Upload batch_8

batch_8/PMC2527489.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2527489",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2527489\nAUTHORS: Lior Dayan, Zohar Keidar, Ora Israel, Victor Milloul, Johnathan Sachs, Giris Jacob\n\nABSTRACT:\nPreserved blood flow to bone and soft tissue is essential for their normal function. To date only numerous methods are suitable for direct bone blood flow (BBF) measurement. Here, we introduce a novel quantitative method for bone and soft tissue blood flow (BBF and SBF, respectively) measurement. It involves a combination of SPECT/CT imaging for blood pool localization in a specific region of interest (\"soft\" and \"hard\" tissues composing a limb) with veno-occlusive plethysmography. Using it, we measured BBF and SBF in the four limbs of 10 healthy subjects. At steady state blood flow measurements in the four limbs were similar, ranging between 5.5 – 6.5 and 1.87–2.48 ml per 100 ml of tissue per minute for BBF and SBF, respectively. Our results are comparable to those in the literature. We concluded that SPECT/CT-plethysmography appears to be a readily available and easy to use method to measure BBF and SBF, and can be added to the armamentarium of methods for BBF measurements.\n\nBODY:\nIntroductionAs with all organs, bone blood flow (BBF) is vital to ongoing skeletal function and growth. BBF preserved at a sufficient degree is a crucial component of normal bone turnover and contributes significantly to the basic metabolic processes preserving bone integrity, as well as to repair mechanisms in pathological conditions such as fractures, infections and osteoporosis [1].Since blood flow to every organ is a dynamic process regulated by internal and external systems, its investigation requires methods that are accurate and reproducible. One method is the positron emission tomography (PET), which is a powerful and widely accepted tool in skeletal muscle perfusion study in humans [2,3]. It has also been utilized for BBF measurement in human [4,5,3], yet this method requires the availability of radioactive substances with extremely short half-life time (e.g. 15O nuclide), which are not readily available in many medical centers, thus rendering it unavailable for routine BBF measurements. Other acceptable methods for BBF assessment are either invasive or involve noninvasive imaging techniques with visual non-quantitative assessment of the transit of various radiotracers through certain anatomical region [6-9]. Laser Doppler technique, one of the most currently acknowledged and accepted techniques for use in BBF measurement in humans, is invasive and provides only a qualitative assessment of BBF. Additional methods of BBF measurement, although accurate and validated in many researches, were developed mainly for animal research (e.g., labeled microspheres, thermocouples and dilution methods) and are not applicable for use in humans [10,7,17].Given the paucity of a suitable quantitative methods for BBF measurement in humans, we were encouraged to develop one that would be relatively safe and available. Using dual modality SPECT/CT imaging devices, it is possible to accurately localize as well as quantify blood pool in regions of interest, i.e within limb compartments. The current study presents a novel method that is based on the combination of two well known clinical tools, strain gauge plethysmography and dual modality SPECT/CT functional and anatomic imaging. The first component of this method enables blood flow measurement in an entire limb. The second enables the highly accurate localization of radiotracer activity to a specific region of interest [18], in our case blood pool in limb different limb's compartments.MethodsSubjectsA group of 10 (4 females and 6 males) healthy subjects aged between 20 and 45, without any history of previous limb fractures, soft tissue damage or major trauma, diseases affecting the vascular system and with no history of any routine medication intake were recruited for the research. No smoking or alcoholic and monoamine containing beverages were allowed 24 hours prior the study. All subjects signed informed and written consent forms approved by the local institutional ethics committee.Experimental designPlethysmography studies were performed in a quiet, darkened room with ambient temperature of ~24°C and following an overnight fast. Thereafter, SPECT/CT studies were acquired. Both studies were performed in a supine position after a 10 minutes supine rest.Plethysmography studiesLimb blood flow (FL) was measured using the well acknowledged venous occlusion plethysmography technique[19]. Briefly, a sphygmomanometer cuff was applied at a predetermined point in the limb under investigation (i.e. 10 cm distal to the tibial tubercle in the leg and 10 cm distal to the olecranon tip in the forearm) and was inflated to 45 mm Hg for 7 seconds to prevent venous egress. During this period, forearm volume changes per time unit (correlates with blood flow changes) were measured by a strain gauge plethysmography (ECR5, Hokanson, Inc, Bellevue WA, USA). A 7-second deflation period was allowed before the subsequent measurement. The flow to the hand and foot was excluded by inflating a cuff above the systolic BP in the wrist or ankle, respectively. Baseline blood flow was the average of at least 4 stable repeated flow measurements. In order to test the reproducibility of the introduced method, all four limbs were measured in the same session. Each limb was taken as control for its contralateral. It is important to note that the plethysmography method measures the whole limb blood. While apparently it is the soft tissue volume that is changed in response to venous occlusion, it is the venous vasculature congestion within the soft tissue rather than soft tissue congestion per-se that is responsible for the volume changes that allow us to measure the blood flow.Quantitative SPECT/CT scintigraphyImmediately following the plethysmography study, a SPECT/CT study of the upper and lower limbs was performed on all patients 10 minutes following intravenous administration of 740 MBq Tc99m in-vitro labeled red blood cells[20]. SPECT/CT was performed using a nuclear medicine dual head variable angle gamma camera system equipped with a low power x-ray imaging system (Infinia & Hawkeye, GE Healthcare Technologies, USA).The x-ray imaging system is composed of an x-ray tube and a set of detectors located opposite the x-ray tube. They are mounted on the same gantry and rotate around the patient with the gamma detectors. SPECT and CT scan acquired sequentially with the patient remaining completely still between the scans. Resolution of the x-ray image is 1 mm, but localization images used for clinical reading are produced with a 1.69 mm pixel size. The x-ray images are acquired and reconstructed using the integrated workstation. The data is then transferred to the nuclear medicine database of the processing workstation (Xeleris, GE Healthcare Technologies, USA). SPECT images were acquired using a dual energy window session providing emission and scatter emission projection. The emission acquisition protocol was performed using a matrix size of 128 × 128, parallel head configuration, 180 degrees rotation per head, with an angle step of 3 degrees. Time per frame was 25 seconds. Reconstruction of SPECT data was performed on the processing workstation using scatter correction and attenuation correction (based on attenuation maps derived from the CT image). CT was also used as anatomical map for the functional NM data. The radiation dose to the patient (i.e. the combination of the radiation dose from the SPECT radiopharmaceutical and the radiation dose from the CT portion of the study) was estimated to 6 mSV.Calculations and statistical analysisBlood flow calculationsSPECT/CT data (volume and scintigraphic readings)1. Volumes and blood pool activity of the bone (including bone marrow) and soft tissue were determined using segmentation based on thresholds within a virtual cylinder consistent of 3–4 slices (slice thickness 7 mm) on the CT image. The height of the virtual cylinder on which measurements were performed was of approximately 1.4 – 2.1 cm. Precise calculation of the entire cylinder volume (VL) is provided by the CT component of the dual modality imaging procedure (see additional file 2).2. Bone volumes, including the bone marrow (VB), were derived from the CT scan using an in-house software that performs segmentation of the bone and soft tissue for each CT slice, subsequently creating corresponding regions of interest. The regions of interest are copied to the registered reformatted SPECT slices in order to correspond to CT voxel size (Figure 1).Figure 1SPECT/CT reconstruction with X-ray image showing volume (in ml) and counts in the bone (red) and total limb (green).3. Data regarding total limb and bone blood pool confined to the virtual cylinder (RL and RB respectively) was derived from scintigraphic data using the corresponding counts confined to VL and VB (raw data shown in additional file 3).4. The limb (and thus the)virtual cylinder is composed of soft tissue (mainly muscles and skin) and \"hard\" tissue (bone and bone marrow). The soft tissue volume (VS) and blood pool (RS) are calculated as follows: VS = VL-VB and the corresponding reading RS = RL-RB.Bone and soft tissue blood flow measurements (FB and FS, respectively) are based on the following considerations:1. Assume that a part of the leg or arm under examination is in a form of a cylinder.The cylinder is composed of two compartments: the bony compartment and the soft tissue compartment.2. We define the blood volumes (units are ml blood) within each compartment:υB – blood volume within bone compartment (including bone marrow).υs – blood volume within soft tissue compartment.υL – blood volume within the volume that is confined within the 100 ml cylinder.3. Based upon plethysmographic measurements, blood flow to the limb (in the selected area) is:FL= υL/min·100 ml tissue (ml blood/min·100 ml tissue)If we determine the portion of limb under examination (i.e. the cylinder) volume is 100 ml, then: υL ml blood pass through it in 1 minute.4. The main assumption is that in a resting state, the vasoregulatory systems are balanced, thus the blood flow in each compartment is constant, and the momentary blood flow can be calculated from the plethysmography. Say that a momentary blood flow through the 100 ml cylinder occurs within a time period dt(t→0), then:υL(dt) = υL·dt/min (note that dt and min are both time units, thus υLdt units are volume units – i.e. ml blood. It means that a momentary volume of υL·dt/min pass during a dt period of time)5. Since only RBC are marked, the scintigraphic readings are proportional to the blood pool within each compartment.Say that:RB – scintigraphic reading within the bony compartment in the virtual cylinder.RS – scintigraphic reading within the soft tissue compartment.RL – scintigraphic reading within the entire limb compartment.6. During the infinitesimal time period dt, the blood volume within the bony compartment are proportional to the scintigraphic readings.υB(dt) = (RB/RL)·υL·dt/min   (units are of volume-i.e ml blood)7. We can measure bone and soft tissue compartments volumes precisely from the CT scans: we take the 3–4 slices within the virtual cylindered shape limb portion under examination. We know the slice thickness (slice thickness 7 mm – the distance between the CT slices). The height of a virtual cylinder that its cross sectional area is equal to the slices' is 1.4 – 2.1 cm.8. The mean measured cylinder volume between the CT slices (VL) is 149 ml and 260 ml for the upper and lower limbs, respectively (see additional file 2). Since these volumes are quite small, we may say that within a 100 ml piece of this virtual cylinder the ratios between the bone and soft tissue volumes is preserved.9. Say that VB/VL is the relative bone volume of the virtual cylinder between the slices. Thus, in order to calculate the momentary blood volume within a 100 ml bonycompartment (υB(dt)100), we need to multiply υB(dt) by the ratio VL/VB:In this way:υB(dt)100=υB(dt)VLVB=(RB/RL)⋅υLdt/min⁡⋅VLVB=(RB/RL)⋅υL⋅dt/min⁡⋅VLVB10. If we assume, again, that in resting position the vasoregulatory systems are in balance, and the ratios VL/VB; υL/VL; and RB/RL remain constant, then we can correct to a minute flow by multiplying υB(dt)100 min/dt, which gives:FB=υB(min⁡⋅100)VB=υL⋅(RB/RL)⋅VLVB11. In a same way, the soft tissue blood flow per 100 ml soft tissue per minute is:Fs=υS(min⁡⋅100)VS=υL⋅(RS/RL)⋅VLVSStatistical AnalysisData are presented as mean ± SEM. Wilcoxon-matched-paired test, which is suitable for comparison between small groups, was used to compare between upper and lower limbs and their contralaterals. The level selected for statistical significance was set at P value < 0.05. Data were analyzed with Excel (Microsoft 2000, USA) and GraphPad Prism (version 3.0, GraphPad Softwarte, Inc., San Diego, CA).ResultsSix men and four women were evaluated. Subject's mean age, weight, height, body mass index (BMI, weight/height in m2), systolic and diastolic blood pressure and heart rates are presented in additional file 1. Raw volume measurements and scintigraphic readings depicted from SPECT/CT are shown in additional file 2. Briefly, the limbs' parts volumes that were examined (referred as a \"virtual cylinder\" in the methods section) were 149 ± 15, 149 ± 16 ml for right and left upper limbs, and 265 ± 22 and 256 ± 20 ml for the right and left lower limbs, respectively.At steady state blood flow measurements in the four limbs ranged between 5.5 – 6.5 and 1.87–2.48 ml per 100 ml of tissue per minute for BBF and SBF, respectively.Blood flows in each limb and its compartments are presented in additional file 3 and in figure 2. FB was significantly higher compared to FS in all four limbs (6.16 ± 0.65 vrs 2.37 ± 0.30 in RUL, p < 0.001; 5.9 ± 1.1 vrs 1.87 ± 0.20 in LUL, p < 0.001;6.28 ± 0.72 vrs 2.48 ± 0.28 in RLL, p < 0.001; 5.63 ± 0.72 vrs 2.36 ± 0.29 in LLL, p < 0.001, units in ml blood per minute per 100 ml). No significant differences in either bone or soft tissue blood flows were measured between right and left limbs, both in the upper and the lower extremities.Figure 2Bone (upper graph) and soft tissue (lower graph) blood flow in each limb (RUL-right upper limb, LUL-left upper limb, RLL-right lower limb, LLL-left lower limb). Blood flow units are expressed in ml/100 ml tissue·min-1 units, mean value for each column is marked with transverse line).DiscussionNormal growth, remodeling and repair of bone require delivery of nutrients and oxygen through blood flow to bone tissue [21]. Interruption of normal bone and soft tissue blood flow has been shown to be responsible for the development of severe and common health problems including diabetic ulcers and osteoporosis [22]. Nevertheless, only limited literature is available on the physiology and pathophysiology of bone and soft tissue blood flow, as compared to other tissues (e.g. renal, brain) that have been thoroughly investigated. Presently, we describe a novel method that which enables noninvasive BBF quantification in humans.Dual modality SPECT/CT imaging enables to quantify, with a high degree of precision, blood pool localized in a specific area of interest (in our study, the \"soft\" and \"hard\" tissues composing a limb). This method, combined with plethysmographic measurements, allows for quantification of blood flow in the tissues being evaluated. In this study, we showed that the BBF in the upper and lower limbs ranges between 5.5 and 6.5 ml per 100 ml of tissue per minute. These results are comparable with previously published data (e.g. Kubo et al., using 15O PET, showed that blood flow in femoral heads correlates with age and ranges between 1.7–6 ml/min per 100 g tissue) [23,5,4].Data from animal studies using labeled microspheres reveals a variation in BBF, in the range of 5–20 ml/min per 100 g, within different regions of the same bone sample [10]. This method requires animal sacrifice for a direct measurement of fluorescence or radioactivity assessment, thus cannot be comparable to methods used in humans.Our research has also shows that soft tissue blood flow (which is mainly a contribution of skeletal muscles) averaged between 1.87–2.48 ml/min per 100 ml tissue, which is comparable of PET measurements (range between 1.43–6.72 ml/min per 100 g muscle [5,4,24,2]. Noteworthy to mention that SPECT/CT-plethysmography revealed a trend towards a higher SBF in the dominant right upper limbs compared with the contralateral. Another interesting observation is that BBF was almost three times higher as compared to the adjacent SBF (per 100 ml tissue). Notice that while data in the literature is expressed as ml per minute per g tissue, ours is expressed as ml per minute per 100 ml tissue, since plethysmographic measurements are based on volume changes. This may be the reason for the small differences of our data from that described in the literature.Venous-occlusive plethysmography is an easy and accurate method for the assessment of total limb blood flow. It cannot however, distinguish between the various tissue components in the limb. It cannot also differentiate between soft tissue components blood flow (i.e. skeletal muscle and skin). In this study, however, an anatomical CT interface such as CT was manually fused with data derived from SPECT studies in order to accurately localize blood flow measurements to the bone. Fusion methods of separately performed functional and structural imaging data are based, as a rule, on extrinsic or intrinsic landmarks. Accurate localization of these markers is, however, difficult and requires considerable operator skill. These drawbacks are more prominent in aligning the nuclear medicine data, which suffer from inherent low resolution. Inaccurate registration of separately acquired scintigraphic and CT data may be due to differences in patient positioning between studies, as well as to differences in organ location and volume at the time of imaging [18]. Sequential acquisition of scintigraphic SPECT and CT data during a single imaging session using SPECT/CT overcomes these limitations by accurate localization of blood pool, represented by uptake of labeled RBC, as demonstrated on SPECT, to specific areas in bone and soft tissues, as delineated by the CT.Study limitations and clinical perspectives1. Plethysmography is a measurement technique that can only be applied to long bones and our method is, at present, only suitable for measuring limb blood flow. Future innovations involving a combination of SPECT/CT with different techniques for the assessment of regional blood flow (e.g., Dupplex) may allow for BBF measurement in flat/small bones.2. Present method did not allow for separation of bone marrow from cortical bone flow. The use of improved devices with higher imaging resolutions may allow in the future studying the specific blood flow distribution within the bone.3. When one comes to compare our results with those of the literature, he need to be aware that in the literature the bone blood flow results are presented in ml blood per minute per 100 grams bone tissueunits. We, however, present the results in units of 100 ml blood per minute per 100 ml bone tissue. In order to compare the values presented in the current paper with those in the literature, one need to correct the units that we used by dividing them in the specific gravity of the tissue. For example: we showed that the mean FB in the RUL is 6.16 ml blood per minute per 100 ml bone tissue. If the specific gravity of bone is (for example) 1.8 gr/ml, then the correction is 6.16/1.8 ml blood/min/100 gr bone.We raised this points in order to precede one expected question regarding our results: the fact that we found bone blood flow much higher than soft tissue blood flow. If you correct our results using the specific gravity of each tissue, you will find them quite similar to those in the literature.4. Our assumption that in resting-supine state is a steady state, where blood hydrodynamic characteristics between bone and muscle are comparable is essential, and the entire theory is based upon it. We could find neither support nor contradiction to this assumption in the literature, yet it seems only intuitive to us.5. Our measurements cannot differentiate muscle from skin blood flow, thus SBF refers to both.6. The resolution of the method, which supposedly determines a metric of blood flow in the bone, would be the smallest difference in blood flow this method can detect. The resolution is usually determined using a phantom simulating the procedure performed on the patient where all parameters are known. We do not believe we can determine this based on our method as no gold standard for osseous blood flow is currently available. The potential noise sources (factors that would influence the measured value that are not related solely to the blood flow in the bone) are:a. Tecnical factors related to the veno-occlusive plethysmography (incorrect placement etc.)b. Poor labeling of RBC.c. Patient motion during acquisition of nuclear medicine study.d. Misregistartion of CT and nuclear medicine portion of study due to motion.e. Metallic devices in bone.f. Operator error during processing of data.ConclusionBone blood flow is a physiologic characteristic that needs yet to be investigated in settings of clinical significances such as atherosclerosis, anti-hypertensive treatment, and osteoporosis, all conditions that are known to affect BBF. Here we offer it not as a replacement, but rather as additional method in the minute armamentarium of methods for BBF measurement.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsLD carried out the research designing, physiologic studies, data analysis and writing. ZK carried out the nuclear scan studies, participated in data processing and writing. OI participated in scan studies and data analysis. VM participated in data analysis. JS participated in scan studies and data analysis. GJ carried out the research designing, physiologic studies, data analysis and writing. All authors read and approved the final manuscript.Supplementary MaterialAdditional file 1Clinical characteristics of subjects (presented as mean ± SEM).Click here for fileAdditional file 2Raw data extracted from SPECT/CT. Volumes (in ml) and scintigraphic readings (in counts) of total \"cylinder\" and bone. RUL-right upper limb, LUL-left upper limb, RLL-right lower limb, LLL-left lower limb. VL and RL-entire \"cylinder\" volume and counts, respectively. VS and RS, volume and counts of soft tissue, respectively. VB and RB, volume and counts of bone compartment, respectively. Data is expressed as mean ± SEM.Click here for fileAdditional file 3Blood flow measurements, resistance and ratios. Blood flow units are expressed in ml/100 ml tissue·min-1 units. RUL-right upper limb, LUL-left upper limb, RLL-right lower limb, LLL-left lower limb. FL-total limb blood flow, FB-bone blood flow, FS blood flow in the soft tissue compartment. Data is expressed as mean ± SEM.Click here for file\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2527497.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2527497",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2527497\nAUTHORS: Jan Paul M Frölke, Almar WA Bruggeman, Frank PAJ Klomp, Joep LRM Smeets\n\nABSTRACT:\nThis case report describes about a young, male patient with persisting syncope during physical therapy for complex regional pain syndrome type 1 after metatarsal fractures.The patient was referred to the Emergency Department, where Brugada syndrome was diagnosed. A cardioverter defibrillator was prophylactically implanted successfully. After this procedure, there were no contraindications for resuming further physical therapy for his painful foot. No clear causal inference with Brugada could be drawn from the complex regional pain syndrome type 1 or physical therapy described in this case report. Hyperthermia may, however, occur during such therapy, which is associated with dysrhythmia in general.\n\nBODY:\nCase presentationBackgroundBrugada syndrome is an autosomal dominant disease that can cause syncope and sudden cardiac death in young individuals with a normal heart [1].It is characterized by an electrocardiographic pattern of complete or incomplete right bundle branch block and ST segment elevation in leads V1–V3. Around 20–25% of patients affected by this syndrome have genetic mutations in SCN5A, which cause a functional reduction in the availability of the cardiac sodium current in Brugada syndrome. The relative male preponderance of the phenotype, despite equal inheritance of the gene in males and females, has led to the speculation of a role for testosterone in the phenotype [2,3]. The disease may manifest first as cardiac arrest without any previous symptoms. The electrocardiographic pattern may be intermittent, requiring a pharmacological challenge with Class I anti-dysrhythmic drugs like flecainide to unmask the ST elevation. Dysrhythmias and sudden death generally occur during sleep or rest, but features of ECGs from Brugada patients also illustrate a cardiac sodium channel mutation. The arhythmogenicity characteristic of this mutation is revealed only at temperatures approaching the physiological range, and this suggests that some patients may be at greater risk during hyperthermia [4].According to the International Association for the Study of Pain (IASP) in 1994, reflex sympathetic dystrophy and its many synonyms are collectively referred to as Complex Regional Pain Syndrome (CRPS). Two types can be distinguished: CRPS-1 may occur in an extremity after even a minor injury or operation, and CRPS-2 may occur only when nerve injury is involved. In the acute phase, signs and symptoms of inflammation or ischemia within the affected extremity characterize CRPS. Although the clinical signs and symptoms of CRPS are well known, the underlying pathophysiology remains unclear despite the knowledge that disuse precedes CRPS in most cases. Many clinicians use the Bruehl criteria to make a proper diagnosis, in which continuing pain disproportionate to the provocative event is the most frequent symptom [5]. Sensory, vasomotor, or sudomotor changes with edema, combined with a decreased range of motion with motor dysfunction and trophic changes, must also be present for diagnosing CRPS-1.An accepted portion of the treatment consists of pain-contingent physical therapy to regain function and strength in the affected limb [6]. Pain Exposure Physical Therapy (PEP-therapy) is a different approach, in which functional physical therapy is combined with a cognitive – behavioral form of treatment for patients who are well-motivated to achieve functional targets. They are invited to use an active coping style to realize the treatment goal during five sessions spanning three months, and they are evaluated three months after the last session (time contingent). For this type of therapy, the use of analgesics, splints, and crutches is reduced. Additionally, treatment does not focus on the existing pain, and patients are instructed that pain should not be interpreted as a signal of tissue damage. This process disrupts the vicious circle of fear of movement, pain avoidance, and dysregulation of the sympathetic and central nervous system. Touching and moving the extremity decreases the fear of movement and increases the extremity's mobility. Clear treatment plans and empathic communication, as well as stimulation and positive feedback, are part of the standard plan. The patient is encouraged to attempt to achieve a realistic treatment goal for the next session by practicing the exercises at home. This approach has been shown to be successful in children [7]. No evidence has been provided yet for treating adult patients with CRPS-1 in this way, but we have been successfully treating patients with therapy-resistant CRPS-1 since 2005. A first report describing our results is in preparation.The purpose of this case report is to describe a patient in whom Brugada syndrome was unmasked during PEP-therapy for CRPS-1 after metatarsal fractures. This report is intended to increase the awareness of healthcare workers, and to help recognizing the risk of strenuous physical therapy in such patients. Hypothetical causal inference should, however, always be interpreted with caution. An association of CRPS-1 with Brugada syndrome has never been reported. It is therefore more likely to assume a relationship with this patient's active physical therapy involvement during the presentation of his symptoms.Patient HistoryA 41-year old, slender, Caucasian male was referred to our outpatient department two years after having sustained midshaft fractures of the second and third metatarsal bone due to the crash of a heavy tent-pole. Both fractures were successfully treated non-operatively. The pain in his foot, however, did not resolve with normal physical therapy, and six months after the accident he presented with symptoms of a CRPS-1 according to the criteria of Bruehl [5]. He was referred to our hospital's department of physical therapy to undergo PEP-therapy in order to regain function of the lower extremity and be able to walk without restrictions.He was otherwise healthy, and no co-morbidities or contra-indications were reported. He was not known with tobacco, drugs or alcohol abuse.ExaminationThis patient was seen in the department of physical therapy for his first treatment. Earlier treatment with pain contingent physical therapy, manual therapy, acupuncture, and action potential stimulation (APS) therapy had resulted in independent gait with orthopaedic shoes that were adapted to relieve his forefoot. The only pain medication he used was paracetamol. Pain relief and improvement of walking ability were the main reasons for which he sought treatment.Physical examination showed a healthy, slender white male with a height of 193 cm and a weight of 80 kgs without further abnormalities. He presented with symptoms of allodynia, hyperesthesia, vasomotor, sudomotor and trophic changes, a decrease in the active and passive range of motion for the right ankle and toes, dystonia during foot extension, and a burning sensation during walking. He avoided placing any weight bearing on his forefoot and presented a passive way of walking.After the painful examination, he was well motivated to begin treatment. His allodynia was treated with resumption of passive mobility during the first visit.TreatmentThe means provided in Pain Exposure Physical therapy (PEP-therapy) are articular movement, including traction, translation, manipulation, massage, desensitization, flooding (experience of the most fearful action), muscle lengthening, and guiding. Exercising of the normal movement patterns (functional training) with posture and relaxation exercises in front of a mirror accompany instructions, information, and advice. The patient is taught to experience the sensation of pain as non-functional and practice the exercises at home. Because pain experience is rational and rational attention is redirected from pain behavior to the ability to move, the patient's pain behavior will decrease as his self-confidence in his physical abilities increases. This progress is encouraged by positive feedback from two therapists and the patient's partner.Touching and massaging the painful foot was initiated with the explanation that the pain sensation is not harmful. The dystonic extensors of the foot were stretched, and the patient was instructed to do so himself. Touching and stretching caused a strong pain perception, which the patient expressed via facial grimaces and tightening of his muscles.At the end of this first foot and ankle treatment session, the patient suddenly blacked out, recovered spontaneously, and began hyperventilating. According to his accompanying wife, these symptoms were common for him and hyperventilation had already occurred several years earlier. Because no one had asked, the patient had not mentioned this fact before. This time, however, the patient was still uneasy after an hour of relaxation and breath control exercises. The physical therapist advised him to attend the hospital's Emergency Ward to further evaluate hyperventilation with syncope [8].OutcomeThis patient presented at the Emergency Ward with symptoms of blurred vision, tingling hands, and numbness of the right cheek. He had shown incidental hyperventilation throughout the last fifteen years. In one instance, he had collapsed when such an episode occurred while he was gardening. Cardiologic evaluation did not reveal any abnormalities at that time. He mentioned a recent sudden collapse in the shower with rapid spontaneous recovery without premonitory symptoms. He denied any other specific (cardiac) complaints such as chest pain, palpitations, or dyspnoea. His medical history was otherwise uneventful except for an inguinal hernia repair during childhood. A family history notes that his father survived a myocardial infarction at the age of fifty years. His fraternal twin brother had fully recovered from epilepsy during his early childhood. One younger brother died shortly after birth from severe spina bifida.Electrocardiography showed a Brugada-like pattern with typical ST segment elevation in the right precordial leads, which has been designated as type II Brugada (fig. 1) [9]. He was admitted for further evaluation. A flecainide provocation test was positive, with a shift from type II to type I, accompanied by symptoms of syncope (fig. 2). Provocation during an electrophysiological study with pacing of the right ventricle produced a ventricular flutter with recognizable symptoms of syncope, which lasted seven seconds and terminated spontaneously (fig. 3). An automated implantable cardioverter defibrillator (AICD) was prophylactically implanted, and the patient's twin brother was advised to undergo screening.Figure 1Electrocardiography at first presentation. Electrocardiography at first presentation with a high take-off ST segment elevation with upward concavity and positive or biphasic T-wave resulting in a saddleback configuration.Figure 2Electrocardiography after provocation with flecainide. Electrocardiography after provocation with flecainide with a shift from type II to type I.Figure 3Electrophysiologic study shows ventricular flutter. Electrophysiologic study shows ventricular flutter after pacing the right ventricle with 500 min-1.Although no contraindications existed for resumption of the physical therapy according to recent guidelines, the patient preferred to accept the current condition of his painful foot [10].DiscussionStrenuous physical therapy may evoke a variety of involuntary patient reactions like hyperventilation. When patients are hyperventilating and do not recover spontaneously or respond to common treatment strategies like paper bag rebreathing, syncope may occur. Any syncope requires thorough clinical evaluation to exclude serious disorders like cardiac arrhythmia and epilepsy [8].Brugada syndrome is a rare cardiac arrhythmia that may cause syncope in otherwise healthy young people. Several conditions producing Brugada-like electrocardiographic patterns, like the 'long QT syndrome,' should be borne in mind and excluded while making a diagnosis of the Brugada syndrome [1]. Treatment of patients suffering from Brugada is difficult, because pharmacological agents are not universally effective. State-of-the-art management involves the implantation of a cardioverter defibrillator. Additionally, symptomatic patients with inducible ventricular dysrhythmias and a positive family history, like the patient described in this report, should be considered for prophylactic implantation of a cardioverter defibrillator. This patient had a history of hyperventilation, which could have been a symptom of masked Brugada. However, hyperventilation may also be the cause of syncope. In addition, this may cause variant angina, which may in turn induce ischemic, life-threatening dysrhythmia. Our patient underwent painful PEP-therapy, which can be considered as strenuous exercise. Strenuous activities and sudden death have also lead to the identification of Brugada syndrome in athletes [1]. It has been assumed that strenuous exercise and its resultant increase in body core temperature can unmask the Brugada syndrome. This could have occurred during PEP-therapy for CRPS-1 in our patient [4].The patient did not mention any form of hyperventilation during our first contact. Due to the presence of his wife, he remembered his history of hyperventilation and syncope during the treatment session. No association between Brugada syndrome and any form of physical therapy or CRPS-1 has ever been described. Therefore, any unexpected symptoms of syncope during strenuous or painful treatments deserve further evaluation to prevent sudden death and must be interpreted as a \"red flag\" or serious warning sign. Most physical therapists are adequately equipped with professional knowledge to detect these warning signs in patients undergoing treatments, such as the case demonstrated in this case report.ConclusionStrenuous physical therapy may evoke a variety of involuntary patient reactions like hyperventilation. When patients are hyperventilating and do not recover spontaneously or respond to common treatment strategies like paper bag rebreathing, syncope may occur. Any symptoms of syncope during strenuous treatment or exercise deserve further evaluation to prevent sudden death.AbbreviationsCRPS: Complex Regional Pain Syndrome Type 1; ECG: Electrocardiogram; SCN5A: Sodium Channel voltage-gated type V alpha subunit; IASP: International Association for the Study of Pain; PEP-therapy: Pain Exposure Physical-therapy; AICD: Automated Implantable Cardioverter Defibrillator.ConsentWritten informed consent was obtained from the patient for publication of this case report and any accompanying images. A copy of the written consent is available for review by the Editor-in-Chief of this journal.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsJF did the initial clinical assessment of the patient and drafted the manuscript. AB wrote the part about CRPS diagnosis. FK treated the patient and wrote the part about CRPS treatmentJS treated the patient and corrected the part about Brugada. All authors read and approved the final manuscript.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2527500.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2527500",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2527500\nAUTHORS: Haridimos Markogiannakis, Dimitrios Theodorou, Konstantinos G Toutouzas, Georgia Gloustianou, Stilianos Katsaragakis, Ioannis Bramis\n\nABSTRACT:\nA 65-year-old woman presented with abdominal pain, weight loss, fatigue, and microcytic anemia. Esophagogastroduodenoscopy, until the second part of duodenum, was normal. Ultrasound and computed tomography demonstrated a solid mass in the distal duodenum. A repeat endoscopy confirmed an ulcerative, intraluminar mass in the third and fourth part of the duodenum. Segmental resection of the third and fourth portion of the duodenum was performed. Histology revealed an adenocarcinoma. On the 4th postoperative day, the patient developed severe acute pancreatitis leading to multiple organ failure and died on the 30th postoperative day.\n\nBODY:\nIntroductionAlthough the small intestine constitutes over 75% of the length and 90% of the mucosal surface of the gastrointestinal tract, small intestine cancer is rare and accounts for only 1% of gastrointestinal malignancies [1,2]. Adenocarcinoma together with carcinoid tumours are the most common histological types of primary malignant tumours of the small bowel but other, including lymphoma and leiomyosarcoma, may less frequently be encountered [1,2]. Adenocarcinomas are predominantly located in the duodenum [1,2].Duodenal adenocarcinomas represent approximately 0.3% of all malignant gastrointestinal tumours and the vast majority of them is found in the second portion of the duodenum [1,2]. Adenocarcinomas of the third and/or fourth portion of the duodenum, however, are very rare [1,2]. A case of adenocarcinoma of the third and fourth part of the duodenum is presented along with a literature review.Case presentationA 65-year-old Greek woman (weight: 80 kgr, height: 160 cm) with a free past and family history, presented with a two month history of intermittent abdominal pain, weight loss, and fatigue. The patient had two normal labours, while she did not smoke, consume alcohol or take any medication. Clinical examination was normal. Blood tests, including tumour markers such as carcinoembryonic antigen (CEA), a-fetoprotein (AFP), carbohydrate antigen 19-9 (CA 19-9), and carbohydrate antigen 125 (CA 125), were within normal limits apart from microcytic anemia (hemoglobin: 8.0 gr/dl). Esophagogastroduodenoscopy, until the second portion of the duodenum, was normal. Abdominal ultrasound (US) finding of a hypoechoic mass with irregular margins in the distal duodenum led to a contrast-enhanced abdominal computed tomography (CT) scan that revealed a solid mass (6 × 5 cm) in the third and fourth part of the duodenum (Figure 1). No proximity to the stomach, pancreas, second portion of the duodenum, duodenojejunal flexure, proximal jejunum or colon was demonstrated. Portal vein, celiac axis, superior mesenteric artery and vein, pancreatic and bile duct were also free of tumour. No lymph node or distant metastasis was identified. Due to US and CT findings, a second endoscopy was performed which confirmed an ulcerative, intraluminar mass in the third and fourth part of the duodenum. Histology demonstrated an adenocarcinoma.Figure 1Abdominal contrast-enhanced CT scan revealing a mass in the third and fourth part of the duodenum.Intraoperatively, a solid mass in the third and fourth part of the duodenum was identified. Local resectability of the tumour was meticulously investigated. Kocher's manoeuvre, mobilisation of the large intestine from the cecum to the midpoint of the transverse colon, mobilisation of the small bowel mesentery, division of the ligament of Treitz, and mobilisation of the third and fourth part of the duodenum along with the duodenojejunal flexure and proximal jejunum was performed. The stomach, pancreas, second portion of the duodenum, duodenojejunal flexure, proximal jejunum, and colon were free of tumour. Portal vein, celiac axis, superior mesenteric artery and vein, pancreatic and bile duct were also free of tumour. No lymph node or distant metastasis was identified. The neoplasm was, thus, considered resectable and a segmental resection of the third and fourth portion of the duodenum along with regional mesentery was performed. Intestinal continuity was then restored by an end-to-end hand sewn duodenojejunal anastomosis. Histopathologic evaluation of the resected specimen verified a moderately differentiated adenocarcinoma, measuring approximately 9 × 6 cm, infiltrating the duodenal wall, without any lymph node involvement (T3N0M0) (Figure 2). The specimen's margins were free of tumour.Figure 2Histopathology verified infiltration of the duodenal wall by an adenocarcinoma. (HEX25).On the 4th postoperative day, the patient developed severe acute pancreatitis and was admitted to the Intensive Care Unit. She developed multiple organ failure and died on the 30th postoperative day.DiscussionFew cases of adenocarcinoma of the third and/or fourth portion of the duodenum have been reported [3-11]. Causative factors have not been clearly identified [5]. Patients with familial adenomatous polyposis (FAP) and Gardner syndrome are considered to have a higher likelihood of developing duodenal cancer [12,13]. Patients who have duodenal polyps without a predisposing family history are also at an increased risk [5].Adenocarcinoma of the third or fourth part of the duodenum presents a diagnostic challenge. Symptoms may be absent until the tumour has progressed leading to a delay, of several months, in presentation [5,7,8]. The most common presenting symptom is abdominal pain; other clinical manifestations are nausea, vomiting, weight loss, anemia, fatigue, weakness, gastrointestinal bleeding or obstruction [4,5,7-11]. Diagnosis is also often delayed due to the vague and non-specific symptoms and the subsequent difficulties in performing the relevant investigation, while most patients undergo a number of diagnostic tests before surgical exploration [2,5-8,11]. Moreover, Cunningham reported that preoperative diagnosis was obtained in 6 of 13 such tumours [8]. The majority of these tumours have infiltrated through the duodenal wall at presentation with many being irresectable due to local and distant invasion [1,5,7-11].Diagnosis is usually made by upper gastrointestinal contrast study and endoscopy [4,5,7-11]. Their location, however, is often inaccessible to endoscopic viewing which may result in failure to diagnose them at endoscopy [5,8,10]. In some cases [6,8], US or CT findings have prompted repeat endoscopy with advancement deeper than usual, into the third and fourth duodenal portion, leading to diagnosis as in our patient. Patients may have at least one negative gastrointestinal contrast study before a positive result on a subsequent study [5]. Endoscopy with extra-long fibre optic scopes may be of benefit.Abdominal US is helpful for diagnosis and evaluation of vascular involvement [6,8]. Lesions appear as irregularly marginated hypoechoic masses but tumours smaller than 2 cm may not be detected [6]. Contrast-enhanced CT scan is useful for diagnosis and determination of malignancy and resectability [4,5,8,10,14]; however, tumours smaller than 2 cm may not be seen [14]. Features indicating malignancy are an exophytic or intramural mass, central necrosis, and ulceration while entirely intraluminal location indicates a benign tumour [14]. These features, though, are sensitive but non-specific. Vascular encasement, invasion of contiguous organs other than the head of the pancreas, distant lymphadenopathy, or metastases precludes curative resection [14]. Endoscopic US and magnetic resonance imaging (MRI), although not frequently used so far, are useful for the diagnosis, staging, and determination of resectability of these tumours.The treatment of choice is radical surgical resection [2,4-11]. The correct operation (pancreaticoduodenectomy, local excision or segmental resection) has been debated. Worldwide there is no general attitude on optimal surgical procedure in treatment of primary non-ampullary adenocarcinoma of the duodenum, especially for early stage disease. Due to its rarity, there is a lack of studies comparing local excision, segmental resection and Whipple procedure in the management of this neoplasm. Some authors prefer local excision or segmental resection while others duodenopancreatic resection, even in the case of early stage duodenal cancer with aim to avoid tumour recurrence, considering pancreaticoduodenectomy the procedure that satisfies the principles of an adequate curative cancer operation. Heniford [7], Santoro [9], and Barnes [10] reviewed 12, 33, and 67 patients with non-ampullary adenocarcinoma of the duodenum, respectively; no significant difference between pancreaticoduodenectomy, wide local excision or segmental resection was observed while tumour stage and resectability were the only predictive factors of survival identified. Moreover, in a study of 47 patients by Tocchi [5], no statistically significant difference was found between patients who underwent duodenal segmentectomy and those undergoing pancreaticoduodenectomy in terms of local recurrence, distant metastases, disease-free survival and overall survival. However, statistically significantly higher blood transfusion requirement, hospital stay, morbidity, and mortality were noted in the pancreaticoduodenectomy group. Factors influencing survival were TNM staging and, particularly, lymph node status. The authors suggested that duodenal segmentectomy may be preferred to pancreaticoduodenectomy because it is associated with low rates of morbidity and mortality, while exerting similar results [5]. It should be noted, though, that local recurrence after wide local excision of early stage disease has been reported [3]. Regardless of the type of surgery, curative resection results in a significant survival advantage compared to noncurative resection or nonoperative management [4,5,7-10]. In advanced disease, and particularly in duodenal obstruction, palliative resection, gastrojejunal bypass or duodenal stents may be indicated. In our case, segmental resection was performed. Our patient died due to severe acute pancreatitis leading to fatal systemic inflammatory response syndrome a complication that has also been reported by others [4].Little is known about the use of radiotherapy and chemotherapy, but most physicians utilise therapeutic strategies modeled on the management of large bowel cancer [2]. Cunningham observed no significant benefit of adjuvant chemotherapy on survival [8]. The prognosis is generally poor and depends on stage at presentation and surgical resectability [2,5-10]. Prolonged survival following complete resection is possible while irresectable disease has a very poor prognosis [4,5,7-11]. Selective, individualised use of pancreaticoduodenectomy, wide local excision or segmental resection as surgical treatment options seems to provide a rational approach to this rare disease.ConclusionAdenocarcinoma of the third and fourth part of the duodenum is very rare. Patients typically present with a long history of variable, vague symptoms and, many, with advanced disease. A higher degree of suspicion and a more aggressive, persistent investigation should lead to earlier treatment, higher curative resectability rate, and, therefore, better long-term results. The treatment of choice is radical surgical resection. The optimal surgical procedure, though, remains controversial. Multi-institutional prospective studies comparing pancreaticoduodenectomy with segmental resection or wide local excision as well as trials of chemotherapy and radiation therapy are needed to identify their impact on prognosis and suggest appropriate treatment recommendations.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsHM contributed to manuscript conception, research, acquisition of data, drafting and writing of the manuscript. DT contributed to organising and drafting of the manuscript, and critically revised the manuscript. KGT contributed to organising and drafting of the manuscript, and critically revised the manuscript. GG carried out the histopathologic evaluation and contributed to writing of the manuscript. SK contributed to organising, drafting and critical review of the manuscript. IB carried out the operation and contributed to acquisition of consent and critical review of the manuscript.All authors read and approved the final manuscript.ConsentWritten, informed consent was obtained from the daughter of the patient for publication of this case report and accompanying images. A copy of the written consent is available for review by the Editor-in-Chief of this journal.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2527566.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2527566",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2527566\nAUTHORS: Dorota Latek, Andrzej Kolinski\n\nABSTRACT:\nBackgroundSeveral different methods for contact prediction succeeded within the Sixth Critical Assessment of Techniques for Protein Structure Prediction (CASP6). The most relevant were non-local contact predictions for targets from the most difficult categories: fold recognition-analogy and new fold. Such contacts could provide valuable structural information in case a template structure cannot be found in the PDB.ResultsWe described comprehensive tests of the effectiveness of contact data in various aspects of de novo modeling with CABS, an algorithm which was used successfully in CASP6 by the Kolinski-Bujnicki group. We used the predicted contacts in a simple scoring function for the post-simulation ranking of protein models and as a soft bias in the folding simulations and in the fold-refinement procedure. The latter approach turned out to be the most successful. The CABS force field used in the Replica Exchange Monte Carlo simulations cooperated with the true contacts and discriminated the false ones, which resulted in an improvement of the majority of Kolinski-Bujnicki's protein models. In the modeling we tested different sets of predicted contact data submitted to the CASP6 server. According to our results, the best performing were the contacts with the accuracy balanced with the coverage, obtained either from the best two predictors only or by a consensus from as many predictors as possible.ConclusionOur tests have shown that theoretically predicted contacts can be very beneficial for protein structure prediction. Depending on the protein modeling method, a contact data set applied should be prepared with differently balanced coverage and accuracy of predicted contacts. Namely, high coverage of contact data is important for the model ranking and high accuracy for the folding simulations.\n\nBODY:\nBackgroundInformation on a pattern of long-range interactions, even very sparse, can be crucial in protein structure prediction[1,2]. Long-range interactions determine a protein structure mainly by placing the building blocks (helices or beta strands) at the appropriate distance between their residues.Whereas short-range information, such as the type of secondary structure, can be predicted in most cases with high accuracy (70–80%)[3] on the basis of a protein sequence, long-range contact predictions are still of rather low accuracy (at most 20%, according to the CASP6 results[4]). Such low accuracy of contact predictions, although well above random (by a factor of more than 11)[5], is not enough for the direct reconstruction of the 3D protein structure using for example distance geometry methods, common in NMR structure determination [6-8]. It is mainly due to the large number of false contact predictions[4]. Nevertheless, if we use this kind of data only as additional and complementary information in protein folding simulations, there is a good possibility that the force field will cooperate with the true contacts and discriminate against the false contacts, biasing the simulations towards the native-like structures. The main aim of this work was to establish to what extent this hypothesis is true.A frequently used definition of a term \"contact\" states that two residues are in contact when the distance between their Cβ atoms is smaller than 8 Å[4,9-11]. There are other, and perhaps more precise, definitions[12]. However, due to the conventions employed by most of the contact-prediction servers, we adhere to the Cβ-based definition. A term called \"sequence separation\" divides contacts into short-range (for which contacting residues are separated by at least 6 and at most 12 residues in a sequence), medium (12–24 residues) and long-range contacts (at least 24 residues)[4]. If we generate a symmetric, two dimensional matrix with rows and columns corresponding to a protein sequence, in which contacts between appropriate residues are binary depicted, we obtain a contact map. Such contact maps can be generated easily from a 3D protein structure, predicted from a protein sequence alone or obtained experimentally from NOESY spectra[13].Early methods for contact map prediction were based on the observation that residues which are closed in space often mutate in tandem. This was called correlated mutations (CMA) and was detected by the analysis of multiple sequence alignments[9]. The addition of other types of data, such as sequence conservation, predicted secondary structure and solvent accessibility, was necessary to improve rather weak performance of contact predictions based on CMA only[14,15]. The major breakthrough in contact prediction resulted from the application of machine learning methods such as neural networks[11,14,16]. Current methods also employ hidden Markov models often combined with threading[4,17], support vector machines [5,18] and genetic programming[19]. Detailed information about the selected contact prediction methods is shown in Table 1.Table 1Description of the selected nine CASP6 contact predictors.Contact predictorsMethodInput dataAccuracy [%]Coverage [%]BakerNeural networkContact predictions from 24 servers, predicted by JUFO secondary structure, amino acid properties, PSI-BLAST generated PSSM matrix, length of a protein sequence25.53.7PROFconFeed-forward neural network with back-propagationevolutionary profiles obtained using PSI-BLAST, predicted secondary structure and solvent accessibility, sequence conservation, biophysical features and \"complexity\" of residues24.23.6Baldi-group-serverRNN – Recursive neural networkPSI-BLAST generated sequence profiles, correlated mutations, predicted secondary structure, solvent accessibility21.92.9GPCPREDGenetic programming with self-organizing mapsPSI-BLAST generated sequence profiles, sequence separation17.42.7KarypisSupport Vector MachinesSequence profiles, correlated mutations from multiple sequence alignment analysis, sequence conservation, sequence separation, predicted secondary structure11.01.5KIASCMA analysisMultiple sequence alignment, hydrophobic packing of residues (data obtained from sequence conservation and hydrophobicity)11.01.7SAM-T04Neural networkAlignments, predicted secondary structure and propensities of residues in contact9.61.43Hamilton-Huber-TordaFeed-forward neural networkMutational correlations from multiple sequence alignments, biophysical class of contacting pair of residues, predicted secondary structure, sequence separation, length of protein sequence9.11.3CORNETNeural networkPSI-BLAST generated sequence profiles, correlated mutations and sequence conservation, sequence separation2.50.34Mean accuracy and coverage was evaluated for all NF and FR-analogy targets. Here, we defined the accuracy and the coverage in the same fashion as in[4] for top N/5 predicted contacts with a sequence separation of 12: Acc=Ntrue positiveNtrue positive+Nfalse positive⋅100%Cov=Ntrue positiveNnative⋅100%.The contact prediction approach is usually not optimized to find the closest homologues of a given protein[4], unlike in the case of comparative modeling or fold recognition methods [20,21]. Consequently, it could be more sensitive in detecting more distant structure similarities with respect to traditional fold recognition or comparative modeling approaches[4,14]. This seems to be especially useful in the structure modeling of new folds, for which meta-predictors (Bioinfo[22] etc.) based on the consensus of fold recognition methods do not yield any reliable templates. Therefore, our work focuses on the contact-based structure prediction of targets from the two categories defined in CASP6[23]: New Fold (NF) and Fold Recognition – Analogy (FR/A), for which producing a reliable template structure was extremely difficult or impossible.Despite the significant potential of contact prediction, typical de novo protein structure prediction methods still do not use this kind of data regularly in modeling pipelines, except for a few examples[4,24-29] Probably the main reason is the difficulty in implementation of such low accuracy data in the reliable protein structure prediction schemes. As we mentioned before, contact maps predicted on the basis of protein sequences are not accurate enough for a direct and thus very fast and simple reconstruction of 3D protein structures. The CABS algorithm, which is used in this work, enables incorporation of different kinds of structural data, in the form of distance or angular restraints, not necessary accurate, such as for example chemical shifts[30]. This feature of CABS encouraged us to test the low accuracy contact data at various stages of protein structure modeling.We focused our test on CASP6 predictions because many contact predictors, after moderate success in the CASP6, abandoned the development of their methods, probably because of the limited usage of such data by structure predictors, which resulted in slightly worse performance in CASP7[31], in which the targets were apparently more difficult (many with good templates difficult to detect) as well. Moreover, the Kolinski-Bujnicki group did not take part in CASP7 and thus reproducing the same conditions, as those typically applied in CASP competitions, could not be possible. The aim of this work is to propose a novel, predicted-contact-assisted approach to protein structure prediction which could be used in high-throughput modeling pipelines which are especially inevitable in CASP – like, large-scale experiments.An important issue in de novo modeling, enhanced by contact prediction, is to establish a minimal level of accuracy and coverage of contact data that could be still useful for structure prediction. Typically, increasing the coverage to obtain the majority of the important contacts decreases the average accuracy of the whole prediction. This interplay between coverage and accuracy seems to be crucial for successful protein structure prediction (see Results and Discussion). It is important to note that most of the false contact predictions are shifted by one or two residues with respect to the true native contacts and the minority of all predicted contacts are completely wrong predictions (see Table 2 in Methods). As it is shown in this work, these moderately shifted contacts can still be useful, provided that they are implemented as very soft biases in the folding algorithms, as in the CABS algorithm[32]. This fact should be taken into consideration while developing contact prediction methods.Table 2Results of the contact-based ranking of Kolinski-Bujnicki's models for NF and FR/A CASP6 targets.Set of contact dataNumber of predicted contactsAccuracy δ = 0 [%]Coverage δ = 0 [%]Accuracy δ = 2 [%]False contacts (δ > 5) [%]ΔRMSD [Å]Spearman corr. coeff. (GDT-TS)Spearman corr. coeff. (RMSD)N/2 top-scoring contacts from each of the best two predictors (a)N19.5812.9149.3719.591.069-0.3000.329N/2 top-scoring contacts from each of the best three predictors(b)1.5 N17.2115.2246.9021.131.453-0.3210.275Consensus of the whole data from the best three predictorsN/223.949.1253.1717.931.393-0.3250.344N18.9014.4449.1720.351.387-0.3330.3401.5 N15.5317.8346.6022.151.453-0.3290.288Consensus of the whole data from the best five predictors(c)N/223.788.9752.4717.671.272-0.1960.360N19.7815.0751.0120.041.498-0.3500.3481.5 N16.6418.9147.7022.111.443-0.3380.400Consensus of the whole data from all nine predictorsN/224.989.5051.6920.561.252-0.2380.432N20.6315.7450.5821.851.365-0.3330.4001.5 N18.0820.3749.3523.391.322-0.3420.440(a) Predictors: Baker and PROFcon(b) Predictors: Baker, PROFcon, GPCPRED(c) Predictors: Baker, PROFcon, GPCPRED, Karypis, SAM-T04Spearman correlation coefficients together with the average difference of Cα RMSD of the best model and the first ranked model (ΔRMSD) were computed for each set of contact data. The average ΔRMSD of the Kolinski-Bujnicki group was 1.484 and the Spearman correlation coefficient: 0.213 (RMSD) and -0.138 (GDT-score). These values were improved for every set of contact data. Apart from the mean accuracy and the mean coverage of contact data, averaged over all targets, we also present the percentage of semi-accurate contacts (shifted with respect to the real by at most two residues) and totally false contacts (shifted by more than five residues).De novo modeling could be enhanced by contact prediction in various ways. First of all, contact maps could be used to generate restraints which bias either de novo folding simulations[28] or refinement simulations which start from preliminary models obtained via more straightforward simulations (see Methods). In the other approach, predicted contacts can be used as a part of various scoring functions for ranking protein models generated by any de novo or fold recognition method[27,33]. In this work, we verify whether these three approaches could enhance protein structure prediction carried out using the CABS modeling tool[32] and if so, to what extent. Our results are compared to the CASP6 predictions of the Kolinski-Bujnicki group[34] who used the CABS algorithm (de novo method)[32] and the Frankenstein-3D (a fold recognition tool)[35] without any information about predicted contacts.MethodsPreparation of data sets for predicted contactsIn this work we used contact data provided in CASP6 by nine predictors performing best and average: Baker group[4], GPCPRED[19], Hamilton-Huber-Torda group[36], KIAS[37], Karypis group[5], SAM-T04[38], baldi-group-server[39], CORNET[40] and PROFcon[11] (see Table 1). Contacts provided by the average-performing groups were added to our data because in the real CASP experiment the final performance of each group would not be known a priori and it would rather be impossible to choose only the best contact predictions for the structure modeling. We used contact data only for targets from NF (New Fold) and FR/A (Fold Recognition Analogy) categories because in the remaining CM (Comparative Modeling) and FR/H (Fold Recognition Homology) categories template structures could be found relatively easy in the PDB and thus the contact prediction did not provide any additional and valuable information.Original data sets contain different numbers of predicted contacts (several, N/2, N; where N is sequence length, or even a few thousands) and assume different minimal sequence separations (from 1 to 12). Such heterogeneous data had to be converted into sets of restraints which could be most beneficial for protein structure prediction. For this reason, we tested different data sets (see Table 2) which contain either contacts from all nine predictors or from the selected best-scoring predictions. All the nine contact predictors used different methods and any consensus selection could perhaps cover-up the imperfections of a single method and thus improve the accuracy of the combined data. Such an approach was successfully exploited by Baker et al. in CASP6[4]. We compared the results of the modeling supported by the consensus contact data with the results obtained using the data set of N/2 top-scoring contacts provided by each of two or three best predictors. In the latter approach, instead of processing the preliminary consensus which could filter out the false contacts, we allow the CABS force field to cooperate with the large sets of predicted contacts in order to obtain some kind of consensus during the simulation. Moreover, in this approach, top-scoring contacts, which were predicted by more than one predictor, were placed in a set of restraints more than once. Consequently, the strength of bias generated by these contacts became somehow multiplied.When generating all tested sets of contact data we chose only those contacts from each predictor for which the contacting residues were separated in the sequence by at least twelve residues. It has been observed in simulations (data not shown) that the short-range contacts with sequence separation below 12 could be obtained using solely the CABS force field due to its well optimized short-rang potentials. Such short-range contacts are typically responsible for the loop formation at the end of a helix or a beta strand. The CABS algorithm is able to reproduce such ending of secondary structure elements itself, provided the predicted secondary structure is of a reasonable accuracy. This was tested in various cases of loop-modeling[32]. Ignoring the predicted contacts with sequence separation below 12 eliminates also their over-expression in the data sets, which is caused by the fact that they are typically much easier to predict than the medium and long-range ones[14]. Those medium and long-range contacts seem to be more important in the structure modeling because they are sometimes difficult to reproduce in template-free CABS simulations.For the consensus sets of contacts we purposely ignore the fact that some predictors optimised their contact prediction to the specific number of contacts (e.g. N/10[19,36] or N/2[11]). The reason for this was the limitation of their methods which could score some relevant contacts very low. Consequently, a minor improvement in accuracy of the final set of contacts obtained by rejecting the low-scoring contacts could be at the serious expense of coverage.Contact-based ranking of CASP6 modelsWe employed the sets of contact data described in Table 2 in various protein structure modeling procedures. First of all, we used the predicted contacts for ranking five models for each target submitted to the CASP6 server by the Kolinski-Bujnicki group. The purpose of this test was to verify whether the selection of the best models (far from perfect in CASP6) could be improved by scoring these 5 submitted models with predicted contacts, i.e. if the modified scoring function would correlate better with the RMSD of the models than the scoring function based solely on the CABS force field and the results of clustering the simulation trajectories. We tested different kinds of the contact-based component of the scoring function (linear, square root and quadratic dependence of the deviations between the observed and predicted Cβ-Cβ distances). The highest correlation coefficients were observed for the linear scoring function and all the results presented here are obtained using the following form of this function:(1)score(i,j)={0d(i,j)≤8Åf⋅d(i,j)d(i,j)>8Åwhere: f is a scaling factor (here equal to 1.0), d(i,j) is the difference between the observed Cβ-Cβ distance and the reference distance of 8 Å (a standard cut-off distance in the contact definition[4]), i and j are residues predicted to be in contact. The final score for the given structure and the predicted contact set is computed as a sum of all i,j pairwise scores.REMC simulations of CASP6 targets with contact-based restraintsThe usefulness of the predicted contacts in protein modeling was also tested in the structure refinement of the final model and in contact-assisted de novo folding. In both cases we used the CABS modeling tool[32]. The CABS algorithm predicts protein structures on the basis of their sequences. It employs a simplified lattice representation of a protein. Each residue is represented by four centers of interactions: Cα united atom, Cβ atom, a united atom in the side-chain center of mass and a united atom representing the peptide bond, located in the center of a Cα-Cα pseudobond. The conformational space of a model protein is explored by the Replica Exchange Monte Carlo method (REMC), a very efficient technique for finding the global energy minimum [41]. The conformational energy of a protein is evaluated by several knowledge-based potentials which bias the model chain towards protein-like conformations. The force field includes long-range orientation-dependent contact-type potentials, short-range sequence-dependent potentials and a hydrogen bonding potential. The details of the CABS force field design and the description of its applications in generalised comparative modeling, folding pathway prediction, docking and de novo modeling and modeling supported by sparse experimental data can be found in other publications[30,42-44].Apart from the protein sequence, for a better performance, CABS also requires some information about the likely secondary structure coded in a three letter code (E-beta structure, H-helix, C-coil or loop). Such a secondary structure is typically predicted quite accurately by different servers (e.g. PSIPRED[45]). In order to reproduce the same conditions of the folding simulations as those applied by the Kolinski-Bujnicki group during the CASP6, except for the additional contact data, we used the same predicted secondary structure and the same version of the knowledge-based potentials as employed during CASP6.The structure refinement supported by the contacts predicted was conducted in several low-temperature REMC simulations with different sets of parameters and different sets of contact data. In every simulation the five protein models submitted to the CASP6 server by the Kolinski-Bujnicki group were used as the initial replicas. The lowest temperature starting replica was the first model submitted in CASP6 by this group. Employing these initial conditions, we refined the structures obtained by the CABS-based de novo modeling combined with evolutionary information from Frankenstein-3D[34].The predicted contacts were used in the refinement as restraints imposed onto Cβ-Cβ distances in the form of square root potential, tested previously[32]:(2)E=∑kf(dkcurrent−dkpredicted−1)fordkcurrent>dkpredicted+1Ek=0fordkcurrent<dkpredicted+1Here, k is a number of a contact between residues i and j and Ek – is a component of the contact potential associated with this contact; dkcurrent-is a current distance between Cβ atoms of residues i and j; dkpredicted – is a predicted maximum distance between these atoms, set to 8 Å; f is a scaling factor (typically between the value of 0.1 and 1). The energy component of the contact potential is added to the total conformational energy only when its value exceeds the certain cutoff depending on the expected quality of the predicted contacts. This feature of the CABS force field is very useful when the restraints data is sparse and inaccurate, because it enables to incorporate restraints into the simulation only when the protein conformation seriously disagree with most of them and discarding them when only a small fraction of restraints (typically 20–30%), which could be false anyway, are not satisfied. The same type of restraint potential was used in de novo folding simulations. Contact-assisted de novo folding simulations were started from random polypeptide conformations. The first stage of de novo folding was processed without contact-based restraints, which were implemented only in the second stage of the simulation. The preliminary simulation with the CABS force field only facilities a very efficient search for the local minima of the free-energy in the enormous conformational space, especially in the case of large protein structures, without any restrictions from restraints [30]. Typically, in the preliminary simulation, a random polypeptide chain yields a protein-like secondary structure, but its overall topology is not set.Preliminary testing and optimisation was processed for the limited number of 14 CASP6 targets (T0198, T0199-3, T0209-1, T0212, T0215, T0230, T0235-2, T0239, T0248-1, T0262-1, T0272-1, T0272-2, T0280-2, T0281) in the series of contacts-supported de novo folding and refinement simulations. After these testing simulations we chose the best performing contact data set on the basis of the RMSD of the obtained models. It was the contact data obtained from the best two predictors (Baker and PROFcon – see Results and Discussion, Table 3). We used this contact data set in simulations for all the remaining targets from NF and FR-analogy categories.Table 3Comparison of different approaches to contact-based modeling tested in this work.Set of contact data(a)Number of predicted contactsavg. Cα RMSD [Å]Contact-based rankingDe novo foldingRefinementFirst (c)Best (d)First (c)Best (d)N/2 top-scoring contacts from each of the best two predictorsN9.588.937.537.697.10N/2 top-scoring contacts from each of the best three predictors1.5 N9.968.698.028.237.36Consensus of the whole data from the best three predictorsN/29.828.157.148.287.17N9.808.827.358.117.301.5 N9.819.218.038.687.73Consensus of the whole data from the best five predictorsN/29.838.927.928.447.02N9.918.947.657.897.211.5 N9.798.707.577.717.11Consensus of the whole data from all nine predictorsN/29.709.267.988.066.44N9.778.797.527.636.411.5 N9.648.427.078.026.99(a) The same as in Table 2.(b) RMSD computed only for models ranked as first.(c) RMSD component computed only for that model of each target which was likely to be selected as the first model (the most probable), because it was a centroid structure of the most populated cluster obtained in a simulation.(d) RMSD component computed only for the best model of each target (a centroid structure of a cluster which was most similar to the native structure).The performance of each method is presented as the Cα RMSD of the final protein model averaged over 14 targets selected, for the sake of shortening computation time, from NF and FR/A categories (i.e. T0198, T0199-3, T0209-1, T0212, T0215, T0230, T0235-2, T0239, T0248-1, T0262-1, T0272-1, T0272-2, T0280-2, T0281). The average Cα RMSD of the first models for this subset of 14 CASP6 targets for the Kolinski-Bujnicki group was 10.27 and we observed improvement of this value for every method and every contact data set. The lowest average RMSD was obtained in the case of the refinement simulations with contact restraints based on the data from the best two predictors.In de novo folding and in the refinement simulations the final protein models were selected from the CABS trajectories as centroids of the most populated clusters using the hierarchical clustering procedure, described elsewhere[46]. The rebuilding of all-atom structures from the CABS Cα only traces was done using the BBQ program from the BioShell package[46].Results and discussionPredicted contact-based ranking of CASP6 modelsThe first question asked was whether the ranking of the models provided by the Kolinski-Bujnicki group could be improved by the ranking based on the contact-dependent scoring function. Because of the limited size of the data set which was too small to assume normal distribution, we employed the Spearman rank order correlation instead of the commonly used Pearson correlation. Spearman correlation coefficients were evaluated for each target separately and then averaged over all targets to obtain the final correlation coefficient for a given set of contacts. The Spearman rank-order correlation was used for example by Feig et al. in the evaluation of CASP4 protein models obtained by different modeling methods, from comparative modeling to de-novo folding[47].The contact-based scoring function improved the ranking of models submitted by the Kolinski-Bujnicki group for every set of contacts considered in this work. As it is shown in Table 2, the more diverse (from different predictors) and the larger data set is considered, the higher correlation coefficients are observed. The highest RMSD-based correlation coefficient (0.440) is for the set of 1.5 N consensus contacts derived from all the nine tested predictions. The correlation coefficient between the RMSD-based ranking and the ranking provided in CASP6 by the Kolinski-Bujnicki group was only 0.213, so the contact-based ranking led to a significant improvement of the best model selection.In the case of protein models which differ significantly from the native structures, the GDT-TS (global distance test) measure is used more often in evaluating the results than the RMSD values[47]. Though RMSD correlates with GDT-TS values, GDT-TS is more sensitive in distinguishing different protein topologies than RMSD[47]. In Table 2, apart from the correlation coefficients computed using the RMSD measure, we also present correlation coefficients between the GDT-TS-based ranking of protein models and the ranking based on the predicted contacts. The GDT-TS scores were obtained using the TMscore program[48]. If a protein model is similar to a native structure, the GDT-TS score is high. Thus, the best correlation would be for the Spearman correlation coefficient equal to -1.The contact-based ranking of models correlates better with the RMSD than with the GDT-TS. The correlation coefficients based on the GDT-TS ranking were obtained in the range from -0.350 to -0.196 (for the Kolinski-Bujnicki group in CASP6: -0.138). Such results, though obtained using a different testing set, are comparable with those of Feig et al. (correlation coefficients for assessing CASP4 models: from -0.407 to -0.218, depending on the type of a physical energy function used)[47].Apart from the correlation coefficients, we also present the ΔRMSD value, which is the difference in Cα RMSD between the first ranked model and the best model, averaged over all targets (see Table 2). Such measure of the performance of ranking methods is widely used[49]. In this case, we improved Kolinski-Bujnicki's results in the best case by about 0.4 Å (from 1.484 to 1.069).Correlation coefficients and average ΔRMSD values were compared in Table 2 with accuracy and coverage of the given contact sets, averaged over all targets. The accuracy of the data was computed considering the δ analysis by Ortiz et al[28]. The value of δ, defined as a maximum shift in the predicted contacts compared to the real ones, was chosen as equal to 0 (accurate contacts), 2 (semi-accurate contacts) and more than 5 (wrong contacts). Generally, the coverage is higher when more contacts are taken into the set but in that case the accuracy decreases. Also, when the coverage improves, the number of completely wrong contacts (shifted by more than 5 residues) increases and the fraction of semi-accurate contacts decreases. The maximum number of shifted residues (δ = 5) which could be still useful in structure modeling is chosen independently of the type of the secondary structure of the target protein. However, it is worth noticing that a 5-residue-long shift in a contact set could be more destructive for extended beta-type structures than for compact helix structures.Generally, models are better ranked (higher correlation coefficients) when the contact data are of high coverage (e.g. 1.5 N contacts instead of N/2). However, if we examine only the difference between the first ranked and the best model (ΔRMSD), better results are obtained when the accuracy is high, but not the coverage (e.g. sets of N/2 and for N contacts instead of 1.5 N).In most cases (see Figure 1a), the Cα RMSD of the models selected as first according to the contact-based criterion was lower or the same as the RMSD values of the Kolinski-Bujnicki's first models. Only one model out of 14 in the FR-analogy category and two models out of 10 in the NF category were worse than the first models selected by the Kolinski-Bujnicki group. We managed to improve the prediction of five models in the FR category and two models in the NF category. The quality of the remaining models did not differ significantly from the CASP6 results.Figure 1Comparison of the CASP6 results of the Kolinski-Bujnicki group with post-CASP contact-based modeling. Results of the contact-based ranking of Kolinski-Bujnicki's models from CASP6 is shown in (a). The scoring function was based on the contact data set from the best two predictors (Baker and PROFcon). The accuracy of the contact data used for scoring 5 models of each target is plotted against the RMSD of the model ranked as first in CASP6 by the Kolinski-Bujnicki group (green squares) and against the RMSD of the protein model ranked as first by the contact-based scoring function (red lines which join corresponding points). Results for NF and FR/A categories are presented separately. The most significant improvement is observed in the case of FR/A targets with the accuracy of the contact prediction range of 15–30%. In a similar way the results of the refinement simulations are presented in the right-hand panels (b). The refinement simulations performed better than the post-simulation ranking of the models.The two targets from the NF category for which the contact-based scoring function failed in the selection of the best models were T0241-2 and T0216-2. The reason for such weak results is that all models provided by the Kolinski-Bujnicki group were of a very poor quality (RMSD = 16–24 Å). In such cases, agreement with the predicted contacts may not correlate with the quality of models.The most significant improvement in the models' ranking was obtained for targets for which the accuracy of contact data was in the medium range of 15–30%. Below this range, the contact-based selection did not improve the quality of models significantly, but, and this is perhaps more important, did not worsen the results. Above this range the improvement in most cases was not significant, probably because such high accuracy of contact data means that contact prediction was straightforward. Such high accuracy of contact prediction was observed only for FR targets, for which some similar protein structures could be detected by a sensitive fold recognition method and was certainly detected by Frankenstein-3D used by the Kolinski-Bujnicki group. Thus, contact prediction in these cases did not provide any additional information.As the results of the contacts-based scoring of the Kolinski-Bujnicki's models were encouraging, we tested the scoring function for models submitted by other groups in CASP6 (see Figure 2). While, the contact-based scoring function is able to distinguish between completely wrong protein models (GDT-TS < 40) and those nearly native (see Figure 2) it needs to be combined with other kinds of data or scoring functions while accessing the models much closer to the near-to-native structures (with GDT-TS > 40).Figure 2Contact-based scoring of NF and FR/A models submitted as first by all groups in CASP6. For each model, the final score computed using the scoring function, which was based on the contacts provided by the Baker group and PROFcon, was plotted as a function of GDT-TS. Although most of wrong or low quality models (with GDT-TS < 40) could be discarded by the contacts based scoring function, it seems inevitable to use some additional discriminating tools for assessing models with GDT-TS > 40.Refinement of CASP6 models with predicted contactsPredicted contacts are more valuable in refinement simulations using CABS than in the post-simulation ranking of the most probable models (see Figure 1). It is a consequence of the fact that, in contrast (for instance) to Rosetta[50], CABS was developed not to generate many distinct protein conformations and rank them after the simulation, but rather to bias the protein conformation towards its single global free energy minimum. The improvement of protein models after the CABS refinement was significant in the case of targets from the FR/A category (11 targets out of 14; even by about 6 Å). In the case of NF targets we managed to decrease the RMSD value of 7 out of 10 targets, but by no more than 2–3 Å. Only 3 models out of all FR/A targets were worse than those selected as the first by the Kolinski-Bujnicki group (by about 0.6–1.5 Å) and only one model out of all NF targets (by 1.4 Å). The accuracy of the contact data for these 4 targets was either below 15% or above 30% and this confirmed our hypothesis formulated in the previous section that the most useful contact predictions are typically of 15–30% accuracy.In Figure 3, we compared the real contact maps with predicted contact maps of selected targets. We also compared the contact maps generated from models before and after the refinement simulations. It is worth noticing that the contact map of the final refined protein model does not necessarily overlap entirely with the predicted contact map used for generating restraints for simulations. It is the consequence of the way the restraint potential is constructed (see Methods section), i.e. to allow a fraction (typically 20–30%) of restraints to be ignored during refinement simulations. Such a form of the restraint potential is especially useful when only a small fraction of predicted contacts is accurate.Figure 3Contact maps presenting results of fold-refinement simulations. The contact maps of the selected CASP6 targets are presented in (a). In the upper triangle in each contact map real (red) and predicted (green) contacts are compared. In the bottom triangle a contact map of Kolinski-Bujnicki's first model (blue) is superposed on a contact map of the final model obtained after the refinement simulations (grey). In most cases we observed improvement of the contact maps for models after the refinement. Some accurate contacts were rebuilt by the CABS despite not being preliminarily predicted (T0209-2). Some falsely predicted contacts in diffused clusters were not observed in the final model (T0281). Predicted contacts in dense and numerous clusters were observed almost in all cases (A and A' in the T0272-1 contact map), contrary to diffuse sparse contact clusters. (b) Lower triangles, contact maps of the T0272-1 models obtained after the simulations with restraints based on the data sets (upper triangles) with either the A or B group of contacts modified. (1) Reduction of the influence of restraints based on the A group of contacts on the simulation with respect to the original contact data in (a) by diffusing these contacts. Intensification of the effect of B contact-based restraints by increasing the number of these contacts (2) and by increasing the scaling factor in the restraint potential corresponding to these contacts(3).In most cases we observed that contact maps of the refined models were more accurate than those of Kolinski-Bujnicki's original models. We also noticed that some contacts which were visible in the contact map of the refined model were not provided by contact predictors (see selected contacts in the T0209-2 map in Figure 3a). This is the consequence of the interplay between the CABS force-field and the contact-based restraints in which CABS plays the key role. For example, the CABS tool is able to reach a protein energy minimum by rebuilding the conformation of one part of the protein chain to satisfy interactions defined either by its force field or by contact-based restraints, imposed on the other, even distant part of the protein (see T0280-2 in Figure 3a).Most of the contacts in the Kolinski-Bujnicki model of the T0198 target were uniformly shifted by a few residues with respect to the real contact map (see Figure 3a). Thanks to rather precisely predicted contact-based restraints the refinement simulation managed to shift these contacts back to the native-like pattern. However, the Cα RMSD of the final model did not improve significantly with respect to Kolinski-Bujnicki's results (from 9.8 to 8.1 Å). As it was tested by Ortiz et al.[28], if all contacts are slightly but uniformly shifted with respect to the native ones, the RMSD of the overall protein model is affected only to a minor extent (1–2 Å). Our results for the T0198 target confirm this observation.Most of the predicted contacts for all targets are precise if they are accurate. The main problem with this kind of data is not the shift of the contact data but the existence of completely mispredicted contacts, very distant from the native ones in the contact map (see the A group of contacts in the T0201 map and A and B in the T0272-1 map in Figure 3a). Such mispredicted contacts, if satisfied in the simulations, could severely worsen the quality of the final model[4]. Of course, if such contacts disagree completely with the CABS-only generated energy landscape or satisfying them would require serious rebuilding of a conformation, not possible in the low-temperature refinement, they may not be observed in the final contact map (e.g. in the case of the A group in the T0201 contact map in Figure 3a). In other cases, their influence could be destructive (see T0280-2 in Figure 3a).The effect of such false contacts could be reduced, as we observed on the basis of contact maps presented in Figure 3, by diffusing the uncertain contact data or decreasing their number. If some predicted contacts are concentrated and, what is more important, numerous in one area of the contact map, the component of the restraint potential corresponding to these contacts is high and significantly influences the folding process. Consequently, concentrated contacts are likely to be observed in the final models (see the dense groups of contacts in T0198, T0209-2 and T0280-2 contact maps in Figure 3a). Of course, this could be useful if contacts are accurate (the T0198 case). However, if they are suspected to be false, like in the case of the T0280-2, they should not be incorporated in the simulations in the form of populated and dense data set. On the other hand, if contacts in a certain area of the predicted contact map are diffuse, they are less numerous and, therefore, they modify the energy landscape to a lesser extent. More diffuse and less numerous contacts could even remain undetected in the final model (see selected contacts in the T0281 contact map in Figure 3a).To confirm this observation, we tested several methods for reducing the influence of false contacts on structure modeling (see Figure 3b). We selected the T0272-1 case, for which two distinct false clusters of contacts were predicted. The A group of contacts forms a dense and numerous cluster and the contacts in the B group are less numerous and less close to each other (see Figure 3a). After the refinement simulation in the final model we observed mainly contacts from the A group (depicted as A'). The B group was barely visible (depicted as B'). To reduce the A group of contacts, the scaling factor of the potential component computed for this group of contacts could be decreased, the number of these contacts could be reduced and finally they could be more diffused (see A\" in Figure 3b-1). To obtain the B contacts in the final protein model, on the other hand, the scaling factor should be increased (see B\" in Figure 3b-3), the number of these contacts should be larger (see B\" in Figure 3b-2) or they should form a more dense cluster. Our results of additional simulations for T0272-1 (see Figure 3b) confirmed that if a cluster of predicted contacts is more numerous and dense, it is more probable that we observe it in a final refined model. Decreasing the scaling factor of the restraint potential affects the simulation, but to a minor extent only.Another important issue which we encountered is the frequent inconsistency of contacts obtained by the combination of different methods. For example, (see Figure 3a), a large group of contacts in the T0201 contact map marked as B, was obtained by combining contacts predicted by two methods. In the refined model this group of contacts was reduced to those contacts which were consistent (B'). As it was mentioned above, the specific form of the restraint potential enables us to suppress the effect of some inconsistent contacts, provided that the scaling factor is not too high. In such a way, accurate contacts could be distinguished from the false ones due to some geometry or physical or knowledge-based criteria defined in the CABS force field. This ability of the CABS algorithm could be especially useful when a diverse set of data from different sources is used.The overall accuracy of predicted contacts may not be crucial in refinement simulations. For example in the case of T0281, T0230 and T0212 targets, for which the accuracy of contact maps was nearly 40%, the quality of the final models did not improve significantly after the simulations. This is a consequence of the fact that the most of the accurate contacts were already observed in the Kolinski-Bujnicki starting model and were provided either by CABS or Frankenstein-3D. The remaining contacts, though diffuse, were mispredicted. The contacts which were crucial for correct structure determination but were not observed in the starting models were also absent in the predicted sets (see the T0281 contact map in Figure 3a). Consequently, mispredicted contacts, despite being sparse, dominated the simulation driving the system into false minima.According to our analysis of Figure 3, it appears that despite their frequent inconsistency more concentrated contact clusters with some accurate and some shifted contacts, but by a few residues only, are more useful for the refinement simulations with the CABS model than a diffuse set of precise contacts of higher overall accuracy.De novo folding supported by predicted contactsIf we compare de novo folding supported by predicted contacts with refinement simulations and the contact-based ranking (see Table 3), it turns out that the performance of this approach is in between these two above discussed methods. The results of the contact-based ranking are better in the case of contact sets of high coverage (1.5 N contacts or consensus from data for all predictors). On the contrary, the results of the refinement simulations are best for the data set of moderate, and thus well balanced, coverage and accuracy. Namely, the results for sets of N contacts are better than for 1.5 N contacts. Also, the best results of the refinement simulations are for the consensus from data from the best five predictors, for which the accuracy is better than in the case of consensus from three predictors, and the fraction of false contacts is smaller than for the consensus data from all predictors (see Table 2). It seems that during the REMC simulations the accuracy of contact data is more important than the coverage, because a large fraction of false contacts may disturb the protein folding pathway, and thus a near-native structure cannot be achieved. Regardless of which modeling method we chose, contact data obtained by the consensus of contact predictions performed better than simple combining together the top-scoring contacts of each prediction (compare results for N/2 top-scoring contacts from each of the best three predictors and for the consensus of the whole data from the best three predictors in Table 3).In the case of T0198, mispredicted contacts (see selected contacts in the T0198 contact map in Figure 3a) were observed in the final model of de novo folding simulation (data not shown), but not in the model obtained in the refinement simulation. It is the consequence of the fact that refinement simulations enable us to neglect some mispredicted contacts if they diverge significantly from the starting model and require too radical rebuilding of the whole conformation, which is impossible due to too low temperature and too loose restraints. In de novo folding simulations, in which the starting temperature is significantly higher than the folding transition temperature, all contacts, both mispredicted and accurate, can be equally satisfied because the starting conformation can be freely rebuilt. Consequently, if the predicted contacts are not accurate enough, de novo folding simulation can lead to the completely false fold, especially when contact-based restraints are tight and thus satisfied in majority. On the other hand, if the prediction of contacts is very accurate and substantial, such de novo simulations exploit it better than refinement simulations, because the energy landscape is properly modified by these contacts nearly from the beginning of the folding process. In Figure 4 we present two distinct situations as an example of the most favourable approach to protein modeling: refinement simulations with low-quality contact data (target T0215) and de novo folding with accurate contact data (T0248-1).Figure 4Distinctive results of two simulation methods involving the predicted contacts: de novo folding and refinement. Results of de novo folding (a) are represented by the best model of the T0248-1 target superimposed on the native structure (RMSD = 2.3 Å) and by the contact map with depicted real and quite accurate and precisely predicted contacts (upper triangle) and contacts of the best model obtained in the folding simulation and the first model of the Kolinski-Bujnicki group (lower triangle). Significant improvement of model quality and its contact map with respect to the native is observed. Results of the refinement simulations (b) are represented by the best model of the T0215 target (green) superimposed on the native structure (blue) (RMSD = 5.5 Å) and by the contact map constructed in the same fashion as in (a). Despite the low quality of the contact data predicted for the T0215 target the quality of the final refined model improved (but not significantly) in comparison to the original Kolinski-Bujnicki results (RMSD = 7.9 Å from the crystallographic structure Cα-trace).The accuracy of contact prediction for T0248-1 is 25.8% and it covers all the important regions of the real contact map (mainly contacts between helices) with a tolerance of a few residues. Such high quality data are better exploited when used as restraints in de novo folding (the RMSD of the representative structure of the best cluster is 2.3 Å) rather than in refinement simulations (RMSD = 3.0 Å). For comparison, the best result of the Kolinski-Bujnicki group was 7.9 Å from the native structure (this model was submitted as the second). The accuracy of contact data for T0215 is only 11.8% and the restraints based on such contact data, if fully satisfied, could worsen the prediction. The refinement simulation, however, takes advantage of part of the restraints only rejecting those inconsistent with the starting model and improves the RMSD from 6.2 Å to 5.5 Å (the best model). In the case of de novo folding the best model obtained was 6.8 Å from the native structure.In the case of poorly predicted contacts, e.g. for T0215, the quality of the final model does not deteriorate in the refinement as much as in the case of restrained de novo folding. As the average accuracy of contact predictions is still rather low, the refinement-based approach to protein structure modeling seems to be more useful.ConclusionIn this work we explored various methods for improving template-free modeling by using contact prediction. In the straightforward contact-based ranking of protein models, the best way is to combine as many predicted contacts as can be collected from different sources into a consensus set of high coverage and at least medium accuracy. Such combination of data obtained by different methods leads to a significant reduction of the effects of limitations and errors of each method.Introducing contact-based restraints into the template-free REMC simulations requires high accuracy or at least a significant number of accurate and semi-accurate contacts and only few false ones. Covering the most of the real contact map by predicted contacts improves structure prediction but overall, coverage is of lesser importance than accuracy. Contacts which are predicted with low probability and therefore can be false should be more diffuse and less numerous in the data set to reduce their influence on the conformational energy. The CABS force field is able to suppress the effect of the wrongly predicted contacts during the REMC simulations, provided that they do not form a densely-populated clusters in the predicted contact map.In general, we have shown that the theoretically predicted contacts could be useful in protein modeling, providing some other high-performance modeling tool, such as CABS, is also used. The usage of additional modeling tool is inevitable due to rather low accuracy of contact data, insufficient for the direct reconstruction of the 3D model. Predicted contacts can be used in simple and straightforward model ranking, refinement of crude models and de novo folding simulations. On average, all approaches lead to the improvement of the quality of predicted models. Sometimes the improvements are of a qualitative nature. This study provides a guideline how to use the contact prediction methods at various stages of protein structure prediction.The method described here is not restricted to the use of data from contact predictors. It is possible to employ contacts provided by any other method, theoretical or experimental. They can be extracted from very rough or coarse structural alignments or from fuzzy experimental data, for example from ambiguous NOEs or cross-link data. The initial structures for the most successful refinement simulations we obtained can be provided by any available modeling method from fold-recognition to entirely de novo methods.Although we tested various methods for contact-assisted model building, refining and ranking using only the CABS generated models, it seems to be rather safe to expect similar applicability of predicted contacts in other modeling techniques.AbbreviationsNF: New Fold; FR: Fold Recognition; FR/A: Fold Recognition – Analogy; FR/H: Fold Recognition – Homology; CM: Comparative Modeling; CASP6: the Sixth Critical Assessment of Techniques for Protein Structure Prediction; CABS: algorithm based on Cα, Cβ and the side group centre of an amino acid; REMC: Replica Exchange Monte Carlo method.Authors' contributionsDL prepared the contact data sets and performed the refinement and de novo folding simulations, the ranking of CASP6 models, statistical analysis, assembled figures and wrote the draft of the manuscript. AK provided CASP6 decoys generated by the Kolinski-Bujnicki group, conceived the study, and participated in its design. Both authors read and approved the final manuscript.