationship between these conditions in terms of demographic and clinical variables. Methods Subjects Two hundred and seventy eight OCD patients (n = 278: 148 male; 130 female), and 54 TTM patients (n = 54; 5 male; 49 female), ranging in age between 8 and 75 years, took part in the study (Table 1 ). These patients were referred to our research unit from a wide range of sources (including the OCD Association of South Africa, community based primary care practitioners, and psychiatrists). Either a clinical psychologist or a psychiatrist with expertise in the field interviewed participants. Participants met the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) criteria [ 1 ] for either a primary diagnosis of OCD or TTM on the Structured Clinical Interview for Axis I Disorders (SCID-I) [ 20 ]. Patients were included irrespective of whether they were at baseline (i.e. not receiving any form of treatment for their primary psychiatric disorder), or were receiving treatment for OCD / TTM, but those with comorbid OCD and TTM (N = 25) were excluded from subsequent analysis. A history of psychosis was also an exclusion criterion. Referring clinicians were contacted to establish, where possible, a longitudinal expert evaluation of the diagnostic status of the patient. All subjects gave informed written consent to participate after confidentiality was guaranteed and risks and benefits had been fully explained. The study was approved by the Institutional Review Board of the University of Stellenbosch. Table 1 Demographic information: OCD and TTM Variables OCD (N = 278) TTM (N = 54) χ 2 P Gender 148 male 130 female 5 male 49 female 40.7 <.001 Age (SD) 33.1 (14.4) 31.3 (12.5) NS Population group 86.6% Caucasian 79.6% Caucasian NS Level of education 50.7% completed high school or higher level education 44.4% completed high school or higher level education NS Employment 6.8% unemployed 3.7% unemployed NS NS = non-significant Interview Specific demographic data, including age when interviewed, age of onset of OCD/TTM, highest level of education, current employment status, and population group were obtained from all participants. In addition to the SCID-I, and selected parts of the SCID-II (obsessive-compulsive, avoidant, schizotypal, borderline personality disorders) for adult patients (aged 18 or older) [ 20 ], the interview also included the Structured Clinical Interview for Obsessive-Compulsive Spectrum Disorders (SCID-OCSD) to determine the presence of other obsessive-compulsive related conditions [ 21 ]. The Yale-Brown Obsessive-Compulsive Severity Scale (Y-BOCS) [ 22 ] was implemented to assess the severity of OCD symptoms. Severity of hair-pulling symptoms was assessed with the Massachusetts General Hospital Hair-pulling Scale [ 23 ]. The Trichotillomania Behaviour Profile (TBP, available from the first author on request) was administered to TTM patients to assess hair-pulling phenomenology. Patients' level of insight into the senselessness or excessiveness of their symptoms was assessed on the relevant YBOCS item. When an adequate trial of pharmacotherapy with an SRI (i.e. for both OCD and TTM groups, at least 10 weeks on the medication with a minimum of 6 weeks on mid-range dose) had been undertaken, response to pharmacotherapy was assessed using the global improvement item of the Clinical Global Impression (CGI) scale; subjects with CGI scores of 1 ('very much improved') or 2 ('much improved') were defined as responders [ 24 ]. Similarly, when patients received an adequate trial of cognitive behavioural therapy (CBT) (i.e. for both OCD and TTM groups, 8 or more sessions with an expert CBT psychotherapist), response to treatment was rated using the CGI. The Disability Profile questionnaire (DP) [ 25 ] was included in the interview to assess current (i.e. past two weeks) and lifetime impairment in eight domains. The DP was initially developed for use in patients with social anxiety disorder; nevertheless, the scale has since been used to assess disability in patients with other anxiety disorders as well [ 26 ]. Questions addressing potential precipitating or exacerbating factors, including the impact of menstrual/reproductive cycle changes, brain trauma and history of autoimmune infections on OCD/TTM symptom fluctuations, were included in the interview. Self-report questionnaires Severity of comorbid depression was evaluated with the Beck Depression Inventory (BDI) [ 27 ]. The Childhood Trauma Questionnaire (CTQ) [ 28 ], a scale proven to be a valid and reliable measure of past traumatic experiences [ 29 ], was used as a self-report questionnaire to assess the nature and severity of childhood trauma. Sub-scales of the CTQ include measures of emotional abuse, physical abuse, sexual abuse, emotional neglect and physical neglect. The self-report Temperament and Character Inventory (TCI) [ 30 ] was also used to measure behaviours associated with seven personality dimensions, namely novelty seeking, harm avoidance, reward dependence, persistence, self-directedness, cooperativeness, and self-transcendence. In addition, participants completed the self-report Young Schema Questionnaire (YSQ) [ 31 ] to assess the current profile of fundamental maladaptive beliefs (cognitive schemas) in OCD and TTM. For each item of the 75-item "short form" of the YSQ (which includes 15 schemas), the answer is required to be placed on a 6-point Likert-type scale (1= 'completely untrue of me', 2 = 'mostly untrue of me', 3 = 'slightly more true than untrue', 4 = 'moderately true of me', 5 = 'mostly true of me', 6 = 'describes me perfectly'). Data analysis As there were few males with TTM (Table 2 ), only clinical data from females with OCD and TTM were analyzed. Chi-square and t-tests were performed to investigate the differences in OCD/TTM phenomenology where appropriate. A one-way analysis of variance (ANOVA) was done to investigate the effects of the primary and comorbid disorders on disability. Subsequently, a two-way fixed effects ANOVA was used to assess the main interactions between primary diagnosis and comorbidity on disability and to test for the main (fixed) effects of the different diagnoses on disability. Residuals of ANOVA's of cognitive schema data in OCD and TTM groups suggested non-normality of the data. As a result, pair-wise comparison tests (Mann-Whitney U) were implemented to compare the two groups on cognitive schemas. Table 2 Comparison of symptomatology: OCD vs TTM Variables OCD TTM P Age of onset (SD) 19.3 (12.0) 11.8 (7.6) <.001 Symptom severity (SD) YBOCS score: 20.1 (8.0); range: 0 – 39 MGHHPS score: 16.1 (6.5); range: 0 – 26 Severity of depressive symptoms (BDI score) 8.9 (11.3) 5.5 (7.2) .04 Poor insight 13.6% 0% NS Treatment response SRI: 90.7% SRI: 42.9% .003 CBT: 73.3% CBT: 33.3% .02 Tics 12.3% 6.1% NS Results Demographics Gender distribution of the sample differed significantly, with a marked predominance of female participants with TTM compared to almost equal numbers of male and female participants with OCD. TTM patients had an earlier age of onset of illness compared to patients with OCD (Table 2 ). Clinical features Comorbidity A number of disorders were more frequent in females with OCD: major depressive disorder (MDD), dysthymia, panic disorder, hypochondriasis and intermittent explosive disorder. In terms of the selected Axis II disorders, obsessive-compulsive personality disorder (OCPD) was more frequent in females with OCD (Table 3 ). Table 3 Lifetime comorbidity: OCD vs TTM Disorder OCD (N = 130) TTM (N = 49) χ 2 P Major depressive disorder 66.9% 49.0% 4.8 .03* Dysthymia 13.8% 2.0% 6.8 .009** Bipolar disorder 0.8% 0% 0.6 0.4 Panic disorder 20.8% 6.1% 6.4 .01* Alcohol abuse 5.4% 2.0% 1.1 0.3 Alcohol dependence 0.8% 0% 0.6 0.4 Substance abuse 0.8% 4.1% 2.0 0.2 Substance dependence 1.5% 2.0% 0.1 0.8 Social phobia 10.0% 8.2% 0.1 0.7 Specific phobia 18.5% 18.4% 0.0 <1.0 Posttraumatic stress disorder 3.1% 0% 2.6 0.1 Generalized anxiety disorder 13.1% 20.4% 1.4 0.2 Body dysmorphic disorder 6.2% 6.1% 0.0 <1.0 Anorexia Nervosa 8.5% 2.0% 2.9 0.1 Bulimia Nervosa 7.7% 6.1% 0.1 0.7 Binge-eating disorder 4.6% 10.2% 1.8 0.2 Hypochondriasis 4.6% 0% 3.9 <.05* Stereotypic movement disorder 3.5% (N = 85) 0% (N = 18) 1.2 0.3 Tourette's disorder 3.8% 2% 0.4 0.5 Tics 12.3% 6.1% 1.6 0.2 Kleptomania 4.6% 4.1% 0.0 0.9 Pyromania 0% 2.0% 2.6 0.1 Compulsive shopping 6.9% 4.1% 0.5 0.5 Hypersexual disorder 1.5% 0% 1.3 0.3 Intermittent explosive disorder 16.2% 6.1% 3.5 .06 OCPD 39.2% 13.3% 11.3 .001** Avoidant personality disorder 21.2% (N = 99) 0.03% (N = 26) 2.9 0.9 Schizotypal personality disorder 5.3% (N = 94) 0% (N = 25) 2.4 0.1 Borderline personality disorder 22.3% (N = 94) 8% (N = 49) 3.0 0.1 * p < .05 (2-tailed) ** p < .01 (2-tailed) Symptom severity The severity of OC symptoms in OCD patients, as measured by the YBOCS severity scale, was 20.1 (± 8.0). TTM patients scored 16.1 (± 6.5) on average on the MGHHPS. Compared to TTM patients, females with OCD had significantly higher depressive symptom scores on the BDI (Table 2 ). Disability The DP was administered to a total of 95 OCD and 30 TTM patients (Table 4 ). OCD patients reported significantly more lifetime impairment due to their illness than TTM patients. More specifically, OCD patients were more impaired in terms of work-related functioning, family functioning, marriage / dating, activities of daily life, and other activities (which included religious activities, membership of clubs, having hobbies, participation in sports etc.) and had more suicidality. One-way ANOVA's showed that the primary diagnosis (either OCD or TTM) (F = 11.84; p = 0.001) and panic disorder (F = 6.73; p = 0.01) had a significant effect on the levels of disability. However, there was no significant interaction effect between the primary diagnosis and panic disorder, suggesting that the levels of disability were dependent on primary diagnosis (either OCD or TTM) (F = 5.79; p = 0.02) and not influenced by the absence or presence of panic disorder (F = 0.001; p = 0.98). Table 4 Disability profile: OCD vs TTM DOMAIN OCD (N = 95) TTM (N = 30) Mann-Whitney U Median Min Max Median Min Max Z P School 1.0 0 4.0 1.0 0 4.0 -1.2 NS Work 2.0 0 4.0 1.0 0 3.0 3.7 <0.001 Family 2.0 0 4.0 1.0 0 4.0 2.4 0.02 Marriage / dating 2.0 0 4.0 1.0 0 4.0 3.1 0.002 Friendships 1.0 0 4.0 1.0 0 3.0 1.1 NS Other activities 2.0 0 4.0 0 0 3.0 2.3 0.02 Activities of daily life 2.0 0 4.0 0 0 3.0 5.5 <0.001 Suicide 1.0 0 4.0 0 0 3.0 2.6 0.009 Total disability 12.0 1.0 26.0 6.5 0 18.0 3.3 <0.001 Character / Temperament Compared to OCD patients, patients with TTM scored significantly higher on novelty seeking, whereas OCD patients had significantly greater harm avoidance (Table 5 ). Table 5 Temperament and Character Inventory: OCD vs TTM TEMPERAMENT / CHARACTER TRAITS* OCD (N = 68) TTM (N = 21) F P NS 17.6 (6.8) 21.6 (6.0) .5 .02 HA 22.3 (7.9) 15.8 (7.3) .4 .001 RD 22.3 (4.2) 21.8 (4.9) .4 NS SD 25.1 (8.2) 28.2 (9.6) .7 NS C 30 (5.1) 29.3 (6.8) 2.8 NS ST 15.2 (6.7) 21.3 (18.3) 5.0 NS * NS = novelty seeking total score SD = self-directedness total score HA = harm avoidance total score C = cooperativeness total score RD = reward dependence total score ST = self-transcendence total score Schemas Fifty-nine OCD and 26 TTM patients fully completed the YSQ. Pair-wise comparison tests (Mann-Whitney U) indicated that OCD and TTM patients differed significantly on 5 schemas, i.e. mistrust / abuse, social isolation, shame / defectiveness, subjugation and emotional inhibition (Table 7 ). More specifically, OCD patients had significantly higher scores on each of these schemas compared to TTM patients. Table 7 OCD and TTM scores on the YSQ subscales Schemas OCD (n = 59) TTM (n = 26) Mann-Whitney U Median Min Max Median Min Max Z P Emotional deprivation 2.4 1.0 5.6 2.3 1.0 5.4 -0.6 NS Abandonment 2.8 1.0 6.0 2.1 1.2 6.0 -0.5 NS Mistrust / abuse 2.6 1.0 5.8 1.9 1.9 5.4 -2.3 .02 Social isolation 2.4 1.0 6.0 1.9 1.9 6.0 -2.7 .007 Shame / defectiveness 2.2 1.0 6.0 1.4 1.4 4.8 -3.0 .003 Failure to achieve 2.0 1.0 6.0 1.9 1.9 4.8 -0.9 NS Incompetence 2.0 1.0 4.8 1.8 1.8 4.4 -1.2 NS Vulnerability to harm 2.2 1.0 6.0 1.6 1.6 5.2 -1.8 NS Enmeshment 1.8 1.0 6.0 1.7 1.7 4.2 -0.6 NS Subjugation 2.2 1.0 6.0 1.7 1.7 5.6 -2.8 .005 Self-sacrifice 3.4 1.2 6.0 3.2 3.2 5.8 -1.0 NS Emotional inhibition 2.4 1.0 5.0 1.4 1.4 4.4 -3.8 <.001 Unrelenting standards 3.8 1.2 6.0 3.6 3.6 6.0 -0.9 NS Entitlement 2.6 1.0 5.8 2.7 2.7 6.0 -0.7 NS Self-discipline 3.0 1.2 6.0 2.8 2.8 4.8 -1.3 NS Precipitating factors Interpersonal trauma history OCD patients reported more childhood sexual abuse than did TTM patients (p = .04). Brain trauma history OCD and TTM patients did not differ significantly in terms of a history of serious head injury associated with the onset of OCD or TTM. History of autoimmune infections Compared to none in the TTM group, 9 OCD patients reported onset of their OCD with an episode of bacterial pharyngitis (p = .06). In terms of other autoimmune infections, OCD and TTM patients did not differ significantly. Hormonal influence Female OCD and TTM patients did not differ significantly in terms of the impact of premenstrual/menstrual/menopausal symptoms on their illness. Compared to 42 (38.5%) of 109 OCD patients who reported OC symptom changes in the premenstrual/menstrual period, 16 of 32 TTM patients (50%) reported regular changes in their symptoms during this time. Seventeen (n = 17) OCD patients were menopausal and 35.3% (n = 6) of these women reported that their OC symptoms only started with menopause. One of the TTM patients had gone through menopause with no effect on her hair-pulling symptoms. However, OCD and TTM patients differed significantly in terms of the temporal association between pregnancy/puerperium and onset of illness: 42.6% (26 of 61) OCD patients reported OCD onset while pregnant or within a month of childbirth, compared to 7.7% (1 of 13) of TTM patients (χ 2 = 6.8; p = .009). Treatment response Significantly fewer TTM patients reported a clinical response to either CBT- or SRI-treatment than did OCD patients (Table 2 ). Discussion A comparison of women with TTM and with OCD found significant differences in clinical variables; OCD patients had more comorbidity, greater disability, increased childhood interpersonal trauma (specifically sexual abuse) and more maladaptive schemas. Fewer TTM patients, however, reported having responded to treatment. The gender ratio findings in both OCD and TTM groups were similar to other surveys where a mean female:male ratio of 1.5:1.0 in OCD [ 11 , 32 , 33 ] and approaching 10:1 in TTM [ 34 ] were documented. OCD patients' mean total score on the YBOCS (i.e. 20.1 ± 8.0) puts them in the "moderate" severity category [ 22 ]. The mean hair-pulling severity score on the MGHHPS (16.1 ± 6.5) was similar to that reported in other studies [ 35 , 36 ]. Taken together, these data suggest that our patients are not dissimilar from those assessed at other sites. Our comorbidity findings are consistent with existing data suggesting that depressive and anxiety disorders are highly prevalent in both OCD and TTM, and significantly more prevalent in OCD [ 3 , 16 , 18 ]. Indeed, compared to TTM, comorbidity in OCD is greater across a range of different diagnostic categories including mood (MDD, dysthymia), anxiety (panic disorder), OCD-related (hypochondriasis) and personality disorders (OCPD). Such comorbidity appears to extend also to impulse control disorders (intermittent explosive disorder), a finding which argues against the current classification of trichotillomania as a member of this spectrum of conditions. Our findings of increased disability in OCD is consistent with studies on OCD suggesting it is one of the most impairing of all medical disorders [ 37 ]. A number of clinical studies have emphasized the burden of OCD across different domains, including higher rates of divorce and separation than in subjects without OCD [ 13 ] and significantly impaired instrumental functioning (work, school, home making and family life) [ 38 - 41 ]. However, the impairment and distress due to TTM should not be underestimated: TTM can be associated with serious sociological and psychological effects (e.g. strong feelings of shame and embarrassment [ 42 ], as well as avoidance behaviour including potentially dangerous avoidance of medical care [ 43 ]) resulting in a significant decline in quality of life (QOL) for patients, their family members and significant others [ 44 , 45 ]. OCD patients reported significantly more sexual abuse than TTM patients (p = .04). This finding differs from our previous data suggesting similar rates of childhood interpersonal trauma (CIT) in OCD and TTM [ 46 ]. However, the current sample size is much increased, resulting in more power to detect smaller differences. Indeed, increased rates of OCD (and other anxiety disorders) have previously been linked with a history of physical and sexual abuse during childhood [ 47 , 48 ]. Nevertheless, in both OCD and TTM, dissociative symptoms – which are present in a minority of patients in both conditions – are positively correlated with a history of childhood interpersonal trauma [ 49 ], so that a potential role for CIT in some TTM patients should not be ignored [ 50 ]. TTM patients had significantly more novelty seeking (NS) than OCD patients, whereas OCD patients scored significantly higher on harm avoidance (HA) compared to TTM. Our findings are consistent with previous work on temperament / character in OCD, showing increased HA and decreased NS [ 51 - 53 ]. Of note, compared with mean temperament scores obtained in a normal community sample [ 30 ], both TTM and OCD scored high on HA. NS scores in the TTM sample were higher than in the OCD sample, but compared to normal controls, these fell in the "medium" range. The higher NS in TTM may however point to greater dopaminergic involvement in this disorder, and might also be used to argue that TTM lies closer to the more impulsive risk-/novelty-seeking pole of an impulsive-compulsive (IC) spectrum of disorders [ 54 ]. OCD patients had more maladaptive cognitive schemas than TTM, i.e. mistrust / abuse, social isolation, shame / defectiveness, subjugation and emotional inhibition. The schemas that OCD and TTM patients differed on are included in 2 of the 4 higher order factors (i.e. "impaired autonomy" and "disconnection") described by Lee and colleagues' YSQ factor model [ 55 ]. While schemas are thought to represent responses to life experience, including the experience of a disorder, they may also reflect underlying symptoms. Given that maladaptive schemas in OCD were not reminiscent of its characteristic symptoms, it is likely that they at least partly reflect life experience. An increased number of maladaptive schemas in OCD is consistent with higher rates of comorbidity, disability, and functional impairment. Nevertheless, further empirical investigation is needed to assess the relationship between schemas and illness course. Hormonal influences have previously been investigated in OCD [ 56 ] and TTM [ 57 ]. For example, it has been noted that menarche, premenstruum, pregnancy [ 58 ], and menopause [ 59 ] may be related to onset or relapse in OCD. Similarly, in a study that investigated the relationship of the menstrual cycle and pregnancy to compulsive hair-pulling, premenstrual symptom exacerbation was reported for actual hair-pulling, urge intensity and frequency, and ability to control pulling [ 57 ]. In that study the impact of pregnancy on TTM was less clear. Our findings suggest that significantly more OCD patients than TTM patients report an association between pregnancy/puerperium and the onset of illness. This finding is in part consistent with previous work suggesting that the postpartum may constitute a risk for the onset of OCD in women [ 60 ]. Taken together our data suggest both similarities and differences in the role of sexual hormones in the mediation of OCD and TTM. Although rare, brain injury may play a role in some cases of OCD [ 61 , 62 ]; in only one OCD patient (and none of the TTM patients), head injury was associated with onset of obsessive-compulsive symptoms. No data could be found on the potential role of brain injury in the etiology of hair-pulling. A number of patients associated the onset of their OCD onset with an infection, possibly bacterial pharyngitis. This finding is consistent with a body of data suggesting post-streptococcal disease is a cause of OCD in children and adolescents [ 63 ], and perhaps also adults [ 64 , 65 ]. There is less work demonstrating a role for autoimmune factors in TTM [ 66 ]. Notably, the data on bacterial pharyngitis were based on retrospective assessment and could have been contaminated by memory bias. In our study, more OCD patients reported a positive response to treatment (with CBT or SRI's) than TTM patients. These data should be interpreted cautiously given the retrospective assessments. Nevertheless, there is evidence that SRI's in TTM may not be as effective over the long-term as in OCD [ 2 ]. About 40–60% of OCD patients respond to the first trial of an SRI [ 67 ], with a proportion of non-responders to a single SRI responding to administration of a second SRI [ 68 ]. In comparison, it has been suggested that TTM patients judge their treatment (including pharmacotherapy, psychotherapy, and behaviour modification) to be relatively ineffective [ 69 ]. The usefulness of SRI's in TTM has been investigated in a number of studies with results so far being equivocal. For example, Christenson et al [ 70 ] were unable to document efficacy for fluoxetine in a placebo-controlled trial in which patients received 6 weeks of the active agent in doses of up to 80 mg/day. Anecdotal evidence also suggests that the effectiveness of SRI's in TTM may wane with time [ 71 ]. Although there is evidence for the usefulness of behaviour therapy in both OCD [ 72 ] and TTM [ 73 ], the focus of the treatment differs in the two disorders (exposure and response prevention in OCD versus habit reversal in TTM). Keuthen et al [ 74 ] have suggested that "state-of-the-art" behavioral and pharmacological treatments offer substantial clinical benefit to patients with TTM, but in general clinics there may be relatively little experience with highly specialized interventions. Several limitations of the current study should be acknowledged. First, interviewers were not blind to the patients' psychiatric diagnosis, so potentially biasing clinician's assessments. Nevertheless, a structured diagnostic instrument ensures a reasonable degree of reliability. Second, instruments employed in the current study are intended for use in adults rather than younger subjects. However, in the case of children and adolescents, the SCID-I was supplemented with a clinical interview of parents or guardians, and self-report data was included only when it was clear that questionnaires had been completed meaningfully. Third, source of referral, and the duration of OCD/TTM, were not controlled for in the analysis, so potentially biasing the analyses. However, given the chronicity of both conditions, this is unlikely to have materially affected the findings. Fourth, males, as well as patients with comorbid OCD and TTM were excluded from the investigation; so that the results here may not be generalizable to all OCD or TTM subjects. Given evidence that the phenotype of OCD [ 75 ] and of TTM [ 10 ] varies with gender, additional work on male subjects should be undertaken in the future. Conclusions In conclusion, our data suggest that despite some overlap, TTM differs from OCD in terms of demographics (gender distribution), associated clinical variables (e.g. comorbidity, cognitive schemas, temperament/character profiles and disability), precipitating factors (trauma history) and treatment response. It has been suggested that although TTM is not the same as OCD, it lies on a compulsive-impulsive spectrum of disorders [ 54 ]. However, it is notable that impulsivity may be an important component of OCD [ 76 ], and rather than viewing OCD and TTM on a single dimension, compulsivity and impulsivity should arguably therefore be seen as lying on orthogonal dimensions. Although TTM patients had more novelty seeking, OCD patients were more likely to have intermittent explosive disorder; such data support a view that TTM should not be classified as an impulse control disorder. Indeed, TTM may have more in common with conditions characterized by stereotypical self-injurious symptoms, such as skin-picking [ 77 ]. Differences between OCD and TTM may reflect contrasts in underlying psychobiology, and may necessitate contrasting treatment approaches. Competing interests The author(s) declare that they have no competing interests. Authors' contributions CL: • participated in the design of the study, • was responsible for patient recruitment, • did most of the clinical assessments, • performed the statistical analyses, • helped to obtain funds for the research, and • was responsible for the final writing up of data. SS: • participated in the design of the study, • assisted with recruitment of patients, and • supervised writing and statistical analyses. PLduT: • participated in the design of the study, • assisted with recruitment of patients, and • did some of the clinical assessments. DGN: • was the primary statistical consultant for analyses. DJHN: • participated in the design and coordination of the study, • was responsible for patient recruitment, and • assisted with clinical assessments. RS: • assisted with literature review of cognitive schema data, and • asisted statistical analysis of schema data. DJS: • conceived of the study • supervised coordination, statistical analysis, and writing, • did the final revision of paper before submission, and • assisted with obtaining of funds. All authors read and approved the final manuscript. Table 6 TCI-temperament cut-off scores: A normal community sample* TCI – TEMPERAMENT TRAITS NS HA LOW MEDIUM HIGH LOW MEDIUM HIGH 16 19.5 22 8 12.5 16 *from Cloninger et al, 1994 (reference nr. 31) Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546013.xml
512287	Meningitis due to Fusobacterium necrophorum in an adult	Background Fusobacterium necrophorum may cause a number of clinical syndromes, collectively known as necrobacillosis. Meningitis is a significant cause of mortality, rarely reported in the adult population. Case presentation We report a fatal case of meningitis, caused by Fusobacterium necrophorum , secondary to otitis media in an alcoholic male. Diagnosis was delayed due to the typical slow growth of the organism. The clinical course was complicated by encephalitis and by hydrocephalus. The patient failed to respond to metronidazole and penicillin. The patient died on day 12 from increased intracranial pressure and brain stem infarction. Conclusions This case emphasizes the need for a high index of clinical suspicion to make the diagnosis of Fusobacterium necrophorum meningitis. We recommend the use of appropriate anaerobic culture techniques and antimicrobial coverage for anaerobic organisms when the gram stain shows gram negative bacilli.	Background Fusobacterium necrophorum may cause a number of clinical syndromes, collectively known as necrobacillosis. Meningitis is a significant cause of mortality, rarely reported in the adult population. Diagnosis is often delayed by difficulties encountered in isolating the organism. Here, we report a fatal case of meningitis from complicated otitis media caused by this organism. Case presentation A 51 year male was brought to the emergency department (ED) by his family for confusion and shaking episodes. The patient was very lethargic in the ED and was intubated for airway protection. Family reported that he had not been well for several months, but the family was not able to define any specific symptoms until the past few days when he reported right ear pain. The patient had been given a prescription for erythromycin within the previous week for a diagnosis of otitis media; the patient had not taken any doses for at least two days. The only significant past medical history was of ongoing alcohol abuse without intravenous drug use. The family reported that he had no alcohol intake over the two days prior to presentation. Vitals on presentation: temperature 98.5°F (36.9°C), heart rate 125/min, respirations 25/min, blood pressure 219/121, Oxygen saturation 96 % on room air. On physical exam he was lethargic, the right tympanic membrane was erythematous with decreased movement on pneumatic otoscopy. The neurological exam revealed no focal deficits. A non-contrasted head CT scan showed no abnormalities; there was no evidence of an intracranial bleed or elevated intracranial pressure. Lumbar puncture revealed markedly purulent CSF with 338,400 white cells per cubic millimeter (94% neutrophils), glucose was less than 2 mg/dl and protein was 3544 mg/dl. Gram stain showed numerous gram negative bacilli. He was initially treated with high dose dexamethasone, along with ceftriaxone, ampicillin and vancomycin. When gram negative bacilli were identified in the CSF, the antibiotic regimen was revised to ceftazidime, levofloxacin and metronidazole. There was a decline in mental status despite aggressive antibiotic therapy. He continued to require complete ventilatory and pressor support to maintain adequate blood pressure and oxygenation. Physical examination revealed up-going plantar reflexes on day 6. MRI brain showed extensive infarction of the brain stem, obstructive hydrocephalus, severe basilar meningitis, and ventriculitis (see figures 1 , 2 , 3 , 4 ). A ventricular drain was placed in an attempt to decrease intracranial pressure. Repeat CSF analysis on day 9 showed 1800 white cells per cubic millimeter and protein was 118 mg/dl. At this time, initial culture of the CSF was reported to be growing F. necrophorum. Penicillin G was added with discontinuation of ceftazidime and levofloxacin. The patient further deteriorated and the neurological exam revealed absent corneal reflexes, and no response to painful stimulus. EEG was done to assess brain function, which revealed evidence of brain death. The caloric test also confirmed brain death. Discussion Fusobacterium is an anaerobic, non spore forming gram negative rod which belongs to the family of Bacteroidaceae . It is a part of normal flora which is found in mouth, upper respiratory tract, gastrointestinal tract and vagina. It can cause local infections like pharyngitis, tonsillitis, mastoiditis or can cause severe bacteremic illness like meningitis. Other central nervous system complications caused by Fusobacterium include cranial nerve palsy, sinus venous thrombosis, and brain abscess [ 1 , 2 ]. In one recent series of brain abscesses, F. necrophorum was the most common anaerobe isolated, found in 33% of patients [ 2 ]. The term "necrobacillosis" or "Lemierre syndrome" is used for the severe bacteremic illness caused by F. necrophorum . Lemierre syndrome has been described to progress through three stages [ 3 ]. The primary infection is pharyngitis in the majority of patients. The second stage is invasion into the pharyngeal space with the development of internal jugular septic thrombophlebitis. The third stage is metastatic spread of the infection. In one series of F. necrophorum meningitis, middle ear infection was the source for 75% of the cases [ 4 ]. Other predisposing infections include sinusitis, pharyngitis and lung infections [ 5 ]. Diagnosis is often delayed by the difficulties in isolating and identifying the organism. A high index of suspicion is necessary in the diagnosis of this infection. There have been over 20 reported cases of meningitis due to Fusobacterium [ 1 , 4 , 6 - 19 ], only one of which was in an adult [ 6 ]. Despite appropriate antibiotic therapy, the outcome is poor with the mortality rate from meningitis due to Fusobacterium as high as 33% with residual sequelae common among survivors (60%) [ 7 , 8 ]. Although the antibiotic regimen of choice has not been established metronidazole seems to be a useful agent. Some authors have suggested the addition of penicillin G to treat this infection [ 8 ]. It has been recommended that metronidazole be administered for at least 6 weeks [ 6 ]. Relapse is possible if the treatment is discontinued prematurely [ 4 , 6 ]. Conclusions This case shows the severity of illness that can result from infection with F. necrophorum . Anaerobic organisms should be considered as potential causative agents of meningitis when routine cultures are negative. Routine cultures of cerebrospinal fluid do not include the use of anaerobic growth media. Therefore, appropriate anaerobic culture techniques should be employed when sinus, otitic or mastoid symptoms precede or accompany the onset of meningitis in children or adults. The presence of irregularly stained gram negative rods in the CSF or meningitis unresponsive to empiric antibiotics should also raise the suspicion of anaerobic infection. The addition of metronidazole should be considered in these cases. Competing interests None declared. Authors' contributions AI cared for the patient. SG drafted the manuscript. TH cared for the patient. All authors reviewed and approved the final manuscript. Pre-publication history The pre-publication history for this paper can be accessed here:	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC512287.xml
515311	What does my patient's coronary artery calcium score mean? Combining information from the coronary artery calcium score with information from conventional risk factors to estimate coronary heart disease risk	Background The coronary artery calcium (CAC) score is an independent predictor of coronary heart disease. We sought to combine information from the CAC score with information from conventional cardiac risk factors to produce post-test risk estimates, and to determine whether the score may add clinically useful information. Methods We measured the independent cross-sectional associations between conventional cardiac risk factors and the CAC score among asymptomatic persons referred for non-contrast electron beam computed tomography. Using the resulting multivariable models and published CAC score-specific relative risk estimates, we estimated post-test coronary heart disease risk in a number of different scenarios. Results Among 9341 asymptomatic study participants (age 35–88 years, 40% female), we found that conventional coronary heart disease risk factors including age, male sex, self-reported hypertension, diabetes and high cholesterol were independent predictors of the CAC score, and we used the resulting multivariable models for predicting post-test risk in a variety of scenarios. Our models predicted, for example, that a 60-year-old non-smoking non-diabetic women with hypertension and high cholesterol would have a 47% chance of having a CAC score of zero, reducing her 10-year risk estimate from 15% (per Framingham) to 6–9%; if her score were over 100, however (a 17% chance), her risk estimate would be markedly higher (25–51% in 10 years). In low risk scenarios, the CAC score is very likely to be zero or low, and unlikely to change management. Conclusion Combining information from the CAC score with information from conventional risk factors can change assessment of coronary heart disease risk to an extent that may be clinically important, especially when the pre-test 10-year risk estimate is intermediate. The attached spreadsheet makes these calculations easy.	Background Aggressive primary prevention of coronary heart disease (CHD) is most appropriate in patients at relatively high risk of CHD events [ 1 , 2 ]. The coronary artery calcium (CAC) score is an independent predictor of coronary heart disease risk [ 3 - 7 ], and therefore may help in deciding how aggressively to pursue cholesterol-lowering, anti-platelet therapy and other primary prevention strategies. To use a given CAC score result, however, one must know how that score compares with the score of an average person of the same sex, age and CHD risk factor profile. A CAC score of 50, for example, may be unusually high for a 40-year-old woman without other CHD risk factors, but unusually low for a 70-year-old man with hypertension. The same score, therefore, affects risk assessment in opposite directions for these two patients. How should a clinician use this CAC score (or any other) when assessing the CHD risk of a more typical patient, say a 60-year-old woman with hypertension and high cholesterol? To answer this question, we need to know the effects of age, sex and other CHD risk factors on the expected distribution of CAC scores. Several large cross-sectional studies have described the prevalence and extent of CAC among different age/sex groups [ 6 , 8 - 10 ] without accounting for conventional CHD risk factors that may strongly influence predicted CAC scores. Five previous studies examined how CAC relates to conventional CHD risk factors [ 11 - 15 ]. Only one of these was adequately powered [ 15 ], none adequately accounted for the abnormal distribution of CAC scores, and none yielded estimates usable for clinical decision-making. We identified a large sample of men and women without clinical CHD who presented for electron beam computed tomography scanning. Using questionnaire data collected from these patients about smoking habits and medical history (hypertension, high cholesterol and diabetes), we determined how conventional CHD risk factors, along with age and sex, affect CAC scores. We then developed a method for combining information from conventional risk factors and the CAC score (easy spreadsheet calculator attached), and we present several examples illustrating how that method may be applied in common clinical situations. Methods Study sample All persons referred by their physician to an electron beam computed tomography (EBCT) scanning center in Nashville, Tennessee for measurement of coronary artery calcification between May 15, 1995 and December 31, 1997 were eligible for inclusion. Subjects with a history of CHD or complaining currently of any chest pain were excluded, as were subjects for whom CHD risk factor data were incomplete or missing. Only the first CAC score was included for those who received more than one EBCT scan. Measurement of coronary heart disease risk factors Current age, sex and presence of CHD risk factors were elicited by questionnaire from subjects and referring physicians. Each subject was labeled with hypertension, high cholesterol and/or diabetes mellitus if they answered affirmatively to the question, "Has your physician ever told you that you needed medicine for X?", or if their physician confirmed that such a condition was documented in their medical records. Patients were labeled as smokers if they currently smoked or had quit smoking within the preceding 3 months. No direct measurements of blood pressure, lipids or glucose were taken for the purposes of this study. Estimation of the 10-year risk of coronary heart disease events We estimated the 10-year risk of a first CHD event using published mathematical models based on the Framingham study [ 16 ]. For this purpose, we assumed that subjects reporting hypertension had systolic blood pressures of 140–160 mmHg and/or diastolic blood pressures of 90–100 mmHg (Stage I hypertension), and that subjects without hypertension had systolic pressures of 120–130 and diastolic pressures of 80–85 mmHg. We also assumed that patients with high cholesterol had low-density lipoprotein (LDL) cholesterol levels of 130–159 mg/dl and high density lipoprotein (HDL) cholesterol levels of 35–44 mg/dl, whereas patients without high cholesterol had LDL cholesterol levels of 100–129 mg/dl and HDL cholesterol levels of 45–49 mg/dl (for men) or 50–59 mg/dl (for women). Smoking and diabetes mellitus were dichotomous variables in both Framingham models [ 16 ] and our data set. We then used published model coefficients [ 16 ] to estimate the 10-year risk for each patient in our study. Measurement of the CAC score Each subject underwent electron beam computed tomography scanning with an Imatron C-100 or C-150 scanner (Imatron, South San Francisco, California) after giving written informed consent. During a single breath hold, 40 consecutive slices of 3 mm thickness were obtained starting at the level of the carina and proceeding to the level of the diaphragm. Scans were obtained within 100 ms and were electrocardiographically triggered at 60–80% of the R-R interval. Coronary calcification was defined as a plaque of at least 3 consecutive pixels (area = 1.03 mm 2 ) with density ≥ 130 Hounsfield units. The CAC score was calculated according to the method described by Agatston [ 17 ]. Statistical analysis We categorized patients according to age and sex, and examined histograms, quantile plots and box plots in each category to evaluate distributional normality. The CAC score is fundamentally not normally distributed because of the large percentage of zero measurements, and hence is not amenable to a normalizing transformation, as noted by others [ 13 ]. We also considered a censored normal distribution, which would have allowed a one-step Tobit regression analysis. However, even after square- and cube-root transformations, the zero scores were distributed in a manner inconsistent with the Tobit regression model. After exclusion of zero values, however, the log-transformed CAC score was approximately normally distributed (Figure 1 ). This led us naturally to a two-stage modeling approach. We first applied logistic regression to model the probability of a non-zero score, and then used linear regression to model the actual CAC score, log-transformed, for the subset of patients with non-zero values. Using this methodology, we assessed the independent effects of CHD risk factors on both the presence and extent of CAC. We considered three sets of predictors: 1) age and sex, 2) age, sex, hypertension, high cholesterol, smoking, and diabetes, and 3) the Framingham 10-year CHD risk estimate. We examined whether the effects of age were linear (as opposed to J-shaped, for example) by testing a quadratic term in the model containing only age and sex. We evaluated the ability of each logistic model to discriminate subjects at high and low risk for CAC using the C-statistic, and estimated the proportion of variability in the extent of CAC explained in each linear regression model using the adjusted-R 2 statistic. Finally, we used coefficients, intercepts and residual variance from logistic and linear models to estimate the probability that the CAC score of an individual with known risk factors would fall into each of four standard CAC score categories (0, 1–100, 101–400, and >400). We estimated these probabilities, using models containing the 10-year risk estimate as the only predictor, for a range of 10-year risk estimates. We also estimated these probabilities, using models with all CHD risk factor predictors, for the specific clinical scenario described in the Introduction (a 60-year-old woman with hypertension and high cholesterol) and for several other scenarios. We compared the actual distribution of CAC scores among 58–62-year-old women with hypertension and high cholesterol in our sample (n = 130) with predictions from 1) our two-stage model, 2) a one-stage model using Ln(CAC score + 1) as a continuous outcome in a linear regression model, and 3) a one-stage model using a censored normal distribution of cube-root transformed CAC scores (a Tobit regression model). This comparison was made both graphically and statistically, using X 2 tests with 3 degrees of freedom to compare the expected frequencies based on each model with the observed frequencies. Lower p values, in this case, indicate a poorer fit of the model to the observed data. All statistical analyses were performed with Stata 7.0 (College Station, Texas). Combining information from conventional risk factors and the CAC score First, we calculated the Framingham 10-year CHD risk estimate (and corresponding 1-year risk estimate assuming an equal event rate each year) according to published models [ 16 ]. Next, we calculated the probability, as described above, that that individual's