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2527608.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2527608",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2527608\nAUTHORS: Bunyamin Dikici, Hakan Uzun, Ebru Yilmaz-Keskin, Taskin Tas, Ali Gunes, Halil Kocamaz, Capan Konca, Mehmet A Tas\n\nABSTRACT:\nBackgroundNeonatal tetanus (NT) is still considered as one of the major causes of neonatal death in many developing countries. The aim of the present study was to assess the characteristics of sixty-seven infants with the diagnosis of neonatal tetanus followed-up in the Pediatric Infectious Diseases Ward of Dicle University Hospital, Diyarbakir, between 1991 and 2006, and to draw attention to factors that may contribute (or may have contributed) to the elimination of the disease in Diyarbakir.MethodsThe data of sixty-seven infants whose epidemiological and clinical findings were compatible with neonatal tetanus were reviewed. Patients were stratified into two groups according to whether they survived or not to assess the effect of certain factors in the prognosis. Factors having a contribution to the higher rate of tetanus among newborn infants were discussed.ResultsA total of 55 cases of NT had been hospitalized between 1991 and 1996 whereas only 12 patients admitted in the last decade. All of the infants had been delivered at home by untrained traditional birth attendants (TBA), and none of the mothers had been immunized with tetanus toxoid during her pregnancy. Twenty-eight (41.8%) of the infants died during their follow-up. Lower birth weight, younger age at onset of symptoms and at the time admission, the presence of opisthotonus, risus sardonicus and were associated with a higher mortality rate.ConclusionAlthough the number of neonatal tetanus cases admitted to our clinic in recent years is lower than in the last decade efforts including appropriate health education of the masses, ensurement of access to antenatal sevices and increasing the rate of tetanus immunization among mothers still should be made in our region to achieve the goal of neonatal tetanus elimination.\n\nBODY:\nBackgroundNeonatal tetanus (NT), a preventable disease, remains one of the major causes of neonatal death in many developing countries [1-3]. It is an acute disease characterized by generalized rigidity and convulsive spasms of skeletal muscles, and is caused by an exotoxin of the bacterium Clostridium tetani which is a gram-positive and anaerobic rod forming spores that are very resistant to heat and the usual antiseptics [4].Tetanus in neonates is primarily caused by lack of hygiene during delivery, and it usually occurs when the umbilical cord is contaminated while it is being cut with a non-sterile instrument, or dressed. Complete eradication of tetanus is not possible as the spores of C. tetani are widespread in soil and in the stools of people and animals, and can be transmitted without human contact. The disease can be prevented by immunizing pregnant women and women of childbearing age with tetanus toxoid. Following appropriate vaccination of the mother, antibodies pass to the fetus across the placenta and provide protection against NT [1,3,5].NT has been reported to be virtually eliminated from developed countries [2]. However, it continues to be a leading cause of neonatal mortality in developing countries. Even with treatment, the case fatality rate can be as high as 80–90% [6,7]. As the second leading cause of death from vaccine-preventable diseases among children world-wide, NT is responsible for 14 percent (215,000) of all neonatal deaths [1,8].In 1989, the World Health Assembly called for the elimination of NT [9]. The year 2005 was set as the target date for worldwide elimination of the disease by UNICEF, WHO and the United Nations Population Fund. In 2000, WHO reported 57 countries, including Turkey, not to have achieved elimination of the disease (Global elimination of NT is defined as the reduction of cases to less than 1 per 1000 live births in every district in every country). In its report, WHO stressed that district assessment must confirm the achievement of NT elimination in the countries seeming to have done so [1].Effective implementation of public health measures in the past decade has reduced the incidence of NT in Turkey [10-12]. However, as of mid-2000, Turkey was reported by WHO not to have achieved elimination of the disease yet although it was ranked among the six countries to have potentially eliminated NT [1]. The aim of the present study was to evaluate the clinical findings, risk factors and prognostic characteristics of sixty-seven infants with the diagnosis of NT hospitalized and followed-up in the Pediatric Infectious Diseases Ward of Dicle University Hospital, Diyarbakir, between 1991 and 2006, and to draw attention to factors that may contribute (or may have contributed) to the elimination of the disease in our region.MethodsThe characteristics of sixty-seven infants diagnosed as NT and followed-up in Pediatric Infectious Diseases Ward of Dicle University Hospital, Diyarbakir (the largest tertiary center in the Southeast Anatolian Region of Turkey) from 1991 to 2006 were reviewed using their clinical charts. Most of the patients had been referred from state hospitals and primary health centers. The epidemiological and clinical findings of all the patients were compatible with neonatal tetanus. Nuchal rigidity, spasticity, muscular spasm, trismus and opisthotonus were included in the criteriae for clinical diagnosis, and the presence of at least three of these was accepted as compatible with tetanus [13]. The diagnosis was maintained on the basis of WHO criteriae [14]. In order to investigate metabolic diseases, serum lactic acid, pyruvic acid, and ammonia levels were assessed and urine/blood aminoacid chromatography was performed. In all cases, metabolic screening yielded negative results.The infants received treatment in specially designed rooms inside the neonatal care unit meeting the requirement of a quiet and dark place. They received all treatment with human tetanus immunoglobulin (3000 U intramuscular injection) or equine tetanus antitoxin (50000 U with half given intramuscularly and half intravenously), and, depending on which was available after appropriate testing, antimicrobial therapy with penicillin G (100,000 IU/kg/day) or with metronidazole, and high dose diazepam (40 mg/kg/day) intravenously. Intravenous fluid was administered until the patients could tolerate nasogastric feeding. Routine microbiologic and biochemical tests were ordered. Besides the age, sex and birth weight of the patients, data including the age at the onset of symptoms, the age at hospital admission, instruments used to cut the umbilical cord, place of delivery (home or hospital), the vaccination status of the mothers with tetanus toxoid (TT), clinical signs and symptoms and treatment procedures were all recorded.Patients were stratified into two groups according to whether they survived or not (group I and II, group I consisting of surviving infants) to assess the effect of certain factors in the prognosis. Factors having a contribution to the occurrence of high rate of tetanus among newborn infants were discussed. For statistical analysis, SPSS (Statistical Package for Social Sciences) 10.0 program was used. Normally distributed data were expressed as mean ± SD and analyzed with Student's t test. Categorical data were analyzed by chi-square test. For the correlation analysis, Spearman's correlation test, and for risk assessments, odds ratio (OR; 95%CI: confidence interval) and logistic regression analysis were done. Results are given as mean ± SD, and p < 0.05 is accepted as significant.ResultsSixty-seven infants (48 males and 19 females) were included in the study. In their article published in European Journal of Epidemiology in 1999, our colleques had reported a total of 55 cases of NT to have been hospitalized in the Pediatric Infectious Diseases Ward of our hospital between 1991 and 1996 [10]. However, in the last decade (between 1996 and 2006), only 12 new cases have been documented in the same clinic showing a tremendous decline in the incidence. The mean age of the patients was 8.9 ± 4.3 days at admission. The mean birth weight of the cases was 3125 ± 420 g. Twenty-eight (41.8%) of the infants [20 of the boys (41.7%), and 8 of the girls (42.1%)] died during their follow-up. The mean age at death in deceased infants was 7.4 ± 3.4 days. For the group of surviving infants, the mean birth weight was found 3402 ± 515 g whereas the birth weight of the deceased cases was 3153 ± 462 g, the difference being statistically significant (p = 0.014) (Table 1). The sex of the infants was not found to be statistically significant (Table 1).Table 1Clinical signs, symptoms and characteristics of the patients with neonatal tetanusVariablesSurvivors (n 39)Deaths (n 28)p valueBirth weight, g, (SD)*3402 ± 5153153 ± 462= 0.014Age at admission, days, (SD)*10.3 ± 4.77.4 ± 3.4= 0.001Age at onset of symptoms, days, (SD)*7.0 ± 4.04.9 ± 1.9< 0.0001Male/female(n)∞28/1120/8NSSpasticity(n)∞3024NSLack of sucking(n)∞2821NSTrismus(n)∞2216NSFever (n)∞1513NSOmphalitis∞1611NSJaundice∞118NSRisus sardonicus∞38< 0.0001Irritability∞86NSOpisthotonus∞18< 0.0001Cyanosis∞54NS* Student t test∞ Chi-square testNS: non-significantThe mean age at the onset of symptoms was 5.8 ± 3.1 days. Infants who survived had a mean age of 7.0 ± 4.0 days when the symptoms first occurred, whereas the mean age at the onset of the symptoms was 4.9 ± 1.9 days in the non-surviving group. The mean age when the symptoms first appeared was significantly lower in the non-surviving infants than in the surviving ones (p < 0.0001) (Table 1). Among the surviving infants, the mean age at admission was noted as 10.3 ± 4.7 days, whereas it was 7.4 ± 3.4 days for the non-surviving ones. This difference was also found to be statistically significant (p = 0.001) (Table 1).The main clinical signs and symptoms of the infants were spasticity (54 infants; 81%), lack of sucking (49 infants; 73%), trismus (38 infants; 57%), fever (28 infants; 42%), omphalitis (27 infants; 40%), jaundice (19 infants; 28%), risus sardonicus (17 infants; 26%), irritability (14 infants; 21%), cyanosis (9 infants; 13%) and opisthotonus (9 infants; 13%) in decreasing order. Signs which were found to be related with fatal outcome were opisthotonus (p = 0.001) and risus sardonicus (p = 0.001) (Table 1).There was negative correlation between age and mortality (r = -0.63, p < 0.0001) whereas there were positive correlations between opisthotonus and mortality (r = 0.44, p < 0.0001), and risus sardonicus and mortality (r = 0.45, p < 0.0001). Opisthotonus was seen 13,9 times more in non-survivors [odds ratio (OR) = 13.9, 95% CI = 1.890–102.660]. Additionally, risus sardonicus was detected 6 times less in survivors [OR= 6.04, 95%CI = 1.896–19.209]. However, in logistic regression analysis in which age at admission (days), gender, birth weight, spasticity, fever, opisthotonus, lack of sucking, trismus, omphalitis, irritability, cyanosis, jaundice, risus sardonicus and initial day of symptoms were included in the analysis as confounding factors, only age at admission (days) was found as an independent risk factor for mortality [odds ratio (OR) = 5.288, 95% CI = 1.730–16.165] (Table 2).Table 2Multivariate analysis of risk factors neonatal tetanus.VariablesOdds ratio95% CIBirth weight, g0.90.1–1.5Age at admission5.2*1.7–16Age at onset of symptoms0.60.3–1Gender1.20.3–4Spasticity1.10.2–7Lack of sucking0.80.2–3Trismus0.70.1–1.2Fever0.40.4–5Omphalitis0.50.1–5Jaundice0.10.1–15Risus sardonicus1.80.1–20Irritability0.90.5–30Opisthotonus1.30.5–1.9Cyanosis0.40.4–3* Statistically significant, p < 0.05.All of the patients came from rural areas, and all of them had been delivered by untrained traditional birth attendants (TBA). None of the mothers had received antenatal care in a clinic before delivery, and none had been immunized with TT during her pregnancy. The instrument to cut the umbilical cord was noted as razor blade in 37 infants (55.2%), scissors in 18 infants (26.9%) and knife in 12 patients (17.9%). All of these instruments had been used in non-hygienic conditions before.DiscussionReducing deaths from NT has been regarded as one of the simplest and most cost-effective ways to reduce the neonatal mortality rate. The disease can be effectively prevented through the use of tetanus toxoid. In addition to immunization, promotion of clean deliveries and improvement of surveillance are the main strategies to achieve its elimination [1].NT occurs most commonly in the lowest income countries and those with the least developed health infrastructure [1]. Over the past decade, progress in reducing the incidence of tetanus has been substantial. According to WHO, 309,000 people are estimated to have died of tetanus in 2000 [15], including 200,000 cases of NT [16], a reduction of about 75% when compared to 1988.The disease has also been reduced geographically: whereas in 1994 a total of 82 countries were accepted as not having eliminated NT [16], the focus has recently been on 57 countries that were reported not to have achieved elimination of the disease in 2000 by WHO. Turkey is included among these 57 countries. According to 1999 WHO estimates, 231 cases of maternal and NT occurred in this country. However in 2000, Turkey was ranked by WHO as one of the six countries to have potentially eliminated the disease. However, district assessments are necessary to confirm that elimination has been achieved [1].With specific therapy for neonatal tetanus, lethality rates have been reported to range from 25 to 90% [17]. The overall case fatality rate was found 41.8% in our patients with similar death rates for male and female infants (a male/female death ratio of 0.99:1.00). In their survey in a district of Pakistan, Quddus et al found the case fatality rate among 43 NT cases admitted to the district hospital as 62% (18) whereas Asekun-Olerinmoye et al found a case fatality rate of 79.4% in their study which was conducted on 120 cases and included all secondary and tertiary hospitals in Ibadan, Nigeria [19]. In two other studies, both carried out in Turkey, Ertem et al and Totan et al found fatality rates of 40.0 and 70.4%, respectively [20,21]. In the United States where 201 cases of NT were reported during 1991–94, fatality rate was found to be relatively low (25%) [22].In the studies carried out by Basu et al and Anita et al, low body weight was found to be a risk factor for mortality in neonatal tetanus [6,23]. Although there was significant difference between surviving infants and deceased ones regarding body weight in our study we could not show that it constituted a risk factor for mortality.Male infants constituted 71.6% of our patients. Higher proportion of neonatal deaths due to tetanus among males than females were reported previously [20-26]. Studies conducted in Pakistan, Sudan and Egypt showed that unhygienic practices during circumcision were responsible for this disparity in the proportion of male and female NT deaths [24,25]. Besides, we think that the lack of records due to underrated and neglected female neonates in underdeveloped populations may contribute to such a disparity. Regarding the case fatality rates, there was no difference between male and female patients in our study.There are many reports stating that mortality increased considerably when the incubation period was 5–10 days or less [6,19,20,23]. Short incubation period may show increased virulence of the infectious agent, or it may designate decreased defense mechanisms of the host against the disease. Both of these conditions may lead to an increase in mortality in neonatal tetanus cases. In our study, there was negative correlation between age of the patient and mortality. The difference regarding the mean age at admission between survivors and non-survivors was statistically significant, and young age at admission was shown to be a risk factor for mortality. [odds ratio (OR) = 5.288, 95% CI = 1.730–16.165]. This coincides with the findings of previous studies.Another important result of this study is the finding of a powerful correlation between risus sardonicus and mortality, and opisthotonus and mortality. Opisthotonus was seen 13.9 times more in non-survivors (OR = 13.9, 95% CI = 1.890–102.660). Additionally, risus sardonicus was detected 6 times less in survivors (OR= 6.04, 95%CI = 1.896–19.209). A powerful correlation between both of these factors and mortality has been pointed out in few studies before. In their study, Basu et al considered risus sardonicus (OR = 4.5, 95%CI = 2.4–8.6). and generalized rigidity (OR = 3.8, 95%CI = 2.3–6.2) to be a negative risk factor for prognosis [6].Illiteracy of the parents, poor antenatal care, unhygienic delivery practices and low immunization coverage with tetanus toxoid have all been shown to contribute to the occurrance of NT cases [1,6,18,20,21,27,28]. In our study, none of the mothers had received antenatal care before, and none of the cases was born by a clean delivery which is defined as a birth attended by professional health staff [1]. The unhygienic use of instruments to cut the cord is alarming and likely to be a cause of contamination by tetanus spores in our cases. It is of critical importance that the health workers giving tetanus toxoid vaccinations to these women also inform them about the components of clean delivery and post-delivery practices, especially umbilical cord care and discourage harmful traditional practices [1].All of the infants in our study came from rural areas. However, in the survey of Ertem et al [21] which analysed the data of 56 children with NT admitted to Diyarbakir Children State Hospital between 1994 and 2001, 17 (30.4%) of the patients came from urban areas. This reflects the persisting risk for NT in city centres as well although the vaccination coverage is higher, and the availability of appropriate delivery practices is easier in the city centres. However, in accordance with our findings, all of the NT cases in that study were born at home with the help of unqualified midwives [21]. These observations pinpoint the need for health education in all communities, both rural and urban.In 1994, the Turkish Ministry of Health launched an immunization program that targeted pregnant women [29]. This program required that all non-immunized women be vaccinated with two doses of TT during their first pregnancy and with one dose during each subsequent pregnancy. Strikingly, none of the mothers of our cases had been immunized with TT during her pregnancy. According to the Ministry of Health, reported coverages of vaccination for pregnant women with TT2 (the second dose of TT) in 2000 and 2001 were 52 and 50% in Samsun, 65 and 65% in Antalya, and 8 and 11% in Diyarbakir, respectively [30,31]. In a recent study carried out in three selected provinces of Turkey (Samsun, Antalya and Diyarbakir), Kurtoglu et al. found that the presence of protective antibody levels against tetanus in women of childbearing age ranged between 56.5–91.4% in different age groups [32], with lowest percentages having been recorded in Diyarbakir. These data may indicate the insufficiency of the immunization service in Diyarbakir during last years, and clearly elucidate the need to expand immunization services in our province in order to achieve the goals of the neonatal elimination program.In their article published in European Journal of Epidemiology in 1999, our colleques had reported a total of 55 cases of NT to had been hospitalized in the Pediatric Infectious Diseases Ward of our hospital between 1991 and 1996 [10]. However, between 1996 and 2006, only 12 new cases have been documented in the same clinic. Turkey has in fact experienced a reduction in the incidence of NT through effective implementation of public health measures in the past decade [10-12]. Though, in spite of the observation that the number of NT cases admitted to our clinic has declined compared to the last decade [10,27], we think that efforts still should be made in our region to achieve the goal of NT elimination.Limitations of the studyAmong our patients, there is a significantly higher percentage of males presenting than females, although no difference was found in case fatality. This could be due to gender bias in care seeking. Since about half the infants survived, we examined differences in those who survived and those who did not. However, there could still be biases associated with presentation that would make it difficult to know whether these were real differences. Another limitation to this study is the fact there is no control group.ConclusionSurveillance of the disease to identify high risk districts and areas, providing the masses with appropriate health education in addition to ensuring access to antenatal services, improving the quality of such services, and extending efforts to provide tetanus immunisation to mothers should be included and maintained among the strategies to be followed in order to attain the goal of elimination of this preventable disease.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsBD: performed the analysis, writing and preparation of manuscript. HU: contributed to the writing of the manuscript. EY–K: contributed to the writing of the manuscript. TT: contributed to data collection. AG: contributed to data collection. HK: contributed to statistical analysis. CK: contributed to data collection. MAT: contributed to writing and preparation of manuscript.Pre-publication historyThe pre-publication history for this paper can be accessed here:\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2527609.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2527609",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2527609\nAUTHORS: Ye Wang, Simona Celebrini, Yves Trotter, Pascal Barone\n\nABSTRACT:\nBackgroundVisual, tactile and auditory information is processed from the periphery to the cortical level through separate channels that target primary sensory cortices, from which it is further distributed to functionally specialized areas. Multisensory integration is classically assigned to higher hierarchical cortical areas, but there is growing electrophysiological evidence in man and monkey of multimodal interactions in areas thought to be unimodal, interactions that can occur at very short latencies. Such fast timing of multisensory interactions rules out the possibility of an origin in the polymodal areas mediated through back projections, but is rather in favor of heteromodal connections such as the direct projections observed in the monkey, from auditory areas (including the primary auditory cortex AI) directly to the primary visual cortex V1. Based on the existence of such AI to V1 projections, we looked for modulation of neuronal visual responses in V1 by an auditory stimulus in the awake behaving monkey.ResultsBehavioral or electrophysiological data were obtained from two behaving monkeys. One monkey was trained to maintain a passive central fixation while a peripheral visual (V) or visuo-auditory (AV) stimulus was presented. From a population of 45 V1 neurons, there was no difference in the mean latencies or strength of visual responses when comparing V and AV conditions. In a second active task, the monkey was required to orient his gaze toward the visual or visuo-auditory stimulus. From a population of 49 cells recorded during this saccadic task, we observed a significant reduction in response latencies in the visuo-auditory condition compared to the visual condition (mean 61.0 vs. 64.5 ms) only when the visual stimulus was at midlevel contrast. No effect was observed at high contrast.ConclusionOur data show that single neurons from a primary sensory cortex such as V1 can integrate sensory information of a different modality, a result that argues against a strict hierarchical model of multisensory integration. Multisensory interaction in V1 is, in our experiment, expressed by a significant reduction in visual response latencies specifically in suboptimal conditions and depending on the task demand. This suggests that neuronal mechanisms of multisensory integration are specific and adapted to the perceptual features of behavior.\n\nBODY:\nBackgroundThe classical view of multisensory integration, based on anatomical grounds [1], proposes that each sensory modality is processed through separate channels from the sensory receptors to the primary sensory areas and then further integrated into associative unimodal areas converging at the level of cognitive polymodal areas [2]. Indeed, in primates, neuronal responses to more than one sensory modality have been described in areas higher-up in the hierarchy like the frontal, temporal and parietal lobes [3-9]. While these polysensory areas are the best candidates to support sensory fusion, recent studies in humans have surprisingly revealed that multisensory interactions can take place in early stages of sensory processing, in regions thought to be involved in only one modality [10,11]. This result has led to a reappraisal of the cortical regions involved in multisensory integration [12]. In humans for example, imaging [13-16] and EEG studies [17,18] have clearly shown heteromodal responses in sensory areas even at the level of the primary sensory fields. Furthermore, the discovery of heteromodal connections directly linking areas involved in different sensory modalities could be the anatomical support of such interactions [19-21]. For example, in the monkey, the core of the auditory cortex receives direct inputs from both somatosensory and visual areas [19]. It can be inferred that these cortical heteromodal connections, as well as the thalamo-cortical loop [22,23], could be the anatomical pathway responsible for the visual [24-26], somatosensory [27,28] or proprioceptive [29] influences observed in the monkey auditory cortex [30].In the normal adult cat, some early electrophysiological studies have reported auditory responses in visual areas [31-33], a result which is still controversial [34]. Multisensory integration in the primary visual cortex (V1) of the monkey has not been established, apart from a clear influence of a non visual eye position signal on visual activity [35,36]. However, auditory or visuo-auditory responses in area V1 are highly probable since we have demonstrated direct projections from the auditory cortex (including A1) and the polymodal area STP to area V1 in the calcarine sulcus [21]. Furthermore, the auditory system is activated more precociously that the visual one, and for example the latencies of auditory responses recorded in areas AI and STP are about 35 and 45 ms respectively [37,38]. Consequently it conceivable that an auditory stimulus can modulate the visual responses in V1 where the mean onset latencies are longer, between 50–70 ms [39,40]. As some authors have reported even shorter latencies in V1 when using high contrast stimuli [41], one could expect that an auditory stimulus would affect mostly late visual responses such as the one obtained using non-optimal stimuli (ie. low visual contrast).We thus conducted an electrophysiological study to look for visuo-auditory interactions at the single cell level in primary visual cortex. Because the auditory projections to V1 are more dense at the representation of the peripheral visual field, a region of space that encompasses most of the auditory receptive field in AI [42,43], our electrophysiological recordings targetted visual cells with RF located between 10 and 20° of retinal eccentricity.MethodsThe present study is based on data obtained from two monkeys (Macaca mulatta) trained to performed a visual or visuo-auditory oculomotor task. A detailed description of the general methods used in the electrophysiological recording has been reported in a previous study [44]. All experimental protocols, including care, surgery, and training of animals, were performed according to the Public Health Service policy on the use of laboratory animals and complied with guidelines of the European Ethics Committee on Use and Care of Animals.Behavioral taskThe core of the present study concerns two monkeys (Mk1 and Mk2) trained to perform a visually guided saccadic task during which the visual target could be accompanied by an auditory stimulus (V/VA active task). A trial was initiated by the appearance of a fixation point (FP) located at the center of the video screen and of a size of 0.2 degree. The monkey had to direct its gaze and to maintain fixation at this central point. The duration of presentation of the FP was randomized between trials and lasted between 1500 to 1800 ms. Simultaneously, with the extinction of the FP, a peripheral visual target was flashed for 50 ms. The monkey was required to perform a saccade in the direction to the locus of the visual target within 250 ms of its appearance. Responses were considered as correct when the saccades were performed within a window of 4 × 4 degrees centered on the visual target, and in these cases a few drops of fruit juice were delivered to the monkeys as a reward. In half of the trials, presented randomly, a 25 ms sound (a white noise) was delivered from a speaker located at the same eccentricity on the azimuth as the visual stimulus. In such visuo-auditory trials (VA), the visual and the auditory stimuli were presented at the exactly same time. In both conditions (V and VA) the monkey was required to perform a saccade directed toward the visual target and consequently, the auditory stimulus had no behavioral meaning for the animal. Note that we did not train the monkeys to perform a saccade toward the auditory stimulus alone.The first monkey engaged in the present study (Mk1) was first trained to perform two control tasks before the V/VA active task. In a first stage, the monkey was trained to perform a simple passive fixation task (V/VA passive task). Following the presentation of the FP (of variable duration from 1500 to 1800 ms), a visual or visuo-auditory stimulus was presented for 500 ms together with the FP. To get rewarded, the monkey had to maintain its fixation until the FP was extinguished.Further, Mk1 was trained in a visual control task (V-only control task), during which the color of the FP informed the animal whether he had to maintain a central fixation (blue FP) or to make a saccade toward a visual peripheral stimulus (Red FP). In this case the visual stimulus was never accompanied by an auditory stimulus. The timing of stimulus presentation was identical to that described for the active task (50 ms).The monkey Mk1 was engaged successively in each of these different protocols for several months, a period during which electrophysiological recordings were performed in the primary visual cortex (see below). Mk2 was trained from the beginning to do the VA active task.Apparatus and electrophysiological recordingThe visual stimuli, delivered by a Vision Research Graphics system with a refresh rate of 120 Hz, were presented in total darkness on a video screen located 50 cm in front of the monkey. The FP consisted of a single dot of 0.2° of size. Peripheral visual stimuli consisted of dynamic random dots (dRD, density 20%, and dot size 3.5 min of arc) presented at either side of the central FP in a random order. Two or three contrast levels of the dRD were used, from low (15%) to medium (55%) or high values (88%). During the behavioral sessions, the visual stimuli were 5° of size and presented at 10 or 20° of retinal eccentricity in azimuth, and -5 or -10° in elevation. During the electrophysiological recording sessions, the size and the location of the dRD were adjusted to the size and location of the cell receptive fields. Auditory stimuli (white noise, 72 dB SPL) were presented through a multi-channel sound card to one of 6 speakers located just below the video screen (at about 3 degrees below the lower border of the monitor). The horizontal position of the speakers was adjustable to be spatially adjusted in the horizontal plan just below the visual dRD. Thus the auditory stimuli are matching the horizontal location of the cell receptive fields, but not the vertical one (see figure 1). Further, to minimize the reflection of the sound, the animal was placed in a restricted space (1 by 1 meter for each sides) covered of thick black curtains. Acoustical reflection could be problematic for sound localization. However in our protocol the sound had no meaning to perform correctly the oculomotor task of the monkey and we believe that any sound reflection did not affect the monkey performance. Further, based on previous electrophysiological data obtained in the auditory pathway [45], in situation of reverberent environment, a cell's response to the leading source is always stronger than that observed to the lagging one, especially at short lag/lead delays [46]. In consequence, in absence of a sound-attenuated chamber, we think that in our experimental set-up the main auditory effect observed on V1 cells will be produced by the leading direct sound source located in front of the animal.Figure 1Experimental protocol.A. In the active Visuo-Auditory task, the monkey had to perform a saccade toward a visual cue located on the right or left side of the central fixation point. The visual stimulus was placed inside the neurons RF or at the opposite location. Auditory speakers were placed just below the visual location. Dashed squares represent the control windows regulating the eye movement for both the fixation and the saccadic periods. B. Temporal succession of the visual and auditory events in the active (upper panel) or passive (lower) tasks. C. EEG recordings of a monkey (Mk1) during a saccadic task oriented toward a visual cue located at 16 degrees on the left or right site of the fixation point. The upper and lower panels correspond to the horizontal or vertical eye movements respectively. Red dots represent the saccade latencies as described in the method section. Note that the anticipatory saccades obtained at latencies lower that 30 ms were not included into the analysis. D. Polar plot showing the receptive fields location of the recorded single cells. Most of the cells have RFs located at retinal eccentricities beyond 10°.Protocols control and data acquisition were executed under REX system. To avoid a jittering when generating the auditory stimulus through Windows system, all stimuli were pre-generated and stored in memory. Then we added a buffer silent time before each auditory stimulus. Triggers corresponding to the beginning of auditory stimuli (from the buffer) and the first visual frame (through VRG) were sent to REX system. The audio buffer length was adjusted to synchronize the visual and auditory stimuli at expected delay.Aseptic surgery was performed to attach a head-post to the skull and to implant a scleral search coil in both eyes. Single-unit recordings were made in one of the two monkeys (Mk1). Once the monkey had reached a high level of performance, a second surgery was performed to implant a recording chamber above the peripheral visual field representation in V1 located in the calcarine sulcus [47]. The skull was removed within the chamber, and a fixed grid was placed, so that the electrode penetrations were spaced 1 mm apart. Guide tubes were used to help to penetrate the dura. Sterile, tungsten-in-glass electrodes of ~1 MΩ impedance were inserted with a hydraulic microdrive fixed to the recording chamber, perpendicular to the cortical surface. Extracellular recordings were carried out in both hemispheres of the monkey from which the visual responses were previously analyzed for disparity selectivity (see [48]). Action potential waveforms were sorted online with the help of a spike sorting software (AlphaOmega MSD®) and only single units recorded through complete trials were selected for analysis.Data analysisThe behavioral analysis was derived from the performance of the two monkeys trained to perform the visually and visuo-auditory guided saccadic task (V/VA active task). For each trial we determined the saccade latency defined as the first point when the eye position was significantly different from the average eye position signal during the 300 ms prior to stimulus offset. This corresponded to the time at which the difference between the current position and the mean was 2 times greater than the maximum range observed during the fixation period. Then we performed a statistical analysis (Multifactor Anova) to compare the saccade latencies obtained during the V and VA conditions. We used a multi-way ANOVA test to the saccade latency obtained in each monkey. Contrast (3 groups in MK1, 2 groups in MK2), eccentricity (2 groups), and V or VA stimulation (2 groups) were treated as different factors. We checked both single factor and two-factors interactions. In case of p value too low to be computed, as we know it show very high significance, we also indicate F value as references. Post-hoc test was then applied to compare the saccade latencies between individual pair of conditions.For each neuron in each condition, the neuronal activity was recorded for 20–40 correct trials. Two parameters were studied to analyze the effect of visuo-auditory interactions in V1 cells : the amplitude and the latency of the visual responses. To measure the visual response latency, we first computed histograms of neuronal activity aligned on the stimulus onset. As previously described [48,49], we further smoothed the accumulative line by simulating each spike as a mini gauss function (Amp. = 1; Sigma = 4). So within each 1 ms bin, we got statistical spike numbers of 40 (trial number)*10(gauss summation) ms window. Then we measured the baseline average spike number per bin in the 200 ms prior to stimulus onset, and used it as the Poisson distribution lambda parameter of spontaneous activity. So the threshold of response activity was the smallest number n that the Poisson cumulative density function evaluated which equaled or exceeded 0.99. Thus, if the firing property obeyed the same Poisson distribution as the baseline, the spike number within each bin would not exceed this value at 99% confidence. We calculated this number n by using the Matlab Poisson function. We then used a detection window of 4 ms and measured activity starting from visual stimulus onset. If the minimum value inside this window was greater than n, we determined the response latency as the first point of the window. Because we could only get one latency value for a group of trials, we used bootstrap methods to compare the activity between conditions : shuffle latencies were calculated from the same number of trials in a sample taken randomly from both conditions. This was performed 4000 times to obtain 2000 randomly grouped pairs, from which we calculated the individual difference within pairs. The bootstrap p value is the ratio of pairs for which differences were no less than the values obtained from the experimental data. At the population level, after normalization of the responses, we used an Anova test to analyze the factor effect on response amplitude or latency, and paired t-tests for post hoc comparison.ResultsVisuo-auditory interaction: behavioral evidenceIn a first stage we analyzed the effect of visual conditions on the saccadic reaction times (sRT) performed by the two monkeys (Fig 2). The present data concern behavioral latencies obtained in highly trained monkeys and are based on several thousands of trials. Across the different conditions of stimulation (uni- or bimodal, at different contrasts or eccentricities), sRT values were on average 155,0 ms for Mk1 and 167,6 ms for Mk2 which correspond to the range of values reported in other studies using similar experimental conditions)[50]. First in uni- and bimodal conditions we observed a decrease in sRT in both monkeys when the contrast of the visual cue increased (Anova, Mk1, F = 4012, p = 0; Mk2, F = 20.39, p = 1.64E-9). This was particularly evident for Mk1 during a 20° task in visual-only conditions, with sRT of 172 ms when the contrast was low and of only 144 ms at high contrast. This is in agreement with numerous similar psychophysical studies on RT in both man and monkey [51-53]. Furthermore, in both sensory conditions, the monkeys had much shorter sRT when the saccades were directed toward the more eccentric (20°) peripheral target (Mk1, F = 527.65, Mk2, F = 148.17, p = 0 both cases). On average, independent of the visual contrast, saccade latencies toward eccentric cues located at 20° were 10% shorter than those toward a cue at 10°. This difference in latencies tended to be larger in the visuo-auditory (12.6% shorter at 20°) compared to the visual-only condition (8.6%). Again, this effect of the eccentricity of a target on the saccadic reaction time is similar to that observed in humans [54].Figure 2Saccadic reaction times (± sd) of the individual monkeys (A: monkey 1; B: Monkey 2) according to the eccentricity of the visual target (10° left panels; 20° right panels) presented at different contrasts and during the visual-only () or visuo-auditory conditions (○). Saccadic latencies are shortened both when increasing the visual contrast and when the visual cue is presented simultaneously with the auditory stimulus (VA conditions).We analyzed next the effect of bimodal stimulation by comparing the sRT in the V-only and VA conditions (Fig 2). As classically reported and resulting from multisensory integration [55], we observed a strong reduction in the saccade latencies in the VA condition compared to the V-only situation. This reduction was observed at all eccentricities (Anova, 10°: Mk1 F = 231.17,, Mk2 F = 66.67, both p = 0; 20° Mk1 F = 404.76, Mk2 F = 271.45, both p = 0) and at all contrasts of the visual target (Anova, Mk1, 15% F = 516, 55% F = 329.7, both p = 0; Mk2 30% F = 107.3, p = 0, 55% F = 69.38, p = 3.33E-16, 88% F = 117.45, p = 0). On average, when combining all conditions, the decrease in sRT ranged from 10% (Mk1) to 15% (Mk2) when saccades were made toward the VA stimulus. The rule of inverse effectiveness [55], proposes that the higher benefits resulting from multisensory integration should be obtained in sensory conditions of low saliency. Thus we searched for an effect of visual contrast on the reduction of sRT during visuo-auditory saccades. In Mk2, for which data were obtained at 3 different contrasts (30–55 and 88%), there was a tendency toward a more pronounced shortening of sRT at low contrast. When saccades were made at 20°, sRT in VA conditions were 19% shorter at a low contrast (187 ms in V-only vs. 152 ms in VA, p = 6.38E-21) while the decrease was only 14% at high contrast (174 ms vs. 150 ms, p = 7.29E-23). However we did not replicate these results in the second monkey or in all conditions. In Mk1, at 20°, the reduction was similar at low (11.1% decrease) and high contrast (11.9% decrease). Thus, we observed a constant decrease in sRT at all the visual contrasts used, data which seem to contradict the rule of inverse effectiveness. However, this could be due to the level of training. When analyzing the data during the first sessions of the behavioral training of Mk1 (not shown), we found a stronger decrease in sRT at low contrast, but at that time the monkey was not performing at an efficient level and his saccade latencies were much longer. This effect disappeared after extensive training over several weeks.Visuo-auditory interaction: electrophysiological evidenceThe present study is based on three sets of visual responsive single units (total n = 136) recorded in the primary visual area V1 of one monkey (Mk1). Each set of cells was obtained during a single behavioral condition (V/VA active task n = 49; V/VA passive task n = 45; V-only control task n = 42), all cells were recorded in peripheral V1 and most of them (69%) were located in the upper bank of the Calcarine sulcus (Fig 1) and present a receptive field located over 10° of eccentricity in the lower visual field. The size of the receptive fields were ranging between 1 and 4° (see [48]) characteristic of those cells recorded in the peripheral representation of V1.Auditory modulation of visual responses in V/VA active taskThe visual responses (discharge rate and latency) of 49 isolated V1 neurons were analyzed during the active visual and visuo-auditory tasks (Table 1). In the V-only conditions, the cells showed a strong phasic activity in response to the dRD and both the magnitude (Anova, F = 4.5, p = 0.0135) and latencies (Anova, F = 58.36, p < 0.0001) of the responses were affected by the contrast level as classically reported for V1 cells [56]. When comparing the response rates, we observed that while the discharge rates were similar at high (88%) and medium (55%) contrasts (54.5 spk/s and 52.2 spk/=s respectively, paired t-test p = 0.18, ns), the neuronal activity was, on average, much lower at the low contrast of 15% (33.9 spk/s, paired t-test, p < 0.001, both comparisons). The cell latencies were also sensitive to the contrast level and we observed a progressive increase in the mean latency (table 1, paired t-test p < 0.001 for all comparisons) when presenting stimuli from the high (49.2 ms), medium (64.5 ms) or low contrasts (100.7 ms). In the bimodal condition (VA active task), we observed exactly the same influence of visual contrast on the neuronal responses as expressed by an increase of discharge rate (Anova, F = 3.65 p = 0.0296) and a decrease of latencies when increasing the contrast levels (Anova, F = 52.31 p = 3.33E-16, Table 1). Thus the simultaneous presentation of an auditory stimulus has no effect on the contrast dependent relationships of the visual responses of V1 neurons.Table 1Response rates and latency values (± se) of V1 single units recorded during the V/VA active tasks using three different contrast levels.Response Rate (spk/s)Latency (ms)V-onlyAVV-onlyAVLow level (n = 17)33.9 ± 5.035.7 ± 5.1100.7 ± 6.296.8 ± 6.2Mid-level (n = 39)52.2 ± 4.052.7 ± 4.164.5 ± 2.561.0 ± 2.3High level (n = 45)54.4 ± 3.954.1 ± 3.849.2 ± 1.849.0 ± 2.0When we compared the V and AV conditions at constant contrasts, the simultaneous presentation of a spatially congruent auditory stimulus affects the cells activity by a change in the response latency. This is illustrated in two examples in figure 3. Both cells showed a characteristic phasic response to a dRD presented in their RF. While the frequency discharge was similar during the V-only and AV conditions (paired t-tests, both cells, p > 0.05), the two neurons showed a significant decrease in latency. For example, cell #33 (left panel) when stimulated at a midlevel contrast, had a mean latency of 56 ms during the visual task, a value that was reduced to 45 ms in the visuo-auditory conditions (bootstrap, p = 0.0415), while the spike discharge remained constant (35 and 38 spk/s respectively, paired t-tests, p = 0.4874, ns). This general effect of the auditory stimulus on the visual responses held when the analysis was performed at the population level. First when comparing the V-only and VA conditions, we did not see any significant change in the response rate of the cells at high and midlevel contrasts (Table 2, paired t-test p > 0.5 ns for both conditions). However, at low contrasts we observed a slight increase in the response rate from 33.9 spk/s to 35.7 spk/s a difference that is just below the significance level (paired t-test, p = 0.04). The middle panel in fig 4 shows the distribution of the relative differences of response rates (in %) between the uni- and bimodal conditions of all cells at each visual contrast. In the two higher contrast conditions, the distribution is centered at 0, corresponding to an absence of variation of the neuronal discharge between the V and AV presentations.Figure 3Examples of activity of two V1 single cells that present a significant reduction of their visual responses latency during the bimodal visuo-auditory conditions (AV). A and D represent rasters of the cells activity in the visual-only condition, while B and E show the activity of the same cells in the visuo-auditory conditions. The red dots indicate the time at which the monkey is making a saccade toward the visual target. C and F represent the response peristimulus time histogram to visual (blue) or visuo-auditory (red) stimuli. In both cells, the AV response latency is shorter compared to the V-only condition.Figure 4Effects of a visuo-auditory stimulus on the visual responses obtained from 49 single units recorded in V1 during a visuo-auditory saccadic task performed at three contrast levels (15%, upper row, 55%, middle row and 88%, lower row). The right panels represent the averaged normalized responses of the cells population during the V-only (blue) or visuo-auditory (red) conditions. The other panels show the distribution of the relative difference in the response rates (left) or latencies (middle) of the cells when the V-only and VA conditions are compared for each individual cell. A value of \"0\" means no difference, while negative values represent a decrease in the VA compared to V-only condition. At high and mid-contrasts no effects are observed concerning the response rate, while we observed a shift of the distribution of latencies toward negative values, indicating a shortening of the visual latency during the bimodal condition. The numbers of cells showing a reduction in latency are indicated in the brackets.Table 2Response rates and latency values (± se) of V1 single units recorded during the V/VA passive tasks and the V-only control task using a middle (55%) contrast value.Passive taskVisual control taskV-onlyAVVisual FixationVisual SaccadeResponse Rate (spk/s)29.5 ± 2.629.8 ± 2.828.7 ± 2.329.4 2.2Latency (ms)65.0 ± 4.765.2 ± 4.344.0 ± 1.443.5 ± 1.5However, the cell response latency was globally reduced when the auditory stimulus was delivered simultaneously with the visual target (Table 1). At the population level and at each contrast, the cells latency tended to be shorter. This is illustrated in Fig 4 by a leftward shift toward negative values in the distribution of the relative differences (in ms) when comparing the A and VA conditions. This effect reached a statistically significant level only for the 55% middle visual contrast (paired t-test, p = 0.009). In this condition, the mean latency was 64.5 ms in the unimodal visual stimulation against 61.0 ms in the VA condition, corresponding to a global decrease of more than 5%.At 15% contrast, the VA stimulation lead to similar values of latency compared to the V-only task (96.8 ms vs 100.7 ms respectively), a difference which was not statistically significant (paired t-test, p = 0.43 ns) probably because of a greater variability in the measured latency due to a strong reduction in the cell discharge (see above) when presenting this low contrast visual stimulus.Finally, at high contrast (88%) the neurons showed very comparable latencies (49.0 in VA vs 49.2 ms in V-only, paired t-test, p = 0.82 ns).The decrease in latency at the mid contrast level did not similarly affect all visual cells in V1. A correlation analysis between the absolute latency values in the V-only condition and the relative change (in ms) that occurred during the VA stimulation, revealed an inverse relationship (r = -0.4, p = 0.01, Pearson test, see Figure 5). This means that the cells with the longer latency showed a greater reduction in the bimodal condition, a mechanism that consequently should globally increase the rate of visual processing in area V1. In the other visual conditions (high and low contrasts), the correlation analysis did not reach a statistically significant level (both cases, p > 0.05).Figure 5Relationship between the latencies of the cells in the V-only active task and their respective changes (in ms) when tested in the AV active task. We observed a statistical inverse relation (P = 0.01; Pearson test) corresponding to a larger reduction for the cells showing longer response latencies.To conclude, we observed that the concomitant presentation of an auditory signal simultaneously with a saccadic visual target induced a reduction of the latency of V1 cells that depended on the contrast of the visual target. Furthermore, the discharge rates of the neurons remained unaffected by bimodal stimulation except during visual conditions that approach the perceptive threshold.Absence of visuo-auditory interactions in the V/VA passive tasksWe analyzed the effect of a visuo-auditory stimulus on the activity of a different set of 45 V1 neurons in a passive task during which the monkey maintained a central fixation while the peripheral stimulus was delivered in the cell's RF. In this case, the results were quite simple in that at all contrasts tested, we did not observe a change in the visual response with the spatially congruent auditory stimulus (Table 2, fig 6). First, the response rate remained unchanged between the V-only and VA conditions (29.5 and 29.8 spk/s respectively, paired t-test, p = 0.69 ns) as reported in the active task. However, in contrast to the effects observed in the saccadic task, the cell visual latencies were the same in the two conditions (65.0 ms in V-only vs. 65.2 ms in VA condition, paired t-test, p = 0.95 ns). These results are presented in Fig 6 as the relative changes in discharge rate and latency values, and the distributions are well centered on zero, indicating no difference between the two conditions.Figure 6Left: effects of a visuo-auditory stimulus on the visual responses obtained from 45 V1 single units recorded during a passive visuo-auditory task. No effects are observed either on the response rate or on the visual latency. Right: absence of an effect of the behavioral paradigm on the activity of 42 V1 neurons during the visual-only condition. When comparing the neuronal activity of the same cells during a visual passive fixation task to a visual saccadic task, we did not observe a modulation of the response rate or response latency. Conventions as in Figure 5.In a subset of neurons (n = 29), we also searched for an effect of the auditory stimulus alone on the neurons activity during a simple central fixation (Fig 7). We did not observe any auditory response. Following a single sound presentation the firing rate of the single cells remained at the same level as the spontaneous activity (paired t-test, p = 0.9056, non significant).Figure 7Average activity of 29 V1 single cells following the passive presentation of an auditory stimulus. The upper graph shows the average activity of the entire population. No response can be observed following the 25 ms presentation of the broad band noise at time 0. This is also apparent in the lower graph that compares the relative difference of the neuronal activity before and during the auditory presentation. No statistical differences were obtained (see text) confirming the lack of auditory response.Visual responses in the V-only control taskPrevious studies in the behaving monkey have shown that neuronal responses in striate and extrastriate cortical areas can be modulated by the behavioral meaning of the stimulus [57,58]. Consequently, we compared the visual responses of a third set of V1 single units (n = 42) during a dual task, a passive central fixation and an active visually guided saccadic task. As explained in the methods, the type of task was indicated to the animal by the color of the fixation point. In this case (Table 2, Fig 6), we did not observe an effect of the task (passive vs. active) either on the frequency discharge (28.7 and 29.4 spk/s respectively, paired t-test, p = 0.25 ns) or on the visual latency (44.0 and 43.5 ms respectively, paired t-test, p = 0.70 ns). This last part suggests that the visuo-auditory interactions that differentially affected the cells in the previous conditions were probably not due to the oculomotor demands of the task in which the monkey was engaged.DiscussionThe present results demonstrate that in behaving monkeys visuo-auditory interaction can occur at the single cell level at the first cortical stage of processing of visual information, the primary visual cortex V1. Multisensory interactions in V1 are characterized in our experiment by a modulation of V1 responses corresponding to a reduction of the neuronal onset latency. Moreover this effect was dependent on the perceptual charge of the task in which the animal was engaged.Visuo-auditory interaction: behavioral evidenceWe show that the simultaneous presentation of a sound during a visually guided saccade, induces a reduction of about 10 to 15% in the saccade latency depending on the animal and on the visual stimulus contrast level. Such behavioral improvement resulting from a bimodal visuo-auditory stimulation has been already reported during similar paradigms of spatially oriented behavior in humans [54,59-61], monkeys [50,59], carnivores [62,63] and even in rodents or birds [64,65]. Numerous studies have established the beneficial effect of bimodal stimulation [66] when the experimental sensory conditions respect the rules of spatial and temporal congruencies [55]. In these cases, multisensory integration results in perceptual improvements by reducing ambiguity in various tasks, from simple detections to complex discriminations, memory or learning tasks [67-71]. The decrease in reaction times during a bimodal paradigm has been explained by a co-activation system [72] that violates the race model of independent sensory channels in which the faster modality initiates the motor response. In our study, we did not train the animals to make a saccade toward an isolated auditory cue, so we cannot conclude on the race model. However, we recently reported evidence that such a converging model can account for a shortening in RT in visuo-auditory detection task in the monkey [23].Multisensory integration is supposed to obey the rule of inverse effectiveness that proposes a higher multisensory benefit when the unisensory stimuli are weak [62,73]. We did not observe such effects and the decrease in sRT was identical when comparing visuo-auditory performances at low or high visual saliencies, a result comparable to that recently reported in a similar behavioral study in the monkey)[50]. We cannot rule out the possibility that if we had used a weaker auditory stimulus it would have produced a change in bimodal gains [23], but it is very likely that this lack of inverse effectiveness is due to the fact that our experiments were performed on highly trained monkeys. It has been shown in monkey, that a continuous training strongly decreases the saccade latency [74], probably reducing the potential range of facilitation induced by the mechanisms of multisensory integration.Multisensory interaction at early stages of sensory processingThe delimitation of the polymodal areas associated with multisensory integration was until recently, generally circumscribed to cortical areas in the parietal, frontal and inferotemporal regions of the monkey [5,38,75-77]. However, electrophysiological and functional imaging studies in humans have recently revealed that visual, somatosensory or auditory areas defined originally as unimodal can be the locus of interactions between other non-specific sensory modalities [13,16-18,78-81]. In the monkey, electrophysiological recordings have confirmed that unimodal areas, located at the first stages of the sensory processing hierarchy, can integrate information from a different sensory channel [11].Until recently, this heteromodal activity had been observed in primates principally in the auditory system. For example, recordings of neuronal activity in the auditory cortex have revealed visual and somatosensory responses in the associative areas of the belt and parabelt [25,27,28,82,83]. In the primary auditory cortex, electrophysiological recordings (current source density) suggest that non-auditory events are of a rather modulatory influence and do not drive activation at the spiking level [84]. For example, proprioceptive information (eye position) can induce changes in the strength of the neuronal discharge in response to a spatially defined sound [29]. Furthermore, it has been proposed that the effect of non-auditory stimuli on AI activity is performed through a modulation of cortical oscillations to allow either enhancement or depression, depending on the timing of the bimodal stimulation [84]. Our results are in agreement with this notion of a modulatory effect and we did not find any auditory response in the single units we tested. The lack of pure auditory response in spite of an auditory modulation of the visual latency suggests that in V1, multisensory interaction could be a subthreshold phenomenon as hypothetized for multisensory interactions in the auditory cortex [24,83]. Because the auditory system is activated faster than the visual one, the auditory stimulus can depolarize the membrane potential of the visual V1 cells, inducing an earlier spiking response compared to the visual-only condition. Such multisensory interaction on cortical sensitivity has been recently suggested by TMS studies in human at both perceptual [85] and behavioral [86] levels.The main visuo-auditory effect we observed, was a shortening of the visual latencies but only in specific behavioral situations. All together these results suggest that in primates, multisensory integration mechanisms differentially affect sensory responses when they occur in primary or secondary sensory areas [11,24,87-90].Most of the neuronal rules of interactions between sensory modalities have been established in the Superior colliculus (SC) which is considered to be the key structure for multisensory integration [55]. In the SC, the convergence of different sensory modalities is reflected mainly by an enhancement in neuronal activity in response to a combined multimodal stimulus when spatial and temporal congruencies are respected [91-94]. A modulation (enhancement or depression) of the strength of the unimodal response by bimodal stimulation has been also reported in higher order polymodal areas of the monkey such as the prefrontal, parietal or inferotemporal areas [75,77,95,96] and even in the primary auditory cortex [24,84,97]. However, the proportion of neurons showing enhancement or depression varies strongly across cortical areas. When presenting middle or high contrast visual stimuli, we did not observe such an effect on the response rate in the large sample of visual cells recorded in V1, irrespective of the behavioral paradigm, suggesting that the neuronal mechanisms of multisensory integration are based on rules which are specific to each individual area. However, at low (15%) contrast, the slight increase of the responsiveness of V1 neurons in the AV condition suggests that the rule of inverse effectiveness could apply to V1. We cannot exclude that such effect on the response rate would be more prominent for visual stimuli of even lower perceptive saliency.In addition, as described in the methods, the visual and auditory stimuli are only spatially congruent in the horizontal azimuth dimension. While the receptive fields of the auditory neurons are large [98] and cover probably the offset that separate the two stimuli, one can speculate that a better spatial congruency between the auditory and visual stimuli would lead to greater effects on V1 cells during bimodal stimulation.Finally, a rule common to several cerebral loci of multisensory integration is an effect on the response onset latency [75,99]. We observed that in the active task, the visual latency was reduced by about 5%, a result very similar to that reported in the SC [50,99]. This decrease in neuronal response onset, which is in line with a shortening of the visuo-auditory bold response in human V1 assessed by fMRI [13], could participate in the speeding up of the behavioral saccadic responses during bimodal presentation (see below, [62]).As developed in the introduction, previous anatomical studies have established that sensory fusion was processed through the convergence of the different sensory channels at the level of associative cortical areas [1,2]. The numerous reports of multisensory interactions at low level of sensory processing (present data, [24,27,84]) and acting on early sensory responses, favor a modulatory influence through heteromodal connections linking directly unisensory areas [19-21]. However, such modulatory effect could also originate from non-specific thalamic nuclei that integrate different sensory processing [100]. A cortico-thalamic loop that bypass cortico-cortical connections could thus support fast transmission and provide multisensory and sensory-motor information to unimodal areas [22,101].Visuo-auditory interaction: role of the behavioral contextIn the alert monkey, we have shown that visual neurons in V1 showed a decrease in response onset when the visual stimuli were presented simultaneously with a sound. However, our main result is that this effect on the visual responses is dependent on the behavioral context : we did not see any changes in V1 neuron latency in a passive situation when the monkey did not perform an oriented saccade toward the spatial location where the auditory stimulus was presented. It could be argued that this difference simply reflects a process of visual spatial attention [102] due to the oculomotor task and not a modulation specifically due to the integration of the auditory stimulus at the neuronal level. In V1 and extra-striate areas, it has been shown that attentional mechanisms [103,104] or behavioral relevance [57,105,106] can affect the characteristics of the neuronal visual response such as the discharge rate, the latency or the neurons selectivity. We did not observe a change in the cell firing rate when comparing neuronal activity in a visual passive and active task without any auditory stimuli. While our comparisons are performed on a different set of neurons, it strongly suggests that the shortening in latency depends specifically on the bimodal conditions in a particular behavioral situation, and not on visual attentional processes linked to the oculomotor demand of the task. However, we are aware that the three paradigms differ in term of attentional loads but in both passive and active AV tasks, the auditory stimulus can involve similar mechanisms of exogenous attention. The distinction between exogenous spatial attention and crossmodal interactions (or integration) is still an open question [107] as both mechanisms result in an improvement in sensory perception [102].Our results are in complete agreement with studies in humans and animals showing different patterns of multisensory integration according to the behavioral context. First in humans, the detection or discrimination of bimodal objects, as well as the perceptual expertise of subjects, differentially affect both the temporal aspects and the cortical areas at which multisensory interactions occur [18,108]. Similarly the index of multisensory integration computed from the activity of neurons in the deep layers of the Superior Colliculus, is also dependent on the oculomotor behavior of the animal [109]. Finally, while heteromodal visual or somatosensory responses can be obtained in the auditory cortex of a passive or anaesthetized monkey [24,27,87], some authors have reported that some visual responses can be related to task in which the animal is engaged [25].All together these findings suggests that the neuronal network involved in multisensory integration as well as its expression at the level of the neuronal activity is highly dependent on the perceptual task in which the subject is engaged. Thus multisensory interactions can underly from active perception to attentional mechanisms. This hypothesis is supported by the anatomical pattern of heteromodal connections that directly link areas involved in different modalities. In monkey, such heteromodal connections either link specific sensory representations (retinotopy or somatotopy) of interconnected areas or specific functional regions in each modality [19,21,110].Such an influence of the perceptual context on the neuronal expression of multisensory interaction has further consequences on the phenomena of cross-modal compensation that occurs after sensory deprivation in animals [111] or humans [112,113]. In blind subjects [114], the efficiency of somatosensory stimulation on the activation of the visual cortex, is maximum during an active discrimination task (Braille reading). This suggests that the mechanisms of multisensory interaction at early stages of sensory processing and the cross-modal compensatory mechanisms are probably mediated through common neuronal pathways.Role of visuo-auditory integration in the primary visual cortexThe effect of an auditory stimulus on V1 responses is probably supported through the direct projections that originate in the auditory (A1 and belt) and multimodal (STP) areas and target V1 [20,21]. As discussed previously [21], the auditory projections to V1 originate mainly from the dorsal auditory stream, specialized in processing spatial information, and reach the peripheral representation of V1. The characteristics of this heteromodal connectivity suggest that this pathway is probably involved in rapidly orienting the gaze toward a sound source located in the peripheral field for which visual acuity is poor. In situations of spatial and temporal congruency, multisensory integration has been shown to facilitate the neuronal responses of neurons of the superior colliculus [115,116], both at the sensory and motor levels [50,59]. Consequently, multisensory integration at the collicular level will allow a direct influence on motor output because the SC is directly involved in the control of oculomotor behavior [117]. A large number of visual areas project directly down to the SC, but in the monkey, the main inputs are originating from the primary visual cortex which constitutes about 20 to 30% of the SC cortical afferents [118]. Consequently the decrease in V1 response latencies during bimodal stimulation can act directly on the response of cells in the SC and speed up the initiation of the saccadic command by the brain stem oculomotor nucleus. However, because the reduction in V1 latencies (5% decrease) does not match the amount of facilitation at the saccadic level (10 to 15% reduction in saccade latency), one should consider other mechanisms outside V1, to transfer the facilitation from the sensory to the motor level.A remaining question is whether the visuo-auditory interactions reported here at the level of V1 and expressed as a reduction in neuronal latency, represent a real multisensory integration or only a sensory combination [119]. In our protocol, auditory and visual stimuli are not redundant signals as the sound has no meaning to perform the task and thus in this way, we should refer to bimodal interactions in V1. However, at the behavioral level, the observed shortening of saccade latency in the bimodal conditions is a phenomenon generally attributed to multisensory integration processing [72]. It is possible that the reduction in latency, especially because it affects mainly the longer ones, will induce a higher temporal coherence of the visual responses across V1. Such a processing has been suggested to increase the cortical synchronization which in turn enhances the speed and reliability of the visual responses [120], and thus could participate to the reduction of RT in bimodal conditions.ConclusionTo conclude, our results provide further evidence of the various roles of monkey area V1 in visual perception. Area V1 receives feedback projections from a large number of cortical areas [121]. V1 is connected with areas located at higher levels of the visual processing hierarchy [122,123], with non-visual sensory areas as described above, as well as with the area prostriata [124] which might constitute a gateway to the motor system [125]. This connectivity pattern could be the anatomical support of the neuronal modulation of V1 responses by higher cognitive processes such as attention mechanisms [126,127] or memory tasks [128,129]. The present results suggest that multisensory integration should be added to the list of cognitive processes performed in V1.Authors' contributionsYW carried out all experiments and performed statistical analysis. YW, SC, YT and BP participated in the design of the study, the data analysis and wrote the manuscript. All authors read and approved the final manuscript.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2527682.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2527682",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2527682\nAUTHORS: Thomas Tully, Régis Ferrière\n\nABSTRACT:\nIn a variable yet predictable world, organisms may use environmental cues to make adaptive adjustments to their phenotype. Such phenotypic flexibility is expected commonly to evolve in life history traits, which are closely tied to Darwinian fitness. Yet adaptive life history flexibility remains poorly documented. Here we introduce the collembolan Folsomia candida, a soil-dweller, parthenogenetic (all-female) microarthropod, as a model organism to study the phenotypic expression, genetic variation, fitness consequences and long-term evolution of life history flexibility. We demonstrate that collembola have a remarkable adaptive ability for adjusting their reproductive phenotype: when transferred from harsh to good conditions (in terms of food ration and crowding), a mother can fine-tune the number and the size of her eggs from one clutch to the next. The comparative analysis of eleven clonal populations of worldwide origins reveals (i) genetic variation in mean egg size under both good and bad conditions; (ii) no genetic variation in egg size flexibility, consistent with convergent evolution to a common physiological limit; (iii) genetic variation of both mean reproductive investment and reproductive investment flexibility, associated with a reversal of the genetic correlation between egg size and clutch size between environmental conditions ; (iv) a negative genetic correlation between reproductive investment flexibility and adult lifespan. Phylogenetic reconstruction shows that two life history strategies, called HIFLEX and LOFLEX, evolved early in evolutionary history. HIFLEX includes six of our 11 clones, and is characterized by large mean egg size and reproductive investment, high reproductive investment flexibility, and low adult survival. LOFLEX (the other five clones) has small mean egg size and low reproductive investment, low reproductive investment flexibility, and high adult survival. The divergence of HIFLEX and LOFLEX could represent different adaptations to environments differing in mean quality and variability, or indicate that a genetic polymorphism of reproductive investment reaction norms has evolved under a physiological tradeoff between reproductive investment flexibility and adult lifespan.\n\nBODY:\nIntroductionAll organisms experience environmental variation, and environmental variation is a fundamental ingredient of the evolution of organismal diversity. Life history attributes are, by definition, closely tied to Darwinian fitness and they occur in extraordinarily diverse combinations [1], [2]; therefore life history evolution should be particularly revealing about the relation between environmental variation and evolutionary change [1], [3].How environmental variation influences the evolution of life history traits depends on the scale over which environmental conditions vary [2]–[5]. When environmental variation operates on large temporal and/or spatial scales compared to population persistence or dispersion, constant, genetically fixed traits are expected to evolve within populations, and variation to evolve between populations. When the temporal/spatial scale of environmental variation is commensurate to the organism's generation time or home range, the evolution of developmental plasticity is expected, whereby the individual's traits are fixed by the environmental conditions experienced during ontogeny.When environmental variation occurs on even faster/shorter scales, an individual is likely to experience different environmental conditions during its lifetime. Fast/short-scale environmental variation can select for life history strategies that consist in genetically determined rules by which single individuals respond to environmental fluctuations. The strategy may be purely probabilistic, as with so-called bet-hedging strategies [6], [7], where the rule is reduced to expressing a certain trait (or trait value) with a genetically determined probability. When environmental variation has some degree of predictability, another type of adaptation is expected: ‘phenotypic flexibility’ (also called ‘flexible phenotypic plasticity’ [8], ‘reversible phenotypic plasticity’ [9], ‘facultative adjustment’ [10] and ‘context dependence’ [11], [12]), i.e. the rapid adjustment of labile phenotypic traits in response to fast/short scale variation in environmental conditions. One would expect life history flexibility to be a common adaptation to microenvironmental variability. Yet surprisingly little is known, both theoretically and empirically, about the occurrence and evolution of adaptive life history flexibility [13]–[15].Phenotypic flexibility is expected to evolve in fitness traits, of which egg size has received much attention as a form of pre-natal maternal care enhancing the chance of offspring survival under adverse environmental conditions [2], [16]. Egg size along with clutch size and reproductive investment form a complex of functionally related traits [16], [17]. Our research program aims at investigating the phenotypic expression, genetic architecture, and long-term evolution of flexibility in this complex of reproductive traits, in response to rapid changes in food and social conditions. Arthropods have provided outstanding model systems for the study of life history evolution [16]. Here we introduce the collembola Folsomia candida, a widespread parthogenetic microarthropod, as a new model organism with several interesting features for the evolutionary analysis of life history traits in variable environments: an asexual reproductive system, a relatively short generation time, a high sensitivity to environmental conditions, including food availability, and the feasibility of non-invasive, semi-automated counting and measurements of individuals and eggs.In this report, we address the following questions: (1) How flexible are egg size, clutch size, and maternal reproductive investment in response to sudden changes in dietary and social conditions? (2) Does the degree of flexibility differ between traits? (3) What fitness benefits do reproductive adjustments carry? (4) How much genetic variation is there in the mean and flexibility of reproductive traits? (5) What are the consequences of different amounts of genetic variation in the flexibility of different traits on the genetic correlations observed under different environmental conditions? (6) How did contemporary variation in reproductive flexibility evolve?Our experimental investigation of reproductive flexibility uses the parthenogenetic (all-female) springtail Folsomia candida Willem (Collembola, Isotomidae) [18] as model organism (see Materials and Methods, section A). Springtails from eleven genetically distinct clones [19] were kept in harsh environmental conditions set by high density and low food ration. After about three months, individuals were isolated. Then they were fed ad libitum and their body size and reproductive behavior (egg size, clutch size, reproductive investment) were monitored for two weeks (see Materials and Methods, section B, Experiments 1 and 2, and section C). This experimental design enabled us to study how the reproductive traits covaried plastically in response to the environmental change between ‘bad’ and ‘good’ conditions. By comparing the eleven clones, we could also measure the genetic variability (heritability) of these traits and their flexibility (see Materials and Methods, sections C and D). A separate experiment was performed to assess the adaptive value of flexible adjustments, by measuring the relationship between egg size, offspring size and offspring survival under either bad or good conditions (see Materials and Methods, section B, Experiment 3, and section D). We looked for potential costs of flexibility by correlating a measure of reproductive flexibility with mortality rates among clones. Finally, we used the phylogeny of the clones to perform a comparative analysis of their flexible traits, in order to gain insight into the origin and diversification of reproductive flexibility (see Materials and Methods, section D).ResultsReproductive traits are flexibleA marked decrease in egg size associated with increasing clutch size occurs 6 days after the release of crowding and dietary restriction (Figure 1a, b)— a time lag that exactly equals the minimal inter-clutch interval (mean inter-clutch interval = 6.7 days, 95% confidence interval = [5.9; 8.9], n = 51). Clutches laid during the first period (P1, day 1 to 6) come from a reproductive cycle that began in the crowded-dietary restricted environment. Clutches laid during the second period (P2, from day 7 onward) are on average composed of smaller (−7.5%, χ2\n1 = 30.7, P<0.001) but more eggs (+231%, χ2\n1 = 89.8, P<0.001) than in P1 (Figure 1).10.1371/journal.pone.0003207.g001Figure 1Reproductive adjustments after release of crowding and dietary restriction: (a) egg size (mean per clutch), (b) individual clutch size.Solid line: smooth spline function fitted to data. The effect of time on egg size and clutch size was analyzed by contrasting two linear models: Egg size (or Clutch size) = Body length+Clone+Time (model 1), and Egg size (or Clutch size) = Body length+Clone+Period (model 2). Model 2 involved two consecutive periods; by varying the limit between the two periods, we could examine whether specific parameterization of model 2 made discrete time (period effect) a better model than continuous time. The two models were compared by means of the ratio of the residual sum of square (dashed line). For both egg size and clutch size, model 2 became superior to model 1 when the period limit was close to 6 days (ratio<1). Note that the plotted values of clutch size and egg size are values corrected for female body length (i.e. the measurements are standardized for a 1.6 mm long individual).In the control experiment (see Materials and Methods, section B, experiment 2), when controlling for clone, food ration and body size, we found no effect of maternal age on egg size (χ2\n1 = 0.39, P = 0.53) and a negative effect of maternal age on clutch size (−0.2 egg/day, χ2\n1 = 82, P<0.001) which is much smaller than, and opposite to the treatment effect evidenced in the main experiment. Thus, the period effect is likely to be due to the sudden change in environmental conditions rather than to a confounded effect of maternal age.Reproductive adjustments confer fitness benefitsIn order to probe the adaptive significance of reproductive flexibility, we assessed the effect of environmental conditions (crowded and dietarily restricted conditions versus isolation and full feeding) on maternal reproductive investment and the relation between egg size and juvenile quality (see Materials and Methods, section B, experiment 3). The Winkler-Wallin optimality model [2], [20] makes a key prediction from the adaptive hypothesis: under poor environmental conditions [16], [21], low maternal reproductive investment is expected while bigger eggs associated with greater nutritional provisions should result in larger larvae that survive better.The prediction is upheld in F. candida. Reproductive investment is uniformly low among clones in period P1, and rises significantly in P2 (clutch size multiplied by 2.5, χ2\n1 = 9.67, P = 0.0018). Body length measured within 20 h after birth on 210 neonates from 41 clutches was positively correlated to the mean volume of the eggs from which they hatched (cor = 0.64, 95%CI = [0.41; 0.79], t39 = 5.2, P<0.001, Figure 2a). Offspring survival was affected by dietary and crowding conditions: the mortality rate was multiplied by 12 under high density and starvation (95%CI = [6.2; 23.4], |z| = 7.3, P<0.001, Figure 2b). Moreover, survival was affected by an interaction between dietary/crowding conditions and mean egg size (|z| = 2.8, P = 0.005, Figure 2c, d). During the first month of life, under high density and food deprivation, clutches containing larger eggs produced individuals surviving longer than clutches with smaller eggs (|z| = 3.94, P<0.001, Figure 2c): a 10% increase in egg volume decreased the mortality rate by 31% (95%CI = [17%; 43%]). In contrast, under low density and full feeding variation in egg size did not affect survival (|z| = 1.02, P = 0.31, Figure 2d).10.1371/journal.pone.0003207.g002Figure 2Egg size, offspring size, and offspring survival in the two environments (Experiment 3).