CAC score would fall into each one of four standard CAC score categories [ 15 , 18 , 19 ] (0, 1–100, 101–400, and >400). We obtained risk factor-adjusted relative risk (RR) estimates from a meta-analysis [ 7 ] comparing the risk of a CHD event among persons with CAC scores of 1–100 (RR = 2.1), 101–400 (RR = 5.4) and <400 (RR = 10) to the risk in a person with a CAC score of zero. The analysis was repeated using more conservative estimates from the same paper: RR = 1.7 (for CAC 1–100), RR = 3.0 (for CAC 101–400), and RR = 4.3 (for CAC>400). The post-test CHD risk estimates for each CAC score category were then calculated algebraically by assuming that the overall 1-year CHD risk estimate represents an average of the 1-year risk estimates from the four CAC score categories, weighted by the probabilities that an individual's score would fall into each category. A spreadsheet that automates these calculations is attached. Results Study population We identified 9341 persons without chest pain or a history of CHD presenting for their first EBCT scan between 4/15/95 and 12/31/97. Our sample was mostly middle-aged, but included persons as young as 35 years and as old as 88 years of age. Forty percent were women. The proportion with cardiac risk factors was high, though only 9% were diabetic (Table 1 ). Framingham 10-year CHD risk estimates ranged widely, mostly dependent on age, but most were between 7% and 15%. Coronary artery calcium score distributions Coronary artery calcium scores ranged from 0 to 4058. The mean score (± standard deviation) was 135 (± 377), and the median was 4 (25 th –75 th percentile: 0 – 87). The prevalence of zero scores ranged from 80% among women younger than 50 years to 5% among men 70 years old or older. After excluding zero scores, log-transformed CAC scores were approximately normally distributed, and appeared to be strongly associated with age and sex (Figure 1 ). Predictors of the presence and extent of coronary artery calcification Age and sex were strong predictors of the presence of CAC in logistic regression models (Table 2 ). There was no evidence that the effects of age were non-linear (i.e. J- or U-shaped) (p-value = 0.32 for a quadratic age term). Conventional CHD risk factors were also independent predictors of the presence of CAC (p < 0.001 in all cases). The logistic model with age, sex and all CHD risk factors produced the most accurate model (C-statistic = 0.78). The Framingham 10-year CHD risk estimate was also a very strong predictor of coronary artery calcification, though the model containing the 10-year risk estimate as the only predictor was slightly less accurate (C-statistic = 0.74). Among patients with non-zero CAC scores, age and sex remained strong predictors of the extent of coronary artery calcification, as measured by the Ln(CAC score) (Table 3 ). Again, the effects of age appeared to be linear (p = 0.16 for the quadratic age term). All conventional CHD risk factors remained statistically significant predictors of the extent of coronary artery calcification (p < 0.001 for all predictors except high cholesterol at p = 0.004). Again, the Framingham 10-year CHD risk estimate was a very strong predictor of the extent of calcification, though when used alone in a model, it explained somewhat less of the variance (R 2 = 0.11) than the full model (R 2 = 0.17). Coronary artery calcium distribution predictions Using these models, we estimated the probability of measuring a CAC score in each of four standard CAC score categories (0, 1–100, 101–400, and >400) using the Framingham 10-year CHD risk estimate, a value easily calculated from conventional CHD risk factors using accessible web- or handheld computer-based software. These probabilities ranged widely based on the value of the 10-year risk estimate, with the probability of measuring a zero CAC score going from 75% (at a 10-year risk of 2.5%) to 13% (at a 10-year risk of 25%) (Table 4 ). Risk integration example Using the case example presented in the Background section, we calculated that a 60-year-old woman with Stage I hypertension (140/90 mmHg) and high cholesterol (LDL cholesterol = 155 mg/dl, HDL cholesterol = 40 mg/dl) will have a 15% risk of suffering a CHD event in 10 years, according to the Framingham equation. If this women undergoes EBCT scanning, our models predict a 47% chance that her CAC score will be zero, a 36% chance that it will be between 1–100, a 12% chance that it will be between 101–400, and a 5% chance that it will be greater than 400. By integrating this information with previously published relative risk estimates (see Additional File 1 ), we estimate her 10-year CHD risk to be as low as 6% (if her CAC score is 0), or as high as 51% (if her CAC score is >400). These estimates are only moderately sensitive to variation in the relative risk assumptions (Table 5 ), and may be easily calculated in any clinical scenario in which CHD risk factor data is available; see Table 5 for several other examples. Comparing predictions from different modeling strategies Our strategy outperformed two other modeling strategies in predicting the actual CAC distribution among the 58–62-year-old non-smoking non-diabetic women with hypertension and high cholesterol in our study sample (n = 127) (Figure 2 ). The one-stage regression model using Ln(CAC score +1) as the outcome, which has been utilized extensively in previous research [ 11 , 12 , 14 , 20 ], performed particularly poorly. Discussion In this article, we present a clinically useful method of combining information from the CAC score with pre-test coronary risk estimates. To fully appreciate the utility of this analysis, it may be worthwhile to discuss the example from the Background section further. According to current guidelines, this 60-year-old woman, whose 10-year CHD risk estimate is about 15%, should receive both aspirin and cholesterol-lowering drug therapy, aiming for a goal LDL cholesterol of 130 mg/dl [ 1 , 2 ]. After measuring her CAC score, however, there is a good chance (64%) that our recommendations would change. If her CAC score were zero (47% chance), our estimate of her 10-year CHD risk would be approximately halved (6–9%). Given this information, we would continue to recommend a healthy diet and exercise, but might decide that cholesterol-lowering medication is unnecessary [ 1 ], and that the benefits of aspirin in terms of CHD prevention do not outweigh the risk of hemorrhagic stroke associated with aspirin use [ 2 ]. On the other hand, if her CAC score were over 100 (17% chance), our estimate of her CHD risk would be approximately doubled (25–31% if CAC score = 101–400) or tripled (34–51% if CAC score > 400). In such a case, we would certainly recommend both aspirin [ 2 ] and cholesterol-lowering medication [ 1 ] and would probably aim for a more aggressive LDL cholesterol goal of < 100 mg/dl [ 1 ]. The probability that her treatment plan would be altered by measurement of her CAC score, therefore, is approximately 64% (the probability that her score is either 0 or >100 = 47% + 17%), indicating likely usefulness of the test in this situation. The third and fourth clinical scenarios presented in Table 5 , on the other hand, provide examples where the test is unlikely to change management. The 40-year-old woman who smokes, for example, has a very low pre-test 10-year CHD risk (3%). It is very likely her CAC score will be zero (89%) or less than 100 (10%), in which case her post-test 10-year CHD risk will still be low (≤ 5%) and her management would not change. The 80-year-old man with high cholesterol has a high pre-test 10-year CHD risk (26%) and a high probability of having a high CAC score (70% will have a score > 100), in which case his post-test 10-year CHD risk would remain over 20% and his management would have to remain aggressive. In these cases, and others in which the risk factor profile indicates very low or very high pre-test risk, the test is not likely to provide useful results, and the clinician might decide not to order the test. We have provided a simple spreadsheet (see Additional File 1 ) that may be used by readers of this article to replicate these analyses and apply our models to other clinical scenarios. While others have proposed similar Bayesian approaches to use of the CAC score for coronary risk prediction [ 6 , 21 - 24 ], ours has advantages. Previous approaches do generally take into account the pre-test probability of coronary heart disease, but none consider the expected distribution of CAC scores in the tested population after adjustment for conventional CHD risk factors. Raggi et al advocate use of an age- and sex-adjusted calcium score percentile, but this ignores both persons with zero scores and the strong effects of other risk factors such as hypertension and hypercholesterolemia [ 6 ]. Some approaches use only sensitivity and specificity from dichotomized CAC score cutoffs [ 21 , 23 ], and others use CAC score-specific relative risks generated from a single study population [ 6 , 24 ]. Only two provide actual post-test risk estimates for specific clinical situations [ 23 , 24 ]. Our approach takes into account the pre-test coronary risk, the expected distribution of CAC scores adjusted for all conventional CHD risk factors, and summary adjusted relative risks from a recent meta-analysis, and provides clinically relevant post-test risk estimates that may be directly useful to primary care physicians, cardiologists and patients as they decide whether or not to take medications for primary prevention of CHD. This analysis confirms that conventional risk factors for CHD (hypertension, diabetes, smoking and high cholesterol, as well as increasing age and male sex) are independent predictors of coronary artery calcification. This finding is consistent with previous studies [ 11 - 15 ]. We also present expected CAC score distributions for a variety of clinical situations, which are not easily calculated from other studies, via Tables 4 and 5 and the attached spreadsheet calculator. Our finding that high cholesterol was less strongly associated with the extent of CAC than other CHD risk factors is consistent with the other large study addressing this issue [ 15 ], and perhaps reflects effective medical treatment for hypercholesterolemia. Male sex was a very strong predictor of the presence and extent of CAC – women with the same CHD risk factor profile would be expected to develop CAC approximately 12 years later than men, and remain approximately 11 years behind men in the extent of their calcification. Finally, our analysis provides a guide for how to use the CAC score as a surrogate outcome when studying causes of coronary artery disease (a widely used study design [ 25 - 27 ]). The central problem with this approach is the fundamentally non-normal distribution of CAC scores, which makes parametric statistic testing (including both simple t-tests and multivariable linear regression) invalid. In dealing with this issue, some researchers have used the Ln(CAC score +1) as an outcome in linear regression analyses [ 11 , 12 , 14 , 20 ]. This approach is not ideal, as the Ln(CAC score +1) is still grossly non-normal – there are too many zero scores. Adding 1 to the CAC score makes the log-transformation possible (yielding zeroes instead of negative infinity), but it does not solve the distributional problem, and leads to predictions that misrepresent actual CAC score distributions (Figure 2 ). This observation has led others to present only non-parametric percentile data without multivariable modeling [ 6 , 8 - 10 ], but this approach does not allow adjustment for conventional CHD risk factors that we have shown are strong predictors of the CAC score. One other group used ordinal logistic regression analysis to analyze CAC scores categorized into four ordinal categories (quartiles in their study sample) [ 13 ]. While such an approach does allow multivariable modeling with ordinal logistic regression, it does not take full advantage of the continuous nature of the CAC score and may blur the important distinction between zero and non-zero scores. Our analysis suggests that a two-step approach (using first logistic regression to model the risk of having a non-zero score, then linear regression of log-transformed non-zero CAC scores to model the extent of coronary calcification) will allow multivariable analysis of the interval data provided by the CAC score without violating the basic assumptions of parametric statistics. Our analysis has a number of limitations, perhaps the most important being a lack of clinical detail about participants. While we had information about conventional risk factors (hypertension, high cholesterol, diabetes mellitus and tobacco use), the data were only available from a questionnaire, and were not confirmed by direct measurement. Only dichotomous indicators of such conditions were used. Furthermore, other conditions and indicators of high CHD risk such as family history of CHD, obesity, physical activity, income, education, and levels of C-reactive protein, triglycerides and Lp(a), for example, were unavailable. Whether such factors are important predictors of the presence and extent of coronary artery calcification is unknown. On the other hand, CHD risk assessment is often based on the same type of limited information we had available on each of our patients, so the models we present are perhaps more easily applicable to common clinical situations than models based on more detailed clinical data. Furthermore, a historical indicator of past exposure to high blood pressure or high cholesterol, as we had access to in this study, may actually be more useful as a predictor of CAC than treated blood pressure measured at one point in time. Another important limitation of this study is our lack of data on race/ethnicity – our results may not apply to all ethnic groups. Finally, our data are limited in application to CAC scores measured by electron beam computed tomography with 3 mm slice thickness and the described protocol. While CAC scores measured by the latest spiral computed tomography scanners appear to be similar to those generated by electron beam computed tomography [ 28 ], we cannot guarantee that our results apply to such scores. Our models should be applied to other similar cohorts for validation, and also applied in cohorts that include different racial/ethnic groups and different ways of measuring the CAC score before being used in these clinical situations. Conclusions The Clinical Research Roundtable at the Institute of Medicine has identified translation of clinical research findings into improvements in medical care as the "next scientific frontier" [ 29 ]. While our analysis has some limitations, it provides methodology that will directly assist in the translation of research into practice. Our models, once validated, can be used directly by patients and clinicians to decide when it might be useful to order this potentially expensive test, and what to do with the results. Competing interests MP has received speaking and consulting fees from Bayer. Authors' contributions MJP conceived the idea for the study, performed the analysis and drafted the manuscript. JAT and MP helped design and interpret the analysis. CM provided statistical guidance and interpretation. TQC recruited the patients and collected the data. WSB provided senior guidance in all aspects. All authors reviewed and commented on multiple drafts of the manuscript and approved the final draft. Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 This spreadsheet is used for combining information from conventional risk factors and the coronary artery calcium score to estimate coronary heart disease risk in an individual patient. Step 1: Enter your patient's clinical information (the red numbers). Step 2: Choose an assumption about the coronary artery calcium score relative risks (optimistic or conservative). Step 3: Find the following results: 1) "Pre-test" 10-year risk of coronary heart disease (CHD) based on Framingham equations; 2) The probability of having a coronary artery calcium (CAC) score that falls within 4 standard CAC score categories; and 3) The "post-test" 10-year risk of CHD for each CAC score category. Step 4: Use the results to interpret a CAC score, or to decide whether or not to order a coronary artery calcium scan. If a score that would change your management is unlikely to occur, it may not be worth the money. Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC515311.xml
516793	Paradoxes of Difference	A new play tackles the politics around race and gender in science	Can “race” be gleaned from our genes? Concern over the emerging trajectories of genetics research has led to ongoing debates about how to characterize and interpret genetic variation. Despite the mantra that humans share 99.9% of our genetic makeup, there is increasing interest in identifying the relatively small percentage difference that distinguishes individuals. Therein lies the paradox: if we are all the same, why do we continue to search for the ways in which we differ from one another? We can take an essential first step toward addressing this paradox by acknowledging the often conflicting stakes for individuals and groups in debates that center around genes, science, and race. Whose genes will be studied? For what purpose? And who has the authority to decide? These stakes are laid bare by Cassandra Medley in her play “Relativity,” which ran at the Magic Theater in San Francisco this spring, in a production directed by Edris Cooper-Anifowoshe. The play focuses on the inner conflict of a young scientist, an African American woman who tries unsuccessfully to straddle the opposing world views of her profession and her family. In “Relativity,” Kalima Davis is a postdoctoral fellow in a prestigious stem cell laboratory on the East Coast. Her impeccable academic pedigree distinguishes Kalima as a rising researcher, and she is one of the few women and African American scientists working in the field. Kalima is also the devoted daughter of the charismatic psychotherapist Claire Reid, who directs the “Leon Davis Foundation,” which Kalima's late father dedicated to the belief that neuropeptide melanin, found in “people of color,” enhances intelligence, athleticism, and emotional sensitivity. This theory also points to the lack of melanin among lighter skinned individuals as a cause of “white racism.” Kalima, who has inherited co-directorship of her father's foundation, is asked by her mother to offer “scientific proof” of the melanin theories and to discount the assertion that all groups are genetically similar. At issue in “Relativity” is the struggle over what constitutes a valid belief. Reid suggests to Kalima that “science is not the sole province of what the ‘West’ defines it to be,” and refers to Chinese acupuncture, Hindu Chakra, and tribal African shamanism as examples of legitimate “sciences.” But the power of Western science to trump other interpretations of lived experience has become all too clear in the genomic era. Truth is excavated from the human body, where genes emerge as the iconographic oracles of our past, present, and future. As Kalima recounts, “We can't get around it. DNA is fact.” Nonetheless, Kalima naively attempts to find a way to retain these opposing epistemologies. The futility of this dual position is made apparent by the arrival of Iris Preston, an African American senior scientist who has taken over as the new head of Kalima's lab. Preston, a highly vocal critic of Claire Reid and melanin theories, serves as a formidable apostle of the scientific method. Shortly after her arrival, Preston convenes members of her new lab to filter those seeking financial and other derivative rewards from the truly devoted, who are motivated solely by their “sense of wonder and amazement” and their desire to “cultivate what Einstein referred to as ‘holy curiosity.’” She makes plain that science, like all ideologies, demands consummate faith and unwavering piety. Kalima's struggle to claim her “true lineage” is much more than a simple choice between her biological mother (Reid) and her intellectual mentor (Preston). Her plight forces her to explore the meaning of justice. The contrast between melanin theories and genomic research, initially stark, blurs as it becomes increasingly clear that both Reid and Preston seek to use their “science” to redress race-related disparities. Citing a history of racism, including the historically well-documented Tuskegee Syphilis Study, Reid asks rhetorically whether genetics research will result in a “genetically modified” white upper class and a lower, dark-skinned “natural birth class”? At stake are issues of power and trust, and the question of whether new genetic technologies will close the gaps between groups or make them wider. Some postulate that the “new genetics” will render conventional notions of race obsolete. However, it is doubtful that such a color-blind utopia will be won through the sequencing of genes without a serious engagement with the differences that lie outside cell walls. A social infrastructure of inequality that mediates race through nutritional deficits, exposure to pollutants, and the use of the emergency room as the sole venue for healthcare will not be rehabilitated through gene therapy or pharmacogenomics. Preston seeks to use her stature in science to focus on the inequities within the academy. Encouraging Kalima to appear with her on a television program about stem-cell research, Preston urges her to imagine the milestone of “not just one, but two black women scientists, holding forth among the usual cadre of white males.” As one of the rare, highly scrutinized, “minority” scientists, Kalima embodies a “double jeopardy.” She must prove that she is worthy of her position—that she is as good, if not better, than her “white” peers—yet she must always be “remembering from whence she came.” When Dan, a white colleague in Kalima's lab who is also her boyfriend, hears that Preston has chosen Kalima to appear with her on television, he jealously accuses her of benefiting from preferential treatment. The play's depiction of this assertion of “reverse racism” questions the legitimate use of race in evaluating promotion and achievement in science. What is apparent is that race, while an ever-present subject, is often presented as the antithesis of conventional notions of a color-blind meritocracy. Left to linger is the critical question: does merely identifying differences among groups constitute an act of racism? Can statements of racial differences be neutral, unfettered by a relative hierarchy? Medley throws into stark relief the politics around race and gender in science. She illuminates how an individual's dilemma transcends the private realm; personal decisions are never truly “personal,” but are inherently public because they always have wider social repercussions. In the fateful confrontation between Kalima and her mother, Reid challenges her daughter to “grow up” and to risk her disapproval. A similar challenge can be issued to those who have inherited the “new genetics”—to vanquish the continuing paradox of reciting a “mantra of sameness” all the while searching for meaningful differences. In “Relativity,” Medley serves us a cautionary tale of the costs of our chronic ambivalence about the critical issues of race and justice in science. A good first step is to recognize that the search for meaning in human difference is inseparable from the struggle over the moral order in which we live. To underestimate the power of science to define our social agenda is to lose an opportunity to determine our future course. We only need to look to history as our proof.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC516793.xml