(a) Correlation between egg size and newborn body length. Egg size and newborn body length size (mean per clutch +/− SE) are positively correlated. Measurements from 20 clutches laid during the first period (open circles) and 21 clutches laid during the second period (closed circles). (b) Survival curves (and 95% confidence intervals) in the no food (and high density) and ad libitum (and low density) food treatments. In the survival analysis we used the death events that occurred over the dotted horizontal (80% limit, see Materials and Methods, section D). Some individuals in the no food treatment survived very long probably because they could scavenge on dead bodies. (c) and (d) Association between egg size (mean per clutch) and offspring survival, depending on food availability. Martingale residuals are computed from Cox proportional hazard models not including egg size as a covariable (see methods for details). Non randomness in the residuals is evidenced by a local polynomial regression fit (curves, computed using scatter.smooth function in software R 2.1). (c) In the no food environment, residuals decrease with egg size: neonate issued from large eggs were more represented among old survivors, whereas those that hatched from small eggs had a higher mortality rate. (d) With food ad libitum, mortality rate tend to increase (not significantly) with egg size.How do reproductive adjustments vary among individuals?In order to analyze the structure of variation of reproductive adjustments among individuals, we begin with an examination of within-environment patterns. Egg size is related to clutch size (controlled for mother's body length) differently among periods (χ2\n1 = 6.14, P = 0.013, Figure 3a): in P1, egg size shows a negative yet non-significant correlation with clutch size (cor = −0.19, 95%CI = [−0.44; 0.08], t52 = −1.4, P = 0.16), whereas in P2, females that produce larger clutches also lay bigger eggs (cor = 0.27, 95%CI = [0.06; 0.45], t86 = 2.61, P = 0.01). These results contrast with the classic assumption of a negative correlation (tradeoff) between egg size and clutch size. In fact, many studies have demonstrated a phenotypic tradeoff between offspring size and number [16], but few of them have controlled for underlying genetic differences between individuals [22]. How much does genetic variation contribute to variation in egg size, clutch size, and reproductive investment within each period? When taking within-period genetic variation into account, no physiological tradeoff between egg size and clutch size could be detected: within-clones residuals for egg size and clutch size are not correlated, neither in P1 (t52 = 0.005, P = 0.99) nor in P2 (t86 = 0.33, P = 0.74, Figure 3b).10.1371/journal.pone.0003207.g003Figure 3Phenotypic correlation structure of egg size and clutch size.Open circles: data from period P1, closed circles: period P2. The 90% concentration ellipses are indicated for both periods. For each measurement of clutch and egg size, maternal body length is taken into account and standardized to 1.6 mm. (a) Global phenotypic correlations between egg size and clutch size. (b) Within-clones residuals correlations between egg size and clutch.Despite a high level of intra-clutch egg volume variation that account for 50% of total variance, egg volume expressed in each environment was found to be highly heritable (H2 = 25%, 95%CI = [21], [29], χ2\n1 = 49.8, P<0.001). During P1 genetic variation had no effect on clutch size (χ2\n1 = 0.79, P = 0.37, Figure 4a) or reproductive investment (χ2\n1 = 0.26, P = 0.88, Figure 4b) whereas in P2 both traits were found to be heritable (clutch size: H2 = 42%, 95%CI = [14; 67], χ2\n1 = 10.6, P = 0.001; reproductive investment: H2 = 48%, 95%CI = [30; 64], χ2\n1 = 34.6, P<0.001). Egg size and clutch size are genetically correlated within each period, and, remarkably, these genetic correlations are reversed between periods: from negative in P1 to positive in P2 (Figure 4a; in P1: cor = −0.81 [−0.96; −0.31], t7 = −3.64, P = 0.008; in P2: cor = +0.70 [0.18; 0.92], t10 = 2.96, P = 0.016). Genetic correlations between reproductive investment and egg size or clutch size are nonsignificant in P1, but are strongly positive in P2 (Figure 4b, c. Egg size and reproductive investment in P1: cor = +0.14 [−0.50; 0.68], t9 = 0.44, P = 0.67; in P2: cor = +0.87 [0.56; 0.96], t9 = 5.22, P<0.001. Clutch size and reproductive investment in P1: cor = −0.18 [−0.75; 0.55], t7 = −0.48, P = 0.64; in P2: cor = +0.84 [0.49; 0.96], t9 = 4.71, P = 0.001).10.1371/journal.pone.0003207.g004Figure 4Genetic correlations between egg size, clutch size and reproductive investment.Bivariate reaction norms (grey lines) and 90% concentration ellipses for period P1 (open circles) and P2 (filled circles). (a) Egg size and clutch size. (b) Clutch size and reproductive investment. (c) Egg size and reproductive investment. Only data from period P2 have been plotted for clones BV and HA because these clones laid too few eggs in period P1. The measurements are standardized for a 1.6 mm long female.Genetic variation in reproductive flexibilityThe great variation of genetic correlations between egg size, clutch size and reproductive investment is the consequence of flexibility in these traits, and an amount of genetic variation in flexibility that differs among traits [23]. There is no genetic variation in egg size flexibility (χ 2\n1 = 0.01, P = 0.92) whereas there is strong genetic variation in the flexibility of reproductive investment (H2 = 34.5%, 95%CI = [18.2; 49.8], χ 2\n1 = 23.0, P<0.001). In effect, the degree of flexibility in reproductive investment varies from no increase in clone BR to an 8-fold increase in clone US (Figure 4b, c). Thus, whereas all genotypes show a similar response in egg size to environmental change, the degree to which clutch size is affected is not simply determined by a physiological trade-off with egg size—it also integrates the flexibility of maternal investment in reproduction. In genotypes producing consistently (i.e. on average across periods) bigger eggs, reproductive investment is more flexible (Figure 5a; correlation of genetic values of mean egg size over both periods with genetic values of reproductive investment flexibility: cor = +0.74 [0.26; 0.93], t9 = 3.36, P = 0.008), and disproportionately larger clutches are produced under favourable conditions, as permitted by a larger increase of reproductive investment.10.1371/journal.pone.0003207.g005Figure 5Genetic correlations between flexibility of reproductive investment and (a) egg size, (b) adult mortality.90% concentration ellipses are indicated. Genetic values of relative risk of mortality (clone AP is taken as a reference with a relative risk of one) come from an independent experiment where the longevity of 20 individuals per clone was measured and analyzed through a Cox proportional hazard model. Mortality risk differs among clones (χ2\n1 = 109, P<0.001). For each measurement of egg size and flexibility of reproductive investment, maternal body length is taken into account and standardized to 1.6 mm.This suggests that resource acquisition strategies may differ among clones [17], [24]. According to this interpretation, under crowded conditions and food deprivation, genetic variation in resource acquisition is weakly expressed and only genetic variation in resource allocation is detected, leading to the negative genetic correlation between clutch size and egg size (Figure 4a). In contrast, under isolated conditions and full feeding, genetic variation in resource acquisition is fully expressed, thus masking genetic variation in resource allocation and leading to positive correlations between egg size, clutch size and reproductive investment (Figure 4).The maintenance of genetic variation in reproductive flexibility could thus be explained by the tradeoff that life history theory predicts between resource acquisition strategies and survival [25], [26]. Specifically, the tradeoff hypothesis implies that the resource acquisition strategy underlying high flexibility in reproductive investment should suffer the genetic cost of higher adult mortality. This hypothesis is supported by our data: the adult risk of mortality is higher in clones that cumulate the benefits of larger egg size in both periods (correlation between genetic values of mean egg size and mortality risk (relative to clone AP): cor = +0.79 [0.37; 0.94], t9 = 3.9, P = 0.003) and high flexibility in reproductive investment (correlation between genetic values of reproductive investment flexibility and mortality risk: cor = +0.84 [0.48; 0.96], t9 = 4.6, P = 0.001, Figure 5b).Genetic costs and long-term evolution of reproductive flexibilityA hierarchical cluster analysis (see Materials and Methods, Section D) made on the genetic values of egg size and reproductive investment highlights the existence of two genetically distinct reproductive strategies (Figure 4c, Figure 6): a high-flexibility strategy, HIFLEX, characterized by larger egg size and highly flexible reproductive investment (clones DK, US, GM, PB, TO, WI), and a low-flexibility strategy, LOFLEX, which produces small eggs in both periods and barely increases its reproductive investment in response to the environmental amelioration (clones BR, BV, HA, GB, AP).10.1371/journal.pone.0003207.g006Figure 6Phylogeny and life history evolution.The phylogeny (left) is a strict consensus cladogram with proportional branch lengths obtained by analyzing two types of molecular characters (RAPD markers and rRNA sequences [19]). The collembola Isotoma viridis Bourlet was used as an outgroup. The topology of the upper clade is unresolved due to contradictory signals—not because of lack of genetic variation. The life history distance tree (right) was derived from a hierarchical cluster analysis performed on the genetic values of egg size and reproductive investment expressed in P2 (cf. Figure 4c and methods). The two trees are highly congruent (comparison of the two associated distance matrices, 1000 permutations, Friedman's χ2\n1 = 96, P<0.001).This genetic clustering into HIFLEX and LOFLEX strategies shows remarkable congruence with the clones' phylogeny (Figure 6). The two strategies arose once along with the early divergence of two major branches of the evolutionary tree, and the distribution of genetic trait values measured in P2 (Figure 4c) almost perfectly matches the subsequent branching structure of the tree.DiscussionPhenotypic plasticity can exist in various guises, which are encapsulated theoretically by the concept of ‘reaction norm’—the potential phenotypic response to different environments (see [8] and [27] for reviews). Reaction norms can be either inflexible, in which a characteristic once determined is never changed later in the organism's life, or they can be flexible, in which a characteristic can be altered more than once in the development of the same individual. To date, life history theory has focused on life history traits, such as growth rate or age at maturity, whose phenotypic variation is described by inflexible reaction norms [8], [9], [28]. In animals, the evolutionary analysis of life history flexibility has been limited chiefly to maternal adjustment of sex ratio [10] and sex allocation [28], and to the context-dependent expression of sexual traits [11], [29] and offspring dispersal [30], [31].Here we have shown that collembola are capable of remarkably fast and large adjustments of their reproductive traits (reproductive investment, clutch size, egg size) in response to sudden environmental change in food and density conditions. We documented the phenotypic expression, genetic variation, and long-term evolution of reproductive flexibility by means of a comparative analysis of eleven clones from different origins worldwide. Our results (i) provide evidence for the adaptiveness of reproductive flexibility, (ii) reveal that genetic variation in flexibility differs between traits (which has consequences for the observed genetic correlations between traits in the different environments experienced by individuals during their lifetime), and (iii) suggest the importance of resource acquisition tradeoffs to understand the origin, maintenance and evolution of genetic variation in the flexibility of resource allocation traits.Hereafter we discuss our results mostly in the light of recently published analyses of adaptive plasticity of egg size, which in many cases might pertain to the flexible kind documented in our study system. Thus, the growing understanding of the evolution of egg size plasticity provides a useful background for interpreting and discussing our results.Adaptive flexibility of egg sizeAs for any biological trait, the adaptive hypothesis implies heritable variation, and differential costs and benefits. In Folsomia candida, egg size strongly correlates with offspring size, and offspring size has a marked, positive effect on juvenile survival under poor food conditions, thus providing evidence for a fitness benefit from egg size adjustment. Although there is no genetic difference between clones in egg size plasticity, the high heritability of egg size in both periods supports the genetic basis of the egg size reaction norm. Thus, our analysis add to a relatively short list of experimental studies that have demonstrated cross-generational adaptive plasticity via maternal manipulation of offspring size, mainly in invertebrate model systems—Daphnia\n[32], the seed beetle Stator limbatus\n[33], the tropical butterfly Bicyclus anynana\n[34], and the bryozoan Bugula neritina\n[31]—and in the Trinidadian guppy Poecilia reticulata\n[25], [35]. The reversible plasticity, i.e. flexibility, of egg size has been documented in the Ural owl Aegolius funereus, a long-lived bird that preys on highly fluctuating populations of voles; pedigree analysis and strong correlative evidence show that egg size is heritable and adjusted in response to variation in prey density, and supplementary experiments suggest that these adjustments do confer fitness benefits [15]. In the common lizard Lacerta vivipara, life history flexibility manifests itself in response to multiple cues, but its putative fitness benefits remain elusive [36], [37].Because of the effect of egg size on individual fitness, egg size has long been viewed as a relatively canalized trait in animals (see references in [38], [39])—an assumption that has been revisited in the light of growing evidence for genetic variation in egg size [38]. In beetles [40] and guppies [35], there is genetic variation for egg size mean and plasticity. A selection experiment in beetles found that selection for increased egg size resulted in increased egg size plasticity, but only in one particular environment [41], whereas in guppies, increased offspring size plasticity was associated with decreased offspring size [35]. In our collembola, lack of genetic variation in egg size flexibility may indicate canalization or convergent evolution. In either case, our results suggest that the evolution of mean egg size can be relatively decoupled from the evolution of egg size flexibility, due to constraints (e.g. egg size flexibility hit its physiological limit, as discussed below) or because the determination of mean egg size and the regulation of egg size flexibility involve different genes or genetic pathways [41], [42].Environmental variation and genetic correlationsWithin periods, egg size genetically correlates with clutch size and reproductive investment. These genetic correlations show a striking reversal between periods, from strong negative in the bad period to strong positive in the good period. This finding adds to growing empirical evidence that genetic correlations can shift, even switch sign, across environments [22], [43]; our results are distinctive as they demonstrate reversals of genetic correlations within the individual lifetime.The bad period is characterized by uniformly low reproductive investment among clones. This is consistent with the hypothesis of unfavorable conditions decreasing heritability as a consequence of selection favoring alleles (in loci promoting resource allocation to reproduction) that are not expressed in periods of food shortage 44, 45. Poor environmental conditions generate strong viability selection on egg size. Thus, the negative genetic correlation between egg size and clutch size in the bad period is consistent with the classic hypothesis that egg size and clutch size are optimized by selection, with harsher environmental conditions favoring larger eggs in smaller clutches [21]. The intriguing result, however, is that the physiological tradeoff expected to constrain the optimization process [2], [21], [46] could not be detected. When controlling for period, maternal size and genotype, no relation exists between egg size and clutch size in either environment (Figure 3b). This puzzling result warrants further investigation.In the good period, genetic variation is expressed in all three reproductive traits: egg size, clutch size, and reproductive investment. The evolution of reproductive investment has long been regarded as decoupled from the evolution of clutch size and egg size [21], but recent empirical studies have cast doubt on this fundamental assumption [22], [34], [47]. Yet even for constant environments, surprisingly little theory is available to predict the outcome of the joint evolution of egg size, clutch size, and reproductive investment. The Winkler-Wallin model [20] remains the chief theory for the joint evolution of all three traits; it predicts that better environmental conditions should select for larger reproductive investment, smaller eggs, and larger clutches—disproportionately so as a consequence of larger reproductive investment. The genetic correlation found in the good period conforms only partly to that prediction: larger reproductive investment is associated with larger clutches but also larger, rather than smaller eggs; and a larger offspring size is expected to evolve under harsher, not milder environments. How can we resolve these discrepancies? Variation in the expression of flexible traits across environments cannot be fully understood without considering the evolution of flexibility itself [5], [24], [48].Limits and costs of reproductive flexibilityIn contrast with egg size flexibility, the flexibility of reproductive investment shows substantial genetic variation in Folsomia candida. Reproductive investment flexibility is genetically and positively correlated with mean reproductive investment and mean egg size. A harsher environment that selects for larger mean egg size may also promote a greater ability to adjust reproductive investment in response to more intense or more frequent environmental fluctuations, e.g. a higher rate of transition from good to bad conditions [9], [45]. Larger mean reproductive investment may then evolve simply as a consequence of a steeper reaction norm [49].How limited or constrained would the evolution of reproductive flexibility be? Egg size flexibility does not seem limited by a response time lag [48]: reproductive traits can be adjusted even once the individual's reproductive cycle has started, which suggests that more energy can be channelled into reproduction as soon as new resources become available (see Figure 1b: those clutches laid during P1 follow a trend for larger size as the laying date advances: +2.9 eggs/day, χ 2\n1 = 8.9, P = 0.003). But the common pattern of egg size adjustments across clones might reveal the ‘phenotypic range’ limit of plasticity [48]. Thus, the lack of genetic variation in egg size flexibility would be consistent with a common physiological limit hit by the evolution of egg size flexibility in all populations. As a consequence, the minimum egg size expressed in good environments would be consistently greater in populations evolving higher mean egg size. Alternatively (yet non exclusively), larger size at birth might evolve as a correlated response to selection for larger reproductive investment [35].Higher adult mortality has been hypothesized as a potential genetic cost for increasing reproductive plasticity [26]; our finding of a strong positive genetic correlation between reproductive investment flexibility and adult mortality upholds this prediction. Consistently with our experimental results, correlational data in Ural owls also show that the most reproductively flexible individuals have shorter reproductive lifespan [15]. In contrast, the experimental analysis of the mean and plasticity of survival, growth, and reproductive effort in the Pacific oyster Crassostrea gigas raised under different food conditions, revealed substantial genetic variation in reproductive effort plasticity and in mean survival; but the degree of plasticity in reproductive effort and mean survival covaried positively [24] —a pattern explained by hypothesizing that reproductive effort plasticity trades off with sensitivity to random factors of mortality [24]. We also expect that in more variable environments, higher adult mortality and higher reproductive investment co-evolve along a basic reproduction/survival trade-off [50], [51]. In collembola, the positive genetic correlation between mean egg size and mean reproductive investment, and the negative covariation of mean egg size or mean reproductive investment with adult survival are also compatible with that prediction. The complexity of the picture exemplifies how challenging the measure of genetic costs of phenotypic plasticity remains [52]–[54].Evolutionary scenariosGenetic correlations obtained from clones of vastly different origins (when it is known ) may reveal patterns of adaptations to a range of selective environments experienced by each of the original populations. We know from laboratory experiments that collembola population dynamics can respond dramatically to changes in patterns of environmental variation and autocorrelation [55] and trophic interactions [56]. Thus, changes in environmental harshness and variability are likely to affect the outcome of competition between genetic variants. In this context, our interpretations of genetic correlations yield two main adaptive scenarios (Figure 7), in which different life history adaptations evolve in response to different degrees of environmental harshness and variability. Resolving these alternate scenarios requires that we learn more about the ecology and population genetics of natural collembola populations. We also need to elucidate the physiological basis of resource allocation between life history traits and their flexibility [41], [57]. Indeed, genetic correlations that reflect different adaptations among populations provide little insight into the structure of physiological tradeoffs that prevail in each population. To this end, the expression of different traits values across environments by the same individuals during their lifetime may present new and fruitful opportunities [13].10.1371/journal.pone.0003207.g007Figure 7Adaptive scenarios for the evolution of reproductive investment flexibility.Harsher and more variable environments select for higher mean egg size and higher flexibility in reproductive investment. Larger mean reproductive investment and shorter adult life span evolve as correlated responses (gray). Scenario (a) emphasizes a tradeoff (dotted arrow) between adult lifespan and reproductive investment flexibility. Alternatively, scenario (b) emphasizes a tradeoff between adult lifespan and reproductive investment mean.The interpretation of genetic differentiation as a response to different selective environments is tantalizing but remains hypothetical. A contrasting view would assume that the genetic variation documented here actually reflects the genetic polymorphism of natural populations. In this case, the single origin and evolutionary divergence of HIFLEX and LOFLEX could be interpreted as the result of disruptive selection operating on a single (i.e. common to all populations) tradeoff between adult survival and reproductive investment mean or flexibility [58]. The breakdown of genetic correlations at smaller phylogenetic scale might indicate that the fitness landscape over which phenotypes evolve becomes flatter away from the original branching phenotype – an assumption that is consistent with theory [58], the empirical analysis of egg size flexibility in Ural owls [15], and the hypothesis of nonlinear selection to explain the breakdown of genetic correlations in laboratory evolution of Drosophila\n[59]. A similar effect – dependence upon phylogenetic scale of the tradeoff underlying variation in reaction norms – was suggested by data on thermal reaction norms of body growth in fish [60].PerspectivesThe evolution of life history flexibility, i.e. the adaptive, context-dependent adjustment of fitness traits by individuals during their lifetime, raises exciting challenges at the crossroads of genetics, physiology, ecology and evolution. While future work on the collembola system may afford further insights into how life history traits evolve as reaction norms, there is an urgent need to develop general models and theory that will form the conceptual framework of empirical studies. Simple theoretical models of the evolution of life history traits are of limited value in heterogeneous environments in which complexes of traits covary and thus co-evolve, and the complex of traits that coevolve varies with environmental conditions [22]. There are still very few general models of the evolution of reversible plasticity [9], [49], and to our knowledge, none that involves the population physiological structure needed to address the evolution of flexibility in life history traits.The development of an evolutionary theory for life history reaction norms will be useful to address the multidimensionality of environmental and physiological cues [61]–[64], to dissect the physiological and genetic architecture of flexibility in complexes of functionally related traits [13], [57], [65], and to investigate the reciprocal influence of phenotypic flexibility and evolutionary dynamics [66]–[71]. Mirroring research perspectives on developmental plasticity [72], one of the next frontiers will be to disentangle the web of ecological and evolutionary feedbacks between life history flexibility and the community and ecosystem contexts of population adaptation.Materials and MethodsA. Folsomia candida as a model organism\nFolsomia candida Willem 1912 (Collembola, Isotomidae) is a widespread parthenogenetic springtail [18] that is typically found in leaf litter, in caves [73], [74] and also in anthropic environments such as the dirt of plant pots [75]. Its natural density is known to vary greatly [76]. Individuals mature within two weeks and lay a clutch about once a week [77], [78]. Clutch size varies from less than ten eggs to more than 100; body length [79] and ration [75], [80], [81] are major influences of egg production.Clonal populations issued from one single female for each strain are maintained in our laboratory. All populations and single individuals monitored during the experiments were maintained in standard containers made of a polyethylene vial (diameter 52 mm, height 65 mm) filled with a 30 mm layer of plaster of Paris mixed with 600 µL of Pebeo® graphic Chinese ink to increase visual detectability of individuals and eggs against their background. The surface of the plaster was sandpapered and covered with a thin layer of a mixture of clay, Chinese ink and charcoal in order to fill up all tiny holes in the plaster that springtails could have used to lay eggs. All direct manipulations were done by using a pooter (for individuals) and a thin moisturized brush (for eggs).Food is provided in the form of small pellets of a mixture of dried yeast and agar in standardized concentration and volume (5000 µL water+80 mg agar+800 mg dried yeast, to produce pellets of 2 µL). All our stock cultures are provided with the same amount of food. Stock cultures and experimental populations are kept in incubators at 21±0.5°C, with a 12 h∶12 h light∶dark cycle and constant humidity (∼100%).B. Experimental designWe used eleven clones of Folsomia candida characterized by molecular markers [19] – nine from Europe (clones AP, BR, BV DK, GB, GM, HA, PB, TO) and two from North America (US, WI).Experiment 1: Transfer experiment to measure reproductive flexibilityThis experiment aimed at measuring the response of reproductive investment (egg size and clutch size, see section C for methodological details) to transfer from crowded and dietarily restricted conditions to isolation and full feeding. For each clone, four replicates of high density populations (ca. ∼40–50 ind./cm2) were provided with low food ration (∼1 µg dried yeast/ind/week) during three months. To mimic environmental amelioration, ten adult females of each clone were then isolated and fully fed (food pellets provided ad libitum). To reduce the influence of uncontrolled factors, we sampled young adults of similar size (size homogeneity between clones: F10–99 = 1.52, P = 0.14; mean body length = 1.47 mm, SE = 0.021).Experiment 2: Control for age effect on egg size and clutch sizeBecause there is no monitoring the reproductive characteristics of single individuals in high density populations, a simple control experiment was not feasible. Therefore, to test for any confounding effect of age with environmental change, we performed a complementary experiment by measuring egg size and clutch size produced by 20 isolated individuals raised at two contrasted rations (low food and ad libitum food), for each clone and over four months. In the low food treatment, a food pellet was available one day per week whereas in the high food treatment, food was provided ad libitum seven days per week. These females were of the same age as those of the reported experiment (younger than four months).Experiment 3: The effect of egg size on offspring survival in ‘good’ and ‘bad’ environmentsThe relationship between egg size and neonate size was documented by measuring body size in 210 neonates from 41 clutches within 20 h after hatching. The relation between egg size, juvenile size, and juvenile quality was assessed by measuring the survival of neonates in two contrasting environments: in the ‘bad’ environment, no food was provided to a cohort of ∼20 individuals; in the ‘good’ environment, food was provided ad libitum to isolated individuals. For each clone, the ‘bad’ environment treatment was carried out by isolating ca. 20 developed eggs obtained from four clutches laid by four females in the second week of the main experiment. For the clones GB and BR, only three clutches could be used, and only one for clone BV. The mortality curve of 811 neonates coming from these 39 clutches was estimated by monitoring the number of collembola still alive at regular time intervals. Each container was inspected twice a day until all the eggs had hatched, then every other day during one month.In the ‘good’ environment treatment, 10 neonates issued from at least four different clutches were isolated for each clone immediately after birth and transferred to fresh rearing boxes. Unlimited food was provided to these 110 individuals by providing and regularly replacing food pellets (these individuals were also used in the ad libitum food treatment of the control experiment). The mortality curve was established by checking the boxes every day during three weeks, and every two to four days during the following three months. From month 4 to month 8 the boxes were inspected weekly.Because we were unable to assign an individual egg size to each neonate, only the mean egg size of the corresponding clutch could be analyzed as a factor of juvenile body length or survival; intra-clutch egg size variation was not taken into account.C. Measurements and data collectionIn the main experiment, rearing boxes were visually inspected twice a day (morning and evening) for clutches. When a new clutch was found, fecundity was measured by counting the eggs. Each clutch was then photographed with a digital camera (Nikon ® Coolpix 990) connected to an Olympus ® SZX12 stereomicroscope, after carefully spreading the eggs with a thin brush to facilitate egg contour detection through image analysis. Pictures were taken and egg size measurements (mean diameter and surface) were performed soon after the clutch had been laid (within 24 hours) to take advantage of the spherical shape of eggs (they become ovoid after the chorion tears, i.e. ∼after 3 days [75]. Egg size measurements were then converted into egg volume under the assumption of spherical shape. Digital pictures and image processing were also used to measure the body length of all females (from the front of the head to the rear of the abdomen) at the start and at the end of the first experiment and every week during the second experiment (control). We used the same method to measure the body length of new born individuals. Most females grew up during the experiments. We therefore estimated the body length of a female at each time she laid eggs by considering a linear body length growth trajectory during the intervals between two body length measurements.We used the ImageJ software for image analysis [82]. The repeatability of egg size measurement was assessed in an independent experiment, by measuring 67 eggs issued from four clutches, each of which was shot four times yielding a total of 268 measurements. Likewise, 400 measurements of body length were obtained from ten pictures of eight adults, analyzed five times. Repeatability is defined as the proportion of variance associated with differences between individuals [83]. Repeatability scored very high for both egg size (79%), and body length (96%).Overall, 93 of the 110 sampled females of the main transfer experiment laid at least one clutch; 51 laid two clutches. Of the 6627 eggs laid in the 144 clutches, 3377 were measured. Each clutch was assigned a maternal body length by assuming linear growth of the mother during the experiment. Maternal volume was estimated under a cylindrical shape approximation, by using body length and the relationship between body length and abdomen width estimated from an independent dataset from a preliminary experiment (abdomen width (mm) = 0.272*body length (mm) – 0.0536, R2 = 0.87, based on body size measurements made on 68 individuals ranging from 1.0 mm to 2.0 mm).Reproductive investment was defined for each individual as the total volume of eggs produced during each period divided by the duration of the period and by the mean volume of the female during that period (%volume.day−1). The total volume of eggs was measured for each female by the sum over clutches of clutch volume, the latter being estimated by the product of egg number by mean egg volume. Our analysis of reproductive investment thus takes into account females' reproduction schedule and females that did not reproduce during one or both periods.D. Statistical analysisBroad-sense heritabilities of reproductive traits' mean and flexibility\nEgg size, clutch size and reproductive investment were analyzed by using hierarchical mixed linear models (lme function of nlme package, R 2.1) with clone, mother (for egg size, clutch size and reproductive investment) and clutch nested within mother (for egg size) as random effects [84]. For clonal organisms, the relevant measure of genetic variance is the broad-sense heritability defined as the ratio of the among-clone component of variance to the total phenotypic variance: H2 = σ2\nG/σ2\nT\n[85]. Broad-sense heritabilities of the traits and of their flexibility, defined as the proportion of genetic to expected phenotypic variance for controlled body size, were calculated by using models with clone – for heritability of the mean trait – and interaction between clone and environment (period) – for heritability of the trait's flexibility – treated as random effects, and by comparing the variance component of these effects to the total variance. In the models used for computing heritabilities, variables of interest (egg size, clutch size and reproductive investment) are corrected for maternal body length; thus, broad-sense heritabilities are defined here as the proportion of genetic to expected phenotypic variance when body size is kept constant between individuals. Statistical significance for heritability of the traits or of their flexibility were assessed by comparing the full model to a model with no clone or period*clone random effect (likelihood ratio test, library lme\n[84]). Bootstrapping was used to compute mean values and confidence intervals for significant heritabilities (1000 resampling with replacement [86]).\nClutch size (for the analysis of egg size), maternal body length and period were treated as fixed effects. Statistical significance was assessed with log likelihood ratio tests [84] and model parameters were estimated by the restricted log-likelihood method. The lme package [84] was used to check the assumptions of models including mixed effects; variables were transformed whenever necessary. Robustness to outliers was tested by removing observations with large Cook distances; only robust results are presented here.Phenotypic and genetic correlationsCorrelations between egg size and clutch size were analysed by modelling egg size and clutch size with two independent linear mixed models, using maternal body length and period as fixed effects, and an interaction between clone and period as a random effect. Both traits were dependent additively on maternal body length and period (Figure 8). For each period, phenotypic correlations between egg size and clutch size were studied by correcting these variables for maternal body length (they were scaled to 1.6 mm, mean female length during the experiment, see Figure 3a). Similarly, the variables plotted and analyzed in Figure 1 and Figure 4 are controlled for maternal body length and scaled to a 1.6 mm long female. Within-clone phenotypic correlations were computed using model residuals, thus controlling for both maternal body length and genetic variation (Figure 3b). For each period, genetic correlations were sought between the genetic values of the traits, computed as the sum of the residuals of the models' random parts with the predicted value of the dependent variable for a 1.6 mm female. Genetic values of the flexibility of reproductive investment were computed as the difference between the genetic values of reproductive investment in each period.10.1371/journal.pone.0003207.g008Figure 8Clutch size, egg size and maternal body length.Clutch size (A) and egg size (B) as a function of maternal body length per period (open circles: P1; closed circles: P2). Egg size increases with maternal body length (χ2\n1 = 4.73, P = 0.029) similarly in both periods (χ2\n1 = 0.819, P = 0.36). These relations are materialized by regression lines (dotted lines) and lowess nonparametric regression lines (continuous line) for both P1 (black) and P2 (gray).Survival analysisIn the offspring survival experiment, mortality was analyzed with a Cox proportional hazards model (Coxph function from package survival, R 2.1 [87]). In order to fulfil the Cox proportional hazard assumptions, only the first 80% death events were included in the analysis for each food treatment; this threshold was reached within 50 days for the no-food treatment, and within 115 days for the ad libitum food treatment (Figure 2b). The potential for mortality correlation among groups of sisters within clutches was taken into account by computing a robust variance (cluster option). Ration was treated as a stratum variable (strata option) to allow for non proportional hazards when comparing the effect of egg size between the two food treatments [88]. The effect of egg size on survival is illustrated in Figure 2b & 2c by means of a graphical method that consists of plotting measurements of egg size against the residuals of a Cox model (known as martingale residuals [89]) that does not include egg size as a covariate. Martingale residuals can be interpreted as an excess of death given the model: positive values mean that the corresponding data have a shorter lifespan than predicted by the model whereas measurements with negative residuals have a longer lifespan than predicted. Therefore plotting these residuals against egg size reveals the underlying relationship between this variable (egg size) and the hazard rate (mortality).Cluster analysisTo build-up a life history distance tree (Figure 6), we used a hierarchical cluster analysis (hclust function in program R 2.1, single linkage method) performed on the genetic values (centred and standardized) of egg size and reproductive investment expressed in the second period (cf. Figure 4c).\n\nREFERENCES:\n1. StearnsSCHoekstraRF\n2000\nEvolution: An Introduction\nOxford\nOxford University Press\n2. RoffDA\n2001\nLife History Evolution\nSunderland, Massachusetts\nSinauer Associates, Inc\n527\n3. MeyersLABullJJ\n2002\nFighting change with change: adaptive variation in an uncertain world.\nTrends in Ecology & Evolution\n17\n551\n557\n4. LevinsR\n1968\nEvolution in Changing Environments\nPrinceton University Press\n5. MoranNA\n1992\nThe evolutionary maintenance of alternative phenotypes.\nAmerican Naturalist\n139\n971\n989\n6. EvansMEKFerrièreRKaneMJVenableDL\n2007\nBet hedging via seed banking in desert evening primroses (Oenothera, Onagraceae): demographic evidence from natural populations.\nAmerican Naturalist\n169\n184\n194\n7. EvansMEKDennehyJJ\n2005\nGerm banking: bet-hedging and variable release from egg and seed dormancy.\nQuarterly Review of Biology\n80\n431\n451\n16519139\n8. StearnsSC\n1989\nThe evolutionary significance of phenotypic plasticity - Phenotypic sources of variation among organisms can be described by developmental switches and reaction norms.\nBioscience\n39\n436\n445\n9. GabrielW\n2005\nHow stress selects for reversible phenotypic plasticity.\nJournal of Evolutionary Biology\n18\n873\n883\n16033559\n10. WestSASheldonBC\n2002\nConstraints in the evolution of sex ratio adjustment.\nScience\n295\n1685\n1688\n11823605\n11. SvenssonESheldonBC\n1998\nThe social context of life history evolution.\nOikos\n83\n466\n477\n12. BadyaevAVHillGE\n2002\nPaternal care as a conditional strategy: distinct reproductive tactics associated with elaboration of plumage ornamentation in the house finch.\nBehavioral Ecology\n13\n591\n597\n13. PiersmaTDrentJ\n2003\nPhenotypic flexibility and the evolution of organismal design.\nTrends in Ecology & Evolution\n18\n228\n233\n14. TullyT\n2004\nFacteurs maternels, génétiques et environnementaux de l'expression des traits d'histoire de vie chez le collembole Folsomia candida\nWillem\nUniversité Pierre et Marie Curie, Paris 6\n15. KontiainenPBrommerJEKarellPPietiainenH\n2008\nHeritability, plasticity and canalization of Ural owl egg size in a cyclic environment.\nJournal of Evolutionary Biology\n21\n88\n96\n18034804\n16. FoxCWCzesakME\n2000\nEvolutionary ecology of progeny size in arthropods.\nAnnual Review of Entomology\n45\n341\n369\n17. de JongGvan NoordwijkAJ\n1992\nAcquisition and allocation of resources: genetic (CO) variances, selection and life histories.\nAmerican Naturalist\n139\n749\n770\n18. FountainMTHopkinSP\n2005\n\nFolsomia candida (Collembola): a “standard” soil arthropod.\nAnnual Review of Entomology\n50\n201\n222\n19. TullyTD'HaeseCRichardMFerrièreR\n2006\nTwo major evolutionary lineages revealed by molecular phylogeny in the parthenogenetic collembolan Folsomia candida.\nPedobiologia\n50\n95\n104\n20. WinklerDWWallinK\n1987\nOffspring size and number: a life history model linking effort per offspring and total effort.\nAmerican Naturalist\n129\n708\n720\n21. SmithCCFretwellSD\n1974\nOptimal balance between size and number of offspring.\nAmerican Naturalist\n108\n499\n506\n22. CzesakMEFoxCW\n2003\nEvolutionary ecology of egg size and number in a seed beetle: Genetic trade-off differs between environments.\nEvolution\n57\n1121\n1132\n12836828\n23. StearnsSde JongGNewmanB\n1991\nThe effects of phenotypic plasticity on genetic correlations.\nTrends in Ecology & Evolution\n6\n122\n126\n21232440\n24. ErnandeBBoudryPClobertJHaureJ\n2004\nPlasticity in resource allocation based life history traits in the Pacific oyster, Crassostrea gigas. I. Spatial variation in food abundance.\nJournal of Evolutionary Biology\n17\n342\n356\n15009268\n25. ReznickDYangAP\n1993\nThe influence of fluctuating resources on life history patterns of allocation and plasticity in female guppies.\nEcology\n74\n2011\n2019\n26. ReznickDNunneyLTessierA\n2000\nBig houses, big cars, superfleas and the costs of reproduction.\nTrends in Ecology & Evolution\n15\n421\n425\n10998520\n27. ViaSGomulkiewiczRDejongGScheinerSMSchlichtingCD\n1995\nAdaptive phenotypic plasticity - Consensus and controversy.\nTrends in Ecology & Evolution\n10\n212\n217\n21237012\n28. BrauerVSScharerLMichielsNK\n2007\nPhenotypically flexible sex allocation in a simultaneous hermaphrodite.\nEvolution\n61\n216\n222\n17300440\n29. BadyaevAVDuckworthRA\n2003\nContext-dependent sexual advertisement: plasticity in development of sexual ornamentation throughout the lifetime of a passerine bird.\nJournal of Evolutionary Biology\n16\n1065\n1076\n14640398\n30. MassotMClobertJ\n1995\nInfluence of maternal food availability on offspring dispersal.\nBehavioral ecology and sociobiology\n37\n413\n418\n31. AllenRMBuckleyYMMarshallDJ\n2008\nOffspring size plasticity in response to intraspecific competition: An adaptive maternal effect across life-history stages.\nAmerican Naturalist\n171\n225\n237\n32. GliwiczZMGuisandeC\n1992\nFamily-planning in Daphnia - Resistance to starvation in offspring born to mothers grown at different food levels.\nOecologia\n91\n463\n467\n28313496\n33. FoxCWThakarMSMousseauTA\n1997\nEgg size plasticity in a seed beetle: An adaptive maternal effect.\nAmerican Naturalist\n149\n149\n163\n34. FischerKBotANMBrakefieldPMZwaanBJ\n2006\nDo mothers producing large offspring have to sacrifice fecundity?\nJournal of Evolutionary Biology\n19\n380\n391\n16599914\n35. BasheyF\n2006\nCross-generational environmental effects and the evolution of offspring size in the Trinidadian guppy Poecilia reticulata.\nEvolution\n60\n348\n361\n16610325\n36. LorenzonPClobertJMassotM\n2001\nThe contribution of phenotypic plasticity to adaptation in Lacerta vivipara.\nEvolution\n55\n392\n404\n11308095\n37. MeylanSClobertJSinervoB\n2007\nAdaptive significance of maternal induction of density-dependent phenotypes.\nOikos\n116\n650\n661\n38. BernardoJ\n1996\nThe particular maternal effect of propagule size, especially egg size: Patterns, models, quality of evidence and interpretations.\nAmerican zoologist\n36\n216\n236\n39. JordanMASnellHL\n2002\nLife history trade-offs and phenotypic plasticity in the reproduction of Galapagos lava lizards (Microlophus delanonis).\nOecologia\n130\n44\n52\n28547024\n40. FoxCWCzesakMEMousseauTARoffDA\n1999\nThe evolutionary genetics of an adaptive maternal effect: Egg size plasticity in a seed beetle.\nEvolution\n53\n552\n560\n28565419\n41. CzesakMEFoxCWWolfJB\n2006\nExperimental evolution of phenotypic plasticity: How predictive are cross-environment genetic correlations?\nAmerican Naturalist\n168\n323\n335\n42. GuttelingEWRiksenJAGBakkerJKammengaJE\n2007\nMapping phenotypic plasticity and genotype-environment interactions affecting life-history traits in Caenorhabditis elegans.\nHeredity\n98\n28\n37\n16955112\n43. SgròCMHoffmannAA\n2004\nGenetic correlations, tradeoffs and environmental variation.\nHeredity\n93\n241\n248\n15280897\n44. GebhardthenrichSGVannoordwijkAJ\n1991\nNestling growth in the Great Tit .1. Heritability estimates under different environmental conditions.\nJournal of Evolutionary Biology\n4\n341\n362\n45. HoffmannAAMerilaJ\n1999\nHeritable variation and evolution under favourable and unfavourable conditions.\nTrends in Ecology & Evolution\n14\n96\n101\n10322508\n46. SinervoBLichtP\n1991\nProximate constraints on the evolution of egg size, number, and total clutch mass in Lizards.\nScience\n252\n1300\n1302\n17842955\n47. CaleyMJSchwarzkopfLShineR\n2001\nDoes total reproductive effort evolve independently of offspring size?\nEvolution\n55\n1245\n1248\n11475060\n48. DeWittTJSihAWilsonDS\n1998\nCosts and limits of phenotypic plasticity.\nTrends in Ecology & Evolution\n13\n77\n81\n21238209\n49. GomulkiewiczRKirkpatrickM\n1992\nQuantitative genetics and the evolution of reaction norms.\nEvolution\n46\n390\n411\n28564020\n50. SchafferWM\n1974\nOptimal reproductive effort in fluctuating environments.\nAmerican Naturalist\n108\n783\n790\n51. RicklefsRE\n1998\nEvolutionary theories of aging: Confirmation of a fundamental prediction, with implications for the genetic basis and evolution of life span.\nAmerican Naturalist\n152\n24\n44\n52. RelyeaRA\n2002\nCosts of phenotypic plasticity.\nAmerican Naturalist\n159\n272\n282\n53. MerilaJLaurilaALindgrenB\n2004\nVariation in the degree and costs of adaptive phenotypic plasticity among Rana temporaria populations.\nJournal of Evolutionary Biology\n17\n1132\n1140\n15312085\n54. PigliucciM\n2005\nEvolution of phenotypic plasticity: where are we going now?\nTrends in Ecology & Evolution\n20\n481\n486\n16701424\n55. PikeNTullyTHaccouPFerriereR\n2004\nThe effect of autocorrelation in environmental variability on the persistence of populations: an experimental test.\nProceedings of the Royal Society B-Biological Sciences\n271\n2143\n2148\n56. TullyTCasseyPFerriereR\n2005\nFunctional response: rigorous estimation and sensitivity to genetic variation in prey.\nOikos\n111\n479\n487\n57. ZeraAJHarshmanLG\n2001\nThe physiology of life history trade-offs in animals.\nAnnual Review of Ecology and Systematics\n32\n95\n126\n58. RuefflerCVan DoorenTJMMetzJAJ\n2004\nAdaptive walks on changing landscapes: Levins' approach extended.\nTheoretical Population Biology\n65\n165\n178\n14766190\n59. PhelanJPArcherMABeckmanKAChippindaleAKNusbaumTJ\n2003\nBreakdown in correlations during laboratory evolution. I. Comparative analyses of Drosophila populations.\nEvolution\n57\n527\n535\n12703942\n60. YamahiraKConoverDO\n2002\nIntra- vs. interspecific latitudinal variation in growth: adaptation to temperature or seasonality?\nEcology\n83\n1252\n1262\n61. RelyeaRA\n2004\nFine-tuned phenotypes: Tadpole plasticity under 16 combinations of predators and competitors.\nEcology\n85\n172\n179\n62. KingsolverJGShlichtaJGRaglandGJMassieKR\n2006\nThermal reaction norms for caterpillar growth depend on diet.\nEvolutionary Ecology Research\n8\n703\n715\n63. StillwellRCWallinWGHitchcockLJFoxCW\n2007\nPhenotypic plasticity in a complex world: interactive effects of food and temperature on fitness components of a seed beetle.\nOecologia\n153\n309\n321\n17486371\n64. RisNAllemandRFouilletPFleuryF\n2004\nThe joint effect of temperature and host species induce complex genotype-by-environment interactions in the larval parasitoid of Drosophila, Leptopilina heterotoma (Hymenoptera : Figitidae).\nOikos\n106\n451\n456\n65. GuttelingEWDoroszukARiksenJAGProkopZReszkaJ\n2007\nEnvironmental influence on the genetic correlations between life-history traits in Caenorhabditis elegans.\nHeredity\n98\n206\n213\n17203010\n66. West-EberhardMJ\n2003\nDevelopmental plasticity and evolution\nNew York\nOxford University Press\ni\nxx\n1–794\n67. PriceTDQvarnstromAIrwinDE\n2003\nThe role of phenotypic plasticity in driving genetic evolution.\nProceedings of the Royal Society B-Biological Sciences\n270\n1433\n1440\n68. PriceTDQvarnstromAIrwinDE\n2003\nThe role of phenotypic plasticity in driving genetic evolution.\nProceedings Of The Royal Society Of London Series B-Biological Sciences\n270\n1433\n1440\n69. HueyRBHertzPESinervoB\n2003\nBehavioral drive versus behavioral inertia in evolution: a null model approach.\nAmerican Naturalist\n161\n357\n366\n70. BadyaevAV\n2005\nStress-induced variation in evolution: from behavioural plasticity to genetic assimilation.\nProceedings of the Royal Society B-Biological Sciences\n272\n877\n886\n71. WilsonAJPembertonJMPilkingtonJGColtmanDWMifsudDV\n2006\nEnvironmental coupling of selection and heritability limits evolution.\nPlos Biology\n4\n1270\n1275\n72. AgrawalAA\n2001\nPhenotypic plasticity in the interactions and evolution of species.\nScience\n294\n321\n326\n11598291\n73. MilneS\n1960\nStudies on the life histories of various species of Arthropleone Collembola.\nProceedings of the Royal Entomological Society of London\n35A\n133\n140\n74. PotapowM\n2001\nDrungerW\nIsotomidae\nGörlitz\nStaatliches Museum für Naturkunde Görlitz\n603\n75. MarshallVGKevanDKM\n1962\nPreliminary observations on the biology of Folsomia candida Willem, 1902 (Collembola: Isotomidae).\nThe Canadian Entomologist\n94\n575\n586\n76. HopkinSP\n1997\nBiology of the springtails (Insecta: Collembola)\nOxford\nOxford University Press\n330\n77. PalevodyC\n1974\nRelations chronologiques entre cycles de mue et cycles de ponte chez Folsomia candida (Collembola, Isotomidae).\nPedobiologia\n14\n196\n198\n78. OdaGACaldasILPiqueiraJRCWaterhouseJMMarquesMD\n2000\nCoupled biological oscillators in a cave insect.\nJournal of Theoretical Biology\n206\n515\n524\n11013112\n79. StamEMVan De LeemkuleMAErnstingG\n1996\nTrade-offs in the life history and energy budget of the parthenogenetic collembolan Folsomia candida (Willem).\nOecologia\n107\n283\n292\n28307256\n80. VanamelsvoortPAMUsherMB\n1989\nEgg-production related to food quality in Folsomia candida (Collembola, Isotomidae) - Effects on life-history strategies.\nPedobiologia\n33\n61\n66\n81. DraheimRLarinkO\n1995\nEffects of differently cultured fungi as a diet of Collembola.\nActa Zoologica Fennica\n0\n168\n170\n82. RasbandW\n2003\nImageJ http://rsb.info.nih.gov/ij/. 1.29 ed: National Institutes of health\n83. LessellsCMBoagPT\n1987\nUnrepeatable repeatabilities: a common mistake.\nAuk\n104\n116\n121\n84. PinheiroJCBatesDM\n2000\nMixed-Effects Models in S and S-PLUS\nNew York\nSpringer-Verlag\n528\n85. LynchMWalshB\n1997\nGenetic and analysis of quantitative traits\nSunderland, Massachusetts\nSinauer Associates\n980\n86. ManlyBFJ\n1997\nChatfieldCZidekJ\nRandomization, Bootstrap and Monte Carlo Methods in Biology\nBoca raton\nChapman & Hall\n399\n87. IhakaRGentlemanR\n1996\nR: A language for data analysis and graphics.\nJournal of Computational and Graphical Statistics\n5\n299\n314\n88. TherneauTMGrambschPM\n2000\nModeling survival data.\nDietzKGailMKrickebergKSametJTsiatisA\nExtending the Cox model\nNew York\nSpringer-Verlag\n350\n89. TherneauTMGrambschPMFlemingTR\n1990\nMartingale-based residuals for survival models.\nBiometrika\n77\n147\n160"
+}

batch_8/PMC2527684.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2527684",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2527684\nAUTHORS: Chrissa Kioussi, Michael K. Gross\n\nABSTRACT:\nBackgroundGenetic regulatory networks of sequence specific transcription factors underlie pattern formation in multicellular organisms. Deciphering and representing the mammalian networks is a central problem in development, neurobiology, and regenerative medicine. Transcriptional networks specify intermingled embryonic cell populations during pattern formation in the vertebrate neural tube. Each embryonic population gives rise to a distinct type of adult neuron. The homeodomain transcription factor Lbx1 is expressed in five such populations and loss of Lbx1 leads to distinct respecifications in each of the five populations.Methodology/Principal FindingsWe have purified normal and respecified pools of these five populations from embryos bearing one or two copies of the null Lbx1GFP allele, respectively. Microarrays were used to show that expression levels of 8% of all transcription factor genes were altered in the respecified pool. These transcription factor genes constitute 20–30% of the active nodes of the transcriptional network that governs neural tube patterning. Half of the 141 regulated nodes were located in the top 150 clusters of ultraconserved non-coding regions. Generally, Lbx1 repressed genes that have expression patterns outside of the Lbx1-expressing domain and activated genes that have expression patterns inside the Lbx1-expressing domain.Conclusions/SignificanceConstraining epistasis analysis of Lbx1 to only those cells that normally express Lbx1 allowed unprecedented sensitivity in identifying Lbx1 network interactions and allowed the interactions to be assigned to a specific set of cell populations. We call this method ANCEA, or active node constrained epistasis analysis, and think that it will be generally useful in discovering and assigning network interactions to specific populations. We discuss how ANCEA, coupled with population partitioning analysis, can greatly facilitate the systematic dissection of transcriptional networks that underlie mammalian patterning.\n\nBODY:\nIntroductionThe patterning and specification process that generates distinct neuronal cell types in the spinal cord begins as the neural tube is formed from the proliferative neuroepithelium. Signaling centers induce asymmetric expression patterns of sequence specific transcription factors (SSTFs) along the dorsal-ventral axis of the early neural tube. The expression patterns overlap and form discrete boundaries so that eleven progenitor laminae, each of which expresses a distinct combination of SSTFs, can be defined in the ventricular zone. The proliferating cells of the ventricular zone shed postmitotic cells into the marginal layer from embryonic day (E) 9.5 to E13 of mouse development. Each progenitor lamina produces at least one postmitotic cell population. that is defined by a new combinatorial code of SSTF expression. The eleven postmitotic populations that emerge are named dI1-dI6, V0-V3, and M [1]–[8].Additional mechanisms contribute to the diversification of cell types in the developing neural tube. For example, individual progenitor layers either produce different postmitotic populations at different developmental times, or postmitotic mechanisms produce different SSTF codes, and hence new populations, from single, nascent, postmitotic populations [9]–[15]. Furthermore, differential expression of Hox genes along the anterior-posterior (A–P) axis produces different neuronal populations from a given dorsal-ventral (D–V) lamina at different axial levels [16]–[18]. Although the full complement of populations is not completely characterized, it appears they can be represented by SSTF “expression codes”.At least 66 SSTFs have been invoked in the neural tube patterning process. These include 42 homeodomain, 11 basic helix-loop-helix, and 8 zinc finger SSTFs. Functional perturbations such as gene knock-outs in mice or overexpression in chick embryos have been performed for at least 47 of these SSTFs and many genetic interactions among these SSTFs have been defined. A high degree of recursive linkage between SSTFs in this system appears to exist. However, a population partitioning analysis (PPA) identified 200 additional SSTFs with the same degree of differential expression as the known set and estimated that 500–700 of the 1700 annotated SSTFs in the genome are active nodes in the genetic regulatory network (GRN) of neural tube patterning [19].Network models are developed to understand the functional organization of complex systems [20]. Specialized software allows complex and evolving datasets, of expression and epistasis information, to be accurately tracked, and aids in decoding the underlying logic of developmental GRNs [21]–[25]. GRNs contain evolutionarily inflexible subcircuits, called kernels, which consist of SSTF nodes with highly recursive linkages and which specify spatial domains in which a body part will form [26]. The SSTF expression codes that are used to spatially define transient neural tube populations are transiently stable in spatial domains during development. Thus, the expressed SSTFs that define a population are predicted to be nodes of a specific network kernel. Transitions between SSTF expression codes, such as those that occur between progenitor laminae and the emergent postmitotic populations, therefore represent transitions between kernels. Removal of one SSTF that participates in a kernel destroys the linkages that stabilize the kernel, and has a catastrophic effect on the development of the respective body part. In line with this model, knocking out SSTFs that contribute to the SSTF codes in the neural tube frequently results in ablation of the respective body part. For example, removing Isl1 results in the loss of motor neurons [27] and removing Lbx1 results in the loss of the substantia gelatinosa [9].Active nodes, in a transcriptional network model of a biological system, represent SSTFs that are differentially expressed in the system. The large number of active SSTF nodes in neural tube patterning suggests that either there are far more kernels than those currently described, or the number of nodes in each kernel is larger than those currently reported in other systems (5–10 SSTFs). We have previously reported how PPA provides a high throughput method to systematically identify active SSTF nodes in a given developmental system [19]. In principle, PPA can be reiterated in progressively constrained sub-systems until no single SSTF is differentially expressed in the sub-system. Such a sub-system is expected to represent a “body part”, specified cell type, network kernel, or stable regulatory state, and can be described by a unique SSTF code.Active nodes should ultimately be connected by regulatory interactions that reflect the direct interaction of SSTFs with cis-regulatory modules in each stable regulatory state. Developmental cis-regulatory modules in the sea urchin system generally have four to eight diverse inputs [28]. The tools needed to demonstrate direct interactions are currently not amenable to the rapidly changing network states of embryonic mammals and require a priori knowledge of all the cis- and trans-acting components of an interaction. The use of null knock-in alleles allows investigators to compare cells that express a SSTF, in heterozygotes, to the equivalent cells, in mutants, that “should have” expressed the SSTF. Such analyses establish epistatic, rather than direct, interactions between the active nodes in the network model.Epistatic interactions are useful in organizing and constraining draft network models, which, in turn, can be used to generate specific testable hypotheses about which components are involved in specific, direct regulatory interactions. The rate of discovery of epistatic interactions is currently limited by availability of in situ probes and antibodies, as well as by manpower, and will therefore be accelerated by applying high-throughput genomic tools. In this report, we describe how the same genetic tools that are developed for PPA can be employed to define epistatic interactions between active SSTF nodes in a high throughput manner. Acquisition of epistasis data from population pools that are defined by active node expression in embryos assures that only physiologically relevant interactions are recorded into network data sets. In tandem, PPA and active node constrained epistasis analysis (ANCEA) provide a systematic approach to characterize the many kernels created by the patterning mechanism to produce a mammalian body.ResultsFlow-Sorting Population Pools by Active Node ConstraintsThe Lbx1\nGFP mouse line [29] provides a robust system for developing genome-wide analyses of epistatic interactions in mammalian embryos. First, the fluorescent cells from E12.5 embryos are abundant and are predominantly from two closely related populations (Fig. 1A). Approximately 80% of Lbx1 expressing neurons in E12.5 neural tubes belong to the two late populations. Second, these two populations have expressed Lbx1, or GFP, for less than 24 hours. Thus, comparisons between mutant and heterozygous sources preferentially reveal immediate molecular consequences of Lbx1 expression, rather than delayed secondary effects. Third, loss of Lbx1 function leads to known changes in SSTF expression that provide positive controls for the analysis [9], [30].10.1371/journal.pone.0002179.g001Figure 1Flow Sorting of Dissociated E12.5 Neural Tubes.(A) Locations of the eight dorsal neural tube populations at E12.5 (right side). The dI1-dI3 populations do not express Lbx1, are born in the most dorsal ventricular zone, and rapidly migrate toward the floor plate (yellow, tangerine, and orange traces). Three small early Lbx1+populations, dI4, dI5, and dI6, are born in distinct layers from the middle-dorsal ventricular zone between E10.5 and early E11.5 and move to the regions outlined by magenta and cyan traces by E12.5. Two large late Lbx1+populations, dI4LA and dI4LB, are born intermingled from the dorsal half of the ventricular zone between late E11.5 and E13 occupy the areas hatched in cyan and magenta. (B) Flow cytometry profiles from neural tubes of heterozygotes (+/−) or mutants (−/−) prior to (top panels) and after (lower panels) sorting. Green or white cell pools are labeled with the four array conditions (hG, hW, mG, mW) they give rise to. (C) Cross sections of heterozygous and mutant neural tubes at the forelimb level at E12.5 stained with GFP antibody.Fluorescence activated cell sorting (FACS) was used to purify the GFP+ (green) and GFP− (white) cells from pools of ten neural tubes of mutant and heterozygote Lbx1GFP embryos at E12.5 (Fig. 1B). The one hour, serum-free procedure was repeated on eight separate occasions. Four runs showed a high level of reproducibility at the level of timing (Table S1) and RNA quality (data not shown). Green cells constituted 41±6% of the sorted events in both mutants and heterozygotes, supporting the idea that cells are re-specified but not yet apoptotic at this stage. Terminal Transferase dUTP Nick End Labeling (TUNEL) assays have shown that apoptosis in mutants occurs at E13.5 and E14.5 [9]. White cells constituted 53±6% and 52±6% of the sorted events in mutants and heterozygotes, respectively. The ratio of green to white cells accurately reflected the GFP expression observed by histology (Fig. 1C).Total RNA from three biological replicates of each of the four conditions, heterozygous green (hG), heterozygous white (hW), mutant green (mG) and mutant white (mW), was used to probe Affymetrix Mouse 430 arrays. Data from all twelve arrays were normalized using GC robust multiarray averaging in Genespring software. The analysis focused on probe sets corresponding to SSTFs , which collectively form the key interface between the genetic regulatory information on the DNA and the RNA polymerase II transcriptional machinery and its coregulators [31] and which make up transcriptional network kernels [26]. The annotated SSTF set used in this analysis includes 177 basic, 749 zinc-coordinated, 512 helix-turn-helix, 116 ß-scaffold with minor groove contacts, and 138 other SSTFs and is an updated version of the set described earlier [19]. These 1691 genes were collectively monitored by 3574 probe sets.Lbx1 Regulates SSTF Genes in a Cell Autonomous MannerRegion, field, compartment, or cell-type specific selector genes of Drosophila are SSTFs. They function cell autonomously in the morphologically distinct, spatial domains of the embryo where they are expressed. Mammalian homologs of these genes are also expressed in spatially constrained ways in embryos. However, the more fluid anatomy of developing mammals generally limits our ability to delineate regions, fields, and compartments on the basis of morphology alone. Consequently, it is usually not possible to determine to what extent the putative mammalian selector genes are functioning cell autonomously.The average signal intensities of triplicate heterozygous and mutant arrays were compared in scatter plots to reveal Lbx1-dependent changes in the expression of SSTF genes (Fig. 2). Heterozygous versus mutant comparisons gave markedly different results in green cells (Fig. 2A), which represent populations that normally express Lbx1, and white cells (Fig. 2B), which represent populations that normally do not express Lbx1. The scatter from the diagonal unity line in the green cell comparison indicates that Lbx1 regulated many SSTF genes in populations where it is expressed. In contrast, little scatter from the unity line was observed in the white cell comparison, indicating that few SSTF genes are regulated in cells where Lbx1 is not expressed. These results demonstrated that SSTF gene regulation by Lbx1 in green cell populations have little effect on SSTF gene expression in adjoining white cell populations. Thus, SSTF regulation by Lbx1 appears largely, or entirely, cell autonomous, consistent with the idea that Lbx1 functions as a selector gene in a field or compartment defined by its own expression, rather than by morphological boundaries. Any non-cell autonomous regulation is either subtle or occurs between green cell populations. The absence of significant cross-talk between cell populations is consistent with the idea that neural tube pattern formation, during this phase, consists of a series of cell autonomous transitions between stable transcriptional network states.10.1371/journal.pone.0002179.g002Figure 2Identification of Lbx1-dependent SSTF Targets.Comparisons were between heterozygotes (x-axis) and mutants (y-axis) in green (left panel) and white (right panel) cell pools. The average intensity values from three independent arrays of three independent isolates are shown. Values for 3574 probe sets corresponding to 1691 SSTF genes are plotted. All 12 arrays were normalized using GCRMA. Color coding, given to each probe set based on their behavior in the green comparison, was maintained in the white comparison. Colored dots represent probe sets that change in a uniform direction in all of the 9 possible 2-way comparisons between the three mutant and three heterozygous arrays of green cells. Probe sets are color coded to show those that passed the 1.3 fold threshold in 35–84 (red), 1–34 (cyan), or zero (green ) permutations of three pair wise, logical AND, comparisons in green cells. Red probe sets were selected to create the tables. Note that red probe sets change in the green comparison but not in the white comparison. The positive controls Lbx1, Lmx1b, Isl1, and Foxd3 are indicated. No probe sets exists for Pax2. However, both Pax8 and Pax5 are regulated in the direction predicted for Pax2.Only the probe sets for Lbx1 and its corepressor, Lbxcor1, showed significant signal reductions in the white cell comparison, albeit from much lower basal signals than in the green cell comparison. The loss of residual signal from Lbx1 probe sets is expected in null mutants. The concomitant reduction of residual Lbxcor1 RNA likely reflects a particularly high affinity interaction of Lbx1 with a cis-regulatory module of this gene.Lbx1 Regulates at Least 6% of SSTF Probe SetsThe large number of SSTF genes that were regulated by Lbx1 in green cells was surprising in light of the relatively simple genetic diagrams that currently describe neural tube development and the lack of non-cell autonomous regulation. The low level of scatter in the white cell comparison indicated that the noise between biological replicates was very low (Fig. 2B). However, it remained possible that there was simply more variability in gene expression in green samples. The nonparametric permutation fold-scanning method [19] was applied to measure the false discovery rate (FDR) at different fold-cutoffs in green (Fig. 3A) or white (Fig. 3B) cell pools. The details of this method are described under Materials and Methods.10.1371/journal.pone.0002179.g003Figure 3Limiting the Total Target Number by FDR.A) Array data from GFP expressing cells of neural tubes (Green) was compared. These are cells that normally express Lbx1. B) Array data from neural tube cells lacking GFP (White) was compared. These are cells that normally lack Lbx1. In each panel, the results from individual arrays was compared, in a pair-wise manner, between replicate arrays of the same condition (internal), and between arrays of mutant and heterozygote conditions (cross). Triplicate array measurements allowed three pair wise internal-comparisons to be made for each of the four conditions (hG, mG, hW, mW) and nine pair wise cross-comparisons to be made between mutant and heterozygote conditions. The number of probe sets with fold changes at or above a given fold cutoff in three specific internal-comparisons was measured and represents the number of false positives because the comparison was between biological replicates (open squares). The numbers of probe sets with fold changes at or above a given fold cutoff for three specific cross-comparisons between heterozygous and mutant arrays was measured in a comparable manner. Each of the 84 possible permutations of three specific cross-comparisons (of the nine available cross-comparisons) was evaluated at each fold cutoff. The average value obtained from all 84 permutations (filled squares) was plotted in each panel and represents the total number of positives, true plus false. By this method, the number of probe sets above a given fold cutoff was determined in an equivalent manner in cross- and internal-comparisons. The FDR (blue circles) was calculated by dividing the false positives by the total positives. The number of true positives (triangles) was calculated at each fold cutoff by subtracting the false positives from the total positives (see Materials and Methods for a more detailed description of the analysis) .Internal comparisons between triplicate arrays reveal false discoveries arising from combined noise, biological and array processing, of the measurements. The number of false discoveries at each cutoff was very similar in each of the four conditions, as would be expected for noise. Cross comparisons between heterozygous and mutant showed more changes than internal comparisons at each cutoff in both green and white cells. However, in green cells the difference between cross and internal comparisons was far greater than in white cells at all cutoffs and at all FDRs. For example, the FDR reached 5% at the 1.3 fold cutoff in the green comparison. At this cutoff, 180 of the 3108 probe sets (6%) were true positives. In striking contrast, the FDR reached 5% at the 2.1 fold cutoff in the white comparison. At this cutoff, only the Lbx1 and Lbxcor1 probe sets were true positives. At the 1.3 fold cutoff in white cells, the FDR was 50% and only 20 SSTFs are true positives. Taken together, these analyses provide a quantitative basis for the assertion that Lbx1 regulates 6% of SSTF probe sets in Lbx1 expressing populations. The large number of SSTF probe sets with altered expression cannot be dismissed as measurement noise.Selecting Interactions for the Network ModelLists of SSTF genes that are targets of Lbx1 in green cells (Tables 1, 2) were established by applying similar algorithms as in the permutation fold-scanning analyses above. The database was first queried for probe sets that change in a uniform direction in all nine cross comparisons (Fig. 2, colored dots). This condition was satisfied in 697 of the 3574 probe sets. These 697 probe sets were queried for those that pass the 1.3 fold cutoff in all three cross comparisons of at least one permutation of three cross comparisons (Fig. 2, cyan and red dots). This condition was satisfied by 426 probe sets. These probe sets were ranked by the number of permutations that satisfy the 1.3 fold cutoff. The 1.3 fold cutoff was satisfied for all 84 permutations for 145 probe sets, for 43–83 permutations for 24 probe sets, and for 1–42 permutations in 257 probe sets. The list was cut at the permutation break nearest the expected number of true positives at a 5% FDR (i.e. 180 probe sets). Thus, only the 203 probe sets with ≥35 permutations passing the 1.3 fold cutoff were selected (Fig. 2, red dots). These constitute 6% of all probe sets and corresponded to 8% of known SSTF genes. Lbx1 repressed 70 and activated 71 SSTF genes (Tables 1, 2). Only 5% are expected to be incorrectly identified interactions. Furthermore, 65 were regulated at least 2 fold and 33 were regulated at least 3 fold. Thus, Lbx1 causes widespread, and large, perturbations in the transcriptional network of developing dorsal horn neurons in a relatively short time after it begins to be expressed.10.1371/journal.pone.0002179.t001Table 1SSTF Genes (1691 genes; 3574 probe sets) Repressed by Lbx1 in the Neural Tube (4%)Classa\n#b\nSSTFc\nNT d\nUCRe\nHeterof (n = 3)Mutantf (n = 3)Fold Δg\n(Highest)Average Signal of PS(Lowest)\n1. Basic (177 genes; 362 probe sets) 9% repressed\nbHLH\n1\n\nOlig3\nd,vz137\n56±20\n\n2207±274\n\n39.1\n2Bhlhb4v\n9±2\n\n129±26\n\n14.2\n\n3\n\nNeurog2\nd,vz\n931±178\n\n4346±156\n\n4.7\n\n4\nBhlhb5dc, vz33/+\n729±50\n\n2147±245\n\n2.9\n5Nhlh1dc, vz\n1288±149\n\n2927±600\n\n2.3\n\n6\nNpas3d, vz125/+\n290±18\n\n678±41\n\n2.3\n\n2.4\n\n2.1\n1.37Neurod1d\n895±79\n\n1752±211\n\n2.0\n1.6*\n8\nNeurod2pm121\n336±128\n\n497±50\n1.5*\n1.6*\n\n9\nNeurod4pm\n351±63\n\n1541±477\n\n4.4\n\n5.0\n1.2*†10Tcf15vz (?)\n140±27\n\n215±13\n\n1.5\n\n11\nEbf2pm\n413±87\n\n617±23\n1.1*1.2*\n1.5\n\n12\n\nEbf1\npm77/+\n1804±\n79\n\n2328±\n189\n1.31.7*\n1.6\n1.6*\n13\nPtf1avz,pm130\n69±41\n\n133±30\n2.0*\n14\n\nAscl1\nd, vz\n491±65\n\n654±58\n1.3\n1.6*\nbHLH-ZIP\n15\nMycns\n101±14\n\n150±3\n\n1.5\nbHSH\n16\nTcfap2adc,pm30/+\n4175±172\n\n6205±890\n\n1.5\n1.51.1*\n2. Zinc-coordinated (749 genes; 1627 probe sets) 2% repressed\nC4\n17\nNr4a2pm16/+\n526±139\n\n7015±615\n\n13.3\n\n29.0\n\n16.6\n\n3.4\n\n18\nEsrrgpm53\n284±38\n\n1767±274\n\n6.2\n1.0*†C2H2\n19\nZfp503+5/+\n1784±236\n\n4523±698\n\n2.5\n\n3.0\n20Prdm8nd44\n574±139\n\n1258±48\n\n2.2\n21Bnc2nd25/+\n55±17\n\n103±14\n\n1.9\n\n1.9\n1.1*†1.0*†\n22\nSall1d,vz,pm+\n711±70\n\n1324±68\n\n1.9\n\n2.4\n\n23\nRepin1ns\n17±7\n\n30±6\n1.2*1.1*\n1.7*\n1.2*†24Zbtb20d112/+\n1037±49\n\n1498±164\n\n1.4\n1.5*\n1.7\n1.5*1.4*\n1.6\n1.41.7*25BC035954nd+\n141±25\n\n222±35\n\n1.6\n26Klf7++\n596±81\n\n958±93\n\n1.6\n\n1.5\n\n2.0\n\n1.8\n\n27\nZfp703nd52\n379±\n53\n\n638±\n16\n\n1.7\n1.0*†\n28\nZfp319ns\n279±29\n\n411±49\n1.41.2*\n1.5\n1.4*29Zfp787ns\n19±\n2\n\n27±\n4\n1.3*1.2*\n1.4†\n1.0*†1.1*†1.0*†\n30\nBtbd4nd\n342±26\n\n497±34\n\n1.5\n1.2*1.1*1.4*†\n31\nGlis2vz\n239±10\n\n326±20\n\n1.4\n1.1*CCHC32Peg10nd\n14±6\n\n35±9\n\n2.4†\nDHHC\n33\nZdhhc2ns\n319±30\n\n565±73\n\n1.8\n\n1.9\n\n2.2\n\n2.3\n\n3. Helix-turn-helix (512 genes; 1026 probe sets) 7% repressed\nHD\n34\nHmx3dc29\n9±1\n\n460±96\n\n53.3\n\n35\nHmx2dc29\n10±4\n\n425±20\n\n41.2\n\n36\n\nIsl1\nv, dc+\n233±141\n\n6128±941\n\n26.3\n\n26.7\n\n37\nOtpv, dc26\n244±65\n\n6059±182\n\n24.9\n1.1*†\n38\nPhox2bdc64/+\n467±350\n\n4194±324\n\n9.0\n\n3.0\n\n39\n\nLhx9\ndc+\n21±14\n\n54±13\n\n2.6\n1.1*1.3*†\n40\nDlx1ns\n34±6\n\n53±11\n\n1.6*†\n41Pou4f2d,pm48/+\n582±23\n\n3914±246\n\n6.7\n\n42\nPou3f1dc59\n339±105\n\n1728±160\n\n5.1\n\n5.0\n\n43\n\nPou4f1\nd,pm79\n5556±492\n\n10500±887\n\n1.9\n\n2.0\n\n44\nHoxa7pm56\n540±77\n\n921±17\n\n1.7\n45Hoxc13v, pm69\n8±1\n\n14±5\n\n1.8*†\n46Hoxc10v, pm69\n284±96\n\n825±64\n\n2.9\n47Hoxc9v, pm69\n629±121\n\n1157±54\n\n1.8\n\n48\n\nHoxc8\nv, pm69\n2607±346\n\n4088±305\n\n1.6\n\n49\nHoxc6v, pm69\n512±41\n\n850±23\n\n1.7\n\n1.6\n1.3\n50\n\nHoxd10\nv, pm15\n275±118\n\n618±121\n\n2.2\n\n51\n\nHoxd9\nv, pm15\n492±167\n\n1157±110\n\n2.4\n52Hoxd8+15\n588±147\n\n1153±226\n\n2.0\n\n1.8\n53Hoxd1d (?)15\n22±5\n\n41±5\n\n1.9†\n\n54\n\nPax3\nd, vz118\n366±77\n\n972±190\n\n2.7\n\n2.6\n\n3.0*\n\n55\n\nPax7\nd, vz128\n29±\n2\n\n63±14\n\n2.2\n1.8*\n56\n\nGsh2\nd, vz\n49±8\n\n104±15\n\n2.1\n\n57\n\nGsh1\nd, vz72\n33±7\n\n61±16\n\n1.9*\n\n58\nTgif2+\n293±35\n\n391±36\n1.3\n1.5\n\n59\nPbx3pm7\n8701±256\n\n12106±1173\n1.4\n1.5\n\n60\nShox2pm, dc42\n143±71\n\n634±451\n4.4*2.4*\n61\n\nBarHl1\npm,dc120\n27±1\n\n69±39\n2.1*2.6*FH\n62\n\nFoxd3\ndc,v17\n15±2\n\n715±33\n\n46.9\n\n63\nFoxp2vc9/+\n397±121\n\n2898±273\n\n7.3\n\n8.3\n\n8.0\n\n8.5\n\n3.9\n1.2*†\n64\nFoxp1m,vc,vz12/+\n349±41\n\n1626±167\n\n4.7\n\n5.3\n\n4.7\n\n4.8\n\n5.2\n\n4.5\n\n4.2\n65Foxp4pm128\n77±23\n\n324±77\n\n4.2\n\n5.2\n1.2*†TC\n66\nNcor2v\n250±27\n\n487±100\n\n1.9\n\n1.8\n\n67\nDll3vz\n1318±198\n\n2253±36\n\n1.7\n68Aste1nd\n30±\n5\n\n42±\n8\n\n1.4*\n\n4. ß-scaffold (116 genes; 281 probe sets) 0.9% repressed\nHMG\n69\nSox1vz\n55±5\n\n100±25\n\n1.8\n1.0*†\n5. Other (138 genes; 279 probe sets) 0.7% repressed\n\n70\nDach2pm+\n86±21\n\n196±50\n\n2.3\n\n1.8*†\naCategories and classes according to TRANSFAC (Braunschweig, Germany)bBold indicates known nodes or predicted active nodes that behave like known nodes (Kioussi et al., 2006). Underline indicates other predicted active nodes with greater than 1.3 fold partitioning.cNames according to Mouse Genome Informatics (MGI). Bold indicates SSTFs of the known set (Kioussi et al., 2006).dEstimated regional expression in the developing neural tube. Estimates are based on expression data linked to the SSTF's MGI website. Expression was observed somewhere between E9.5 and E13.5. Accurate RNA in situ data at E12.5 and double-labeling immunohistochemical data are often not available. Codesare as follows: “+”, region specific expression observed; vz, ventricular zone; pm, postmitotic layer or mantle zone; svz, subventricular zone between vz and pm; d, dorsal; v, ventral; dc, dorsalcommissural; da; dorsal association; ns, not seen; nd, no data available.eSandelin et al., 2004 lists the 150 largest clusters of ultra conserved regions (non-coding) in the entire mammalian genome (1 is the largest ,150 is the smallest on their list). The number shown indicates the position of this SSTF on their list. “+” indicates that the UCR cluster is enriched in Sox, Pou and homeodomain binding sites (Bailey et al., 2006).fValues are the average and standard deviation from three microarray values. Data shown is from the probe set with the highest average signal, in all 12 arrays, of those that passed the t-test, or, if none passed t-test, of all.gProbe sets that passed the 1.3 fold threshold 35 to 84 possible permutations of three pairwise, logical AND comparisons (bold) were selected to establish the initial SSTF gene list. All other probe sets corresponding to these genes were identified in the UCSC genome browser and were added to the table (plain text). Multiple probe sets for a given gene were ranked by their average signal in all 12 arrays an their fold change listed. Asterisk (*) indicates that a probe set failed the t-test at the 95% confidence interval. Dagger (†) indicates that the average signal intensity of the probe set was below 30 (or <0.1% of maximum).10.1371/journal.pone.0002179.t002Table 2SSTF Genes (1691 genes; 3574 probe sets) Activated by Lbx1 in the Neural Tube (4.1%)Classa\n#b\nSSTFc\nNTd\nUCRe\nHeterof (n = 3)Mutantf (n = 3)Fold Δg\n(Highest)Average Signal of PS(Lowest)\n1. Basic (177 genes; 362 probe set) 5% activated\nbZIP\n1\nMafa+\n487±\n115\n\n48±\n12\n\n−10.2\n\n2\nMafbda\n271±\n33\n\n64±\n8\n\n−4.2\n\n−2.6\n\n3\nJun+\n255±\n52\n\n129±\n13\n\n−2.0\n\n−2.1\n\n4\nJundm2ns75\n442±\n12\n\n313±\n23\n\n−1.4\nbHLH\n5\n\nNeurog3\nvz\n25±4\n\n12±5\n\n−2.1\n−1.1*†\n\n6\nId4d+\n7606±306\n\n3816±930\n\n−2.0\n\n−1.9\n\n−1.9\n\n−2.5\n\n7\nId3d\n324±42\n\n191±58\n\n−1.7\n\n8\nHes5vz\n369±62\n\n289±47\n−1.3*\n−1.5*\n\n2. Zinc-coordinated (749 genes; 1627 probe sets) 3% activated\nC4\n9\nNr2f2+81/+\n833±23\n\n412±66\n\n−2.0\n\n−2.4\n\n−2.0\n\n−2.4\n\n10\nTrps1d, vz+\n158±32\n\n113±19\n−1.4*−1.2*−1.2*\n−1.8\n11Sall3da,vz28\n777±184\n\n77±16\n\n−10.1\n\n−4.9\n\n12\nSall4+76\n95±28\n\n26±8\n\n−3.7\n−1.5*†\n13\n\nZic5\nda63\n477±113\n\n63±12\n\n−7.6\n\n−4.5\n\n−1.8\n\n14\n\nZic4\nda62/+\n414±157\n\n105±18\n\n−3.9\n\n−3.4\n\n15\nZic3da95\n1625±212\n\n601±34\n\n−2.7\n\n−3.2\n\n16\n\nZic2\nda63/+\n327±71\n\n110±7\n\n−3.0\n\n17\n\nZic1\nda62\n14774±804\n\n6851±417\n\n−2.2\n\n−3.8\n\n18\nWt1+\n35±18\n\n10±3\n\n−3.5*†\n\n19\nZfp804and\n584±72\n\n195±19\n\n−3.0\n−1.6*†\n20\nKlf5ns\n33±6\n\n12±2\n\n−2.9\n\n−1.8†\n−1.5*†\n21\nZfpm2pm21/+\n199±54\n\n97±16\n\n−2.1\n−1.4\n22\nIkzf4+\n197±\n12\n\n142±\n16\n−1.4\n−2.2\n−1.1*†\n23\nBcl11apm13/+\n5272±514\n\n2881±141\n\n−1.8\n\n−1.9\n\n−2.1\n\n24\nZcchc11nd\n1899±34\n\n1838±86\n−1.0*−1.2*\n−1.8*\n−1.1*\n25\nHivep2vz\n153±20\n\n95±5\n\n−1.6\n\n−1.8\n\n26\nRestda\n187±9\n\n108±13\n\n−1.7\n−1.4*−1.1*†−1.0*†27Zfp704nd\n361±52\n\n226±15\n\n−1.6\n−1.2*†28Zfp467ns\n227±21\n\n138±9\n\n−1.6\n\n−1.7\n−1.3*\n29\nZfp775nd\n208±29\n\n142±11\n\n−1.5\n\n30\nB930008K04Riknd\n62±5\n\n43±6\n\n−1.4\n\n31\nZfp784nd\n54±5\n\n38±2\n\n−1.4\nCCHC32Zcchc12pm\n337±71\n\n151±36\n\n−2.2\nCXXC\n33\nCxxc5nd\n3114±127\n\n1975±7\n\n−1.6\n\n−1.4\n\n34\nCxxc4nd\n1398±118\n\n1105±129\n−1.3\n��1.4\n1.1*†\n3. Helix-turn-helix (512 genes; 1026 probe sets) 5% activated\nHD\n35\nGbx1da\n1422±\n288\n\n67±\n29\n\n−21.2\n\n36\nGbx2da86\n3027±237\n\n1216±169\n\n−2.5\n\n37\n\nLbx1\nda\n3374±137\n\n331±7\n\n−10.2\n\n−7.6\n\n38\nTshz2d,pm76\n1372±238\n\n209±24\n\n−6.6\n\n−8.1\n\n−5.3\n\n39\n\nLhx2\nd,pm104\n240±179\n\n34±3\n\n−7.0*\n\n40\n\nLhx1\npm50/+\n5763±271\n\n2575±1214\n\n−2.2\n\n−2.4\n\n41\nPknox2d, pm\n1865±198\n\n437±43\n\n−4.3\n\n−2.5\n\n42\nSatb2++\n216±28\n\n70±12\n\n−3.1\n−1.0*\n43\n\nLmx1b\nda+\n601±\n135\n\n286±\n50\n\n−2.1\n−1.1*†\n44\n\nNkx6-1\nvz,svz31\n81±27\n\n47±12\n\n−1.7*\n\n45\nHoxa4pm56\n149±15\n\n99±4\n\n−1.7*\n\n−1.5\n\n46\nHoxb2d,pm82\n2517±160\n\n1526±130\n\n−1.6\n\n47\nHoxb8d,pm82\n4850±249\n\n3297±476\n\n−1.5\n\n48\n\nMsx1\nd,vz\n77±14\n\n46±7\n\n−1.7\n1.1*\n49\n\nMsx2\nd,vz\n14±1\n\n10±2\n\n−1.5\n\n50\nZfhx1bd,svz2/+\n853±100\n\n519±90\n\n−1.6\n\n−1.8\n−1.7*−1.2*†\n51\n\nPax6\nvz,da83/+\n72±11\n\n51±5\n\n−1.4\n1.0*−1.1*\n52\n\nPrrxl1\nda101/+PD\n53\nPax8pm\n861±224\n\n84±24\n\n−10.2\n\n54\nPax5+90\n350±45\n\n209±\n24\n\n−1.7\n1.2*†1.0*†FH\n55\nE2F8nd\n53±11\n\n33±7\n\n−1.6\n1.0*†TC56Ets2d\n346±90\n\n140±22\n\n−2.5\n\n57\nElk3+\n103±29\n\n48±14\n\n−2.1\n1.1*−1.3*\n58\nEtv5d\n46±5\n\n23±4\n\n−2.0\n1.4*1.0*†WH\n59\nDepdc1and\n31±3\n\n16±3\n\n−2.0\n\n60\nCdc6nd\n36±9\n\n20±4\n\n−1.8\n\n61\nRfx4d\n524±50\n\n323±86\n\n−1.6\n1.1*†\n62\nMyst4d\n1420±39\n\n992±80\n\n−1.4\n\n−1.1*†\n\n63\nRfxdc2d\n2433±54\n\n2177±121\n−1.1−1.1*†\n−1.6*†1.0*†\n4. ß-scaffold (116 genes; 281 probe sets) 2% activated\nSTAT64Stat1ns\n76±6\n\n53±10\n\n−1.4\n−1.5*−1.3*HMG65Sox13pm\n127±28\n\n66±7\n\n−1.9\n\n5. Other (138 genes; 279 probe sets) 4% activated\n\n66\nDmrt3pm148\n63±31\n\n12±1\n\n−5.5\n\n67\nLbxcor1pm+\n25659+\n1614\n\n6284±\n653\n\n−4.1\n\n68\nDmrtb1nd\n357±63\n\n169±11\n\n−2.1\n\n69\nNotch3vz\n282±82\n\n143±23\n\n−2.0\n−1.3*\n70\nNotch2vz\n266±32\n\n158±33\n\n−1.7\n1.0*†\n71\nObfc2and\n186±41\n\n104±15\n\n−1.8\n−1.3*−1.5*−1.3*†1.0*†1.0*†a–gAs in Table 1.With few exceptions (asterisks), the differences observed in these probe sets were significant at the 95% confidence interval in t-tests. For comparison, 18% of probe sets that showed an average fold change greater than 1.3 and passed the t-test at the 95% confidence interval were not selected by the permutation analysis. In contrast , only 8% of the 203 selected probe sets failed the t-test at the 95% confidence interval (Tables 1, 2; asterisks). Thus, selection by permutation analysis is more stringent than selection by average fold change and t-test. Both methods produce similar lists of targets.Microarray analysis provided a detailed snapshot of the flux in the transcriptional network when Lbx1 is removed from the system. However, the permutation analysis allows the resolution of the snapshot to be understood intuitively by quantitatively linking the number of selected targets to a FDR. The resolution of the snapshot is limited by the FDR that one considers acceptable. A lower FDR translates to a higher effective cutoff and produces a shorter target list. Nevertheless, demanding excessively low FDRs is counterproductive because it eliminates many true interactions and subtle influences on the network that produce reasonable constraints on an evolving network model. The ability of a given FDR to produce a low cutoff is limited by the ability to reproduce data in biological replicates. For example, the green and cyan dots are closer to the unity line in the white cell comparison (Fig. 2B) than in the green cell comparison (Fig. 2A). These dots represent probe sets whose signals changes in a uniform direction in all single array comparisons, but which fail the 1.3 fold comparison in ≥49 permutations. The genes corresponding to these probe sets were not included in the tables, but potentially have Lbx1-dependent expression. One may expect that a larger number of replicate arrays would decrease the cutoff corresponding to a 5% FDR and would allow almost all probe sets to be identified as targets. However, probe sets need to change in a uniform direction in all single array comparisons to be considered in fold scanning analyses. Larger numbers of replicate arrays increase the stringency of this criterion and more probe sets, whose changes are not reproducible or are due to stochastic noise, are thereby eliminated. Permutation analyses allow the quality of different epistasis data sets obtained by microarray analyses to be quantified so that their use in assembling network models can be appropriately weighted.Repressed Target GenesThree target genes, Foxd3, Isl1, and Pou4f1, are known to be repressed by Lbx1 [9], [30] and appear in Table 1. Foxd3 and Isl1 are normally expressed in postmitotic populations that do not express Lbx1. They are normally not expressed in any of the Lbx1 expressing populations. In Lbx1 mutants, Foxd3 is specifically upregulated in the dI4LA and I4 populations, and Isl1 is specifically upregulated in the I4LB and I5 populations, respectively The upregulation of each of these genes in only half of the green cells resulted in some of the strongest effects observed. The effects were so strong because none of the green populations initially expressed these genes. Similarly strong effects were observed for the Hmx2, Hmx3, and Otp homeobox genes that have not yet been implicated in the neural tube GRN. The RNA in situ patterns of these genes show that their expression is restricted to regions corresponding to the most dorsal postmitotic neural tube populations [32], [33]. Taken together, the results suggest that these three new genes: (1) are expressed in few, if any, of the Lbx1 expressing populations, (2) are strong candidates to play a role in establishing the kernels that specify dorsal cell types, and (3) must be shut down by Lbx1 in the same way Foxd3 and Isl1 are shut down to create the network kernels that specify the dI4-dI6, dI4LA and dI4LB cell types.The data for the known Lbx1 target gene Pou4f1 (Brn3a) illustrates how striking effects observed in specific populations by immunohistochemistry become moderate effects in microarray analyses because of the pooling of populations. Pou4f1 is normally expressed in dI5 and dI4LB, but not in dI4, dI6, or dI4LA\n[2]. Pou4f1 is specifically upregulated in the dI4 and dI4LA populations of mutants [30]. The early populations each contribute approximately 7% of the green cells whereas the late populations each contribute approximately 40% of green cells. The heterozygous green cells therefore contain 47% cells that express Pou4f1 and 53% that do not. In mutants, only dI6 does not express Brn3a. Thus, 93% of green cells express Pou4f1. The moderate 2-fold increase observed in the microarrays of mutant green populations is consistent with these observations. It becomes clear that the smaller changes observed for many SSTFs may be due to large changes in specific populations that are ameliorated by the pooling of populations in the analysis.A large number of SSTF genes that have known functions in the establishment of the dorsal progenitor domains show higher expression levels in Lbx1 mutants (Table 2). These include Olig3, Neurog2(Ngn2), Ascl1(Mash1), Gsh1, Gsh2, Math1, Pax3, Pax7, and Ptf1a. These genes are generally expressed in the dorsal ventricular zone and are not co-expressed with Lbx1, which appears shortly after the last cell division [9].Notably absent from the list of repressed targets are SSTFs that are known to be expressed specifically in nascent postmitotic ventral interneurons (Evx1, Evx2, En1, Chx10, Sim1, Sox14, Etv1, Etv4, Gata2, Gata3), motor neurons (Isl2, Lhx3, Lhx4, Hlxb9 (Hb9, MNR2)), or the ventral progenitor laminae (Dbx1, Dbx2, Nkx6.2, Nkx6.1, Irx3, Olig1, Olig2, Nkx2.2, Nkx2.9, Pax6, Gli1, Gli2, Gli3). Comparisons between mutant and heterozygote embryos with Evx1, En1, and Chx10 antibodies were used to confirm this observation (data not shown).Taken together, our data suggest that Lbx1 serves two general roles in downregulating SSTFs in the network. First, it represses SSTF genes expressed in the progenitor pools and thereby destroys network states that confer progenitor cell characteristics. Second, it represses SSTF genes expressed in dorsal commissural interneurons and thereby prevents the establishment of network states that confer properties of relay interneurons. Lbx1 appears to have little influence on SSTF genes that are expressed in network states that define ventral cell types.Activated Target GenesFive target genes (Pax2, Lmx1b, Lhx1/5, Satb2, Gbx1) are known to be activated by Lbx1 in the developing neural tube [9], [30], [34], [35]. Table 2 shows that Gbx1, Lmx1b, Satb2 and Lhx1 were expressed at significantly lower levels in mutant green cells. The probe set for Lhx5 produced a robust, but Lbx1-independent, signal. There is no probe set for Pax2 on the array. However, real time PCR analyses showed that Pax2 RNA declines three fold in mutant green cells (data not shown). Moreover, the two paralogs of Pax2, namely Pax8 and Pax5, were expressed at 10.2 and 1.7 fold lower levels in mutant green cells. The present analysis therefore fully confirms and extends the information about known activated targets.The known activated SSTF genes are generally expressed in the nascent dorsal association interneurons that make up the substantia gelatinosa and parts of the nucleus proprius. The expression patterns of Prxxl1 (DRG11), Lbxcor1(Corl1), Zic1-5, Tsh2, Sall3, Zfhx1b, Bcl11a indicate that these SSTFs are expressed in association interneurons [9], [36]–[38]. All of these genes were selected as activated targets (Table 2). Immunohistochemistry was used to confirm that Lbxcor1 was activated by Lbx1 (Fig. 4A). Thus, Lbx1 activates SSTF genes expressed in the nascent dorsal association interneurons and thereby promotes the establishment of network states that eventually confer properties of association, or pain-gating, interneurons.10.1371/journal.pone.0002179.g004Figure 4Validation of Microarray Measurements.A) Effect of Lbx1 mutation on Lbxcor1 and Isl1 expression in the thoracic neural tube at E12.5 B) Average fold change observed between hG and mG population pools in three replicate microarrays is plotted against the average fold change measured by quantitative real time PCR (qRTPCR) in at least four replicates. No genes were culled from the initially selected set of 26 genes. Gene regions amplified in qRTPCR generally differed from gene regions detected by Affymetrix probe sets. The outlier in the top right quadrant corresponds to Mafa, which gave erroneous low values in qRTPCR because the crossing thresholds occurred after more than 30 cycles. R = 0.99 if this point is disregarded.Lbx1 Represses Ventrally- and Activates Dorsally-Expressed Hox GenesOne striking observation that emerged from the present analysis was the abundance of Hox genes and their co-binding cofactors (Tgif2, Pbx3, and Pknox2) on the target lists. Three Hox genes (a4, b2, and b8) were activated and ten Hox genes (c6, c8, c9, c10, c13, d1, d8, d9, and d10) were repressed by Lbx1. It has long been known that Hoxb expression is higher in the dorsal half of the spinal cord [39]. Closer examination of all Hoxb probe sets showed that Hoxb3, b4, and b5 also had significant decreases in expression in the green cells of Lbx1 mutants. It is also known that Hoxc6, c8, c9, d9, and d10 gene expression is restricted to postmitotic cells in the ventral region of the neural tube at E12.5 [40]–[43]. The significant increase in the expression levels of these Hox genes in Lbx1 mutants indicates that Lbx1 represses the expression of these Hox genes in at least some dorsal cell populations. The small changes are likely due to the fact that Lbx1 mediated control of Hox genes only occurs in an anterior to posterior (A–P) restricted subset of the green populations for each Hox gene. Taken together, the data indicate that Lbx1 plays a key role in controlling which Hox genes can function in dorsal cell types and thereby helps to coordinates patterning along the A–P and D–V axes in the neural tube.Validation of Target GenesQuantitative real time PCR was used to validate 25, or approximately 20%, of the identified target set. Primer pairs for Lbx1, activated targets (Pax2, Lmx1b, Pax8, Pax2?e6, Mafa1, Pknox2, Sall3, Tshz2, Lbxcor1, Gbx2, Satb2, Sal4, and Zic1-5), repressed targets (Isl1, Foxd3, Olig3, Otp, FoxP2, Hmx2, Hmx3, Nr4a2(Nurr1), Sall1), and neutral genes (Sall2, Uncx4.1, Bcl11b) were designed according to the online resource PRIMER BANK. Total RNA from flow sorted green cells from mutant and heterozygote embryos was reverse transcribed and quantitative real time PCR was performed (Fig. 4B). The fold change from four replicate measurements was plotted against the fold change observed for the probe set with the highest absolute signal. The data clearly validate the fold changes observed in the microarray analyses (Tables 1, 2).Lbx1 Targets Reside in Clusters of Ultra-Conserved Non-Coding RegionsPerhaps more striking than the large number of Lbx1-affected nodes was the observation that 65 of the affected nodes were in the 150 most prominent clusters of ultra-conserved non-coding regions (UCRs) in the mammalian genome [44] and 29 were associated with highly conserved noncoding regions (HCNRs) enriched in Hox, Sox, and POU binding sites [45]. Thus, over half of the Lbx1-regulated nodes have previously been associated with genomic regions that are rich in conserved non-coding sequences (Tables 1, 2; UCR column). Many Lbx1 targets are therefore located in those genomic regions that are richest in evolutionarily conserved regulatory elements, consistent with the idea that Lbx1 participates in the evolutionarily inflexible transcription circuits that are called network kernels.The clusters of UCRs reported by Sandelin et al. [44]are ranked by density of UCRs within a 500 kb window. The frequency of Lbx1 targets at the top of the list is higher than at the bottom. Thus, 70% of the first 10, 60% of the first 40, or 50% of all 150 clusters contain SSTF genes with significantly regulated probe sets (t-test; 95% confidence). The SSTF genes located in the genomic regions that are richest in UCRs are likely to have the most cis-regulatory modules, and can therefore participate in more, or more diverse, network kernels. Those genes at the top of the list are more likely to have more different roles in development as a whole. This may explain why the frequency of Lbx1 epistasis rises with UCR density.DiscussionThe present analysis shows that a single node, Lbx1, in a very discrete phase of neural tube development, in a very limited set of neural tube populations, alters expression of 20–30% of the 500–700 active nodes of the transcriptional network of neural tube patterning. Thirty of the 141 SSTF target genes have already been implicated in neural tube development and 85 were predicted active nodes [19]. Only 26 targets were neither known nor predicted nodes. Lbx1 regulated 46 of the 229 homeodomain and 18 of the 96 basic helix-loop-helix (bHLH) containing genes in the genome (20%). It is known that 82% of the SSTFs used to describe specification in neural tube pattern formation contain either a homeodomain (65%) or a bHLH domain (17%) [19]. The large number of targets and their close association with clusters of UCRs indicates that SSTF nodes may be frequently re-used by different network kernels during development.The interaction between Lbx1 and a high fraction of the available SSTF nodes, the strong association of Lbx1 target SSTFs with the densest UCR clusters, and the ability to observe combinatorial codes by immunohistochemistry are inconsistent with standard hierarchical models of cell lineage specification, in which “master regulators” at the top of the lineage control hierarchies of lineage-specific SSTFs. Instead, they support the view that a central conserved network exists that assumes many different stable states during pattern formation. These stable states are defined by the regulatory interactions between the expressed SSTFs and the kernel-specific cis-regulatory modules of these expressed SSTFs. The dense UCR clusters associated with many of the Lbx1 targets are likely to contain such cis-regulatory modules. Collectively, these regulatory interactions stabilize the expression levels of the SSTFs that define the stable network state, which can also be described as a body part, cell type, or network kernel. Transitions between such stable states are likely to be internally clocked or brought on by inductive cues that alter the expression of one or more of the kernel's SSTFs.During the pattern formation phase of neural tube development, the “combinatorial codes of SSTFs” observed by immunohistochemistry are more consistent with “stable states that rapidly convert to other stable states” than with “smooth or gradual modifications of state” (see below). This paradigm likely reverses at later stages of development when the “stable state” or cell type has been more firmly established.It is not immediately obvious how a single SSTF perturbation can affect the expression of so many other SSTFs in so few populations in such a short time. One possibility is that rapid signaling cross talk amongst green cell populations alters gene expression levels. However, the lack of cross talk between green and white populations strongly suggests that little signaling cross talk between populations exists at this stage.A second attractive model to explain the large number of genetic dependencies is that they are not all direct molecular targets. In this model, Lbx1 would regulate the transcriptional output from a small number of the tabulated SSTF genes by direct interactions with their enhancers and these direct SSTFs targets would, in turn, regulate some of the other tabulated SSTFs, as secondary targets. Each level of direct control must alter the protein levels produced from the target gene and must therefore require time for transcription and translation. Approximately 20% of the analyzed cells came from early populations (dI4-dI6) and had therefore expressed Lbx1 for more than 24 hours, which may be sufficient for secondary target regulation. It should be noted that the five fold dilution of RNA from early cells, by RNA from late born cells (dI4LA, dI4LB), could reduce the ability to detect early Lbx1 targets.At least one modeling study suggests that SSTFs do not need to reach steady state levels to alter developmental network state [25], suggesting that 24 hours may be sufficient to allow secondary or higher order effects to be observed. If this were true, then standard immunohistochemistry, which can be done in half-day time scales, would only reveal relatively stable GRN features. Microarray snapshots of epistasis such as those generated by ANCEA, which quantitatively record fold changes, may be the only current way to detect GRN features that change in more rapid timescales.A final way to explain the large number of SSTF targets is by noting that Lbx1 regulates different sets of target genes in each population where it is expressed. While it is formally possible that all of the tabulated SSTFs are regulated in all green cells, it is known that Foxd3 and Isl1 are upregulated, and Pax2 and Lmx1b are downregulated, in different populations [9]. It is also formally possible that each cell population uses a discrete non-overlapping set of the target genes. However, Lbx1 regulates Foxd3 and Isl1 in three (dI4, dI6, dI4LA) and two (dI5, dI4LB) populations, respectively. It is therefore likely that the set of Lbx1 target genes in each of the five different populations overlaps somewhat with the sets used by each of the other populations. The degree of overlap is likely to be higher in populations that have closer lineage relationships. Thus, it is even possible that some of these target genes function in limb muscle precursors, which also express Lbx1. However, due to the remote lineage relationship a relatively small overlap is expected. Based on these considerations one would predict that Lbx1 regulates between 25 and 141 SSTF genes in each D–V cell type. Hox codes could create more cell types by subdividing the D–V cell types along the A–P axis. Thus, even fewer targets per cell type would be predicted.Assembly of a Network ModelConstructing and managing a 500–700 node GRN that simulates the emergent properties of neural tube pattern formation requires computer software such as Biotapestry [22] in which expression and epistasis data are integrated into an evolving model as they are acquired. As a first step, one must assemble a “view from the genome”. This requires all nodes and their known epistatic relationships to be entered on one page. Typically nodes that are expressed in related cell types are spatially clustered so that the future “views from the cell” will be spatially coherent.