516787	The Menopause Rating Scale (MRS) scale: A methodological review	Background This paper compiles data from different sources to get a first comprehensive picture of psychometric and other methodological characteristics of the Menopause Rating Scale (MRS) scale. The scale was designed and standardized as a self-administered scale to (a) to assess symptoms/complaints of aging women under different conditions, (b) to evaluate the severity of symptoms over time, and (c) to measure changes pre- and postmenopause replacement therapy. The scale became widespread used (available in 10 languages). Method A large multinational survey (9 countries in 4 continents) from 2001/ 2002 is the basis for in depth analyses on reliability and validity of the MRS. Additional small convenience samples were used to get first impressions about test-retest reliability. The data were centrally analyzed. Data from a postmarketing HRT study were used to estimate discriminative validity. Results Reliability measures (consistency and test-retest stability) were found to be good across countries, although the sample size for test-retest reliability was small. Validity: The internal structure of the MRS across countries was astonishingly similar to conclude that the scale really measures the same phenomenon in symptomatic women. The sub-scores and total score correlations were high (0.7–0.9) but lower among the sub-scales (0.5–0.7). This however suggests that the subscales are not fully independent. Norm values from different populations were presented showing that a direct comparison between Europe and North America is possible, but caution recommended with comparisons of data from Latin America and Indonesia. But this will not affect intra-individual comparisons within clinical trials. The comparison with the Kupperman Index showed sufficiently good correlations, illustrating an adept criterion-oriented validity. The same is true for the comparison with the generic quality-of-life scale SF-36 where also a sufficiently close association has been shown. Conclusion The currently available methodological evidence points towards a high quality of the MRS scale to measure and to compare HRQoL of aging women in different regions and over time, it suggests a high reliability and high validity as far as the process of construct validation could be completed yet.	Background The interest of clinical research in aging women and males increased in recent years and thereby the interest to measure health-related quality of life and symptoms. Women, as do men, experience an age-related decline of physical and mental capacity. They observe symptoms such as periodic sweating or hot flushes, impaired memory, lack of concentration, nervousness, depression, insomnia, and bone – joint complaints. The Menopause Rating Scale (MRS) is a health-related quality of life scale (HRQoL) and was developed in response to the lack of standardized scales to measure the severity of aging-symptoms and their impact on the HRQoL in the early 1990s. Actually, the first version of the MRS was to be filled out by the treating physician but methodological critics lead to a new scale which can easily be completed by women, not by their physician [ 1 , 2 ]. The validation of the MRS started some years ago [ 2 - 6 ] aiming at establishing an instrument to measure HRQoL that can easily be completed. The aims of the MRS were (1) to enable comparisons of the symptoms of aging between groups of women under different conditions, (2) to compare severity of symptoms over time, and (3) to measure changes pre- and post-treatment [ 4 - 6 ]. The MRS was formally standardized according to psychometric rules and initially published in German [ 2 ]. During the standardization of this instrument, three independent dimensions were identified explaining 59% of the total variance (factor analysis): psychological, somato-vegetative, and urogenital sub-scale. The MRS consists of a list of 11 items (symptoms or complaints). Each of the eleven symptoms contained in the scale can get 0 (no complaints) or up to 4 scoring points (severe symptoms) depending on the severity of the complaints perceived by the women completing the scale (an appropriate box is to be ticked). The scoring scheme is simple, i.e. the score increases point by point with increasing severity of subjectively perceived symptoms in each of the 11 items (severity 0 [no complaints]...4 scoring points [very severe symptoms]). The respondent provides her personal perception by checking one of 5 possible boxes of "severity" for each of the items. This can be seen in the questionnaires in the additional files linked to this publication. The composite scores for each of the dimensions (sub-scales) is based on adding up the scores of each item of the respective dimensions. The composite score (total score) is the sum of the dimension scores. The three dimensions, their corresponding questions and the evaluation are detailed and summarized in an attached file linked to this publication [see Additional file 1]. The MRS scale became internationally well accepted. The first translation was into English [ 7 ]. Other translations followed [ 8 ], i.e. taking international methodological recommendations [ 9 , 10 ] into consideration. Currently, the following versions are available: Brazilian, English, French, German, Indonesian, Italian, Mexican/Argentine, Spanish, Swedish, and Turkish language. These versions are available in a published form, and can be downloaded in PDF-format from the internet (see reference 8 and ). Like in other QoL scales, it is a challenge to satisfy the demands of a clinical utility and outcomes sensitivity, and this in addition to the conventional psychometric requirements of test reliability and validity. The aim of this paper is to present additional psychometric data to discuss the methodologically relevant characteristics of the MRS scale. Methods The development of the scale, instrument characteristics (item selection, scaling), and norms and standardized scores have been published elsewhere [ 2 - 5 ]. This applies also for a few data that have been published on test-retest stability and criterion-dependent validity [ 3 , 6 ]. During the last two years a number of smaller and larger investigations were made from different groups to further check methodological features of the scale. We performed recently a large, multinational survey to represent the situation across nine countries and cultures using existing and for the respective countries representative panels between November 2001 and February 2002 to get information about knowledge, attitudes and behaviour related to hormonal treatment in women aged 40–70 years: Europe (Germany, France, Spain, Sweden), North America (USA), Latin America (Mexico, Argentine, Brazil), and as example for Asia – Indonesia. Study participants were accrued as a random sample of females aged 40 to 70 years from existing population panels. The sample size in each of the countries was about 1000 females aged 40–70 years, with exception of USA (n = 1500). The participation rates ranged between 46 and 94% across countries. The demographic details of the sample are: On average, about tertiles of the respondents were under 50 years, between 50–59, and over 60 years old in most of the countries, however, about 50% were less than 50 years in Indonesia and in Brazil. The majority of respondents reported a Christian religion in Europe (range: 74% (Germany) to 96% (Spain), 85% in USA, and in Latin America (range: 95% (Argentina) to 97% (Mexico). The use of the MRS was part of this survey, i.e. multinational data became available to reconsider methodological issues more thoroughly such as internal structure of the scale, reliability (internal consistency alpha), and reference values for different population. For the purposes of reliability assessment we performed a few preliminary studies with a test-retest approach. These small, descriptive studies of community samples of women aged 40–70 were done in summer and fall 2002 by local collaborators in the respective countries, but they were done separately and independent from the main study. These studies were done just for orientation with convenience samples – not representative for the respective population. There is only one intervention study (before and after hormonal treatment) available to our knowledge. This study has been published [ 6 ] but not with regard to methodologically relevant results of the MRS. These data will be published soon. With these data available, we were able to scrutinize many methodological characteristics of the MRS scale to review most fundamental psychometric characteristics as well validity parameters. Results and Discussion Reliability The assessment of scientific measurements depends first of all on the evidence of replicability (consistency) and test-retest reliability. In contrast to systematic and random variation, reliability gives an estimate of method-related measurement error which should be low not to hide or dilute intended systematic changes – due to treatment for example. Table 1 show the internal consistency measured with Cronbach's Alpha. The consistency coefficients range between 0.6 and 0.9 across countries for the total score as well the scores in the three domains. This is indicative for a very acceptable consistency of the MRS scale in our opinion. Moreover, there is no evidence that the scale works different in so many different countries in four continents. Table 1 Internal consistency coefficients (alpha) for the MRS scale across countries: total score and scores for the psychological, somatic, and urogenital domain. Data from the Nine-Country Study International Europe North America Latin America Asia Total Overall Sweden Germany France Spain USA Overall Mexico Brazil Argentina Indonesia N 9907 4465 1490 1050 941 984 1440 3002 1002 1000 1000 1000 Total score 0.83 0.86 0.85 0.84 0.86 0.86 0.88 0.86 0.87 0.86 0.83 0.84 Psychologic ascore 0.87 0.88 0.88 0.86 0.89 0.86 0.90 0.85 0.86 0.87 0.81 0.79 Somatic score 0.66 0.64 0.65 0.64 0.64 0.61 0.70 0.66 0.65 0.69 0.64 0.69 Urogenital score 0.65 0.65 0.64 0.63 0.67 0.67 0.70 0.60 0.62 0.55 0.62 0.65 The test-retest correlation coefficients (Pearson's correlation) support the suggestion of a good temporal stability of the total scale and its three sub-scales (Table 2 ), although most of the assessments across countries are based on very small numbers and convenience samples not claiming to be representative for the respective population. The intention of these pilot studies was to get a preliminary idea about retest stability. Larger sample sizes are required to permit final conclusions for individual countries / languages. Table 2 Test-retest coefficients (Pearson's correlation coefficient "r") for the MRS scale across countries: total score and scores for the psychological, somatic, and sexual sub-scale. All Europe Latin America Asia Overall Overall Germany UK France Spain Portugal Sweden Turkey Overall Argentina Brazil Overall Indonesia N 349 259 73 30 36 30 30 30 30 60 30 30 30 30 Total score 0,90 0,92 0,82 0,80 0,89 0,93 0,96 0,90 0,90 0,81 0,78 0,82 0,84 0,84 Psychological score 0,84 0,87 0,76 0,72 0,88 0,92 0,91 0,79 0,82 0,72 0,66 0,76 0,71 0,71 Somatic score 0,89 0,90 0,80 0,85 0,82 0,88 0,97 0,95 0,93 0,85 0,86 0,85 0,81 0,81 Urogen. score 0,86 0,89 0,77 0,82 0,94 0,98 0,95 0,87 0,89 0,73 0,67 0,74 0,50 0,50 The test-retest coefficients of the total score range between 0.8 and 0.96 across Europe, North and Latin America, and Asia. When it comes to the subscales with much fewer items, the variation increased and some of the coefficients went down to 0.5 (urogenital domain in Indonesia). Altogether, the test-retest stability over a time period of two weeks aggregated at the international level supports the notion of a very acceptable test-retest reliability of the total scale and their three sub-scales. Although there is an impressive set of information currently available concerning the reliability of the MRS scale, there are also limitations: Small sample sizes prevent a final conclusion regarding test-retest reliability in some of the languages the scale has been translated in. Validity Similar to reliability which assesses the consistency of measurement, the validity estimates if a scale measures what it intends to measure. But whereas reliability can be determined straight forward with very few indicators, the validity is almost always a continuous process (construct validation). It is a process of accumulating evidence for a valid measurement of what is purposed. Therefore, the currently available data are already fairly comprehensive and do pave the way for a focussed and continuing validation process. Internal structure of the MRS across countries The first step of validation is usually to multivariately demonstrate a similar internal structure ("dimensions") of a given scale through factor analysis. The first factorial analysis in 1996 was applied to identify the dimensions of the scale. Three dimensions of symptoms/complaints were identified [ 2 ]: a psychological, a somato-vegetative, and a urogenital factor that explained 58.8% of the total variance. The recent large, multinational survey in nine countries of four continents provided data to compare with the initial standardisation sample of the MRS. The question was: Is the internal structure of the MRS results comparable among different countries or cultures. Astonishingly similar factor loadings of the 11 items of the 3 domains of the MRS were observed (Table 3 ). The same applies for the individual countries of the respective regions (data not shown). Although the prevalence of menopausal symptoms may slightly differ among regions/cultures (see later), the structure of complaints/symptoms seems to be pretty much the same. It suggests that the scale measures constantly the same phenomenon which speaks in favour of the translation/cultural adaptation of the scale. Table 3 Internal structure of the MRS scale across countries in 9 countries of four continents (2002) compared with the initial analysis of a German sample in 1996. Principal component analysis, Varimax rotation. Complaints (item number in MRS), numbers, and country groups. Only factor loadings ≥ 0.5 are shown. Complaints (item number) N Flushes(1) Heart(2) Sleep(3) Joints(11) Depress(4) Irritabil(5) Anxiety(6) Exhaust(7) Sexual(8) Bladder(9) Dryness(10) Germany, 1996 479 somatic 0.8 0.7 0.6 0.5 psychologic 0.8 0.7 0.8 0.6 urogenital 0.7 0.8 0.8 All countries, 2002 10297 somatic 0.7 0.8 0.5 0.5 psychologic 0.8 0.8 0.8 0.7 urogenital 0.7 0.6 0.8 Europe, 2002 4791 somatic 0.7 0.7 0.6 0.6 psychologic 0.8 0.9 0.8 0.6 urogenital 0.7 0.6 0.8 N.-America (USA) 1500 somatic 0.7 0.7 0.7 0.5 psychologic 0.8 0.9 0.8 0.6 urogenital 0.5 0.8 0.6 0.8 Latin America, 2002 3006 somatic 0.5 0.9 0.5 0.4 psychologic 0.5 0.8 0.8 0.8 0.7 urogenital 0.6 0.7 0.8 Asia (Indonesia) 1000 somatic 0.9 0.6 0.5 0.5 psychologic 0.8 0.8 0.8 0.5 urogenital 0.8 0.5 0.9 However, there are indications that the domains could be somewhat overlapping and not as entirely independent as the statistical model suggests: Muscle or joint problems got a loading of 0.5 in the somatic and urogenital domain (USA), and sleep disturbances both 0.5 in somatic and psychological domain (Latin America). These two items had similar problems in Spain, Mexico, and Brazil but not in other countries (data not presented in table 3 ). In clinical studies intra-individual comparisons over time (before/after treatment) will be the main criterion which might not be affected by potential slight differences in the patient reported outcome structure. Therefore the general agreement in the internal structure of the MRS scale across country groups, even accepting the possibility of slight differences in two items (cf. Table 3 ), suggests that the scale can very well be used in clinical studies – even including different countries. Sub-scores and total score correlations The relations among the sub-scales and the aggregate total scale are patterns that are important in the methodological assessment of a scale. In an ideal world, the correlations between subscales (supposed to be independent due to the statistical model) would be closer to 0 than the correlations with the construct of the aggregate total score to which all sub-scales should significantly contribute. But that is theory; Table 4 shows only somewhat lower correlations among sub-scales (0.4–0.7) as compared with correlation of sub-scales with the total score (0.7–0.9). This is less different than one would have wished. It suggests that the sub-scales are not as independent from each other as one would expect them to be – based on a factorial analysis with orthogonal factors. The situation was similar in the four regions listed in Table 4 and in the individual countries belonging to these regions. It is important to realize how similar these correlation coefficients are among countries/aggregates. This is suggestive of pretty similar features of the MRS scale across the countries of this review. Table 4 Domain score – total score correlations of the MRS scale across four country groups. Community samples. Domains Psychological score Somatic score Urogenital score Europe (n = 4246) Total score 0.9 0.9 0.7 Psychological score -- 0.6 0.5 Somatic score -- -- 0.5 North America (n = 1376) Total score 0.9 0.9 0.8 Psychological score -- 0.7 0.5 Somatic score -- -- 0.5 Latin America (n = 3001) Total score 0.9 0.9 0.7 Psychological score -- 0.7 0.5 Somatic score -- -- 0.5 Asia (n = 1000) Total score 0.9 0.9 0.7 Psychological score -- 0.7 0.4 Somatic score -- -- 0.4 Compatibility of MRS reference values for different population There are different categories of severity of complaints or problems with QoL. For the comparison of these categories across countries or cultures it is important to understand the prevalence of complaints. Currently, there is only one table with reference values and definitions published – for the German population from the initial standardization of the MRS [ 2 ]. Are these reference values applicable for other countries/cultures? The data of our large multicultural survey permitted such comparisons. The mean values (SD) of the MRS total score and the three domains are depicted in Table 5 . The mean values of the total score and the 3 domain scores are not statistically significantly different between Europe and North America. Thus, there is no evidence yet to exclude direct comparisons of MRS values between these regions. Table 5 Mean values and standard deviation of MRS total score and 3 domains. Results from a large, multinational survey (see methods) Total score Psychological Score Somato-vegetative Score Urogenital Score n Mean (SD) n Mean (SD) n Mean (SD) n Mean (SD) Europe 4246 8.8 (7.1) 4453 3.4 (3.4) 4465 3.6 (2.9) 4465 1.9 (2.2) N.-America (USA) 1376 9.1 (7.6) 1426 3.4 (3.5) 1440 3.8 (3.1) 1437 2.0 (2.3) Lat.-America 3001 10.4 (8.8) 3002 4.9 (4.5) 3006 4.1 (3.6) 3005 1.4 (2.2) Asia 1000 7.2 (6.0) 1000 2.9 (2.9) 1000 3.3 (2.7) 1000 1.0 (1.6) However, the total, psychological and somatic scores were systematically higher in Latin America, and systematically (significantly) lower in Asia (Indonesia) than in Europe/North America. The urogenital scores were significantly lower in Latin America/Indonesia than in Europe/US. Obviously the perception of the prevalent symptoms depends on cultural factors – or the symptoms show real differences in prevalence. Thus, direct comparisons of MRS scores between Europe/North America on the one side and regions in Latin America and Asia cannot are not recommended. This does not effect intra-individual comparisons (e.g., pre/ post therapy) within these countries and it may also very little affect the comparison of relative changes (pre/post treatment) among different countries. The latter is a hypothesis and needs further evidence form research/experience. Similar mean values could still have a different distribution across the proposed four categories of "severity of complaints": no/little symptoms, mild, moderate, and severe complaints, i.e. for the total scale and the three domains. The prevalence of these categories across the four regions studied is seen in Table 6 . The comparison of the prevalence (and 95% confidence interval) showed that the above discussed differences between Europe/US and Latin-America or Indonesia very much depend on the severity of complaints. Whereas the differences in the psychological domain were less impressive, the dissimilarity was most pronounced in the urogenital domain and less also in the somatic domain. Whether this is due to different perception of identical symptoms (differences in the appearance of symptoms or both) remains a speculation. This however needs to be considered when direct comparisons among different cultures are intended. The prevalence of different "degrees of severity" of menopausal symptoms measured with the MRS was found to be almost identical in the aggregate of Europe and North America. Table 6 Comparison of "degree of severity" of the MRS and its domains. Prevalence in percent (%) and 95% confidence interval (in parenthesis) in the population sample studied in the respective regions (see methods) Europe North America Latin America Asia Total score No, little (0–4) 28.8 (+/-1.3) 28.0 (+/-2.3) 31.0 (+/-1.7) 40.2 (+/-3.0) Mild (5–8) 21.9 (+/- 1.2) 23.9 (+/-2.2) 20.2 (+/-1.4) 27.5 (+/-2.8) Moderate (9–16) 25.1 (+/-1.2) 25.7 (+/-2.2) 26.2 (+/-1.6) 22.8 (+/-2.6) Severe (17+) 24.3 (+/- 1.2) 22.5 (+/-2.1) 22.7 (+/-1.5) 9.5 (+/-1.8) Psychological domain No, little (0–1) 35.4 (+/-1.4) 36.8 (+/-2.4) 36.8 (+/-1.6) 41.3 (+/-3.1) Mild (2–3) 21.8 (+/-1.2) 21.9 (+/-2.1) 21.9 (+/-1.4) 25.4 (+/-2.7) Moderate (4–6) 19.5 (+/-1.1) 18.7 (+/-2.0) 18.7 (+/-1.4) 21.3 (+/-2.6) Severe (7+) 23.4 (+/-1.2) 22.5 (+/-2.1) 22.5 (+/-1.7) 12.0 (+/-2.0) Somato-vegetative domain No, little (0–2) 39.5 (+/-1.4) 37.9 (+/-2.4) 42.1 (+/-1.8) 46.8 (+/-3.1) Mild (3–4) 22.6 (+/-1.2) 25.6 (+/-2.2) 19.4 (+/-1.4) 27.0 (+/-2.8) Moderate (5–8) 24.2 (+/-1.2) 24.3 (+/-2.2) 25.6 (+/-1.6) 20.8 (+/-2.5) Severe (9+) 13.7 (+/-1.0) 12.1 (+/-1.7) 12.9 (+/-1.2) 5.4 (+/-1.4) Urogenital domain No, little (0) 34.3 (+/-1.3) 33.4 (+/-2.4) 28.2 (+/-1.8) 55.6 (+/-3.1) Mild (1) 17.2 (+/-1.1) 17.0 (+/-2.0) 18.6 (+/-1.1) 18.6 (+/-2.4) Moderate (2–3) 23.0 (+/- 1.2) 24.2 (+/-2.2) 21.8 (+/-1.3) 17.0 (+/-2.3) Severe (4+) 25.6 (+/-1.2) 25.4 (+/-2.2) 31.4 (+/-1.3) 8.8 (+/-1.8) Criterion-oriented validity: correlation with other scales In fact, the comparison with other scales of similar purpose is important. It is known from other quality of life scales that comparisons with scales with similar purposes are much more important than comparisons with so-called objective parameters such as exercise tests, physiological or chemical parameters – in our case with hormones. Health related quality of life should be validated against quality of life measured with other generic QoL scales (e.g., SF-36), and against specific instruments to measure symptoms in aging women (e.g. Kupperman index). These data were published elsewhere [ 6 , 11 ] but will be briefly summarized in the context of this review. Kupperman Index Although the Kupperman Index was not validated according to psychometric standards it is still in use in the medical practice to monitor menopausal symptoms. Therefore a comparison with the fully standardized MRS seems to be reasonable. If one divides the distribution of both scales into quartiles and compares the frequencies, both instruments were found to be closely associated: Kendall's tau-b coefficient 0.75 (95% CI 0.71–0.80) [ 6 ]. Similar was the Pearson correlation coefficient with r= 0.91(95% CI 0.89–0.93). The two scales can be regarded as measuring the same phenomena. However, some methodological problems of the Kupperman Index were identified in this comparison (see [ 6 ] for details). Generic QoL Scale SF-36 Two sub-scales of the multi-domain quality of life scale SF-36 was compared with the MRS: the somatic sum score (with somatic domain of MRS) and the psychologic sub-scales of both instruments. Both somatic domains were sufficiently well and significantly associated: Kendall's tau-b = 0.43 (95% CI 0.52–0.35); Pearson correlation coefficient r= 0.48 (95% CI 0.58–0.37). That means, the higher the score in the somatic dimension of the MRS, the lower the quality of life according to the somatic sum-score of the SF-36 [ 6 , 11 ]. Similar was the results of the comparison of the psychological scores of both instruments: Kendall's tau-b = 0.49 (95% CI 0.56–0.41); Pearson correlation coefficient r= 0.73 (95% CI 0.81–0.65). Discriminative validity i.e., the ability of the scale to accurately measure treatment effects and to predict the clinically based assessment of physicians, was not analysed so far. At present, there is one post-marketing study that can be used to preliminary assess discriminative validity. The results will be published soon elsewhere. To this end, many clinicians understand the term "validity" and mean high utility for clinical work or research. Conclusions The MRS scale is a standardized HRQoL scale with good psychometric characteristics. The use in many countries offered the possibility to compare the test characteristics across countries. Reliability measures (consistency and test-retest stability) were found to be good in all countries where data were obtained – however, some samples were very small and therefore considered as preliminary information. The validity was measured in its various forms: The internal structure of the MRS across countries was sufficiently similar to conclude that the scale really measures the same phenomenon in women with complaints. The sub-sores and total score correlations showed high coefficients with the total score and less among the sub-scales. This however indicates that the subscales are not fully independent in practice. Comparisons of reference values from different populations showed that the MRS scores can easily be compared between Europe and North America/US. Direct comparisons between Europe/North America and Latin American countries and Asia (Indonesia) should be considered with caution because the severity of reported symptoms seems to differ. The reasons are not clear, further research is needed. The comparison with other scales for menopausal symptoms (Kupperman Index) showed a sufficiently close association and correlation coefficients, i.e. illustrating a good criterion-oriented validity. The same is true for the comparison with the generic QoL scale SF-36 where also high correlation coefficients have been shown. Thus, the currently available methodological evidence points towards a high quality of the MRS scale to measure and to compare HRQoL of aging males over time or intervention. It suggests a high reliability and high validity as far as the process of construct validation could be completed. Authors' contributions KH: responsible for drafting the manuscript and running analyses. AR: responsible for designing and overseeing the multinational survey (2001/2002), contributed to writing and revising of the paper. PP: co-ordination of the field work of the multinational survey, setting up the initial database, and contributed to writing of the paper, responsibility in developing/validating the MRS scale. HPGS: Major responsibility in developing the MRS scale, contributed to writing/revision of the manuscript. FS: Provided data of a clinical study, contributed to writing of the manuscript. LAJH: responsible for the collection and evaluation of the data, and involved in writing/revising the paper. DMT: responsible for checking the integrated database, responsible for several analyses regarding validity, and contributed to writing of the paper.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC516787.xml
544943	Epidemiology in conflict – A call to arms	In this first special theme issue, Emerging Themes in Epidemiology publishes a collection of articles on the theme of Epidemiology in conflict . Violent conflict is an issue of great sensitivity within public health, but more structured research and reasoned discussion will allow us to better mitigate the public health impacts of war, and place the public health community in a more informed position in discussions about possible interventions in future conflicts.	"And there went out another horse that was red: and power was given to him that sat thereon to take peace from the earth, and that they should kill one another: and there was given unto him a great sword." – Revelations 6:4 Perhaps more than any other previous conflict, the recent war in Iraq has stirred the public health community, with numerous pages in general medical and scientific journals devoted to contributions condemning the basis of the war, condemning those condemning the war and condemning editors for delving into politics by publishing these condemnations [ 1 ]. For all this self-castigation, however, the public health message was notably absent from mainstream media and political discussion. The issue of war is particularly sensitive in the field of public health, some might even say taboo. It is not seen as appropriate to openly denounce war. Perhaps we feel that becoming involved in what is largely perceived to be a political issue constitutes a threat to that most precious of our ideals, that of objectivity. Or maybe we are uncomfortable with the thought of being associated with military activities. Specifically, the opportunistic use of military language in public health issues has not been particularly welcome. Political calls for war on the societal problems of drugs and cancer have been accused of victimizing individuals and setting implausible goals for their prevention and control [ 2 - 6 ]. Ironically, we have yet to publicly declare war on the greatest of societal ills, war itself. The dilemma facing a scientific editor attempting to circumnavigate the turbulent waters of politics is eloquently summarized in a recent British Medical Journal editorial: to do nothing is as much a political decision as to challenge an issue head-on [ 7 ]. Given such an environment, readers may find it surprising that we should devote our first special issue to the subject of Epidemiology in conflict . Our reason for doing so is simple: regardless of the political context, war is bad for your health. The politicization of war within public health is unfortunate; re-framing public health questions within a political context prevents us from conducting informed discussion and finding rational solutions to them. Here, we draw a parallel with the early efforts to communicate the link between smoking and lung cancer. In his reflections on the subject [ 8 ], Ernst Wynder describes the opposition he encountered when trying to relay the findings from the first studies establishing the epidemiologic link. Opposition came not just from the conflicted interests of governments, the media and tobacco companies, but also from within the public health profession, which was at the time dominated by physicians, many of whom were smokers, were unused to the interpretation of epidemiologic data or simply thought the association to be implausible. Smoking was (and continues to be) a highly political issue, yet few would argue in retrospect that Doll, Hill, Wynder and the early proponents of the smoking-lung cancer association should, in the face of incontrovertible evidence, have done anything else than to publicize this link. Here the parallel ends, however. Given adequate knowledge about the risks to their health, an individual may freely choose to smoke and accept responsibility for any ensuing personal health consequences. Equally, they may choose to avoid these risks by not smoking altogether (although we stipulate that the effect of passive smoking is contentious here). An individual cannot choose not to go to war or have war inflicted upon them. Such decisions are carried out by governments, insurgents or other groups, many of which are not accountable to individuals. In this scenario, the role of academics and professionals, as well as professional associations and non-governmental organizations in informing governments and the public and raising issues that affect society as a whole becomes even more important. The duty of health professionals as advocates for public health is emphasized by Wynder and is equally applicable today and perhaps even more so to the issue of war: " ...the consensus of opinion among experts is not sufficient to create action unless such consensus is translated into preventive or control measures.... Scientists and physicians cannot be content with discoveries until their beneficial or protective outcome for the population has been fully realized. This means that the members of the scientific and medical community must become more proactive in public health matters [ 8 ]." The issue of consensus is important, but difficult. Some will argue that war is inherently bad for public health and should always be opposed. Others will consider some wars justifiable if they address gross injustice or human rights abuses. And yet others, in line with the Geneva Conventions, will see wars, just or not, as inevitable and will want to focus on mitigation of human suffering. Regardless of one's viewpoint, each of these positions needs to be supported by an evidence base with answers to the questions of when, how and why war is bad for public health, as well as how the adverse health effects of war may be prevented. Therein lies the greatest challenge for epidemiologists. Conflict situations deny us access to data and dissolve the health infrastructures on which we rely for the collection of epidemiologic information. In the face of such challenges, the authors of the articles in this special issue deal with a broad range of issues of great relevance to epidemiologists. Mock et al . [ 9 ] argue that the interface of HIV/AIDS and conflict is more complex than is usually assumed. It is often said that war exacerbates the HIV epidemic, but the ecologic evidence suggests that this is not always the case. The authors examine the complexities of this issue and analyze how conflict can both exacerbate and retard the spread of HIV. McDonnell et al . [ 10 ] evaluate the role of epidemiologists in conflict settings. Present barriers to effective engagement stem from the fact that epidemiologists do not receive training on issues pertinent to their operating in conflict-affected areas. Perhaps most important are appropriate communication skills to enable epidemiologists to present their message clearly as health related rather than political. Roberts and Hofmann [ 11 ] place the work of humanitarian agencies under the epidemiologist's gaze. All too often such agencies, with the best intentions, measure success in terms of process – how many meals were handed out, how many vaccinations were given? From a health perspective this is only part of the story. Did these actions really have a positive impact on health? The difficulties of collecting such information mean that humanitarian interventions rarely incorporate tangible, impact-driven outcomes as priorities. This article proposes a framework for assessing the impact of aid on health. These papers provide a foundation on which we hope authors will continue to build so that a comprehensive range of relevant topics on this subject may be compiled. There remain many issues to be addressed by the epidemiology community (see figure). Some of these issues involve challenges so great that we have perhaps not even begun to think about how we might start tackling them. We encourage authors to continue submitting articles on the theme of Epidemiology in conflict , so that we may be kept informed of developments in the field and promote the public health perspective in discussions about future conflicts. Figure Conceptual framework for Epidemiology in conflict We hope readers intending to take up these challenges will be informed, inspired and provoked into action. This is a call to arms, not against the barriers of physical inactivity and excessive caloric intake, the adaptability of infectious agents or the subtleties of gene-environment interactions, but against the sheer brutality of human beings killing each other. We encourage readers to research and discuss the humanitarian and public health consequences of this social disease. The knowledge gained will allow us to better mitigate the public health impacts of war, and place the public health community in a more informed position in discussions about possible interventions in future conflicts. The pen may yet prove to be mightier than the sword, but only as long as it keeps writing.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544943.xml
340956	World on Fire	Fires are increasing in severity and incidence around the globe and affect many different ecosystems; will the new generation of fire science tools help managers retain biodiversity?	From Bob Clark's snug office in Boise, Idaho, where he manages the United States government's Joint Fire Science Program (JFSP), he figures his computer provides fingertip reach to just about everybody who's anybody in wildfire research. This points to a primary need of nations worldwide in combating the scourge of recurrent wildfires: tools and technology suited to the job. It's no small order in places as economically, socially, and ecologically varied as, say, Brazil, South Africa, Australia, Indonesia, and the United States, which are among the countries where wildfire creates the greatest havoc. More than 750,000 acres (303,500 hectares) were burned in southern California alone during last year's wildfires. The 2000 season was one of the country's worst on record, destroying 8.4 million acres (3.4 million hectares), more than double the decade's 10-year average. Australia's summer months around the turn of 2002–2003 brought perhaps the worst drought in a century to the populous southeast and the biggest fire season for two decades. Mountain forests were extensively burned and more than 500 houses were lost. In 2002, Brazil suffered 217,000 wildfires, a number that is almost certainly too low because remote imaging cannot detect many fires under the forest canopy. In Indonesia, wildfires that burned for months during 1997–1998 were later estimated to have released the equivalent to 13%–40% of annual global carbon emissions from fossil fuels, inflicting smoke-related ailments on thousands. Where wildfire is concerned, the many differences between such countries can perhaps be pinned down to two essentials. The first is whether a blaze occurs in temperate or tropical forest, and the second is whether the nation is developed or developing. “The science can be rock solid, but it can only go so far before social, economic, and political pressures take over,” Clark says. “That's what a forest service manager's job is, picking the best option based on all those considerations.” Unfortunately, having science-based options that are applicable to local conditions is largely a luxury for developed countries. Managers there can choose to let a fire burn under hopefully contained conditions, a policy known in the United States as “wildland fire use.” They can set experimental crown fires to study their effects, as was done recently in"journal" Canada ( Figure 1 ). And they can take preemptive measures, such as reducing fuel in the forest to lower fire hazard. Figure 1 Northwest Crown Fire Experiment (Photograph used by permission of the USDA Forest Service.) The two main fuel-reduction methods are mechanical removal of combustible materials and controlled or “prescribed” burning ( Figure 2 ). During Bill Clinton's administration, prescribed burns were encouraged in protected areas, but thinning was allowed only for trees with trunks of nine inches (22.8 cm) in diameter or less. Under George W. Bush, prescribed burning remains a choice, but the United States Department of Agriculture's (USDA) Forest Service policy is much more focused on mechanical means. The argument runs that there's been too much concern about removing trees, when what counts most is the enhanced fire-resistance of the thinned habitat. Figure 2 Prescribed Burns in the Intermountain Region of the United States (Photograph used by permission of the USDA Forest Service.) Fire hazard reduction methods must be tailored to an understanding of fuel characteristics in a given area, says David Peterson of the Forest Service's Pacific Wildland Fire Sciences Laboratory in Seattle, Washington. “There's no uniform way of doing it, partly because, as scientists, we haven't given the management folks any quantitative guidelines.” Working with other ecologists, social scientists, and economists, he's currently producing just such guidelines for the dry interior forests of the Pacific Northwest. “One thing we don't want to do is take choices out of the hands of field managers working at the local level.” Forecasting Tools: Models and Simulations For those choices to be meaningful, managers need reliable information on the risk of wildfire outbreaks and on the future behavior of existing fires. This requires models and simulations that incorporate climatic conditions, particularly wind ( Figure 3 ). At the Forest Service's Fire Sciences Laboratory in Missoula, Montana, researchers have created a “gridded wind” tool based on the engineering discipline of computational fluid dynamics. The program maps wind speed and direction using a digital elevation model, which is a grid of elevation points every 30–100 feet (9–30.5 meters) over a terrain 10–40 square miles (25.9–103.6 square kilometers) in size. This map forms the floor of a box extending up to five miles (eight kilometers) high, which is subdivided into a million or more cubes. Wind flow from either real observations or estimates can be entered into the software, and the layer of cubes nearest the grid floor is used to create surface wind maps at resolutions of every 100 meters (109 yards) or less. In contrast, the usual resolution of weather forecasts is 12 kilometers (7.5 miles), down to 4 kilometers (2.5 miles) in some urban areas. Figure 3 Shaded Surface Images of Areas in Northwestern Montana That Suffered Wildfires during 2003 The winds are from the west at 20 mph/32 kph. The white lines represent fire perimeters. (Image used by permission of the USDA Forest Service, Fire Sciences Laboratory.) “Two or three years ago, we