\nFig. 5 shows the “view from the genome” of the current network model of neural tube patterning. We have placed all of the known nodes (blue) and new nodes (black) according to their approximate zone of expression (shaded regions). Tables 1 and 2 list many SSTFs whose function in neural tube development has not been explored. Online resources in Mouse Genome Informatics (MGI) and published literature were searched for expression analyses for these target genes in the developing neural tube. Spatially restricted expression patterns in the developing neural tube were found for 57 of the 70 genes in Table 1 and for 54 of the 71 genes in Table 2. Expression patterns from the literature generally were not from the desired cross sections of E12.5 neural tubes. However, crude evaluations of their expression zone were made (column NT in tables). Most of the genes in Table 1 appeared to be expressed outside of the zone of Lbx1 expression, consistent with the idea that they are repressed by Lbx1. Similarly, most of the genes in Table 2 appeared to be expressed within the zone of Lbx1 expression, consistent with the idea that they are activated by Lbx1. Those genes for which no expression information could be obtained were grayed out and placed at the periphery of the model.10.1371/journal.pone.0002179.g005Figure 5Draft GRN for Neural Tube Patterning.Biotapestry was used to create a View from the Genome from known (blue) and Lbx1-responsive (black or grey) SSTF nodes and their epistatic interactions. The shaded areas represent different regions of expression: ventricular zone (yellow); ventral postmitotic (green); dorsal commissural (salmon); dorsal association (blue); no expression information (grey). Epistatic relationships between nodes that have been published are shown in black. The epistatic relationships of Lbx1 that were discovered in this work are shown in green (activating) or red (repressing). Those that were previously known and confirmed in this work are shown as bold red or green lines. We emphasize that the interactions shown are genetic and do not represent direct molecular interactions. Few, if any, direct molecular interactions have been demonstrated in the neural tube literature.The background network of Fig. 5 attempts to capture the knowledge base of known nodes and their interactions from the literature. Superimposed in color on this background are the new interactions revealed by this study. The number of new nodes and interactions is far greater than the known ones, demonstrating the efficiency of the ANCEA method. There is also a clear interlocking of the two data sets because 30 of the known nodes are Lbx1 targets and because virtually all of the known Lbx1 interactions were recapitulated in the ANCEA dataset. This interlocking provides further validation for the ANCEA method.The model, it should be emphasized, is only a first crude estimate of the network. The sea urchin endomesoderm models, in which network kernels were discovered, reveal how expression and epistasis information can be coupled in Biotapestry, to give a series of “views from the cell” that show how the network progresses from state to state in time and space [22]. The expression level of each SSTF node is entered for each cell type, at each time in development. The relatively simple and rigid anatomy of the sea urchin embryo allows cell type to be independently identified by anatomical position. In contrast, cell types, or populations, in neural tube development are primarily defined by combinatorial codes of SSTF expression. It follows that a change in the expression status at SSTF node would formally represent the establishment of a new population, cell type, or network state. Given that the set of Lbx1 targets differs between known populations, it is reasonable to assume that each network state in which Lbx1 is expressed has a different set of Lbx1 targets. If this is a general property of SSTFs, or only of homeodomain SSTFs, it will be impractical, or perhaps impossible, to construct the network model using only immunohistochemical techniques.Coupling ANCEA with PPA to Resolve Networks in Fluid AnatomyIn rigid anatomies, cell populations can be identified by location so that changes in SSTF expression can be tracked over time, or after genetic perturbations. In fluid anatomies such as those of mammals, landmarks that allow populations to be identified only by location generally do not exist during embryonic pattern formation. This problem is particularly acute in the developing central nervous system, where cell populations move and position themselves relative to other populations based on their specifications. Thus, location becomes almost useless as a means to identify populations.As a result, populations are typically defined by the combination of SSTFs they express. However, the SSTFs they express regulate each other and also define the function of the population. Thus, both the definition and function of a population are inextricably linked to the transcriptional network state of that population. Independent or singular markers of populations do not exist. Mutation of a SSTF to discover its “function” in a population, will generally alter expression of the other SSTF genes used to identify the population. How can one disentangle such a fluid system and resolve it into a branching time series of states without the aid of spatial landmarks?Given that the states of the system are typically defined by combinations of SSTF expression, it seems logical to first determine which SSTFs are differentially expressed within the system. Any SSTFs that are uniformly expressed or silent throughout the system will not contribute to the dynamic description of the states of the system and should be eliminated from consideration. We will call these SSTFs passive nodes of the network model of the system. In contrast, SSTFs that are differentially expressed within the system can contribute to the dynamic description of the states of the system and we will call these active nodes.In our previous work [19] we describe a method, called population partitioning analysis (PPA), to systematically identify the active nodes of a system. It requires that one know a SSTF that is differentially expressed within the system. Preferably, the expression pattern of this SSTF should divide the system into two roughly similar sized pools of populations, one pool that expresses the SSTF and one pool that does not. The system must then be experimentally separated into these two pools and the expression of all SSTFs compared between the pools. Passive nodes, being uniformly expressed throughout the system, will show no differential expression on a per cell basis no matter how you divide the system. Active nodes, being expressed in some populations but not in others, will show different levels in the two pools. We demonstrated this by comparing green (Lbx1+) and white (Lbx1−) population pools in the heterozygous (Lbx1GFP/+) neural tube system [19].These studies indicated that 500–700 SSTFs were differentially expressed in this system and identified approximately 200 of these active nodes.Once the active nodes of a system have been identified, one needs to know the interactions between the active nodes to set up a dynamic network model of the system. One must bear in mind that the interactions of an active node may differ in different states (i. e. populations) of the system.At the molecular level, interactions between active nodes involve one SSTF protein binding a cis-regulatory module that regulates expression of the transcript of a target SSTF. However, dealing with direct interactions at the molecular level is currently intractable. The location of cis regulatory modules is generally not known and measuring the effect of SSTF occupancy on them in a particular population (i.e. state) of the system is not currently feasible.At the genetic level, interactions between active nodes involve mutating the gene for an upstream SSTF and observing changes in the expression of a target SSTF. This establishes that the mutated SSTF is epistatic to the target SSTF. However, the epistatic interaction established in such an experiment is not universally valid throughout the system. It must be linked to a particular state (i.e. population) of the system. If the target SSTF is one of the SSTFs used to define the population (i.e. state) of the system, then one may not be able to link the epistatic interaction to the population because the population (i.e. state) disappears as a result of the technique (mutation) used to measure it. One must have a way of identifying the equivalent population of cells in the presence or absence of the upstream SSTF.One of the most powerful means to do this is to make a knock-in at the upstream SSTF locus, in which a reporter such as GFP replaces the open reading frame of the SSTF. This allows one to identify equivalent cells in heterozygotes and mutants and therefore allows target gene expression to be compared in these cells in the presence or absence of the SSTF. This is often done by immunohistochemistry or it can be done by the ANCEA method we have developed here. It is critical to do the analysis shortly after the onset of reporter expression so that the measured effects will be primary rather than secondary. In ANCEA one constrains the search for epistatic interactions to the populations (i.e. states) that express the active node one is testing. For example, in this report, we compared gene expression of heterozygote Lbx1GFP/+ green cells with null Lbx1GFP/GFP green cells to discover epistatic interactions between Lbx1 and target SSTFs. The interactions discovered in this report must occur in one or more of the five populations labeled green by the Lbx1GFP allele. Thus, the ANCEA method will typically assign the epistatic interactions it discovers to a pool of populations rather than to an individual population (i.e. state).Ideally, one would like to perform ANCEA on a homogeneous population so that discovered interactions would pertain to only one state and could be input into a “view from the cell” in the network model. Unfortunately, individual populations (i.e. states) are rarely, if ever, defined by the expression of a single SSTF. Instead, individual populations are defined by the combination of SSTFs they express. Most SSTFGFP knock-ins are likely to label several populations (i.e. states) in a local piece of anatomy. One or more additional SSTF markers will generally be needed to identify individual populations.One future refinement would be to purify cells from mice bearing multiple fluorescent knock-ins of different colors (e.g. SSTF1GFP|SSTF2 dsRed). Such an approach would more severely limit the number of populations that interactions could be assigned to in a given ANCEA experiment. The approach would be limited by the fraction of useful embryos in each litter.A second approach would be to mutate one SSTF while sorting cells labeled by another SSTF. For example, the same LbxGFP cells could be purified from embryos that are wild type, heterozygote and mutant at another SSTF locus. This approach would be most useful with active nodes that do not interact with the GFP-tagged node.Applying ANCEA to the active nodes identified by PPA will rapidly identify comprehensive sets of interactions for each active node and assign these interactions to specific pools of populations. Practically this can be done by purifying and analyzing population pools from mutant and heterozygotes of GFP knock-ins at known active nodes, as we have here for Lbx1. The results must be integrated with knowledge about combinatorial codes of SSTF expression to assemble network models computationally.If the large number of genetic dependencies observed for Lbx1 is typical for all, or just homeodomain containing, SSTFs, then the patterning GRN that generates cell diversity in neural tube is far more cross-linked and complex than was previously appreciated. Either Lbx1 is a highly connected hub in some or all of the five populations that are currently defined, or there are far more populations in which Lbx1 participates in only a few of interactions we discovered.Materials and MethodsCell sorting, RNA isolation, probe preparation and quantitative real time PCR (qRTPCR) were described previously [19]. Corl1 antibody was obtained from Yuichi Ono and immunohistochemistry was performed as described [29]. Annotations and permutation analyses were performed by scripts written in the business relational database software FileMaker Pro 6.0. Electronic databases used in this work will be shared.Permutation Fold-scanning AnalysisBelow we will describe only the analysis of the heterozygous (control) and mutant (test) green samples. The analysis for the white samples was done identically (substitute W for G in the sample names in the text below).Internal- and Cross-ComparisonsIn the description below, individual arrays are compared, in a pair-wise manner, within conditions (internal-comparisons; biological replicates) and across conditions (cross-comparisons, test vs. control, between mutant and heterozygote). The six internal-comparisons are hG1 vs hG2, hG1 vs hG3, hG2 vs hG3, mG1 vs mG2, mG1 vs mG3, and mG2 vs mG3. The nine cross-comparisons are hG1 vs mG1, hG1 vs mG2, hG1 vs mG3, hG2 vs mG1, hG2 vs mG2, hG2 vs mG3, hG3 vs mG1, hG3 vs mG2, and hG3 vs mG3.Selecting a Uniformly Changing Set of Probe setsOnly probe sets that changed in a uniform direction (up or down) in all six internal-, or all nine cross-comparisons were considered in the fold-scanning analyses below.Selection of a set of probe sets that changes uniformly in all six internal comparisons was based on the following type of logic: {IF [(hG1>hG2) AND (hG1>hG3) AND (hG2>hG3) AND (mG1>mG2) AND (mG1>mG3) AND (mG2>mG3)]OR [(hG1<hG2) AND (hG1<hG3) AND (hG2<hG3) AND (mG1<mG2) AND (mG1<mG3) AND (mG2<mG3)] THEN “select probe set for further evaluation”}. The three replicate hG arrays can be numbered (1,2,3) in six different ways. Similarly, the three replicate mG arrays can be numbered (1,2,3) in six different ways. As a consequence there are 36 equivalent ways to produce a list of probe sets that change in one uniform direction (only all up or only all down). If one allows two uniform directions (either all up or all down) then half of these 36 ways become redundant. Thus, there are 18 ways to arrange the data and produce a list of probe sets that change in a uniform direction, either all up or all down. For Green internal comparisons, the 18 lists contain an average of 194±109 probe sets. The longest list contains 558 probe sets and the shortest list contains 109 probe sets. {For White internal comparisons, the 18 lists contain an average of 195±74 probe sets The longest list contains 345 probe sets and the shortest list contains 108 probe sets.} Each list produces similar results when subjected to permutation fold scanning analysis (data not shown). However, we conservatively chose the longest list to maximize the measured error. Thus, the uniform list used for generating the internal comparison curves in Fig. 3A and Fig. 3B contained 558 and 345 probe sets (out of the 3574 SSTF probe sets), respectively. These probe sets changed in a uniform direction in all six internal comparisons.In contrast, there is only one way to select a set of probe sets that change in a uniform direction in all nine cross-comparisons. Selection of probe sets was based on the following logic: {IF [(hG1>mG1) AND (hG1>mG2) AND (hG1>mG3) AND [(hG2>mG1) AND (hG2>mG2) AND (hG2>mG3) AND [(hG3>mG1) AND (hG3>mG2) AND (hG3>mG3)]OR [(hG1<mG1) AND (hG1<mG2) AND (hG1<mG3) AND [(hG2<mG1) AND (hG2<mG2) AND (hG2<mG3) AND [(hG3<mG1) AND (hG3<mG2) AND (hG3<mG3)]THEN “select probe set for further evaluation”}. The selected list would be identical regardless of how arrays are assigned to names in this case. There were 697 and 376 uniformly changing SSTF probe sets in the Green (Fig. 3A) and White (Fig. 3B) cross-comparisons , respectively. It should be noted that nine “logical AND” conditions need to be met to make the uniform set in cross comparisons, whereas only six “logical AND” conditions need to be met to make the uniform set in internal comparisons. Thus, the noise measured by the internal comparisons is again, conservatively, overestimated.Permutation Fold-ScanningInternal-comparisons were fold scanned in the arrangement that produced the largest uniform set (G = 558 probe sets; W = 345 probe sets). As noted above, there are six possible internal-comparisons that can be made for each probe set. These comparisons were labeled A,B,C,D, E, and F. The fold scanner script asks, for a specific combination of three comparisons (three of the letters A through F), within the set of uniformly changing SSTF probe sets, how many probe sets pass the fold cutoff in all three comparisons. The six letters A–F (n) can be combined in sets of 3 (k) in 20 ways according to the combinatoric formula nCk = n!/((n-k)!*k!). Thus, there are 20 equivalent ways of performing fold scanning on a given uniform changing set. All twenty are performed by the script and the average and standard deviation at each fold cutoff are plotted in Fig. 3. As the scanner approaches the fold cutoff of 1, the curves rise sharply and closely approach the uniform set size on the Y-axis. This is more apparent as fold scanning is done from 1.1 to 1.01 fold cutoffs (data not shown).Cross-comparisons were fold scanned on the uniform changing set of SSTF probe sets in each case (G = 697 probe sets; W = 376 probe sets). ). As noted above, there are nine possible cross-comparisons that can be made for each probe set. These comparisons were labeled A,B,C,D, E, F, G, H, and I. The fold scanner script asks, for a specific combination of three comparisons (three of the letters A–I), within the set of uniformly changing SSTF probe sets, how many probe sets pass the fold cutoff in all three comparisons. The nine letters A–I (n) can be combined in sets of 3 (k) in 84 ways according to the combinatoric formula nCk = n!/((n-k)!*k!). Thus, there are 84 equivalent ways of performing fold scanning on a given uniform changing set. All 84 are performed by the script and the average and standard deviation at each fold cutoff are plotted in Fig. 3. As the scanner approaches the fold cutoff of 1, the curves rise sharply and closely approach the uniform set size on the Y-axis. This is more apparent as fold scanning is done from 1.1 to 1.01 fold cutoffs (data not shown).Computing False Discovery Rates as a Function of Fold CutoffInternal comparisons should reveal no changes in ideal or perfect replicates. Thus, any differences in our internal comparisons reflect the combined noise, due either to measurement, sample preparation, or to real differences in “identical” biological samples. Two independent measurements of this combined noise, using G internal or W internal comparisons, produced remarkably similar numbers of changes at all fold cutoffs, as would be expected for measurement noise (open boxes in Fig 3A and B). Averages over all 18 Uniform internal sets are even more similar (data not shown).If there were no true biological differences between mutant and heterozygote samples, then the cross comparisons should reveal the same number of changes at each fold cutoff as the internal comparisons. This is not the case. There are clearly far more changes at all fold cutoffs. The changes observed at each fold cutoff in cross comparisons include real and noise-related changes. Because averages of three specific cross-comparisons were evaluated in both internal and cross comparisons (as opposed to 3 of 6 vs. 3 of 9), the number of probe sets that showed changes at or above a given fold cutoff were evaluated in a statistically equivalent manner in cross and internal comparisons and were therefore directly comparable. The cross comparisons are signal plus noise (real plus false positives) and the internal comparisons are noise only (false positives). The method did not introduce investigator bias or require assumption of a statistical model and is therefore nonparametric.The FDR (circles) was calculated by dividing the internal comparison values (false positives) by the cross comparison values (real plus false positives) at each fold cutoff. The number of real positives (triangles) was calculated at each fold cutoff by subtracting the internal comparison values from the cross comparison values.Supporting InformationTable S1Tracking of Flow Sorting and RNA Preparation(0.07 MB DOC)Click here for additional data file.\n\nREFERENCES:\n1. GowanKHelmsAWHunsakerTLCollissonTEbertPJ\n2001\nCrossinhibitory activities of Ngn1 and Math1 allow specification of distinct dorsal interneurons.\nNeuron\n31\n219\n232\n11502254\n2. HelmsAWJohnsonJE\n2003\nSpecification of dorsal spinal cord interneurons.\nCurr Opin Neurobiol\n13\n42\n49\n12593981\n3. MarquardtTPfaffSL\n2001\nCracking the transcriptional code for cell specification in the neural tube.\nCell\n106\n651\n654\n11572771\n4. CasparyTAndersonKV\n2003\nPatterning cell types in the dorsal spinal cord: what the mouse mutants say.\nNat Rev Neurosci\n4\n289\n297\n12671645\n5. GouldingMLanuzaGSapirTNarayanS\n2002\nThe formation of sensorimotor circuits.\nCurr Opin Neurobiol\n12\n508\n515\n12367629\n6. JessellTM\n2000\nNeuronal specification in the spinal cord: inductive signals and transcriptional codes.\nNat Rev Genet\n1\n20\n29\n11262869\n7. ShirasakiRPfaffSL\n2002\nTranscriptional codes and the control of neuronal identity.\nAnnu Rev Neurosci\n25\n251\n281\n12052910\n8. LeeSKPfaffSL\n2001\nTranscriptional networks regulating neuronal identity in the developing spinal cord.\nNat Neurosci\n4\nSuppl\n1183\n1191\n11687828\n9. GrossMKDottoriMGouldingM\n2002\nLbx1 specifies somatosensory association interneurons in the dorsal spinal cord.\nNeuron\n34\n535\n549\n12062038\n10. SmithEHargraveMYamadaTBegleyCGLittleMH\n2002\nCoexpression of SCL and GATA3 in the V2 interneurons of the developing mouse spinal cord.\nDev Dyn\n224\n231\n237\n12112475\n11. HargraveMKarunaratneACoxLWoodSKoopmanP\n2000\nThe HMG box transcription factor gene Sox14 marks a novel subset of ventral interneurons and is regulated by sonic hedgehog.\nDev Biol\n219\n142\n153\n10677261\n12. Moran-RivardLKagawaTSaueressigHGrossMKBurrillJ\n2001\nEvx1 is a postmitotic determinant of v0 interneuron identity in the spinal cord.\nNeuron\n29\n385\n399\n11239430\n13. PieraniAMoran-RivardLSunshineMJLittmanDRGouldingM\n2001\nControl of interneuron fate in the developing spinal cord by the progenitor homeodomain protein Dbx1.\nNeuron\n29\n367\n384\n11239429\n14. SharmaKShengHZLettieriKLiHKaravanovA\n1998\nLIM homeodomain factors Lhx3 and Lhx4 assign subtype identities for motor neurons.\nCell\n95\n817\n828\n9865699\n15. TsuchidaTEnsiniMMortonSBBaldassareMEdlundT\n1994\nTopographic organization of embryonic motor neurons defined by expression of LIM homeobox genes.\nCell\n79\n957\n970\n7528105\n16. DasenJSTiceBCBrenner-MortonSJessellTM\n2005\nA Hox regulatory network establishes motor neuron pool identity and target-muscle connectivity.\nCell\n123\n477\n491\n16269338\n17. DasenJSLiuJPJessellTM\n2003\nMotor neuron columnar fate imposed by sequential phases of Hox-c activity.\nNature\n425\n926\n933\n14586461\n18. EnsiniMTsuchidaTNBeltingHGJessellTM\n1998\nThe control of rostrocaudal pattern in the developing spinal cord: specification of motor neuron subtype identity is initiated by signals from paraxial mesoderm.\nDevelopment\n125\n969\n982\n9463344\n19. KioussiCShihHPLoflinJGrossMK\n2006\nPrediction of active nodes in the transcriptional network of neural tube patterning.\nProc Natl Acad Sci U S A\n103\n18621\n18626\n17132738\n20. BarabasiALOltvaiZN\n2004\nNetwork biology: understanding the cell's functional organization.\nNat Rev Genet\n5\n101\n113\n14735121\n21. LevineMDavidsonEH\n2005\nGene regulatory networks for development.\nProc Natl Acad Sci U S A\n102\n4936\n4942\n15788537\n22. LongabaughWJDavidsonEHBolouriH\n2005\nComputational representation of developmental genetic regulatory networks.\nDev Biol\n283\n1\n16\n15907831\n23. BolouriHDavidsonEH\n2002\nModeling transcriptional regulatory networks.\nBioessays\n24\n1118\n1129\n12447977\n24. BolouriHDavidsonEH\n2002\nModeling DNA sequence-based cis-regulatory gene networks.\nDev Biol\n246\n2\n13\n12027430\n25. BolouriHDavidsonEH\n2003\nTranscriptional regulatory cascades in development: initial rates, not steady state, determine network kinetics.\nProc Natl Acad Sci U S A\n100\n9371\n9376\n12883007\n26. DavidsonEHErwinDH\n2006\nGene regulatory networks and the evolution of animal body plans.\nScience\n311\n796\n800\n16469913\n27. PfaffSLMendelsohnMStewartCLEdlundTJessellTM\n1996\nRequirement for LIM homeobox gene Isl1 in motor neuron generation reveals a motor neuron-dependent step in interneuron differentiation.\nCell\n84\n309\n320\n8565076\n28. DavidsonEHMcClayDRHoodL\n2003\nRegulatory gene networks and the properties of the developmental process.\nProc Natl Acad Sci U S A\n100\n1475\n1480\n12578984\n29. GrossMKMoran-RivardLVelasquezTNakatsuMNJaglaK\n2000\nLbx1 is required for muscle precursor migration along a lateral pathway into the limb.\nDevelopment\n127\n413\n424\n10603357\n30. MullerTBrohmannHPieraniAHeppenstallPALewinGR\n2002\nThe homeodomain factor lbx1 distinguishes two major programs of neuronal differentiation in the dorsal spinal cord.\nNeuron\n34\n551\n562\n12062039\n31. KadonagaJT\n2004\nRegulation of RNA polymerase II transcription by sequence-specific DNA binding factors.\nCell\n116\n247\n257\n14744435\n32. SimeoneAD'ApiceMRNigroVCasanovaJGrazianiF\n1994\nOrthopedia, a novel homeobox-containing gene expressed in the developing CNS of both mouse and Drosophila.\nNeuron\n13\n83\n101\n7913821\n33. WangWLoPFraschMLufkinT\n2000\nHmx: an evolutionary conserved homeobox gene family expressed in the developing nervous system in mice and Drosophila.\nMech Dev\n99\n123\n137\n11091080\n34. BritanovaOAkopovSLukyanovSGrussPTarabykinV\n2005\nNovel transcription factor Satb2 interacts with matrix attachment region DNA elements in a tissue-specific manner and demonstrates cell-type-dependent expression in the developing mouse CNS.\nEur J Neurosci\n21\n658\n668\n15733084\n35. JohnAWildnerHBritschS\n2005\nThe homeodomain transcription factor Gbx1 identifies a subpopulation of late-born GABAergic interneurons in the developing dorsal spinal cord.\nDev Dyn\n234\n767\n771\n16193514\n36. DingYQYinJKaniaAZhaoZQJohnsonRL\n2004\nLmx1b controls the differentiation and migration of the superficial dorsal horn neurons of the spinal cord.\nDevelopment\n131\n3693\n3703\n15229182\n37. MizuharaENakataniTMinakiYSakamotoYOnoY\n2005\nCorl1, a novel neuronal lineage-specific transcriptional corepressor for the homeodomain transcription factor Lbx1.\nJ Biol Chem\n280\n3645\n3655\n15528197\n38. LiMZWangJSJiangDJXiangCXWangFY\n2006\nMolecular mapping of developing dorsal horn-enriched genes by microarray and dorsal/ventral subtractive screening.\nDev Biol\n292\n555\n564\n16516881\n39. GrahamAMadenMKrumlaufR\n1991\nThe murine Hox-2 genes display dynamic dorsoventral patterns of expression during central nervous system development.\nDevelopment\n112\n255\n264\n1685115\n40. ErseliusJRGouldingMDGrussP\n1990\nStructure and expression pattern of the murine Hox-3.2 gene.\nDevelopment\n110\n629\n642\n1723949\n41. OliverGWrightCVHardwickeJDe RobertisEM\n1988\nDifferential antero-posterior expression of two proteins encoded by a homeobox gene in Xenopus and mouse embryos.\nEmbo J\n7\n3199\n3209\n2460338\n42. HedlundEKarstenSLKudoLGeschwindDHCarpenterEM\n2004\nIdentification of a Hoxd10-regulated transcriptional network and combinatorial interactions with Hoxa10 during spinal cord development.\nJ Neurosci Res\n75\n307\n319\n14743444\n43. TiretLLe MouellicHMauryMBruletP\n1998\nIncreased apoptosis of motoneurons and altered somatotopic maps in the brachial spinal cord of Hoxc-8-deficient mice.\nDevelopment\n125\n279\n291\n9486801\n44. SandelinABaileyPBruceSEngstromPGKlosJM\n2004\nArrays of ultraconserved non-coding regions span the loci of key developmental genes in vertebrate genomes.\nBMC Genomics\n5\n99\n15613238\n45. BaileyPJKlosJMAnderssonEKarlenMKallstromM\n2006\nA global genomic transcriptional code associated with CNS-expressed genes.\nExp Cell Res\n312\n3108\n3119\n16919269"
+}

batch_8/PMC2527685.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2527685",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2527685\nAUTHORS: Yuanfang Guan, Chad L. Myers, Rong Lu, Ihor R. Lemischka, Carol J. Bult, Olga G. Troyanskaya\n\nABSTRACT:\nEstablishing a functional network is invaluable to our understanding of gene function, pathways, and systems-level properties of an organism and can be a powerful resource in directing targeted experiments. In this study, we present a functional network for the laboratory mouse based on a Bayesian integration of diverse genetic and functional genomic data. The resulting network includes probabilistic functional linkages among 20,581 protein-coding genes. We show that this network can accurately predict novel functional assignments and network components and present experimental evidence for predictions related to Nanog homeobox (Nanog), a critical gene in mouse embryonic stem cell pluripotency. An analysis of the global topology of the mouse functional network reveals multiple biologically relevant systems-level features of the mouse proteome. Specifically, we identify the clustering coefficient as a critical characteristic of central modulators that affect diverse pathways as well as genes associated with different phenotype traits and diseases. In addition, a cross-species comparison of functional interactomes on a genomic scale revealed distinct functional characteristics of conserved neighborhoods as compared to subnetworks specific to higher organisms. Thus, our global functional network for the laboratory mouse provides the community with a key resource for discovering protein functions and novel pathway components as well as a tool for exploring systems-level topological and evolutionary features of cellular interactomes. To facilitate exploration of this network by the biomedical research community, we illustrate its application in function and disease gene discovery through an interactive, Web-based, publicly available interface at http://mouseNET.princeton.edu.\n\nBODY:\nIntroductionEstablishing a functional network is invaluable to furthering our understanding of gene function, pathways, and systems-level properties of an organism and can be a powerful resource in directing targeted experiments. The availability of diverse genome-scale data enables the prediction of networks encompassing all or at least most of the proteins in an organism. In Saccharomyces cerevisiae, probabilistic models have been used to predict the genomewide protein–protein functional interactions by integrating diverse data types [1]–[6]. Such probabilistic approaches have also been used in mammals to predict physical interactions [7],[8] and to generate expression networks [9]–[13]. In human, functional relationship networks have also been generated by integrating diverse interaction data [14]. However, it is still challenging to predict functional relationships through integrating diverse genomic data in mammalian model systems, due to the intrinsic complexity of these genomes and functional biases in individual datasets. Yet recent accumulation of both traditional targeted experiments, including protein physical interactions [15]–[17], gene-disease/phenotypic associations [18] and genome-scale data including gene expression and tissue localization [19]–[21], phylogenetic and phenotypic profiles [22],[23], as well as data retrieved based on homology [2],[24] provides the basis for establishing a global functional relationship network in the laboratory mouse [25].We describe here a functional network in mouse generated by integrating a wide range of data types. In contrast to interactomes that represent physical interactions, our functional network predicts the probability that two proteins are involved in the same biological process and thus represents a more comprehensive combination of physical, genetic and regulatory linkages (Figure 1A). We demonstrate the utility of our network to predict gene functions and pathway components by both computational and experimental approaches. Further, we demonstrate how it can be used to further our understanding of the systems-level features of the mouse functional network. Our global functional network for the laboratory mouse is a valuable resource for analysis and annotation of the mouse proteome and can be used as a means of generating biological hypotheses for subsequent experimental validation, especially through the interactive public web interface available at http://mouseNET.princeton.edu.10.1371/journal.pcbi.1000165.g001Figure 1Strategy for processing and integration of diverse genomic data.(A) Schematic of the network integration pipeline. We collected five different types of data that are indicative of functional relationships, each of which may consist of multiple datasets (Table 1). We assessed the redundancy of each pair of datasets by comparing likelihood ratios with and without the independence assumption; datasets for which these values differed significantly were deemed mutually redundant and combined as a single input node in the Bayesian network for the purposes of integration. Finally, we systematically grouped continuous data and integrated all data with a naïve Bayes classifier to predict pair-wise functional relationships. (B) Global view of the predicted mouse functional network with higher than 0.8 confidence level of linkage. Nodes of high connectivity (more than 20 interactions) are labeled and highlighted in red.ResultsA Probabilistic Model To Predict Functional Relationships by Integrating Diverse Data TypesBayesian networks have been used successfully for integrating diverse data sources in many biological settings, including protein function prediction [3],[6], prediction of genetic interactions [26], physical interactions [4],[7] and most relevant to this work, prediction of functional networks in S. cerevisiae\n[2],[5],[6] and human [14]. The Bayesian approach is especially well-suited to our problem, where many genome-scale data have missing values and collections of individual investigations may not be a complete representation of genome profiles. Based on a Bayesian framework, we designed a method that combines redundant datasets, processes continuous data, minimizes over-fitting and finally, integrates all experimental evidence (Table 1) in a confidence-based manner to estimate the genomewide pair-wise probabilities of functional linkage (Figure 1A). The resulting mouse interactome includes 20,581 genes, with edges representing the probability of functional relationship between each pair (Figure 1B). As demonstrated below, creation of this functional network through integrating diverse data sources can facilitate identification of novel pathway components and represents a powerful resource for understanding genetic diseases and network evolution.10.1371/journal.pcbi.1000165.t001Table 1Data sources used for functional interactome integration.Data TypeData SourcesDate/VersionNumber of Protein Pairs\nProtein–protein physical interaction data\nBIND [15]\n01/21/072,709DIP [17]\n01/21/0747GRID [16]\n10/16/068,144OPHID [24]\n10/28/0632,342\nPhenotype/disease\nMGI phenotype [18]\n01/22/072,765,378OMIM disease01/22/07\nPhylogenetic profiles\nInparanoid [23]\nVersion 4.0123,284,253BioMart [22]\n01/22/07127,017,891\nHomologous functional relationship predictions\nbioPIXIE (yeast) [2]\n10/30/064,280,332\nExpression and Tissue localization\nSAGE [19]\nVersion 02/14/06139,871,175Zhang et al [34]\nN/A92,011,395Su et al. [42]\nN/A165,756,528MouseNET Recovers Functional RelationshipsA key application of a functional network prediction is to uncover novel pathway components. We first evaluated the accuracy of our predicted network through cross-validation analysis on known functional linkages (co-annotations of proteins to specific Gene Ontology [27] terms), which is the standard for unbiased computational evaluation. In short, cross-validation can be used to assess the accuracy of predictions by evaluating the system's accuracy in recovering subsets of known annotations withheld during the training process. Our integrated network is substantially more successful in predicting known functional linkages than any of the individual datasets and making more correct predictions (demonstrating higher precision) at every confidence cutoff (Figure 2A). This result is robust to using a different annotation standard, i.e., co-annotation to the same Kyoto Encyclopedia of Genes and Genomes [28] (KEGG) pathways (Figure 2B). Notably, although the relative performance of datasets varies with different standards, the consistently good performance of our results suggests that the integrated predictions are robust to variations in the annotation standard.10.1371/journal.pcbi.1000165.g002Figure 2Computational performance analysis of the integrated network to predict functional relationships and the relative performance of different datasets.(A) Five-fold cross-validation of the integrated results applied to predict gold standard pairs defined by co-annotation to specific GO terms. Positive pairs were defined as those having at least one co-annotation to a specific GO term. Negative pairs are those that have a specific annotation, but share no co-annotations. Precision, or the fraction of correct predictions out of all predictions made, is measured across a number of cutoffs in prediction confidence (higher cutoff allows for less predictions of higher quality, and lowering the cutoff allows more predictions to be made at the cost of some decrease in accruacy). MouseNET predictions always have higher accuracy than those of the individual datasets. (B) Performance of the integrated results when evaluated against a different test set where positives are defined as pairs co-annotated to the same KEGG pathways, and negatives are pairs in which both members are annotated in KEGG, but share no co-annotations. Both performance measurements show that the integrated results are better in recovering known functional relationships than individual datasets.A common pitfall of many global integration schemes is the tendency to make precise predictions over only a limited set of biological processes [29]. Thus we evaluated the functional composition of our integrated results using KEGG, which is an accurate representation of our current knowledge of different pathways. The integrated network exhibits a balanced representation of a large group of pathways, even though many individual datasets have significant functional biases (Figure S1, the complete statistics of this functional composition analysis are included in the Dataset S1). For instance, the protein–protein interaction data obtained from the Biomolecular Interaction Network Database (BIND) [15] is significantly skewed towards the processes of focal adhesion. In contrast, given the broad functional coverage of the integrated network, we expect our approach will be useful in further characterization of a variety of pathways.MouseNET Predicts Novel Pathway Components and Gene FunctionsThe high accuracy in predicting co-annotation to KEGG pathways (Figure 2B) by our network and its broad functional coverage (Figure S1) suggest that mouseNET can accurately capture pathway-based functional linkages for a variety of processes. We thus focused specifically on the predicted functional network for the major conserved signaling pathways related to development, including Hedgehog, Wnt, MAPK, TGF-β, Notch, and Toll-like receptor signaling pathways. We find that in addition to recovering known pathway components (Figure S2), these networks include a number of proteins not previously annotated to the pathway. Many of these novel predictions have reasonable experimental support in the literature. For example, in the 40 most tightly connected nodes surrounding known MAPK pathway proteins (Figure 3), 14 of them are annotated as the canonical pathway components in KEGG (p<10−10, hypergeometric distribution). Furthermore, two of the other nodes (Kit, MGI:96677 and Shh, MGI:98297) are not annotated to the MAPK pathway in KEGG but are annotated in the Gene Ontology [27] to be MAPK-related. Another nine unannotated predictions in the cluster of 40 have been suggested in literature to be involved in the MAPK pathway (Table S2 and Text S1). Thus, our system not only recovers well-established knowledge but also implicates novel pathway components, and therefore could be a powerful tool for generating hypotheses for experimental approaches.10.1371/journal.pcbi.1000165.g003Figure 3Analysis of MAPK pathway predictions based on the integrated functional network.Predictions were derived by iteratively sampling 10 proteins from the known MAPK pathway and finding the closest 40 neighbors based on network adjacency. The results shown are based on an aggregation of 300 such samplings. Bright blue denotes proteins annotated to the canonical MAPK pathway in KEGG. Many of the newly predicted components, although not annotated in KEGG, are supported in the literature (Table S2) and are colored in red. Predictions without literature support are colored in purple. Linkages predicted to be above 0.5 confidence level by our integrated network are shown.Our genomewide prediction of protein function based on the integrated network produced 689 novel annotations with an estimated 80% precision. A subset of these new predictions was evaluated through examination of the literature by MGD curators and the precision estimate was confirmed (Dataset S2). Of these, 17 predictions were confirmed based on literature evidence at the level sufficient for annotation in MGI, and another six were found to have some support in the literature, but at a level not yet sufficient for GO annotation. For example, Retn (MGI:1888506), which does not have a GO biological process or KEGG pathway annotation, was predicted with high confidence (over 0.8) to be involved in glucose homeostasis (GO:0042593). The loss of Retn was indeed found to improve glucose homeostasis in leptin deficiency [30], confirming the prediction. This evaluation demonstrates that through integrating information from diverse sources, the system is capable of making accurate novel predictions on genes not previously annotated in GO or KEGG.Experimental Validation by Nanog Down-Regulation Induced Cell DifferentiationTo further validate novel functional relationships predicted by our integrative network, we investigated proteins predicted to cluster around the homeobox transcription factor Nanog (MGI:1919200), which is an essential gene responsible for maintaining embryonic cell fate. Specifically, we experimentally down-regulated the expression of Nanog, and observed the nuclear protein expression changes of the top functional interactors in our predicted network by mass spectrometry. Five of the top 10 Nanog interactors predicted by mouseNET (Figure 4A) were detected in the nuclei and thus, we could evaluate their expression following Nanog down-regulation. We observed that after Nanog down-regulation, expression levels of four of them either significantly increased (DNA (cytosine-5-)-methyltransferase 3-like, Dnmt3l, MGI:1859287 and DNA methyltransferase 3B, Dnmt3b, MGI:1261819) or decreased (transformation related protein 53, Trp53, MGI:98834 and POU domain, class 5, transcription factor 1, Pou5f1, MGI:101893) (p<0.1 when compared to the overall distribution of the nucleus-detected proteins, Figure S8). Of those, Pou5f1 has also been previously shown to be involved in ES cell regulation [31],[32] and it has significant overlap in genomic binding targets with Nanog\n[33],[34]. Furthermore, the change in expression for these four proteins is consistent for different time points after Nanog knock-down, and increases consistently over the time course (Figure 4B). This experimental verification demonstrates that our system is a powerful tool which can aid researchers in generating accurate hypotheses for discovery of proteins involved in a specific cellular process.10.1371/journal.pcbi.1000165.g004Figure 4Validation by Nanog down-regulation experiment.(A) The top 10 neighbors of Nanog as predicted by Bayesian integration. Links with more than 0.1 confidence level are presented in the figure. The colors of Trp53, Dnmt3b, Dnmt3l, Pou5f1, and H3f3a indicate the Log2 changes in protein expression on the fifth day after Nanog knock down compared to day 0. (B) Protein expression changes detected by mass spectrometry after Nanog knock-down. Four of the five top neighbors detected in the nucleus have significant changes in protein expression level, with increasing changes during the time course.Our functional network can also highlight information about physical interactions and transcriptional binding sites. For example, the 17 physical interactions with Nanog identified by Wang et al. were highly enriched in pairs of high functional relationship confidence (Mann-Whitney U test p = 0.00069). In addition, on the transcription level, the Nanog binding loci associated genes [34] were also highly enriched in high confidence functional interactors of Nanog predicted by our network (U test p = 3.98E-18). Therefore, by integrating a diverse collection of data, mouseNET enables users to explore variety types of functional associations, including physical interactions and transcriptional level regulation.Topological Analysis Reveals Distinct Characteristics of Modulators of Diverse ProcessesMouseNET provides a valuable resource to characterize the systems-level features of a model organism, which is a critical issue in understanding the organization and dynamics of the proteome. In the mouseNET network, the majority of proteins have only a small number of connections (Figure 5A), yet the presence of a few highly connected nodes (Figure 1B) implies central modifiers of the proteome. These ‘hub’ genes (at confidence cutoff 0.6) are enriched in regulation of response to stress, DNA metabolic process and cell cycle, (Bonferroni-corrected p<1.0E-9) (Table 2). Additionally, these hubs were significantly enriched (Bonferroni-corrected p = 8.3E-10) for ‘chromosome organization and biogenesis’, which is in agreement with a previous study in C. elegans that identified a class of genetic interaction hubs, all six of which were chromatin regulators [35].10.1371/journal.pcbi.1000165.g005Figure 5Topological properties of the functional network.(A) The degree (node connectivity) distribution of the integrated functional network (log10 scale) for several different edge probability cutoffs. (B) Connectivity (at 0.6 cutoff in confidence) versus clustering coefficient. The color represents the number of processes represented in that gene's local network (top 40 neighbors). At the same level of connectivity, proteins with smaller clustering coefficients tend to participate in more processes. Local networks centered around Nol1 (C) and Pxn (D). While both genes have roughly equivalent node degree (∼50 confident connections), a potential modulator of multiple pathways (D), however, is differentiated from other hub genes (such as (C)) in that it has a lower clustering coefficient and thus the network centered at Pxn is less densely connected.10.1371/journal.pcbi.1000165.t002Table 2GO SLIM (Biological Process) enrichment of potential modulators of several pathways (Ci<0.15, N≥10) and highly connected genes (N≥10).GO TermGO Term Name\nCi<0.15, N≥10\nN≥10GO:0000003Reproduction1.12E-041.88E-08GO:0016043Cell organization and biogenesis<1.01E-12<1.01E-12GO:0016265Death<1.01E-122.66E-11GO:0006950Response to stress<1.01E-12<1.01E-12GO:0009628Response to abiotic stimulus7.63E-071.15E-09GO:0006259DNA metabolic process2.93E-11<1.01E-12GO:0008283Cell proliferation<1.01E-12<1.01E-12GO:0044238Primary metabolic process<1.01E-12<1.01E-12GO:0015031Protein transport1.22E-06<1.01E-12GO:0006810Transport1.53E-016.32E-06GO:0009605Response to external stimulus3.45E-105.52E-05GO:0009653Anatomical structure morphogenesis<1.01E-127.52E-14GO:0007165Signal transduction<1.01E-125.34E-01GO:0008152Metabolic process<1.01E-12<1.01E-12GO:0030154Cell differentiation<1.01E-12<1.01E-12GO:0050789Regulation of biological process<1.01E-12<1.01E-12GO:0007267Cell-cell signaling1.40E-052.82E-03GO:0007154Cell communication<1.01E-123.27E-01GO:0008219Cell death<1.01E-122.66E-11GO:0006139Nucleobase, nucleoside, nucleotide and Nucleic acid metabolic process1.76E-11<1.01E-12GO:0006996Organelle organization and biogenesis4.00E-08<1.01E-12GO:0009719Response to endogenous stimulus1.60E-09<1.01E-12GO:0006464Protein modification<1.01E-122.29E-08GO:0006350Transcription5.48E-117.69E-09GO:0007275Multicellular organismal development<1.01E-12<1.01E-12GO:0019538Protein metabolic process<1.01E-12<1.01E-12GO:0006412Translation3.12E-021.78E-10GO:0007010Cytoskeleton organization and biogenesis2.83E-062.29E-11GO:0009790Embryonic development<1.01E-121.98E-13GO:0040007Growth1.08E-128.69E-09GO:0007049Cell cycle<1.01E-12<1.01E-12GO:0000003Reproduction0.0001123381.88E-08We further analyzed the topology of the functional network surrounding these hubs and found distinct characteristics that correlate with their role in the cell. Proteins with high connectivity may appear in densely connected modules, or alternatively, they could be linkers of multiple functional modules and participate in several pathways [36]. To investigate these two classes, for each gene we computed the clustering coefficient, C, which gives the probability that its interactors are connected to each other. We found that low clustering coefficients, when controlled for node degree, are critical indicators of proteins participating in more biological pathways (Figure 5B). This trend is robust against different confidence cutoff levels for the interactions (Figure S3). For example, both nucleolar protein 1 (Nol1, MGI:107891) and paxillin (Pxn, MGI:108295) have 50 functional linkages with more than 0.6 confidence in interactions (Figure 5C and 5D). However, the former, which has a C of 0.44, is involved in only the rRNA processing pathway, while the latter, with a C of 0.06, is known to be involved in multiple biological processes, including activation of MAPK activity, branching morphogenesis of a tube, cell adhesion and protein folding. Furthermore, we found that the set of proteins with low clustering coefficients, but not the set of all proteins with only high node degree, is highly enriched for ‘signal transduction’ (Table 2), probably because proteins involved in signal transduction are central to cross-talk among multiple pathways and the cell's diverse response to various stimuli. Thus, the topology of the functional network contains important clues to the global organization of the proteome; and in addition to connectivity, we demonstrate that the clustering coefficient is a critical factor characterizing modifiers of multiple biological pathways.Phenotypic and Disease Effects in Relation to Topology and Functional ParticipationGlobal modeling of functional linkages provides a general framework to analyze the relationship between local network properties and functional consequences of individual gene perturbations. For example, previous studies have predicted that the network connectivity is correlated with the propensity of a protein to be essential [37],[38]. Recently, however, there has been debate over whether this relationship is indeed true in yeast or human [39],[40], the main issue being whether high connectivity is truly a property of the underlying network or simply an effect of intense study of the essential gene set (i.e., annotation or investigational bias).To address this question in the mouse functional network and control for investigation bias, we constructed two networks: one including all input data except knock-out phenotype information, and one including only whole-genome datasets. To avoid the caveat that not all gene knock-outs have been constructed, only genes that have been knocked out or targeted were included in all statistical analyses. For the first functional network, essential genes or disease-associated genes are significantly more connected than average (p<10−18 for perinatal lethality, p<10−9 for postnatal lethality, and p<10−6 for disease-associated genes, Mann-Whitney U test) (Figure S4A). However, in the functional network based on only whole-genome datasets, the difference between essential and non-essential sets was not significant, nor was that between disease-related set and the genome average (Figure 6A), suggesting the observed relationships between essentiality and network connectivity are likely to be explained by investigational biases in our case. This result is consistent with a previous study [41] which suggested that the vast majority of disease genes show no tendency to encode physical interaction hubs in human data. We further considered whether connectivity and local topology in our functional network relate to other perturbation phenotypes. Although most phenotype-responsible gene groups (Table S1) have a higher than average connectivity based on all available input data (Figure S4B), only proteins involved in tumorigenesis, embryogenesis still have significantly higher connectivity than average (p<0.05) on the whole-genome-data-only network (Figure 6B). This result highlights that the variation in intensity of study for genes can cause significant biases in the conclusions reached when comparing the connectivity of different groups of genes.10.1371/journal.pcbi.1000165.g006Figure 6Relationship between phenotypic effects and local network configuration.(A) Comparison of connectivity (at 0.6 confidence) between essential and non-essential genes, and between genes whose orthologous mutants cause disease in human and those with no apparent phenotype. Both comparisons are based on a functional network excluding any phenotypic or disease input data to avoid circularity, and excluding any datasets involving individual investigation results to avoid investigational biases. (B) The average number of functional interactions (at 0.6 confidence) for genes within each phenotypic class. (C) Based on a functional network from integration of all available data, the clustering coefficient is consistently lower for genes having diverse categories of phenotypes; the size of the bubble is proportional to the number of processes represented in nearest neighbors (40 closest proteins). This trend holds true in a network where all individual investigations are excluded, suggesting this trend is not an effect of investigational bias.We observed that all groups of phenotype-associated genes have a lower clustering coefficient than average, and most participate in more biological pathways (Figure 6C). This conclusion holds true when controlling for investigational biases. For example, Trp53, with very high connectivity (Figure 1B) and particularly low clustering coefficient (0.02252), is essential during both embryonic perinatal and postnatal stages and plays a role in tumorigenesis, the reproductive system, and has ten other high level phenotypes (Table S1) according to the Mouse Genome Informatics (MGI) database [18]. This result implies that hubs with low clustering coefficient and participating in multiple pathways are important buffers of the genome, and that mutations or other disruptions of these genes are likely to be related to a detrimental phenotypes and, likely, disease.Comparison of Yeast and Mouse Functional NetworksGenome evolution on the sequence level has been studied intensively during the past decades. Studies of functional evolution on the genome-scale, on the other hand, require comprehensive profiling of proteins, which is difficult due to largely incomplete annotation of protein function in most organisms. Here, we demonstrate that mouseNET is a valuable resource for cross-species functional evolution studies by comparing it to the S. cerevisiae network [2]. To avoid circularity caused by integration of sequence similarity information, we generated a functional network that excludes all orthology-based input data. Given these mouse and yeast networks, we first investigated whether functional linkages are conserved between pairs of orthologs as identified through InParanoid [23]. Our results indicate that high-confidence functional linkages in S. cerevisiae are strongly predictive of functional linkages between orthologous gene pairs in mouse (Figure 7A for statistical analysis).10.1371/journal.pcbi.1000165.g007Figure 7Comparison of yeast and mouse interactome and identification of mouse-specific functional linkages.(A) Distribution of functional relationships in mouse for the corresponding interaction between orthologous genes in yeast. For each graph, the range of edge confidences in the yeast network is labeled below, and relative frequency (y-axis) is plotted against confidence of functional relationships for orthologous pairs in mouse. The p-value (Mann-Whitney U test) for each sub-figure indicates the significance of the difference between the distribution of mouse functional relationships in that bin and relationships in the range of 0.0–0.2 yeast interaction confidence (the first graph). (B) Subgraphs of mouse interactome centered at Rpl15 (MGI:1913730), ribosomal protein L15; Slc27a5 (MGI:1347100): solute carrier family 27 (fatty acid transporter), member 5; Htra1 (MGI:1929076): HtrA serine peptidase 1. (C) To visualize how interactions in mouse were evolutionarily acquired, we adapted a method of collapsing paralogous genes [47] in the yeast interactome. Yeast orthologs of mouse genes in (B) appear at the same positions in (C). The links represent the average weight of the interactions between paralogs.We also investigated the conservation of functional neighborhoods in the mouse and yeast networks. To make the datasets comparable, we included only orthologous pairs in the conservation statistical analysis. We found that the two networks vary from a high degree of conservation to almost no conservation (Figure 7B and 7C). Functional linkages between proteins involved in response to stress, response to endogenous stimulus, catabolic process, DNA metabolism, cell cycle, and other core biological processes and components were highly conserved between yeast and mouse (Table 3), e.g., the ribosomal protein L15 (Rpl15, MGI:1913730; Figure 7B and 7C). In contrast, functional relationships in processes specific to higher organisms, including, behavior, embryonic development, multicellular organismal development and anatomical structure morphogenesis were limited to the mouse network (Table 4). For example, the HtrA serine peptidase 1 (Htra1, MGI:1929076) plays a role in BMP signaling pathway [42], but its ortholog in yeast, YNL123W (Nma111, SGD: S000005067) is involved in apoptosis and lipid metabolic process [43],[44] (Figure 7B and 7C). The newly generated interactions for these mouse-specific functional networks originated through a combination of orthologous pairs in yeast and novel connections with existing genes or genes that have no ortholog in yeast (Figure 7B and 7C). Interestingly, ion transport was among the list of enriched processes for both conserved and unconserved subgraphs. We found that in conserved subgraphs, these genes were enriched in energy-coupled proton transport, which is conserved from yeast to mammals. In contrast, in the unconserved subgraphs, this enrichment of ion transport was due to genes involved in metal-ion or chloride transport, probably because of their involvement in the neural system. Details regarding the enrichment statistics are available in the Dataset S3.10.1371/journal.pcbi.1000165.t003Table 3Conservation between yeast and mouse functional relationships.GO TermGO Term NameBonferroni-Corrected p ValueGO:0006950Response to stress2.27E-09GO:0009719Response to endogenous stimulus2.36E-09GO:0009056Catabolic process1.37E-08GO:0006259DNA metabolic process2.26E-08GO:0007049Cell cycle1.01E-05GO:0006091Generation of precursor metabolites and energy0.00266GO:0006811Ion transport0.00600GO SLIM (Biological Process) enrichment in mouse for genes of conserved interactions (higher than 0.6 confidence of functional relationship in both S. cerevisiae and in mouse) against all orthologous genes.10.1371/journal.pcbi.1000165.t004Table 4Divergence between yeast and mouse functional relationships.GO TermGO Term NameBonferroni-Corrected p ValueGO:0015031Protein transport1.29E-05GO:0006811Ion transport<E-06GO:0005975Carbohydrate metabolic process<E-06GO:0009607Response to biotic stimulus<E-06GO:0006519Amino acid and derivative metabolic process<E-06GO:0009628Response to abiotic stimulus<E-06GO:0006464Protein modification<E-06GO:0007275Multicellular organismal development<E-06GO:0007165Signal transduction<E-06GO:0007610Behavior<E-06GO:0009653Anatomical structure morphogenesis<E-06GO:0050789Regulation of biological process<E-06GO:0007010Cytoskeleton organization and biogenesis<E-06GO:0007154Cell communication6.15E-07GO:0006810Transport1.34E-06GO:0009790Embryonic development3.67E-05GO SLIM (Biological Process) enrichment in mouse for genes whose interactions are not conserved from yeast (higher than 0.6 confidence of functional relationship in S. cerevisiae but less than prior in mouse) against all orthologous genes.Comparative analysis of interactomes between species, such as that presented above, is no doubt a promising approach for answering a number of fundamental biological questions [45]. Previous studies, e.g., [40], have demonstrated the sparsity of our current knowledge of physical interactions in many organisms, which has led to a very limited set of identified conserved interactions. As demonstrated here, the comparison of higher-coverage functional networks based on probabilistic models for integrating diverse genomic data provide an alternative solution for studying the evolution of functional linkages between proteins.Example Application of the MouseNET Web InterfaceGenerating hypotheses for biological functions for a protein of interest based on integrating diverse data sourcesAn important application of the network analysis is to identify, for a protein of interest, which biological processes and pathways it participates in. Here, we use the mouseNET online query system to identify two different biological processes involving Ace (angiotensin I converting enzyme 1, MGI:87874). Ace is currently only annotated to metabolic process (GO:0008125) and proteolysis (GO:0006508) biological process terms in the Gene Ontology. Ace has a well-established central role in blood pressure regulation, evidenced by knock-out phenotypes [46], but it currently lacks annotation to the corresponding GO term. When mouseNET is queried with ‘Ace’, the system indeed suggests that the local network is highly enriched in blood pressure regulation (GO:0008217, p = 8.17E-4), including four proteins annotated directly to this term (Agtr1a, Agtr1b, Ren1, and Agt) (Figure S5A). The functional links between Ace and these four genes cannot be confidently surmised from any single input dataset; instead, they are supported by a combination of data from InParanoid [47], phenotype [48], OMIM [24], SAGE [19], and Zhang [21] expression data, indicating the important role of data integration for suggesting accurate functional role for proteins.In the Ace predicted functional network, we also found enrichment for another unrelated process: menstrual cycle phase (GO:0022601), which currently is synonymous to estrous cycle in mouse GO annotation. Three of the top 40 interactors (Stat5a, Nos3 and Agt) were annotated to this term (p = 3.73E-2), with support from InParanoid [23], phenotype [48], OMIM, SAGE [19], Su [20], and Zhang [21] expression data. Indeed, the expression cycle of Ace shown by immunohistochemistry is correlated with menstrual cycle in human [49], suggesting that mouseNET's prediction of Ace participation in the estrous cycle phase process is likely correct. This annotation is missing from existing annotation databases and such prediction would not be made based on genome scale pair-wise physical interaction studies. Because our system integrates diverse data sources and presents them in a network context, it can quickly allow biology researchers to reveal multiple independent roles of a single gene. mouseNET can thus serve both as a source of functional information for genes that have been previously investigated, but not yet annotated in public databases, as well as a method for directing experiments by hypothesizing novel roles for previously uncharacterized proteins.Identifying disease-related genes through multiple queries of the mouseNET networkBecause genes responsible for the same disease are often involved in related pathways, mouseNET provides a valuable resource for identifying novel disease gene candidates though its multiple-query feature. For example, by searching mouseNET with a set of genes (Mapt, Sncaip, Tbp, Drd4, Ndufv2 and Nr4a2) already known to be involved in Parkinson's disease, we are able to extract other genes annotated to this disease and some novel candidates (Figure S5B). The top three interactors returned by mouseNET (Uchl1, Dbh and Snca) are already labeled with Parkinson's disease in OMIM, indicating the ability of our system to accurately identify other disease genes given some known ones. The fourth gene Msx1 (Homeo box, msh-like 1) is not yet annotated to Parkinson's disease. However, its connection to several query genes (Tbp and Mapt) and to several proteins functionally related to the query set (Mdm2, Fyn, Psen1, Apoe, Uchl1, and Dbh) in mouseNET suggests its potential role in Parkinson's disease. Interestingly, Msx1 was found to act as an intrinsic dopamine-neuron determinant during development, and therefore is very likely to be a candidate involved in Parkinson's disease, which leads to mesencephalic dopamine neuron degeneration. In addition, among the top three interactors, experiment using transgenic mice shows that Uchl1 mutant could lead to dopaminergic neuronal loss [50]; Dbh is a critical gene involved in dopamine biosynthesis; and Snca has been suggested to be an essential regulator of dopamine neurotransmission [51]. Notably, query of Tbp alone results in a list of transcription-related genes that has no significance with the particular disease. The novel candidate Msx1 is only identified with multiple disease gene queries and a network including both direct and indirect neighbors. This illustrates the ability of mouseNET to identify novel candidates of disease genes based on its multiple-query feature, which cannot be achieved by existing databases nor can be readily extracted from any single genome-scale dataset.DiscussionIn this study, we combined diverse genetic and genomic data using a probabilistic framework to generate a functional network for the laboratory mouse. Our network accurately predicts functional linkages between mouse genes and covers a broad range of biological processes. We expect this view of the mouse proteome will be an invaluable resource in identifying novel pathway components and understanding system-level organization.We have demonstrated several applications of our network in this study. First, we characterized the topology of the network and demonstrated that local network topology correlates with biological functions. Also, we used this genomewide view of functional linkages to investigate the relationship between diverse phenotypes and the local configuration of subnetworks. Finally, although network comparison across several species is limited by the sparsity of our current knowledge of physical interactions [40], generation of a functional network based on diverse data types also allowed us to examine the conservation of subnetworks on a global system level.We provide a searchable interface for the exploration of the mouse functional network (http://mouseNET.princeton.edu). The interface also presents a full analysis of the functional enrichment of networks surrounding the genes(s) of interest and the disease genes in the local network. Through our interface, users could identify the original evidence supporting for specific functional linkages. The website includes integration results generated for the purpose of topological studies (controlled for investigational biases) and of cross-species network alignment studies (by excluding homology data) (http://mouseNET.princeton.edu/supplement/supplemental_data.htm). In the future, new publicly available genome-scale data will be added to our system, which will provide up-to-date support for hypothesis generation for questions ranging from individual protein function prediction to characterization of diverse system-level features.In this study, we focused on the generation of a global functional network of mouse and demonstrated its wide applicability. Availability of tissue-specific datasets should allow us to generate tissue, cell, and developmental stage-specific network predictions using similar probabilistic frameworks. These tissue or developmental stage-specific networks will be more targeted and will be invaluable to the researchers of individual fields of study.Materials and MethodsFunctional Genomic Data Retrieval and PreprocessingTo build a functional network of proteins, we have collected a diverse set of evidence from several databases (Table 1). In order to predict pair-wise protein–protein relationships, all data were preprocessed, as described below, into pair-wise scores, reflecting the similarity between protein pairs. The databases included in our analysis are:Physical interaction data from the Biomolecular Interaction Network Database (BIND) [15], the Database of Interacting Proteins (DIP) [17] and the General Repository for Interaction Datasets (GRID) [16]. We also mapped the interactions in the Online Predicted Human Interaction Database (OPHID) [24] to mouse orthologs via InParanoid [23]. In this process, members of the interactions that have more than one ortholog in mouse were mapped for each of their orthologs. Because physical interaction data are pair-wise and binary (representing the presence or absence of evidence for a physical interaction between a pair of proteins), these datasets were in the format of pair-wise binary scores and were ready to be input into the Bayesian network.Phenotype and disease data from MGI [18] and the Online Mendelian Inheritance in Man (OMIM). The disease association data were mapped to mouse using InParanoid [23]. Based on independence analysis (see below), we found that different phenotypes are highly conditionally dependent on each other, and that the phenotype data and disease data are dependent on each other as well. Thus treating phenotype and disease data as separate evidence nodes in a naïve Bayesian network would cause significant over-estimation of functional relationships between gene pairs that affect the same multiple phenotypes/diseases. As a result, phenotype and disease data were treated as a single evidence node in our Bayesian network, of which the score for the protein pair j,k will be:(1)Where ai(j) = 1 if protein j has phenotype i and ai(j) = 0 otherwise, and Ni is the number of proteins involved in this phenotype/disease; n is the total number of phenotypes and diseases. In this way, co-occurrence of rare phenotypes or diseases will be given more weight than common ones. Such calculation allows the transformation from original phenotype/disease profiles to pair-wise scores that reflect the similarity level between a pair of proteins.Homologous functional relationship predictions in yeast from the bioPIXIE system. bioPIXIE is a previously established genomewide prediction of S. cerevisiae functional network, which is based on integration of diverse yeast genome-scale datasets [2]. This integrated dataset was used as an input in our mouse interactome by mapping orthologous genes between S. cerevisiae and laboratory mouse using InParanoid [23]. The average was taken in the case that orthology mapping results in multiple mapped pair-wise scores in yeast for a single pair in mouse.Expression and Tissue localization datasets from Su et al., 2004, Zhang et al., 2004, and the SAGE database [19]. We chose these three datasets because they represent expression profiles of a wide range of tissue and developmental stages. In total, they included 333 conditions. To make the data suitable as an input to our Bayesian network, we applied the Pearson correlation coefficient ρ, to assess levels of co-expression between pairs of genes:(2)Where x and y are expression level data vectors of length n for two genes, x̅ and y̅ are means, and σx and σy are standard deviations. The correlation coefficients were further Fisher z-transfored to ensure comparable, normal distribution [52].Filtering Redundant DatasetsIn the following section, we applied a naïve Bayes network to integrate all data sources and to predict pair-wise functional relationships. However, the application of a naïve Bayesian framework requires a non-trivial assumption of independence between individual evidence sources, which correspond to different evidence nodes in the naïve Bayes network. To address this issue, we evaluated the conditional independence between datasets and those with significant dependence were merged into a single evidence node. To determine whether two datasets should be merged, we calculated the likelihood ratio of each combination of datasets with and without the assumption of independence.(3)\n(4)where E is the score of the protein pair in dataset i or j, a FRY means a positive functional relationship (FR = 1) in gold standard, and FRN means a negative functional relationship (FR = 0).Two conditionally independent datasets will have similar likelihood ratios calculated by the above two approaches (Figure S6A). In contrast, highly dependent datasets tend to have erroneously high likelihood ratios (Figure S6B) when they are treated as independent ones. After a complete analysis of the independence properties between every dataset pair, we found that phenotype data from MGI and disease data from OMIM are highly dependent on each other. As a result, we treated these phenotype and disease data as a single evidence node in the Bayesian network, and each of the remaining datasets as an individual evidence node.Bayesian Network ConstructionAs data sources are different in their accuracy of measurement as well as relevance for predicting protein functions, creating an accurate network for functional linkages requires a systematic approach that weights and integrates information from individual datasets. We applied a Bayesian network to integrate diverse data and make the final functional linkage predictions (Figure 1A). Specifically, we computed the posterior probability of a functional relationship given all available evidence as follows:(5)where FR represents functional relationship, Ei represents the score of the pair in each dataset i and Z is a normalization factor. Intuitively, this probability FRij for two proteins i and j represents how likely it is, given existing data and accuracy and coverage of each input dataset, that proteins i and j participate in the same biological process.To learn the parameters in this Bayesian framework, we established a gold standard that approximates a true set of functionally related proteins. Mouse Genome Informatics (MGI) maintains curated annotations of Gene Ontology (GO) for mouse [53]. The sources of these annotations include (1) hand annotation from primary literature, (2) electronic annotation based on gene name and symbols, (3) annotation from SwissProt keywords, (4) Enzyme Commision (EC) numbers. These annotation sources are reasonably accurate for our analysis. We defined positive as pairs of proteins that are co-annotated to a specific Biological Process GO term (less than two hundred genes annotated to this GO term) and negatives as those in which both members of the pair have specific annotations but do not share any of them.To model the posterior distribution given a set of data, we grouped the pair-wise values from each dataset into discrete groups. For binary datasets, for example, physical interactions, it is easy to separate the two categories where 0 means that there is no interaction between the pair, and 1 means that the interaction exists. Continuous pair-wise scores (e.g., expression profiles and phenotype/disease data) require a binning approach for discretization. We observed that for each dataset, the posteriors generally decreases with small fluctuation as the pair-wise score decreases (Figure S7). Thus, to avoid over-fitting to noise in the datasets, discretization was done so as to force the posteriors of the discretized bins to decrease as the average pair-wise score of those bins decreases.Network-Based Pathway Component PredictionAn important application of such a functional network is to predict novel pathway components. We therefore applied our network to predict pathway components in KEGG [28]. For a specific pathway, during each iteration, 10 known genes were seeded into the weighted network and the rest of the genes were treated as unknowns. Thus for every other gene, we compute an adjacency to the 10 seeds. This process was repeated three hundred times with random samplings of the seed set. We then calculated the average adjacency for each gene:(6)where wi represents the weight of each gene and j represents the seed genes, and wijk represents the confidence, as estimated by our integration, of the functional relationship between protein i and j in iteration k. ni is the number of times gene i was not one of the seed genes. The top components and recovery curves were generated based on the ranking of wi.Topological Characterization of the Functional InteractomeTo characterize the topology of the functional network, we calculated the connectivity and clustering coefficient C of all proteins. The clustering coefficient of a protein gives the probability that its neighbors are connected to each other. In a densely connected module or clique, C is close to one. C for each of the proteins was calculated as follows [54]:(7)where n denotes the number of links between k direct interactors.Functional EnrichmentWe obtained GO annotations [27] from the Mouse Genome Informatics (MGI) [18] on Jan 18, 2007. The enrichment of each GO term was found using a hypergeometric distribution. The most enriched GO terms were represented by the lowest Bonferroni-corrected p value [55].Implementation, Publicly Available Interface, and Network-Based Gene Function PredictionsTo facilitate wide access to the integrated functional network by the biology community, we implemented a web interface (http://mouseNET.princeton.edu) that allows the users to browse our predictions based on single or multiple protein queries. We have implemented a probabilistic algorithm that searches the direct or indirect neighbors with the largest adjacency to the query set [2]. GO term enrichment was calculated for the top neighbors, which facilitates fast discovery of unknown gene function.We also provide the community with a list of gene function predictions based on our network for proteins with no currently known function. Specifically, we calculated the GO term enrichment of the top 40 nearest neighbors of each gene using the hypergeometric distribution. Then the per-function enrichment of each gene's top neighbors is reported as a Bonferroni-corrected p-value and thus their putative function is deduced.Experimental VerificationThe Nanog controllable embryonic stem cell lines were set up and tested by Natalia Ivanova, and were cultured as described [56]. The feeder cells, primary mouse embryonic fibroblasts, were removed before use. To down-regulate Nanog, we withdrew the doxycycline (1 g ml−1) from the media, but still supplied the cells with all the routine ES cell nutrients (DMEM with 15% FBS (Hyclone), 100 mM MEM non-essential amino acids, 0.1 mM 2-mercaptoethanol, 1 mM l-glutamine (Invitrogen), and 103 U ml-1 of LIF (Chemicon). For the nuclear protein measurement, nuclear protein samples were prepared with nuclear/cytosol fractionation kit (BioVision, catalog number: K266-100). The samples from four different time points were labeled by different isotope (iTRAQ) and then analyzed at a single run of mass spectrometry. We used ProQUANT (Applied Biosystems) and the ProGROUP (Applied Biosystems) software to identify proteins. The experiment was repeated three times. Proteins detected more than twice were included in the analysis and the average values were used.Supporting InformationDataset S1Functional composition and biases of each data source and the integrated result.(0.41 MB XLS)Click here for additional data file.Dataset S2Expert curation for a selected set of gene function predictions based on the network.(0.07 MB XLS)Click here for additional data file.Dataset S3Functional biases of conserved and non-conserved sub-network.(0.09 MB XLS)Click here for additional data file.Figure S1The functional composition of the integrated results and individual datasets.(0.14 MB TIF)Click here for additional data file.Figure S2Performance of the integrated interactome in predicting the components of six major pathways in development.(1.57 MB TIF)Click here for additional data file.Figure S3Connectivity (at 0.3 cutoff in confidence) versus clustering coefficient.(0.43 MB TIF)Click here for additional data file.Figure S4Connectivity and phenotypic effects in networks integrated using both individual experimental evidence and large-scale genomic data.(0.67 MB TIF)Click here for additional data file.Figure S5Illustration of the mouseNET interface.(1.20 MB TIF)Click here for additional data file.Figure S6Example of a conditionally independent pair of datasets and a conditionally dependent dataset pair.(0.24 MB TIF)Click here for additional data file.Figure S7The general trend of posteriors for continuous datasets.(0.13 MB TIF)Click here for additional data file.Figure S8The distribution of Log2 changes in protein expression level on the fifth day after Nanog knock-down for 1148 proteins detected in the nucleus.(0.11 MB TIF)Click here for additional data file.Table S1Mapping of phenotypes and MP index in MGI.(0.05 MB DOC)Click here for additional data file.Table S2Literature evidence for novel components (not currently annotated to MAPK in KEGG or GO) predicted to be involved in MAPK pathway.(0.09 MB DOC)Click here for additional data file.Text S1Supplementary figure and tables.(5.07 MB DOC)Click here for additional data file.\n\nREFERENCES:\n1. JiangTKeatingAE\n2005\nAVID: an integrative framework for discovering functional relationships among proteins.\nBMC Bioinformatics\n6\n136\n15929793\n2. MyersCLRobsonDWibleAHibbsMAChiriacC\n2005\nDiscovery of biological networks from diverse functional genomic data.\nGenome Biol\n6\nR114\n16420673\n3. ChenYXuD\n2004\nGlobal protein function annotation through mining genome-scale data in yeast Saccharomyces cerevisiae.\nNucleic Acids Res\n32\n6414\n6424\n15585665\n4. JansenRYuHGreenbaumDKlugerYKroganNJ\n2003\nA Bayesian networks approach for predicting protein-protein interactions from genomic data.\nScience\n302\n449\n453\n14564010\n5. LeeIDateSVAdaiATMarcotteEM\n2004\nA probabilistic functional network of yeast genes.\nScience\n306\n1555\n1558\n15567862\n6. TroyanskayaOGDolinskiKOwenABAltmanRBBotsteinD\n2003\nA Bayesian framework for combining heterogeneous data sources for gene function prediction (in Saccharomyces cerevisiae).\nProc Natl Acad Sci U S A\n100\n8348\n8353\n12826619\n7. RhodesDRTomlinsSAVaramballySMahavisnoVBarretteT\n2005\nProbabilistic model of the human protein-protein interaction network.\nNat Biotechnol\n23\n951\n959\n16082366\n8. XiaKDongDHanJD\n2006\nIntNetDB v1.0: an integrated protein-protein interaction network database generated by a probabilistic model.\nBMC Bioinformatics\n7\n508\n17112386\n9. AlaUPiroRMGrassiEDamascoCSilengoL\n2008\nPrediction of human disease genes by human-mouse conserved coexpression analysis.\nPLoS Comput Biol\n4\ne1000043\ndoi:10.1371/journal.pcbi.1000043\n18369433\n10. StuartJMSegalEKollerDKimSK\n2003\nA gene-coexpression network for global discovery of conserved genetic modules.\nScience\n302\n249\n255\n12934013\n11. NovershternNItzhakiZManorOFriedmanNKaminskiN\n2008\nA functional and regulatory map of asthma.\nAm J Respir Cell Mol Biol\n38\n324\n336\n17921359\n12. SchadtEEMolonyCChudinEHaoKYangX\n2008\nMapping the genetic architecture of gene expression in human liver.\nPLoS Biol\n6\ne107\ndoi:10.1371/journal.pbio.0060107\n18462017\n13. TsaparasPMarino-RamirezLBodenreiderOKooninEVJordanIK\n2006\nGlobal similarity and local divergence in human and mouse gene co-expression networks.\nBMC Evol Biol\n6\n70\n16968540\n14. FrankeLvan BakelHFokkensLde JongEDEgmont-PetersenM\n2006\nReconstruction of a functional human gene network, with an application for prioritizing positional candidate genes.\nAm J Hum Genet\n78\n1011\n1025\n16685651\n15. AlfaranoCAndradeCEAnthonyKBahroosNBajecM\n2005\nThe Biomolecular Interaction Network Database and related tools 2005 update.\nNucleic Acids Res\n33\nD418\nD424\n15608229\n16. BreitkreutzBJStarkCTyersM\n2003\nThe GRID: the General Repository for Interaction Datasets.\nGenome Biol\n4\nR23\n12620108\n17. SalwinskiLMillerCSSmithAJPettitFKBowieJU\n2004\nThe Database of Interacting Proteins: 2004 update.\nNucleic Acids Res\n32\nD449\nD451\n14681454\n18. EppigJTBultCJKadinJARichardsonJEBlakeJA\n2005\nThe Mouse Genome Database (MGD): from genes to mice—a community resource for mouse biology.\nNucleic Acids Res\n33\nD471\nD475\n15608240\n19. SiddiquiASKhattraJDelaneyADZhaoYAstellC\n2005\nA mouse atlas of gene expression: large-scale digital gene-expression profiles from precisely defined developing C57BL/6J mouse tissues and cells.\nProc Natl Acad Sci U S A\n102\n18485\n18490\n16352711\n20. SuAIWiltshireTBatalovSLappHChingKA\n2004\nA gene atlas of the mouse and human protein-encoding transcriptomes.\nProc Natl Acad Sci U S A\n101\n6062\n6067\n15075390\n21. ZhangWMorrisQDChangRShaiOBakowskiMA\n2004\nThe functional landscape of mouse gene expression.\nJ Biol\n3\n21\n15588312\n22. DurinckSMoreauYKasprzykADavisSDe MoorB\n2005\nBioMart and Bioconductor: a powerful link between biological databases and microarray data analysis.\nBioinformatics\n21\n3439\n3440\n16082012\n23. O'BrienKPRemmMSonnhammerEL\n2005\nInparanoid: a comprehensive database of eukaryotic orthologs.\nNucleic Acids Res\n33\nD476\nD480\n15608241\n24. BrownKRJurisicaI\n2005\nOnline predicted human interaction database.\nBioinformatics\n21\n2076\n2082\n15657099\n25. Peña-CastilloLTasanMMyersCLeeHJoshiT\n2008\nA critical assessment of M. musculus gene function prediction using integrated genomic evidence.\nGenome Biol\n9\nSuppl 1\nS2\n26. ZhongWSternbergPW\n2006\nGenome-wide prediction of C. elegans genetic interactions.\nScience\n311\n1481\n1484\n16527984\n27. AshburnerMBallCABlakeJABotsteinDButlerH\n2000\nGene ontology: tool for the unification of biology.\nNat Genet\n25\n25\n29\n10802651\n28. KanehisaMGotoSHattoriMAoki-KinoshitaKFItohM\n2006\nFrom genomics to chemical genomics: new developments in KEGG.\nNucleic Acids Res\n34\nD354\nD357\n16381885\n29. MyersCLBarrettDRHibbsMAHuttenhowerCTroyanskayaOG\n2006\nFinding function: evaluation methods for functional genomic data.\nBMC Genomics\n7\n187\n16869964\n30. QiYNieZLeeYSSinghalNSSchererPE\n2006\nLoss of resistin improves glucose homeostasis in leptin deficiency.\nDiabetes\n55\n3083\n3090\n17065346\n31. ChambersIColbyDRobertsonMNicholsJLeeS\n2003\nFunctional expression cloning of Nanog, a pluripotency sustaining factor in embryonic stem cells.\nCell\n113\n643\n655\n12787505\n32. MitsuiKTokuzawaYItohHSegawaKMurakamiM\n2003\nThe homeoprotein Nanog is required for maintenance of pluripotency in mouse epiblast and ES cells.\nCell\n113\n631\n642\n12787504\n33. BoyerLALeeTIColeMFJohnstoneSELevineSS\n2005\nCore transcriptional regulatory circuitry in human embryonic stem cells.\nCell\n122\n947\n956\n16153702\n34. LohYHWuQChewJLVegaVBZhangW\n2006\nThe Oct4 and Nanog transcription network regulates pluripotency in mouse embryonic stem cells.\nNat Genet\n38\n431\n440\n16518401\n35. LehnerBCrombieCTischlerJFortunatoAFraserAG\n2006\nSystematic mapping of genetic interactions in Caenorhabditis elegans identifies common modifiers of diverse signaling pathways.\nNat Genet\n38\n896\n903\n16845399\n36. SnelBBorkPHuynenMA\n2002\nThe identification of functional modules from the genomic association of genes.\nProc Natl Acad Sci U S A\n99\n5890\n5895\n11983890\n37. LeeILehnerBCrombieCWongWFraserAG\n2008\nA single gene network accurately predicts phenotypic effects of gene perturbation in Caenorhabditis elegans.\nNat Genet\n40\n181\n188\n18223650\n38. JeongHMasonSPBarabasiALOltvaiZN\n2001\nLethality and centrality in protein networks.\nNature\n411\n41\n42\n11333967\n39. CoulombSBauerMBernardDMarsolier-KergoatMC\n2005\nGene essentiality and the topology of protein interaction networks.\nProc Biol Sci\n272\n1721\n1725\n16087428\n40. GandhiTKZhongJMathivananSKarthickLChandrikaKN\n2006\nAnalysis of the human protein interactome and comparison with yeast, worm and fly interaction datasets.\nNat Genet\n38\n285\n293\n16501559\n41. GohKICusickMEValleDChildsBVidalM\n2007\nThe human disease network.\nProc Natl Acad Sci U S A\n104\n8685\n8690\n17502601\n42. OkaCTsujimotoRKajikawaMKoshiba-TakeuchiKInaJ\n2004\nHtrA1 serine protease inhibits signaling mediated by Tgfbeta family proteins.\nDevelopment\n131\n1041\n1053\n14973287\n43. FahrenkrogBSauderUAebiU\n2004\nThe S. cerevisiae HtrA-like protein Nma111p is a nuclear serine protease that mediates yeast apoptosis.\nJ Cell Sci\n117\n115\n126\n14657274\n44. TongFBlackPNBivinsLQuackenbushSCtrnactaV\n2006\nDirect interaction of Saccharomyces cerevisiae Faa1p with the Omi/HtrA protease orthologue Ynm3p alters lipid homeostasis.\nMol Genet Genomics\n275\n330\n343\n16470384\n45. SharanRIdekerT\n2006\nModeling cellular machinery through biological network comparison.\nNat Biotechnol\n24\n427\n433\n16601728\n46. BernsteinKEXiaoHDFrenzelKLiPShenXZ\n2005\nSix truisms concerning ACE and the renin-angiotensin system educed from the genetic analysis of mice.\nCirc Res\n96\n1135\n1144\n15947253\n47. BandyopadhyaySSharanRIdekerT\n2006\nSystematic identification of functional orthologs based on protein network comparison.\nGenome Res\n16\n428\n435\n16510899\n48. EppigJTBlakeJABultCJKadinJARichardsonJE\n2007\nThe mouse genome database (MGD): new features facilitating a model system.\nNucleic Acids Res\n35\nD630\nD637\n17135206\n49. HarataTAndoHIwaseANagasakaTMizutaniS\n2006\nLocalization of angiotensin II, the AT1 receptor, angiotensin-converting enzyme, aminopeptidase A, adipocyte-derived leucine aminopeptidase, and vascular endothelial growth factor in the human ovary throughout the menstrual cycle.\nFertil Steril\n86\n433\n439\n16769060\n50. SetsuieRWangYLMochizukiHOsakaHHayakawaH\n2007\nDopaminergic neuronal loss in transgenic mice expressing the Parkinson's disease-associated UCH-L1 I93M mutant.\nNeurochem Int\n50\n119\n129\n16965839\n51. AbeliovichASchmitzYFarinasIChoi-LundbergDHoWH\n2000\nMice lacking α-synuclein display functional deficits in the nigrostriatal dopamine system.\nNeuron\n25\n239\n252\n10707987\n52. FisherRA\n1915\nFrequency distribution of the values of the correlation coefficient in samples from an indefinitely large population.\nBiometrika\n507\n521\n53. HillDPDavisAPRichardsonJECorradiJPRingwaldM\n2001\nProgram description: strategies for biological annotation of mammalian systems: implementing gene ontologies in mouse genome informatics.\nGenomics\n74\n121\n128\n11374909\n54. WattsDJStrogatzSH\n1998\nCollective dynamics of ‘small-world’ networks.\nNature\n393\n440\n442\n9623998\n55. BoyleEIWengSGollubJJinHBotsteinD\n2004\nGO::TermFinder—open source software for accessing Gene Ontology information and finding significantly enriched Gene Ontology terms associated with a list of genes.\nBioinformatics\n20\n3710\n3715\n15297299\n56. IvanovaNDobrinRLuRKotenkoILevorseJ\n2006\nDissecting self-renewal in stem cells with RNA interference.\nNature\n442\n533\n538\n16767105"
+}

batch_8/PMC2527807.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2527807",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2527807\nAUTHORS: L Li, K Mu, G Zhou, L Lan, G Auer, A Zetterberg\n\nABSTRACT:\nThe role of genomic instability and proliferative activity for development of distant metastases in breast cancer was analysed, and the relative contribution of these two risk factors was quantified. A detailed quantitative comparison was performed between Ki67 and cyclin A as proliferative markers. The frequency of Ki67 and cyclin A-positive cells was scored in the same microscopic areas in 428 breast tumours. The frequency of Ki67-positive cells was found to be highly correlated with the frequency of cyclin A-positive cells, and both proliferation markers were equally good to predict risk of distant metastases. The relative contribution of degree of aneuploidy and proliferative activity as risk markers for developing distant metastases was studied independently. Although increased proliferative activity in general was associated with an increased risk of developing distant metastases, ploidy level was found to be an independent and even stronger marker when considering the group of small (T1) node negative tumours. By combining proliferative activity and ploidy level, a large group of low risk breast tumours (39%) could be identified in which only a few percentage of the tumours (5%) developed distant metastases during the 9-year follow-up time period.\n\nBODY:\nNumerous studies published during the last decades have clearly shown that genomic instability in terms of degree of aneuploidy (D- or A-type, Forsslund and Zetterberg, 1990; Forsslund et al, 1996) and chromosomal rearrangements is closely related to tumour development and tumour progression. Breast tumours of the D-type generally progressed more slowly, on the average four times, and were clinically much less aggressive than their highly aneuploid, genomically unstable counterparts of the A-type (Auer et al, 1980; Cornelisse et al, 1987; Fallenius et al, 1988a; Kronenwett et al, 2006). Chromosomal rearrangements in terms of deletions, duplications and amplifications, as studied by comparative genomic hybridization (CGH), were found to be much more frequent in the highly aneuploid breast tumours than in the diploid ones (Kallioniemi et al, 1994; Ried et al, 1995; Blegen et al, 2001). High-resolution microarray-based CGH data have verified and extended these findings, and identified chromosomal regions and novel specific patterns and degree of rearrangements related to aggressive tumour behaviours (Hicks et al, 2006). Taken together, these data clearly indicate that genomic instability is an important factor for tumour development and progression including distant metastases.In addition to genomic instability, proliferative activity is a general property to be considered in the progression of tumours. Tumour cell proliferation has been widely investigated in breast cancer for its association with neoplastic growth, progression, metastatic potential, survival and response to chemotherapy (van Diest et al, 2004; Colozza et al, 2005; Beresford et al, 2006). Proliferative activity could be assessed through immunohistochemical procedures detecting proliferation-associated antigens, such as Ki67 (Gerdes, 1990), or cell-cycle-specific proteins such as cyclin A (Hunt, 1991; Sherr, 1993; Nurse, 1994). Various studies have shown that a high expression of Ki67 or cyclin A is correlated with a worse prognosis in breast cancer (Bukholm et al, 2001; Kuhling et al, 2003; Poikonen et al, 2005; Baldini et al, 2006; Ahlin et al, 2007; de Azambuja et al, 2007; Railo et al, 2007). However, evidence has also been obtained that the prognostic value of proliferation markers varies significantly depending on clinical characteristics of the tumour disease, for example, lymph node status (Jalava et al, 2006; Trere et al, 2006). Thus, to obtain more detailed information regarding the prognostic contribution of proliferation markers in breast cancer, patients have to be subgrouped according to clinical features, for example, tumour size and lymph node status.The specific aim of this study was to investigate and compare in the same individual tumours, the relative influence of genomic instability and proliferative activity as risk factors for development of distant metastases in breast cancer. As one aspect of genomic instability, the degree of aneuploidy was quantified, and the tumours were separated in the two groups (A or D) with respect to ploidy level. The proliferative activity was analysed by immunohistochemistry using antibodies against Ki67 and cyclin A. An important methodological aspect of this paper was the direct quantitative comparison performed between the Ki67 analysis and the cyclin A analysis in the same tumour areas. By combining proliferative activity and ploidy level, a relatively large group of low-risk breast tumours (39%) could be identified in which only a few percentage of the tumours (5%) developed distant metastases during the 9-year follow-up time period. In the remaining 61% of the breast cancers, 35% developed distant metastases during the same follow-up time period.Materials and methodsTumour samplesThis study was based on the data of 428 patients with breast cancer analysed at the department of Oncology–Pathology, Karolinska University Hospital, Solna, at the time of diagnosis (1997–1998). All histological specimens were routinely Ki67- and cyclin A-stained. In the 428 cases, 378 patients available with clinical data were followed up from diagnosis until death or survivors for at least 9 years. All tumours were classified according to the World Health Organization (1981) and graded on the basis of the recommendations of Elston and Ellis (1991). Permission to analyse the samples and correlate the results to patient data was obtained by the Ethical Committee Nord, Karolinska Institutet (Dnr 00-186). The tumour samples were fixed in 4% phosphate-buffered formaldehyde directly after operation and paraffin-embedded. From each specimen, 10 contiguous sections were prepared and used for HE staining and immunohistochemistry (thickness 4 μm).ImmunohistochemistryThe sections were deparaffinized with xylene, rehydrated through a graded alcohol series and microwaved at 500 W for 2 × 5 min in 10 mM citrate buffer (pH 6.0). After rinsing in Tris-buffered saline (TBS, pH 7.6), sections were treated with 3% hydrogen peroxide in methanol to exhaust endogenous peroxidase activity followed by normal horse serum (1 : 20 dilution) in 0.1 M PBS (pH 6.0), and then incubated overnight with the monoclonal primary antibodies diluted in 1% (wt/vol) bovine serum albumin and visualized by standard avidin–biotin–peroxidase complex technique (Vector Laboratories, Burlingame, CA, USA). Counterstaining was performed with Mayer's haematoxylin. The antibodies used were as follows: MIB-1 (antibody against the nuclear proliferation-associated antigen Ki67, Immunotech SA, Marseille, France), dilution 1 : 150; NCL-cyclin A (Cyclin A monoclonal antibody, Novocastra Laboratories Ltd, Claremont Place, Newcastle upon Tyne, UK), dilution 1 : 100.Evaluation of immunoreactivity scoresBy comparison with the haematoxylin-and-eosin-stained sections, images of the same morphology areas expressing Ki67 and cyclin A were taken by digital camera in at least 5–14 high-power fields (10 × 40 magnification). The percentage of positive cells was measured by two experienced pathologists blinded to each other. A minimum of 1000 tumour cells were counted. Only distinct nuclear staining was accepted as a positive reaction for both markers, whereas all cells with simultaneous nuclear and cytoplasmic cyclin A staining were regarded as positive for cyclin A.Image cytometryNuclear DNA was measured by image cytometry on Feulgen-stained imprints as previously described (Auer et al, 1980). DNA histograms were interpreted according to a modified subjective method. The normal control cells were given the value 2c, denoting the normal diploid DNA content, and all tumour-cell DNA values were expressed in relation to that. The histograms were divided into two groups. Cases with a major peak near the 2c region (1.8c–2.2c), and <10% cells exceeding 2.2c were denoted diploid. DNA profiles with a stem line outside the diploid and tetraploid region and distinctly scattered DNA values exceeding the tetraploid region (3.8c–4.2c) were classified as aneuploid. Furthermore, the S-phase fraction (SPF) was measured on the basis of the DNA distribution patterns (Falkmer et al, 1990).Statistical analysisStatistical analyses were performed using the SPSS for Windows version 11. The correlation between cyclin A, Ki67 and SPF were evaluated by Spearman's rank correlation test and the linear correlation test. Fisher's exact test was used to compare the difference between non-continuous variable. Cut-off points of Ki67 and cyclin A in patients with distant metastases were calculated by ROC curves quantitative analysis, and contribution of the risk factors to distant metastases was determined by multivariate analysis with logistic regression. P-value<0.05 was considered to be statistically significant.ResultsTo obtain accurate information about proliferative activity, two independent markers Ki67 and cyclin A were used, and a direct quantitative comparison between these two markers was performed. An important methodological aspect of the approach used in this paper is that the Ki67 and the cyclin A analyses were carried out on identical microscopic areas (5–14 areas in each corresponding tumour) of all of the 428 tumours. This gives particular strength to the accuracy of the quantitative data obtained on proliferative activity.Figure 1 illustrates immunostaining of two tumours, one slowly proliferating near-diploid, D-tumour (Figure 1A and C) and one rapidly proliferating clearly aneuploid, A-tumour (Figure 1B and D). The number of Ki67-positive cells is low (4%) in the D-tumour (Figure 1A) and high (40%) in the A-tumour (Figure 1B). A corresponding result is seen in the same microscopic areas of the tumours stained for cyclin A (2 and 20%, respectively; compare Figure 1A and C and Figure 1B and D). Image cytophotometric S-phase analysis of Feulgen-DNA-stained cell nuclei in the near-diploid D-tumour (Figure 1A and C) showed about 1% cells in S-phase in contrast to 15% in the A-tumour (Figure 1B and D).Figure 2 shows the direct quantitative relationship between Ki67 and cyclin A as proliferative markers. In Figure 2A, the percentage of Ki67-positive cells is plotted against the percentage of cyclin A-positive cells counted in the same 5–14 randomly selected microscopic fields in each of four different tumours exhibiting low, intermediate and high proliferative activity (Figure 2A). The percentage of Ki67-positive cells was highly correlated (correlation coefficient 0.88) with the percentage of cyclin A-positive cells, when considering the same individual microscopic field in each of the four tumours (Figure 2A). A large variation in proliferative activity, in most cases two- to five-fold, was observed between the different microscopic fields in each tumour. This emphasises the importance of counting several different microscopic fields in each tumour to get reliable quantitative information about proliferative activity. When the analysis was performed in such a way on a set of 428 tumours, a very high correlation (correlation coefficient 0.90) was found between percentage of Ki67-positive cells and percentage of cyclin A-positive cells (Figure 2B). The data presented in Figure 2B represent the average of 5–14 randomly chosen microscopic fields in each tumour.A comparison between D-tumours and A-tumours was performed on 375 of the 428 tumours (Figure 2C and D). A similar correlation between Ki67 and cyclin A was found in tumours of both types, a correlation coefficient of 0.90 for the D-tumours and 0.86 for the A-tumours. When comparing D- and A-tumours with respect to proliferative activity, two features could be seen. On average, the proliferative activity was twice as high in A-tumours (median value 21% Ki67-positive cells) as in D-tumours (median value 11%). However, a very large overlap in proliferative activity was found between these two groups. When using a cut-off value of 15% for Ki67-positive cells (see further Materials and Methods), the majority of the A-tumours (76%) showed high proliferative activity, whereas the majority of the D-tumours (61%) showed low proliferative activity.Image cytophotometric S-phase determination was performed on the 375 ploidy analysed tumours. Although only about 100 cells were analysed for each tumour, which is enough for the accurate ploidy determination into D- or A-type, but probably insufficient for accurate S-phase analysis, some information about SPF could be obtained. Thus a lower, but still relatively good, correlation was found between Ki67 and SPF (correlation coefficient 0.61) and cyclin A and SPF (correlation coefficient 0.65).A clearly increased risk of developing distant metastases is seen when the proliferative activity increases (Figure 3). The same result was obtained when either Ki67 or cyclin A was used as a proliferative marker (Figure 3A and B). Among the tumours with low proliferative activity, using a cut-off value of 15% for Ki67-positive cells and 8% for cyclin A-positive cells (see further Materials and Methods and Ahlin et al, 2007), between 5 and 10% of all tumours had developed distant metastases during the 9-year follow-up time (Figure 3A and B). Among the tumours with high proliferative activity, between 25 and 30% of all tumours had developed distant metastases during the same time period (Figure 3A and B). However, when taking node status and tumour size into account, the role of proliferative activity as a risk factor becomes more evident (Figure 4A–D). In node-negative (N0) tumours smaller than 20 mm (T1) in which the proliferative activity was low, only 2–3% of the tumours developed distant metastases during the 9-year follow-up time (Figure 4A and C). For node-positive (N+), T1 tumours with low proliferative activity, the risk of developing distant metastases had increased slightly to between 5 and 7% (Figure 4B and D). In tumours equal to or larger than 20 mm (T2), the risk of developing distant metastasis had increased somewhat further in the node-negative (N0) tumour group to between 8 and 12% (Figure 4A and C), but substantially in the node-positive (N+) tumour group to between 30 and 40% (Figure 4C and D). In tumours of high proliferative activity, the risk of developing distant metastases was considerably increased in all tumour groups. For N0 tumours, the risk was between 20 and 25% (Figure 4A and C), and in N+ tumours it was between 35 and 55% (Figure 4B and D).An increased risk of distant metastases was also seen in the highly aneuploid A-tumours as compared with the near-diploid D-tumours (Figure 4). However, this increased risk could only be demonstrated in small (T1), node-negative (N0) tumours, among which the D-tumours exhibited a risk as low as 2–3% and the A-tumours as high as 20–25% of developing distant metastases (Figure 4E). For the remaining tumour groups T2N0, T1N+ and T2N+, no significant difference between D- and A-tumours could be demonstrated (Figure 4E and F).The relative contributions of genomic instability and proliferative activity as risk factors for distant metastases became very evident when the tumours were divided into four groups, D-tumours with low proliferative activity (D low), D-tumours with high proliferative activity (D high), A-tumours with low proliferative activity (A low) and A-tumours with high proliferative activity (A high), and also taking node status and tumour size into account at the same time (Table 1 and Figure 5A). In the group T1N0, all D-tumours, independently of proliferative activity, showed a very low risk (less than 3%) of developing distant metastases (Table 1 and Figure 5A). For the A-tumours in the group T1N0, the situation was different. A-tumours with low proliferative activity, however, showed the same low risk of developing distant metastases as the D-tumours, but for the A-tumours with high proliferative activity, the risk of developing distant metastases was found to be high (close to 30%) and significantly increased (P<0.01) over that of the D-tumours with high proliferative activity. This means that the A-tumours with high proliferative activity metastasize very early, contrary to D-tumours with high proliferative activity, and that the genomic instability associated with the ploidy type A adds prognostic information in addition to that of proliferative activity per se. For large (T2) node-negative (N0) tumours, or small (T1) node-positive (N+) tumours, the risk of distant metastases was only moderately increased for tumours with low proliferative activity, 0–12% for the D-tumours and 13–17% for the A-tumours (Table 1), again indicating that ploidy gives information in addition to that of proliferative activity. Large (T2) node-positive (N+) tumours were at high risk of metastasis relatively independent of ploidy type or proliferative activity (Figure 5C and Table 1).By combining ploidy type (D or A), proliferative activity (low or high), node-status (N0 or N+) and tumour size (T1 or T2), breast cancers could be divided into two risk groups with respect to development of distant metastases (Figure 5D). For the low-risk group, consisting of all D-tumours in the T1N0 group together with the A-tumours with low proliferative activity in the same group plus all tumours with low proliferative activity in the T2N0 and T1N+ groups, the risk of developing distant metastases was less than 6%. For the high-risk group, consisting of all large (T2) node-positive (N+) tumours and all tumours with high proliferative activity, except the D-tumours in the T1N0 group, the risk of distant metastases was found to be in the range 30–60% (Figure 5C) and around 35% in average (Figure 5D). The low-risk group defined in this way is relatively large and constitutes about 40% of all women with stage T1 and T2 breast cancer (Figure 5B). By defining the low-risk group in a more stringent way, only taking into account all the D-tumours in the T1N0 group plus the A-tumours with low proliferative activity in the same group, the low-risk group is still relatively large and now constitutes around 25% of all women with breast cancer. The risk of developing distant metastases in this more stringently defined group is now as low as around 2% (Table 1). Multivariate analysis showed that proliferation (both Ki67 and cyclin A) was the most critical factor in N+ tumours (P<0.05) and ploidy was the most critical one in N0 tumours (P<0.05).DiscussionIn this study, it was found that both genomic instability and proliferative activity can influence risk of developing distant metastases independently of each other, and that they have different relative impacts depending on tumour size and node status. One important methodological aspect of the approach used in this paper is that the Ki67 and the cyclin A analyses were carried out on identical microscopic areas of each tumour. This is particularly important when considering the large variations in proliferative activity that was found to exist between different areas in the tumours. In addition to obtaining accurate information about proliferative activity by using two independent markers analysed in several identical microscopic areas of each tumour, a direct quantitative comparison between the use of Ki67 and cyclin A could also be performed in an accurate way. A second important methodological aspect of this paper is that proliferative activity and genomic instability could be studied simultaneously as independent risk factors in the same individual tumour. This gives particular strength to the quantitative approach of this study.The frequency of cyclin A-positive cells was highly correlated to the frequency of Ki67 cells, both in the analysis of each individual microscopic area and in the whole set of the 428 tumours. This clearly shows that both Ki67 and cyclin A are equally reliable as markers of proliferative activity. The average frequency value of the cyclin A-stained cells was found to be about 0.4 times that of the corresponding Ki67 values. As Ki67 is expressed throughout most of the cell cycle (Gerdes, 1990), and the expression of cyclin A is restricted to the S- and G2-phases in both normal and tumour cells (Erlandsson et al, 2000), the data indicate that in breast tumours the S- and G2-phases on average occupies about 40% of the whole cell cycle as defined by Ki67.A lower but still relatively good correlation was also found between Ki67 and SPF and between cyclin A and SPF. The true correlation is likely to be much higher taking into account the large variation in proliferative activity seen between different microscopic fields and the fact that the SPF was obtained from cytophotometric measurements of only about 100 cells from just a few microscopic fields in each tumour. Another factor that tends to lower the correlation between SPF and Ki67 or cyclin A is the overrepresentation of normal cells measured in the S-phase analysis. This risk is, however, relatively low when SPF is based on image cytometric analysis, in which only morphologically identified cells are measured. For flow cytometric analysis, this error in determining the SPF can be substantial, particularly in the near-diploid tumours where all cells in the tumour are measured without any possibility to make a discrimination between normal and tumour cells on the basis of DNA content. For the clearly aneuploid tumours, the admixture of normal cells in the tissue sample can to some extent be estimated from the co-existence of cells with diploid DNA values. However, a problem with the clearly aneuploid tumours on the other hand is that some of the cells with DNA values in the S-phase region may in fact represent non-proliferating or growth-arrested cells with numerical chromosomal aberrations. It is thus of decisive importance to be aware of these difficulties when calculating SPFs from flow cytometry data (Falkmer et al, 1990). In spite of these methodological deficiencies, it has been shown that high SPF is a property that is related to poor prognosis for many tumours (Kallioniemi et al, 1987).In contrast to the interpretation of DNA histograms based on flow-cytometric measurements, image cytometry based on DNA histograms were preferentially interpreted by subdividing the histograms in diploid or pseudo-diploid (D-tumours) and highly aneuploid types (A-tumours) (Zetterberg and Esposti, 1980; Cornelisse et al, 1987; Fallenius et al, 1988a, 1988b; Forsslund and Zetterberg, 1990; Forsslund et al, 1996). The large difference in the extent of aneuploidy that exists between the highly aneuploid A-tumours on one hand and the pseudo-diploid D-tumours on the other is most likely generated by two principally different mechanisms. Evidence has been obtained that a defective coordination between the centrosome cycle and the DNA replication cycle leading to abnormal, often tripolar, mitoses and highly unequal segregation of chromosomes between daughter cells is crucial for the generation of high-degree aneuploidy seen in the A-tumours (Kronenwett et al, 2005). The low-degree aneuploidy seen in the pseudo-diploid D-tumours on the other hand is more likely to be generated by some defect primarily in the chromatid segregation mechanism at mitosis.One of the main points of this paper was to quantitate the role of genomic instability on one hand and proliferative activity on the other as independent risk factors for the development of distant metastases. The relative contributions of each of these two risk factors became clearly evident first when the tumours were divided into different subgroups with respect to ploidy type and proliferative activity (D low, D high, A low and A high), and also taking node status and tumour size into account at the same time. Although increased proliferative activity in general was associated with an increased risk of developing distant metastases, multivariate analysis showed that ploidy level was an independent and even stronger marker when considering small (T1) node-negative (N0) tumours. In this group (T1N0), all D-tumours, independently of proliferative activity, showed a very low risk of developing distant metastases. This was also found for grade 1 tumours, although histological grade in general was found to have less impact as risk marker in comparison with proliferative activity and ploidy. For the A-tumours in the same group (T1N0), the situation was different. A-tumours with low proliferative activity showed about the same low risk of developing distant metastasis as the D-tumours, but for the group of A-tumours with high proliferative activity, the risk of developing distant metastases was found to be high and significantly increased over that of the D-tumours with high proliferative activity. This means that the A-tumours with high proliferative activity metastasise very early, contrary to D-tumours with high proliferative activity. Thus the genomic instability associated with the ploidy type A adds prognostic information that is independent of the proliferative activity per se. Genomic instability is therefore likely to be a crucial property in the process of forming distant metastases, and aneuploidy per se as one mechanism of generating gene copy number imbalances in the tumour is one important aspect of genomic instability. This is further supported by our previous findings that chromosomal rearrangements in terms of deletions, duplications and amplifications, as studied by CGH, were much more frequent in the highly aneuploid breast tumours than in the pseudo-diploid ones (Blegen et al, 2001). With high-resolution microarray-based CGH, we could furthermore show that overall the degree of chromosomal rearrangements as well as specific patterns of rearrangements were both related to aggressive tumour behaviour (Hicks et al, 2006). Collectively, these data clearly indicate that genomic instability is an important factor for tumour progression and metastasis.The possibility to classify breast tumours accurately into risk groups on the basis of genomic instability together with information about proliferation is an implication of practical clinical relevance of this study. It is clear from the findings in this paper that by combining ploidy type, proliferative activity and tumour stage, a large group of low-risk breast tumours could be identified, in which only a few percentage of the tumours developed distant metastases. It is obvious that it would have great practical implications when selecting patients for different therapy regimes if risk of distant metastases could be predicted objectively with high degree of precision.\n\nREFERENCES:\n1. Ahlin C, Aaltonen K, Amini R-M, Nevanlinna H, Fjällskog M-L, Blomqvis C (2007) Ki67 and cyclin A as prognostic factors in early breast cancer. What are the optimal cut-off values? Histopathology\n51(4): 491–49817711446\n2. Auer GU, Caspersson TO, Wallgren AS (1980) DNA content and survival in mammary carcinoma. Anal Quant Cytol\n2: 161–1656252802\n3. Baldini E, Camerini A, Sgambato A, Prochilo T, Capodanno A, Pasqualetti F, Orlandini C, Resta L, Bevilacqua G, Collecchi P (2006) Cyclin A and E2F1 overexpression correlate with reduced disease-free survival in node-negative breast cancer patients. Anticancer Res\n26: 4415–442117201163\n4. Beresford MJ, Wilson GD, Makris A (2006) Measuring proliferation in breast cancer: practicalities and applications. Breast Cancer Res\n8: 21617164010\n5. Blegen H, Ghadimi BM, Jauho A, Zetterberg A, Eriksson E, Auer G, Ried T (2001) Genetic instability promotes the acquisition of chromosomal imbalances in T1b and T1c breast adeno-carcinomas. Anal Cell Pathol\n22: 123–13111455031\n6. Bukholm IR, Bukholm G, Nesland JM (2001) Over-expression of cyclin A is highly associated with early relapse and reduced survival in patients with primary breast carcinomas. Int J Cancer\n93: 283–28711410878\n7. Colozza M, Azambuja E, Cardoso F, Sotiriou C, Larsimont D, Piccart MJ (2005) Proliferative markers as prognostic and predictive tools in early breast cancer: where are we now? Ann Oncol\n16: 1723–173915980158\n8. Cornelisse CJ, van de Velde CJ, Caspers RJ, Moolenaar AJ, Hermans J (1987) DNA ploidy and survival in breast cancer patients. Cytometry\n8: 225–2343582068\n9. de Azambuja E, Cardoso F, de Castro Jr G, Colozza M, Mano MS, Durbecq V, Sotiriou C, Larsimont D, Piccart-Gebhart MJ, Paesmans M (2007) Ki-67 as prognostic marker in early breast cancer: a meta-analysis of published studies involving 12 155 patients. Br J Cancer\n96: 1504–151317453008\n10. Elston CW, Ellis JO (1991) Pathological prognostic factors in breast cancer: experience from a long study with long term follow up. Histopathology\n19: 403–4101757079\n11. Erlandsson F, Linnman C, Ekholm S, Bengtsson E, Zetterberg A (2000) A detailed analysis of cyclin A accumulation at the G(1)/S border in normal and transformed cells. Exp Cell Res\n259: 86–9510942581\n12. Falkmer UG, Hagmar T, Auer G (1990) Efficacy of combined image and flow cytometric DNA assessments in human breast cancer: a methodological study based on a routine histopathological material of 2024 excised tumour specimens. Anal Cell Pathol\n2: 297–3122275876\n13. Fallenius AG, Auer GU, Carstensen JM (1988a) Prognostic significance of DNA measurements in 409 consecutive breast cancer patients. Cancer\n62: 331–3413383134\n14. Fallenius AG, Franzen SA, Auer GU (1988b) Predictive value of nuclear DNA content in breast cancer in relation to clinical and morphologic factors. A retrospective study of 227 consecutive cases. Cancer\n62: 521–5303390793\n15. Forsslund G, Nilsson B, Zetterberg A (1996) Near tetraploid prostate carcinoma. Methodological and prognostic aspects. Cancer\n78: 1748–17558859188\n16. Forsslund G, Zetterberg A (1990) Ploidy level determinationsin high-grade and low-grade malignant variants of prostatic carcinoma. Cancer Res\n50: 4281–42851694718\n17. Gerdes J (1990) Seminars in Cancer Biology, Vol. 1, pp 99–206. Saunders Scientific Publications: London, New York\n18. Hicks J, Krasnitz A, Lakshmi B, Navin NE, Riggs M, Leibu E, Esposito D, Alexander J, Troge J, Grubor V, Yoon S, Wigler M, Ye K, Borresen-Dale AL, Naume B, Schlicting E, Norton L, Hagerstrom T, Skoog L, Auer G, Maner S, Lundin P, Zetterberg A (2006) Novel patterns of genome rearrangement and their association with survival in breast cancer. Genome Res\n16: 1465–147917142309\n19. Jalava P, Kuopio T, Juntti-Patinen L, Kotkansalo T, Kronqvist P, Collan Y (2006) Ki67 immunohistochemistry: a valuable marker in prognostication but with a risk of misclassification: proliferation subgroups formed based on Ki67 immunoreactivity and standardized mitotic index. Histopathology\n48: 674–68216681683\n20. Hunt T (1991) Cyclins and their partners: from a simple idea to complicated reality. Semin Cell Biol\n2: 213–2221842340\n21. Kallioniemi A, Kallioniemi OP, Piper J, Tanner M, Stokke T, Chen L, Smith HS, Pinkel D, Gray JW, Waldman FM (1994) Detection and mapping of amplified DNA sequences in breast cancer by comparative genomic hybridization. Proc Natl Acad Sci USA\n91: 2156–21608134364\n22. Kallioniemi OP, Hietanen T, Mattila J, Lehtinen M, Lauslahti K, Koivula T (1987) Aneuploid DNA content and high S-phase fraction of tumour cells are related to poor prognosis in patients with primary breast cancer. Eur J Cancer Clin Oncol\n23: 277–2823595689\n23. Kronenwett U, Huwendiek S, Castro J, Ried T, Auer G (2005) Characterisation of breast fine-needle aspiration biopsies by centrosome aberrations and genomic instability. Br J Cancer\n92: 389–39515558069\n24. Kronenwett U, Ploner A, Zetterberg A, Bergh J, Hall P, Auer G, Pawitan Y (2006) Genomic instability and prognosis in breast carcinomas. Cancer Epidemiol Biomarkers Prev\n15: 1630–163516985023\n25. Kuhling H, Alm P, Olsson H, Ferno M, Baldetorp B, Parwaresch R, Rudolph P (2003) Expression of cyclins E, A, and B, and prognosis in lymph node-negative breast cancer. J Pathol\n199: 424–43112635132\n26. Nurse P (1994) Ordering S phase and M phase in the cell cycle. Cell\n79: 547–5507954820\n27. Poikonen P, Sjostrom J, Amini RM, Villman K, Ahlgren J, Blomqvist C (2005) Cyclin A as a marker for prognosis and chemotherapy response in advanced breast cancer. Br J Cancer\n93: 515–51916091759\n28. Railo M, Lundin J, Haglund C, von Smitten K, Nordling S (2007) Ki-67, p53, ER receptors, ploidy and S phase as long-term prognostic factors in T1 node-negative breast cancer. Tumour Biol\n28: 45–5117143016\n29. Ried T, Just KE, Holtgreve-Grez H, du Manoir S, Speicher MR, Schrock E, Latham C, Blegen H, Zetterberg A, Cremer T, Thomas C, Gert A (1995) Comparative genomic hybridization of formalin-fixed, paraffin-embedded breast tumors reveals different patterns of chromosomal gains and losses in fibroadenomas and diploid and aneuploid carcinomas. Cancer Res\n55: 5415–54237585611\n30. Sherr CJ (1993) Mammalian G1 cyclins. Cell\n73: 1059–10658513492\n31. Trere D, Ceccarelli C, Migaldi M, Santini D, Taffurelli M, Tosti E, Chieco P, Derenzini M (2006) Cell proliferation in breast cancer is a major determinant of clinical outcome in node-positive but not in node-negative patients. Appl Immunohistochem Mol Morphol\n14: 314–32316932023\n32. van Diest PJ, van der Wall E, Baak JP (2004) Prognostic value of proliferation in invasive breast cancer: a review. J Clin Pathol\n57: 675–68115220356\n33. Zetterberg A, Esposti PL (1980) Prognostic significance of nuclear DNA levels in prostatic carcinoma. Scand J Urol Nephrol Suppl\n55: 53–586938036"
+}

batch_8/PMC2527828.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2527828",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2527828\nAUTHORS: K Saeb-Parsy, A Veerakumarasivam, M J Wallard, N Thorne, Y Kawano, G Murphy, D E Neal, I G Mills, J D Kelly\n\nABSTRACT:\nMembrane type-1 matrix metalloproteinase (MT1-MMP) is a zinc-binding endopeptidase, which plays a crucial role in tumour growth, invasion and metastasis. We have shown previously that MT1-MMP has higher expression levels in the human urothelial cell carcinoma (UCC) tissue. We show here that siRNA against MT1-MMP blocks invasion in UCC cell lines. Invasion is also blocked by broad-spectrum protease and MMP inhibitors including tissue inhibitor of metalloproteinase-1 and -2. Membrane type-1-MMP can also regulate transcription. We have used expression arrays to identify genes that are differentially transcribed when siRNA is used to suppress MT1-MMP expression. Upon MT1-MMP knockdown, Dickkopf-3 (DKK3) expression was highly upregulated. The stability of DKK3 mRNA was unaffected under these conditions, suggesting transcriptional regulation of DKK3 by MT1-MMP. Dickkopf-3 has been previously shown to inhibit invasion. We confirm that the overexpression of DKK3 leads to decreased invasive potential as well as delayed wound healing. We show for the first time that the effects of MT1-MMP on cell invasion are mediated in part through changes in DKK3 gene transcription.\n\nBODY:\nMatrix metalloproteinases (MMPs) are a group of zinc-dependent endopeptidases that regulate the pericellular microenvironment and play an important role in development and physiological regulation of the extracellular environment (Woessner, 1991; Nagase et al, 2006). Membrane type-1 MMP (MT1-MMP) was the first member of the membrane-bound MMPs to be identified, and it plays a crucial role in tumour growth, angiogenesis, invasion and, ultimately, development of metastasis (Sato et al, 1994; Zhou et al, 2000; Sabeh et al, 2004; Itoh and Seiki, 2006).Membrane type-1-MMP enhances degradation of collagen IV, a major component of the basement membrane, by forming a complex with tissue inhibitor of metalloproteinase-2 (TIMP-2) to activate pro-MMP-2 (Seiki, 2003). The process of substrate activation is a well-characterised function of MT1-MMP, and other substrates requiring cleavage and shedding for activation, including MMP13, CD44 and α-v-β-3, have been identified (Knauper et al, 1996; Kajita et al, 2001; Beauvais and Rapraeger, 2003, 2004; Endo et al, 2003; Nagano and Saya, 2004). Recently the functional role of MT1-MMP has been extended beyond sheddase activity; MT1-MMP induces VEGFA and Smad1 and the suggested mechanism is transcriptional regulation (Sounni et al, 2004; Freudenberg and Chen, 2007). Induction of VEGFA appears to be through the activation of src kinase pathway, and the presence of catalytic and cytoplasmic domains are crucial in this process, which suggests that MT1-MMP is a transcriptional regulator of multiple targets (Sounni et al, 2004).Bladder cancer or urothelial cell carcinoma (UCC) is the fifth common malignancy in the United Kingdom (Cancer Research UK (CRUK) cancer statistic 2003), and it is anticipated that in 2007 over 67 160 new cases would be diagnosed in the United States of America (Jemal et al, 2007). We have previously demonstrated that of the MMPs, MT1-MMP is highly expressed in UCC, localising to both the epithelial and stromal compartments and associated with stage and grade progression (Wallard et al, 2006). Expression of MT1-MMP is highest in tumours invading the lamina propria (pT1) and detrusor muscle (>pT2). This expression is in keeping with the propensity for up to 20% of pT1 tumours to progress to muscle invasive disease, and for up to 50% of muscle invasive tumours to metastasis, within 2 years (Torti and Lum, 1984).In this study, we explored the potential downstream events following targeted manipulation of MT1-MMP in UCC cell lines. We employed expression microarray to highlight changes in the expression of genes following suppression of MT1-MMP. Among candidate genes, we identified Dickkopf-3 (DKK3) as a putative target, which appears to be transcriptionally suppressed by MT1-MMP and by itself is a potent regulator of cell invasion. Dickkopf-3, also known as reduced expression in immortalised cells (REIC), is a divergent member of a group of four secreted proteins, and emerging evidence suggests that it functions as a tumour suppressor to inhibit cell growth and motility (Hsieh et al, 2004; Kawano et al, 2006). Dickkopf-3 may interact to inhibit wnt-signalling pathways (Hoang et al, 2004), and its loss in UCC and other cancers corresponds to a tumour suppressor effect (Nozaki et al, 2001; Kurose et al, 2004; Kawano et al, 2006; Urakami et al, 2006). In support of an extended role for MT1-MMP, the results of this study offer further evidence of interaction with other targets, specifically with negative regulation of the antimigratory tumour suppressor gene DKK3 in UCC.Materials and methodsCell lines and cultureUrothelial cell carcinoma cell lines RT112, 253JBV and EJ28 were sourced from CRUK. UMUC3 and HT137 were sourced from the American Type Culture Collection. UMUC3, 253JBV and 253JBV were grown in RPMI 1640 supplemented with 10% foetal bovine serum. The remainder of the cell lines were grown in DMEM supplemented with 10% foetal bovine serum. Cells were maintained at 37°C in a humidified incubator containing 95% air and 5% CO2.ConstructspCS2-hDKK3 and pCDNA-T05 empty vectors were kind gifts from Y Kawano (Imperial college, London, UK) and J Girling (Cambridge Research Institute, Cambridge, UK), respectively. FuGENE 6 (Roche, Welwyn Garden City, UK) was used for transfection.Cell culture treatmentsCells were treated for 24 h with synthetic and physiological protease inhibitors: aprotinin (10 μg ml−1), BB94 (5 μM), TIMP-1 (500 nM) and TIMP-2 (500 nM) in growth media. For functional studies, cells were pretreated as above and maintained in media with the inhibitor throughout the experiment. For invasion assays, inhibitors at the above-mentioned concentrations were also added to the Matrigel™ mix. Aprotinin, BB94 and purified recombinant TIMP-1 and TIMP-2 were kind gifts from Professor Murphy.siRNA transfectionsiGENOME Individual Duplex Human MT1-MMP siRNA and siCONTROL non-targeting siRNA were purchased from Dharmacon (Chicago, IL, USA) and used at 20 μM. Opti-MEM™ transfection media and oligofectamine™ (both from Invitrogen, Paisley, UK) were used to transfect the cells once they reached 50% confluency. Knockdown was assessed by both quantitative reverse-transcriptase polymerase chain reaction (qRT-PCR) and western blot analysis. To block transcription, 24 h following the knockdown of MT1-MMP, cells were incubated in serum-free media for 12 h and actinomycin D (10 μg ml−1) was added. Cells were harvested at time 0 as well as time points 1, 2, 4 and 8 h following the addition of actinomycin D, and qRT-PCR analysis was performed for DKK3 and MT1-MMP mRNA expression.Isolation of secreted DKK3 from mediaFollowing knockdown of MT1-MMP, cells were grown in opti-MEM (Invitrogen). Media were concentrated using Vivaspin column (pore size 1 × 105; Vivascience, Munich, Germany).Western blot analysisCell pellets were resuspended in modified RIPA buffer (50 mM Tris (pH 7.8), 150 mM NaCl, 5 mM EDTA, 15 mM MgCl2, 1% NP-40, 0.5% sodium deoxycholate, 1 mM DTT, protease inhibitors 1 : 100, 20 mM\nN-ethylmaleimide) and further disrupted by mechanical shearing through a 19-gauge needle. Soluble proteins were then separated by centrifugation at 4°C for 5 min. Concentration of protein was calculated using a Bio-Rad protein assay kit (Bio-Rad, Hemel Hempstead, UK). Proteins were run on 10% SDS-PAGE and blotted on nitrocellulose membrane and incubated with appropriate primary antibodies. The membranes were incubated with horseradish-peroxidase-labelled secondary antibody. The membranes were washed and developed with ECL Plus western blotting detection system (Amersham, Little Chalfont, UK). The MT1-MMP antibodies were kindly donated by Professor Murphy and used at a concentration of 1.5 μg ml−1; 350-amino-acid isoform of DKK3 antibody (Abcam, Cambridge, UK) was used at 1 μg ml−1.Reverse transcription and qRT-PCRRNEasy (Qiagen, Crawley, UK) extraction and purification kits were used to extract total RNA from cell pellets according to the manufacturer's recommendations. SuperScript™ III first-strand synthesis system (Invitrogen) and random hexamer primers were used to reverse transcribe 2 μg of total RNA according to the manufacturer's recommendations. Amplification by PCR was carried out in a final volume of 10 μl containing 1.5 μl of reverse-transcribed cDNA, 5 μl of SYBER Green PCR master mix (Applied Biosystems, Warrington, UK) and 2 pmol of primers. ABI PRISM 7900 HT sequence detection system (Applied Biosystems) with the following PCR conditions was used: 50°C for 2 min, 95°C for 10 min, 40 cycles of 95°C for 15 s and 60°C for 1 min. As normalisation standards, GAPDH and SDH were used. Dickkopf-3 QuantiTech® primer assay (Qiagen) was used. The MT1-MMP, SDH and GAPDH primers were designed using the Primer Express software (Applied Biosystems). Sequences of the primer pairs were as follows: MT1-MMP, forward primer 5′-TGCCATGCAGAAGTTTTACGG-3′ and reverse primer 5′-TCCTTCGAACATTGGCCTTG-3′; GAPDH, forward primer 5′-GCAAATTCCATGGCACCGT-3′ and reverse primer 5′-TCGCCCCACTTGATTTTGG-3′; SDH, forward primer 5′-TGGGAACAAGAGGGCATCTG-3′ and reverse primer 5′-CCACCACTGCATCAAATTCATG-3′.Expression array analysisAmino Allyl Message Amp kit (Ambion, Warrington, UK) was used according to the manufacturer's recommendation for amplification and labelling of RNA. In brief, 10 μg of amplified RNA was resuspended in 6.7 μl of nuclease-free water, 4.5 μl of DMSO, 3.3 μl of NaHCO3 (pH 9) and 4.5 μl of Cy3 or Cy5. Labelled amplified RNA were hybridised to 22 k CRUK human cDNA expression array slides (ICR Array Facility, Surrey, UK).Matrigel invasion assayA modified Boyden dual chamber was used to assess directional migration of the cells. In brief, 10 μg of growth-factor-reduced Matrigel (BD Biosciences, Oxford, UK) in a total of 40 μl of serum-free media was applied to the upper chamber of 8-micron-pore cell culture insert (Becton Dickinson, Oxford, UK) and allowed to set over 2 h. A total of 5 × 104 cells were suspended in serum-free media and added to the upper chamber. Serum-positive media were used as chemoattractant in the lower chamber.After a 24 h incubation period at 37°C, 5% CO2, the media and cells remaining in the upper chamber were removed using a cotton bud. The insert was fixed in methanol and stained using haematoxylin. The membrane was removed from the insert and the number of invading cells was quantified by counting the cells in four random per high-power fields and calculating the mean number of invading cells. All experiments were performed in triplicate.Scratch testA total of 1 × 106 cells were seeded onto a six-well plate and allowed to reach full confluence. The monolayer was wounded using a cocktail stick. Cells were incubated with appropriate media, as stated. Digital images were taken at times 0, 24 and 48 h. The mean area was calculated using Internet-based software Image J (http://www.rsb.info.nih.gov/ij/download.html) and compared to time 0.MTT cell proliferation assayA total of 5000 cells were plated onto flat-bottomed 96-well plates and maintained overnight. Cells were incubated with 20 μl of MTT (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) for 2 h. A volume of 200 μl of DMSO was added and the absorbance measured at 492 nm.ResultsReducing MT1-MMP through siRNA leads to a reduction of invasive phenotypeWe examined the expression of MT1-MMP mRNA in a bank of UCC cell lines, and detected the highest levels in EJ28 cells (Figure 1A), which is the most invasive cell lines, as demonstrated by Matrigel invasion assay (Figure 2A).We confirmed that EJ28 cell invasion was dependent on MT1-MMP by siRNA transfection. Effective reduction of endogenous MT1-MMP was confirmed by western blot analysis and qRT-PCR over 120 h (Figure 1B and C) and a 70% reduction of cell invasion (P<0.05) was detected, when compared with wild-type and scrambled control (Figure 1D).Protease and MMP inhibition reduces invasive potentialUrothelial cell carcinoma cell lines were cultured with synthetic and physiological protease inhibitors, including BB94, aprotinin, TIMP1 and TIMP2, and invasion was assessed by the Matrigel assay. As shown in Figure 2A, treatment of EJ28 and HT1376 with aprotinin lead to a 45% reduction in invasiveness; treatment of EJ28 and HT1376 with BB94 to a 79% and a 55% reduction in invasiveness, respectively. Protease inhibition reduced the invasive capacity of the EJ28 cells, but had limited effect on 253JBV and UMUC3, which expressed low and moderate levels of MT1-MMP, respectively. Targeting the soluble MMPs, such as MMP2 and 13, by TIMP1 lead to a 44% reduction of invasion in EJ28 cells, whereas inhibition of all MMPs, including MT1-MMP, by TIMP2 lead to a 74% reduction (P<0.0005) (Figure 2B).Expression array identifies potential targets of MT1-MMP in EJ28 cellsTo explore putative targets of MT1-MMP, we employed expression microarray to highlight differential gene transcript expression following knockdown of MT1-MMP.A total of 70 genes were found to have a statistically different expression ratio following downregulation of MT1-MMP in EJ28 cells compared with controls. Functional analysis of these genes using gene ontology annotation through DAVID (data base for annotation, visualization and integrated discovery) revealed involvement in the regulation of cell growth and maintenance, signal transduction as well as regulation of cell cycle. Of the targets identified, our interest was drawn to DKK3, which was among the top five upregulated genes. Dickkopf-3 has been shown to act as a tumour suppressor and inhibit cell invasion (Niehrs, 2006) and is potentially negatively regulated by MT1-MMP. We confirmed this by transient knockdown of MT1-MMP at 48 h. Dickkopf-3 mRNA levels were increased 1.8-fold compared with scrambled control (Figure 3A), and an increase in secreted protein was confirmed in the culture media of MT1-MMP-suppressed EJ28 cells (Figure 3B). Densitomeric analysis of the western blot revealed that a 1.8-fold increase in mRNA levels corresponds to over 3-fold increase in protein level, as compared to a nontransfected control (Figure 3C).Increased levels of DKK3 following the knockdown of MT1-MMP is not as a result of mRNA stabilityTo determine whether knockdown of MT1-MMP in EJ28 cells affects DKK3 mRNA stability, we examined the effect of actinomycin D, an inhibitor of RNA transcription, on DKK3 mRNA levels following knockdown of MT1-MMP. In the absence of actinomycin D, the DKK3 mRNA expression was increased 2.5-fold following knockdown of MT1-MMP compared with scrambled control siRNA. In the presence of actinomycin D, DKK3 mRNA levels decreased progressively over time with minimally detectable levels of DKK3 mRNA after 8 h. The DKK3 mRNA levels at 8 h showed no statistical difference with negative control condition where no mRNA was added (P>0.10). These data suggest that upregulation of DKK3 is not because of enhancement of mRNA stability but as a result of transcriptional activation (Figure 3D).DKK3 reduces directional migration of EJ28 cellsTo explore the effects of DKK3 on tumour cell migration, conditioned medium was enriched for DKK3. Conditioned media were derived from transient transfection of Cos-7 cells with pCS2-hDKK3, a wild-type construct, or pCDNA-T05 empty vector, and migration was assessed through Matrigel and by the scratch test assay. Addition of DKK3 conditioned media lead to a 75% reduction in the invasive capacity of wild-type EJ28 cells (P<0.0005) (Figure 4A and B). In keeping with this finding, DKK3-conditioned media lead to a 37 and 30% reduction in closure of scratched monolayer at 24 and 48 h, respectively, when compared with controls (P<0.01) (Figure 4C and D). Interestingly, addition of broad-spectrum MMP inhibitors aprotinin and BB 94 did not alter the rate of closure of scratched monolayer. However, addition of DKK3 to aprotinin and BB94 lead to a 40 and 37% reduction of the scratched monolayer, respectively (Figure 4E). We have already shown that inhibition of MMP reduces the invasive potential of UCC cell lines, as demonstrated by Matrigel invasion assay. However, these data suggest that in two-dimensional culture, MMPs may not play a significant role in would healing (scratch test), but DKK3 plays a more central role in both invasion and migration.DiscussionWe have previously shown that among MMPs, MT1-MMP is highly expressed and associated with high grade and stage UCC (Wallard et al, 2006). In keeping with our previous report we have now demonstrated that MT1-MMP was expressed in all the UCC cell lines tested and high levels were detected in the more invasive cell lines. In addition, suppression of MT1-MMP using siRNA technology effectively inhibited invasion in these lines. These findings are in accordance with the accepted role in which MT1-MMP modifies the pericellular microenvironment to promote invasion and spreading of tumour cells (Itoh and Seiki, 2006).There is emerging evidence that the cytoplasmic domain of MT1-MMP and its Src-dependent phosphorylation play a role in modulation of cell migration (Nyalendo et al, 2007). Sounni et al and Freudenberg et al have recently suggested that MT1-MMP has additional transcriptional activation properties, which may be involved in angiogenesis as well as in cell invasion and motility. Transcriptional activation of VEGF-A appears to be mediated via Src tyrosine kinase (Sounni et al, 2004), whereas induction of Smad1 is mediated via TGF-β signalling (Freudenberg and Chen, 2007). In invasive cell lines, we confirmed that broad-spectrum protease (aprotinin) and MMP inhibitors (BB94, TIMP1 and TIMP2), as well as direct targeting of MT1-MMP, inhibited invasion. These data support our previous observations in clinical samples and suggest that MT1-MMP is a critical MMP promoting UCC cell invasion. The inhibition of invasion upon targeting MT1-MMP by TIMP2 and the less striking effect of TIMP1 on EJ28 cells lead us to explore potential downstream targets of MT1-MMP contributing to EJ28 cell invasion. To further indirectly support the notion that factors other than activation of MMP2 play a role in invasion of EJ28 cells, profiling of UCC cell lines showed that EJ28 cells express one of the highest levels of MT1-MMP, whereas their MMP2 expression levels are low compared with other cell lines (Supplementary data).Using a corneal myofibroblasts model, Ottino et al (2005) have also shown that activation of MT1-MMP by inflammatory mediator platelet activation factor does not result in increased MMP2 activity, suggesting an alternative pathway for extracellular matrix (ECM) remodelling in the cornea.We explored potential MT1-MMP-related targets and used expression array technology to identify gene alterations following suppression of MT1-MMP in EJ28 cells. Of the targets identified, our interest was drawn to DKK3, which is a tumour suppressor and inhibits cell invasion (Niehrs, 2006).In this study, expression of DKK3 was induced following downregulation of MT1-MMP and elevated levels of the protein were detected in the growth medium of MT1-MMP-repressed cells. Overexpression of DKK3 in conditioned media lead to significant reduction of invasion in UCC. Expression levels of DKK3 mRNA decreased progressively over time following incubation with actinomycin D, suggesting transcriptional regulation of DKK3 by MT1-MMP rather than mRNA stability. These results support studies that point toward a role in which MT1-MMP controls the transcriptional regulation of a number of genes important in the tumour progression machinery. As far as we are aware, modulation of DKK3 by MT1-MMP has not been reported previously. Whether this transcriptional regulation is a direct or an indirect effect remains to be elucidated. One possible explanation for this transcriptional effect observed could be internalisation of MT1-MMP and, in particular, localisation of the cytoplasmic domain in the nucleus and direct effect on transcription (Ip et al, 2007). This mechanism has previously been implicated in the regulation of transcription by the ErbB4 and Notch receptors, which require proteolytic cleavage and nuclear translocation of their cytoplasmic domains. However, this may represent a rather unique case and it is more commonly accepted that degradation of the ECM by MT1-MMP modulates signals generated by integrin–ECM interactions, which consequently affect gene transcription. A further consideration would be transcriptional regulation as an indirect result of activation of other pathways.Future workThere remain many unanswered questions regarding the exact mechanism of modulation of DKK3 by MT1-MMP. Indeed, it would be very interesting to pose the question that in view of the fact that DKK3 is considered to be a tumour suppressor gene whether it can also regulate MT1-MMP, suggesting a feed back loop. To establish the mechanism of the modulation of DKK3 by MT1-MMP, we plan to use various mutants of MT1-MMP in our knockdown system to investigate the contribution made by various elements of MT1-MMP protein.In conclusion, we have shown that MT1-MMP is essential for the invasion of UCC cell lines, and involves hitherto unreported mechanisms. Membrane type-1-MMP appears to transcriptionally modulate DKK3 levels, which is a potent inhibitor of invasion.\n\nREFERENCES:\n1. Beauvais DM, Rapraeger AC (2003) Syndecan-1-mediated cell spreading requires signaling by alphavbeta3 integrins in human breast carcinoma cells. Exp Cell Res\n286: 219–23212749851\n2. Beauvais DM, Rapraeger AC (2004) Syndecans in tumor cell adhesion and signaling. Reprod Biol Endocrinol\n2: 314711376\n3. Endo K, Takino T, Miyamori H, Kinsen H, Yoshizaki T, Furukawa M, Sato H (2003) Cleavage of syndecan-1 by membrane type matrix metalloproteinase-1 stimulates cell migration. J Biol Chem\n278: 40764–4077012904296\n4. Freudenberg JA, Chen WT (2007) Induction of Smad1 by MT1-MMP contributes to tumor growth. Int J Cancer\n121: 966–97717455258\n5. Hoang BH, Kubo T, Healey JH, Yang R, Nathan SS, Kolb EA, Mazza B, Meyers PA, Gorlick R (2004) Dickkopf 3 inhibits invasion and motility of Saos-2 osteosarcoma cells by modulating the Wnt-beta-catenin pathway. Cancer Res\n64: 2734–273915087387\n6. Hsieh SY, Hsieh PS, Chiu CT, Chen WY (2004) Dickkopf-3/REIC functions as a suppressor gene of tumor growth. Oncogene\n23: 9183–918915516983\n7. Ip YC, Cheung ST, Fan ST (2007) Atypical localization of membrane type 1-matrix metalloproteinase in the nucleus is associated with aggressive features of hepatocellular carcinoma. Mol Carcinog\n46: 225–23017219425\n8. Itoh Y, Seiki M (2006) MT1-MMP: a potent modifier of pericellular microenvironment. J Cell Physiol\n206: 1–815920734\n9. Jemal A, Siegel R, Ward E, Murray T, Xu J, Thun MJ (2007) Cancer statistics, 2007. CA Cancer J Clin\n57: 43–6617237035\n10. Kajita M, Itoh Y, Chiba T, Mori H, Okada A, Kinoh H, Seiki M (2001) Membrane-type 1 matrix metalloproteinase cleaves CD44 and promotes cell migration. J Cell Biol\n153: 893–90411381077\n11. Kawano Y, Kitaoka M, Hamada Y, Walker MM, Waxman J, Kypta RM (2006) Regulation of prostate cell growth and morphogenesis by Dickkopf-3. Oncogene\n25: 6528–653716751809\n12. Knauper V, Will H, Lopez-Otin C, Smith B, Atkinson SJ, Stanton H, Hembry RM, Murphy G (1996) Cellular mechanisms for human procollagenase-3 (MMP-13) activation. Evidence that MT1-MMP (MMP-14) and gelatinase (MMP-2) are able to generate active enzyme. J Biol Chem\n271: 17124–171318663255\n13. Kurose K, Sakaguchi M, Nasu Y, Ebara S, Kaku H, Kariyama R, Arao Y, Miyazaki M, Tsushima T, Namba M, Kumon H, Huh NH (2004) Decreased expression of REIC/Dkk-3 in human renal clear cell carcinoma. J Urol\n171: 1314–131814767340\n14. Nagano O, Saya H (2004) Mechanism and biological significance of CD44 cleavage. Cancer Sci\n95: 930–93515596040\n15. Nagase H, Visse R, Murphy G (2006) Structure and function of matrix metalloproteinases and TIMPs. Cardiovasc Res\n69: 562–57316405877\n16. Niehrs C (2006) Function and biological roles of the Dickkopf family of Wnt modulators. Oncogene\n25: 7469–748117143291\n17. Nozaki I, Tsuji T, Iijima O, Ohmura Y, Andou A, Miyazaki M, Shimizu N, Namba M (2001) Reduced expression of REIC/Dkk-3 gene in non-small cell lung cancer. Int J Oncol\n19: 117–12111408931\n18. Nyalendo C, Michaud M, Beaulieu E, Roghi C, Murphy G, Gingras D, Beliveau R (2007) Src-dependent phosphorylation of membrane type I matrix metalloproteinase on cytoplasmic tyrosine 573: role in endothelial and tumor cell migration. J Biol Chem\n282: 15690–1569917389600\n19. Ottino P, He J, Axelrad TW, Bazan HE (2005) PAF-induced furin and MT1-MMP expression is independent of MMP-2 activation in corneal myofibroblasts. Invest Ophthalmol Vis Sci\n46: 487–49615671273\n20. Sabeh F, Ota I, Holmbeck K, Birkedal-Hansen H, Soloway P, Balbin M, Lopez-Otin C, Shapiro S, Inada M, Krane S, Allen E, Chung D, Weiss SJ (2004) Tumor cell traffic through the extracellular matrix is controlled by the membrane-anchored collagenase MT1-MMP. J Cell Biol\n167: 769–78115557125\n21. Sato H, Takino T, Okada Y, Cao J, Shinagawa A, Yamamoto E, Seiki M (1994) A matrix metalloproteinase expressed on the surface of invasive tumour cells. Nature\n370: 61–658015608\n22. Seiki M (2003) Membrane-type 1 matrix metalloproteinase: a key enzyme for tumor invasion. Cancer Lett\n194: 1–1112706853\n23. Sounni NE, Roghi C, Chabottaux V, Janssen M, Munaut C, Maquoi E, Galvez BG, Gilles C, Frankenne F, Murphy G, Foidart JM, Noel A (2004) Up-regulation of vascular endothelial growth factor-A by active membrane-type 1 matrix metalloproteinase through activation of Src-tyrosine kinases. J Biol Chem\n279: 13564–1357414729679\n24. Torti FM, Lum BL (1984) The biology and treatment of superficial bladder cancer. J Clin Oncol\n2: 505–5316427417\n25. Urakami S, Shiina H, Enokida H, Kawakami T, Kawamoto K, Hirata H, Tanaka Y, Kikuno N, Nakagawa M, Igawa M, Dahiya R (2006) Combination analysis of hypermethylated Wnt-antagonist family genes as a novel epigenetic biomarker panel for bladder cancer detection. Clin Cancer Res\n12: 2109–211616609023\n26. Wallard MJ, Pennington CJ, Veerakumarasivam A, Burtt G, Mills IG, Warren A, Leung HY, Murphy G, Edwards DR, Neal DE, Kelly JD (2006) Comprehensive profiling and localisation of the matrix metalloproteinases in urothelial carcinoma. Br J Cancer\n94: 569–57716465195\n27. Woessner Jr JF (1991) Matrix metalloproteinases and their inhibitors in connective tissue remodeling. FASEB J\n5: 2145–21541850705\n28. Zhou Z, Apte SS, Soininen R, Cao R, Baaklini GY, Rauser RW, Wang J, Cao Y, Tryggvason K (2000) Impaired endochondral ossification and angiogenesis in mice deficient in membrane-type matrix metalloproteinase I. Proc Natl Acad Sci USA\n97: 4052–405710737763"
+}

batch_8/PMC2528003.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2528003",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2528003\nAUTHORS: Gerald Kastberger, Evelyn Schmelzer, Ilse Kranner\n\nABSTRACT:\nGiant honeybees (Apis dorsata) nest in the open and have evolved a plethora of defence behaviors. Against predatory wasps, including hornets, they display highly coordinated Mexican wave-like cascades termed ‘shimmering’. Shimmering starts at distinct spots on the nest surface and then spreads across the nest within a split second whereby hundreds of individual bees flip their abdomens upwards. However, so far it is not known whether prey and predator interact and if shimmering has anti-predatory significance. This article reports on the complex spatial and temporal patterns of interaction between Giant honeybee and hornet exemplified in 450 filmed episodes of two A. dorsata colonies and hornets (Vespa sp.). Detailed frame-by-frame analysis showed that shimmering elicits an avoidance response from the hornets showing a strong temporal correlation with the time course of shimmering. In turn, the strength and the rate of the bees' shimmering are modulated by the hornets' flight speed and proximity. The findings suggest that shimmering creates a ‘shelter zone’ of around 50 cm that prevents predatory wasps from foraging bees directly from the nest surface. Thus shimmering appears to be a key defence strategy that supports the Giant honeybees' open-nesting life-style.\n\nBODY:\nIntroductionGiant honeybees (Apis dorsata and A. laboriosa) belong to the oldest honeybee species after the dwarf honeybees (e.g. A. florea), having evolved about five to ten million years ago [1], [2]. Unlike the more recent cave-dwelling honeybees (e.g. A. cerana, A. mellifera), Giant honeybees build their nests predominantly in the open, suspending their roughly semicircular combs from overhead supports such as tree branches, rocks or buildings [for general biology of Giant honeybees see 2]–[12]. In the open, they are directly exposed to a variety of predators, particularly to birds and wasps [3], [8], [13]. This predatory pressure apparently gave rise to the evolution of a series of defence strategies [2], [5], [6], [13], [14], [15].Generally, defence behaviors of honeybees may involve physical contact with aggressors. A prominent example is their capacity to recruit stinging guards [15] and to mobilize a whole army of defenders. In Giant honeybees, mass mobilization of stinging guards may occur within a split second, and gave them the reputation of being the most dangerous stinging insects on earth [6], [8], [13], [16]. Against wasps, which are major predators of bees, honeybees have developed specific defence behaviors such as heat balling of wasps that come into direct contact with the honeybee nest. Heat balling has been reported for A. cerana, A. mellifera\n[17], [18] and A. dorsata\n[19], [20]. For this purpose, the bees heat their thoraces by their flight muscles to above 45°C, a temperature that is lethal to wasps.Honeybee colonies also defend themselves without physical contact with their enemies, minimizing the risk for the defending bees. Examples include the aposomatic coloration, which is characteristic of all hymenopterans [21], the colony aggregation [2], [13], [22], and shimmering behavior [5], [6], [8], [14], [16], [23], [24]. Shimmering has been observed in A. cerana and A. florea\n[2], and in A. dorsata, it is a notable visual cue that is impressive even to humans [3], [8]. Shimmering involves an intriguing capacity for very rapid communication within the nest (movie S1), in which hundreds of individual bees flip their abdomens upwards in a split second forming Mexican wave-like patterns [25]. These wave-like figures are ineffective to stop larger predators such as birds [13] or mammals from feeding directly from the Giant honeybees' nests. Field observations [14] suggest that shimmering is provoked especially by wasps. However, the evolutionary role of shimmering and, in particular, its significance as a defence behavior are so far only hypothesized [for summary see 2], and the precise relationship between Giant honeybees and potential predators regarding shimmering is not understood.This article investigates whether Giant honeybees succeed defending their nests against hornets by shimmering, and how prey and predator interact. The honeybee colony behavior is analyzed concerning the occurrence, strength and repetitiveness of shimmering under the aspects of proximity and velocity of predatory wasps. This proves shimmering as a colony response to approaching wasps. On the other hand, the wasp behavior has been investigated in response of the time course and the strength of shimmering regarding proximity to the honeybee nest. In two different scenarios of experiments, shimmering waves, in particular in their ‘big-scale’ shape, are proved to repel wasps within a specified range around the honeybee nest. ‘Small-scale’ shimmering is effective in preventing wasps from predation by generating ‘confusion’. It is demonstrated that the predation activity of wasps near honeybee nests and the defence responses of Giant honeybees through shimmering base on a reciprocal, mutually adjusted relationship with possibly coevolutionary roots. Lastly, it is discussed how shimmering benefits Giant honeybees, in particular to support their open-nesting habit.Materials and MethodsSpecies, study siteThis article investigates the prey-predator interactions between colonies of Giant honeybees (A. dorsata) and Vespa sp. hornets at two water towers in the Agricultural Campus of the Tribhuvan University of Kathmandu, Rampur, Chitwan, Nepal in November 2004. The aim was to analyze the wave-like shimmering behavior [8], [23], [24] of Giant honeybees in response to the predatory hornets that hovered around the bee nests. Two scenarios were chosen at two experimental nests where hornets had regularly been observed. The first experimental nest (scenario A) consisted of 8000 bees, measuring one meter in the horizontal span and was located at the external rim of the water tower. The colony had arrived just one day before observations had started. It formed a cluster with a small central comb. The second test colony (scenario B) was slightly smaller than the first one; it had also migrated to the ceiling of the same water tower some days before experimentation and was also a cluster without a central comb.Scenarios investigatedIn both scenarios, a single camera recorded the behaviors of hornets and bees in PAL format enabling a frame-by-frame analysis at a rate of 25 images per second. Prior to these experiments documented in this article, the authors had observed thousands of shimmering waves in hundreds of Giant honeybee colonies on several expeditions in India and Nepal over 15 years. Based on this broad experience, both scenarios of the two experimental nests were recognized to typify the shimmering behavior of giant honeybees and the respective flight behaviors of hornets under the given prey–predator relations.In scenario A, predatory hornets were observed hovering around the honeybee nest regularly provoking shimmering behaviors (126 shimmering waves in 77 s). For reference, we analyzed the nest situation before the hornets had appeared (354 shimmering waves in 168 s). The camera obtained views of the front flat of the nest (Fig. 1). The flight trajectories of the hornet were, therefore, documented in the projection of the horizontal and vertical nest-specific real-world coordinates, that is, to the left and the right side of the nest (defined as x-dimensions), and upwards and downwards of the nest (defined as y-dimensions).10.1371/journal.pone.0003141.g001Figure 1The two different perspectives used in the experimental scenarios A and B.The nest-specific axes were defined by their real-world coordinates: x, the horizontal (sideways) directions in regard to the vertical flat of the nest; y, the up- and downward directions; z, towards and off wards the nest as projected on the horizontal plane. The camera (cam) imaged the x-y projection in scenario A and the x-z projection in scenario B. Note, the distances of the hovering wasp from the bee nest (dxz) were addressed in scenario B as projection on the x-z plane.In scenario B, 203 shimmering waves were observed in the presence of two predatory hornets; 88 cases referred to a single hornet, 115 cases to two hornets that were simultaneously around the nest. In total, 318 episodes of hornets were traced under shimmering activity. The camera documented the scenery from the bottom view (Fig. 1), which referred to the horizontal (x-dimension) real-world coordinates and to the z-dimensions, which are defined as the directions ‘toward and off ward’ the flat expansion of the honeybee nest.Assessment of honeybee colony behaviorsThe focus of this article is on shimmering behaviors. Shimmering is made up by abdominal movements of quiescent individuals predominantly at the surface of the honeybee nest, displaying wave-like processes. These abdominal movements of surface bees were detected by image analysis (Image-Pro, Flir), assessing the shimmering waving strength (W) frame by frame. Shimmering was detected in three steps, which were controlled by automated and manual decisions. In a first step, the time course of movement activities was traced. The criterion for automated detection of a peak in movement activity was that the W-value of the reference frame had to exceed the W-values three frames before and after the threshold value, which was defined at 0.9% of the overall maximum of shimmering. Second, the time of the onset of this movement activity was traced by searching for the minimum value of this waving session, 3–10 frames backwards from the peak time. Here, the maximum period of 10 frames considers the time course of shimmering that exhibits a maximum of less than 400 ms (Figs. 2,3). As the third step, the nature of shimmering was identified as a wave-like process. Here, manual control was needed to distinguish shimmering from other movement activities, such locomotor activities as walking, dancing or flying.10.1371/journal.pone.0003141.g002Figure 2(A) Continuous assessment of abdominal thrust activity of the experimental Giant honeybee nest, while a predatory wasp was present in front of it (scenario A); the ordinate value gives the relative strength of thrust activity, the maximum value (1.0) refers to the maximum strength of shimmering (max Wpeak) as observed during 300 s after the onset of waves; the threshold at rW = 0.4 discerns small-scale (blue area) from big-scale (red area) waves; (B) the time course of big-scale (red curve) and small-scale (blue curve) waves; curves show arithmetical means of the respective waves, thin vertical lines denote SEMs; abscissa, rel experimental time, time zero is defined by the onset of the abdominal thrust activity (corresponding to waving); (C) the rate of abdominal thrust activity per min of the experimental nest in two behavioral contexts, (C1) ‘undisturbed by a hornet’ and (C2) ‘disturbed by a hornet’; abscissa gives the categories of abdominal thrust activity as percentage of maximal waving strength; the blue columns refer to small-scale waves, the grey columns and the thin lines in the background give the respective rates of the opposite behavioral context for comparison. The distributions of the rates of abdominal thrust activities differed between C1 and C2 insofar that the proportions of the occurrences of waves regarding both states varied from one category to the other (Chi-square test, P<0.001, f = 49). Furthermore, under the presence of wasps the abdominal thrust activities show generally higher rates (Wilcoxon Signed Rank Test: P = 0.017).10.1371/journal.pone.0003141.g003Figure 3The time courses of shimmering of Giant honeybees in response to approaching hornets (scenario B).The waving strength W ( = number of abdomen-shaking bees per frame) depends on the hornets' distances from the nest dxz (A) and on the hornets' flight velocities vxz (B); time zero defines the onset of the waves; (A) five dxz classes (Cdxz = 1–5; coded in yellow to red; for definition, see Methods and Fig. 4,5) and (B) eight vxz classes (Cvxz = 1–8; coded in green to blue) of hornet flight episodes were considered; dxz and vxz class values were assessed from hornets in the 400 ms interval prior to the start of shimmering. Curves show arithmetical means, thin vertical lines denote SEM. For data details, see table 1 and 2.10.1371/journal.pone.0003141.t001Table 1The modulation of shimmering by hovering hornets regarding their distance to the honeybee nest (scenario B).Categories of distance of the hornets from the nest before shimmeringNumber of episodesDistance of the hornets from the nest before shimmeringShimmering activity as response to the presence of the hornetsCdxz\ndxz [cm]ndxz [cm]Wpeak\nW600\n10–18612.7±1.812.6±1.371.1±4.4218–301925.9±0.712.1±2.067.1±6.9330–404935.8±0.410.0±0.855.2±2.9440–607949.2±0.78.2±0.549.4±2.05>604871.5±1.56.5±0.542.4±2.1201Shimmering activities of 201 episodes in terms of Wpeak and W600 in dependence of the pre-wave distances of the hornets from the nest (definitions, see Methods). The hornet data (dxz) were assessed in the interval 400 ms prior to shimmering. Shimmering responses (means±SEM) were categorized by five classes (Cdxz = 1 to 5) of dxz (cf. Fig. 3A).10.1371/journal.pone.0003141.t002Table 2The modulation of shimmering by hovering hornets regarding their flight velocities (scenario B).Categories of flight velocity of the hornets before shimmeringNumber of episodesDistance of the hornets from the nest before shimmeringShimmering activity as response to the presence of the hornetsCvxz\nvxz [cm s−1]ndxz [cm]Wpeak\nW600\n110–20741.15±6.046.43±0.7237.08±2.63220–252540.27±3.337.49±0.7543.23±3.51325–302941.01±3.389.28±0.9851.36±3.42430–405338.60±2.058.88±0.9053.08±3.23540–503047.23±3.539.06±0.8355.58±3.97650–602147.68±4.718.86±1.0453.36±5.24760–802153.78±4.019.89±0.9558.28±4.208>801656.98±5.3511.14±1.2661.88±5.58202Shimmering activities of 202 episodes in terms of Wpeak and W600 in dependence of the hornets' pre-wave flight velocities. The hornet data (vxz) were assessed in the interval 400 ms prior to shimmering. Shimmering responses (means±SEM) were categorized by eight classes (Cvxz = 1 to 8) of flight velocity vxz of the hornets (cf. Fig. 3B).The wave strength W was originally defined and calibrated as the number of surface bees per frame that were actively ‘shaking’ or ‘lifting’ their abdomens. This ‘wave’ parameter was also applied in a more general sense as one of the ongoing aspects of movement of nest mates, regardless of whether or not shimmering had been identified as the respective process. For explanation, the value Wpeak characterized the intensity of shimmering at its peak, and the value W−400 ms was used to quantify the levels of movement activity 400 ms before the onset of the shimmering waves (which may include residual waving activity).To estimate the number of bees that participated in a shimmering wave, the integral ∫W(t) was calculated and this value was calibrated by a correction factor f\ncorr. Calibration needed the manual counting of the real number of bees that had lifted their abdomen in the referenced images. This calibration procedure was necessary because a single abdominal shaking lasts for 160 ms [14] and in shimmering, the movement detection by image analysis is particularly sensitive to the onset of shaking. As a representative estimate for the shimmering strength, the number of bees participating in the first 600 ms after the start of shimmering was defined by the formula W600 = ∫W(600 ms)/f\ncorr.Assessment of hornet behaviorsThe scenarios A and B provided two different perspectives (Fig. 1) regarding the honeybee nest and referred to different sets of flight parameters of the predatory hornets. Using image analysis (Image-Pro, Flir), behavioral parameters was assessed frame by frame (in steps of 40 ms) in both scenarios. The positional coordinates of the hornet with regard to the x-y plane in scenario A and with regard to the x-z plane in scenario B were measured. In scenario A, the movement components (Δx/40 ms, Δy/40 ms) of the hornets in front of the bee nest were measured and the respective angular parameters of flight direction (αxy/40 ms), turning behavior (θxy/40 ms), and the nondirectional parameter of flight velocity (vxy) were calculated. In scenario B, the shortest distance of the hornet to the nest surface (d\nxz) and its differentiation over time that defines the distance velocity v\ndxz = Δd\nxz/Δt were determined. The turning angle per time θxz/40 ms documented changes in flight direction in the x-z projection, which was calculated as the difference between the flight directions (αxz {1,2}, αxz {2,3}) displayed in two successive pairs of frames ({1,2},{2,3}). Here, the hornets displayed scenario B (nest-)-specific behaviors; they approached the nest predominantly from one direction, that is, from the bottom right side of the image. In the image-based standard, the hornets in scenario B turned counter-clockwise ‘away from the nest’ after shimmering, and turned clockwise ‘toward the nest’. ‘Turning away from the nest’ was coded as positive and ‘turning toward the nest’ as negative, to get the same signs for the shimmering of the bees and the turning behaviors of the hornets in the comparing graphs.The flight velocity v\nf was calculated by vf = ds/dt (with ds as length of the flight path per time interval dt) in the respective projection planes (vf = vxy for scenario A, and vf = vxz for scenario B). The vf value in its time course is a good indicator for changes in the hornets' flight behavior near the honeybee nest, if the focus is on the reactivity of the hornets to the shimmering waves. But in scenario B, its absolute levels were also considered.Analysis of interactions/distinguishing ‘action’ and ‘reaction’The behaviors of both predator and prey have been synchronized to the onset of the shimmering waves. This trigger concept not only associates the shimmering waves of the bee colony to the hornet behaviors considering shimmering as responses to the hornets' behaviors, but also vice versa, that is, it allows considering hornet behaviors as responses to shimmering. The shimmering waves and the flight behaviors of the hornets were monitored from 400 ms prior to until 1000 ms after the onset of shimmering. The question was whether the flight behavior of the hornet in the vicinity of the honeybee nest represented adequate cues for affecting shimmering. In detail, the investigation was on whether the strength of the shimmering waves (W) was dependent on the flight parameters of the hornet (dxz, vdxz, vf) in the pre-wave period, 400 ms before the onset of waving. To the contrary, changes of the hornet behavior in the first 600–1000 ms after the onset of shimmering was tested as obvious responses to shimmering.Categorization of bee-hornet episodesAn episode between the honeybee as prey and the hornet as predator was defined by the shimmering behavior of the bee colony and by the flight behaviors of the hornet in the front of the bee nest. As detailed in earlier sections, the scenarios A and B refer to different geometrical perspectives and therefore, in particular, to different sets of flight parameters of the hornets. Besides that basic difference, both scenarios enabled to provide distinct concepts of evaluation. In scenario A, ‘big-scale’ and ‘small-scale’ waving activities were distinguished according to their peak strength, which was above or below the threshold of 40% of the maximum wave strength W that was monitored in the total experimental session. Big-scale processes represented shimmering waves that spread over the nest surface, affecting hundreds of bees to lift their abdomens sequentially. Small-scale processes were local, wave-like reactions of small groups of abdomen-thrusting bees and have not exceeded the number of 10 active bees.In scenario B, the episodes were categorized in two other ways. First, by the distance dxz of the hornets to the honeybee nest prior to the onset of shimmering. For that, the dxz values in the 400 ms before shimmering was averaged and the episodes were sorted into five classes of pre-wave distances (Cdxz = 1 to 5, table 1), whereas the class widths were chosen according to the statistical incidence of the hornet episodes. If two hornets were present, the dxz value of that hornet that was nearer to the nest was used to categorize the shimmering wave. The second method of categorization considered the flight velocities (vf = vxz) of the hornets in the same 10 frames prior to shimmering. Eight classes of pre-wave flight velocity (Cvxz = 1 to 8, table 2, see Results, chapter ‘Shimmering as response to predatory hornets’) of the hornets were defined. If two hornets (w1,w2) were present, a combined vw1+w2 value was calculated using a linear correction model considering the particular dxz value for each hornet (dw1;dw2): vw1+w2 = vw1*fw1+vw2*fw2 with fw1 = 1−(dw1/(dw1+dw2)) and fw2 = 1−fw1. This procedure weighted the nearer hornet more than the distant one because of the linear relationship between hornet proximity and waving strength (Fig. 4A).10.1371/journal.pone.0003141.g004Figure 4The effect of flight behavior of predatory hornets on the waving strength of Giant honeybees (scenario B).The waving strength W600 gives the numbers of bees which had shaken their abdomens over 15 frames (definition see inset and text); it depends on the hornet's distance from the nest dxz (A) and on the hornet's flight velocity vxz (B); time zero in the insets defines the start of the shimmering waves (cf. Fig. 3); (A,C) red to yellow shaded areas define the five dxz classes, and (B,D) green to blue shaded areas define the eight vxz classes of hornet flights as used in Fig. 3 (for definition see Methods); open circles are arithmetical means, thin vertical and horizontal lines denote SEM; thick lines are regressions of the mean values regarding to 201 wave episodes with 317 flight episodes of two wasps (Cdxz = 1 to 5; Cvxz = 1 to 8); regression A: r = −0.969, P = 0.006; regression B: r = −0.963, P = 0.015. W−400, the amplitude of the waving strength 400 ms before the onset of the consecutive wave (see inset and text for definition), giving the residual shimmering strength under repetitive conditions; W−400 declines with dxz (regression C: r = −0.989; P = 0.022), but inclines regarding vxz at lower velocity levels (Cvxz = 1–5: regression d1: r = 0.975, P = 0.005) and shows an overall nonlinear relation for Cvxz = 1–8 (regression d2: r = 0.857; P = 0.027). All tests refer to Polynomial Regression (SigmaStat).Categorization of ‘reactive’ and ‘nonreactive’ hornetsIn scenario B, the shimmering waves and the flight behaviors of the hornet were categorized by the following two principles: The waving episodes were judged by manual decision based on gross inspection of honeybee and hornet behaviors, whether or not the hornet had the chance to provoke shimmering, and whether or not the hornet was in the position to be affected by the wave. In the video, these behavioral categories were clearly distinguishable (see the ‘reactive’ hornet demonstrated in movie S2,S3). If hornets responded to a wave, independently whether they provoked it or not, the respective episodes (n = 149) and the hornets were termed ‘reactive’. In close vicinity to the nest (dxz<15 cm) all episodes were ‘reactive’.Shimmering is not necessarily a response to wasps, it may also be provoked by homing and departing bees (Fig. 2). It is intricate to judge whether hornets hovering near the honeybee nest are or are not in the position to respond to a shimmering wave. If they would be not, the episodes and the hornets were termed ‘nonreactive’ (n = 167). Three typical examples may illustrate the conditions for the categorization of ‘nonreactive’ episodes: First, if a hornet hovered near the nest but was relatively too far away from the nest (e.g. more than one meter away) and additionally, too slow (e.g. when hovering at the spot). In this case, the hornet was not necessarily disposed to elicit shimmering, and it was also not likely to respond to shimmering provoked by any other sources. Second, if a hornet chased a homing bee the hornet's body length axis was usually directed to the fleeing bee, and mostly away from the honeybee nest. Under this condition, the hornet hardly showed any sign of response to the shimmering wave, which might have been elicited by itself or by other sources. Third, if one hornet interacted with another hornet in front of the giant honeybee nest, both of them were concentrated on each other and were obviously not able to respond to the shimmering which they might have provoked.While the category of ‘reactive’ hornets is unequivocally defined, the category of ‘nonreactive’ hornets remained to be crucial because of its dependence on gross subjective manual criterions, which did not allow deciding in detail about any residual pattern of reactivity of the hornets to e shimmering. To proof whether ‘nonreactive’ hornets did not respond to shimmering or showed any odd reaction patterns, we tested the time courses of the respective behaviors of hornets categorized as ‘nonreactive’ for the existence of any obscure responsiveness.Assessment of ‘direction fidelity’ of predatory hornetsIn scenario A, the flight trajectories of the predatory hornets in the x-y projection were observed and the data during 400 ms before and 1200 ms after the onset of the shimmering wave considered. The mean trajectories of the hornet flight were compiled by integrating the mean positional changes (mΔx, mΔy) regarding two 40 ms time intervals relative to the onset of shimmering, and the relevant mean distances (mDxy = (mΔx2+mΔy2)0.5) per 40 ms were calculated. The value mDxy factually is the result of vector subtraction, because they refer to data of pooled individual flight paths, and therefore mDxy is a function of turning ranges ρxy of the hornet (ρxy = max θxy−min θxy) rather than a function of the flight velocities vxy of the hornet (compare Results, chapter ‘Small-scale and big-scale shimmering’). Assuming that the hornet has, per 40 ms, a maximum turning range of ρxy = 180°, the scalar values mDxy of the mean heading mθxy follow the equation mDxy = cos ρxy. In terms of statistics, mDxy is shorter the more the turning angles θxy of the hornet deviate from one frame to the other. Therefore, mDxy is a useful measure for the range of deviations in the directional flight behavior of the hornet. Calibrated between 0 and 1 (rel mDxy = a* mDxy) for the turning ranges from 180° to 0°, rel mDxy describes the ‘direction fidelity’ of the predatory hornet. In other words, if the hornet keeps its flight directions constant from one frame to the other (that is with a turning range of ρxy = 0°), it performs a high level of direction fidelity, which is reflected by the value rel mDxy = 1. To the contrary, if the hornet turns by θxy = ±90° within 40 ms, the turning range is maximal (ρxy = 180°), and the level of direction fidelity is minimal (rel mDxy = 0).StatisticsGaussian distributed data sequences were compared by parametric tests (t-test). If the normality test failed, the software automatically used nonparametric tests (Chi-square test, Wilcoxon Signed Rank Test). These tests traced differences in behaviors between two experimental states, e.g. how the behaviors of the hornets differ between two time intervals, such as before and after the onset of shimmering. Correlations were characterized by the regressions of the original data values of the respective behavioral classes or facultatively (see below) of their arithmetic means. The regressions were fitted by optimizing their coefficients of determination (R2) and tested by Spearman rank order correlation test.The time courses of the behaviors of honeybees and hornets in three steps were compared, using the One Way Repeated Measures ANOVA (e.g. Friedman test on Ranks, Sigmastat). The time correlations of the original data over a number of discrete time intervals in 40 ms steps were proved (e.g. from time zero to 400 ms after the onset of waving) and was adjusted for ties. In the first step, a test was conducted to see whether the pairs of variables of prey (W) and of predator (vdxz, θxz) tend to increase together during the experimental time after the onset of shimmering in two time intervals (0–400 ms; 400–800 ms). Here, Dunns Method tested the original data per time interval, and the Spearman test their mean values. In a second step, the relative differences among the treatment groups of prey (rel W) and of predator (rel vdxz, rel θxz) were proved. In a third step, if the relative differences were because of random sampling variability, the respective regression of the arithmetical means of the same time intervals as the appropriate description of the correlation was accepted. This three-step statistical procedure estimated the type of interaction between prey and predator, although the original predator data would have been too unfriendly for a straight one-way Repeated Measures ANOVA.ResultsOccurrence of shimmeringIn this study, the conditions under which shimmering waves are produced by giant honeybee colonies were investigated. In scenario A, the waving activities at the surface of the experimental honeybee nest under two conditions were monitored, (a) when a wasp was hovering in front of the nest (observation time: 77 s; Fig. 2, movie S4), and for reference (b) without a wasp around the honeybee nest (observation time: 168 s). These abdominal movement activities (see Methods) were categorized into small-scale (n = 72) and big-scale (n = 37) waves. In small-scale activities, only tens of bees raised their abdomens. The wave strength in the pre-wave period is low because the respective signals are nonrepetitive; its time course (blue curve in Fig. 2B) peaked after 200–250 ms. Big-scale waves (red curve in Fig. 2B) spread over the nest, reached their maximum activity typically after 400 ms and were repetitive. Consequently, the pooled data in the pre-wave phase of the reference wave exhibit residual traces of the preceding wave cycle.Wave-like processes also occur without the presence of a hovering wasp. Foraging nest mates departing from or arriving at the nest mostly are the source for the generation of shimmering in absence of predators. However, in scenario A, big-scale waves only occurred under the presence of predatory wasps (Fig. 2C, red columns). The rates of wave-like processes differed between the states ‘with wasp’ and ‘without wasp’ (Chi-square test, P<0.001, f = 49); here, the proportions of the occurrences of waves varied regarding both states from one wave strength category to the other. Furthermore, under the presence of wasps the wavelike processes showed generally higher rates (Wilcoxon Signed Rank Test: P = 0.017). This finding proved wasps evoked shimmering in close vicinity of the nest.Shimmering as response to predatory hornetsWaving strengthThis study investigates several flight parameters of predatory hornets in the vicinity of Giant honeybee colonies regarding their significance for eliciting shimmering. In scenario B, shimmering behavior was measured by its waving strength W and it was categorized by two parameters of the hornets' flights in the pre-wave phase. The first parameter was the proximity dxz of the hornets in respect to the honeybee nest. Here, five classes (Cdxz = 1–5; Fig. 3A, table 1) were considered. The supplemental regression in Fig. 4A summarizes the overall dependency of the waving strength (W600) from the proximity of the hornets (dxz). It was found that when the hornets were immediately prior to the onset of the wave close to the nest, the strength of the provoked wave was maximal, with a participation of 70 bees in the course of 600 ms (Cdxz = 1, Fig. 4A). Farther away, the hornets caused much smaller waves and aroused fewer bees.The second parameter to categorize the shimmering responses in scenario B was the flight velocity vf = vxz. For that, eight classes (Cvxz = 1–8; Fig. 3B, table 2) were considered. Hornets with a speed of less than 20 cm s−1 provoked only a weak single wave (Cvxz = 1 in Fig. 3B). A more detailed view shows that less than 40 bees were involved (Cvxz = 1 in Fig. 4B) in the average that corresponded to approximately 60% of the maximal shimmering response. If the hornets flew slightly faster, for example, with more than 30 cm s−1, the colony response was much stronger (Cvxz = 4 in Fig. 4B) at more than 80% of the maximal waving strength, which the fastest flying hornets had evoked (Cvxz = 8 in Fig. 4B). This finding demonstrates that wasps elicit strong waves if they fly at moderate speed, but if they hover at the spot the wasps are able to ‘creep’ nearer to the bees without evoking big-scale waves.RepetitivenessThe arousal state of a Giant honeybee colony when predatory wasps are nearby is also expressed by the repetitiveness of shimmering. A good measure of it is the residual waving strength W−400 (lower inset of Fig. 4), assessed 400 ms before the start of the particular wave. The amplitude of the residual wave strength also depended on both flight parameters, dxz and vf of the predatory hornet (Fig. 4C,D). The repetitiveness of waving increased exponentially the nearer the hornet came to the nest (regression in Fig. 4C). Regarding flight velocity, the repetitiveness of shimmering is seemingly complex and follows an optimum distribution (Fig. 4D). Hornets in the lower range of flight categories (Cvxz = 1 to 5) provoked a linear increase of repetitiveness in shimmering with their flight speed, but in the higher range of flight categories (Cvxz = 6 to 8) the repetitiveness of shimmering decreased with increasing vxz (regression d\n1 in Fig. 4D). The overall relation between the repetitiveness of shimmering and the flight speed vxz of predatory hornets is nonlinear (regression d\n2 in Fig. 4D). This complex relation results from the fact that vxz itself is a function of the hornet's proximity dxz. The auxiliary regression in Fig. 5C showed that the individual vxz correlated with the dxz data of the hornets in the average, slower hornets flew closer at the nest than faster ones. Furthermore, most of the hornets flew within the mean hovering distance of dxz = 52 cm (dashed line in Fig. 5A,C) and slower than 50 cm s−1 (Fig. 5B) when they provoked shimmering.10.1371/journal.pone.0003141.g005Figure 5Effect of flight behavior of predatory hornets on the waving strength of Giant honeybees (scenario B).Percentage of hornet episodes (n = 317) observed in the respective five dxz classes (A) and eight vxz classes (B) of hornets' flights; (C) the relationship between the flight velocities vxz (abscissa) of individual hornets and their distances dxz to the nest (ordinate); dashed lines (A,C) give the average hovering distance of the hornets (cf. Fig. 6). Regression of the means in C: dxz = 0.215*vxz+40.596; Cvxz = 1 to 8; r = 0.962; 317 episodes; P<0.001); for definition and color coding, see Figs. 3,4 and Methods.Finally, these data of scenario B (Fig. 3,4) make clear why shimmering waves were less repetitive although the hornets were fast. The reason was that fast wasps flew mostly farther away from the nest (Fig. 5C), and their effect on the honeybees to elicit shimmering was consequently less. In other words, repetitiveness of shimmering is primarily a function of proximity.Hornet behavior in response to shimmeringLong-lasting experiences over more than one decade with Giant honeybee colonies have enabled the authors hypothesize that shimmering waves do have antipredatory goals. If so, it should be possible to observe that shimmering lowers the chances of the hornets to prey on the curtain bees on the surface of the Giant honeybee nests. In the following sections, this surmise is investigated and questioned whether shimmering is able to distract wasps from grabbing bees, whether it is able to repel wasps or is even able to make wasps turn away from the nest.Shimmering drives hornets away from the nestThe primary question for an obvious antipredator impact of shimmering on wasps is whether wasps respond to shimmering. In scenario B, the hornets hovered in the average at a distance of dhov = dxz = 52.1±0.53 cm (mean±SE; 9757 images; dashed lines in Figs. 5,6) in front of the nest. At this mean hovering distance, hornets elicited only weak waves (cf. Cdxz = 4 in Fig. 3A); but if the hornets were nearer than dhov they not only elicited bigger waves, they also withdrew from the nest after the start of shimmering. They increased their distance from the nest more than they were before the wave had started (Cdxz = 1–3 in Fig. 6). However, when they were outside the mean hovering distance (Cdxz = 5 in Fig. 6), they came significantly nearer to the nest as soon shimmering had started.10.1371/journal.pone.0003141.g006Figure 6The hornets' responses to shimmering depend on the distance from the bee nest dxz (scenario B).Hornet behaviors were categorized in five distance classes Cdxz = 1–5 (see inset, Figs. 3–\n5 and Methods). Thick lines connect the arithmetical means, thin vertical lines denote SEM;. (A) Wasp behavior monitored for 1600 ms, starting 400 ms prior time 0, the onset of shimmering; Δdxz (1s) values give the changes in the position of the hornet regarding its distance to the honeybee nest within 1 s after the onset of the wave (significance levels: *, P<0.05; **, P<0.01; ***, P<0.001; t-test); the horizontal dashed line, the average hovering distance (dhov = 52.10±0.53 cm; n = 9757 images). (B) Correlation between Δdxz (1s) and dxz, the thick line gives the regression of means (Δdxz (1s) = −0.449* dxz+22.078; r = 0.998; Cdxz = 1–5; P<0.001; 326 episodes); positive values of Δdxz (1s) at Cdxz = 1–3 represent movements of the hornets away from the bee nest and indicate avoidance responses; the response shown for Cdxz = 4 is neutral, and the negative values of Δdxz (1s) at Cdxz = 5 outside dhov illustrate that the hornets usually approached the nest when shimmering started.To quantify this responsiveness of the wasp to shimmering, the change in distance of the hornet within one second after the onset of waving (Δdxz(1s)) was chosen. Taken together the responses at all five proximity categories (Cdxz = 1 to 5) the data correlated linearly with the proximity of the hornets to the nest prior to the shimmering wave (regression in Fig. 6B).Summarizing, the authors conclude that at distances greater than the mean hovering distance, hornets tended to approach the nest under the influence of shimmering. This suggests first, that hornets were attracted by shimmering if they were further away from the nest by at least half a meter. Second, at the mean hovering distance, hornets were not affected by shimmering; they stayed neutral with regard to approaching or leaving the nest site. Third, when the hornets were closer than the mean hovering distance, they withdrew from the nest in the course of shimmering. This hornet behavior is indicative of avoidance behavior, suggesting that shimmering plays a role in repelling predatory hornets, but only when close to the nest.Shimmering makes ‘reactive’ wasps turn away from the nestIn scenario B, the time courses of the flight behaviour of ‘reactive’ (see Methods) hornets were analyzed when they show the avoidance response to shimmering (Fig. 7). For that, two groups were distinguished regarding their distance to the nest, that is, when they were (a) nearer than the hovering distance (dxz<45 cm) or (b) when they were outside the mean hovering distance (dxz>45 cm). The average position dxz of the hornet in the 400 ms interval prior to the start of shimmering was again the criterion for the classification of both, the shimmering waves and the hornets' flights.10.1371/journal.pone.0003141.g007Figure 7The hornets' behaviors during shimmering (scenario B).The time courses of shimmering (A,D) and of the hornets' flights (B–C,E–F), ‘reactive’ (A–C) and ‘non-reactive’ (D–F) episodes (for definition, see text). Honeybee and hornet behaviors were synchronized to the start of the shimmering waves (abscissas give the time in milliseconds after the start of the shimmering waves). The hornets' behaviors are shown in terms of distance velocity vdxz (B,E) and turning angle θxz (C,F); see inset and text for definition. Two classes of hornets were defined according to their distance to the nest in the 400 ms interval prior to shimmering: dxz<45 cm (red circles, 84 ‘reactive’ episodes; 59 ‘non-reactive’ episodes), and dxz>45 cm (yellow circles, 65 ‘reactive’ episodes; 108 ‘non-reactive’ episodes). Different brown-shaded areas define two test intervals in relation to the time course of shimmering (brown-shaded: 0–400 ms, grey-brown shaded: 400–1000 ms). Circles and bars give arithmetical means±SEM. Big full (red or yellow) circles give significant differences of the data in relation to the starting time of the wave at t = 0 ms (P<0.05; Holm-Sidak test, Friedman Repeated Measures ANOVA on Ranks).‘Reactive’ hornets in close vicinity to the nest (dxz<45 cm) showed a strong reaction to shimmering and turned away from the nest with a maximum speed vdxz of around 18 cm s−1 (Fig. 7A,B, red symbols) and by a turning angle of θxz∼50° (Fig. 7C, red symbols), showing a strong and immediate avoidance reaction. ‘Reactive’ hornets that were more distant from the nest than dhov (dxz>45 cm), usually approached the nest when shimmering had started. This is expressed by the negative vdxz and θxz values in Fig. 7B,C (yellow symbols), which coincide with the values of the wasp outside the mean hovering distance (Cdxz = 5) in Fig. 6. As expected, 300 ms after the start of shimmering, they turned themselves away from the nest displaying positive vdxz and θxz values. For comparison, the respective time courses of shimmering under the presence of ‘reactive’ hornets display that the shimmering waves were stronger and more repetitive if the hornets were nearer to the nest (Fig. 7A, red symbols) than further away (Fig. 7A, yellow symbols).The additional analysis in Fig. 8 shows that the shimmering waves and the behaviors vdxz and θxz of ‘reactive’ hornets' correlate in both phases of the shimmering waves, that is, in the increasing (Fig. 8A,C) and decreasing part (Fig. 8B,D) of the wave, at least if the means of shimmering and hornets' behaviors were taken in pairs per time interval for correlation (Table S1, see section Methods for the discussion about test strategies). The data suggest that the hornets are not only driven away from the nest by the shimmering wave as documented in Fig. 7, but they are also urged to do so over the entire time course of shimmering. Although the original data give only subsignificant trends, the correlations in Fig. 8 based on the time-interval related means allow assuming that the hornets increase their avoidance behavior in the course of 400 ms, as long as the shimmering wave also steps up in strength. In contrast to ‘reactive’ hornets, ‘nonreactive’ hornets did not show any residual responsiveness to shimmering (Fig. 7D–F, 8E–H).10.1371/journal.pone.0003141.g008Figure 8The correlations of the waving strength with the flight behaviors of ‘reactive’ (A–D) and ‘non-reactive’ (E–H) hornets over the time course of shimmering (scenario B).Abscissas, the waving strength W assessed by the number of abdomen-shaking bees per frame; ordinates, the hornets' behaviors measured by the parameter distance velocity vdxz (A–B;E–F) and turning angle θxz (C–D;G–H) using the data of ‘reactive’(A–D) and ‘non-reactive’ (E-H) episodes (cf Fig. 7). Red circles and red lines refer to hornets which were near the nest (dxz<45 cm) in the pre-wave period, yellow circles and yellow lines refer to hornets which were further away from the nest (dxz>45 cm). Two time intervals were defined in relation to the time course of shimmering (0–400 ms: A,C,E,G; 400–1000 ms: B,D,F,H; for the coding of the brown shaded areas, see Fig. 7). Circles and bars give arithmetical means and their SEM; thick lines give the regression functions (test parameter, see Table S1), which refer to Multiple Linear Regression (Sigmastat) of the arithmetical means (see Methods).Small-scale and big-scale shimmeringIn scenario A, the flight trajectories of the predatory wasps were sorted according to two classes of arousal conditions for the hovering wasps, in particular to small-scale and big-scale shimmering (Fig. 2,9). To illustrate the hornet's responsiveness to both levels of shimmering in more statistical way, the trajectories of the hornets were synchronized to the onset of shimmering, and their positional x- and y-data pooled in 40 ms intervals. Four nest sectors (dashed lines in Fig. 9A,B) allowed defining four directional classes of flight trajectories according to the mean positions of the wasp in the 400 ms prior to the onset of the wave (Fig. 9A,B).10.1371/journal.pone.0003141.g009Figure 9‘Confusion’ of hornets hovering in front of the experimental honeybee nest as observed in big-scale (red color code) and small-scale (blue color code) episodes in scenario A.(A,B) Mean trajectories of the approaching hornet (for compilation of trajectories, see text); the sectors between the dashed lines define the four divisions of the mean pre-wave flight directions of the hornets for pooling the x- and y-values of the positions of the hornets approaching to the nest in 40 ms intervals. Thick lines, arithmetical means of x- and y-values of the hornet's position; horizontal and vertical bars, SEM. Note, that the trajectories before the onset of the waves (coded by black thick lines) are straighter than after the onset of the waves (coded by red and blue thick lines). (C) Hornet flight behavior under the influence of big-scale (full red circles) and small-scale (open blue circles) shimmering waves; ordinate, mDxy (see inset for definition of the mean vector length mDxy; grey segment, the defined turning range), calibrated between 0 and 1, the resulting scalar rel mDxy is a measure of direction fidelity of the hovering wasp (circles and vertical bars, arithmetical means±SEM). The data show that the direction fidelity of predatory hornets is lowered by shimmering waves, by small-scale waves stronger than by big-scale waves; stars refer to significant (P<0.05, one-way ANOVA test) differences between the responses of the hornet to big-scale (n = 20) and small-scale (n = 33) waves per time interval. (D) The mean time courses of big-scale (red line) and small-scale (blue line) waves (cf. Fig. 2); ordinate, relative abdominal thrust activity (mean±SEM); abszissa, the relative experimental time (C,D) at time zero occurred the onset of shimmering; pre-wave sessions are coded by gray or black, shimmering sessions are coded by blue or red (A–D).\nSmall-scale shimmering lowered the direction fidelity of the predatory hornet. The shimmering waves strongly influenced the flight trajectories of the wasp in both, small-scale and big-scale conditions (Fig. 9). During shimmering, the resulting flight paths of the hornet were generally shorter as compared to the pre-wave phase. The reason was not that the hornet would have decelerated its flights, but its turning tendencies deviated much stronger, so that its direction fidelity dropped after the onset of shimmering, which is documented in the x-y plane of scenario A (Fig. 9C, blue symbols). Without any directional preference, the hornet turned away from those bees, which it had obviously decided before to prey on. To the contrary, here, big-scale shimmering had less effect on the hornet (Fig. 9C, red symbols).