couldn't have done this simulation on a single-processor laptop,” says one of its developers, physical engineer Bret Butler. “It would have taken two or three days. Now we can do it in a matter of hours.” The ability of these maps to show varying wind flow in valleys, at midslope and on ridgetops, is just the beginning. The next step is to feed these data into models that predict wildfire spread. Butler and colleagues have coupled their gridded wind technology to a fire growth model and tested it against the actual spread of several wildfires, including in Southern California last summer. Maps of actual and predicted surface winds showed strong similarities, encouraging Butler to foresee an ideal scenario in which fire fighting teams enter wind flow data online or by telephone to a central base where gridded wind maps and fire growth simulations are generated within hours, before operational decisions are made. Yet he admits that challenges remain, including the current inability of fire behavior simulation to account for diurnal winds in addition to cold front-driven flow. In mountainous terrain, for example, winds often move up-canyon in the morning and down-canyon in the evening. Moreover, the effect of vegetation on wind is not yet included in such models. Those issues and others are being addressed by researchers working on improvements to the regional weather forecasts of so-called mesoscale models. At the Forest Service's Rocky Mountain Research Center in Fort Collins, Colorado, meteorologist Karl Zeller and colleagues are contributing calculations of biological processes to mesoscale weather models. Their algorithms not only can account for diurnal winds but can predict the effects on local weather when vegetation takes in carbon dioxide and releases water vapor. This process can produce different fluxes of carbon dioxide drawn into the canopy and water vapor coming out, depending largely on the type of vegetation and its canopy density. Zeller's group has analyzed current mesoscale forecasts in the Rocky Mountains and found that in the daytime, they often are too hot in the high country and too cold in the plains. Water vapor estimates are too low in the mountains and too high in the plains, which Zeller thinks is because the models feed off soil moisture estimates, not off vegetation. In coupling his team's new biophysical interface to gridded wind and mesoscale forecast models, Zeller says “point forecasts” are being developed that can focus on a prescribed burn area or even a single house. Wildfire and Species Diversity Fitting the appropriate mix of strategies to a given situation is an issue that has also received close attention in Australia. After the bushfires of 2002–2003, media commentators called for increased “hazard reduction burning” in national parks, prompting ecologists around the country to distribute a joint statement declaring that such a strategy would not further reduce bushfire risk, but would actually threaten biodiversity. Australian species are often well-adapted to fire, and researchers have learned that different fire regimes—meaning the type of fire, its intensity, severity, extent, season, and frequency—favor different species ( Box 1 ). In the southeast of Australia, prescribed burns of high frequency and low intensity can alter the habitat in ways that therefore threaten survival of numerous plant and animal species. “A generic problem or conundrum seems to be that species which do not prosper under relatively frequent fires can be found in most fire-prone environments,” notes Ross Bradstock, principal research scientist in the New South Wales Department of Environment and Conservation. He says it's very difficult to determine how human interventions in various habitats can foster the coexistence of species that have different fire regime requirements. Fire Suppression and Tropical Forests As tough as such questions are to answer in developed countries, they pale compared to the problems of tropical forest wildfire researchers and managers in developing countries. In these countries, a destructive cycle of human behavior begins with land-clearing and burning for farming, logging, mining, road-building, and other uses that open gaps in the rainforest's canopy cover. This lets in sunlight and air, reducing the forest's ability to smother fire by trapping moisture, and it encourages the growth of smaller, more fire-prone plants. The first wildfires that occur are bad, but successive ones can eventually transform tropical forest to scrub savanna ( Figure 4 ). Of course, the remaining forest is thereby broken into fragments that continue to suffer incursions at their edges, as the cycle continues. Figure 4 Forest Regeneration (A) Dense understory regeneration three years after a low-intensity fire. (B) The almost total loss of live, above-ground biomass six months after a forest burnt for the third time in 15 years. (Photographs by Jos Barlow; used by permission.) In a recent paper in Science, Michigan State University Amazon expert Mark Cochrane pointed out that prescribed burning is ineffective in tropical forests, because the collateral damage outweighs any benefits. Indeed, tools and technologies employed in temperate conditions can seldom be applied usefully to tropical forests without significant alterations. “One of the main issues in fire science is that the U.S. has no capacity to develop new tools,” charges Ernesto Alvarado, a research scientist at the University of Washington in Seattle. He's been working for several years with United States Forest Service and Brazilian scientists on field studies in Mata Grosso, the southernmost state of the Brazilian Amazon. He says that fire prediction simulations developed decades ago have not yet been replaced by ones that account for tropical wildfire extremes, including either large-scale crown fires or surface fires, which often reach only 10 centimeters (3.9 inches) in height and move slowly but can burn for weeks and kill many trees. Fire behavior models don't work for tropical surface fires because the physics are different from those in temperate forests, he explains. A slow wind generated from the unburned forest blows toward the fire, forcing the small flames to advance against, rather than with, the wind. Another difference is that the fuel is mostly leaf litter, not conifer needles or sticks. Alvarado and colleagues light experimental fires in clear-cuts to determine factors limiting ignition and spread. Such experimental work is rare in tropical forests, where observation and description still predominate. But the team also monitors surface wildfires, measuring fire length, spread, and heat release. “We're trying to find applications that people can use to control fires or to explain implications of fire policy,” he says. Most wildfires originate from deliberately set burns. For example, many farmers still clear land by the ancient method of slash-and-burn, in which forest is chopped, left to dry, and then burned. These farmers are now banned by Brazilian federal law from burning at the height of the dry season, mid-July to mid-September. They cut in May, but if the rains come early in September, they can't burn after the ban ends and must wait until the next season, with nowhere to grow their crops in the meantime. Alvarado thinks a more flexible burning schedule is a solution. The challenge is to pass on technological understanding to decision-makers. For example, even ranchers in Mata Grosso's economic elite usually haven't heard of fire management techniques, says Amazonian ecologist Carlos Peres at the University of East Anglia in the United Kingdom. Educational projects from nongovernmental organizations have helped to turn some farmers away from heavy reliance on slash-and-burn techniques, but fire suppression information remains to be distributed on the frontiers. “What we really need are very large areas of primary forest that effectively serve as fire breaks,” he says. Conservation plans have been made by the federal government in collaboration with international agencies, but implementation remains a question, particularly given the high level of economic pressure from multinational resource developers eager to enter the Amazon. Major roads through the jungle are also on the drawing board. “Different categories of conservation units can be gazetted on paper, but in practice they're a long way from working. Someone draws lines on a map high in an office in Brasilia, but when you go out to that place in the forest, no one knows it's a conservation zone.” Fire Prevention: Developing the Technology Information transfer faces similar barriers in much of Southeast Asia, as Canadian forestry researchers discovered during a five-year project (now winding up) to create a computerized early warning tool for wildfire outbreaks. The program was instigated after the 1997–1998 fires created a regional haze hazard, largely because of peat deposits up to 20 meters (21.8 yards) thick that had become susceptible to burning in swampy forests drained and cleared for development. Michael Brady, who managed the Canadian project in Jakarta, points out that headmen in remote communities are still likely to believe that wildfires start spontaneously, by grasses rubbing together or even by magic. A fire scientist whose doctorate is in tropical forest peat dynamics, Brady sees the project as a medium to strengthen regional fire ecology in general. “In some ways, that's more important to me than the tool itself.” The tool is a variation of the Fire Danger Rating System used in Canada and, with various permutations, in many other countries. The Canadian system has two components, one for indexing fire weather and another predicting fire behavior. The weather component models moisture input and output in fuels generically classed as fine, moderate, and heavy. Brady and Indonesian university scientists grouped grasslands in the fine fuel category, fallen leaves and litter as medium, and peat and woody materials as heavy. They spent three years calibrating these fuels to local weather conditions, examining moisture dynamics and performing ignition tests. In developed countries, fuels are further specified in numerous classes for fire behavior prediction, but that requires decades of field work. Brady's team concentrated instead on helping key agencies in seven Southeast Asian countries, especially Indonesia and Malaysia, to obtain and use the appropriate computing tools. Brady doesn't expect immediate results in terms of reducing acreage burned. “Canada and the U.S. still have huge fire problems after working on it for a century.” But he does hope for a change of thinking, away from a current fascination in the region with satellite imaging of “hot spots” where fires are likely to be occurring. Fire danger rating concentrates on where fires are most likely to begin. “It allows you to add prevention into your management program.” Beyond Prevention In South Africa, “retention” is a conservation buzzword referring to strategies that, in a sense, go beyond prevention of problems. What ecologists hope to retain is biodiversity in the midst of changes that can't be stopped, and their methods are producing major repercussions throughout government. The work is centered on the Cape Floristic Region of Africa's southwestern tip. Almost 90,000 square kilometers (34,750 square miles) in area, it's the world's smallest floral kingdom. A conservation plan was launched in 1998 that has drawn cooperation from tourism, mining, water use, agricultural, and land use planning groups. The project has the ambitious goal of protecting not only the usual biodiversity patterns of conservation areas but also the spatial components of evolutionary processes that enable species to adapt to potentially harmful changes. This entails a complex effort to determine which parts of developed and undeveloped lands are most necessary to such processes, including rivers, sand movement corridors, gradients from uplands to coastal lowlands, and major wilderness areas. University of Port Elizabeth botanist Richard Cowling, one of the scheme's principal architects, estimates that it might require 60%–70% of the region's landscape. As in Australia, fires are important to the Cape's biodiversity, but too-frequent burns are a problem. Cowling thinks that by consolidating mountainous megawilderness under the project's plan and protecting spatial transitions between fire-prone areas and those that resist fire, managers could move toward allowance of natural fire regimes. The current problem, he says, is that protected areas usually stop short of the transition to semidesert areas that are privately owned. When fire spreads from public to private land, the government often gets sued. Under the evolving Cape plan, landowners will sit on governing boards, and property that they contract for conservation will be tax-exempt. The Cape plan has attracted millions of dollars in support from the World Bank and other international sources, but Cowling regards that achievement as much less important than the progress made in gaining support from various interest groups. “The key issue is the extent to which you can get biodiversity concerns mainstreamed to other sectors,” he says. Threats to habitat retention, not least of which is wildfire, endanger every species. “It's about making people realize that biodiversity is the basis upon which all other things will succeed.” Box 1. Fire-Adapted Species Plants and animals of many countries evolved for millennia with wildfire as a natural occurrence, but when human interventions increase the frequency of fire, species suffer. African fire lilies and Australian “grass trees” are among plants that are stimulated to flower by smoke constituents such as ethylene. Plant seeds in fire-prone landscapes of Australia and South Africa often require fire to stimulate their germination, but it can take more than a decade for new seed banks to mature in some species. If a second fire arrives before then, the species could die out. Animals can be similarly affected. For example, a threatened marsupial called the potoroo is capable of surviving a high-intensity wildfire, but cannot tolerate the habitat changes caused by frequent, low-intensity fires. Likewise, some species of Australian honeyeaters are threatened with extinction because too-frequent fires have changed the proportion of mature and immature nectar plants. On the other hand, ecosystems can also be transformed by fire suppression. In southern Africa, decades of such activity have encouraged forests to replace grasslands. Yet the lovely marsh rose almost disappeared from the Cape before land managers realized that fire suppression was preventing its seeds from germinating. In such ways, biodiversity must find its place among the goals and tradeoffs of human intervention.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC340956.xml
546198	Lessons learnt from the painful shoulder; a case series of malignant shoulder girdle tumours misdiagnosed as frozen shoulder	Adhesive capsulitis or frozen shoulder is a common condition characterized by shoulder pain and stiffness. In patients in whom conservative measures have failed, more invasive interventions such as arthrographic or arthroscopic distension can be very effective in relieving symptoms and improving range of movement. However, absolute contraindications to these procedures include the presence of neoplasia around the shoulder girdle. We present five cases referred to our institution where the diagnosis of shoulder joint malignancy was delayed, following prolonged, ineffective treatment for frozen shoulder. These cases highlight the importance of careful review of the radiology and the need for reconsideration of the diagnosis in refractory "frozen shoulder".	Introduction Frozen shoulder was first described by Codman in 1934, as an idiopathic painful restriction in the range of shoulder joint movement, in the presence of normal plain radiographs [ 1 ]. It is also known as "adhesive capsulitis", based on the presence of chronic synovitis and a contracted, thickened joint capsule seen during open surgery of the shoulder joint [ 2 ]. It is usually a self-limiting condition, with a mean duration of one to three years [ 3 ]. The natural clinical course involves an initial painful phase, followed by progressive stiffness, with a gradual return of functional range of motion [ 4 ]. However, between 15 – 50% of patients have persisting severe refractory pain that is unresponsive to conservative management involving physiotherapy, non-steroidal anti-inflammatories and subacromial corticosteroid injections [ 5 , 6 ]. More aggressive treatment options for these patients include manipulation under anaesthesia, arthrographic capsular distension (hydrodilatation), and arthroscopic or open capsular release [ 7 , 8 ]. Hydrodilatation is commonly performed as treatment for frozen shoulder as it is minimally invasive, inexpensive, does not require an anaesthetic, and is effective [ 9 - 12 ]. The procedure involves insertion of a needle into the glenohumeral joint under radiologic guidance, followed by gradual distension of the capsule with a combination of local anaesthetic, corticosteroid and normal saline, until lysis of adhesions and capsular rupture are achieved [ 13 ]. Arthrography performed at the beginning of the procedure by injecting radio-opaque contrast material into the shoulder joint is the definitive diagnostic investigation for frozen shoulder, and is associated with decreased joint volume and obliteration of the axillary fold and subscapular bursa. Tumours around the shoulder girdle are uncommon causes of shoulder pain and stiffness, but often present with symptoms and a clinical history identical to that of a frozen shoulder. A strict contraindication to arthrographic or arthroscopic distension of the shoulder is the presence of a local oncological process. Such procedures may change the surgical management from being a limb-preserving resection to a forequarter amputation. In the past month, five patients have been referred to us with malignant tumours around the shoulder joint, all previously diagnosed as having a frozen shoulder. All patients had undergone prolonged conservative management and hydrodilatation, with persistence of symptoms. Two of the patients had also undergone arthroscopic surgery. The following cases illustrate the importance of reconsidering the diagnosis in refractory frozen shoulder and the value of a detailed clinical history and examination and careful consideration of radiologic imaging in assessing recalcitrant "frozen shoulder". Case Reports Case 1 A 60 year old woman presented to her local medical officer with an eighteen month history of worsening right shoulder pain and stiffness. She was initially treated with oral analgesia followed by a cortisone injection without improvement. Two months later she had a hydrodilatation of the shoulder but her symptoms persisted. MRI was then performed, which demonstrated a large permeative tumour arising from the scapula (Figure 1 ). She was subsequently referred to us, and underwent staging studies and needle biopsy. Histologic sections were consistent with Ewing's sarcoma. Figure 1 A Plain radiograph of the right shoulder, showing an irregular, mixed lytic and sclerotic lesion in the glenoid (arrow), that was not appreciated by the reporting radiologist. B Arthrogram performed prior to hydrodilatation. C Coronal TSE post-contrast MR image, showing a diffusely enhansive scapular lesion extending into the inferior aspect of the gleno-humeral joint. D Axial TSE post-contrast MR image showing diffuse enhancement of the tumour extending on either side of the scapular blade with bony destruction. E Bone scan showing increased uptake in the area of the lesion on the delayed static image. F Thallium functional scanning showing retained thallium activity in the glenoid region at 4 hrs. Subsequent biopsy was consistent with Ewing's sarcoma. Case 2 A 42 year old man was referred to an orthopaedic specialist with a history of sudden onset left shoulder pain following a work related activity. He was initially diagnosed with rotator cuff tendinopathy and subacromial impingement, and had a course of intensive physiotherapy followed by arthroscopic shoulder surgery, without improvement in symptoms. Two months later a minor incident involving his left shoulder led to an increase in pain and swelling and reduction in movement. Hydrodilatation was then performed but pain and function of the shoulder continued to worsen. On retrospective review of plain x-rays of the shoulder, a destructive lesion at the metaphysis with a cortical breach medially in the region of the surgical neck of the humerus was realized (Figure 2 ). Further anatomic imaging showed an aggressive tumour mass in the proximal diaphysis of the humerus and humeral head extending into the adjacent soft tissues. He was referred to us, and subsequent biopsy of the lesion was consistent with a high-grade pleomorphic sarcoma. After a course of chemotherapy, the patient underwent en bloc resection of the tumour, via an extra-articular approach. Definitive histopathologic diagnosis was malignant fibrous hystiocytoma. Figure 2 A Plain radiograph of the left shoulder showing a lytic lesion affecting the proximal humerus, with cortical irregularity medially (arrow), that was not initially recognized. B At the time of arthrographic distension, the lesion (arrow) was more apparent, but remained unnoticed. C Sagittal TSE post-contrast MR image showing an enhancing lesion within the proximal humerus extending outside the bone. D Axial TSE post-contrast MR image showing the tumour destroying the humeral head and extending into the gleno-humeral articulation. E Bone scan showing increased uptake in the area of the lesion on the delayed static image. F Thallium functional scanning showing retained thallium activity in the proximal humerus at 4 hrs. Histological sections from the biopsy and surgical resection specimen were consistent with a malignant fibrous