\nBig-scale shimmering drove the hornet to accelerate. Under the influence of big-scale waves, the hornet in scenario A speeded up, small-scale waves had practically no effect in this respect (Fig. 10). The acceleration pulses terminated after about 400 ms, at the same time when the shimmering waves decreased in amplitude. In other words, big-scale shimmering waves irritate wasps, which speed up for some hundreds of milliseconds. Consequently, these acceleration pulses drive the wasps away from the targets region to which the wasps had directed their predation flights to prey on.10.1371/journal.pone.0003141.g010Figure 10Hornets' responses regarding the flight velocity vxy to big-scale (A,B) and small-scale (C,D) shimmering waves (scenario A).Big-scale and small-scale waves were categorized according to a threshold of 40% of the maximal waving strength (cf. Fig. 2). The peak in flight velocity vxy of big-scale wave episodes (A) from 240 to 360 ms after the onset of wave significantly differs from pre-wave vxy values (*, P<0.05, One-way Repeated Measures ANOVA; 15 episodes). Small-scale waves (C,D) obviously do not affect the flight speed vxy of the hornet. Note, that the acceleration pulse of the hornet (A) coincides with the time courses of big-scale waves (B). Abscissa, experimental time in ms; time zero defines the start of shimmering; lines connect arithmetical means, vertical bars give SEM. The sketches above the graphs symbolize big-scale waves (red-orange areas) as spreading over the nest, while small-scale waves (blue area) remain local processes.DiscussionThe data obtained from this study provide new insights into the complex spatial and temporal patterns of interaction between bee-hawking hornets and Giant honeybees under defence. In support of their open-nesting life-style, Giant honeybees have evolved a set of defence strategies that keep predatory animals, birds, and wasps in particular, off the nest. The obviously most spectacular defence action refers to the recruitment and release of flying defenders [2], [8], [13], [19], which chase vertebrate disturbers or predators away from the nest by counter-attacking them through their stinging behavior. In addition, honeybee colonies pose an even deadly peril to Vespine, Ropalidiini and Polybiini wasps [26] that touch the nest surface. A group of immediately recruited bees seize such intruders, draw them into the bee curtain and heat-ball them to death [15], [17], [19], [20].It has been the conventional view [2], [8], [16] that shimmering has been evolved in particular for colony defence in Giant honeybees. It belongs to categories of strategies [27] that pose practically no risk to the defenders. It consumes far less energy than emergency reactions that are released when defence turns to a matter of physical contact with the enemy. This study first presents quantitative proof that shimmering is an anti-predatory response of giant honeybee colonies to the presence of hornets and demonstrated this for both sides of this prey–predator interaction by a series of finely shaped details: The study showed that shimmering waves became stronger and more frequent the nearer a predatory hornet came to the nest and the faster the hornet flew there. In turn, hornets were more affected by shimmering the nearer they came to the honeybee colony. In this intermezzo, substantial evidence was gained that shimmering does have anti-predatory impact on wasps. While local small-scale shimmering may confuse wasps, which had approached the nest into touching reach, big-scale shimmering that may spread over the whole honeybee nest does have the capacity to repel predatory wasps, but only within a restricted limit away from the bee nest.Shimmering repels hornetsBee-hawking hornets incessantly approach honeybee nests, again and again, to prey on them, without showing the tiniest sign of habituation (which was tested in 335 episodes of shimmering waves). The data obtained do not provide any support for the ‘proximity-avoidance’ hypothesis that would propose that hornets are deterred by the honeybee nests and would avoid its vicinity. Nevertheless, the factor ‘proximity to honeybee nest’ apparently modulates their responsiveness to shimmering, essentially because the honeybee colony also alters its defence response depending upon the distance of the intruder.However, three aspects have been proved in support of the ‘shimmering-repels-wasps’ hypothesis that assumes that bee-hawking hornets show an avoidance response to shimmering when they come too close to the honeybee nest. First, it was proved, particularly in scenario A, that shimmering forces the hornets to accelerate their predation flight for some hundreds of milliseconds, which happened just at the peak time of shimmering. This acceleration pulse is sufficiently long to drive the predator away from the region where it had previously intended to prey on. Second, when the hornets were close to the nest, in particular inside the mean hovering distance, the hornets turned off from the honeybee nest as soon as the shimmering wave became sturdy (scenario B). Third, the hornets were not only affected by the onset of shimmering, their reactivity correlated with the waving during its whole course. When the number of abdomen-thrusting bees increased, the hornet enforced its avoidance reaction by turning stronger and flying faster away from the nest, and when shimmering declined, the hornet reduced and terminated its shimmering-specific avoidance reactions. However, when the hornets were only slightly further away from the honeybee nest, when shimmering occurred, just outside the zone defined as the ‘mean hovering distance’, this repelling goal of shimmering was reversed. Then, shimmering even attracted the hornets, which were inclined to approach the nest as a region of prey.The ultimate goal of shimmering is likely to shelter the nestThus, the capacity of shimmering to repel wasps is limited to a distance of around half a meter from the nest, which equals the mean hovering distance. This restriction in the defensive coverage of colony of Giant honeybees is likely to be associated with the obvious ultimate goal of colony defence to generate a safety zone around the nest that should keep predatory wasps away from the nest, preventing them from catching bees directly from the nest surface. If the wasps, nevertheless, succeed in intruding this shelter zone they should not stay long. This is exactly what was observed during shimmering: When wasps approached the nest, the honeybee colony continuously generated shimmering waves that repeatedly repelled the wasp. Shimmering evidently benefits the honeybee colony because it lowers, factually to zero, the hunting success of those predatory wasps that want to seize bees from the nest surface. The chance to observe wasps in trying to seize giant honeybees from the surface of the nests is quite rare. In a total of estimated 30 minutes of own observation over years and of a hundred of trials by the wasp to catch surface bees, we have not observed any successful grasp.On the other hand, the hornets hardly elicited shimmering when they were more than 50 cm away from the nest. The question here is whether and why giant honeybees do not recognize hornets outside the distance of 50 cm as a threatening peril. It is known that Apis dorsata colonies, which had mobilized their guards for a potential counter attack, recognize predatory birds at much greater distances from the nest than the mean hovering distance of wasps. One of the experimental Apis dorsata nests in Chitwan instantaneously released hundreds of flying defenders when a kite approached it, although this bird was still far more than twenty meters away. Thus, it seems that Giant honeybee colonies have developed specific distance measures for predatory wasps.Visual and pheromone cues of shimmeringAnother question associated to the concept of a shelter zone of Giant honey bee nests arises here: does shimmering deliver only visual cues to the wasps or does it also utilize pheromone channels? It is known that shimmering is linked to chemical scenting [14], but there are arguments that make chemical scenting extremely unlikely to trigger the avoidance response of wasps. First, the release of alarm pheromones in honeybees is accompanied by sting protrusion [15], [27]. Stinging activities do not occur during shimmering, but otherwise, alarm pheromones of honeybees do not prevent hornets from hunting bees [2], [17]. Second, shimmering is accompanied by the release of Nasonov pheromone [14]. After a series of repetitious waves, Giant honeybees open their last inter-tergital gaps of their abdomens, exposing the Nasonov glands. Nasonov scent is a social pheromone and signals to the bees to ‘stay together’ [14], thus preventing single bees from changing their roles into those of guard bees (flying defenders) that would fly off to attack the predator. However, there are reasons that make it is impossible for Nasonov pheromone to trigger the avoidance response of an approaching hornet. Firstly, the exposure of Nasonov glands has been only observed after a series of shimmering episodes [14], but hornets were disturbed by shimmering from the first wave onwards. Additionally, and more important is that the latency of the avoidance reaction of the wasp after the onset of shimmering is less than 100 ms, and is therefore by several orders of magnitude faster than the exposure of Nasonov glands and also faster than the obvious spreading of the pheromone would take. Summarizing, the hornet's avoidance behavior appears to be triggered solely by visual cues of shimmering.Why social waves against wasps have been evolved?If wasps should be hindered to feed directly from the honeybee nest it would be sufficient to organize local groups of surface bees to confuse or misguide them [28]–[30]. Bee-hawking wasps should then learn that it is impossible to catch bees directly from the nest, and they will try to find easier prey near the honeybee nest, such as homing or departing worker bees. Any Mexican wave-like synchronisations of hundreds or even thousands of honeybees would then be seemingly too much ado for this defence purpose. Why then have Giant honeybees evolved shimmering that is an extraordinarily complex trait of group defence unique in the whole animal kingdom? The findings of this study allow assuming that the goal of waving must be associated with the open-nesting life style of the Giant honeybees.As demonstrated earlier (in scenario A), local small-scale waves increased the deviation of turning angles of predatory wasps (lowering their direction fidelity) when they hovered close to a Giant honeybee nest (<20 cm). Here, they may well discern single bees, but such local waves make it difficult for the wasps to concentrate upon them. This aspect is termed ‘confusion’, an anti-predatory strategy often cited as an important mechanism in predatory interactions [28], [31]–[33] as the reduced attack-to-kill ratio experienced by a predator resulting from an inability to single out and attack individual prey in a group. However, ‘confusion’ has been proved only for very few predatory interactions [31]. Thus, small-scale waves as a local response of Giant honeybees on the nest surface would suffice to confuse wasps and to prevent predation. To the contrary, big-scale waves were typically provoked by wasps, which were further away from the nest and flew faster. The results show that big-scale shimmering effects much less ‘confusion’ for the hornets, which is plausible because ‘confusion’ of wasps that are further away from the nest would hardly benefit the colony.Therefore, the authors propose that the wave-like character of shimmering has been evolved obviously not primarily to confuse wasps, but, as shown above, to repel wasps. A possible explanation for this striking capacity probably has two aspects. First, shimmering may reinforce in wasps innate and not habituating fixed action patterns of avoidance. Second, waving possesses a further sophisticated and strikingly ‘convenient’ effect that may also enforce the innate avoidance of the addressee: When the wave of abdomen-thrusting bees spreads over the nest, the wave front stays indeed ‘behind’ the wasp, it factually press-gangs the wasp away from the place it originally wanted to prey on (see movie S4). Subsequently, the wasp is strongly inclined to retreat and fly away from the ‘threatening’ wave front.Hunting outside the shelter zone of Giant honeybee nestsAlthough hornets are continuously attracted to the honeybee nests (by their rich resources of protein and sugar), shimmering effectively prevents the potential predators from collecting bees from the nest surface. Hunting episodes in which hornets continuously attempted to ambush flying bees in front of the honeybee nest were recorded. In thousands of wasp episodes in several honeybee colonies, a single case of successful hunt of a hornet having caught a bee from the nest curtain was not observed. However, hornets do have another kind of hunting success if they focus on ingoing and outgoing bees that cross their hovering range. Bees threatened by the bee-hawking wasp are unprotected by the colony-bound collective defence, but are still able to escape by dodging and fast flight (vxz = 2.25±0.03 ms−1, n = 1855 images, n = 107 flights of bees; scenario B). Most flying bees escaped the wasps successfully; they either flew off the nest at maximal speed or landed as fast as possible on the nest.The movies S2,S3 show one example of an unsuccessful trial of the hornet to catch a flying bee. It was attracted by a homing bee; it chased after her, and directed its flight course and its body's length axis exactly toward its target. The bee was totally upset, made an escape round and tried to land as fast as possible. During this manoeuvre, the hornet came closer to the honeybee nest and was repelled by a shimmering wave. This colony response was provoked by the homing bee in union with the hornet. Within 6 min of observation and after 67 trials of hunting, homing or departing bees, the success rate was 3% (corresponding to two bees), which was not a big loss for the bee colony, but still a benefit for the hornets.The significance of evolving shimmering in the course of evolutionTheoretically, there is a fundamental problem for a prey–predator relation if a defence action of a potential prey, such as shimmering, does not lead to any physical contact with the enemy. Of course, such traits are less risky for the defenders, but they are obviously less dangerous for the predators, which may learn to ignore ‘unperilous’ signals of the potential prey. However, observations clearly demonstrate that repetitious shimmering efficaciously repels the same hornet again and again. Any habituation effect in hornets can be excluded; obviously, they cannot ignore shimmering, although they repeatedly try, without showing any sign of habituation, to hunt their prey from the honeybee nest. Although the wasps decelerate and approach the nest, shimmering interrupts their landing operations, and elicits avoidance reactions, which take the wasps away from the spot of prey. Mostly they are repelled off the nest, at least half a meter or more, from where they start the next hunting episode.Because of their persisting and nonhabituating bee-hawking quirks, it is assumed that wasps envisage honeybee nests as a prey of extraordinary attractiveness. Obviously to avoid widespread wasp predation, honeybees have acquired cavity-nesting abilities (in Southeast Asia: Apis cerana, A. nuluensis; in Eurasia and Africa: Apis mellifera). In particular, Apis cerana and A. nuluensis have strong defence lines against bee-hawking wasps and at their nest entrance they also exhibit shimmering against wasps, although at a far lower level than Giant honeybees [30], [34]. To the contrary, the European honeybee (Apis mellifera) has acquired far less effective abilities to thwart the predation of wasps. This can be demonstrated in direct comparison with A. cerana\n[30], because the European honeybee has been introduced from Europe to South East Asia, and it fails under the strange conditions of widespread wasp predation in Southeast Asia. This particularly illuminates a lack of adaptation in the predator–prey relationship to defend against non-European wasps. Therefore, the co-evolution between Apis cerana and their autochthon bee-hawking wasp predators must have been very intense [17], [18], [30].Hence, it has been extremely important for the open-nesting Giant honeybees during their five million years [1] of coexistence with their wasp predators to evolve defence traits that effectively support their life-style. In this article, it is proved that the first defence line of the giant honeybees against wasp includes shimmering behavior. The reciprocal interactions between Giant honeybees and hornets, during their trials to catch bees from the nest and during the subsequent shimmering of the honeybees, are far more complex than mere stimulus–response behaviors would allow to expect. It seems extremely unlikely that the finely shaped, mutually adjusted behaviors (shown in Figs. 3–\n\n\n\n8) have developed by chance as a kind of general response flexibility. In particular, in view of the observation of experiments with Apis cerana and A. mellifera specimens in the same apiary [30], these mutual responses between Giant honeybees and wasps are suggestive of co-evolutionary adaptation in a predator–prey relationship. The Giant honeybees as potential prey, have acquired the ability to continuously signal to their wasp predators through shimmering ‘keep distance and do not expect a free meal’. The visual cue of shimmering may thus have a combined impact of signaling vigilance of the prey and of an unprofitability [35] of exploiting the honeybee nest. However, the capacity of shimmering goes beyond this goal, as shimmering is proved to actively repel hornets to prevent predation.Supporting InformationTable S1Regressions of the correlations between shimmering behavior and the hornets' behaviors (see Fig. 7).(0.04 MB DOC)Click here for additional data file.Movie S1This video shows two 130 cm-wide nests of the Asian Giant Honeybee Apis dorsata attached to a thick branch of a tree in Assam. The nest in the foreground displays shimmering: a Mexican-wave-like, spiral or circular, pattern. The ‘mouth’ zone of the nest is at the left bottom rim, where forager bees depart, arrive and dance. In contrast, the bees in the periphery are quiescent, but in response to hornet attacks, they produce shimmering (QuickTime; 7.8 MB).(7.94 MB MOV)Click here for additional data file.Movie S2This video shows a nest attached to the ceiling of a water tower in Chitwan, Nepal. A typical, unsuccessful, hunting episode is shown in which a hornet chases a flying bee, but the bee escapes and lands on the nest. The landing of the bee and the manoeuvre of the hornet provokes shimmering, which makes the hornet turn off the nest. The film documents this in original speed (QuickTime; 1.6 MB).(1.59 MB MOV)Click here for additional data file.Movie S3This video shows the same scene as ‘Movie 2’, but in slow motion and contains explanatory text and arrows and the trajectories of the hornet and the flying bee. Light green represents the flying bee, and the shimmering (shaking) nest bees are shown in dark green. The hunting hornet is indicated in red, and in violet when repelled by shimmering. Dark green arrows point out that shimmering repels the hornet. The violet arrow shows the new flight course of the hornet in response to shimmering (QuickTime; 0.9 MB).(0.90 MB MOV)Click here for additional data file.Movie S4This video shows the nest of scenario A under the presence of a wasp. The movement activity of the bees from one frame to the other, assessed by image analysis, were displayed as white areas. The red spot marks the thorax of the hovering wasp. Shimmering waves are repetitively produced in different strengths. The yellow line at the bottom traces this shimmering strength by the sum of the white areas; the numbers on the right bottom side give time of observation in seconds. Note, how the wasp is driven under the influence of the shimmering wave to the upper rim of the nest after 6s (QuickTime; 3.2 MB).(3.27 MB MPG)Click here for additional data file.\n\nREFERENCES:\n1. RuttnerF\n1988\nBiogeography and Taxonomy of Honeybees\nBerlin, Germany\nSpringer Verlag\n2. OldroydBPWongsiriS\n2006\nAsian honey bees\nHarvard University Press\n3. RoepkeW\n1930\nBeobachtungen an indischen Honigbienen, insbesondere an Apis dorsata\nWageningen\nH. Veenman&Zonen\n4. LindauerM\n1952\nÜber die Verständigung bei indischen Bienen.\nZeitschrift für vergleichende Physiologie\nBd.38\nS.521\n557\n5. ButlerCG\n1962\nThe world of the honeybee\nLondon\nCollins\n6. MorseRALaigoFM\n1969\nApis dorsata in the Phillippines.\nPhillip Assoc Entomol\n7. KoenigerNKoenigerG\n1980\nObservations and experiments on migration and dance communication of Apis dorsata in Sri Lanka.\nJ Apic Res\n19[1]\n21\n34\n8. SeeleyTDSeeleyRHAratanakulP\n1982\nColony defence strategies of the honeybees in Thailand.\nEcological monographs\n52\n43\n63\n9. UnderwoodBA\n1990\nSesonal Nesting Cycle and Migration Patterns of the Himalayan Honey Bee Apis laboriosa.\nNational Geographic Research\n6[3]\n276\n290\n10. KastbergerGWinderOHötzlTRaspotnigG\n1996\nBehavioural features of a periodic form of massed flight activity in the Giant honeybee Apis dorsata.\nApidologie\n27\n381\n395\n11. PaarJOldroydBKastbergerG\n2000\nGiant honey bees return to their nest sites.\nNature\n406\n475\n10952300\n12. PaarJOldroyd BPHuettingerEKastbergerGSchaefferS\n2004\nGenetic Structure of an Apis dorsata Population: The significance of migration and colony aggregation.\nJournal of Heredity\n95(2)\n119\n126\n15073227\n13. KastbergerGSharmaDK\n2000\nThe predator-prey interaction between blue-bearded bee eaters (Nyctyonis athertoni) and Giant honeybees (Apis dorsata).\nApidologie\n31\n727\n736\n14. KastbergerGRaspotnigGBiswasSWinderO\n1998\nEvidence of Nasonov scenting in colony defence of the Giant honeybee Apis dorsata.\nEthology\n104\n27\n37\n15. BreedMDGuzmán-NovoaEHuntGJ\n2004\nDefensive Behavior of Honey Bees: Organisation, Genetics, and Comparison with Other Bees.\nAnn Rev Entomol\n49\n271\n298\n14651465\n16. KoenigerNFuchsS\n1975\nZur Kolonieverteidigung der asiatischen Honigbienen.\nTierpsychologie\n37\n99\n106\n17. OnoMIgarashiTOhnoESasakiM\n1995\nUnusual thermal defense by a honeybee against mass attack by hornets.\nNature\n377\n334\n36\n18. KenTHepburnHRRadloffSEYushengYYiqiuLDanyinZNeumannP\n2005\nHeat-balling wasps by honeybees.\nNaturwissenschaften\n92\n492\n495\n16151794\n19. KastbergerGWinderOSteindlK\n2001\nDefence strategies in the Giant honeybee Apis dorsata.\nProceedings of the Deutsche Zoologische Gesellschaft, Osnabrück\n94.1\n7\n20. KastbergerGStachlR\n2003\nInfrared imaging technology and biological applications.\nBehavior Research Methods, Instruments, & Computers\n35\n429\n439\n21. WicklerW\n1968\nMimicry in plants and animals\nNew York\nMcGraw-Hill\n22. OldroydBSmolenskiALawlerSEstoupACrozierR\n1995\nColony aggregations in Apis mellifera L.\nApidologie\n26\n119\n130\n23. SakagamiSF\n1960\nPreliminary report on the specific difference of behaviour and other ecological characters between European and Japanese honeybees.\nActa Hymenoptera\n1\n171\n198\n24. SchmelzerEHartbauerMKastbergerG\n2004\nDefence waving in the Giant honeybee (Apis dorsata).\nProceedings of the Deutsche Zoologische Gesellschaft, Osnabrück\n94.1\n71\n25. FarkasIHelbingDVicsekT\n2002\nSocial behaviour: Mexican waves in an excitable medium.\nNature\n419\n131\n132\n12226653\n26. DasBPGuptaVK\n1989\nThe social wasps of India and the adjacant countries (Hymenoptera: Vespidae).\nOriental Insects Monograph 11, The Association for the study of oriental insects Gainesville Florida\n27. CreweRMHastingsH\n1976\nProduction of pheromones by workers of Apis mellifera adansonii.\nJ Api Res\n15\n149\n154\n28. Eibl-EibesfeldtJ\n1978\nGrundriß der vergleichenden Verhaltensforschung\nMünchen\nPiper\n29. ImmelmannK\n1982\nWörterbuch der Verhaltensforschung\nBerlin, Hamburg\nVerlag Paul Parey\n30. TanKRadloffSELiJJHepburnHRYangMXZhangLJNeumannP\n2007\nBee-hawking by the wasp Vespa velutina, on the honeybees Apis cerana and A. mellifera.\nNaturwissenschaften\n94\n469\n472\n17235596\n31. JeschkeJMTollrianR\n2007\nPrey swarming: which predators become confused and why?\nAnimal Behaviour\n74\n387\n393\n32. KrauseJRuxtonGD\n2002\nLiving in Groups\nOxford\nOxford University Press\n33. LandeauLTerborghJ\n1986\nOddity and the ‘confusion effect’ in predation.\nAnimal Behaviour\n34\n1372\n1380\n34. KoenigerNKoenigerGGriesMTingekSKelituA\n1996\nObservations on colony defense of Apis nuluensis Tingek, Koeniger and Koeniger, 1996 and predatory behavior of the hornet, Vespa multimaculata Pérez, 1910.\nApidologie\n27(5)\n341\n352\n35. CaroT\n2005\nAntipredator defenses in birds and mammals\nChicago (IL)\nUniversity of Chicago Press"
+}

batch_8/PMC2528027.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2528027",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2528027\nAUTHORS: Deb Kumar Mojumder, Theodore G. Wensel, Laura J. Frishman\n\nABSTRACT:\nPurposeIntracellular free calcium ions (Ca2+) are an important element in retinal ganglion cell response. Two major EF-hand (E-helix-loop-F-helix-hand) calcium binding proteins in the retina, calretinin and calbindin-28 kDa, are important buffers of intracellular free Ca2+ in neurons, and may also serve as Ca2+-dependent regulators of enzymes and ion channels.MethodsThis study used immunohistochemistry to investigate the subcellular expression patterns of calretinin and calbindin-28 kDa, in the soma, dendrites, and the axonal compartment of rat retinal ganglion cells.ResultsAntibodies for calretinin and calbindin-28 kDa labeled different cell populations in the retinal ganglion cell layer. In this layer, calretinin labeled a larger number of cells compared to calbindin-28 kDa, many, but not all, of which were displaced amacrine cells. The calbindin-28 kDa immunopositive neurons were distinct in that their somata were peripherally encircled by microtubule associated protein 1 (MAP-1) or neurofilament-200 kDa subunit (NF-200 kDa) immunofluorescence. Although somata of retinal ganglion cells contained these calcium binding proteins, neither protein was found in the dendrites or initial segments of the axons. However, both were expressed in the ganglion cell axons in nerve fiber layer. Calretinin and calbindin-28 kDa staining overlapped in some fibers and not in others. Calretinin immunofluorescence was concentrated in discrete axonal regions, which showed limited staining for calbindin-28 kDa or for NF200 kDa, suggesting its close proximity to the plasma membrane.ConclusionsThere is a clear compartmentalization of calbindin-28 kDa and calretinin distribution in retinal ganglion cells. This suggests that the two calcium binding proteins perform distinct functions in localized calcium signaling. It also indicates that rather than freely diffusing through the cytoplasm to attain a homogeneous distribution, calbindin-28 kDa and calretinin must be bound to cellular structures through interactions that are likely important for their functions.\n\nBODY:\nIntroductionRetinal ganglion cells (RGCs), the final output neurons of the retina, gather visual information from bipolar cells and amacrine cells by synaptic inputs from these neurons. They encode visual signals into Na+-dependent action-potentials that are transmitted along the optic nerve to higher visual centers in the brain. Both low-threshold and high-threshold Ca2+ channels present in RGCs contribute to their responses (for a review, see [1]). Indirectly, Ca2+ via Ca2+-activated K+ channels present in RGCs [2,3] can contribute to K+-dependent after-hyperpolarization following action potentials, which in turn can control excitability and firing patterns of neurons [4,5]. In the dendrites of RGCs, synaptic currents have been found to activate T-type calcium channels [6,7] which can augment and shape transient synaptic responses [8]. Changes in intracellular Ca2+ can also modulate ion channels, signaling cascades, and neurotransmitter receptors locally [2,9-17]. Impaired regulation of Ca2+ by calcium-binding proteins has been suggested to contribute to neurodegenerative processes [18,19], and changes in intracellular Ca2+ in RGCs have been proposed to play a role in excitatory neurotoxicity [20], inactivation of calpain [21] and other proteases, and in apoptotic cell death [22,23].Changes in intracellular Ca2+ are modulated by calcium binding proteins (CBPs) that act as Ca2+ buffers, and these buffers are the major determinants of the kinetics of fluctuations in intracellular Ca2+ (for a review, see [24]). Calretinin and calbindin-28 kDa belong to a family of low molecular weight CBPs expressed in the retina and nervous system of vertebrates [25-30]. These proteins share approximately 59% sequence identity and 77% similarity (Figure 1B). Each has six E-helix-loop-F-helix-hand (EF)-hand motifs (Figure 1A), but only four are functional in calbindin-28 kDa and only five are active in calretinin [31,32].Figure 1Schematic representation of calretinin and calbindin-28 kDa proteins and their sequence identity. A: Shown is a schematic representation of calretinin and calbindin-28 kDa proteins. The red blocks mark the E-helix-loop-F-helix-hand (EF) hand regions within each molecule. B: Alignment of the amino acid sequences of rat calretinin and calbindin-28 kDa molecules is based on NCBI accession numbers P47728 and P07171 respectively. Protein sequences were obtained from the NCBI protein database. C: Western blots for different calbindin-28 kDa (CB) and calretinin (CR) antibodies for rabbit (R) and mouse (M) are shown. Both calbindin-28 kDa and calretinin antibodies recognized a single protein band close to 26 kDa. The blot on the far right used antibodies for both calretinin (AB148) and calbindin-28 kDa (300). The arrow indicates the putative calbindin-28 kDa-positive band below the thicker calretinin positive band.Despite their similar amino-acid sequence, these two proteins are different in many respects. Structurally, they have disparate domain organizations of their EF-hand motifs [31], and functionally, they interact with different partners in various cells. For example, in calcium signaling pathways, calbindin-28 kDa interacts with caspase-3 [33] whereas calretinin interacts with cytoskeletal components [34] and basic helix–loop–helix transcription factors [35]. Under pathological conditions, such as in response to ischemia and reperfusion, their levels in RGCs are differentially altered [36]. Their distinctive functions are highlighted by their presence in distinct neuronal populations in the central nervous system (for example [27,37-39]), where they may serve unique roles.The goal of the present study was to go beyond the previous studies that investigated the distribution of calretinin and calbindin-28 kDa in the rat retina [27,40] and to examine the cellular and subcellular distributions of these proteins in the ganglion cell layer. This study shows that calbindin-28 kDa and calretinin have distinct compartmentalization in RGCs. This suggests that structurally certain intracellular quantities of these two CBPS must be bound to cellular structures. Functionally these bound proteins could influence subcellular Ca2+ signaling and local Ca2+ dynamics.MethodsAnimals for immunohistochemical studiesStudies were performed using 16 1-1.5-year-old Brown Norway rats (Rattus norvegicus; Charles River Laboratories, Wilmington, MA) and two three-month-old Sprague-Dawley rats (Rattus norvegicus; Charles River Laboratories, Wilmington, MA). All animals were maintained on a 12 h:12 h light-dark cycle. Food and water were available ad libitum. Animals were light-adapted before euthanasia—in room light that was rod saturating, at least 1 scot cd.m2. The Brown Norway rats were initially anesthetized using an intramuscular injection of 86 mg/kg ketamine and 6.5 mg/kg xylazine (Both drugs were from Vedco, St. Joseph, MO) and then euthanized by an intramuscular injection of 150–200 mg/kg pentobarbital. The two Sprague-Dawley rats were used for immunoblotting, and were euthanized by CO2 inhalation. All animal procedures and protocols conformed to the United States Public Health Service and Institute for Laboratory Animal Research guidelines and were approved by both the University of Houston Institutional Animal Care and Use Committee as well as the Baylor College of Medicine Institutional Animal Care and Use Committee.Tissue preparation for immunohistochemistryFollowing euthanasia, the eyes were rapidly excised from the orbit. A portion of the superior rectus muscle was left to indicate the superior pole of the globe. The corneas were slit with a razor blade, the lens was removed, and eyes were immersed in 4% paraformaldehyde in 0.1 M sodium cacodylate buffer (pH 7.4) for 1 h at 4 °C. Following fixation, eyes were rinsed in phosphate buffered saline (PBS; pH 7.4) and cryoprotected in 30% sucrose in 1X PBS overnight. The next day, the eyes were embedded in Tissue-Tek OCT Compound (Tissue-Tek, Hatfield, PA) and fast frozen in liquid nitrogen. Vertical cryostat sections of 10–12 µm thickness were cut parallel to a plane tangent to the corneal surface at the center of the pupil and collected onto Superfrost/Plus microscope slides (Fisherbrand; Fisher Scientific, Pittsburgh, PA). Sections were stored at −20 °C until use.For retinal whole-mounts, the eye was excised and the cornea and lens were removed as described in the previous section. Following removal of the sclera and the retinal pigmented epithelium, the neural retina along with some overlying vitreous was rapidly removed. The retina was rinsed in cold Ames’ medium (Sigma-Aldrich, St. Louis, MO; 4 °C, pH 7.4) and then immediately fixed in 4% paraformaldehyde in 0.1 M cacodylate buffer (pH 7.4) for 5 min at 4 °C. The vitreous humor was removed, and relaxing cuts were made in the retinal margin to allow the retina to flatten. The retina was rinsed in 1X PBS and subsequently incubated with the antibodies diluted to their appropriate concentration (Table 1). The details of the immunolabeling procedure has been described in detail below.Table 1Primary antibodies and antiseraAntigenHostDilutionSourceReferenceCalbindin-28kDaMouse1:1000-1:5000SWANT, Bellinzona, Switzerland (Cat# 300)[37], [27]Calbindin-28kDaRabbit1:1000-1:5000SWANT, Bellinzona, Switzerland (Cat# CB38)[27]CalretininRabbit1:1000-1:5000Chemicon International, Temecula, CA (Cat# AB148)[60], [27]Microtubule Associated Protein-1 (MAP-1)Mouse1:300Sigma-Aldrich, St. Louis, MO (Cat# M4278; Clone HM-1)[58], [59]Neurofilament-200 kDaMouse1:1000Chemicon International, Temecula, CA (Cat# MAB5266; Clone N52)[52], [59]NaV α-subunit, pan-specific (Pan-NaV)Mouse1:500-1:1000Sigma-Aldrich, St. Louis, MO (Cat# S8809; Clone K58/35)[49]NaV1.1 α-subunitMouse1:500-1:1000G. Matthews (SUNY-Stony Brook), J. Trimmer (UC Davis), Antibodies, Inc./NeuroMab, Davis, CA (Cat# 75-023; Clone K74/71)[50]NaV1.2 α-subunitMouse1:500-1:1000J. Trimmer (UC Davis); Antibodies, Inc./NeuroMab, Davis, CA (Cat# 75-024; Clone K69/3)[51]NaV1.6 α-subunitMouse1:500-1:1000J. Trimmer (UC Davis), Antibodies, Inc./NeuroMab, Davis, CA (Cat# 75-026; Clone K87A/10)[48]Specificity, host, dilution, and source information for the primary antibodies and antisera that were used in these studies.Antibodies and antiseraThe antibodies and antisera used are detailed in Table 1. Binding of primary antibodies was detected using fluorescent antisera. The secondary antisera used were raised in goat and specific for either mouse or sheep immunoglobulins and were conjugated to a 1:200 dilution for AlexaFluor488 or AlexaFluor546 (Invitrogen Corporation, Carlsbad, CA).ImmunoblottingAntibody specificity was tested using immunoblots of membrane homogenates. After the Sprague Dawley rats were euthanized by CO2 inhalation, one retina from each rat was rapidly extracted, rapidly frozen in dry ice, and powdered with a pestle. Added to the retina was a homogenization buffer composed of the following: 20 mM HEPES, pH 7.0, 150 mM NaCl, 3 mM MgCl2, 1 mM CaCl2, 1 mM beta-mercaptoethanol, 0.01% NaN3, 1 mM EDTA, 1X protease inhibitors, and solid phenylmethylsulphonyl fluoride (PMSF). The tissue was sonicated and then centrifuged at 12,000x g for 10 min. Protein concentration of the retinal homogenate was quantified using Bradford assay against a known BSA standard [41]. Supernatant containing approximately 70 µg of protein were loaded into each well of 12% SDS polyacrylamide gel. Molecular weight standards (BenchMark™ Protein Ladder; Invitrogen) were run on adjacent lanes. The gels were run at constant current to separate proteins. Proteins were transferred to nitrocellulose membranes and blocked with 5% fat-free milk in Tris-buffered saline with Tween-20 (TBST) buffer with 0.02% NaN3. Each membrane was then incubated with primary antibody, which had been diluted in 5% fat-free milk in TBST with 0.02% NaN3. A 1:1,000 dilution was used for all antibodies. The nitrocellulose membrane was rinsed and incubated in secondary antibody conjugated to 1:10,000 HRP. Protein bands were visualized by enhanced chemiluminescence.ImmunolabelingImmunofluorescent methods used in this study for immunolabeling frozen sections and retinal whole-mounts are described previously [42-45]. Frozen sections were thawed, rinsed in deionized water, treated with 1%–2% NaBH4 to reduce autofluorescence, rinsed in deionized water, followed by 1X PBS. Nonspecific labeling was attenuated with 10% normal goat serum, 5% BSA, 0.5%–1% fish gelatin,and 0.1%–0.5% Triton X-100 in PBS (“blocker”).After removal of excess blocker, the primary antibody was incubated for 24–48 h at 4 °C in blocker. A combination of primary antibodies was applied simultaneously for double labeling experiments. Subsequently, sections were rinsed with 1X PBS, blocked for 30 min at room temperature, and secondary antibody was applied for 1 h at room temperature in blocker. An appropriate combination of secondary antisera was applied simultaneously for double labeling experiments. Sections were rinsed and coverslipped in a fade-retardant mounting medium (Vectashield; Vector Labs or Prolong Gold; Invitrogen) and examined with the microscope. As anticipated, there was no labeling in sections processed substituting normal rabbit serum for rabbit polyclonal primary antisera, nonspecific mouse IgGs, or in the absence of primary antibodies.Rat retinal whole-mounts were immunolabeled free-floating (i.e., were incubated with the appropriate antibodies diluted in “blocker” in an eppendorpf tube, such that they were capable of free movement within the tube). Whole-mounts were treated with 1%–2% NaBH4 for 1–2 min, rinsed in deionized water followed by 1X PBS and incubated in blocker solution for 1 h at room temperature to block nonspecific labeling. Retinas were incubated in primary antibody for 5 days at 4 °C. Retinas were rinsed in 1X PBS for 2 h at room temperature and then incubated free-floating in secondary antibody at room temperature for 1–2 h. Retinas were then rinsed in 1X PBS for 2 h at room temperature, flattened onto microscope slides with the ganglion cell side up, coverslipped with a fade-retardant mounting medium (Prolong Gold; Invitrogen) and examined in the confocal microscope. For each antibody a minimum of three retinas from three different animals were tested. The results of this study were consistent for the antibody concentrations (Table 1) and detergent concentrations of 0.1 to 0.5% Triton X-100 used.ImagingImages were acquired using a Leica TCS SP2 confocal microscope and LCS software (Leica Microsystems, Exton, PA). Images were captured using 20x (NA, 0.7), 63x oil (NA, 1.32), or 63x water immersion (NA, 1.2) objective lenses. Stacks of serial optical sections spaced from 1.5 µm to 6 µm apart in the Z plane were collected. For assessment of labeling in single optical planes, we used 63x objectives to achieve a maximal Z-plane resolution. Images in each fluorescent channel were collected sequentially with laser power and detector sensitivity adjusted to prevent bleed-through of signals between fluorescence channels. The absence of bleed-through between channels was confirmed in sections treated with a single primary antibody and a combination of secondary antibodies imaged using identical settings to verify that only the channel corresponding to the primary antibody showed labeling.Figures were prepared by importing images into Adobe Photoshop 6.0 (Adobe Systems, Inc., Mountain View, CA) and calibrating image scale. To estimate colocalization of immunofluorescence from two different antibodies in the nerve fiber layer, we plotted the channel intensity of each label along its long axis by using the ImageJ software (W.S. Rasband, NIH, Bethesda, MD) and its red-green-blue (RGB)_Profiler plugin (Laummonerie and Mutterer, Institut de Biologie Moléculaire des Plantes, Strasbourg, France). To estimate cell soma size in the RGC layer, the outer border of the cell membrane, stained for NaV, was traced manually from confocal projection of the RGC layer. The area was then measured using ImageJ software (W.S. Rasband, NIH, Bethesda, MD).ResultsImmunoblottingWestern blots using the different primary antibodies against calbindin-28 kDa (300 and CB-38a) and calretinin (AB148) each labeled a single protein-band close to the molecular weight marker band of 26 kDa (Figure 1C), indicating that the antibodies were specific for proteins of that molecular weight. The mouse monoclonal antibody for calbindin-28 kDa, 300, (Figure 1C) used together with the rabbit polyclonal antibody for calretinin, AB148, revealed that the two antibodies recognized specific proteins that had different molecular weights. The single protein band labeled by the calbindin-28 kDa rabbit polyclonal antibody (CB-38a) was too broad to rule out cross-reactivity with calretinin.Labeling for calretinin and calbindin-28 kDaStaining for calretinin and calbindin-28 kDa in vertical sections of the retina was similar to previous results in rats (for example, see [40]). The somata and descending processes of some amacrine cells whose somata resided in the proximal inner nuclear layer (INL) close to the inner plexiform layer (IPL) were brightly stained by calretinin (Figure 2A). A large number of cell bodies in the RGC layer (GCL) were also stained with calretinin antibodies. The calretinin immunopositive cell bodies in the GCL are known to include the displaced amacrine cells in the rat retina [46]. Dendrites from some of the calretinin-positive neurons in the INL and GCL projected toward the narrow calretinin-positive bands in the IPL (arrows in Figure 2A). These calretinin immunopositive dendrites were found to originate from the displaced amacrine cells in the ganglion cell layer in the rat retina [46]. The nerve fiber layer (NFL), where the unmyelinated axons of the RGCs are located, was also immunoreactive for calretinin.Figure 2Calretinin and calbindin-28 kDa immunolabeling of a vertical cryosection. A: Calretinin immunolabeling was present in cell bodies and processes of amacrine cells at the inner nuclear layer (INL)-inner plexiform layer (IPL) border. Calretinin labeling was also present in cell bodies and processes (arrows) in the ganglion cell layer (GCL). Calretinin labeling is also found in three distinct bands in the IPL and retinal ganglion cell (RGC) axons in the nerve fiber layer (NFL; arrowhead). B: Calbindin-28 kDa immunolabeling was present in cell bodies and processes of horizontal cells at the outer plexiform layer (OPL)-INL border. Calbindin-28 kDa also labeled amacrine cells at the INL-IPL border. Some descending processes were seen for some of these neurons (arrow). There are also diffuse calbindin-28 kDa-positive punctate in the IPL. Calbindin-28 kDa labeling is seen in few cells in the GCL as well as in the RGC axons in the NFL (arrowhead). Scale bars represent 20 µm.Intense calbindin-28 kDa labeling was found in horizontal cells and their processes in the outer plexiform layer- inner nuclear layer (OPL-INL) border (Figure 2B) as observed in previous studies [27]. There was no significant staining for calbindin-28 kDa in the mid-INL where the cell bodies of the bipolar cells are located. Amacrine cells at the INL-IPL border showed staining for calbindin-28 kDa. Many of these cells had a single, stout apical process descending into the IPL. These cells did not have any calretinin labeling (see next section). In the IPL, the mouse monoclonal calbindin-28 kDa antibody showed punctate labeling. Calbindin-28 kDa labeling of somata in GCL with the mouse monoclonal antibody was sparse but axons in the NFL were labeled. On double-stained whole-mounts, calretinin and calbindin-28 kDa antibodies labeled different subsets of amacrine cells in the proximal INL (Figure 3) with almost no overlap, indicating that each antibody identified a specific amacrine cell type.Figure 3Calretinin and calbindin-28 kDa are differentially distributed in amacrine cells as seen in retinal whole-mounts. A-C: Double labeling for calretinin (red) and calbindin-28 kDa (green) in the inner nuclear layer shows that labeling for each was present in a distinct set of amacrine cells For the apparent region of overlap (arrowhead) in the overlay of this confocal plane, examination of different z-planes revealed that these were disparate cells located at different depths. Arrows indicate calbindin-28 kDa-immunopositive cells. Scale bar represents 20 µm. Abbreviations: bv is blood vessel.Differential distribution of calretinin and calbindin-28 kDa in GCL/NFLDouble labeling immunofluorescence with mouse monoclonal anticalbindin-28 kDa and rabbit anticalretinin in whole-mounts showed that the two calcium binding proteins were present in the somata of random sets of the neurons in the GCL (Figure 4). In the whole-mounts, processes emanating from some of the calretinin immunopositive cells in the GCL were also stained. When followed through multiple optical sections, it was apparent that these processes were directed away from the GCL and NFL, and toward the IPL, where they merged with the proximal calretinin positive plexus in the IPL (Figure 4 D-F) as was observed in the vertical sections (see Figure 2A). Because this proximal calretinin positive plexus in the IPL also corresponds with the stratification of displaced amacrine cells it is evident that these dendrites originated from the displaced amacrine cells in the GCL and merged with the proximal cholinergic band in the IPL [46]. Calbindin-28 kDa immunopositive somata in the GCL were fewer in number and never showed any colocalization with calretinin. None of the somata staining for calbindin-28 kDa showed calbindin-28 kDa-immunopositive-processes emanating directly from the cell bodyFigure 4Calretinin and calbindin-28 kDa are distinctly distributed in the ganglion cell and nerve fiber layer as seen in retinal whole-mounts. A-C: Double labeling for calretinin (red) and calbindin-28 kDa (green) in the ganglion cell layer (GCL)/nerve fiber layer (NFL) shows that labeling for each was present in a distinct set of neurons. Calbindin-28 kDa positive cell bodies are indicated by arrowheads. Some calretinin-positive neurons show processes (arrow) that ascend distally. Note the discontinuous staining pattern of calretinin in the NFL in contrast to a smoother staining pattern for calbindin-28 kDa. D,E: A single confocal optical section distal to that of A-C shows that the calretinin positive processes (arrow) are directed in the inner plexiform layer (IPL) distally toward a calretinin-immunopositive plexus, a characteristic of displaced amacrine cells. G: Representative calretinin and calbindin-28 kDa double staining in the GCL/NFL is shown. Channel intensity profiles for the red and green channels for straight lines along the long axis (lines a and b in G shown in H and I respectively) show different intensity profiles for calretinin and calbindin-28kDa immunofluorescence. Scale bar represents 20 µm.Calretinin and calbindin-28 kDa antibodies both stained the RGC axons in the NFL, but for the most part the staining patterns did not overlap. In the axonal compartment, calretinin antibodies showed punctuate staining at intermittent locations along the axons, on a background of diffuse immunofluorescence (Figure 4 A,G). The axons that were immunopositive for calbindin-28 kDa were fairly uniformly labeled along their length. The channel intensity profiles along the long axes of two nerve fiber bundles show that relative levels of staining for the two CBPs varied greatly from axon to axon, (Figure 4G-I), indicating that these two proteins are present in differing amounts in different axons. Sharp peaks in the calretinin intensity profile demonstrated that calretinin was concentrated at distinct locations in the axons unlike calbindin-28 kDa (Figure 4H).Dendritic compartments of retinal ganglion cells are devoid of immunostaining for calretinin and calbindin-28kDaCalbindin-28 kDa did not label any processes that ascended distally from the GCL. The processes labeled by calretinin that ascended distally from the GCL were likely from displaced amacrine cells in the GCL, as noted in the previous section.. To characterize further the morphology of the ganglion cells whose somata were stained for calretinin and calbindin-28 kDa, double labeling for microtubule associated protein 1 (MAP-1) and calretinin or calbindin-28 kDa was performed in whole-mounts. MAP-1 is known to label the dendrites of RGCs in rats [47]. In the IPL/GCL MAP-1 labeled the dendrites of RGCs (Figure 5) but for almost their entire length these did not stain for either CBP. Calbindin was found to be present at the very proximal portions of the RGCs but absent more distally (Figure 5G-I). These results indicate that neither calretinin nor calbindin-28 kDa was present in detectable levels in the dendritic compartment of the RGCs.Figure 5Calretinin and calbindin-28 kDa immunofluorescence is not present in the distal dendritic compartment of retinal ganglion cells as seen in retinal whole-mounts. A-C: Double labeling for calretinin (red) and microtubule-associated protein 1 (MAP-1; green) in the ganglion cell layer (GCL)/nerve fiber layer (NFL) shows that MAP-1 positive dendrites are not colabeled with calretinin. Some large retinal ganglion cells (RGCs) that are completely ringed by MAP-1 staining (arrowhead) are not positive for calretinin. RGCs with smaller somata partially ringed with MAP-1 staining are immunopositive for calretinin (arrows). Other brightly stained somata not showing MAP-1 immunofluorescence are the displaced amacrine cells that also stained for calretinin. D-F: Confocal plane showing that the MAP-1 positive (green) dendrites (arrow) do not merge with the calretinin-positive plexus (red) in the inner plexiform layer. G-I: Double labeling for calbindin-28 kDa (red) and MAP-1 (green) in the GCL/NFL shows that some large RGCs that are completely ringed by MAP1 staining (arrowhead) are positive for calbindin-28 kDa. Some RGCs with smaller somata that are incompletely ringed with MAP-1 are also calbindin-28 kDa positive while others are not (arrow). Some neurons incompletely ringed with MAP-1 are also not calbindin-28 kDa positive (arrow). Scale bar represents 20 µm.MAP-1 immunopositive dendrites did not stratify extensively with the calretinin-positive plexuses in the IPL (Figure 5D-F). However, calretinin-positive dendrites emanating from the displaced amacrine cells in the proximal INL and GCL did stratify in the calretinin-positive bands in IPL (Figure 4D-F).MAP-1 completely encircled the periphery of the cell body in some large RGCs (Figure 5A-C,G-I arrowhead). These large RGCs somata were not calretinin positive (Figure 6A-C), but some were calbindin-28 kDa immunopositive (Figure 5G-I). Yet, many smaller RGC somata that were incompletely encircled by MAP-1 immunofluorescence were found to be calretinin positive (arrows, Figure 5A-C) and calbindin-28 kDa negative (arrow, Figure 5G-I). Displaced amacrine cells in the GCL that were brightly stained for calretinin did not have MAP-1 staining in their dendritic processes.Figure 6Comparison of immunofluorescence for the calcium binding proteins, calretinin and calbindin-28 kDa, and voltage-gated sodium channel antibodies in the ganglion cell layer (GCL)/nerve fiber layer (NFL) as seen in retinal whole-mounts. A-C: Calcium binding proteins (CBP) in the NFL are extensively colocalized with Pan-NaV. Some Pan-NaV stained retinal ganglion cell (RGC) somata were also immunopositive for CBPs (arrowhead) while others were not (double arrows). Initial segments of RGCs, (arrow) some of which can be seen emerging from the RGC somata, were immunopositive for Pan-NaV but not colabeled with CBPs. D-F: NaV1.1-immunopositive (green) RGC nerve fiber bundles in the nerve fiber layer (NFL) were colabeled with CBPs (red), but the axon initial segments (arrow) were not. G-I: NaV1.2 immunopositive (green) RGC nerve fiber bundles in the NFL were colabeled with CBPs (red) but not the axon initial segments (arrow). J-K: NaV1.6 immunopositive (green) axon initial segments (arrow) were not colabeled with CBPs (red). Scale bar equals 20 µm. Abbreviations: bv is blood vessel.Immunofluorescence for calretinin and calbindin-28 kDa are absent in the RGC initial segmentsRGC initial segments that emanate from the RGC somata show clustering of specific isoforms of voltage-gated sodium channels [48]. To label these initial segments an antibody specific for all NaV1 α-subunits (Pan-NaV) [49] and those for NaV1.1 [50], NaV1.2 [51], or NaV1.6 [48] α-subunits were used. NaV1.1 and NaV1.2 α-subunit antibodies are also known to label some processes in the IPL, whereas, NaV1.6 α-subunit antibody does not [50]. The initial segments in the GCL/NFL that were immunopositive for Pan-NaV, NaV1.1, NaV1.2, or NaV1.6 α-subunit antibodies did not show immunofluorescence for either calretinin or calbindin-28 kDa, indicating their absence in the initial segments of the RGCs. A combination of staining for calretinin and calbindin-28 kDa is seen in the red channel (Figure 6).Some large neuronal somata were immunopositive for the NaV α-subunits (for example, Figure 6A-C, double arrow); these did not show immunofluorescence for calretinin or calbindin-28 kDa. The Pan-NaV antibody delineated the cell membrane, making it possible to trace the outer margin of the cell soma and calculate its surface area of projection in whole-mounts as an estimate of cell size. Of the cell counted, sizes varied for CBP immunopositive and CBP immunonegative neurons (Figure 7). The smaller sized somata most likely included displaced amacrine cells. In addition, 15% of CBP immunonegative neurons and 7.5% of CBP immunopositive neurons showed somatic sizes that were greater than 300 µm2. Based on somatic size both the CBP and non-CBP immunopositive cells represented a heterogeneous population of cells,Figure 7Size distributions of Calcium-binding protein (CBP) immunopositive and immunonegative neurons are shown. The histogram shows the surface area of projection of the somata of neurons in the retinal ganglion cell (RGC) layer from retinal whole mounts that were immunopositive for Pan-NaV only (n=40, red) or Pan-NaV and calcium binding proteins (CBP; n=40, black). The histogram is based on projections of all optical planes corresponding to the RGC layer from five midperipheral retinal areas (256 μm × 256 μm).Distribution of calretinin and calbindin-28 kDa relative to NF-200 kDa in the axonal compartmentNF-200 kDa [52] was used together with the CBP antibodies to label the axons of the RGCs. In Figure 8A-C, the discrete regions in the axons where calretinin was concentrated showed limited immunofluorescence for NF 200 kDa, indicating different subcellular localization for these proteins. In contrast, patterns of calbindin-28 kDa immunofluorescence colocalized with NF-200 kDa in several locations (Figure 8J), but showed limited colocalization in other areas (Figure 8I). This demonstrates that calbindin-28 kDa was present in significant quantities in many but not all axons, where they were present in close proximity to neurofilaments in the cytoplasm. No CBP was visible in portions of axons immediately juxtaposed to the cell body, again indicating their absence in the initial segments of the axons. These results show that the differential distribution of calretinin and calbindin-28 kDa, noted in Figure 4G, could be because of their different subcellular locations in the RGC axons. High concentrations of calbindin-28 kDa together with lower diffuse concentrations of calretinin were most likely present in the cytoplasmic core in close proximity to NF-200 kDa, whereas high concentrations of calretinin were most likely present at discrete locations on the RGC membrane.Figure 8Calretinin, calbindin-28 kDa, and NF-200 kDa immunofluorescence in the ganglion cell layer (GCL)/nerve fiber layer (NFL) as seen in retinal whole-mounts. A-C: Double labeling for calretinin (red) and NF-200 kDa (green) shows retinal ganglion cell (RGC) somata, surrounded peripherally by light NF-200 kDa immunofluorescence (arrows), that were not immunopositive for calretinin. Calretinin immunofluorescence was present at discrete locations intermittently along the long axis of the RGC nerve fiber, whereas the NF-200 kDa immunofluorescence was uniform. Channel intensity profiles for the red and green channels along the long axis (lines a and b in C shown in D and E respectively) revealed that for regions on the long axis where staining for calretinin was prominent, staining for NF-200 kDa was less prominent (D) and vice versa (E). F-H: Double labeling for calbindin-28 kDa (red) and NF-200 kDa (green) showed that calbindin-28 kDa-positive immunofluorescence was smoothly distributed in the nerve fibers similar to NF-200 kDa. RGC somata that were surrounded peripherally by light NF-200 kDa immunofluorescence (arrow) were also stained with calbindin-28 kDa while for others (arrowhead) staining was less prominent. Channel intensity profiles for the red and green channels for straight lines along the long axis (lines d and e in H shown in I and J respectively) presented some region where immunofluorescence for NF-200 kDa was prominent while that for calbindin-28 kDa was less prominent (I) and others where the intensity profiles were similar (J). Scale bar represents 20 µm.Abbreviations: bv is blood vessel.Similar to MAP-1 immunofluorescence, NF-200 kDa immunofluorescence was detected circumferentially around the somata of some ganglion cells (Figure 8B,G). The somata of these ganglion cells showed scant calretinin immunofluorescence; however, some (arrow, Figure 8F-H), but not others (arrowhead, Figure 8F-H), expressed calbindin-28 kDa, indicating another morphological difference between the RGCs that expressed calbindin-28 kDa and those that expressed calretinin. Along the length of axons in the NFL, relatively uniform neurofilament staining was largely coincident with calbindin-28 kDa staining but contrasted with the very punctate calretinin staining.Antibody specificityAlthough some cross reactivity of the calbindin-28 kDa rabbit polyclonal antibody (CB-38a) with calretinin cannot be ruled out by the immunoblotting results, our interpretation that calbindin-28 kDa labeled a specific subset of ganglion cells is straightforward; because the calretinin antibody did not show cross-reactivity for calbindin-28kDa as revealed by its failure to label the calbindin-28 kDa immunopositive horizontal cells or the calbindin-28 kDa-immunopositive amacrine cells (Figure 3). In addition, the calbindin-28 kDa-immunolabeled RGCs have morphology clearly distinct from those staining with calretinin, as seen by colabeling with MAP-1 and NF-200 kDa antibodies. Thus, under the conditions used for immunostaining, the antibodies appear to be highly specific, and clearly reveal distinct staining patterns in the RGC axons.DiscussionThe major new finding reported here is that the subcellular distribution of both calretinin and calbindin-28 kDa are non-uniform in RGCs. In both cases, even in cells where they brightly stain the soma, staining is excluded from the dendritic arbors, with, in most cases, the CBP excluded even from the dendritic trunks closest to the soma. Similarly, although staining was found for both CBP on either side of the initial segments of RGC axons—that is, in the somata and nerve fibers—they were excluded from the initial segments. Thus even though the cytoplasmic staining for these CBP seems reasonably uniform, suggestive of freely diffusing soluble proteins, there are clearly mechanisms, which must involve either local binding sites or active transport, that exclude these CBP from certain regions of the cell and concentrate them in others. This phenomenon is particularly striking in the case of calretinin staining of RGC axons, which reveals punctate spots of high concentration superimposed on a diffuse background of what we presume to be cytoplasmic staining. These results have been summarized as a schematic in Figure 9.Figure 9Schematic summarizing the subcellular distribution of calbindin-28 kDa and calretinin in the different compartments of the retinal ganglion cell. Labeling for calbindin-28 kDa or calretinin was absent from the dendritic compartment of the retinal ganglion cells (white). Many, but not all, retinal ganglion cell soma show labeling for either calbindin-28 kDa or calretinin (red). The initial segments of the retinal ganglion cells where the voltage-gated sodium channels (NaVs in blue) are clustered are not labeled by either calbindin-28 kDa or calretinin. Calretinin immunolabeling is concentrated at distinct locations on the retinal ganglion cell axon (dashed red line) where they are most likely membrane bound. Calbindin-28 kDa and calretinin are either diffusely coexpressed or differentially expressed in the axons (diffuse red dots), but are both absent from the initial segments.The mechanisms by which these proteins are localized are not known, but it seems likely that their localized distributions are linked to their functions. Both proteins are often regarded as contributing to shaping neuronal responses largely by serving as Ca2+ buffers [24]. Such buffers can determine the kinetics of changes in local Ca2+ concentrations; in the case of transient fluxes of Ca2+ into the cytoplasm through channels in the plasma membrane or endoplasmic reticulum, buffers can govern the amplitudes of such changes, as well as their time courses. The kinetics and buffering power are determined both by the intrinsic Ca2+-binding properties of the CBP (numbers of sites, kinetic and equilibrium constants for Ca2+ binding), and by the local concentrations. Calcium binding kinetics for calretinin have not been fully elucidated [53,54] but are expected to resemble calbindin-D28k for its relatively low-affinity buffering capacity and its fast calcium-binding kinetics [55,56]. It is possible that calretinin and calbindin-28 kDa may have different binding properties to other proteins which enable it to be localized in distinct loci in the RGCs to influence local fluxes of Ca2+. Thus it seems likely that the RGC dendrites and axonal initial segments either require lower Ca2+ buffering capacity than the other regions of the cell, or that alternative CBPs serve to replace calretinin and calbindin-28 kDa in these regions. Clearly, the same must be true of those RGCs that do not contain either of these proteins in detectable amounts. It is important to note here that other calcium-binding proteins, such as parvalbumin [57], are also known to be expressed in rat RGC’s but these were not tested in this study.In addition to simply buffering Ca2+, CBPs can be involved in direct regulation of signaling pathways by binding to and modifying the activities of cellular proteins in a Ca2+-dependent way. Whether calretinin and calbindin-28 kDa play such roles in RGCs is not known, but their irregular distribution can only be explained in terms of relatively high affinity binding to other proteins. It is possible that CBPs regulate these proteins, rather than simply being localized by them.\n\nREFERENCES:\n2. WangGYRobinsonDWChalupaLMCalcium-activated potassium conductances in retinal ganglion cells of the ferret.J Neurophysiol19987915189425186\n3. KlockerNOliverDRuppersbergJPKnausHGFaklerBDevelopmental expression of the small-conductance Ca(2+)-activated potassium channel SK2 in the rat retina.Mol Cell Neurosci2001175142011273646\n4. VergaraCLatorreRMarrionNVAdelmanJPCalcium-activated potassium channels.Curr Opin Neurobiol1998832199687354\n5. SahPCa(2+)-activated K+ currents in neurones: types, physiological roles and modulation.Trends Neurosci19961915048658599\n6. HendersonDMillerRFLow-voltage activated calcium currents in ganglion cells of the tiger salamander retina: Experiment and simulation.Vis Neurosci200724375117430608\n7. HendersonDMillerRFEvidence for low-voltage-activated (LVA) calcium currents in the dendrites of tiger salamander retinal ganglion cells.Vis Neurosci2003201415212916736\n8. MillerRFStenbackKHendersonDSikoraMHow voltage-gated ion channels alter the functional properties of ganglion and amacrine cell dendrites.Arch Ital Biol20021403475912228988\n9. WangGYOlshausenBAChalupaLMDifferential effects of apamin- and charybdotoxin-sensitive K+ conductances on spontaneous discharge patterns of developing retinal ganglion cells.J Neurosci19991926091810087074\n10. Abdel-MajidRMTremblayFBaldridgeWHLocalization of adenylyl cyclase proteins in the rodent retina.Brain Res Mol Brain Res2002101627012007833\n11. AhlijanianMKWestenbroekRECatterallWASubunit structure and localization of dihydropyridine-sensitive calcium channels in mammalian brain, spinal cord, and retina.Neuron19904819322163262\n12. AkopianAGabrielRWitkovskyPCalcium released from intracellular stores inhibits GABAA-mediated currents in ganglion cells of the turtle retina.J Neurophysiol1998801105159744925\n13. KarschinALiptonSACalcium channels in solitary retinal ganglion cells from post-natal rat.J Physiol1989418379962559971\n14. LiuYLasaterEMCalcium currents in turtle retinal ganglion cells. I. The properties of T- and L-type currents.J Neurophysiol199471733428176435\n15. LohmannCMyhrKLWongROTransmitter-evoked local calcium release stabilizes developing dendrites.Nature20024181778112110889\n16. ShenWSlaughterMMMetabotropic and ionotropic glutamate receptors regulate calcium channel currents in salamander retinal ganglion cells.J Physiol1998510815289660896\n17. LiptonSATauckDLVoltage-dependent conductances of solitary ganglion cells dissociated from the rat retina.J Physiol1987385361912443669\n18. HeizmannCWBraunKChanges in Ca(2+)-binding proteins in human neurodegenerative disorders.Trends Neurosci199215259641381122\n19. SchaferBWHeizmannCWThe S100 family of EF-hand calcium-binding proteins: functions and pathology.Trends Biochem Sci199621134408701470\n20. SucherNJLiptonSADreyerEBMolecular basis of glutamate toxicity in retinal ganglion cells.Vision Res1997373483939425525\n21. McKernanDPGuerinMBO'BrienCJCotterTGA key role for calpains in retinal ganglion cell death.Invest Ophthalmol Vis Sci20074854203018055788\n22. DasAGarnerDPDel ReAMWoodwardJJKumarDMAgarwalNBanikNLRaySKCalpeptin provides functional neuroprotection to rat retinal ganglion cells following Ca2+ influx.Brain Res200610841465716600192\n23. OkaTTamadaYNakajimaEShearerTRAzumaMPresence of calpain-induced proteolysis in retinal degeneration and dysfunction in a rat model of acute ocular hypertension.J Neurosci Res20068313425116528750\n24. BaimbridgeKGCelioMRRogersJHCalcium-binding proteins in the nervous system.Trends Neurosci19921530381384200\n25. EllisJHRichardsDERogersJHCalretinin and calbindin in the retina of the developing chick.Cell Tissue Res19912641972081878940\n26. HamanoKKiyamaHEmsonPCManabeRNakauchiMTohyamaMLocalization of two calcium binding proteins, calbindin (28 kD) and parvalbumin (12 kD), in the vertebrate retina.J Comp Neurol1990302417242289978\n27. PasteelsBRogersJBlachierFPochetRCalbindin and calretinin localization in retina from different species.Vis Neurosci199051162125465\n28. PochetRPasteelsBSeto-OhshimaABastianelliEKitajimaSVan EldikLJCalmodulin and calbindin localization in retina from six vertebrate species.J Comp Neurol1991314750621816273\n29. RohrenbeckJWassleHHeizmannCWImmunocytochemical labelling of horizontal cells in mammalian retina using antibodies against calcium-binding proteins.Neurosci Lett198777255603302765\n30. UesugiRYamadaMMizuguchiMBaimbridgeKGKimSUCalbindin D-28k and parvalbumin immunohistochemistry in developing rat retina.Exp Eye Res19925449191623935\n31. PalczewskaMGrovesPBattaGHeiseBKuznickiJCalretinin and calbindin D28k have different domain organizations.Protein Sci200312180412493841\n32. LinseSThulinEGiffordLKRadzewskyDHaganJWilkRRAkerfeldtKSDomain organization of calbindin D28k as determined from the association of six synthetic EF-hand fragments.Protein Sci199762385969385641\n33. BellidoTHueningMRaval-PandyaMManolagasSCChristakosSCalbindin-D28k is expressed in osteoblastic cells and suppresses their apoptosis by inhibiting caspase-3 activity.J Biol Chem2000275263283210835428\n34. MarilleyDSchwallerBAssociation between the calcium-binding protein calretinin and cytoskeletal components in the human colon adenocarcinoma cell line WiDr.Exp Cell Res2000259122210942575\n35. ZimmermannLSchwallerBMonoclonal antibodies recognizing epitopes of calretinins: dependence on Ca2+-binding status and differences in antigen accessibility in colon cancer cells.Cell Calcium200231132511990296\n36. KwonOJKimJYKimSYJeonCJAlterations in the localization of calbindin D28K-, calretinin-, and parvalbumin-immunoreactive neurons of rabbit retinal ganglion cell layer from ischemia and reperfusion.Mol Cells2005193829015995355\n37. CelioMRCalbindin D-28k and parvalbumin in the rat nervous system.Neuroscience1990353754752199841\n38. HofPRGlezerIICondéFFlaggRARubinMBNimchinskyEAVogt WeisenhornDMCellular distribution of the calcium-binding proteins parvalbumin, calbindin, and calretinin in the neocortex of mammals: phylogenetic and developmental patterns.J Chem Neuroanat1999167711610223310\n39. BastianelliEDistribution of calcium-binding proteins in the cerebellum.Cerebellum200322426214964684\n40. MengerNSeidenbecherCIGundelfingerEDKreutzMRThe cytoskeleton-associated neuronal calcium-binding protein caldendrin is expressed in a subset of amacrine, bipolar and ganglion cells of the rat retina.Cell Tissue Res1999298213210555536\n41. BradfordMMA rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding.Anal Biochem19767224854942051\n42. SherryDMWangMMFrishmanLJDifferential distribution of vesicle associated membrane protein isoforms in the mouse retina.Mol Vis200396738814685145\n43. SherryDMWangMMBatesJFrishmanLJExpression of vesicular glutamate transporter 1 in the mouse retina reveals temporal ordering in development of rod vs. cone and ON vs. OFF circuits.J Comp Neurol20034654809812975811\n44. SherryDMMitchellRStandiferKMdu PlessisBDistribution of plasma membrane-associated syntaxins 1 through 4 indicates distinct trafficking functions in the synaptic layers of the mouse retina.BMC Neurosci200675416839421\n45. MojumderDKFrishmanLJOttesonDCSherryDMVoltage-gated sodium channel alpha-subunits Na(v)1.1, Na(v)1.2, and Na(v)1.6 in the distal mammalian retina.Mol Vis20071321638218079688\n46. GabrielRWitkovskyPCholinergic, but not the rod pathway-related glycinergic (All), amacrine cells contain calretinin in the rat retina.Neurosci Lett1998247179829655622\n47. OkabeSShiomuraYHirokawaNImmunocytochemical localization of microtubule-associated proteins 1A and 2 in the rat retina.Brain Res1989483335462706525\n48. BoikoTVan WartACaldwellJHLevinsonSRTrimmerJSMatthewsGFunctional specialization of the axon initial segment by isoform-specific sodium channel targeting.J Neurosci20032323061312657689\n49. RasbandMNPelesETrimmerJSLevinsonSRLuxSEShragerPDependence of nodal sodium channel clustering on paranodal axoglial contact in the developing CNS.J Neurosci19991975162810460258\n50. Van WartABoikoTTrimmerJSMatthewsGNovel clustering of sodium channel Na(v)1.1 with ankyrin-G and neurofascin at discrete sites in the inner plexiform layer of the retina.Mol Cell Neurosci2005286617315797713\n51. RasbandMNTaylorCMBansalRParanodal transverse bands are required for maintenance but not initiation of Nav1.6 sodium channel clustering in CNS optic nerve axons.Glia2003441738214515333\n52. ShawGWeberKThe structure and development of the rat retina: an immunofluorescence microscopical study using antibodies specific for intermediate filament proteins.Eur J Cell Biol1983302193211596496\n53. SchwallerBMeyerMSchiffmannS'New' functions for 'old' proteins: the role of the calcium-binding proteins calbindin D-28k, calretinin and parvalbumin, in cerebellar physiology. Studies with knockout mice.Cerebellum200212415812879963\n54. EdmondsBReyesRSchwallerBRobertsWMCalretinin modifies presynaptic calcium signaling in frog saccular hair cells.Nat Neurosci200037869010903571\n55. NagerlUVNovoDModyIVergaraJLBinding kinetics of calbindin-D(28k) determined by flash photolysis of caged Ca(2+).Biophys J20007930091811106608\n56. HackneyCMMahendrasingamSPennAFettiplaceRThe concentrations of calcium buffering proteins in mammalian cochlear hair cells.J Neurosci20052578677516120789\n57. OkoyamaSMoriizumiTOnset of calbindin-D 28K and parvalbumin expression in the lateral geniculate complex and olivary pretectal nucleus during postnatal development of the rat.Int J Dev Neurosci2001196556111705670\n58. HuberGMatusAImmunocytochemical localization of microtubule-associated protein 1 in rat cerebellum using monoclonal antibodies.J Cell Biol198498777816363428\n59. MojumderDKSherryDMFrishmanLJContribution of voltage-gated sodium channels to the b-wave of the mammalian flash electroretinogram.J Physiol200858625518018388140\n60. WinskyLHarveyJAMcMasterSEJacobowitzDMA study of proteins in the auditory system of rabbits using two-dimensional gels: identification of glial fibrillary acidic protein and vitamin D-dependent calcium binding protein.Brain Res1989493136462776001"
+}

batch_8/PMC2528895.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2528895",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2528895\nAUTHORS: Michael R Law, Sumit R Majumdar, Stephen B Soumerai\n\nABSTRACT:\nObjective To assess the impact of direct to consumer advertising of prescription drugs in the United States on Canadian prescribing rates for three heavily marketed drugs—etanercept, mometasone, and tegaserod.Design Controlled quasi-experimental study using interrupted time series analysis.Population Representative sample of 2700 Canadian pharmacies and prescription data from 50 US Medicaid programmes.Main outcome measures Differences in number of filled prescriptions per 10 000 population per month between English speaking and French speaking (control) Canadian provinces before and after the start of direct to consumer advertising in the United States.Results Spending on direct to consumer advertising for study drugs ranged from $194m to $314m (£104m-£169m; €131m-€212m) over the study period. Prescription rates for etanercept and mometasone did not increase in English speaking provinces relative to French speaking controls after the start of direct to consumer advertising. In contrast, tegaserod prescriptions increased 42% (0.56 prescriptions/10 000 residents, 95% confidence interval 0.37 to 0.76) in English speaking provinces immediately after the start of US direct to consumer advertising. Uncontrolled analysis of US Medicaid data showed a larger 56% increase in tegaserod prescriptions. However, this increase did not persist over time in either country, despite continued advertising.Conclusions Exposure to US direct to consumer advertising transiently influenced both Canadian and US prescribing rates for tegaserod, a drug later withdrawn owing to safety concerns. The impact of direct to consumer advertising on drug use seems to be highly variable and probably depends on the characteristics of the advertised drug, the level of exposure to direct to consumer advertising, and the cultural context.\n\nBODY:\nIntroductionDirect to consumer advertising is a major component of drug promotion in the United States; manufacturers spent an estimated $4.24bn (£2.28bn; €2.88bn) in 2005—a 330% increase since 1996.1 The merits of direct to consumer advertising have been extensively debated, which has led to differing regulations across countries.2\n3 Regulatory disputes continue worldwide, with ongoing debate about the introduction of direct to consumer advertising in the European Union and Canada; at the same time, the US Senate has recently considered legislation prohibiting such advertising during the first two years after the release of a new drug.4\n5\n6 Although the debate includes a broad range of concerns, many assertions assume that direct to consumer advertising increases the use of particular types of drugs. For example, proponents argue that it increases use of effective treatments for undertreated conditions, such as depression.2 Opponents, however, suggest that it drives up demand for newer drugs with higher costs, marginal benefits, and unknown safety profiles.3Both sides of the argument assume that direct to consumer advertising increases use. However, the effectiveness of drug advertising campaigns is unclear and no extant studies use a concurrent control group and quantify the impact on use of marketed drugs.7\n8 Previous uncontrolled longitudinal studies have found that expenditure on direct to consumer advertising was associated with higher sales of antidepressants, proton pump inhibitors, antihistamines, and nasal sprays but non-significant or very small association with sales of statins and cyclo-oxygenase-2 selective non-steroidal anti-inflammatory drugs.9\n10\n11\n12 How these associations might be confounded by selection bias is unclear from these previous studies. For example, drugs with a larger pool of potential users or that are more innovative are more likely to be promoted through both direct to consumer advertising and physician directed campaigns using detailing, journal advertisements, and free samples.13 Moreover, previous studies have not controlled for pre-advertising trends in use or evaluated comparable markets that are unexposed to such advertising.In the absence of firm evidence describing the effect of direct to consumer advertising on use of prescription drugs, policy makers in the United States and New Zealand have permitted it whereas their counterparts in Europe, Canada, and Australia have prohibited it. The extent of benefits or harms attributable to direct to consumer advertising will be directly proportional to how effectively it increases use of particular advertised drugs and at what cost. We studied the impact of US direct to consumer advertising campaigns on Canadian prescribing rates for three heavily marketed drugs by using a controlled longitudinal study design. Because Canadians are regularly exposed to “illicit” English language direct to consumer advertising from the United States, we hypothesised that these campaigns would increase use of the marketed drugs in English speaking Canadian provinces. For any campaigns associated with increased use in our Canadian analysis, we examined US Medicaid data without a control group to investigate whether the effects were greater with increased exposure to direct to consumer advertising.MethodsStudy settingExamination of US data alone to delineate the impact of direct to consumer advertising is limited by two factors. Firstly, near universal exposure to advertising makes it almost impossible to find a comparable unexposed control group within the United States.14 Secondly, manufacturers start many direct to consumer advertising campaigns shortly after the launch of a drug—precisely when detailing to physicians and coverage in the medical literature are likely to be at their highest. We sought to limit these threats to validity by examining the impact of US direct to consumer advertising campaigns on Canadian patterns of drug use in provinces with and without substantial exposure to such advertising—that is, in predominantly English speaking provinces compared with predominantly French speaking Quebec. For drugs for which we found an impact on Canadian prescribing rates, we used data from nationwide US Medicaid programmes to assess whether a dose-response relation might exist between greater exposure to direct to consumer advertising in the United States and more marked increases in drug use.Although Canada prohibits direct to consumer advertising that includes both a brand name and indications, substantial cross border exposure to US advertising occurs through cable and satellite television, radio, print media, and internet advertising.15 Statistics Canada estimates that around 30% of television watched by English speaking Canadians is foreign sourced, most of which is probably US cable and satellite stations.16 Previous Canadian survey work suggested that more than 85% of English speaking patients had seen drug advertisements in the previous year and half had seen advertisements for six or more different products.15 Moreover, their primary care physicians filled nearly three quarters of patients’ requests for specific drugs.15 Thus, English speaking Canadians are regularly exposed to considerable amounts of US advertising and have the means to obtain advertised drugs.Data sourcesOur primary analysis used monthly drug use data from the nationally representative CompuScript audit from IMS Health Canada, an independent health information company, from January 2002 to December 2006. This audit uses a panel of approximately 2700 pharmacies (roughly 34% of all community pharmacies in Canada) to estimate total Canadian use of each drug. The major outcome of interest was the number of dispensed prescriptions of each drug per 10 000 residents per month. To calculate these rates, we used population estimates from Statistics Canada.17 We also obtained IMS Health Canada data estimating Canadian expenditure on detailing and distribution of free samples for the study drugs, to assess whether other marketing increased coincidently with US direct to consumer advertising. We found no evidence of such changes. Our analysis in the United States used quarterly data from 50 US Medicaid programmes.18 Using state level enrolment numbers, we calculated dispensed prescription rates per 10 000 Medicaid enrolees per quarter.19 These data provide estimates up to the end of 2005, when many patients were transferred to the new Medicare Drug Benefit.The start month and total spending on US direct to consumer advertising campaigns came from TNS Media Intelligence. The dataset tracks advertising and estimates expenditure across several media, including television, radio, and print media, and has been used in previous research on direct to consumer advertising.9 We also searched the Vanderbilt Television news archive to ascertain when particular drugs were advertised during major US national news broadcasts.20 Finally, we assessed whether manufacturers aired television advertising in Canada mentioning a brand name by reviewing the databases of Eloda, an independent company that provides monitoring and verification services for North American advertising.Study drugsDifferences exist between the United States and Canada in terms of availability and approval dates for drugs.21 Consequently, we sought out drugs that were included in US marketing campaigns started between January 2003 and December 2005; not advertised on Canadian television with a brand name; and approved for use in Canada before US advertising, to allow estimation of the marginal effect of direct to consumer advertising on prescribing.On the basis of these characteristics, we identified three study drugs. The first eligible drug was etanercept (Enbrel), a biological agent approved in Canada for the treatment of symptom refractory rheumatoid arthritis. Direct to consumer advertising for etanercept started in January 2003, and US network news advertising started in March 2003.20 The second eligible drug was mometasone (Nasonex), an inhaled nasal steroid spray for symptoms of allergy. Direct to consumer advertising for mometasone started in December 2004, and the Vanderbilt database showed extensive US news advertising starting the same month.20 Thirdly, tegaserod (Zelnorm) is a serotonin receptor agonist approved for the treatment of constipation predominant irritable bowel syndrome in women. When released, it was the only drug approved specifically for this indication in Canada. Although direct to consumer advertising began in February 2003, tegaserod’s most influential and major campaign first aired in August 2003 and featured memorable written messages such as “I feel better” on actresses’ stomachs.6 This later campaign was considered very successful from a marketing perspective, and even won major advertising industry awards, before the drug was withdrawn in both Canada and the United States owing to concern about cardiac side effects.6\n22 The Vanderbilt database indicates that US newscast advertising for tegaserod occurred in 1-12 September 2003 and subsequently in March 2004.20AnalysisWe used regional differences in exposure to investigate the impact of direct to consumer advertising. As all US advertising was in English, we hypothesised that changes in prescribing in Canada would be concentrated in predominantly English speaking provinces. Although French speaking Canadians watch a similar amount of television, they view much less foreign sourced television, estimated at less than 5% of all viewing.16 Consequently, we analysed the difference in prescribing rates between predominantly English speaking provinces (n=8) and Quebec, where French is the mother tongue for more than 80% of the population.23 Quebec is also attractive as a control as it has one of the least restrictive public drug formularies in Canada but has comparable universal health insurance coverage, age, sex, and income profiles to the other provinces.24\n25We used interrupted time series analysis, one of the strongest quasi-experimental designs available, to examine longitudinal changes in Canadian prescribing rates.26 Firstly, we calculated the difference in the prescribing rate per 10 000 population by subtracting the rate in French speaking provinces from that in English speaking provinces. We then fitted time series models to test whether a statistically significant change occurred in the level or trend of the difference after the start of US advertising or US national network news advertising, controlling for the pre-direct to consumer advertising level and trend. This method simultaneously controlled for any pre-advertising differences in the absolute level of prescribing between the provinces as well as any differences in pre-advertising temporal trends related to changes in the rates of prescribing between provinces. We also did a sensitivity analysis using the ratio of English and French prescribing rates instead of the difference. The results and interpretation of this analysis (not shown) were consistent with those shown below. For drugs that showed any significant impact of direct to consumer advertising in Canada, we did a sensitivity analysis using data from US Medicaid programmes. We also did this with an interrupted time series analysis but without an “unexposed” US control group. Although this method is uncontrolled compared with the Canadian analyses, it still controlled for pre-direct to consumer advertising trends in drug use. We used a generalised least squares model allowing for a first order autoregressive correlation between consecutive months or quarters and excluded the advertising start month in Canada. We validated our use of this autocorrelation structure by using likelihood ratio tests. Moreover, alternative models with no or longer autocorrelation structures led to results with very similar estimates and identical interpretations.ResultsTable 1 describes the US advertising campaigns and Canadian approval dates for the three study drugs. All three drugs had large direct to consumer advertising expenditures, ranging from US$194 million to $314 million during the study period. Pre-advertising trends in use for each of the study drugs were generally comparable between English speaking and French speaking provinces (figs 1, 2 and 3). We found that US direct to consumer advertising led to increased Canadian prescribing rates for only one of the three drugs, tegaserod.US approval and advertising dates and Canadian approval dates for study drugsDrugUnited StatesCanadaGeneric nameBrand nameApprovalAdvertising startDTCA spending to 2006 ($m)ApprovalEtanerceptEnbrelNovember 1998January 2003$194December 2000MometasoneNasonexOctober 1997December 2004$235July 1998TegaserodZelnormJuly 2002February 2003$314March 2002DTCA=direct to consumer advertising.Start dates and US advertising values are from TNS Media Intelligence. Data include spending on network and cable television, magazine, newspaper, radio, and billboard advertising.Fig 1 Number of etanercept prescriptions per 10 000 population per month in Canadian provinces that are predominantly English speaking (n=8) or French speaking (n=1). Vertical line indicates start of US advertising in January 2003. Difference between rates shown at bottom of chart; fitted trend line shows predicted differences from interrupted time series regression. DTCA=direct to consumer advertisingFig 2 Number of mometasone prescriptions per 10 000 population per month in Canadian provinces that are predominantly English speaking (n=8) or French speaking (n=1). Vertical line indicates start of US advertising in December 2004. Difference between rates shown at bottom of chart; fitted trend line shows predicted differences from interrupted time series regressionFig 3 Number of tegaserod prescriptions per 10 000 population per month in Canadian provinces that are predominantly English speaking (n=8) or French speaking (n=1). Vertical lines indicate start of US advertising in February 2003 and start of new TV advertising campaign in August 2003. Difference between rates shown at bottom of chart; fitted trend line shows predicted differences from interrupted time series regression. DTCA=direct to consumer advertisingEtanerceptFigure 1 shows the times series of monthly prescribing rates of etanercept in Canada, which were very similar in both language regions. We found that advertising had no statistically significant impact on the level or trend of differences in prescribing rate between English speaking and French speaking provinces (level change −0.18 prescriptions per 10 000 population, 95% confidence interval −0.39 to 0.04, P=0.10; trend change −0.03 prescriptions per 10 000 population per month, −0.06 to 0.003, P=0.07).MometasoneFigure 2 shows the monthly prescribing rates for mometasone in English speaking and French speaking Canadian provinces. As with etanercept, we saw no clinically important or statistically significant change in the level or trend of differences in prescribing rate between English speaking and French speaking provinces (level change −3.61 prescriptions per 10 000 population, −10.51 to 3.29, P=0.30; trend change −0.08 prescriptions per 10 000 population per month, −0.57 to 0.40, P=0.73).TegaserodIn contrast to the first two drugs described, US direct to consumer advertising for tegaserod seemed to have a strong influence on Canadian prescribing. Figure 3 shows the monthly prescribing rates for tegaserod. The February 2003 campaign, which contained no US network news advertising, had no significant impact on prescribing rates and was incorporated into the pre-advertising period. In contrast, a level increase of 0.56 prescriptions per 10 000 population (0.37 to 0.76, P<0.001) in the difference in prescribing rate between English speaking and French speaking provinces occurred immediately after the August 2003 campaign. We found no statistically significant change in trend (−0.003 prescriptions per 10 000 population per month, −0.03 to 0.02, P=0.77). Overall, this represents an estimated 42% increase in the first month after direct to consumer advertising. However, this difference did not persist despite continued advertising throughout the study period. Within two years of direct to consumer advertising, prescribing rates were again virtually identical between English speaking and French speaking regions.Using the same start date for direct to consumer advertising, we found a similar increase in Medicaid prescription rates of tegaserod. Figure 4 shows that the pre-advertising upward trend in tegaserod use was substantially higher in US Medicaid than in Canada. After national network news direct to consumer advertising, we saw an increase in the level of prescribing in the United States; the number of prescriptions per 10 000 enrolees increased by 5.70 (3.65 to 7.75, P<0.001). As in Canada, we found no statistically significant change in prescribing trends (−0.62 prescriptions per 10 000 enrolees per quarter, −1.52 to 0.27, P=0.15). Overall, the estimated increase in prescribing in the first quarter of direct to consumer advertising was 56% higher than would have been expected and greater than the 42% increase seen in Canada.Fig 4 Number of tegaserod prescriptions per 10 000 enrolees per quarter in US Medicaid programmes. Vertical line indicates start of new TV advertising campaign in third quarter of 2003. Fitted trend line represents fitted interrupted time series analysis for rate of use in MedicaidDiscussionDuring the past decade, drug manufacturers have substantially increased spending on direct to consumer advertising.1 To our knowledge, this study is the first analysis that uses a concurrent control group to evaluate the impact of such advertising on use of specific drugs. We found that for two of three drugs the US direct to consumer advertising had no apparent impact on Canadian prescribing rates, and for one drug (tegaserod) we saw a short lived effect. These mixed findings are surprising, as we included several expensive advertising campaigns that were highly recalled by consumers.27\n28 Our empirical results raise important questions about whether and how prescribing trends for specific drugs respond to advertising directed at consumers. Thus, they have important implications for the ongoing debate about the benefits and harms of direct to consumer advertising.Possible explanationsWe believe that the differential responses to direct to consumer advertising that we saw may be related to the characteristics of the drugs examined. Although all of the study drugs are primarily used for relieving symptoms, they differ in important ways. For example, etanercept requires referral to a specialist and intravenous administration, making the pathway between direct to consumer advertising and drug use complicated. Thus, the effect of advertising probably differs substantially from that of drugs prescribed predominantly in primary care settings. Furthermore, tegaserod, unlike the other study drugs, was the only drug approved for its indication in Canada.29\n30 In contrast, the other drugs studied all had competitors within the same drug class. In such markets, direct to consumer advertising might protect against drops in levels of use, rather than expanding use. Other characteristics, such as effectiveness, may also be important. A meta-analysis of short term placebo controlled trials of tegaserod indicates that the number needed to treat for one patient to have some improvement in their gastrointestinal symptoms is about 17, suggesting that most patients trying tegaserod for the first time were unlikely to derive symptomatic benefit.30 This may explain, in part, why the changes in use for this drug were short lived.Our results also suggest that when direct to consumer advertising does increase use, a dose-response relation with the level of exposure to advertising exists. Our results in US Medicaid programmes estimated a larger increase than in Canada, in both absolute and percentage terms. Although the immediate change in use in the United States was larger than in Canada, assessing the comparative long term effect of advertising in the United States is difficult, because no concurrent control group is available. Nevertheless, the observed US Medicaid prescribing rates returned to the pre-direct to consumer advertising trend around the same time as in Canada (mid-2005). Furthermore, use of tegaserod was both higher and growing faster in Medicaid before direct to consumer advertising, suggesting that other factors were driving these differential trends. For example, we cannot rule out between country differences in physician directed marketing activities.31Strengths and limitationsThe major strength of our study is the use of a strong quasi-experimental design with a comparable and concurrent control group. Moreover, our study design controlled for difference in both pre-existing level and trend and explicitly considered the timing of advertising campaigns. This method controls for differences in characteristics between language regions of Canada that remained constant or changed predictably over time, such as culture or patterns of general medical practice. Indeed, other differences such as variation in provincial drug reimbursement plans would bias our results only if they coincidentally changed when the individual direct to consumer advertising campaigns started. We could find no evidence that this occurred for any of the drugs studied. None the less, exclusion from provincial formularies might constrain the effects of successful advertising campaigns. However, most private insurance plans in Canada do not have formularies and cover most of the population.32 Moreover, although Ontario and Alberta both excluded tegaserod from their public drug programmes, the effect of direct to consumer advertising was apparent in both provinces (data not shown).The study has other limitations. Firstly, generalising beyond the three drugs that met our inclusion criteria is difficult. Secondly, we do not have information on whether these drugs were subject to disease awareness advertising by companies that did not mention the brand name. However, this would bias our results only if it was similarly timed, and we found no indication for mometasone or etanercept of increased use coincident with branded direct to consumer advertising, thus making it unlikely. Thirdly, variation in drug coverage, the overall health system, culture, levels of exposure to advertising, or television viewing patterns might result in the effect of direct to consumer advertising differing between drugs and between countries. However, the percentage increase in and duration of effect for tegaserod was similar in both countries. In terms of drug coverage, more than 60% of Canadians are covered by generally unrestrictive employer based private drug plans, and less than 20% of these plans use formularies to limit access to specific drugs.32 Although the Canadian and US health systems vary substantially, studies indicate similar access to primary care and willingness of physicians to fulfil patients’ requests for specific prescription drugs in the two countries.15\n33 Overall, the striking initial effect of direct to consumer advertising on tegaserod prescribing rates provides evidence that exposure to US advertising is sufficient to influence Canadian prescribing.ImplicationsThe implications of our analysis are threefold. Firstly, it indicates that illicit cross border exposure to direct to consumer advertising has the potential to modify drug use, even where such advertising is technically prohibited. As advertising over global mediums such as the internet increases, this phenomenon may grow in importance. Secondly, to our knowledge, these results are the strongest evidence that direct to consumer advertising can increase use of a drug that was removed from the market as a result of concerns about safety. Finally, our findings suggest that the impact of direct to consumer advertising campaigns is mixed, as they seem to work for some drugs and not others. If the overall impact of direct to consumer advertising is limited or variable, then a substantial portion of expenditure on such advertising—borne by governments, insurers, and patients in the form of higher costs or by companies as reduced profits—may be better spent elsewhere. Previous commentary may have overemphasised the impact of direct to consumer advertising for many individual drugs for which evidence that it increases use is either weak or non-existent.2 Until we better understand how direct to consumer advertising modifies prescribing for particular drugs, debates about its positive and negative consequences will continue to be based on conjecture rather than strong evidence.What is already known on this topicAlthough direct to consumer advertising (DTCA) of prescription drugs remains controversial, no controlled studies have investigated its impact on prescribingIn the absence of such evidence, both opponents and proponents of DTCA have generally assumed it to be highly effective at increasing the use of advertised drugsWhat this study addsDTCA campaigns seem to have mixed effectiveness; drug use did not increase for two of three drugs studiedDespite prohibitions, DTCA can influence prescribing across national bordersThe drug (tegaserod) for which use increased with DTCA was eventually withdrawn owing to safety concerns\n\nREFERENCES:\n1. Donohue JM, Cevasco M, Rosenthal MB. A decade of direct-to-consumer advertising of prescription drugs. N Engl J Med2007;357:673-81.17699817\n2. Holmer AF. Direct-to-consumer prescription drug advertising builds bridges between patients and physicians. JAMA1999;281:380-2.9929095\n3. Hollon MF. Direct-to-consumer advertising: a haphazard approach to health promotion. JAMA2005;293:2030-33.15855439\n4. Meek C. Europe reconsidering DTCA. CMAJ2007;176:1405.\n5. Cassels A. Canada may be forced to allow direct to consumer advertising. BMJ2006, 10.1136/bmj.332.7556.1469-a\n6. Shuchman M. Drug risks and free speech: can Congress ban consumer drug ads? N Engl J Med2007;356:2236-9.17476002\n7. Mansfield PR. Do advertisements in clinical software influence prescribing? Med J Aust2008;188:13-4.18205555\n8. Gilbody S, Wilson P, Watt I. Benefits and harms of direct to consumer advertising: a systematic review. Qual Saf Health Care2005;14:246-50.16076787\n9. Rosenthal MB, Berndt ER, Donohue JM, Epstein AM, Frank RG. Demand effects of recent changes in prescription drug promotion. In: Cutler DM, Garber AM, eds. Frontiers in health policy research. Cambridge, MA: MIT Press, 2003:1-26.\n10. Donohue JM, Berndt ER, Rosenthal M, Epstein AM, Frank RG. Effects of pharmaceutical promotion on adherence to the treatment guidelines for depression. Med Care2004;42:1176-85.15550797\n11. Calfee JE, Winston C, Stempski R. Direct-to-consumer advertising and the demand for cholesterol-reducing drugs. J Law Econ2002;45:673-90.\n12. Bradford WD, Kleit AN, Nietert PJ, Steyer T, McIlwain T, Ornstein S. How direct-to-consumer television advertising for osteoarthritis drugs affects physicians’ prescribing behavior. Health Aff (Millwood)2006;25:1371-7.16966735\n13. Iizuka T. What explains the use of direct-to-consumer advertising of prescription drugs? Journal of Industrial Economics2004;LII:349-79.\n14. Weissman JS, Blumenthal D, Silk AJ, Zapert K, Newman M, Leitman R. Consumers’ reports on the health effects of direct-to-consumer drug advertising. Health Aff (Millwood)2003;(suppl):W3-95.\n15. Mintzes B, Barer ML, Kravitz RL, Bassett K, Lexchin J, Kazanjian A, et al. How does direct-to-consumer advertising (DTCA) affect prescribing? A survey in primary care environments with and without legal DTCA. CMAJ2003;169:405.12952801\n16. Statistics Canada. Television viewing: data tables. Ottawa: Statistics Canada, 2006.\n17. Statistics Canada. Quarterly demographic estimates. Ottawa: Statistics Canada, 2007. (Catalogue no. 91-002-XIE.)\n18. Centers for Medicare and Medicaid Services. State drug utilization data. Baltimore, MD: Centers for Medicare and Medicaid Services, 2007.\n19. Centers for Medicare and Medicaid Services. Medicaid managed care enrollment report. Baltimore, MD: Centers for Medicare and Medicaid Services, 2005.\n20. Vanderbilt Television News Archive. 2007. http://tvnews.vanderbilt.edu.\n21. Lexchin J. A comparison of new drug availability in Canada and the United States and potential therapeutic implications of differences. Health Policy2006;79:214-20.16472888\n22. New York American Marketing Association. Effie awards brief of effectiveness: Zelnorm “Tummies”. 2005. http://s3.amazonaws.com/effie_assets/2005/573/2005_573_pdf_1.pdf.\n23. Statistics Canada. Language composition of Canada: highlight tables, 2001 census. http://www12.statcan.ca/english/census01/products/highlight/LanguageComposition/Page.cfm?Lang=E&Geo=PR&View=1a&Table=1a&StartRec=1&Sort=2&B1=Counts&B2=Both.\n24. Statistics Canada. Median total income, by family type, by province and territory. 2007. http://www40.statcan.ca/l01/cst01/famil108a.htm.\n25. Statistics Canada. Population by sex and age group, by province and territory. 2007. http://www40.statcan.ca/l01/cst01/demo31a.htm.\n26. Wagner AK, Soumerai SB, Zhang F, Ross-Degnan D. Segmented regression analysis of interrupted time series studies in medication use research. J Clin Pharm Ther2002;27:299-309.12174032\n27. IAG Research. Advertising for prescription drugs is memorable and motivating: latest rankings for Rx Ads on television released by IAG Research. 2005. www.iagr.net/ne_press/pr_112905.jsp.\n28. IAG Research. Prescription drug advertising break-through: IAG Research reveals the best prescription drug advertising from this past TV season. 2006. www.iagr.net/ne_press/pr_112906.jsp.\n29. Rogers EM. Diffusion of innovations. 5th ed. New York: Free Press, 2003.\n30. Evans BW, Clark WK, Moore DJ, Whorwell PJ. Tegaserod for the treatment of irritable bowel syndrome. Cochrane Database Syst Rev2007;(4):CD003960.\n31. Majumdar SR, McAlister FA, Soumerai SB. Synergy between publication and promotion: comparing adoption of new evidence in Canada and the United States. Am J Med2003;115:467-72.14567371\n32. Applied Management. Canadians’ access to insurance for prescription medicines. 2000. www.hc-sc.gc.ca/hcs-sss/pharma/acces/pubs_e.html.\n33. Schoen C, Osborn R, Huynh PT, Doty M, Davis K, Zapert K et al. Primary care and health system performance: adults’ experiences in five countries. Health Aff (Millwood)2004;(suppl):W4-503."
+}

batch_8/PMC2528964.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2528964",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2528964\nAUTHORS: Patrick Flaherty, Mala L. Radhakrishnan, Tuan Dinh, Robert A. Rebres, Tamara I. Roach, Michael I. Jordan, Adam P. Arkin\n\nABSTRACT:\nMacrophage cells that are stimulated by two different ligands that bind to G-protein-coupled receptors (GPCRs) usually respond as if the stimulus effects are additive, but for a minority of ligand combinations the response is synergistic. The G-protein-coupled receptor system integrates signaling cues from the environment to actuate cell morphology, gene expression, ion homeostasis, and other physiological states. We analyze the effects of the two signaling molecules complement factors 5a (C5a) and uridine diphosphate (UDP) on the intracellular second messenger calcium to elucidate the principles that govern the processing of multiple signals by GPCRs. We have developed a formal hypothesis, in the form of a kinetic model, for the mechanism of action of this GPCR signal transduction system using data obtained from RAW264.7 macrophage cells. Bayesian statistical methods are employed to represent uncertainty in both data and model parameters and formally tie the model to experimental data. When the model is also used as a tool in the design of experiments, it predicts a synergistic region in the calcium peak height dose response that results when cells are simultaneously stimulated by C5a and UDP. An analysis of the model reveals a potential mechanism for crosstalk between the Gαi-coupled C5a receptor and the Gαq-coupled UDP receptor signaling systems that results in synergistic calcium release.\n\nBODY:\nIntroductionThe G-protein-coupled signal transduction system integrates a wide range of intercellular signals and actuates downstream pathways. G-protein-coupled receptors (GPCRs) are composed of seven α-helices that span the plasma membrane, an extracellular domain that is activated by an agonist and an intracellular domain that binds a guanine nucleotide heterotrimer made up of different α, β, and γ subunit isoforms. This receptor system accounts for 40–50% of modern medicinal drug targets but only 10% of the known receptors are targeted by drugs [1]. Though the system is physiologically and pharmacologically important, the mechanism by which the system integrates multiple signals is not well understood [2].We address the G-protein-mediated route to calcium release in RAW264.7 cells. When activated by a specific ligand, the G protein heterotrimer dissociates to free Gα-GTP and Gβγ. Specific Gα and Gβγ isoforms are able to bind specific isoforms of phospholipase C β (PLCβ) and catalyze the synthesis of inositol (1,4,5)-triphosphate (IP3) and diacylglycerol (DAG) from phosphatidylinositol (4,5)-bisphosphate (PIP2) [3],[4]. In addition to its catalytic activity, PLCβ acts as a GTPase for Gα-GTP [5]. IP3 binds to specific receptor-channels on the membrane of the ER to release Ca2+ into the cytosol [6]. DAG and Ca2+ bind to and activate protein kinase C (PKC) which may phosphorylate and inactivate specific PLCβ isoforms [7]. G protein receptor kinase (GRK) is activated once it is phosphorylated by PKC [8] and is localized to the plasma membrane by Gβγ [9]. Though phosphorylation has not been shown to be necessary for GRK activation, we have assumed so in our model because phosphorylation by PKC may release the inhibition of GRK2 by being bound to calmodulin [8]. Activated GRK can then phosphorylate specific GPCRs which leads to receptor inactivation—perhaps directly or by arrestin activity [8]. In this complex signal transduction network, Gα and Gβγ subunits have different patterns of specificity for PLCβ isoforms and calcium is an important cofactor in several important feedback loops [10].The two extracellular signaling ligands we consider here are C5a and UDP. The small peptide C5a is a potent anaphylotoxin and a strong chemoattractant for many immune system components [11]. The calcium response due to stimulation by C5a is predominantly coupled through Gαi-linked heterotrimers. Macrophage cells and their precursors, monocytes, express several receptors that are specific to extracellular nucleotides and it has been shown that the P2Y6 receptor, which is sensitive to UDP, regulates the production and secretion of the chemokine interleukin 8 (IL-8) in monocytes [12]. The UDP response is mediated by Gαq-linked heterotrimers, but other receptors in the P2Y family may respond to UDP and couple the signal through other G protein isoforms [13].Four recent models have sought to explore various aspects of the G protein coupled signal transduction system in detail. Lukas et al. compare measured calcium response over a range of bradykinin doses to their model predictions [14]. Mishra and Bhalla built a model to investigate the role of IP4 as a signal coincidence detector in the GPCR pathway [15]. The model by Lemon et al. predicts the calcium response to UTP stimulation and is the closest in focus to our model [16]. A recent model of calcium dynamics in RAW cells has been proposed that is quite similar to this model, but does not deal with crosstalk between receptors or formal statistical uncertainty in model predictions [17],[18].Several hypotheses for the mechanism of crosstalk and synergy among GPCR-mediated pathways have been proposed. Crosstalk among GPCR-mediated pathways is important both physiologically and pharmaceutically. Quitterer et al. propose that crosstalk is mediated by Gβγ exchange between Gαi-coupled and Gαq-coupled receptors [19]. Zhu et al. speculated that PLC is under either conditional or dual regulation of Gβγ and Gα [20]. Though these hypothetical mechanisms for crosstalk among G protein coupled receptor systems are conceptually plausible we have not found these or any other of the many competing hypothetical mechanisms tested in the context of a quantitative mathematical model [2].In this paper Bayesian statistical inference is used to provide a rigorous connection between the mathematical model derived from mass-action kinetics, prior information from in-vitro biochemical studies and heterogeneous experimental data. The prior distribution over the parameters represents our uncertainty before observing a set of experimental data. A broad, high variance, prior distribution means we are quite uncertain and a concentrated, low variance, prior means we are more certain about the parameter a priori. The objective of our inference is the posterior distribution over the parameters because it is an informed estimate of both the value of the parameter and the uncertainty in the parameter value. The posterior distribution over the parameters is then used as a tool for experiment design to estimate the model-based posterior distribution over observable quantities such as the cytosolic calcium concentration and to drive the design of new experiments. This statistical approach is possible in a model of this size because of the abundance and quality of the data collected for this study.ResultsThere are two main features of the structure of our model, shown in Figure 1, which contribute to crosstalk in the system and produce the key dynamical features in the calcium response: isoform specificity and calcium-dependent feedback. As we will show, by including multiple isoforms of PLCβ and Gα as well as the negative feedback mediated by PKC, GRK and the IP3 receptor itself, we are able to predict the synergistic interaction between C5a and UDP observed in the experimental data.10.1371/journal.pcbi.1000185.g001Figure 1The model for crosstalk between the Gαi and Gαq pathways depends on both differential specificity and activity for Gαi, Gαq, and Gβγ interactions with PLCβ3 and PLCβ4 to catalyze PIP2 hydrolysis and calcium-dependent feedback control mediated by GRK and PKC.Selected model parameters are informed by calcium measurements taken for various ligand doses on wild-type and cell lines with shRNAi knockdowns on the proteins shown in red.Our representation of the G-protein-coupled signal transduction system includes C5a and P2Y6 receptors, Gαi2, Gαq, Gβγ, PLCβ3, PLCβ4, PIP2, DAG, IP3, PKC, GRK2, calcium buffer, a Na2+/Ca2+ exchanger, a sarco(endo)plasmic reticulum Ca2+-ATPase (SERCA) pump, IP3 receptors and RGS. The model is composed of 53 coupled ordinary differential equations with 84 parameters and 24 non-zero initial conditions. The complete model equations are shown in Figure S7 and a more detailed model diagram is shown in Figure S6. The parameters and initial conditions are in Table S2 and Table S1, respectively. Where available, we have relied on in-vitro or in-vivo biochemical experiments for the reactions and parameter values (see Supporting Information). In cases where the biochemical parameter values were not known, we chose physically reasonable values. Twenty of the 84 parameters most relevant to the knock-down and wild-type data were estimated from cytosolic calcium measurements as described in the Methods section. Most reactions were assumed to be governed by mass-action kinetics, but for a few proteins—such as RGS—the mechanism of regulation is not known in enough detail and we have approximated with Michaelis-Menten kinetics or a phenomenological function.We briefly discuss the reactions involving the Na2+/Ca2+ exchanger, SERCA pump, IP3 receptors, RGS and calcium buffer because they are important for the faithful representation of the system in our model. Regulators of G protein signaling (RGS) are GTPase proteins that down-regulate the extent of signaling [21]; RGS2 at least is expressed in RAW264.7 macrophage cells and therefore an RGS activity is included in our model. The mechanism of activation of RGS2 as it relates to Gαi and Gαq signaling is not entirely known and is difficult to assess because antibodies that specifically recognize RGS2 are not widely available [22]–[24]. We have assumed constitutive activity and expect as more information becomes available a more accurate model of the regulation of RGS2 and other RGS isoforms will be possible. The SERCA pump helps to bring the cytosolic Ca2+ concentration back to the resting level after stimulation. We have modeled the SERCA pump as in the Keizer and DeYoung model [25]. The IP3 dependent opening of ER calcium channels was found to be cooperative [26] and we have used the Meyer and Stryer model for the IP3-gated channel with a Hill coefficient of four [25],[27]. Finally, many other proteins such as calmodulin and the fluorescent indicator Fura-2 bind Ca2+. Because our measurements reflect these effects, we have included a general buffer for cytosolic calcium.Isoform SpecificityComplement factor 5a activates the C5a receptor which is a Gαi-coupled receptor [28]. The released Gβγ dimer activates PLCβ2 and PLCβ3 which are lumped and called PLCβ3 in our model because: (i) the activity of Gβγ-activated PLCβ3 has been shown to be greater than Gβγ-activated PLCβ2 in in-vitro studies and (ii) Gαq activates both PLCβ2 and PLCβ3 so the structural connections from Gβγ and Gαq to PLCβ2 and PLCβ3 in the model are identical [4],[29]. PLCβ1 is activated by Gβγ and Gαq, but RAW264.7 macrophage cells do not express this isoform, so we have not included it in the model. PLCβ3 then catalyzes the hydrolysis of phosphatidylinositol (4,5)-bisphosphate (PIP2) into inositol 1,4,5-trisphosphate (IP3) and diacylglycerol (DAG).UDP stimulates the P2Y6 receptor and the associated Gαq-GTP activates both PLCβ3 [30] and PLCβ4 [31]. The GTPase rate of Gαq is increased 1000-fold when bound to PLCβ [5]. Due to this rapid hydrolysis rate, we have assumed, in our model, that PLCβ3 or PLCβ4 bound Gαq-GTP may only hydrolyze one molecule of PIP2 before releasing Gαq-GDP. Additionally, the Gβγ released by the P2Y6 receptor also activates PLCβ3 [30], but does not activate PLCβ4 [32].Our model assumes that PLCβ3 does not simultaneously bind Gβγ and Gαq. Indeed, a biochemical study of PLCβ2 activity in reconstituted membrane fractions strongly argues that Gαq and Gβγ do not simultaneously bind this effector [33]. While this was specifically demonstrated for PLCβ2, we implicitly assume the same holds for PLCβ3 because we lump the two in our model. This is a mechanistic assumption of our model and an interesting issue for future testing with directed experiments.Calcium-Dependent FeedbackThough important for response specificity, the dynamical control of calcium release is not limited to the forward pathway in this system. Calcium participates in feedback processes that both enhance and inhibit its own release at multiple points in the pathway. There are four main nodes of calcium-dependent feedback control in our model: PLCβ, IP3 receptor, protein kinase C (PKC) and G protein receptor kinase (GRK).Calcium enhances its own release by binding to the EF-hand domain on PLCβ and is required for PLCβ to hydrolyze PIP2 into IP3 and DAG [34]. Because the dissociation constant for PLCβ-Ca2+ in our model is larger than the basal concentration of cytosolic calcium, as more Ca2+ is released from the ER, more PLCβ-Ca2+ becomes available to bind Gαq or Gβγ. This positive feedback mechanism accelerates the release of Ca2+.In our model, Ca2+ and IP3 cooperatively open the channel between the ER and the cytosol. It is believed that Ca2+ initially stimulates the IP3 receptor with maximal stimulatory effect at 100–300 nM [6]. At higher concentrations, Ca2+ has an inhibitory effect. We use the IP3 receptor model structure in the Keizer and DeYoung model for this component [25].Protein kinase C (PKC) has been shown to phosphorylate PLCβ3 which inhibits PLCβ3 activation due to Gαq and Gβγ [35],[36]. PKC is activated when bound to DAG and Ca2+\n[7],[37]. Because the preferred order of binding is not entirely known, PKC, DAG and Ca2+ form a thermodynamic cycle of reversible reaction with only the PKC-DAG-Ca2+ form active. In our model, the dissociation constant of PKC and Ca2+ is much greater than the basal Ca2+ concentration, and upon binding DAG, the PKC-DAG complex has a higher affinity for Ca2+ making the order of binding preferentially PKC to DAG then PKC-DAG to Ca2+. It is not known whether PLCβ4 is also regulated by PKC. We have assumed, in our model, the same mechanism of PKC regulation of PLCβ3 and PLCβ4.The final key calcium-dependent feedback loop in our model is mediated by G protein receptor kinase (GRK). GRK2 phosphorylates and inactivates ligand-bound C5a receptors when activated by PKC and Gβγ. In sequence, PKC phosphorylates GRK2 which causes translocation to the plasma membrane [8]. When properly localized, GRK2 may bind Gβγ and then phosphorylate the C5a-C5a receptor complex to inactivate it [38]. This simplified representation of the receptor desensitization mechanism does not include arrestin activity, multiple receptor phosphorylation sites and other fine grain or slower biochemical interactions that may be present in-vivo.\nSingle Ligand Experiments\nHaving specified the structure of our model, we direct our attention to the parameters. We estimate 20 of the 84 parameters in our model using a dataset composed of 96 Fura-2 time series measurements as described in the Materials and Methods section. Each experiment consists of 3–4 samples from different wells in a 96 well plate. There are 15 experiments spanning 9 doses of C5a and 14 experiments spanning 11 doses of UDP on wild-type cells in the dataset (see Figure S3). The dataset also contains calcium measurements on 5 different shRNAi knockdown cell lines constructed by lentiviral infection (see Figure S4). The time interval between samples is approximately 3–4 seconds and each time series is approximately 100–300 seconds of post-stimulation data. Table 1 shows a summary of the knockdown data used for statistical parameter estimation for this model in addition to the wild-type experiments.10.1371/journal.pcbi.1000185.t001Table 1Dataset used for parameter estimation.Cell LineMeasured Fraction KnockdownModel ValueSample SizeC5aUDPqRT-PCRWesternNominalLowerUpper<10 nM10–100 nM>100 nM<1 µM1–10 µM>10 µMWild-type–––––483554GRK2 (2)90%±7%, n = 540%±6%, n = 640.0%22.0%58.0%2122315Gai2 (3)83%±5%, n = 473%±6%, n = 573.0%55.0%91.0%–5–5–7Gaq (3)70%±8%, n = 766%±23%, n = 266.0%0.0%95.0%–3–1–3PLCβ3 (1)–83%±15%, n = 383.0%38.0%100.0%–3–––3PLCβ4 (1)87%±6%, n = 5–87.0%69.0%100.0%–4–4–4Five different cell lines that have a perturbation in the level of a key signal transduction protein were constructed by shRNAi lentiviral infection. The calcium response from these cell lines in addition to the wild-type cell line were used to fit relevant parameters in the model. Because shRNAi does not entirely remove the protein product, the fraction knockdown was estimated by qRT-PCR and by Western blot analysis. The standard error (se) was computed for each estimate and the upper and lower confidence intervals were computed as ±3·se. The knockdown confidence intervals are used in the GPCR model to construct prediction confidence intervals for the calcium response. Where several cell lines were constructed for each knockdown, the best was selected and reported in parenthesis. The sample size for each knockdown-ligand dose combination is shown in the last 6 columns.We find that our model is generally quantitatively consistent with the experimental data within measurement uncertainty. Where the model is less consistent with the data – specifically for the GRK knockdown experiment – we find the deviation has a reasonable biological explanation. The summary of the dataset and the fit of the model to each single ligand experiment are available in the Supporting Information. We briefly discuss some issues relating to goodness of fit and the Bayesian parameter estimation here.While most optimization procedures produce a point estimate of the parameters that maximize the goodness of fit of the model to the observed data, the Bayesian procedure we have employed here estimates the entire posterior distribution of the parameters given the data. This information is valuable for qualitatively and quantitatively evaluating the precision of the parameters estimates. Figure 2 shows, as a qualitative evaluation, that while the a-priori forward and reverse binding rates for the receptors (C5aR and P2YR) are uncorrelated they are correlated in the posterior distribution. The calcium measurements have informed and constrained the posterior estimates of the dissociation constants to be approximately 5 nM and 250 nM for the C5aR and P2YR respectively. We have quantitatively computed marginal highest posterior density (HPD) confidence intervals for each of the twenty parameters we have estimated from the data. Those estimates are shown in Table S3. Those parameters with large HPD intervals are not well informed by the measurements and are candidates for directed biochemical experiments.10.1371/journal.pcbi.1000185.g002Figure 2This figure shows that the single and pairwise marginal posterior distributions for the ligand binding reactions for the P2YR and C5aR receptors.The vertical line in the single marginal posterior distributions shows the point estimate that were selected. The posterior distributions show the dissociation constants for the reactions are tightly constrained by the data, while the values of the forward and reverse rates that make up the ratio are not as well constrained by the data. Additionally, as expected the UDP binding rates are not correlated with the C5a binding rates. Marginal posterior distributions for all parameters and a discussion of the point estimate selection can be found in Figure S2.Wild-Type ExperimentsThe calcium response to C5a adapts and returns to the basal level, but the UDP response has a sustained elevated calcium level that slowly decays. Figure 3 shows two representative experiments of the response of the wild-type cell to stimulation with C5a and UDP. We expect that the fit to this data will be good because 20 key model parameters were fit using an experimental dataset that included these experiments – the fit is indeed accurate. The point estimate curve is constructed from the maximum a-posteriori parameters from an MCMC chain. The prediction intervals are estimated by Monte Carlo sampling from the posterior parameter distribution and the measurement error distribution conditional on the parameters. The prediction confidence intervals generally cover the observed data.10.1371/journal.pcbi.1000185.g003Figure 3Model simulations are compared to experimental data.The point estimate is computed using the posterior distribution of the parameter as estimated by Markov chain Monte Carlo given the data from 96 experiments on C5a and UDP at various doses in combination with 5 different shRNAi knockdown cell lines. The 95% posterior predictive intervals are estimated by Monte Carlo simulations including both parameter and measurement uncertainty. The measured mean and approximate 95% confidence intervals of four replicates is shown by a black dot and error bar. (A) C5a at 250 nM was introduced at 20s and the experimentally observed pulse in cytosolic calcium concentration is shown. (B) The qualitative shape of the calcium pulse for 25 µM UDP is different than for 250 nM C5a. The pulse does not completely adapt and return to the prestimulated level. For both ligands, the model prediction confidence intervals overlap the data error bars that indicate the model fit is consistent with the data within the measurement uncertainty.Knockdown ExperimentsLentiviral infection is used to introduce small hairpin RNAs to interfere with the translation of the key signaling proteins GRK2, Gαi2, Gαq, PLCβ3 and PLCβ4 [39]. There are three main sources of uncertainty in the knockdown experiment model predictions: parametric uncertainty, measurement uncertainty and knockdown efficiency uncertainty. We have dealt with the first two sources in the previous section on wild-type experiments. Here we address prediction variability due to knockdown efficiency uncertainty by using nominal parameter values.\nFigure 4 shows simulations and experimental data for three representative knockdown experiments. The upper-left panel of Figure 4 shows a GRK knockdown line stimulated with 250 nM C5a. Because GRK2 desensitizes the C5a receptor, we expect that by eliminating the feedback mechanism, the calcium peak will be higher and more sustained. The experimental data as well as the model indeed show that effect. Quantitatively, the model prediction shows a greater effect than the experimental data. A likely reason is that the model only considers one isoform of GRK while there are four isoforms expressed in the RAW264.7 cell line (GRK1,2,4,6). If more than one isoform can desensitize the C5a receptor, the effective knockdown in desensitization function will be less than as measured by western blot analysis on GRK2.10.1371/journal.pcbi.1000185.g004Figure 4The model simulation results for GRK and PLCβ3 knockdown cell lines stimulated with C5a and UDP are shown.The experimental mean±1 s.d. of 3–4 replicates within one experimental run is shown in black. The knockdown simulation result with nominal knockdown fraction and parameters is shown in red and the wild-type simulation result is shown in green for comparison. Upper and lower model 99% confidence intervals (shown as blue dashed lines) are simulated using the upper and lower knockdown fraction values from Table 1. As expected the Ca2+ response to C5a in the GRK knockdown line (A) was increased compared to wild-type. The quantitative deviation between the model and data is possibly due to the availability of multiple redundant GRK isoforms. (B) The expected effect of the GRK knockdown on the UDP response is an increase in the cytosolic calcium levels. Because GRK2 does not directly desensitize the P2Y receptor in this model, the effect is likely due to a reduction of sequestration of Gβγ by GRK. (C) The signal transduction of the C5a response is predominantly through the PLCβ3 isoform. The effect of the PLCβ3 knockdown is much greater for C5a than for UDP (D).While GRK does not desensitize the P2Y receptor in our model, it is a buffer for Gβγ released from Gαq. Reducing the amount of GRK will shift the equilibrium towards more Gβγ bound to PLCβ3 and thus more calcium release even though GRK does not directly feed back on the P2Y6 receptor. The top-right panel in Figure 4 shows that, based on the model, the peak intracellular calcium concentration is expected to be very slightly higher in the GRK2 knockdown line when stimulated by 25 µM UDP. A comparison of the experimental peak heights of the wild-type and GRK knockdown cell line data by t-test cannot reject the null hypothesis that the peak heights are equal (p = 0.9963). The effect of the GRK knockdown is expected to be so slight that the effect size is overwhelmed by the measurement error in the data. The effect of the uncertainty in the GRK2 knockdown fraction impacts the range of the confidence intervals of the predicted C5a response much more than the confidence intervals of the predicted UDP response which is consistent with GRK2 being a more significant component of the C5a response.Our model structure has PLCβ3 stimulated by either Gβγ or Gαq. Because the C5a response signals only through PLCβ3 the effect of the knockdown is expected to be more pronounced for the C5a response than for the UDP response. The bottom-left panel of Figure 4 confirms that the model prediction is consistent with the representative experiment. The UDP response activates PLCβ3 through Gβγ, but also activates PLCβ3 and PLCβ4 with Gαq. Therefore, we expect that the calcium response should be more robust to perturbations in just one of the PLCβ isoforms. The UDP response in the PLCβ3 knockdown line (bottom right panel of Figure 4) shows that our model predicts the knockdown effect to be small relative to the total magnitude of the response in part due to the redundancy in the use of PLCβ isoforms in the UDP response.Because this dataset was used for parameter estimation, the fit of model to the data may overstate the accuracy of the model. Nonetheless, the good fit does suggest that the model warrants being tested in truly predictive experiments; we describe such experiments in the following section.Double Ligand ExperimentsWe examine our model response to a simultaneous stimulation by C5a and UDP because it has been shown experimentally that macrophage cells respond synergistically to such conditions [40]. To quantify the amount of synergy or non-additivity that is present in the calcium response, a synergy ratio is computed for each ligand dose pair. The numerator of the ratio is the peak offset from baseline of the intracellular calcium concentration. The denominator of the ratio is the sum of the peak offsets when the cell or model is stimulated with only one ligand. A synergy is present when the ratio is greater than one implying the peak height is greater than expected from an additive combination of ligand effects. While this is certainly not the only possible measure of synergy it is widely adopted and has been used in previous studies on calcium synergy [40].The left panel of Figure 5 shows the results of model simulations at nominal parameters for a grid of doses of C5a and UDP. In the dose response surface, there is a ridge of synergistic calcium release for a moderate dose of UDP. We tested the model prediction with the experiment design measuring the synergy ratio at the points denoted as black open circles in the left panel of Figure 5. A χ2 goodness-of fit test comparing the model expected synergy ratio to the observed synergy ratio fails to reject the null hypothesis that the data were generated by the model mechanism (p-value≈1.0). The root-mean-squared error (RMSE) deviation between the predicted and actual experimental data is 0.492. By way of comparison, the RMSE between the data and the null model of no synergy is 1.044. We therefore conclude that the model predictions are consistent with the experimental observations. It should be noted that measurements of synergy in RAW cells are noisy and the ridge occurs at low doses of UDP. Notwithstanding, the phenomenon has been reported [40] and has been observed by us in this cell line.10.1371/journal.pcbi.1000185.g005Figure 5The model is used as a predictive tool to infer the effect of stimulating the cell simultaneously with UDP and C5a that signal through the Gαq and Gαi pathways, respectively.Synergy was measured as the ratio of peak height offset from baseline attained from simultaneous stimulation to the peak height offset calculated by the sum of the responses to each ligand individually. (A) Expected synergy ratio as a function of UDP and C5a dose (truncated at 1.5). The simulations show a ridge of synergy at a moderate UDP dose for most C5a doses. The black circles indicate dose combinations points of experiments that were conducted to test the model. (B) Expected synergy ratio as a function of UDP and C5a dose for a simulated GRK2 knockdown cell line. Without the GRK-mediated negative feedback to keep the IP3 generation from the C5a receptor within the non-linear range of calcium release the ridge in the synergy dose response is diminished. The synergy in the GRK knockdown simulation is not entirely eliminated because the shRNAi knockdown of GRK does not constitute a complete loss-of-function and low concentrations of ligand are still able to synergize. Furthermore, the asymmetric synergy dose response surface is more symmetric in the GRK knockdown simulation because the asymmetric calcium-dependent feedback mechanism is reduced.The right panel of Figure 5 shows the same synergy dose response surface but for a GRK knockdown cell line. The synergy ridge observed in the wild-type cell simulation is changed in the GRK knockdown simulation indicating the C5a receptor desensitization mechanism mediated by GRK is important for the synergistic release of calcium. In the next section we pursue this conclusion in more detail, developing a conceptual explanation of the mechanism of crosstalk and synergy within our model.DiscussionG-protein-coupled receptors form a complex network of interacting proteins that generally exhibits the properties of a system in which each receptor signal is buffered from the others. For a minority of ligand combinations, however, crosstalk between pairs of receptors is apparent. Due to the complexity and importance of the system many hypothetical mechanisms have been proposed to explain the crosstalk [2]. In particular, simultaneous Gβγ and Gαq binding to PLCβ [20] and Gβγ exchange between Gαi and Gαq-coupled receptors have been proposed as potential mechanisms [19]. While our model does not eliminate these potential mechanisms, we do show that the mechanism represented in our model is consistent with a full range of experimental data including a variety of doses of C5a and UDP, C5a and UDP stimulation of five different knockdown cell-lines and double-ligand dose response experiments.To our knowledge, this is the first multireceptor GPCR model and the first to address the complex phenomenon of crosstalk between GPCR receptor pathways that has been statistically estimated and validated with experimental data. This important phenomenon plays a role in processes as diverse as chemotaxis and perhaps drug interactions. In our model, the primary mechanism of synergy is due to the cooperative opening of the IP3 receptor. The robustness of the synergy is due to the feedback of GRK on the C5a receptor and the specificity of the synergy is due to the interaction patterns between specific Gα isoforms and PLCβ isoforms. The simultaneous binding model [20] accounts for the specificity of synergy, but not the robustness pattern of the synergy.We observe in the model that if the Gαq-PLCβ3-Ca2+ and Gαq-PLCβ4-Ca2+ binding reactions are inhibited, the system still exhibits synergy. We conclude from this observation that the crosstalk mechanism is mediated by Gβγ. If the binding reaction of Gβγ to phosphorylated GRK2 is removed, the synergy is eliminated. Furthermore, if the GRK2-mediated phosphorylation of complexed C5a receptors is removed, the double ligand response is additive. We deduce then that the synergy mechanism involves GRK2 phosphorylation of complexed C5a receptors. However, GRK2 phosphorylation does not entirely explain the synergy mechanism.In our model, the calcium released from the IP3 receptor is a function of the number of receptor molecules complexed to IP3 raised to the fourth power [41]. Therefore, for a small range of IP3 concentration, the amount of Ca2+ released is more than additive (see Figure S8). We conclude from our analysis of the model that the synergy ridge in Figure 5 arises because the GRK2 mediated mechanism holds the IP3 concentration in this non-additive region for most concentrations of C5a. The UDP response does not have the GRK2 mediated feedback and thus only shows a synergistic response for a small range of UDP concentration. If the GRK2 desensitization is removed from the model, the synergy ridge is removed and synergy is only present at low doses of C5a and UDP (see Figure 5).The Bayesian method we have used for this model has several advantages for the estimation of model parameters in complex mechanistic system models. We have used an informative prior to exclude negative rate constants from the permitted parameter space. We have also used the prior distribution to center our a priori expectations of the rate constant at values obtained from in-vitro and other biochemical experiments. The Bayesian update rule allowed us to estimate parameters with our best current dataset and then update those estimates as new data became available from the calcium assay. In this way, we were able to iteratively refine and recalibrate our model with the most recent data available during data collection period for this project. The posterior distribution provides not only an estimate of the rate constants, but the entire distribution, from which we can calculate highest posterior confidence intervals and posterior correlations between parameters. For example, the posterior correlation between the binding and unbinding rates for the UDP-P2Y receptor complex were highly correlated, but those two constants were uncorrelated with the corresponding rates for the C5a-C5a receptor complex reaction even though we imposed no correlations a priori. Finally, the algorithmic methods for collecting ensembles of samples from the posterior distribution have improved considerably in recent years in terms of speed and robustnessWe have shown that the signal transduction system as it is represented by our model does not require simultaneous binding of Gαq and Gβγ to PLCβ3 to cause a synergistic Ca2+ response due to simultaneous stimulation by C5a and UDP. We have shown that our representative model is consistent with this experimental dataset in RAW264.7 macrophage cells, but we have not excluded all other potential mechanisms that may be absent or regulated differently in this cell line compared to other macrophage cell lines. Indeed there are a few examples of statistical discrepancies between the model and experiments in our dataset (Table S4). These differences are substrate for further experimentation and modeling. The purpose of our model is to provide a quantitative tool to aid in reasoning about such complex interacting systems so that meaningful experiments can be designed to explore and understand the biological mechanism.Materials and MethodsThe model equations are given in Figure S7. The initial conditions and parameter values are in Table S1 and Table S2, respectively. All the data used in this work and a stand-alone implementation of the model is provided at http://genomics.lbl.gov/supplemental/flaherty-gpcr/. The model was simulated using CVODE [42] and the GNU Scientific Library. Further details on materials and methods are available in Dataset S1.Experimental MethodsIntracellular free calcium in cultured adherent RAW264.7 cells was measured in a 96-well plate format using the Ca2+-sensitive fluorescent dye Fura-2 [43],[44]. A Molecular Devices FLEXstation scanning fluorometer was used to measure fluorescence using a bottom read of a 96-well plate. Each well was sampled approximately every 4 seconds. The measurement protocol is described in AfCS experimental protocol ID #PP00000211 (available from http://www.signaling-gateway.org). The parameters in ligand concentration model were estimated using FITC solution in the FLEXstation scanning fluorometer as described in Molecular Devices Maxline Application Note #45 and in Protocol S1 (see also Figure S5).Statistical InferenceTwenty of the 84 parameters were chosen to be estimated from data based on relevance to the experimental hypothesis. Only those parameters that related to the knockdown experiments in the dataset were estimated and are denoted with a star in Table S2. We used data to estimate only the two forward rate constants in the enzymatic mass-action equations because the forward and reverse rate constants for a given reaction will be highly correlated in the posterior distribution making estimation by Markov chain methods computationally expensive. An analysis of the sensitivity of the model to each parameter is shown in Figure S9.For each estimated parameter we constructed an independent Gaussian prior on a log scale with a mean chosen based on relevant literature and a standard deviation of 0.25. We found that this prior variance was sufficiently permissive to allow exploration of the space while still constraining the rates to be physically reasonable. The prior distribution over the parameters allows the incorporation of both soft and hard constraints in the parameter estimates. Parameter sets with zero measure are not permitted in the posterior distribution and parameter sets with small measure must be assigned a large likelihood in order to have a large posterior probability.The likelihood is a function of the parameters (θ) and links the prior distribution with the posterior distribution under Bayes rulewhere y denotes the observed data.In our model, the likelihood function is a Gaussian distribution according to the non-linear regression equation y = f(θ)+ε, ε∼N(0,σ\n2), where f(θ) is the deterministic model prediction. The posterior distribution is of interest because it informs us as to the most probable setting of the parameters as well as the uncertainty in the values.The Metropolis-Hastings algorithm [45] was used to estimate the posterior density of the parameters Pr(θ|y). Three independent chains were simulated from different initial parameter values (see Figure S1). To assess convergence of the posterior distribution estimate, we used the Gelman-Rubin potential scale reduction factor (PSRF) [46]. The multivariate PSRF is 2.44 and 95% of the individual PSRFs were less than 1.5. A PSRF value of one indicates that the distribution has converged and values near one are close to converged.Posterior prediction confidence intervals were constructed using the percentiles from the predictive distribution approximated with 2000 Monte Carlo samples from Pr(y\nnew|θi) at each of 100 simple random samples from Pr(θ|y) obtained fromwhere Pr(y\nnew|θi)∼N(f(θ),s\n2) and s\n2 is the pooled variance estimate, which is computed as an average of the variances of all the time points in each of the 29 wild-type experiments. These average variances were weighted by the number of technical replicates in each experiment and then averaged to yield the estimate s\n2. A small factor of 1 nM2 was added to each variance estimate to bound variance estimates away from zero.Supporting InformationDataset S1(0.11 MB DOC)Click here for additional data file.Figure S1This figure shows exemplar MCMC realizations for parameter k109f (the UDP+P2YR forward binding rate) from three independent chains. The chains have converged to the stationary distribution which is the posterior distribution as measured by the PSRF (see Materials and Methods).(0.20 MB DOC)Click here for additional data file.Figure S2Posterior distributions and correlations The first figure shows that the pairwise marginal posterior distributions for the ligand binding reactions for P2YR and C5aR. The posterior distributions show the dissociation constants for the reactions are tightly constrained by the data, while the values of the forward and reverse rates that make up the ratio are not as well constrained by the data. Additionally, the UDP binding rates are not correlated with the C5a binding rates. k108f and k108r are the P2YR forward and reverse rates and k101f and k101r are the C5aR rates. The next two figures show the one-way marginal posterior density estimates from three independent MCMC chains with approximately 30,000 samples. The 20 estimates parameters are along the rows and the independent chains are along the columns. In each plot, the light blue density is the prior density and the green, purple and orange densities are the posterior densities. The vertical line shows the parameter value used in the model simulations in the paper and listed in Table S3. All of the densities are plotted on a log scale. Each marginal posterior distribution estimate is constructed from independent MCMC chains. The results from each chain (three of them) are shown in the columns of the second figure below. In some cases the algorithm sampled heavily from one mode that was not explored as heavily by another chain. However, the PSRF criterion used to assay convergence and a visual inspection of overall posterior density correspondence do indicate that the posterior distributions are sufficiently sampled by all three chains in aggregate. Furthermore, the fit of the model to the data as shown in Figure S3 shows that the model point estimates are effective in fitting the actual calcium measurements.(0.55 MB DOC)Click here for additional data file.Figure S3Peak height dose response. This figure shows the single ligand calcium dose responses for C5a and UDP stimulation.(0.21 MB DOC)Click here for additional data file.Figure S4Knockdown simulations. This figure shows representative simulations and data for each knockdown experiment. A complete set of all 96 experiments is provided in a supplementary folder.(0.69 MB DOC)Click here for additional data file.Figure S5Input model fit. This figure shows the input model (described in Materials and Methods) fit to the FITC measurements. The ligand concentration that the cell sees does not transit instantaneously from 0 to the final concentration. The ligand concentration is expected to take an amount of time that is significant on the scale of the measurements made for this study to reach the final concentration.(0.12 MB DOC)Click here for additional data file.Figure S6Large pathway diagram.(0.18 MB DOC)Click here for additional data file.Figure S7System of differential equations. This figure shows the complete set of differential equations used to simulate the model. These equations are also available in the source c code for the model supplied. This system of equations with the initial conditions and nominal parameter values reported in Table S1 and Table S2, respectively, completely define the model and allow for the reproduction of the simulations used in this paper on any platform.(1.62 MB DOC)Click here for additional data file.Figure S8Hill function self-synergy. Consider a Hill function, .  is a dimensionless critical concentration y*, below which self-synergy will occur. Based on the analysis, we conclude that: (i) n must be greater than 1 for self-synergy to occur, (ii) self synergy never occurs if the concentration x exceeds equilibrium constant K (y>1), and (iii) for n>2, there is a large range of concentration for self-synergy. In the G protein model, x, is the concentration of IP3-IP3R, H(x) is the rate of change in cytosolic calcium concentration and n = 4. We have tested the validity of this self synergy hypothesis by stimulating the cells with both 20 nM UDP and 40 nM UDP (data not shown). Though at such low ligand concentrations, the measurement variability is high, we observed that the synergy ratio, on average was 1.17 compared to a value of 1.25 predicted by the model.(0.02 MB DOC)Click here for additional data file.Figure S9Parameter sensitivity analysis. The parameter of interest is varied by 10% while all other parameters are kept constant. The parameters are grouped according to their functionalities. The sensitivity coefficient is the ratio of the relative change in the peak height to the relative change in the parameter value. The four most sensitive parameters (sensitivity coefficient >2) in the Cacyt category are Vqssk50 (IP3+IP3K_a->IP4+IP3K_a (Vmax)), Kqssk50 (IP3+IP3K_a->IP4+IP3K_a (Km)), a1 (Ca leak into the cell from outside), and Kex (Na/Ca exchange activation const). The top 3 most sensitive parameters in the PLCb3 category are: k21bf* (PLCb3_Ca_Gbg_PIP2->PLCb3_Ca_Gbg+IP3+DAG), k20f (Gbg+PLCb3_Ca->PLCb3_Ca_Gbg), k21af* (PLCb3_Ca_Gbg+PIP2->PLCb3_Ca_Gbg_PIP2). A star next to the parameter name indicates it was estimated.(0.04 MB DOC)Click here for additional data file.Protocol S1FITC protocol.(0.03 MB DOC)Click here for additional data file.Table S1Model initial conditions. This table shows the initial conditions used for the model. The model was run for sufficient time for the species states in the model to reach equilibrium before ligand stimulation was added. The number of molecules was calculated using a cell volume of 1 pL.(0.05 MB DOC)Click here for additional data file.Table S2Model parameters. This table shows the nominal parameters used for the model. Parameter distributions that were estimated are shown as shaded rows and with a star next to the parameter name in the table. The prior distribution for each parameter is as described in the Materials and Methods section with mean value specified by the column labeled “prior”.(0.14 MB DOC)Click here for additional data file.Table S3Parameter posterior uncertainty and references. This table shows the HPD intervals as computed by the R CODA library function “hpdinterval”. HPD intervals for each of the three MCMC chains were calculated and the union of those intervals is reported for each parameter in this table. The prior value reported in Table S2 was set using information from references listed in the appropriate column. The references used to form the basis of the parameter estimates are shown in the last column.(0.37 MB DOC)Click here for additional data file.Table S4Goodness of fit evaluation. We use the mean squared error criterion to evaluate the goodness of our model fit to the data. We have used this data in the estimation procedure and thus does not constitute a true validation. However, we show that in general our model fits the bulk of the data. Those areas of lack-of-fit are usually due to extraordinary experiment-to-experiment variation and in some cases point to unaccounted mechanisms. We elaborate on one such mechanism (multiple GRK isoforms) in the text of the article.(0.27 MB DOC)Click here for additional data file.\n\nREFERENCES:\n1. KroezeWKShefflerDJRothBL\n2003\nG-protein-coupled receptors at a glance.\nJ Cell Sci\n116\n4867\n4869\n14625380\n2. WerryTDWilkinsonGFWillarsGB\n2003\nMechanisms of cross-talk between G-protein-coupled receptors resulting in enhanced release of intracellular Ca2+.\nBiochem J\n374\n281\n296\n12790797\n3. CaseyPJGilmanAG\n1988\nG protein involvement in receptor-effector coupling.\nJ Biol Chem\n263\n2577\n2580\n2830256\n4. WuDKatzASimonMI\n1993\nActivation of phospholipase C β2 by the α and βγ subunits of trimeric GTP-binding protein.\nProc Natl Acad Sci U S A\n90\n5297\n5301\n8389480\n5. MukhopadhyaySRossEM\n1999\nRapid GTP binding and hydrolysis by Gq promoted by receptor and GTPase-activating proteins.\nProc Natl Acad Sci U S A\n96\n9539\n9544\n10449728\n6. PattersonRLBoehningDSnyderSH\n2004\nInositol 1,4,5-trisphosphate receptors as signal integrators.\nAnnu Rev Biochem\n73\n437\n465\n15189149\n7. AnanthanarayananBStahelinRVDigmanMAChoW\n2003\nActivation mechanisms of conventional protein kinase C isoforms are determined by the ligand affinity and conformational flexibility of their C1 domains.\nJ Biol Chem\n278\n46886\n46894\n12954613\n8. PenelaPRibasCMayorFJr\n2003\nMechanisms of regulation of the expression and function of G protein-coupled receptor kinases.\nCell Signal\n15\n973\n981\n14499340\n9. PitcherJATouharaKPayneESLefkowitzRJ\n1995\nPleckstrin homology domain-mediated membrane association and activation of the β-adrenergic receptor kinase requires coordinate interaction with Gβγ subunits and lipid.\nJ Biol Chem\n270\n11707\n11710\n7744811\n10. BergJMTymoczkoJLStryerL\n2002\nBiochemistry. 5th ed\nBethesda (Maryland)\nW.H. Freeman; National Center for Biotechnology Information\n11. AllegrettiMMoriconiABeccariARDi BitondoRBizzarriC\n2005\nTargeting C5a: recent advances in drug discovery.\nCurr Med Chem\n12\n217\n236\n15638737\n12. WarnyMAboudolaSRobsonSCSevignyJCommuniD\n2001\nP2Y6 nucleotide receptor mediates monocyte interleukin-8 production in response to UDP or lipopolysaccharide.\nJ Biol Chem\n276\n26051\n26056\n11349132\n13. YoshiokaKSaitohONakataH\n2001\nHeteromeric association creates a P2Y-like adenosine receptor.\nProc Natl Acad Sci U S A\n98\n7617\n7622\n11390975\n14. LukasTJ\n2004\nA signal transduction pathway model prototype I: from agonist to cellular endpoint.\nBiophys J\n87\n1406\n1416\n15345523\n15. MishraJBhallaUS\n2002\nSimulations of inositol phosphate metabolism and its interaction with InsP3-mediated calcium release.\nBiophys J\n83\n1298\n1316\n12202356\n16. LemonGGibsonWGBennettMR\n2003\nMetabotropic receptor activation, desensitization and sequestration—I: modelling calcium and inositol 1,4,5-trisphosphate dynamics following receptor activation.\nJ Theor Biol\n223\n93\n111\n12782119\n17. MauryaMRSubramaniamS\n2007\nA kinetic model for calcium dynamics in RAW 264.7 cells: 2. Knockdown response and long-term response.\nBiophys J\n93\n729\n740\n17483189\n18. MauryaMRSubramaniamS\n2007\nA kinetic model for calcium dynamics in RAW 264.7 Cells: 1. Mechanisms, parameters and sub-populational variability.\nBiophys J\n93\n709\n728\n17483174\n19. QuittererULohseMJ\n1999\nCrosstalk between Gαi- and Gαq-coupled receptors is mediated by Gβγ exchange.\nProc Natl Acad Sci U S A\n96\n10626\n10631\n10485876\n20. ZhuXBirnbaumerL\n1996\nG protein subunits and the stimulation of phospholipase C by Gs- and Gi-coupled receptors: lack of receptor selectivity of Gα16 and evidence for a synergic interaction between Gβ�� and the α subunit of a receptor activated G protein.\nProc Natl Acad Sci U S A\n93\n2827\n2831\n8610126\n21. KehrlJH\n1998\nHeterotrimeric G protein signaling: roles in immune function and fine-tuning by RGS proteins.\nImmunity\n8\n1\n10\n9462506\n22. RossEMWilkieTM\n2000\nGTPase-activating proteins for heterotrimeric G proteins: regulators of G protein signaling (RGS) and RGS-like proteins.\nAnnu Rev Biochem\n69\n795\n827\n10966476\n23. CunninghamMLWaldoGLHollingerSHeplerJRHardenTK\n2001\nProtein kinase C phosphorylates RGS2 and modulates its capacity for negative regulation of Gα11 signaling.\nJ Biol Chem\n276\n5438\n5444\n11063746\n24. KehrlJHSinnarajahS\n2002\nRGS2: a multifunctional regulator of G-protein signaling.\nInt J Biochem Cell Biol\n34\n432\n438\n11906816\n25. KeizerJDe YoungGW\n1992\nTwo roles of Ca2+ in agonist stimulated Ca2+ oscillations.\nBiophys J\n61\n649\n660\n1324019\n26. MeyerTHolowkaDStryerL\n1988\nHighly cooperative opening of calcium channels by inositol 1,4,5-trisphosphate.\nScience\n240\n653\n656\n2452482\n27. MeyerTStryerL\n1988\nMolecular model for receptor-stimulated calcium spiking.\nProc Natl Acad Sci U S A\n85\n5051\n5055\n2455890\n28. JiangHKuangYWuYSmrckaASimonMI\n1996\nPertussis toxin-sensitive activation of phospholipase C by the C5a and fMet-Leu-Phe receptors.\nJ Biol Chem\n271\n13430\n13434\n8662841\n29. ParkDJhonDYLeeCWLeeKHRheeSG\n1993\nActivation of phospholipase C isozymes by G protein βγ subunits.\nJ Biol Chem\n268\n4573\n4576\n8383116\n30. SmrckaAVSternweisPC\n1993\nRegulation of purified subtypes of phosphatidylinositol-specific phospholipase C β by G protein α and βγ subunits.\nJ Biol Chem\n268\n9667\n9674\n8387502\n31. LeeCWLeeKHLeeSBParkDRheeSG\n1994\nRegulation of phospholipase C-β4 by ribonucleotides and the α subunit of Gq.\nJ Biol Chem\n269\n25335\n25338\n7929227\n32. JiangHWuDSimonMI\n1994\nActivation of phospholipase C β4 by heterotrimeric GTP-binding proteins.\nJ Biol Chem\n269\n7593\n7596\n8125982\n33. RunnelsLWScarlataSF\n1999\nDetermination of the affinities between heterotrimeric G protein subunits and their phospholipase C-β effectors.\nBiochemistry\n38\n1488\n1496\n9931014\n34. RheeSG\n2001\nRegulation of phosphoinositide-specific phospholipase C.\nAnnu Rev Biochem\n70\n281\n312\n11395409\n35. YueCKuCYLiuMSimonMISanbornBM\n2000\nMolecular mechanism of the inhibition of phospholipase C β3 by protein kinase C.\nJ Biol Chem\n275\n30220\n30225\n10893237\n36. LitoschI\n2002\nNovel mechanisms for feedback regulation of phospholipase C-β activity.\nIUBMB Life\n54\n253\n260\n12587975\n37. SpitalerMCantrellDA\n2004\nProtein kinase C and beyond.\nNat Immunol\n5\n785\n790\n15282562\n38. LangkabelPZwirnerJOppermannM\n1999\nLigand-induced phosphorylation of anaphylatoxin receptors C3aR and C5aR is mediated by G protein-coupled receptor kinases.\nEur J Immunol\n29\n3035\n3046\n10508278\n39. ShinKJWallEAZavzavadjianJRSantatLALiuJ\n2006\nA single lentiviral vector platform for microRNA-based conditional RNA interference and coordinated transgene expression.\nProc Natl Acad Sci U S A\n103\n13759\n13764\n16945906\n40. NatarajanMLinKMHsuehRCSternweisPCRanganathanR\n2006\nA global analysis of cross-talk in a mammalian cellular signalling network.\nNat Cell Biol\n8\n571\n580\n16699502\n41. De YoungGWKeizerJ\n1992\nA single-pool inositol 1,4,5-trisphosphate-receptor-based model for agonist-stimulated oscillations in Ca2+ concentration.\nProc Natl Acad Sci U S A\n89\n9895\n9899\n1329108\n42. HindmarshACBrownPNGrantKELeeSLSerbanR\n2005\nSUNDIALS: Suite of Nonlinear and Differential/Algebraic Equation Solvers.\nACM Trans Math Softw\n31\n363\n396\n43. TsienRY\n1989\nFluorescent indicators of ion concentrations.\nMethods Cell Biol\n30\n127\n156\n2538708\n44. GrynkiewiczGPoenieMTsienRY\n1985\nA new generation of Ca2+ indicators with greatly improved fluorescence properties.\nJ Biol Chem\n260\n3440\n3450\n3838314\n45. RobertCPCasellaG\n2004\nMonte Carlo Statistical Methods\nNew York\nSpringer\n46. GelmanARubinDB\n1992\nInference from iterative simulation using multiple sequences.\nStat Sci\n7\n457\n472"
+}

batch_8/PMC2528965.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2528965",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2528965\nAUTHORS: Warren J. Ewens, Mingyao Li, Richard S. Spielman\n\nABSTRACT:\nQuantitative trait transmission/disequilibrium tests (quantitative TDTs) are commonly used in family-based genetic association studies of quantitative traits. Despite the availability of various quantitative TDTs, some users are not aware of the properties of these tests and the relationships between them. This review aims at outlining the broad features of the various quantitative TDT procedures carried out in the frequently used QTDT and FBAT packages. Specifically, we discuss the “Rabinowitz” and the “Monks-Kaplan” procedures, as well as the various “Abecasis” and “Allison” regression-based procedures. We focus on the models assumed in these tests and the relationships between them. Moreover, we discuss what hypotheses are tested by the various quantitative TDTs, what testing procedures are best suited to various forms of data, and whether the regression-based tests overcome population stratification problems. Finally, we comment on power considerations in the choice of the test to be used. We hope this brief review will shed light on the similarities and differences of the various quantitative TDTs.\n\nBODY:\nIntroductionThis review aims at outlining the broad features of various frequently used quantitative transmission/disequilibrium tests (quantitative TDTs). We focus on describing the models assumed in these tests and the relationships between the tests. It is impossible in a brief review to describe and compare the great variety of quantitative TDT procedures to be found in the literature and that are available in computer packages, because some of these procedures involve quite complex forms of data and a sophisticated statistical analysis. We have therefore deliberately restricted the scope of this review by considering only one simple form of data—namely, family trios—that is families with a mother, a father, and exactly one child. Further, we consider only the basic features of the procedures that we discuss. Even with these limitations, a number of interesting questions arise, some of which we raise but do not answer. We leave a deeper analysis of the procedures discussed here, and an analysis of the more complex procedures that we do not consider, to another occasion.The aim of the original qualitative TDT procedure [1] was to test for linkage (and linkage disequilibrium) between a marker locus and a disease locus in a way that overcomes problems arising from potential population stratification. We assume the same null hypothesis for the quantitative TDT procedures considered here (see below for details). For convenience, we focus here on quantitative TDT procedures carried out in the frequently used FBAT (http://biosun1.harvard.edu/~fbat/fbat.htm) and QTDT (http://www.sph.umich.edu/csg/abecasis/QTDT/) packages. Specifically we discuss the “Rabinowitz” [2] and the “Monks-Kaplan” [3] procedures, as well as various “Abecasis” [4] and “Allison” [5] regression-based procedures, when applied specifically to family trio data. The questions addressed in this review are as follows:What are the properties of the various Allison/Abecasis regression-based tests and the Rabinowitz and the Monks-Kaplan tests?What is the relation between the regression-based tests and the Rabinowitz and Monks-Kaplan tests?What hypotheses are tested by the various quantitative TDT procedures that we describe?What testing procedures are best suited to various forms of data?Do the regression-based tests that we describe overcome population stratification problems?What power considerations arise in the choice of the test to be used?We note that some of these issues were previously considered by Lange et al. [6] from a different viewpoint.Notation and DataThe notation used in the various quantitative TDT papers on which our comments are based is not consistent from one author to another, and we adopt a unifying notation that is loosely based on that of these papers. In accordance with standard statistical practice, we use upper case notation for random variables and the corresponding lower case notation for the observed values of these random variables. To focus on the main points in this expository review, we assume a specific (and restricted) form of data. We assume that the data concern a marker locus “A,” having two possible alleles, denoted by A and a, and consist of information on n family trios, with complete marker locus genotype information on the two parents and the child in each trio. The value of the quantitative trait of interest is known for the child in each trio but not for the parents. We assume, in line with the original qualitative TDT, that all parental mating types are informative (i.e., contain at least one Aa parent). The observed number of A alleles in the child in trio i is denoted by xi (i = 1, 2,…, n), and the observed value of the quantitative trait of interest in the child in trio i is denoted by yi.We do not consider here the extent to which the comments made below carry over to data other than those described above, for example cases where several children are observed in each family, where parental phenotype information is available, and where the data contain families with noninformative mating types. We restrict our analysis in this way so as to highlight the main features of the testing procedures that we discuss without getting into the analyses required for forms of data more complicated than those we consider.The null hypothesis tested is “no linkage (or no linkage disequilibrium) between the marker locus and a locus involved with the quantitative trait.” Under this null hypothesis, the mean number of A alleles, Xi, in the child in trio i will depend on the parental mating type, being 0.5 if it is aa×Aa, 1.0 if it is Aa×Aa, and 1.5 if it is AA×Aa. The null hypothesis variance  of Xi also depends on the parental mating type, being 0.25 if it is aa×Aa or AA×Aa and 0.5 if it is Aa×Aa. We frequently use the convenient Abecasis [5] notation Wi (“within family”) to describe Xi minus its null hypothesis mean as computed from the mating type in trio i. The null hypothesis mean of Wi is zero and the null hypothesis variance of Wi is the same as that for Xi, and depends on the mating type in trio i. When discussing the typical family we drop the suffix i and use the generic notation W, w, X, x, Y, and y.Quantitative TDT ModelsIn this section, we describe in algebraic terms the various quantitative TDT procedures outlined above and address questions raised in the Introduction.Properties of various quantitative TDT proceduresWe start with five Allison and Abecasis “regression-based” procedures. (The Abecasis “total” test is not a TDT test [as is indicated in the QTDT package documentation], so we ignore it.) These all assume a regression model where the mean of the phenotype Yi of the child in trio i depends on the actual value wi for that child, often along with other information, for example, the parental mating type of trio i. More precisely, the five models that we consider are as follows:The Abecasis “within only” model, denoted here Ab-Wthn. In this model, Y is assumed to depend only on the value of w, the observed difference between the number of A alleles in any child and the null hypothesis mean of this number, given the parental mating type for this child.The Abecasis “orthogonal” model, denoted here Ab-Orth. In this model, Y is assumed to depend on w and also a linear term describing parental mating type.The Abecasis “dominance” model, denoted here by Ab-Dom. This model generalizes Ab-Orth in that Y is assumed also to depend on whether the child in any trio is a homozygote or a heterozygote.The first Allison model, denoted here Al-Lin. This is a “general/linear” model, where Y is assumed to depend on parental mating type in an unspecified way and also on w.The second Allison model (his TDTQ5), denoted here Al-Quad. This is a “general/quadratic” model, and extends Al-Lin by assuming that Y depends also on w\n2. In algebraic terms, the assumptions for the “full,” or alternative, hypothesis case of these models can be written in terms of the three mating types considered as follows:\nAbecasis “within only” model (Ab-Wthn).For all parental mating types: Y = μ+β\n1\nw+E.\nAbecasis “orthogonal” model (Ab-Orth).For aa×Aa parental mating type: Y = μ+β\n1\nw+E;For Aa×Aa parental mating type: Y = μ+α+β\n1\nw+E;For AA×Aa parental mating type: Y = μ+2α+β\n1\nw+E.\nAbecasis “dominance” model (Ab-Dom).For aa×Aa parental mating type: Y = μ+β\n1\nw+γd+E;For Aa×Aa parental mating type: Y = μ+α+β\n1\nw+γd+E;For AA×Aa parental mating type: Y = μ+2α+β\n1\nw+γd+E.(In this model, d = −1 for a homozygous child and +1 for a heterozygous child, and corresponds to Wd in the QTDT package documentation. For the data that we consider, the Bd term in QTDT package documentation for this model is a constant across the three mating types, and is thus absorbed into the constant μ.)\nAllison “general/linear” model (Al-Lin).For aa×Aa parental mating type: Y = μ+β\n1*x+E;For Aa×Aa parental mating type: Y = μ+α\n1+β\n1*x+E;For AA×Aa parental mating type: Y = μ+α\n2+β\n1*x+E.\nAllison “general/quadratic” model (Al-Quad).For aa×Aa parental mating type: Y = μ+β\n1*x+β\n2*x\n2+E;For Aa×Aa parental mating type: Y = μ+α\n1+β\n1*x+β\n2*x\n2+E;For AA×Aa parental mating type: Y = μ+α\n2+β1*x+β\n2*x\n2+E.In all four models Greek symbols describe unknown parameters and E is a random residual term having mean zero and (unknown) variance . Because of the relations w = x−1/2 for Aa×aa matings, w = x−1 for Aa×Aa matings, and w = x−3/2 for Aa×AA matings, Al-Lin and Al-Quad can be re-written conveniently as:\nAllison “general/linear” model (Al-Lin).For aa×Aa parental mating type: Y = μ\n1+β\n1\nw+E;For Aa×Aa parental mating type: Y��= μ\n2+β\n1\nw+E;For AA×Aa parental mating type: Y = μ\n3+β\n1\nw+E.\nAllison “general/quadratic” model (Al-Quad).For aa×Aa parental mating type: Y = μ\n1+(β\n1+β\n2)w+β\n2\nw\n2+E;For Aa×Aa parental mating type: Y = μ\n2+(β\n1+2β\n2)w+β\n2\nw\n2+E;For AA×Aa parental mating type: Y = μ\n3+(β\n1+3β\n2)w+β\n2\nw\n2+E.The null hypothesis tested in the Ab-Wthn, Ab-Orth, and Al-Lin models is β\n1 = 0, and in the Ab-Dom model is β\n1 = γ = 0. The null hypothesis tested in the original Al-Quad model is , and this is equivalent to β\n1 = β\n2 = 0 in the re-written version above. The testing procedures in all five cases follow standard multiple regression methods, with mating type membership denoted with indicator variables. The null hypothesis model in each case removes a certain sum of squares for the phenotypic measurements in the children, and the full model removes a larger (or in rare cases, an equal) sum of squares. The difference between these two sums of squares forms the key component of the numerator of the F statistic used in all testing methods. This component is divided by the respective “model” degrees of freedom, which is equal to the number of extra parameters in each full model compared to the number in the corresponding null hypothesis model. This number takes the value 1 for Ab-Wthn, Ab-Orth, and Al-Lin and takes the value 2 for Ab-Dom and Al-Quad. This division by 2 tends to lead to smaller F ratios for Ab-Dom and Al-Quad, and thus to reduce power, and it is a trade-off against the increased generality of those models. This point is discussed further below.The use of the F distribution to determine the significance of the observed value of the F statistic is appropriate only if the data have a normal distribution. For cases where the data are taken from one extreme tail of some distribution, for example very large values of the quantitative measurement, this might be an unreasonable assumption. This matter is discussed further below.The three Abecasis models are nested, with Ab-Wthn being a special case of Ab-Orth, which in turn is a special case of Ab-Dom. Similarly Al-Lin is a special case of Al-Quad. Ab-Wthn and Ab-Orth are also special cases of Al-Lin. The nesting property is reflected in the residual degrees of freedom for the respective models: under the assumptions we have made concerning the data analyzed, the Ab-Wthn model has n−2 residual degrees of freedom, the Ab-Orth model has n−3 residual degrees of freedom, the Al-Lin and the Ab-Dom models have n−4 residual degrees of freedom, and the Al-Quad model has n−5 residual degrees of freedom.There are regression-based models that are more general than those discussed above. A model more general than Ab-Dom and Al-Lin, and including these as particular cases, is:“General/dominance” model.For aa×Aa parental mating type: Y = μ\n1+β\n1\nw+γδ+E;For Aa×Aa parental mating type: Y = μ\n2+β\n1\nw+γδ+E;For AA×Aa parental mating type: Y = μ\n3+β\n1\nw+γδ+E.This model has n−5 residual degrees of freedom. A model more general than this, with n−6 residual degrees of freedom, and which includes all regression-based models described above, allows a term in w\n2 as well as those in the “general/dominance” model. These more general models are not considered further here.We can also consider testing procedures other than those described above. In particular, we suggest a modification of the Ab-Dom test, in which the null hypothesis is changed from the present β\n1 = γ = 0 to simply β\n1 = 0. This is for two reasons. First, it does not seem natural to test simultaneously the two hypotheses that no dominance phenotypic effects exist and that y does not depend on the transmission values w. Second, in testing the hypothesis β\n1 = 0 instead of β\n1 = γ = 0, one only has a single “model” degree of freedom, leading to increased power (compared to Ab-Dom) in testing for transmission effects.We now describe the Rabinowitz [2] and Monks-Kaplan [3] procedures. In both of these procedures, the phenotype measurements y\n1, y\n2,…,yn in the children in the n trios are taken as given, and used as weights on the transmission random variables W\n1, W\n2,…,Wn. This is in direct contrast to the Abecasis [4] and Allison [5] regression-based procedures, which take the wi as given and the phenotype measurements Y\n1, Y\n2, …, Yn as random variables. It is, however, more in line with the original qualitative TDT, which also uses W\n1, W\n2, …,Wn as the random variables of interest.For the data that we consider, Rabinowitz [2] defines  by , where y̅ is the average of the yi values taken over the n children in the data. Given the observed values w\n1, w\n2,…, wn of W\n1, W\n2,…, Wn, his test statistic z is(1)In this expression, the sum (as with all sums in this article) is taken over i = 1, 2, …, n, and  is as defined above. This statistic is based on the “within family” wi values because if this is done [2], the effects of population stratification are overcome. (This parallels a similar observation in the original qualitative TDT [1].) The Rabinowitz statistic is written as a z rather than a t because, with the  taken as given, the null hypothesis standard deviation of —the term in the denominator of Equation 1 —is known. Central limit theorem arguments then show that if n exceeds about 20, the Rabinowitz statistic has an approximate N(0, 1) distribution when the null hypothesis is true.The Monks-Kaplan statistic is similar to the Rabinowitz statistic, being(2)The numerators in z and tMK are the same, but the denominator in the Monks-Kaplan statistic contains a standard deviation estimate rather than a known standard deviation. (This allows generalizations to handle data more complex than the data considered here.) It is written as a t statistic because of this fact.Relationship between regression-based tests and Rabinowitz and Monks-Kaplan testsThe Rabinowitz and the Monks-Kaplan procedures differ from the Allison and the Abecasis procedures in various ways, of which we mention two. First, and most important, they regard the Wi values as random variables with the yi values taken as given, whereas the Abecasis and Allison procedures regard the Yi values as random variables with the wi values taken as given. Second, unlike the Abecasis and Allison procedures, neither is explicitly based on regression models (see Laird et al. [7] for more details). Despite these differences, it is interesting to consider the hypothesis testing procedure in a “role-reversal” regression model of the form(3)This model, when compared to the Allison and Abecasis procedures, reverses the roles of W and Y in the regression. Because the Rabinowitz and Monks-Kaplan statistics are defined in terms of  rather than yi, it is convenient to reformulate Equation 3 equivalently as(4)The estimate of β in this regression is , and the standard regression t statistic testing for departures of β from zero is(5)where s is the usual regression estimate of the standard deviation of . The Rabinowitz statistic (Equation 1) has the same numerator as that in Equation 5 but has, in the denominator, the known null hypothesis standard deviation of  rather than a regression-based estimate of this standard deviation. The Monks-Kaplan statistic also has the same numerator as that in Equation 5, but has a standard deviation estimate in the denominator different from that in both Equation 1 and Equation 5.Despite this similarity, there are essential differences between the Allison and Abecasis regression procedures and the Rabinowitz and Monks-Kaplan procedures. The Rabinowitz procedure, and in general all FBAT procedures, use “score statistics” based on the alleles transmitted to the children, conditional on the parental genotypes and the offspring phenotype. They are not explicitly based on regression models. Under the score test approach, the null hypothesis distribution of the test statistic is calculated directly from Mendel's laws. The test statistic thus has the correct distribution so long as these laws hold, regardless of any hypothetical model for the mean and variance of the offspring phenotype.Hypotheses tested by various quantitative TDT proceduresThere is an important difference between the hypothesis being tested by all the quantitative TDT procedures described above and the original qualitative TDT. The original TDT assesses whether the sum of the wi values differs significantly from zero. By contrast, none of the quantitative TDT procedures described above assess whether the wi values (or their weighted sum in the Rabinowitz and Monks-Kaplan procedures) differ significantly from zero. This can be seen from the fact that they are all unchanged if an arbitrary constant is added to the wi values. What they do test is whether there is significant change in the value of w as y changes (or of y as w changes). This is explicit in the regression procedures and also applies for the Rabinowitz and Monks-Kaplan procedures. We discuss this fact below in the context of the data that the investigator is analyzing. By contrast, it is the intercept in the estimated regression (Equation 4), namely w̅, that is directly comparable to the qualitative TDT statistic. We now show that, if the Rabinowitz approach of using the known variance of Wi is taken, the test of whether this intercept is zero is identical to the original qualitative TDT procedure.The standard regression test statistic of the null hypothesis α = 0 in the regression model (Equation 4) is the estimated intercept (in this case w̅) divided by the estimated standard deviation of W̅. In this case, the standard deviation of W̅ is known, so the appropriate (z) test statistic is w̅ divided by the standard deviation of W̅. It is equivalent and more convenient to use z\n2 as test statistic, where z\n2 can be written as  divided by the variance of . Under the null hypothesis, z\n2 has an approximate chi-square distribution with one degree of freedom. We now calculate the value of z\n2 in terms of transmission information.We consider first the case of those trios having either Aa×AA or Aa×aa matings. If in any such trio the heterozygous parent transmits the A allele, the value of w for that trio is +1/2. We write the number of such trios as m\n1. If in any such trio the heterozygous parent transmits the a allele, the value of w for that trio is −1/2. We write the number of such trios as m\n2. The value of w in any trio where the parental mating type is Aa×Aa is +1 if both parents transmit the A allele, 0 if one parent transmits the a allele and the other parent transmits the A allele, and −1 if both transmit the a allele. We write the respective numbers of these trios as m\n3, m\n4, and m\n5. Thus  is m\n1/2−m\n2/2+m\n3−m\n5 = m\n1−m\n2+2m\n3−2m\n5/2. But this is just (b−c)/2, where b is the total number of transmissions of A from heterozygous parents and c is the total number of transmissions of a from heterozygous parents. The numerator in z\n2 is thus (b−c)2/4.We now turn to the denominator of z\n2. Suppose that in the n trios, there are exactly n\n1 where the parental mating type is Aa×Aa. Since the variance of W in any such trio is 1/2 and for all other mating types is 1/4, the variance of  is n\n1/2+(n−n\n1)/4 = [2n\n1+(n−n\n1)]/4. This may be written, using the notation of the preceding paragraph, as (b+c)/4, because b+c is the total number of transmissions from heterozygous parents. It follows from the above that z\n2 = (b−c)2/(b+c), and this is the standard qualitative TDT statistic of Spielman et al. [1].Following a similar line of reasoning, if the variance of  is estimated from the data, (as is done in the qualitative TDT procedure of Martin et al. [8], where such estimation is needed for non-trio data) the test of the hypothesis α = 0 in the regression (Equation 4) can be shown to be identical to the Martin et al. procedure [8].Testing procedures best suited to various forms of dataThe above considerations lead to a discussion of the data being analyzed. The original qualitative TDT of Spielman et al. [1] uses data only from children affected by some disease. Spielman et al. [1] also discuss an alternative to the qualitative TDT procedure when segregation distortion at the marker locus is suspected. In this alternative procedure, the proportion of transmissions of the A allele from heterozygous parents is compared not with 1/2, as in the original “standard” TDT, but with the corresponding proportion in nonaffected individuals. The analogues of “affected and not affected” in the quantitative TDT context might be “extreme and nonextreme phenotype values.” If the yi values in a quantitative TDT procedure are derived from a random sample, and if the null hypothesis is not true, one might expect extreme and nonextreme y values to tend to correspond to different w values. This might lead to a significant dependence of w on y (or equivalently of y on w). Thus for a random sample, a “slope” test such as those carried out by all quantitative TDT procedures described above appears to be appropriate. These procedures are analogous to the alternative qualitative TDT testing procedure of Spielman et al. [1].Thus if the data analyzed concern either only extremely low or extremely high yi values (but not both), which might be thought of as corresponding to “affected” children, it might be more appropriate to carry out the qualitative TDT test that uses only data from such children. As shown above, this is identical to an “intercept” regression test. One may, if desired, carry out both this procedure and a (slope) quantitative TDT test, since in a regression procedure, the test of the slope and the test of the intercept in a regression line are independent. However, extreme phenotypes might well not have a normal distribution, so that those regression-based procedures that use F tests might be unreliable. The Rabinowitz and the Monks-Kaplan procedures are not subject to this problem. The information provided jointly by the qualitative TDT and the Rabinowitz or the Monks-Kaplan tests would show whether there is significant absolute linkage disequilibrium and also a significant change in linkage disequilibrium as the phenotypic value varies.Population stratification in regression-based testsThe aim of the original qualitative TDT was to overcome potential problems arising from population stratification, and this was done by using the transmission values wi. By design, the Rabinowitz and Monks-Kaplan procedures also overcome population stratification problems, using the same approach. The situation is not, however, so straightforward for the regression models.The simplest of the regression models considered above, namely Ab-Wthn, is a regression of Y on w. Once parental mating type information has been factored out of the regression, Ab-Orth, Al-Lin, and Al-Quad also have this property. The same is true of our suggested modification of the Ab-Dom procedure. It is thus tempting to argue that these regression procedures are immune to problems caused by population stratification, using the claim that it is sufficient to overcome stratification problems by using only w in the testing procedure. This conclusion is, however, not necessarily correct.We illustrate this by considering an extreme case where the parental mating type Aa×aa occurs only in one stratum in the population, parental mating type Aa×Aa occurs only in another stratum, and parental mating type Aa×AA occurs only in a third stratum in the population. Suppose also that for reasons not connected with the marker locus, the null hypothesis mean phenotypes in the three strata are μ\n1 in the first stratum, μ\n2 in the second and μ\n3 in the third. Then of the models considered above, the Al-Lin model most closely reflects this situation. Suppose finally that for reasons having nothing to do with the marker locus, the three means—μ\n1, μ\n2,and μ\n3—are not all equal.Suppose that despite this, the investigator uses the Ab-Wthn test, which in effect assumes equality of μ\n1, μ\n2, and μ\n3. The regression sum of squares (used in the numerator of the F ratio for this test) isand the residual sum of squares (used in the numerator of the F ratio) iswhere b\n1 is the standard regression estimate of β\n1. If μ\n1 = μ\n2 = μ\n3, then under the null hypothesis, these sums of squares have expected values σ\n2 and (n−2)σ\n2 respectively, the corresponding mean squares both have expected values σ\n2, and (assuming a normal distribution for the phenotype) the F statistic has the F distribution. If however μ\n1, μ\n2, and μ\n3 are not all equal, the null hypothesis mean values of these two mean squares areandrespectively. In these expressions all sums are taken over the n trios,  is the value of μ (either μ\n1, μ\n2, or μ\n3) appropriate for trio i, and  is the average of the  values. The term in square brackets in the first expression is clearly non-negative, and the term in square brackets in the second expression can be shown, via the Cauchy-Schwartz inequality, also to be non-negative. Thus the F statistic does not now have the F distribution when the null hypothesis (β\n1 = 0) is true, so that the type I error of the Ab-Wthn procedure will now not be at the assumed value. Simulations also show that it generally exceeds the assumed value. In this sense, and in this example, the Ab-Wthn procedure is not immune to population stratification. A similar observation was also reported by Yu et al. [9]. This is of course an extreme example, which might seldom arise in practice. It nevertheless shows that caution is needed in assuming that a test based on the wi values only is automatically immune to stratification problems for quantitative traits. It also indicates that the practitioner should make an assessment of which regression model most closely reflects the situation from which the data were obtained and use the testing procedure for that model.Power considerationsGiven the various quantitative TDT procedures, it is important to address the power comparisons between them. First it has to be noted that the power comparison between any two tests is only meaningful if both are “anchored” so as to have the same type I error. In the situation described in the previous paragraph, a power comparison between the Ab-Wthn and the Al-Lin test is not meaningful. (The Al-Lin test is valid in this situation, and has the assumed type I error.)Suppose on the other hand that there is no population stratification associated with parental mating type and the Ab-Wthn test can be taken as appropriate. Then both the Ab-Wthn and Al-Lin procedures are valid tests, and the F ratios in the two tests both have the F distribution under the null hypothesis. The Al-Lin test will lose a small amount of power because of an unnecessary decrease in the residual number of degrees of freedom. It follows from all the above that no uniform statement about power can be made, and that the investigator has to use his/her judgment about the most appropriate test to use.ConclusionsThese notes indicate that there are several matters that the investigator should keep in mind in his/her data analysis. First, as noted above, all the procedures described here test for changes in the phenotype value Y as a function of W (or equivalently changes in W as a function of Y). This implies that these procedures are best suited either to a random sample of data or to data only comprising both low and high values of the phenotype under discussion. If the data relate only to “extremely low” or to “extremely high” values of Y, the qualitative TDT procedure is perhaps more appropriate. Second, if the investigator suspects population stratification associated with mating types, careful consideration should be given to the test that is to be used. Third, the regression-based methods are more susceptible to departure from the normality assumption, but the Rabinowitz and Monks-Kaplan procedures are not. We suggest that users be cautious when interpreting results from different tests, especially when the distribution of the trait is non-normal.In this brief review there are many topics that we have not covered. On the practical side, we have purposely not tried to recommend particular tests for specific kinds of data. This was not our goal, and in any case, would require considering a very large number of possible situations. Similarly, we have not discussed approaches to handle missing genotypes, although there are standard ways to do this [10]–[12]. On the theoretical side, we have not discussed the statistical theory behind the procedures described above. “Optimal” procedures often use score statistics, but the choice of the appropriate statistic relies on a choice of model that is felt best to describe the data. Next, we have considered only informative mating types, whereas some procedures use data from uninformative mating types, which may cause inflation of type I error rate if the phenotype distributions are different for different mating types. These and other theoretical questions will be taken up elsewhere.\n\nREFERENCES:\n1. SpielmanRSMcGinnisREEwensWJ\n1993\nTransmission test for linkage disequilibrium: the insulin gene region and insulin-dependent diabetes mellitus (IDDM).\nAm J Hum Genet\n52\n506\n516\n8447318\n2. RabinowitzD\n1997\nA transmission disequilibrium test for quantitative trait loci.\nHum Hered\n47\n342\n350\n9391826\n3. MonksSAKaplanNL\n2000\nRemoving the sampling restrictions from family-based tests of association for a quantitative-trait locus.\nAm J Hum Genet\n66\n576\n592\n10677318\n4. AbecasisGRCardonLRCooksonWO\n2000\nA general test of association for quantitative traits in nuclear families.\nAm J Hum Genet\n66\n279\n292\n10631157\n5. AllisonDB\n1997\nTransmission-disequilibrium tests for quantitative traits.\nAm J Hum Genet\n60\n676\n690\n9042929\n6. LangeCDeMeoDLLairdNM\n2002\nPower and design considerations for a general class of family-based association tests: quantitative traits.\nAm J Hum Genet\n71\n1330\n1441\n12454799\n7. LairdNMHorvathSXuX\n2000\nImplementing a unified approach to family based tests of association.\nGenet Epidemiol\n19\nSupplement 1\nS36\nS42\n11055368\n8. MartinERKaplanNLWeirBS\n1997\nTests for linkage and association in nuclear families.\nAm J Hum Genet\n61\n439\n448\n9311750\n9. YuJPressoirGBriggsWHVroh BiIYamasakiM\n2006\nA unified mixed-model method for association mapping that accounts for multiple levels of relatedness.\nNat Genet\n38\n203\n208\n16380716\n10. BurdickJTChenWMAbecasisGRCheungVG\n2006\nIn silico method for inferring genotypes in pedigrees.\nNat Genet\n38\n1002\n1004\n16921375\n11. MarchiniJHowieBMyersSMcVeanGDonnellyP\n2007\nA new multipoint method for genome-wide association studies by imputation of genotypes.\nNat Genet\n39\n906\n913\n17572673\n12. LiYDingJAbecasisGR\n2007\nMarkov model for rapid haplotyping and genotype imputation in genome wide studies.\nAm J Hum Genet\nS79\n2290"