histiocytoma. Case 3 A 50 year old women was referred to an orthopaedic specialist with a 6 month history of episodic pain in the right shoulder, with associated decreased range of movement. Initial plain x-rays were unremarkable. She then underwent a variety of procedures, which included repeated subacromial corticosteroid injections, arthrographic distension, manipulation under anaesthetic, and arthroscopic debridement and acromioplasty. On arthroscopy, a marked synovitis was observed, to which her ongoing symptoms and the development of a palpable mass on the anterior aspect of the shoulder was initially attributed. Repeat plain radiographs two years after the onset of her symptoms demonstrated a large lesion extending from the glenoid cartilage into the base of the coracoid process (Figure 3 ). She was then referred to us, where staging radiologic imaging studies and CT-guided biopsy was consistent with a low-grade chondrosarcoma. The patient subsequently underwent en bloc resection of the tumour. Figure 3 A Initial plain radiographs of the shoulder were unremarkable. B Repeat radiographs after two years of failed treatment, showing an irregular mixed lytic and sclerotic lesion destroying the coracoid process of the scapula (arrow), which was not appreciated. C Arthrogram performed prior to hydrodilatation similarly showing the destructive process, which remained unnoticed. D Axial T1-weighted post-contrast MR image showing a heterogenous contrast-enhancing lesion destroying the glenoid and extending into the gleno-humeral joint. The lesion is lobulated and loculated with central areas of lower signal intensity, suggestive of a chondroid lesion. E Bone scan showing increased uptake in the area of the lesion on the delayed static image. F Thallium functional scanning showing no retention of thallium by the lesion at 4 hrs. Biopsy confirmed low-grade chondrosarcoma. Case 4 A 68 year old man had previously had a squamous cell carcinoma of the upper back excised, after which he had a three year history of shoulder pain and stiffness. He was treated for a frozen shoulder and received intensive physiotherapy and multiple subacromial corticosteroid injections, followed by hydrodilatation. Initially this seemed to settle his symptoms, although a month later pain and stiffness recurred, with marked reduction in shoulder function, and he was referred to us. An MRI was performed, which showed lesions in the supraspinatus and trapezius muscles, which were consistent with metastatic deposits (Figure 4 ). The patient underwent a course of palliative radiation therapy. Figure 4 Patient had persistent pain and stiffness following hydrodilatation. A Plain shoulder radiographs were normal. B STIR MR image showing multiple high signal intensity lesions in the supraspinatus muscle. A presumptive diagnosis of metastatic squamous cell carcinoma was made. Case 5 A 55 year old female with a past history of malignant fibrous histiocytoma of the left thigh resected five years previously, presented to her local medical officer with right shoulder pain. Plain films were performed at the time which appeared normal (Figure 5 ). Subsequent treatment included multiple intra-articular injections of corticosteroid and local anaesthetic, however her pain and associated restricted movement worsened. She was referred for an orthopaedic surgical opinion and shoulder ultrasound. An arthroscope was performed which demonstrated rotator cuff pathology but failed to reveal the actual cause of the patient's symptoms. One year after the onset of her original symptoms, repeat plain films showed destruction of the glenoid and coracoid process. A bone scan demonstrated increased osteoblastic activity involving the coracoid process and right humeral head and relative photopaenia of the glenoid. Functional thallium scintigraphy showed increased metabolic activity around the right shoulder joint with CT and MRI scanning confirming destruction of the glenoid with an associated soft tissue mass and involvement of the humeral head (Figure 5 ). CT-guided percutaneous biopsy was performed and diagnosis of malignant fibrous histiocytoma made. Figure 5 A Initial plain radiographs of the right shoulder appeared normal. B CT scan was performed after a year of progressive shoulder pain and stiffness, showing a destructive lesion involving the glenoid (arrow). C Axial T1-weighted and post-contrast ( D ) MR images showing destruction of the glenoid with an associated soft tissue mass and involvement of the humeral head. E Bone scan showing increased osteoblastic activity involving the coracoid process and right humeral head and relative photopaenia of the glenoid. F CT-guided percutaneous biopsy was able to obtain a histological diagnosis of malignant fibrous histiocytoma. Discussion Adhesive capsulitis or frozen shoulder is a common condition that may affect up to 5% of the general population in their lifetime. Although the aetiology of frozen shoulder is unknown, it has been associated with diabetes mellitus, thyroid disease, ischaemic heart disease and various autoimmune conditions [ 14 ]. Other causes of shoulder pain and stiffness that need to be excluded include rotator cuff pathology, arthritis, fractures, infection and local tumours [ 15 , 16 ]. Arthrographic or arthroscopic distension with shoulder capsular rupture are effective treatment modalities in well-selected patients with refractory frozen shoulder symptoms despite intensive conservative management [ 5 - 7 ]. In a recent randomised, double blinded study, Buchbinder et al . [ 12 ] demonstrated a significant improvement in both pain and range of motion in patients treated with hydrodilation compared with arthrogram alone. However, absolute contraindications to surgical intervention for frozen shoulder include neurological abnormalities originating from the cervical spine, presence of infection, and an ongoing oncological process. Tumours of the shoulder girdle are uncommon causes of shoulder pain and restricted movement. In most cases, they are diagnosed based on the presence of a soft tissue mass on clinical examination, as well as characteristic radiographic changes. Robinson et al . [ 17 ] suggested that younger patients with bony tenderness elicited by gentle tapping are more likely to have a shoulder neoplasm. However, in up to 10% of shoulder neoplasms, plain x-rays are normal, and these patients may present with painful limitation of shoulder motion that can be difficult to distinguish from primary frozen shoulder. Indeed, in one series of 140 patients with frozen shoulder referred for manipulation, 2% had a primary chest wall tumour [ 18 ]. Misdiagnosis, inappropriate surgery and delayed therapy for shoulder symptoms due to malignancy may potentially have grave consequences. Our five patients had locally invasive malignant tumours, and received prolonged conservative and interventional treatment for "frozen shoulder" before the definitive diagnosis of tumour was made. In all cases of recalcitrant frozen shoulder resistant to conventional treatment, less common causes for shoulder pain and stiffness such as an ongoing oncological process must be considered. A detailed clinical history and examination is critical in the assessment of a painful, stiff shoulder. Plain antero-posterior and axillary lateral radiographs of the shoulder should be performed as a routine, and these films then require careful review by an experienced radiologist prior to undertaking any invasive procedures. More sensitive radiological investigations such as radionucleotide scanning and CT scanning or MRI should be considered when shoulder symptoms are atypical or progress despite invasive management, if there is suspicion of malignancy, or if there are any bony abnormalities evident on plain radiographs. Abbreviations MRI: magnetic resonance imaging, CT: computed tomography, TSE: turbo spin echo, STIR: short tau inversion recovery.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC546198.xml
518966	What is the impact of the ACE gene insertion/deletion (I/D) polymorphism on the clinical effectiveness and adverse events of ACE inhibitors? – Protocol of a systematic review	Background The Angiotensin Converting Enzyme (ACE) insertion/deletion (I/D) polymorphism has received much attention in pharmacogenetic research because observed variations in response to ACE inhibitors might be associated with this polymorphism. Pharmacogenetic testing raises the hope to individualise ACE inhibitor therapy in order to optimise its effectiveness and to reduce adverse effects for genetically different subgroups. However, the extent of its effect modification in patients treated with ACE inhibitors remains inconclusive. Therefore our objective is to quantify the effect modification of the insertion/deletion polymorphism of the angiotensin converting enzyme gene on any surrogate and clinically relevant parameters in patients with cardiovascular diseases, diabetes, renal transplantation and/or renal failure. Methods Systematic Review. We will perform literature searches in six electronic databases to identify randomised controlled trials comparing the effectiveness and occurrence of adverse events of ACE inhibitor therapy against placebo or any active treatment stratified by the I/D gene polymorphism. In addition, authors of trials, experts in pharmacogenetics and pharmaceutical companies will be contacted for further published or unpublished data. Hand searching will be accomplished by reviewing the reference lists of all included studies. The methodological quality of included papers will be assessed. Data analyses will be performed in clinically and methodologically cogent subgroups. The results of the quantitative assessment will be pooled statistically where appropriate to produce an estimate of the differences in the effect of ACE inhibitors observed between the three ACE genotypes. Discussion This protocol describes a strategy to quantify the effect modification of the ACE polymorphism on ACE inhibitors in relevant clinical domains using meta-epidemiological research methods. The results may provide evidence for the usefulness of pharmacogenetic testing for individualised ACE inhibitor therapy.	Background Research in genetics and genome sequencing has led to a better understanding of the molecular genetic mechanisms and to the detection of inter-individual genetic differences, so-called polymorphisms, which may have a functional consequence on the response to drugs. Pharmacogenetic tests provide information to better predict and prevent therapeutic failures and adverse drug reaction and raise the hope for an individualised pharmacotherapy [ 1 - 3 ]. Although pharmacogenetic research has moved into several branches of medicine such as cardiology [ 4 ], oncology [ 5 ] and respiratory medicine [ 6 ], the implementation of pharmacogenetic testing into clinical practice is still at the very beginning [ 7 ]. The ACE polymorphism identified in 1990 by Rigat and co-workers [ 8 ] is one of the best-researched polymorphisms. This polymorphism of the ACE gene is based on the presence or absence of a 287-bp element on intron 16 on chromosome 17. Rigat et al. have shown that the level of circulating ACE enzymes depends on the insertion/deletion (I/D) polymorphism. Since then it has been speculated that these differences in plasma ACE activity associated with the ACE genotype might affect the therapeutic response of ACE inhibitors explaining interindividual variability in cardiovascular or renal response to equivalent doses of ACE inhibitors [ 9 ]. Several studies have investigated the extent of this effect modification on response to ACE inhibitors for different indications such as hypertension [ 10 ], diabetic nephropathy [ 11 , 12 ] and coronary artery disease [ 4 , 13 ]. There are however inconsistencies in trial findings [ 14 - 16 ] and as a result the extent of effect modification of this polymorphism remains unclear. Therefore, our objective is to systematically review all randomised controlled trials that assessed to what extent the insertion/deletion polymorphism of the angiotensin converting enzyme gene influences the effect and adverse events of angiotensin converting enzyme inhibitors on any surrogate and clinically relevant parameters in patients with cardiovascular disease, diabetes, renal transplantation and/or renal failure. Methods Search strategy We will perform literature searches in (Pre-) MEDLINE (DataStar version, Cary North Carolina), EMBASE (DataStar version, Cary North Carolina), Biosis (Ovid version "Previews 1989 to 2003", New York, New York), the Cochrane Controlled Trials Register (CCTR <3rd Quarter 2003>, Oxford, United Kingdom) and the Science citation index. A preliminary literature search in Medline has been carried out to estimate the range of relevant literature. Out of the citations of the pilot searches (172 citations) we identified articles that met our inclusion criteria. Keywords of these articles were used to refine our search strategies. In collaboration with an information specialist we designed the final search strategies for the six databases avoiding any language restrictions [see additional file 1 : the search strategies]. In addition, authors of trials identified in the literature search will be contacted for additional published or unpublished data. Particular efforts will be made to obtain unpublished data on genetic test information and effect measures stratified according to the genetic subtypes examined. We will send our requests and subsequent reminders for additional data to the first and last authors. Other contacts will include the relevant collaborative review groups of the Cochrane Collaboration, pharmaceutical companies and manufacturers and researchers known to have published pharmacogenetic analyses in the area of cardiovascular disease, diabetes, renal transplantation and/or renal failure. Hand searching will be accomplished by reviewing the bibliographies of all included studies to identify additional relevant articles as well as by using the "related articles" function of PubMed and the citation index of ISI Web of Science. Anticipating that subgroup analyses investigating gene polymorphisms may not be specifically mentioned in titles or abstracts, we will study the full text of all randomised placebo-controlled trials (RCT) that assessed the effectiveness of ACE inhibitors in order to identify subgroup analyses investigating gene polymorphisms. Inclusion criteria Two reviewers will independently assess all obtained titles and abstracts of the literature search for inclusion. The criteria to be used to identify relevant studies will be 1) randomised controlled trials 2) the investigation of an angiotensin converting enzyme inhibitor used for one of the clinical domains mentioned below and 3) the determination of the deletion/insertion polymorphism of patients. The two reviewers will then examine the full texts of all potentially relevant citations. The decision on in- and exclusion will be based on the following, more explicit inclusion criteria. Clinical domains We will include studies investigating ACE inhibitors in the four major clinical domains namely cardiovascular diseases, diabetes, renal transplantation and/or renal failure. Patients Studies should include patients with the following indications for ACE inhibitor therapy: Heart failure, primary and secondary hypertension, coronary artery disease, diabetic nephropathy, primary nephropathy and status after renal transplantation. Intervention All licensed or unlicensed ACE inhibitors identified through the literature search will be included. We will prefer placebo as control intervention in order to study the effect modification. However to assess all available evidence, we will also include pragmatic trials where patients with active treatments (e.g. usual care with any antihypertensive medication) served as controls. Co-intervention We are mainly interested in studies investigating a single drug exposure with ACE-inhibitors. However we will also include those studies, which allowed co-medications. Description of the pharmacogenetic test Studies must include a description on how determination of the angiotensin converting enzymes genotypes (DD/ DI/ II) has been performed. If a study does not report details of testing but provides relevant results, authors will be contacted to obtain information of testing technique. Outcomes We will secure data on any reported outcome, surrogate endpoints (e.g. decrease in blood pressure, changes in hemodynamic parameters, proteinuria, creatinine levels, microalbuminuria) and clinically relevant outcomes (e.g. total and disease specific mortality, morbidity (none fatal myocardial infarction, reinfarction, stroke, transient ischemic attack, rehospitalisation kidney failure or end-stage renal disease)). The two reviewers will resolve any discrepancies about in- or exclusion by discussion. If agreement cannot be achieved, a third reviewer will make the decision. Data extraction strategy We will use a pre-designed data extraction form that includes different items to assess the studies' external validity [see additional file 2 : Data extraction and quality assessment sheet]. Details on study design, treatment, patients and pharmacogenetic tests as well as outcome parameters will be registered onto the data extraction form independently by two reviewers. Also, bibliographic details such as author, journal, year of publication and language, will be registered. This list will be pre-tested on a small sample of included and excluded studies addressing the appraisal topic. A third reviewer will resolve any discrepancies. The data extraction shows the extent of insufficient reporting and authors will be contacted to obtain missing information. Quality assessment strategy All trials included in the review will be assessed using a list of selected quality items indicating components of internal validity and descriptive information [ 17 ]. In principle these selected items will enable us to define any process at any stage of inference tending to produce results that differ systematically from the true values (bias) [ 18 ]. We will also assess additional methodological aspects that might bias the results of pharmacogenetic studies (e.g. blinding of laboratory assessor of outcomes, blinding of outcome assessor for genotypes, blinding of treatment provider for genotypes. See Data extraction and quality assessment sheet). In addition, we will assess the description of the methods to determine genotypes. Angiotensin converting enzymes genotypes (DD/ DI/ II) are traditionally determined using polymerase chain reaction (PCR) amplification according to previously published protocols [ 19 ]. The D allele is preferentially amplified; therefore each sample found to have the DD genotype should be confirmed in a second independent PCR amplification by the use of an insertion specific primer to avoid the misclassification of the 4–5 percent of samples with DI genotypes as DD genotypes [ 19 ]. Beyond the use of the standard PCR with/without the second round of PCR using an insertion-specific primer, there is also a "tri-primer" method, which has been shown to be the proper method to be used in genotyping ACE I/D polymorphism [ 20 ]. The methodology of the ACE genotyping will be considered as an explanatory variable for heterogeneity between studies [see additional file 2 : Data extraction and quality assessment sheet]. We will pre-test these quality assessment items on a small sample of studies in duplicate and if necessary add missing descriptive items. Two reviewers will independently score the internal and descriptive validity. The initial degree of discordance between the reviewers will be reported. Discordant scores based on obvious reading errors will be corrected. Discordant scores based on real differences in interpretation will be resolved through consensus. A third party will be sought if necessary. The reviewers will not be blinded for names of authors, institutions, journals or the outcomes of the studies. These detailed quality assessment will be used to describe the methodological quality of selected studies, to explore potential sources of heterogeneity, to make informed decisions regarding suitability of meta-analysis and to weigh the strength of any conclusions. Methods of analysis and synthesis Description of data The results of the data extraction and assessment of study validity will be presented in different structured tables and in a narrative description [see additional file 2 : Data extraction and quality assessment sheet]. This will allow us to display variation in patient characteristics, study quality and results. Thus, the description will include the details about the clinical domain in which the ACE inhibitors have been assessed, information about the study design and quality, a list reporting co-interventions during the study period, details about the study population (baseline characteristics, e.g. severity of disease, ethnic groups, environmental and social characteristics) and a description of the outcome measures that were applied. Finally the tables will provide the individual study results (all reported outcomes) of the different genotypes in the intervention and control group. Continuous outcomes (e.g. blood pressure) will be summarised in the table as mean differences between baseline and follow up measures. For data of dichotomous outcomes (e.g. cardiovascular death) the relative risks between the results of the DD genotype and the II genotype will be calculated and described in the table. A relative risk of one indicates no difference between two genotypes, where as relative risks lower respectively higher than 1 indicates variations in the treatment effect. Heterogeneity assessment The heterogeneity assessments help us to examine study characteristics that might be related to variability in the observed outcome. Within each subgroup potential sources of heterogeneity that may affect the imprecision in the estimate of treatment effect such as the study methodology, population characteristics, intensity of intervention, co-medications and risk factors will be examined. We will perform multiple linear regression analyses (meta-regression) to explore sources of between-study heterogeneity. The log transformed odds ratio for dichotomous outcomes (myocardial infarction) and continuous outcomes (blood pressure) measurements will be used as dependent variables and the clinical and methodological items of the extraction sheet as described above will be entered into the model as independent variables. When a factor is strongly associated with the variation in ln odds ratio or on the continuous outcome, we will stratify the studies on that variable and inspect residual heterogeneity using forest plots. If a meta-analysis seems appropriate, that is when the p-value of the chi-squared test for heterogeneity is greater than 0.10, a fixed effects model will be used for pooling. Within clinically and methodologically cogent subgroups relative risks for dichotomous outcomes and weighted mean differences for continuous outcomes will be calculated comparing the contrast between the intervention and the control group within genotypes. The results between the different genotypes will be presented in a forest plot as shown in Figure 1 and differences will be statistically assessed. The pooled results of a cogent subgroup will produce an estimate of the differences in the average effect of ACE inhibitors observed between ACE genotypes. Thus the treatment effect of each genotype could be compared to the overall effect of ACE inhibitors regardless of the genotype. All statistical analyses will be performed using the Stata statistical software package (StataCorp. 2004. Stata ® Statistical Software: Release 8.2 College Station, Texas, USA). Figure 1 Example of analysis using virtual data . Forest plot: For clinically and methodologically cogent subgroups weighted mean differences (95% confidence interval) for reduction of systolic blood pressure in patients with hypertension have been assessed. This graph displays differences of the ACE inhibitor effect within genotypes. The diamonds below each of the three genotypes indicate the pooled results. The lowest (forth) diamond reflects the overall effect of ACE inhibitors across all genotypes. In this example, the DD genotype shows the largest ACE inhibitor effect and the II genotype shows the smallest effect. The size of the box is related to the number of studied patients. Discussion This review shows an efficient approach to quantify the effect modification of the ACE polymorphism on ACE inhibitors when applied in different clinical occasions. We aim to resolve part of the controversy in the literature by quantifying the influence of the three genotypes (DD/DI/II) on different outcomes and in the light of study methodology and participants characteristics. These results should inform clinicians about the potential of pharmacogenetic testing to individualise ACE inhibitor treatment. Competing interests None declared. Authors' contributions LMB, JS and MS initiated the project. MS wrote the first draft of the protocol. MP, JS and LMB critically reviewed and revised the manuscript. All authors approved the manuscript. Pre-publication history The pre-publication history for this paper can be accessed here: Supplementary Material Additional File 1 Search strategies: Search terms and number of citations listed for five electronic databases Click here for file Additional File 2 Data extraction and quality assessment sheet: Example of data extraction and quality assessment for the study of Hernandez [21] Click here for file	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC518966.xml
544957	Association between monocyte Fcγ subclass expression and acute coronary syndrome	Background Atherosclerosis lesions contain abundant immunoglobulins complexed with oxidized LDL (OxLDL) that are endocytosed by macrophages to form foam cells. While recent evidence supports a role for the macrophage scavenger receptor pathway in 75–90% of OxLDL uptake, in vitro evidence suggests another potential uptake pathway could involve autoantibody binding to IgG subclass-specific Fc receptors. Objective and Methods To address this mechanism from an in vivo standpoint, the objective of this study was to utilize flow cytometry to prospectively determine monocyte Fcγ (FcR) I, II, and III receptor expression levels in patients with acute coronary syndrome (ACS, n = 48), diabetes mellitus (DM, n = 59), or neither (C, n = 88). Results Increased FcR I expression was found in the ACS versus DM groups [geometric mean, (95% CI) = 2.26 (2.07, 2.47) versus 1.83 (1.69, 1.98) (p < 0.001)] and versus C [1.90 (1.78, 2.03) (p = 0.005)]. Similar relationships were found with both the FcR II receptor [ACS mean = 4.57 (4.02, 5.19) versus DM 3.61 (3.22, 4.05) (p = 0.021) and versus C 3.86 (3.51, 4.24) (p = 0.09)] and FcR III receptor [ACS mean = 1.55 (1.44, 1.68) versus DM 1.36 (1.27, 1.46) (p = 0.038) and versus C 1.37 (1.30, 1.45) (p = 0.032)]. There was no difference between DM and C groups in FcR I, II or III expression. Conclusions This in vivo data supports a possible second OxLDL-autoantibody macrophage uptake mechanism through an Fc receptor-mediated pathway and a potential relationship between atherosclerotic plaque macrophage FcR levels and ACS.	Introduction Atherosclerosis is a chronic inflammatory process that results from hyperlipidemia and complex interactions involving other genetic and environmental factors. OxLDL plays a central role in the atherogenic process through generation of highly immunogenic neodeterminants for the immune system [ 1 ]. Natural autoantibody titer to a number of these epitopes and extent of immune complex formation may correlate with plaque size and rate of progression, and plaques have been shown to contain OxLDL/autoantibody immune complexes [ 2 - 5 ]. It is clear that both innate and adaptive immunity can modulate lesion progression and composition, and most studies to date have indicated a proatherogenic influence of the immune system on this process [ 1 , 4 ]. Recent evidence supports the macrophage scavenger receptors SR-A and CD36 as a mechanism responsible for up to 90% of uptake of OxLDL that leads to foam cell formation with no other scavenger receptors compensating for their absence in a knockout mouse model [ 6 ]. Earlier evidence involving in vitro incubation of both human monocyte-derived macrophages and the monocytic cell line THP-1 with human LDL-rabbit anti-apo B immune complexes demonstrated a potential role for the FcγRI receptor in its uptake [ 7 ]. A second in vitro study also suggested a potential Fc receptor role through inhibition of immune complex uptake when Fab or F(ab') 2 fragments were substituted for an intact anti-apo B antibody [ 8 ]. Though findings from the latter two studies may have been partly explained by contributions from the scavenger pathway, it is reasonable to speculate the Fc receptor pathway maybe playing a small but important role as well [ 9 - 11 ]. Immune complexes with modified lipoproteins have recently emerged as an important coronary artery and macrovascular disease risk factor in DM [ 12 , 13 ]. Evidence supports an increased content of macrophages in the atherosclerotic lesions of persons with DM that is thought to be due to altered levels of cytokines [ 12 ]. Furthermore, while DM itself does not increase levels of LDL, the small dense LDL particles found in type 2 DM are more atherogenic because they are more easily glycated and are thought to be more susceptible to oxidation [ 14 , 15 ]. In recent work our group has shown FcγRII expression to be increased in the platelets of patients experiencing an acute atherothrombotic event, or who are healthy with two or more atherosclerosis risk factors [ 16 ]. Non-acutely ill diabetes patients have significantly elevated expression levels and this may play a role in the increased sensitivity of their platelets to activation by subendothelial collagen [ 16 - 19 ]. We speculate that Fc expression levels and activity on macrophages and platelets may represent another link between the immune system and atherosclerosis progression and plaque disruption. In view of the controversy regarding the mechanism of cholesterol uptake by monocyte-macrophages in atherosclerosis and diabetes [ 20 , 21 ] and the previous lack of in vivo data to help elucidate any role of the Fc receptors in this process, we have prospectively determined IgG-binding receptor expression levels for each Fcγ receptor subclass on the monocytes of three groups: (1) patients admitted to the hospital with ACS, (2) well patients with no history of heart disease but one or more atherosclerosis risk factors (ARF's) that included DM, and (3) control patients (with no history of ACS or DM). Materials and Methods All 195 patients were randomly chosen for study participation from a larger group who fit study inclusion criteria and gave written informed consent. Forty-eight patients in the study had heart disease (HD) and were within 2 hours of onset of an ACS (myocardial infarction or unstable angina), 59 were DM outpatients (both type 1 and type 2 were included) with no known history of HD, and an additional 88 outpatients without HD or DM were randomly chosen as controls (C). The number and nature of ARF's was documented for each group (Table 1 ). Table 1 Demographic characteristics at enrollment Characteristic Group 1 (ACS) Group 2 (DM) Group 3 (C) p-value 1 Total patients (% with MI Group 1) 48 (52) 59 88 Mean age (years) 56 55 56 0.86 2 Male (%) 35 (73) 24 (41) 55 (63) 0.002 Positive family history (%) 3 6 (13) 21 (36) 21 (24) 0.018 Current cigarette smoker (%) 16 (33) 10 (17) 5 (6) <0.001 Hypertension (%) 29 (60) 29 (49) 20 (23) <0.001 Abnormal lipid profile present (%) 24 (50) 12 (21) 17 (19) <0.001 Diabetes mellitus present (%) 15 (31) 59 (100) 0 (0) <0.001 1. Based on chi-square test. 2. Based on ANOVA F-test 3. One patient in group 2 had missing data with positive family history (Abbreviations: ACS: acute coronary syndrome; DM: diabetes mellitus; C: control; MI: myocardial infarction) Blood was collected in 3.8% trisodium citrate and divided into 50 μl aliquots to which 5 μl of a saturating concentration of anti-FcR I (32.2), anti-FcR II (IV.3), anti-FcR III (3G8), or a negative class-specific control antibody (MOPC-141, Sigma) was added. Following a 15 minute incubation, 5 μl of FITC-sheep anti-mouse antibody (Sigma) was added and a second 15 minute incubation done. The pellet was washed twice before 5 μl of phycoerythrin-conjugated anti-CD14 (Becton Dickinson) was added and incubated for 15 minutes. The solution was diluted with 1 ml ammonium chloride lysing solution and incubated for 10 minutes or until the solution was clear, and the pellet washed twice before flow cytometry to determine relative receptor expression levels was carried out according to the manufacturer's specifications (Becton Dickinson). Monocytes were readily identifiable from other blood cells by their forward and side scatter properties along with CD14 expression. Following establishment of saturating concentrations for each antibody, mean inter and intra-individual coefficients of variation for each of the three Fc receptors were calculated in the antibody labeling assay employing blood samples from five healthy laboratory volunteers who met control patient criteria and FcR I values found to be 3.2 and 9.7%, FcR II 11.7 and 16.1% and FcR III 4.1 and 26.9% respectively. The ratios of FcR I, FcR II, FcR III and MOPC-141 antibody expression were calculated for each patient and the logarithm of the ratios were used to analyze the results. An analysis of variance (ANOVA) was performed to compare the groups of (1) these 3 ratios with the 3 groups of patients, (2) the 3 ratios with the total number of major ARFs, (3) the 3 ratios with each ARF, such as hypertension, diabetes, and smoking, and (4) the 3 ratios with MI, and unstable angina in group 1. The overall p-values were based on the ANOVA F-test. If the overall F-test p-value < 0.05, the LSD method (least significant difference) [ 22 ] was used for multiple comparison. The geometric means and the associated 95% confidence intervals were calculated to summarize the data. Patient baseline data between groups was analyzed using the chi-square test. To perform statistical analysis, SAS software, version 8.2, was used (SAS Institute, Inc., Cary, NC). Results Significantly increased FcR I expression was found in ACS patients compared with DM patients [geometric mean FcR I expression, (95% CI) = 2.26 (2.07, 2.47) versus 1.83 (1.69, 1.98) (p < 0.001)] and compared with C [1.90 (1.78, 2.03) (p = 0.005)] (Table 2 , Figure 1 ). Similar relationships between the three groups were found to exist employing antibodies specific to the FcR II receptor: ACS geometric mean (CI) = 4.57 (4.02, 5.19) versus DM 3.61 (3.22, 4.05) (p = 0.021) and versus C 3.86 (3.51, 4.24) (p = 0.09) and the FcR III receptor: ACS geometric mean (CI) = 1.55 (1.44, 1.68) versus DM 1.36 (1.27, 1.46) (p = 0.038) and versus C 1.37 (1.30, 1.45) (p = 0.032). There was no difference between DM and C groups in FcR I, II or III expression (p = 0.73, 0.66, and 0.99 respectively). Table 2 Mean monocyte FcR expression in 195 study patients FcR I FcR II FcR III N Geometric mean (95% CI) p-value Geometric mean (95% CI) p-value Geometric mean (95% CI) p-value Unadjusted analysis Groups <0.001 2 0.024 2 0.021 2 Group 1 (ACS) 48 2.26 (2.07, 2.47) group 1 vs. 2: 0.001 3 4.57 (4.02, 5.19) group 1 vs. 2: 0.021 3 1.55 (1.44,1.68) group 1 vs. 2: 0.038 3 Group 2 (DM) 59 1.83 (1.69, 1.98) group 1 vs. 3: 0.005 3 3.61 (3.22, 4.05) group 1 vs. 3: 0.09 3 1.36 (1.27, 1.46) group 1 vs. 3: 0.032 3 Group 3 (C) 88 1.90 (1.78, 2.03) group 2 vs. 3: 0.73 3 3.86 (3.51, 4.24) group 2 vs. 3: 0.66 3 1.37 (1.30, 1.45) group 2 vs. 3: 0.99 Adjusted analysis 1 Groups 0.010 2 0.061 2 0.10 2 Group 1 (ACS) 48 2.33 (2.14, 2.55) group 1 vs. 2: 0.007 3 4.57 (4.00, 5.22) group 1 vs. 2: 0.050 3 1.58 (1.45,1.71) group 1 vs. 2: 0.10 3 Group 2 (DM) 59 1.95 (1.78, 2.13) group 1 vs. 3: 0.09 3 3.71 (3.24, 4.24) group 1 vs. 3: 0.47 3 1.41 (1.30, 1.53) group 1 vs. 3: 0.22 3 Group 3 (C) 88 2.05 (1.87, 2.25) group 2 vs. 3: 0.56 3 4.11 (3.59, 4.71) group 2 vs. 3: 0.38 3 1.44 (1.32, 1.56) group 2 vs. 3: 0.43 3 1. Adjusted by smoking and hypertension status 2. p-value based on F-test 3. p-value based on Tukey-Kramer test Figure 1 Monocyte Fcγ receptor subclass expression levels of 48 patients with heart disease (HD) and 59 patients with DM compared with 88 control patients with neither HD nor DM. HD patients display significantly increased expression levels of all 3 subclasses versus controls. * p = 0.002, 0.037, and 0.014 for FcR I, FcR II, and FcR III respectively when HD group is compared to control group. There were no statistically significant associations with increased expression of any Fc receptor and gender, family history of premature coronary disease, diabetes, or abnormal lipid profiles. Current cigarette smoking significantly increased expression of FcR I, 2.24 (2.01, 2.50), compared to absence of current cigarette smoking, 1.91 (1.82, 2.01) (p = 0.010) (Table 2 adjusted analysis). FcR II was significantly increased among the patients with hypertension, 4.29 (3.88, 4.74) compared to those without hypertension 3.73 (3.44, 4.05) (p = 0.034). There was a slight association between age and FcR I, II or III expression (p = 0.042, 0.050, 0.022 respectively). Expression levels in younger (age < 45) and older (age > 55) groups were higher than the middle age group. No difference was found in FcR expression with respect to ARF number (with the lone exception of increased FcR III in patients with 2 or more ARFs compared with less than two), or between the ACS subgroups of acute MI and unstable angina. When FcR expression is compared between all diabetes and non-diabetes patients in the 3 groups, there is no difference in monocyte FcR I, II, or III expression. Discussion It can be speculated from this in vivo data that phagocytosis of OxLDL-autoantibody immune complexes by plaque-associated macrophage through an Fc-mediated pathway could be a second uptake mechanism in addition to that involving the scavenger receptors. The potential clinical implication behind these findings is that while marrow and blood monocyte scavenger receptors SR-A and CD36 have not demonstrated inter-individual variability in their basal expression levels (prior to initial uptake of OxLDL or differentiation to macrophages) [ 23 , 24 ], the variable expression of Fcγ receptors found in this series of patients maybe playing a role in the extent of OxLDL immune complex uptake by atherosclerosis plaques. The fact we were able to document relatively increased surface expression of all three receptor classes in patients with ACS, along with increased FcR I in those who smoked and FcR II in those with hypertension, supports this hypothesis. Any precise pathophysiological implication behind these findings, though, or any cause and effect relationship between monocyte Fc expression and ACS is presently uncertain. No difference was noted in expression of any Fc receptor between the diabetes and control groups. Given that the control group was the largest of the three and that there was a difference noted in the expression levels of all three receptors between the two smaller groups, it can be concluded that increasing the sample size of either of the two groups with similar expression levels could possibly lead to an increase in the difference between them, but this difference is still unlikely to be of any significance compared with that between the ACS group and the other two groups. There was an interesting trend in both the unadjusted and adjusted analyses (Table 2 ) in which the control groups had consistently higher monocyte Fc expression levels compared with the diabetes groups. One potential explanation for this would be the hypothesis that those with diabetes may have reduced FcR expression levels (with a possible consequent decreased uptake in oxidized LDL) compared with non-diabetes subjects in response to relatively higher levels of molecular mediators that support atherosclerosis progression and a pro-inflammatory, pro-thrombotic environment. This may reflect a biological ying-yang type of response that leads to an attempt at dampening the effects of molecular players capable of contributing to atherothrombotic events. An issue pertinent to this study would be the possible effects of inflammation on monocyte FcR expression levels. In diabetes patients it would be reasonable to speculate a significant number of activated cells in the circulation would be unlikely since in most patients with atherosclerosis the inflammatory reaction is circumscribed to the vessel wall. Overexpression of monocyte Fc receptors may have been a possibility in ACS however. Figure 1 shows the ratio of ACS/control mean FcR expression levels to be very similar between Fc receptor subtypes. The extent of increased expression associated with inflammation-associated monocyte activation has been shown to be variable between FcR subtypes [ 25 ]. In this respect FcR II and III represent the receptors primarily involved in the inflammatory response in vivo. Since all three receptors had uniform increases in expression levels in ACS compared with controls (with the FcR I ratio being the highest of the three), it may be reasonable to attribute a relatively minimal effect of acute inflammation to the ACS data. Circulating monocyte activation may also have returned relatively close to baseline as a consequence of blood being drawn around 2 hours after symptom onset in most cases. The average circulation time of blood monocytes in response to an inflammatory stimulus may fall to as little as 30 minutes [ 26 ]. Observational studies of this nature have certain limitations related to their design and patient selection. As an example, selection bias needs to be considered in any study involving a population of volunteers associated with an atherosclerosis prevention trial (Groups 2 and 3). Even though the ethnic composition of the groups was the same, the implication of this selection bias is that the results are not generalizable to the population at large and this is attested to by the demographic characteristics in Table 1 . Through bridging innate and adaptive immune processes, macrophages play an important role in the progression of atherosclerosis and mediating plaque disruption that is considered to be the inciting event in the majority of coronary thrombi [ 27 , 28 ]. In this respect there is continual migration of monocytes between neighboring endothelial cells as well as two-way migration of monocytes between blood and OxLDL-containing foam cells when there is separation of endothelial cells associated with the fatty streak [ 29 ]. This observation and the inherent difficulty in isolating monocytes from tissues compared with serum both served as justification for utilizing blood monocyte Fc expression as a surrogate for plaque macrophage expression [ 30 ]. The overall effect of the humoral immune response on atherogenesis is likely to be complex. Of note, for example, is that the FcγRII receptor has an inhibitory role on B cells that are rarely seen in plaques, while it mediates phagocytosis and release of inflammatory mediators from cells of the myeloid lineage when cross-linked by immune complexes [ 31 , 32 ]. Thus FcR binding by opsonized OxLDL could induce either negative or positive regulation of immune cell responses. Elucidation of the immune mechanisms involved in atherogenesis will continue to evolve and lead to new insights into the molecular pathways associated with disease progression. Ultimately these insights will contribute towards the full explanation behind the clinical diversity of atherosclerosis expression in patients who appear to have equal risk. Competing Interests The authors declare that they have no competing interests.	/Users/keerthanasridhar/biomedlm/data/PMC000xxxxxx/PMC544957.xml

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.