+}

batch_8/PMC2529266.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2529266",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2529266\nAUTHORS: Vidushi S Patel, Steven JB Cooper, Janine E Deakin, Bob Fulton, Tina Graves, Wesley C Warren, Richard K Wilson, Jennifer AM Graves\n\nABSTRACT:\nBackgroundVertebrate alpha (α)- and beta (β)-globin gene families exemplify the way in which genomes evolve to produce functional complexity. From tandem duplication of a single globin locus, the α- and β-globin clusters expanded, and then were separated onto different chromosomes. The previous finding of a fossil β-globin gene (ω) in the marsupial α-cluster, however, suggested that duplication of the α-β cluster onto two chromosomes, followed by lineage-specific gene loss and duplication, produced paralogous α- and β-globin clusters in birds and mammals. Here we analyse genomic data from an egg-laying monotreme mammal, the platypus (Ornithorhynchus anatinus), to explore haemoglobin evolution at the stem of the mammalian radiation.ResultsThe platypus α-globin cluster (chromosome 21) contains embryonic and adult α- globin genes, a β-like ω-globin gene, and the GBY globin gene with homology to cytoglobin, arranged as 5'-ζ-ζ'-αD-α3-α2-α1-ω-GBY-3'. The platypus β-globin cluster (chromosome 2) contains single embryonic and adult globin genes arranged as 5'-ε-β-3'. Surprisingly, all of these globin genes were expressed in some adult tissues. Comparison of flanking sequences revealed that all jawed vertebrate α-globin clusters are flanked by MPG-C16orf35 and LUC7L, whereas all bird and mammal β-globin clusters are embedded in olfactory genes. Thus, the mammalian α- and β-globin clusters are orthologous to the bird α- and β-globin clusters respectively.ConclusionWe propose that α- and β-globin clusters evolved from an ancient MPG-C16orf35-α-β-GBY-LUC7L arrangement 410 million years ago. A copy of the original β (represented by ω in marsupials and monotremes) was inserted into an array of olfactory genes before the amniote radiation (>315 million years ago), then duplicated and diverged to form orthologous clusters of β-globin genes with different expression profiles in different lineages.\n\nBODY:\nBackgroundThe evolution of the vertebrate globin superfamily has been extensively studied for many decades by comparing the structure and function of members of the gene families. These are principally haemoglobin, myoglobin, cytoglobin and neuroglobin and, more recently, globin X (in fish and amphibians [1]) and globin Y (specific to amphibians [2]).Haemoglobin genes (alpha- and beta-globin) are of particular interest because of their critical role in oxygen transportation from the respiratory surfaces to the inner organs, and because of the dire effects of mutations in human globin genes that cause haemoglobinopathies [3]. The genes contained in the alpha (α)- and beta (β)-globin clusters are expressed at different stages of development and in different tissues. Together, gene products from both clusters form the functional tetrameric haemoglobin molecules needed to fulfil oxygen requirements.The evolutionary history of α- and β-globin genes can be traced back to the common ancestors of fish, amphibians and amniotes (reptiles, birds and mammals), by comparing gene structure and composition of α- and β-globin clusters across vertebrates. In the amphibians Xenopus laevis and X. tropicalis, α- and β-globin genes are tightly juxtaposed as 5'-α-β-3' [2,4-6]. In the Antarctic notothenioid fish (Notothenia coriiceps, N. angustata, Trematomus hansoni, T. pennellii), there is also a single 5'-α-β-3' locus [7], although in pufferfish (Fugu rubripes) there are two globin clusters (one with α-globin genes and the other with both α- and β-globin genes), which are located on different chromosomes [8].In amniotes, α- and β-globin clusters are located on different chromosomes. It was proposed that the ancestral α- and β-globin genes were located together in the common ancestor of amniotes, as they are in fish and amphibians, but became separated, either by chromosome fission or translocation between α- and β-genes, or by chromosome/genome or in trans duplication and gene loss [5].Further duplications then occurred in amniote lineages. The ancestral α-globin gene is thought to have duplicated twice before the divergence of the bird-mammalian lineages, to produce progenitors of embryonic globin genes π/ζ, and adult αD and αA, all of which are present in birds (for example, the chicken Gallus gallus) [9-11] and mammals [12,13]. The order and timing of these duplications is still debated, as is their origin: for instance, αD may have evolved by duplication either of adult αA (see [12]), or of an embryonic α-like gene [14]. After the avian and mammalian lineages diverged, there were further tandem duplications of the π/ζ and αA lineages to produce more complex marsupial and eutherian ('placental') mammalian α-globin clusters, 5'-ζ-ψζ'-αD-ψα3-α2-α1-θ-3' (see [12,15-18]). The timing of these duplication events is also uncertain, because we do not know whether these seven α-like globin genes all existed at the stem of the mammalian radiation.As for many other gene families [19], comparisons of globin genes between distantly related mammals have provided unique insight into the evolution and function of the mammalian globins. Marsupials diverged from eutherian mammals about 148 million years ago (MYA), and mammalian Subclass Theria that contains these groups diverged from monotremes (Subclass Prototheria) about 166 MYA [20], so comparisons between these major mammal groups provide depth for evolutionary comparisons. Monotremes retain many anatomical and developmental features shared with birds and reptiles. Their small genome, too, and disjunct chromosome size classes are reminiscent of reptile genomes, and the 10 sex chromosomes in a karyotype of 52 chromosomes is unique among mammals [21-23]. Their importance for comparative studies is now increasingly recognised after the sequencing of the genome of a monotreme, Ornithorhynchus anatinus (platypus), to a depth of six to eight times by the Washington University Genome Centre, St Louis [24].Indeed, studies of marsupial globins have clarified the timing of some of the duplications. The finding of single ε- (embryonic) and β-globin (adult) genes together in the marsupial β-globin cluster indicated that a two-gene cluster (ε-β) was present in the common therian ancestor [25-28]. Genes in the cluster were further duplicated to produce the ancestral eutherian β-globin cluster of 5'-ε-γ-η-δ-β-3' (see [29-32]), which then underwent further tandem duplication events. In contrast, the bird (G. gallus) β-like globin genes (ε-βH-βA-ρ) show very little homology to the mammalian β-like globin genes [33,34].The discovery of a β-like globin gene (ω -globin) adjacent (3') to the α-globin cluster in marsupials led to a re-interpretation of globin evolution in birds and mammals [35,36]. Comparative sequence and phylogenetic analysis suggested that the ω-globin gene was more closely related to bird β-like globin genes than to other mammalian β-like globin genes. The specific function of the ω-globin gene is not yet known, but it is expressed just before birth and in the early stages of pouch young development [37]. In addition, the ω-globin product binds to α-like globin chains to form functional haemoglobin, so it is likely to be involved in oxygen transportation [35-37].This finding of a remnant β-like globin gene (ω -globin) beside the α-globin cluster in marsupials [35,36] provided some support for the alternative hypothesis [5] that the α- and β-globin clusters in birds and mammals arose by in trans duplication of a chromosomal region, rather than simply by separation of the ancestral α-β globin cluster by chromosome fission or translocation. Wheeler et al. [35,36] proposed that before the divergence of birds and mammals (>315 MYA), the chromosome region bearing the ancestral α-β clusters duplicated to form two clusters (α1-β1 and α2-β2) on different chromosomes, and their contents diverged independently in mammals and birds by silencing of some genes within each cluster (Figure 1). To account for the apparent orthology of the marsupial ω-globin gene and bird β-like globin genes, Wheeler et al. [35,36] suggested that the α1 and β2 were silenced in the eutherian lineage, but β2 was retained in marsupials as the ω-globin. In contrast, α2 and β1 were silenced in the bird lineage (Figure 1). On this hypothesis, then, both the α clusters and the β clusters of birds and mammals are paralogous (that is, evolved independently from ancient duplicates in an amniote ancestor) rather than orthologous (that is, diverged from the same ancestral cluster in an amniote ancestor).Figure 1Current proposed model for the evolution of α- and β-globin clusters from paralogous clusters in different lineages. The unlinked α- and β-globin clusters in birds and mammals evolved from an ancient in trans duplication of the ancestral linked α-β cluster, followed by differential gene silencing (marked with X). This resulted in bird β-like globin genes (β2) orthologous to the marsupial ω-globin gene (β2 beside the α-globin cluster) but paralogous to mammalian β-like globin genes (β1). Adapted from Wheeler et al. [36].This paralogy hypothesis (which rests on the rather weak orthology between the chicken β and marsupial ω), as well as the dates and types of other duplications, could be further tested by studying globin genes of monotreme mammals, and using comparative data to infer the ancestral globin gene arrangement of a mammal ancestor 166 MYA. The availability of platypus genomic sequences now provides an efficient way to discover all of the globin genes and regulatory signals, and to understand their function and evolution. Studies of globin genes in monotremes are also interesting because the specialized features and lifestyle of these unique mammals may have given rise to special adaptations of globin genes to fulfil unusual oxygen requirements. These features include the need for oxygen by diffusion through the egg membrane to the embryo after birth and the physiological response to hypoxic conditions during hibernation, burrowing and diving [38-40].Little is known about monotreme α- and β-globin families. More than 30 years ago, studies of adult blood revealed a single adult α and β globin protein in the platypus [41,42] and echidna (Tachyglossus aculeatus [43,44]). Lee et al. [28] later isolated an adult β-globin gene in the echidna that encoded a polypeptide identical to the previously isolated echidna β-globin [44]. To date, there is no evidence of any monotreme embryonic ζ- or ε-globin genes.We used platypus genomic sequences from bacterial artificial chromosomes (BACs) to characterise the α- and β-globin gene families of the platypus and investigate their molecular evolution. In particular, we searched for embryonic and ω-globin genes and any novel globin genes that might fulfil the requirements for oxygen transport under hypoxic conditions. We investigated the genome context in order to infer the structure and origin of the ancestral α- and β-globin clusters at the stem of the mammalian radiation. Our results strongly support the hypothesis that the mammalian α- and β-globin clusters are orthologous to the avian α- and β-globin clusters, respectively, and that the β cluster evolved by transposition of a copy of the beta-like ω-globin gene in an amniote ancestor.ResultsIdentification of BAC clones containing the α- and β-globin clustersThe draft sequence assembly of platypus [24] is readily available on the University of California Santa Cruz (UCSC) Genome Browser [45]. However, currently the assembly is incomplete for the α- and β-globin clusters, as individual globin genes appear on different contigs. There are also sequences of the platypus BAC clones available in NCBI GenBank that are not yet annotated and assembled, nor is part of the platypus genome assembly. Two of these are Oa_Bb-2L7 [GenBank:AC195438] and Oa_Bb-131M24 [AC203513], which were identified from the Encyclopaedia of DNA Elements Project to contain parts of the α-globin cluster (see Methods). The BAC clone Oa_Bb-484F22 [GenBank: AC192436] containing the β-globin cluster was obtained by screening a male platypus BAC library (Clemson University Genomic Institute, USA) and was subsequently fully sequenced and assembled by the Washington University Genome Sequencing Centre (St Louis, USA). These sequences were therefore used in this study to characterise the whole α- and β-globin clusters in the platypus.Genes in these sequenced BAC clones were predicted by programs GENSCAN [46] and GenomeScan [47]. Many genes were predicted, which were then used for BLAST searches of nucleotide (BlastN) and amino acid (BlastP) databases to help identify them (data not shown). Phylogenetic analyses were also conducted for the platypus α- and β-like globin genes to further verify the identity of each gene (see below and also Figures 2, 3 and 4 below). With only one exception (platypus ε-globin, see below), the identities of all of the genes inferred by BLAST analyses were supported by phylogenetic analyses with high posterior probabilities and bootstrap support values.Figure 2Evolutionary relationships among vertebrate α-like globin genes using a 50% majority rule consensus phylogram from an analysis using Bayesian Inference. The tree was constructed using mixed models of evolution for each codon position (see methods) and estimated base frequencies in an unlinked analysis using MrBayes (v. 3.1.2). Numbers adjacent to branches refer to % posterior probabilities. GenBank accession numbers for sequences are: Virginian Opossum (Didelphis virginiana) ζ1, ζ2, α1, α2, θ [AC139599.2, AC148752.1]; Stripe-faced Dunnart (Sminthopsis macroura) αD, α2, θ [AC146781]; Brazilian Opossum (Monodelphis domestica) α [TI# 453585430]; Tammar wallaby (Macropus eugenii) θ [AY459590], α [AY459589]; ζ [AY789121], ζ' [AY789122]; Horse (Equus caballus) θ (ψ α) [Y00284], α1 [M17902], ζ [X07051]; pig (Sus scrofa) αD [AC145444]; cat (Felis catus) αD [AC130194]; cow (Bos taurus) αD [AC150547]; Goat (Capra hircus) α [J00043]; Human (Homo sapiens) α1 [V00491], θ [X06482], ζ [NM_005332]; mu/αD chain [AY698022]; Mouse (Mus muscularis) α1 [NM_008218], ζ [X62302]; Rabbit (Oryctolagus cuniculus)α [X04751]; Eastern Quoll (Dasyurus viverrinus) α [M14567]; Chicken (Gallus gallus) αA, π, αD [AF098919]; Duck (Cairina moschata) αD [X01831]; Pigeon (Columba livia) αD [AB001981]; Turtle (Geochelone nigra) αD [SEG# AB1165195]; Zebrafish (Danio rerio) α1 [NM_131257]; Salamander (Hynobius retardatus) larval α [AB034756]; Salamander (Pleurodeles waltlii) α [M13365]; Frog (Xenopus laevis) α I [X14259], larval (tadpole) α T5 [X02798]; Yellowtail (Seriola quinqueradiata) αA [AB034639]; Salmon (Salmo salar) α [X97289]; Southern Puffer (Sphoeroides nephelus) α2 [AY016023]; Platypus (Ornithorhynchus anatinus) ζ, ζ', αD, α3, α2, α1 [AC203513].Figure 3Evolutionary relationships among vertebrate β-like globin genes using a 50% majority rule consensus phylogram from an analysis using Bayesian inference. The tree was constructed using mixed models of evolution for each codon position (see methods) and estimated base frequencies in an unlinked analysis using MrBayes (v. 3.1.2). Numbers adjacent to branches refer to % posterior probabilities. GenBank accession numbers for sequences are: Fat-tailed Dunnart (Sminthopsis crassicaudata) β [Z69592], ε [Z48632], ω [AY014770]; Stripe-faced Dunnart (S. macroura) β, ε [AC148754]; Virginian Opossum (Didelphis virginiana) β [J03643], ε [J03642]; Brazilian Opossum (Monodelphis domestica) β [XM_001365299], ε [XM_001364448], ω [XM_001364828]; Tammar Wallaby (Macropus eugenii) β [AY450928], ε [AY450927], ω [AY014769]; African clawed frog (Xenopus laevis) larval β I [NM_001086273], larval βII [NM_001088028]; Western clawed frog (X. tropicalis) β [NM_203528], larval ε1 [NM_001016495]; Chicken (Gallus gallus) β (βA) [NM_205489], ε [NM_001004390], γ (βA) [NM_001031489]; Duck (Cairina moschata) β [J00926], ε [X15740]; Human (Homo sapiens) β [NM_000518], γ [BC130459], ε [NM_005330]; Mouse (Mus musculus) β (β1) [NM_008220], γ (β h0) [NW_001030869], ε (εy) [M26897]; Goat (Capra hirus)β (βA) [DQ350619], ε (εI) [X01912], γ [M15388]; Rabbit (Oryctolagus cuniculus) β, γ, ε [M18818]; Echidna (Tachyglossus aculeatus) β [L23800]; Pufferfish (Fugu rubripes) β [AY170464]; Zebrafish (Danio rerio) ε1 [NM_001103130]; Platypus β, ε [AC192436], ω [AC203513].Figure 4Evolutionary relationships among vertebrate β-like globin genes analysed by maximum parsimony (MP) trees of length 926 (one of eight trees). Third position in codons were excluded in the MP analyses, which were conducted using a heuristic search in PAUP* v.4.0b10 [65]. The tree is rooted using pufferfish β-globin. Numbers adjacent to branches represent % bootstrap values (>50%) from MP heuristic analyses of 1000 pseudoreplicates. Accession numbers for sequences are given in the caption of Figure 3.Predictions and characterisation of genes in the platypus α-globin clusterOne BAC (Oa_Bb-2L7) contained two embryonic α-like globin genes, and a second BAC (Oa_Bb-131M24) contained six α-like globin genes and a β-like globin gene (see Additional file 1). These two BACs were found to overlap by 10,066 base pairs (bp), resulting in a contig of 330,126 bp that contained the entire platypus α-globin cluster and flanking genes.The 330,126 bp α-globin contig was found to contain six α-like globin genes, a β-like globin gene, and a gene that bore little similarity to α- and β-like globin genes but some similarity to cytoglobins (Figure 5A). These six α-like globin genes have a three-exon/two-intron structure and conserved donor/acceptor splice sites (GT/AG) typical of all vertebrate α-like globin genes. They are separated from each other by 2 to 6 kilobase pairs (kb). Full details of the exon/intron lengths, location of the putative poly-A addition site (AATAAA) and the lengths of the coding domains with the predicted encoded polypeptide for each predicted gene are given in Table 1. Figure 5B shows the predictions for some of the well-characterised protein-binding sites in the 5' promoter region (about 200 bp 5' to the cap site of each gene). These include CACCC [48], CAAT [49], TATA [50], GATA 1 [51], EKLF (Erythroid Krüppel-like Factor; [52]) and have been experimentally shown to control the stage- and tissue-specific expression of α- and β-like globin genes in other mammals [50,53-55].Table 1Gene-structure of the predicted platypus α- and β-like globin genes and GBYLength/GenesExon 1 (bp)Intron 1 (bp)Exon 2 (bp)Intron 2 (bp)Exon 3 (bp)Position of Poly-ACDS (bp)Poly- peptide (aa)ζ95337205114129+119429142ζ195336205102129+133429142αD9214502051610129+77426141α1/α392405205151129+94426141α295720205155129+115429142ω92256223111129+69444147GBY9833642233053144+141465154ε92143223474129+96444147β92153223438129+71444147For each predicted gene, the length of the exons and introns, position of poly-A addition site (AATAAA) from the stop codon, and the length of their putative coding domain (CDS) and encoded polypeptide are shown. All genes contained consensus splice sites (GT/AG) in both introns.Figure 5Gene structure of the platypus α- and β-globin clusters and flanking loci, and comparisons of their promoter regions with other mammals. (A) The platypus α-globin cluster contains six α-like globin genes (red), a β-like (ω) globin gene (blue) and a distantly related globin gene, GBY (green), which are flanked by IL9RP3-POLR3K-C16orf33-C16orf8-MPG-C16orf35 on the 5' end and LUC7L-ITFG3-RGS11-ARHGDIG-PDIA2-AXIN1 on the 3' end (black). The platypus β-globin cluster contains only two genes, ε and β (blue), which are flanked on both sides by ORG genes (black). (B) Relative positions of the putative transcription factor binding sites in the 200 bp promoter region located upstream of the predicted platypus, marsupial (Didelphis virginiana ζ and ψζ', and Sminthopsis macroura αD, ψα3, α2, α1, ω, ε and β) and human α- and β-like globin genes. For the platypus GBY no data was available from other species, including Xenopus tropicalis, for comparisons.Two genes at the 5' end of the α-globin cluster were both identified as ζ-like (referred to here as ζ and ζ') and predicted to encode polypeptides of 142 amino acids (aa), which are typical of known functional mammalian α-like globin genes. The amino acid sequence alignment of ζ and ζ' shows 95% identity. In the promoter region of both genes, CACCC and CAAT consensus boxes are conserved at similar positions, and in comparable order to that of human ζ and ζ' (Figure 5B).Adjoining the two ζ-like globin genes, four other α-like globin genes were identified. One was an orthologue of bird and reptilian αD, and the other three were orthologues of adult α genes (here called α3, α2 and α1). The long and uninterrupted open reading frame (ORF) of αD strongly suggests that it encodes a functional polypeptide of 141 aa, typical of known functional αD globin genes. The platypus αD globin gene contains introns of 1450 bp (intron 1) and 1610 bp (intron 2) that are very large compared with those of other α-like globins, which are usually less than 1000 bp.Analyses of the platypus adult α-like globin genes reveal three adult (α3, α2 and α1) globin genes in the α-globin cluster. The sequence of α3 (the most 5' gene, adjacent to αD) was found to be almost identical to α1 (the most 3' gene) in their exon and intron regions, as well as in flanking regions of about 130 bp on both sides. The coding region was 100% identical, and just two sites in intron 1 were found to be different between the two genes. In order to confirm that identification of these two identical genes was not due to an error in the assembly of the original sequence data, the boundaries of the region containing the homology between α1 and α3 was further analysed by a BLAST search of the platypus whole-genome shotgun (WGS) database (data not shown). Two contigs were identified with homology to α1 and α3; these had identical sequences on one side of the boundary but different sequences on the other, confirming the presence of two separate genes. Further confirmation was obtained by performing a Southern blot on the α-globin-containing BACs, digested with an enzyme (EcoRV) that does not cut within the α1, α2and α3 (data not shown). Probing with α1/α3 revealed two bright bands, corresponding to α1 and α3, and one fainter band between them, corresponding to α2. Probing with α2 produced the same three bands, but in this case the middle one was brighter, corresponding to α2, and the outer bands were fainter, corresponding to α1 and α3. These analyses confirmed the existence of separate genes α1 and α3 in the platypus α-globin cluster. The α2 gene, located between α1 and α3, was distinct from both genes in the coding sequence (with 83% homology), in intron lengths (intron 1: 405 bp in α1/α3 and 720 bp in α2; intron 2: 151 bp in α1/α3 and 155 bp in α2) and in the promoter region (Figure 5B).The amino acid sequence encoded by α1 and α3 was identical to the platypus adult α-chain previously identified by Whittaker and Thompson [41], implying that at least one of these genes is expressed in the adult platypus. The coding domain of α1 and α3 is shorter (426 bp) than that of α2 (429 bp), because it lacks the first three nucleotides of exon 1. The ORF of α2 gives a strong indication that it is translated into a functional polypeptide of 142 aa, typical of known functional mammalian α-like globin genes.On the 3' side of the six α-like globin genes, a β-like globin gene was predicted, which was identified as the orthologue of the marsupial ω-globin gene. This platypus ω-globin gene has a typical three-exon/two-intron structure, conserved donor/acceptor splice sites, and encodes a polypeptide of 146 aa, typical of all vertebrate β-like globin genes (Table 1). The promoter region located 5' of the ω-globin initiation codon contains conserved sites for CAAT-EKLF-CACCC in an order identical to that of marsupial ω-globin gene.Unexpectedly, GenomeScan predicted a gene based on the protein similarities with the α- and β-polypeptide chains, approximately 1.5 kb 3' of the ω-globin gene. Like other α- and β-globins, this gene also has a three-exon/two-intron structure and conserved donor/acceptor splice sites (Table 1). The lengths of its exons 1, 2, and 3 are 98, 223 and 144 bp, respectively, compared with 92, 223 and 129 bp in other β-like globin genes. However, it has much larger introns of 3364 bp (intron 1) and 3053 bp (intron 2). The long and uninterrupted ORF of this gene can be translated into a polypeptide of 154 aa, which is atypical of any known α- or β-like globin genes. A BLAST search of the amino acid sequence of this gene obtained the best hit with Globin Y (gby) of the amphibian X. laevis (identity score of 39%), and weaker identity scores with Cytoglobins (cygb) of other species, such as the fish Danio rerio (27%), X. tropicalis (26%), chicken (28%) and human (25%) at the protein level. We designated this gene 'GBY' based on similarities with X. laevis gby, and its similar position adjoining the globin cluster [2]. The predicted polypeptide of platypus GBY (154 aa) was shorter than X. laevis gby (156 aa), and quite different from X. laevis cygb (179 aa), D. rerio cygb1 (174 aa) and cygb2 (179 aa), and human CYGB (190 aa). Using the Expressed Sequence Tag (EST) database, a BLAST search of the platypus GBY also obtained an identity score of 38% with X. tropicalis gby that was expressed in both tadpoles and adults, but produced no significant matches with any other mammalian genes. The present work was the first opportunity to analyse the promoter region of any GBY gene (Figure 5B).Predictions and characterisation of genes in the platypus β-globin clusterIn the platypus, only two β-like globin genes were predicted within the 129,521 bp BAC clone (Oa_Bb-484F22) by GENSCAN and GenomeScan (see Additional file 1). When the predicted amino acid sequences were subjected to BLAST search, the 5' gene had best hits with mammalian embryonic ε-globin genes. Although the phylogenetic analyses using Bayesian inference (BI; see below) indicated that this gene was more closely related to the platypus and echidna adult β-globin genes than to therian ε-globin genes, the position of this gene on the 5' end of the β-globin cluster and expression data (see below) supports its orthology with mammalian embryonic ε-globin genes, and is henceforth referred to as ε. The 3' gene encoded a protein identical to the previously identified platypus adult β-chain [42], and is henceforth referred to as β.Both genes encode polypeptides of 146 aa, typical of known functional mammalian β-like globin genes. The promoter region of the platypus β has conserved sites of CACCC and CAAT in all three extant of mammals. However, the promoter region of the platypus ε appears to be quite different from other mammalian ε-globin genes and even from the platypus β (Figure 5B). The promoter of platypus ε contains only one predicted motif (CAAT), whereas the promoters of other mammalian ε, β and the platypus β contain many predicted motifs.Expression studies of the platypus α- and β-like globin genesTranscription studies were performed to gain insight into the expression and function of all of the predicted platypus globin genes. Adult liver, kidney, spleen, testis, lung and brain were obtained for this project: no embryonic samples were available (or are ever likely to be available) for this vulnerable and iconic species. Observation of the expression of any of the predicted genes would constitute a good indication that the gene is transcriptionally active and functional.Reverse-transcriptase polymerase chain reaction (RT-PCR) of all predicted platypus genes showed that they are all expressed in at least some of these adult platypus tissues (Figure 6). Platypus genes α1/α3, α2 and β, whose orthologues are usually expressed in the bone marrow of an adult human, were expressed in almost all platypus tissues tested, suggesting a broader expression of these genes in the monotreme lineage. Surprisingly, the genes ζ, ζ' and ε, whose therian orthologues are expressed only at embryonic stages of development, were expressed in adult spleen and testis, but not in the other tissues of adult platypus. This suggests that persistent expression of these genes in some adult tissues was selected for in the platypus, perhaps in response to its aquatic lifestyle and the hypoxic conditions of a confined burrow. Also, the expression pattern of platypus ε is similar to embryonic α-like ζ and ζ' but different from that of adult globin genes (α1/α3, α2 and β). The ω and αD globin genes, whose functions are unknown, were also expressed mainly in the spleen. GBY was expressed in all adult platypus tissues, most strongly in testis.Figure 6Expression of all predicted α- and β-like globin genes including GBY in an adult platypus. For each of the platypus predicted genes, expression was investigated by reverse transcriptase polymerase chain reaction in adult liver, kidney, spleen, testis, brain and lung. Primers for each gene were designed between two exons so that it would result in a product distinguishable from genomic contamination of cDNA. The negative control (last lane) contained no cDNA. All genes were expressed in one or more tissues, indicating that they are transcriptionally active and might be functional.Phylogenetic analysesPhylogenetic analyses of the α-like globin genes using BI and maximum parsimony (MP) produced several noteworthy results. The platypus adult α globin genes (α1/α3 and α2) grouped closely together to the exclusion of eutherian and marsupial α- and θ-globin genes for all analyses, although posterior probability (69%) and bootstrap support (66%) for this arrangement were relatively weak (Figure 2). This finding suggests that the duplication leading to the marsupial and eutherian θ-globin lineage occurred after the divergence of the monotreme and therian lineages. This is consistent with the absence of a θ-globin gene from the region between platypus α1- and ω-globin, its expected location based on its position in marsupial α-globin clusters [12,56].Both platypus ζ-globin genes grouped closely together and formed a sister group relationship with chicken π, supported by a high posterior probability of 97% (Figure 2). A sister group relationship was also found in MP trees for analyses of the entire platypus coding region (bootstrap support <50%), and when third positions in the codon were excluded, was supported by 73% bootstrap pseudoreplicates (data not shown). This differs from the expectation that platypus ζ-globin genes would group with other mammalian ζ-globin genes to the exclusion of chicken π, suggesting that other factors (for example, purifying selection) operated to maintain a similar sequence in birds and monotremes.There is still considerable uncertainty in the phylogenetic position of the αD-globin clade. It has recently been proposed that the αD globin lineage resulted from duplication of the embryonic α-globin lineage, with phylogenetic analyses supporting a sister lineage relationship of these lineages to the exclusion of the adult α-globin lineage [14]. However, this arrangement was not supported in BI analyses of the data set used here, and the position of the αD lineage was different in the different analyses. Analyses using BI (Figure 2) supported the sister lineage relationship of the αD and adult α-globin lineages (as proposed by Cooper et al. [12]), with 87% posterior probability support. In contrast, all MP analyses supported the sister lineage status of αD and embryonic α-globin genes, indicating an uncertainty in the phylogenetic position of the αD-globin clade.Phylogenetic analyses of the β-globin genes provided results similar to recently reported phylogenetic analyses [35,36], with one notable exception. The BI analyses of coding sequence data (Figure 3) provided strong support (99% posterior probability) for the sister relationship of bird and mammalian β-like globin genes, contradicting previously published phylogenies of mammalian β-globin genes showing a sister relationship of marsupial ω-globin and bird β-like globin genes [35,36]. MP analyses (Figure 4), excluding third position in the codon, gave a similar tree arrangement, albeit with very low bootstrap support (<50%). In marked contrast to the BI analyses of DNA sequence data, BI protein analyses (data not shown) supported the sister relationship of bird β-like globin and mammal ω-globin lineages with a high posterior probability (99%).Lastly, phylogenetic analyses using BI indicated that the platypus ε gene was more closely related to the platypus and echidna adult β-globin genes than to therian ε-globin genes, suggesting it may not be orthologous to marsupial and eutherian ε-globin (Figure 3). BI analyses of β-globin protein data and MP analyses of the coding sequence data, with third codon positions excluded, grouped the gene as an ancestral lineage to eutherian and monotreme adult β-globin genes (see Figure 4). This ancestral position suggests that the lineage evolved following duplication of an ancestral β-globin gene prior to the divergence of monotremes and therians.Location of the α- and β-globin clusters in the platypusThe location of the verified BAC clones containing the α- (Oa_Bb-2L7) and β-globin (Oa_Bb-484F22) clusters in the platypus was determined by fluorescence in situ hybridisation (FISH) (Figure 7). The β-globin cluster localised to one of the largest autosomes, giving unambiguous signals on the long arm of chromosome 2 (2q5.1). The α-globin cluster localised to the smallest autosome, 21, whose two arms are not distinguishable by size or DAPI banding pattern [21]. This is the first gene that has been localised on the platypus chromosome 21.Figure 7Chromosomal location of the platypus α- and β-globin clusters. Two-colour fluorescence in situ hybridisation showing the location of the α-globin cluster on chromosome 21 (green) and the β-globin cluster on chromosome 2q5.1 (red). The chromosomes are counterstained with DAPI (blue).Loci flanking the α- and β-globin clusters in the platypus and other vertebratesTo explore the genome context of the α- and β-globin clusters in the platypus and other vertebrates, the platypus BAC sequences and the genomes of other sequenced species were searched for loci residing beside the α- and β-globin clusters.As well as globin genes, GENSCAN predicted within the platypus α-globin 330,126 bp contig many genes that flank the platypus α-globin cluster (Figure 5A), which were identified by BLAST analyses. These include IL9RP3-POLR3K-C16orf33-C16orf8-MPG-C16orf35 upstream (5') of the α-globin cluster, and, LUC7L-ITFG3-RGS11-ARHGDIG-PDIA2-AXIN1 downstream (3') of the α-globin cluster (Figure 5A).To compare the α-globin flanking loci of the platypus and other vertebrates, the genes closest to the α-globin cluster, MPG, C16orf35 and LUC7L were searched for in the human, opossum (Monodelphis domestica), chicken, frog (X. tropicalis) and zebrafish (D. rerio) genomes that were accessible from Ensembl [57]. Figure 8A shows that the locations of MPG, C16orf35 and LUC7L are conserved adjacent to the α-globin cluster of birds and mammals, and in the same position adjacent to the α-β cluster of amphibians, and all but LUC7L were also present in fish. These results are consistent with the previous analyses of Flint et al. [58] and Hughes et al. [59]. Thus the flanking loci analyses reveal that the genome context of the platypus α-globin cluster is the same as the α-globin clusters in therian mammals and birds, and this is the same as for the α-β cluster of fish and frogs.Figure 8Loci flanking vertebrate α-globin (A) and β-globin (B) clusters. The relative locations of flanking loci (A) MPG, C16orf35, LUC7L and GBY and (B) RRM1, CCKBR, ILK and ORG genes were searched for beside the α-β globin cluster in zebrafish (Danio rerio) and frog (Xenopus tropicalis), and beside the separate α-globin and β-globin clusters in chicken (Gallus gallus), opossum (Monodelphis domestica) and human (Homo sapiens) from Ensembl [57]. The pufferfish (Fugu rubripes) flanking loci shown here were adapted from Gillemans et al. [8]. For the platypus, the α-globin flanking loci were characterised in this study, and ORG genes surrounding the platypus β-globin cluster were discovered: however, the BAC clone (484F22) was too small to cover the region containing the loci RRM1, CCKBR and ILK. In X. tropicalis LUC7L was found on another scaffold (466 from Ensembl) but sequence analyses by Fuchs et al. [2] suggested that LUC7L resides 3' to the frog α-β-GBY cluster. The flanking loci as well as the α- and β-globin clusters are differentiated by colour.GENSCAN also predicted numerous genes other than globin genes in the platypus β-globin BAC (484F22). These were identified by a BLAST search as members of the olfactory receptor gene (ORG) family that are responsible for odour detection. Three conserved ORG members were identified at the 5' end of the platypus β-globin cluster and one conserved ORG member at the 3' end (Figure 5A).To compare β-globin flanking loci, ORG genes, as well as other genes that are closest to the β-globin cluster in other species, RRM1, CCKBR and ILK were searched for in the human, opossum, chicken and zebrafish genomes that were accessible from Ensembl [57]. Data from frog (X. tropicalis) was not useful since all of these loci lie on different contigs or scaffolds due to assembly problems. The locations of multiple ORG genes, RRM1, CCKBR and ILK were found to be conserved adjacent to β-globin cluster of birds and mammals [60,61], but not for the α-β cluster of fish and frogs, nor beside the second α-β cluster of zebrafish and pufferfish (Figure 8B). Thus the genome context of the platypus β-globin cluster is the same as in therian mammals and birds, but this is different from the α-β cluster of fish and frogs.DiscussionThe phylogenetic position of monotremes makes comparisons with platypus of special value for exploring the organization, function and evolution of mammalian genes and genomes. The availability of platypus genome sequence data now makes many such studies possible, and have been used here to characterise the platypus α- and β-globin gene clusters and explore their evolutionary history.The platypus α-globin gene clusterThe platypus α-globin cluster contains at least eight genes within more than 40 kb, including six α-like globin genes (including the identical α1 and α3), one β-like globin gene (ω-globin) and a gene belonging to another member of the globin super-family (GBY) arranged in the order 5'-ζ-ζ'-αD-α3-α2-α1-ω-GBY-3' (Figure 5A). The cluster maps to chromosome 21, the smallest autosome in platypus. All eight genes are likely to be functional since their expression was detected in tissues of an adult platypus.The platypus α-globin cluster is almost identical to the arrangement of α-like globin genes in the ancestral therian cluster reported by Cooper et al. [12]. The one exception is the absence of a θ-globin gene from the platypus cluster. Phylogenetic analyses support the basal position of the monotreme adult α-globin lineage relative to marsupial and eutherian α- and θ-globin lineages, implying that the duplication of an adult α-globin to produce θ-globin occurred in the therian lineage after its divergence from the monotreme lineage (Figure 9B). However, although the numbers and arrangements of genes is so similar in platypus and therians, the presence of three adult α-globin genes and two embryonic ζ-globin genes in their common ancestor was not supported by phylogenetic analyses, which showed independent groupings of the three adult and embryonic genes within each separate mammalian lineage (Figure 2 and see Cooper et al. [12]). This result can be interpreted literally as resulting from independent duplications in each mammalian lineage to produce three adult and two embryonic genes in each. However, this seems unlikely to explain the convergence in gene number of the α-globin cluster in these distantly related mammalian lineages. We suggest that a more parsimonious explanation is that the common ancestor of monotremes and therians contained three adult α-globin genes and two ζ-globin genes, which were homogenised by ongoing gene conversion events, leading to the gene tree that does not match the duplication history of the individual genes. The close similarity of the platypus α3 and α1 loci suggests a very recent gene conversion event that homogenised their sequences. Therefore, we propose that the platypus α-globin cluster of eight genes (ζ-ζ'-αD-α3-α2-α1-ω-GBY) represents the ancestral mammalian α-globin cluster arrangement (Figure 9B), in which all genes were transcriptionally active.Figure 9Proposed model for the evolution of the α- and β-globin clusters in vertebrate lineages. (A) A region containing MPG-C16orf35-α-β-GBY-LUC7L represented the ancient α-β globin cluster of jawed vertebrates (>450 MYA), which is seen in the amphibian lineage. This region further duplicated and underwent some gene silencing in teleost fish. In an amniote ancestor of reptiles, birds and mammals (>315 MYA), a copy of an ancestral β-globin gene from this region was inserted into a different chromosome within a region replete with multiple copies of ORG genes. The original amniote β-globin gene survives as the ω-globin gene (β1) in the α-globin cluster of marsupials and monotremes, whereas the transposed β-globin gene (β2) duplicated several times to form different clusters in the different lineages. (B) Tandem duplications of the ancestral amniote α-globin gene produced a three-gene (π-αD-αA) cluster in the avian lineage. In the mammalian lineage, further duplications gave rise to a six-gene (ζ-ζ'-αD-α3-α2-α1) cluster with ongoing gene conversion events homogenising the embryonic and adult genes. In monotremes, the ancestral ω (β1) and GBY are retained. After the divergence of monotreme and therian mammals, there was an additional duplication of α2 to form θ, giving rise to the seven-gene cluster (ζ-ζ'-αD-α3-α2-α1-θ) in marsupials and eutherians. Marsupials also retain the ancestral ω but may have lost GBY gene; eutherians retain no identifiable remnant of either gene. Furthermore, the ancestral transposed β2-globin gene duplicated independently in birds and mammals. Before the mammalian radiation, we propose that the ancestral β2 gene duplicated to form a two-gene β-globin cluster (ε-β) as seen in monotremes and marsupials, except that ongoing gene conversion events homogenised platypus ε to group with monotreme β genes. After the divergence of marsupial and eutherian mammals, there were further tandem duplications of these two genes to produce complex β-globin cluster (ε-γ-η-δ-β) in eutherians.Importantly, the platypus α-globin cluster contains a copy of the β-like ω-globin gene, also found in the marsupial α-globin cluster, but absent in humans, supporting the hypothesis that ω-globin was present in the common ancestor of all mammals. Phylogenetic analyses also confirm the ancient ancestry of the ω-globin gene, as concluded by Wheeler et al. [35,36]. Among adult platypus tissues this gene was expressed only in the spleen. In marsupials, expression of the ω-globin gene was detected just prior to birth and during early pouch young development [37], although the site of expression was not studied, and there was no evidence of adult expression in blood cells.Discovery of a mammalian GBY globin gene adjoining the α-globin clusterWe discovered a globin gene GBY in the platypus that is adjacent (3') to ω in the α-globin cluster. It has a typical three-exon/two-intron structure like other α/β-globin genes, contains an ORF encoding a polypeptide chain of 154 aa, and is expressed in almost all adult tissues, most strongly in testis. The amino acid sequence is unrelated to any of the other globin genes in the cluster, so it is unlikely to be derived by duplication of α- or ω-globin within the monotreme lineage. Rather, it shows sequence similarity to gby of X. tropicalis and X. laevis, a gene thought to be related to cytoglobins [2].Little is known of the function of amphibian gby, or its relationship with other globins. Fuchs et al. [2] reported that amphibian gby encodes a bona fide globin of 156 aa, having all of the sequence features of a functional respiratory protein. gby was expressed in all adult tissues tested in X. laevis, most strongly in ovary, kidney and eye, and was present in 20 expressed sequence tag clones from different stages of X. laevis and X. tropicalis embryonic and adult development [2], suggesting that it is expressed in embryonic as well as adult stages. Phylogenetic analysis of all vertebrate globins [2] showed that the gby lineage diverged at the base of two separate clades, one comprising all vertebrate cytoglobins, myoglobins, agnathan globins and bird globin E, and the other comprising the haemoglobin α- and β-chains.The position of platypus GBY adjacent to the α-globin cluster and flanked by LUC7L mirrors its position in X. tropicalis between the main α-β cluster and LUC7L [2]. Another common feature of both was strong expression in gonads (ovary in X. laevis [2] and testis in platypus), so GBY has sex-related expression in both lineages. Thus GBY is not specific to amphibians, as was thought, but was a component of the cluster in an ancient tetrapod, and has been lost, or has diverged beyond recognition, in birds and therian mammals.The platypus β-globin gene clusterCharacterisation of the platypus β-globin cluster revealed two β-like globin genes over about 13.2 kb that are arranged in the same order as marsupials, 5'-ε-β-3' (Figure 5A). This cluster is located on platypus chromosome 2q5.1. Both genes appear to be transcriptionally active and are likely to be functional.At the time of revising this paper, an independent paper on monotreme β-like globin genes was published by Opazo et al. [62] in which they reported the presence of ω, εP and βP in the platypus. Largely on the basis of phylogenetic analyses of flanking and coding sequence data, they proposed that platypus εP and βP were not 1:1 orthologues of therian ε and β, respectively, and arose by independent duplication of an ancestral β-globin gene in the monotreme lineage, with a separate duplication event, just prior to the divergence of therians, producing the progenitors of ε and β of therians. This hypothesis was strongly supported by our BI phylogenetic (Figure 3) analyses, but not by MP analyses of coding sequence data, with third codon sites excluded (Figure 4), or BI analyses of protein sequence data (not shown). These contradictory analyses highlight the difficulty in resolving deep relationships among globin genes, particularly when the time periods between duplication and speciation events are relatively small, the phylogenetic signal at third codon positions is potentially saturated, and non-synonymous sites may be subjected to purifying or positive selection. Despite a very high posterior probability (100%) for the grouping of platypus ε with monotreme β, this value is a Bayesian probability and depends on the model adequately representing the evolution of the gene. Furthermore, although it was reported [62] that the 5' flanking sequences of platypus ε and β were similar, we found no evidence for similarity of the promoter signals of these two genes (Figure 5B).We consider that a more parsimonious explanation is that the platypus ε is orthologous to the marsupial and eutherian embryonic β-like globin lineages (ε and γ), and arose by duplication of an ancestral β-globin gene prior to the mammalian radiation (166 MYA; Figure 9B). The sequence of platypus ε may have been homogenised by some gene conversion events, leading it to group with other monotreme adult β-like globin genes. In addition to the MP analyses reported above, this explanation is further supported by the conserved position of ε to the 5' side of the adult β-globin gene in the platypus cluster, which is similar to that found in other therian β-globin gene clusters [26]; see also [29]). Amino acid sequence analyses (BlastP) also provided additional support for the orthology of platypus ε to other mammalian ε-globin genes. Although we were unable to examine the expression of the genes in embryonic tissues, it was found that the expression profile of the platypus ε was similar to the embryonic α-like globins ζ and ζ' of the platypus, but not to the adult β-globin gene, supporting its potential role as an embryonic β-like globin gene.The ω-globin gene and the evolution of the β-globin clusterThe discovery of the marsupial ω-globin gene in the α-globin cluster [35,36] was critical in re-interpreting the relationships of the α- and β-globin clusters in amniotes (reptiles, birds and mammals) to favour the hypothesis that these clusters in birds and mammals are paralogous, having diverged independently from different ancestral copies of the vertebrate α-β-globin locus [63].Our observation of an ω-globin gene in the α-globin cluster in the platypus, as well as in the marsupials, confirms that the ancestral mammal α-globin cluster contained a β-like globin gene that was lost in eutherians, as proposed by Wheeler et al. [35,36]. However, the position of monotreme and marsupial ω in the phylogeny (Figure 3) is more consistent with the original hypothesis [5] that mammal and bird β-globin are orthologous, having descended from the same β-globin progenitor in an amniote ancestor, and this is strongly supported by flanking sequence data (see below). Our data support the proposition that the ω -globin gene represents an ancient β-like globin gene lineage that is ancestral to a group containing both mammalian and bird β-globins with a high posterior probability (99%). This arrangement, however, was not supported by analyses of amino acid sequence data, indicating that there is uncertainty in the phylogenetic position of ω-globin relative to bird β-globins, or that convergent evolution of bird β-globin genes and ω-globin resulted in their similarity at the protein level. To further resolve the key question of whether bird and mammal β-globin gene clusters are orthologous we carried out comparative analyses of flanking loci of the α- and β-globin clusters.Genome context of vertebrate α- and β-globin clustersWe found that the platypus α-globin cluster is flanked by MPG, C16orf35, GBY and LUC7L, and that the same genes (except GBY) flank the α-globin cluster in mammals and birds [58,59]. The same genes flank the α-β cluster of frog, and even zebrafish and the α-cluster of pufferfish [8] (except GBY and LUC7L), implying that a very ancient region containing these genes (5'-MPG-C16orf35-α-β-GBY-LUC7L-3'), or perhaps an even larger region, was present in their common ancestor and has been conserved since the evolution of jawed vertebrates more than 450 MYA.In contrast, the amniote β-globin clusters reside in a very different genome, sharing none of the flanking loci with the mammal and bird α-globin clusters, or the α-β cluster of frogs and fish. In platypus, as well as in therian mammals [60,61], the β-globin clusters are flanked by numerous ORG genes on both sides. In birds, also, the β-globin cluster is embedded in ORG genes [60]. Even the outside loci RRM1, CCKBR and ILK lie in the same orientation with respect to the bird and mammalian β-globin clusters [60], suggesting that the 5'-RRM1-ORG-β (cluster)-ORG-CCKBR-ILK-3' arrangement has been conserved since before the divergence of birds and mammals, more than 315 MYA. Therefore, the bird β-globin cluster is orthologous to the β-globin clusters of mammals.ConclusionNew model for the evolution of α- and β-globin clusters in amniotesThis analysis of flanking loci, in addition to the phylogenetic analyses reported above, refutes the prevailing hypothesis that mammal and bird α- and β-globin clusters evolved from different (paralogous) copies of an ancestral α-β-globin region containing MPG-C16orf35-α (cluster)-β (cluster)-GBY-LUC7L. Rather, the context of β-globin clusters within olfactory receptor genes in birds as well as mammals suggests that a copy of a β-globin locus was moved into a region replete with ORG genes before the divergence of birds and mammals 315 MYA. The precise mechanism for this translocation is unknown, but is likely to be either by transposition of a tandem duplicate of an ancestral β-globin gene, or retrotransposition of an intron-containing primary transcript. Phylogenetic analyses suggest that this ancestral β-globin gene within the α-globin cluster is represented by the platypus and marsupial ω-globin gene. The transposed β-globin gene then independently duplicated several times within the avian and mammalian lineages to form the different clusters of differentially expressed β-globin genes. Full details of this new model are given in Figure 9A and 9B.This hypothesis could be further tested by investigating the gene organization of the α- and β-globin clusters in reptiles such as lizards and snakes, which form a sister group to birds. Our hypothesis predicts that reptiles should possess a MPG-C16orf35-α (cluster)-β (cluster)-GBY-LUC7L cluster, and an unlinked RRM1-ORG-β (cluster)-ORG-CCBKR-ILK cluster like birds and mammals. The full genome sequence of the first reptilian species,Anolis carolinensis, will provide an opportunity to test this hypothesis.MethodsIsolation and purification of probes to screen for platypus β-globinsAt the start of this project there were no trace sequences available for any globin genes in the platypus trace archive. We therefore designed probes to screen the platypus male Oa_Bb BAC library (Clemson University Genomic Institute, USA). The platypus β-globin-specific primers OaBGF (5'-tggacccagaggttctttgac-3') and OaBGR (5'-tgcaattcactcagcttggag-3') were designed from the reference tammar β-globin sequence [GenBank: AY450928] using Primer3 [64]. Amplification by PCR was performed in a final volume of 25 μl, with 40 ng genomic DNA, 1× Buffer (Roche, Australia), 0.2 mM dNTPs, 0.05 U Taq (Roche, Australia) and 1 μM each of forward and reverse primers. PCR cycling conditions were: 94°C for 2 minutes, then 35 cycles of 94°C for 30 seconds, 50 to 60°C for 30 seconds, 72°C for 1 minute, followed by 72°C for 10 minutes. The PCR products were sub-cloned according to the TOPO TA cloning® Kit Protocol (Invitrogen, Australia) and the resulting plasmids were purified according to the centrifugation protocol of Wizard® Plus SV Minipreps DNA Purification System (Promega, Australia). The purified plasmids were confirmed to contain PCR products of a partial platypus β-globin gene (167 bp) by sequencing at the Australian Genome Research Facility (AGRF, Brisbane, Australia) using M13 forward (5'-gtaaaacgacggccag-3') and M13 reverse (5'-caggaaacagctatgac-3') primers. Once confirmed, they were used as probes to screen the platypus BAC library.Screening the platypus BAC library for β-like globin genesThe platypus BAC library filters were pre-hybridised at 65°C with Church Buffer (1 mM EDTA, 0.5 M phosphate buffer, 7% (w/v) SDS) including 1% BSA for 4 hours. The platypus β-globin probes (25 ng) were labelled with 32P-dATP using the Megaprime DNA labelling System (GE Healthcare, Australia) following the manufacturer's instructions. The probes were allowed to hybridise to the filters at stringent conditions (65°C with the above buffer) for 24 hours and then washed twice for 15 minutes each in 2 × SSC/0.1%SDS and 1 × SSC/0.1%SDS. Autoradiography was carried out for 14 days at -80°C with an intensifying cassette.Identification of platypus BAC clones containing α-like globin genesUnlike β, BACs were not screened for α-like globin genes. Instead they were identified directly from the Encyclopaedia Of DNA Elements Project [65], in which the α-globin cluster is one of the targeted regions [12,66]. Two platypus BAC clones (Oa_Bb-2L7 and Oa_Bb-131M24), which were sequenced but not yet annotated, were identified by computational analysis (below) to contain parts of the α-globin cluster and a ω-globin gene.Isolation and purification of DNA from BAC clonesDNA from the identified BAC clones (including those that were screened) was extracted using Wizard® Plus SV Minipreps DNA Purification System (Promega, Australia). The purified BAC clones were then subjected to Dot or Southern Blot to confirm the presence of α- or β-globin genes respectively.Confirmation of BACs containing globin genesDot blot methods were used to verify the presence of the α-like globin genes. In a plate containing Luria broth agar with chloramphenicol, a Hybond N+ (GE Healthcare, Australia) filter was placed and multiple 1 μl of liquid culture BAC clones were spotted onto the filter. The plate was incubated at 37°C overnight and then the filter was soaked in Denaturation Solution (0.5 M NaOH and 1.5 M NaCl) for 5 minutes, followed by soaking twice in Neutralisation Solution (0.5 M Tris-Cl pH 7.4 and 1.5 M NaOH) for 5 minutes each. The filter was then rinsed in 2 × SSC, soaked in 0.4 M NaOH for 20 minutes and washed with 6 × SSC to remove all cellular debris. The filters were then screened with the platypus α-globin probes using the standard library screening procedure (above).Southern blotting was used to verify the presence of the β-like globin genes. In a 40 μl reaction, 20 to 40 ng BAC DNA was digested with 10 U of restriction enzyme,HIND III (Roche, Australia). The reaction was incubated at 37°C for at least 4 hours and separated by electrophoresis on a 0.8% agarose gel overnight at 40 V. The DNA fragments were transferred onto a Hybond N+ (GE Healthcare, Australia) nylon filter overnight by capillary action following the manufacturer's instructions, and cross-linked in 0.4 M NaOH for 20 minutes. These filters were then screened with the platypus β-globin probes using the standard library screening procedure (above).Fluorescence in situ hybridisation (FISH)Male platypus metaphase spreads were prepared and in situ mapping was performed using two-colour FISH as described previously by McMillan et al. [21]. The verified BACs containing the α-like globin genes (ζ and ζ': Oa_Bb-2L7) and β-like globin genes (ε and β: Oa_Bb-484F22) were labelled with different fluorochromes and then hybridised to the chromosomes. The signals were detected by fluorescent microscopy, where at least twenty metaphase images were captured and analysed.Sequence data of the platypus BAC clones containing the α- and β-globin clustersInformation about the platypus BAC clones containing the α-like globin genes along with the ω-globin gene was obtained directly from the ENCODE Project [66]. Their sequence information was obtained from GenBank; accession numbers: AC195438 (Oa_Bb-2L7) and AC203513 (Oa_Bb-131M24).The BAC clone containing the β-like globin genes that were found from the library screening procedure were sequenced at the Washington University Genome Sequencing Centre (St Louis, USA). The sequence information for this BAC clone was obtained from GenBank: AC192436 (Oa_Bb-484F22).Computational characterisation of the α- and β-globin clusters in the platypusUsing sequence information of AC195438, AC203513 and AC192436, genes were predicted by GENSCAN [46] and GenomeScan [47] using default settings. All predicted gene sequences were then subjected to a BLAST search of the translated nucleotide acid (BlastX) and protein (BlastP) databases to confirm their identities.Promoter analysesTranscription factor binding motifs were predicted in the 200 bp promoter region located 5' to the predicted platypus α- and β-like genes and GBY by rVista 2.0 [67] using user-defined consensus sequences for 'CACCC', 'CAAT', 'TATA', GATA1 ('WGATAR' [51]) and EKLF ('NGNGTGGGN' [51]). The same criteria were used to predict the same motifs in marsupials (Didelphis virginiana [ζ and ψζ': AC139599] and Sminthopsis macroura [αD, ψα3, α2, α1, ω: AC146781; and ε, β: AC148754]) and in humancs [ζ, ψζ', αD, ψα3, α2, α1: NG_000006; and ε, β: NG_000007] for consistency in comparison.Confirmation of α1 and α3 by BLAST search and Southern blotTo confirm that the presence of two almost identical genes (α1 and α3) was real rather than an assembly error, the boundaries (~300 bp) of the homologous regions were investigated by a BLAST search against the platypus WGS database. The raw sequences of best hits were extracted from NCBI GenBank, cleaned and aligned in Sequencher v4.8 (Gene Codes Corporation, Michigan) using default settings.Southern blotting was also used to verify the presence of α1 and α3 genes. In a 30 μl reaction, 100 μg BAC DNA (Oa_Bb: 131M24, 130N2, 150K14 and 223I12) was digested with 10 U of restriction enzyme,EcoRV (Roche, Australia). The reaction was incubated at 37°C for at least 4 hours and separated by electrophoresis on a 0.8% agarose gel overnight at 40 V. The DNA fragments were transferred onto a Hybond N+ (GE Healthcare, Australia) nylon filter overnight by capillary action following the manufacturer's instructions, and cross-linked in 0.4 M NaOH for 20 minutes. These filters were then screened with the platypus α1/α3 (test) and α2 (control) probes using the standard library screening procedure (above).RT-PCR analysesTo remove DNA contamination, RNAs derived from adult male platypus liver, kidney, spleen, testis, brain and lungs were DNase treated using a DNA-free™ kit according to the manufacturer's instructions (Applied Biosystems, Australia). Treated RNAs were then reverse transcribed using Superscript III (Invitrogen, Australia) following the manufacturer's instructions. Primers were designed against predicted α- and β- like and GBY globin gene sequences using Primer3 [64]. In each case, the region amplified spanned an intron so that the origin of the template (gDNA or cDNA) was immediately obvious. Primer sequences and the expected sizes of amplified cDNA and gDNA bands are shown in Table 2. PCR reactions and cycling conditions were the same as for screening for the platypus β-globin genes, above. The positive bands were directly sequenced by AGRF (Brisbane, Australia) to confirm their identities. The blood contamination in the tested samples had minimal effect on the observed expression pattern, as some tissues (for example, lung and liver) showed no amplification despite containing large quantities of blood.Table 2PCR primers used for amplification of the α- and β-like globin genes including GBY from the platypus gDNA and cDNAGeneForward PrimerReverse PrimergDNA(bp)cDNA (bp)ζGGCCGACAAGACCGCAGTCATCTCCCCCCGATGGCGCTGATGACT527190ζ1TGACCAAAGGCGACAAGACCTCCCCGATGGCACCGATGACC534198αDGAGGCTGTGAAGAACCTGGAGGTGTACTCCCCTTGCAGAT1793153α2TGGCCCACCTCGATGACCTGGGGGAAGGTGTCTGGCCACC289134α1/α3GCAAGGCCGCCGGTCACGGCCGCTGTCCATGTCATCGAAGTGCC597192ωATTGTGTCCATCTGGGGAAAGCTTGGCAAAGTTGCTCTTC488232GBYCTGGAAACAGGTGTGCAAGACTATCTCCGGGGTGTAGCAG3202149εATCTGAGCGCTGAGGAGAAGGACAGGTTGCCGAAGGAGTCA285142βCTGTGGGGGAAAGTGAACATGGTCAGCACCTTAGCACCAT321168Phylogenetic analysesPhylogenetic analyses were employed to verify the identities of the platypus globin genes and study the evolutionary relationships of the different members of the α- and β-globin gene families. This study was restricted to the coding domains of the α- and β-globin members and the accession numbers of the sequences used are given in the legends of Figures 2 and 3. Phylogenetic analyses were conducted using MP in PAUP* v.4.0b10 [68], and a BI approach using MrBayes v.3.1.2 [69]. Concordance of trees from each of the different methods, bootstrap proportions and posterior probability estimates were used to examine the robustness of nodes.MP analyses were conducted for the entire coding sequence matrix and after excluding third codon positions using a heuristic search option and default options (TBR branch swapping), with the exception of using random stepwise addition repeated 100 times. Character state optimisation for MP trees used the DELTRAN option. MP bootstrap analyses [70] were carried out using 1000 bootstrap pseudoreplicates, employing a heuristic search option with random stepwise addition.The program MODELTEST [71] and the Akaike information criterion (AIC) were used to assess the most appropriate model for BI analyses. The MODELTEST analyses were facilitated using the program MrMTgui v1.0 [72]. The MODELTEST analysis was carried out on separate codon positions for α- and β-globin data sets. For α-globin sequences, a general time reversible (GTR) model [73], with a proportion of invariant sites (I) and unequal rates among sites [74], modelled with a gamma distribution (G) was found to be the most appropriate model to use for first and second codon positions, and a GTR+G model was appropriate for third codon positions under the AIC. For β-globin sequences a GTR+I+G model was considered appropriate for first positions, and a GTR+G model was found to be appropriate for second and third codon positions. The MrBayes analysis was carried out applying these different models to each codon position using an unlinked analysis, with default uninformative priors. Four chains were run simultaneously for 2 million generations in two independent runs, sampling trees every 100 generations. The program TRACER (version 1.3; [75]) was used to assess tree and parameter convergence. For both the α-globin and β-globin analyses all effective sample sizes for all parameters were >1297, indicating a sufficient sample of the parameter space had been taken. A burn-in of 2000 trees (equivalent to 200,000 generations) was chosen for each independent run of MrBayes, with a >50% posterior probability consensus tree constructed from the remaining 36,002 trees (18,001 trees each run).A BI analysis using MrBayes (version 3.1.2) was also carried out using protein sequence data from β-globin genes. A mixed protein model was used, allowing the optimum model of protein evolution to be assessed from a selection of nine fixed-rate models. The optimum model was found to be the Dayhoff model with a posterior probability of 1.0. The analyses were conducted using two million generations in two independent runs, sampling trees every 100 generations. A burn-in of 2,000 trees was used for each run with a 50% consensus tree constructed from the remaining 36,002 trees.Authors' contributionsVSP designed and performed most of the experiments and analysed the data. VSP also drafted the main manuscript. SJBC conducted phylogenetic analyses and contributed to the writing of the manuscript. JED helped in designing the experiments and trouble-shooting experiments. BF, TG, WCW and RKW were involved in sequencing the platypus BAC clone (Oa_Bb-484F22). JAMG conceived and supervised the research and contributed to the writing of the manuscript. All authors read and approved the final manuscript.Competing InterestsThe authors declare that they have no competing interests.Supplementary MaterialAdditional file 1Annotation of the platypus α- and β-like and GBY globin genes. This table shows the predicted positions of six α-like, ω and GBY globin genes in the platypus BAC clone Oa_Bb-131M24 [GenBank: AC203513], two α-like globin genes in BAC clone Oa_Bb-2L7 [GenBank: AC195438], and two β-like globin genes in BAC clone Oa_Bb-484F22 [GenBank: AC192436 reverse direction].Click here for file\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2529281.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2529281",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2529281\nAUTHORS: Neha Patel, David Hoang, Nathan Miller, Sara Ansaloni, Qihong Huang, Jack T Rogers, Jeremy C Lee, Aleister J Saunders\n\nABSTRACT:\nA number of studies have shown that increased APP levels, resulting from either a genomic locus duplication or alteration in APP regulatory sequences, can lead to development of early-onset dementias, including Alzheimer's disease (AD). Therefore, understanding how APP levels are regulated could provide valuable insight into the genetic basis of AD and illuminate novel therapeutic avenues for AD. Here we test the hypothesis that APP protein levels can be regulated by miRNAs, evolutionarily conserved small noncoding RNA molecules that play an important role in regulating gene expression. Utilizing human cell lines, we demonstrate that miRNAs hsa-mir-106a and hsa-mir-520c bind to their predicted target sequences in the APP 3'UTR and negatively regulate reporter gene expression. Over-expression of these miRNAs, but not control miRNAs, results in translational repression of APP mRNA and significantly reduces APP protein levels. These results are the first to demonstrate that levels of human APP can be regulated by miRNAs.\n\nBODY:\nResultsAccumulating evidence suggests that increased expression of the amyloid precursor protein gene (APP) increases Alzheimer's disease (AD) risk. The resulting increase in APP protein levels results in increased Aβ levels, leading to synaptic dysfunction, neurodegeneration and, eventually, cognitive decline.APP levels can be regulated at the genomic, transcriptional or translational level. At the genomic level, Down's Syndrome (Trisomy 21) patients have three copies of the APP gene and develop AD symptoms early in life [1]. Similarly, duplication of the APP locus, in the absence of a full trisomy 21, also leads to early-onset AD [2]. Dysregulation of APP transcription can also increase the risk of AD. Genetic variants in the APP promoter increase APP transcription by ~2–3 fold and have been reported to increase AD risk [3]. Growth factors have been reported to control APP mRNA half-life [4]. These growth factors effects are dependent on a 29 bp sequence in the APP 3' UTR [4,5]. APP translation is also regulated; for example, IL-1 can induce an increase in APP translation [6]. IL-1 is a pro-inflammatory cytokine and genetic variants have been linked to increased AD risk [7,8]. Taken together, these findings provide strong evidence that increased APP levels increase AD risk.MicroRNAs (miRNAs) are small noncoding RNAs that control gene expression post-transcriptionally. Complementary binding between miRNAs and sequences within the 3' UTR of target genes results in repression of target gene expression by translational inhibition or mRNA degradation [9]. Approximately 700 miRNA genes are encoded in the human genome and recent evidence demonstrates that some miRNAs are differentially expressed in AD patients compared to age-matched controls [10]. These differences in miRNA expression may play an important role in AD pathogenesis. In an attempt to address this possibility, we test the hypothesis that miRNAs can regulate APP levels.Bioinformatic analysis predicts that the 3' UTR of human APP contains 28 unique miRNA target sites [11,12]. To experimentally confirm that APP levels can be regulated by miRNAs, we chose to initially study miRNA hsa-mir-106a (mir-106a; Figure 1A) since (i) the putative target site in the APP 3'UTR is 100% complementary to the seed region of the miRNA, (ii) it has a large free energy of seed region binding, and (iii) it is expressed in human brain [13]. To determine if the putative mir-106a target site in the APP 3'UTR is capable of regulating gene expression, we cloned it into the 3' UTR of firefly luciferase. We co-transfected this reporter into naïve HEK-293 cells along with a mir-106a over-expression vector [14] and measured luciferase activity (Figure 1B). We observed a significant ~50% decrease (p < 0.0001) in luciferase activity when the putative mir-106a target site was included in the reporter compared to either a reporter lacking the putative target site or reporter carrying a seed-region mutant of the putative mir-106a target site. To determine if this effect was simply due to over-expressing miRNAs, we repeated the experiment while over-expressing mir-373, a miRNA not predicted to target the APP 3'UTR. We observed no change in luciferase activity. Another miRNA, mir-520c, shares the same seed region target sequence as mir-106a but is not expressed in human brain (Figure 1A) [13]. Therefore we tested mir-520c, we observed that mir-520c over-expression significantly decreased luciferase activity when the putative mir-106a target site was included in the reporter compared to either a reporter lacking the putative target site or a reporter carrying a seed-region mutant of the putative mir-106a target site (Figure 1B). We repeated these experiments in the human neuroblastoma cell line SH-SY5Y and observed similar results (data not shown). To confirm that the miRNAs were being over-expressed, we utilized RT-QPCR to quantify miRNA levels and observed significant increases in both mir106a and mir-520c (Table 1) levels (p < 0.0001).Figure 1mir-106a target sequence regulates reporter gene expression. (A) Predicted mir-106a and mir-520c target sites in the 3'UTR of APP. (B) Over-expression of mir-106a or mir-520c, but not mir-373, significantly reduced luciferase expression (p = 0.0006) controlled by the putative mir-106a APP 3'UTR target sequence. This reduction is not observed when a seed region mutant of mir-106a (106a*) is utilized. For all experiments, three independent trials were performed. Error bars represent standard deviation. *p < 0.05; **p < 0.01, compared to the appropriate control.Table 1Relative miRNA levelsFold Change 2-ΔΔCtp-valueVector1mir-106a30.1 ± 1.2< 0.0001mir-520c1964.6 ± 1.1< 0.0001QPCR results demonstrate a significant increase in mir-106a and mir-520c levels in response to over-expression compared to cells transfected with an empty vector.Having demonstrated that over-expression of mir-106a or mir-520c was capable of repressing reporter gene expression via interaction with its putative target site, we investigated whether over-expression of these miRNAs could decrease endogenous APP levels in human cell lines. We transfected naïve HEK-293 with mir-106a and mir-125b over-expression vectors and then performed quantitative Western blot analysis to determine APP steady state levels. We utilized mir-125b as a negative control since it is not predicted to target APP but has increased expression in AD brain [10]. We observed that mir-106a over-expression significantly decreased APP levels (Figure 2A). Both APP isoforms expressed in this cell line, APP770 and APP751, were significantly and similarly affected. Mir-106a over-expression reduced APP levels by ~50% (p < 0.01) compared to cells transfected with the empty vector (Figure 2C). Over-expression of mir-125b had no significant effect on APP levels. Over-expression of mir-520c had a similar effect on APP levels as mir-106a (Figure 2B and 2D).Figure 2mir-106a and mir-520c can regulate APP levels post-transcriptionally. APP770 and APP751 levels are reduced in cells over-expressing (A) mir-106a compared to cells expressing either mir-125b or the empty vector and (B) mir-520c compared to cells expressing the empty vector. (C) Quantification of these Western blot results reveals that mir-106a over-expression significantly decreases APP levels (p < 0.01) compared to cells expressing mir-125b or cells transfected with the empty vector. (D) Quantification of these Western blot results reveals that mir-520c over-expression significantly decreases APP levels (p < 0.01) compared to cells transfected with the empty vector. (E) QPCR results show that APP mRNA levels are not altered by mir-106a or mir-520c over-expression. For all experiments, three independent trials were performed. Error bars represent standard deviation. *p < 0.05; **p < 0.01, compared to the appropriate control.Most human miRNAs repress gene expression by inhibiting translation and do not affect target gene mRNA levels [15,16]. This seems to be the case in our experimental setting. We utilized RT-QPCR to determine if miRNA over-expression resulted in decreased APP mRNA levels. Over-expression of mir-106a or mir-520c had no effect on APP mRNA levels (Figure 2E). Mir-106a and mir-520c, therefore, appear to inhibit translation of the APP transcript.Our results are the first to experimentally demonstrate that human APP levels can be regulated by miRNAs. In 2004, it was predicted that APP levels could be regulated by miRNAs [12]; recently it was shown that expression of the C. elegans orthologue of APP, APL-1, is regulated by developmentally-timed miRNAs [17].In human neurons the APP695 isoform is the predominant expressed isoform. All APP isoforms (APP695, APP751 and APP770) share the same 3' UTR [18] therefore we expect that the mir-106a mediated regulation of APP levels that we observe in the HEK-293 cell line should also occur in neurons given that mir-106a is expressed in the brain. We do not expect mir-520c mediated APP regulation to occur in neurons since this miRNA is not expressed in brain. It is important to test if miRNA regulation is a normal aspect of APP metabolism in neurons. If so, it will be important to determine whether AD pathogenesis is affected by alterations in miRNA function and/or expression. It is possible that aging- or environment-induced changes in miRNA expression, and/or sequence variation in miRNAs or their targets, contribute to increased APP levels and increased AD risk. Recently, it was demonstrated that expression of the β-secretase BACE can be regulated by miR-29a/b-1 and mir-107; furthermore, increased BACE levels correlated with decreased miR-29a/b-1 and mir-107 levels in AD patients [19,20].Regardless of the biological roles of miRNA in APP metabolism, therapeutics based on miRNA-induced decrease in APP levels would offer a treatment targeting the underlying pathophysiology of the disease. In the near future, substantial progress will be made in understanding the role of miRNAs in AD pathogenesis and in therapeutic approaches to treating AD.MethodsReporter Vectors and DNA constructsReporter vectors containing the putative miRNA target sites from the APP 3'UTR, were synthesized with double-stranded oligos perfectly complementary to putative miRNA target sites and oligos in which the seed regions were mutated. The mir-106a target oligos had the sequence (seed region bolded):5' CTAGTAATCCCTGTTCATTGTAAGCACTTTTGCTCAGCA 3'3' ATTAGGGACAAGTAACATTCGTGAAAACGAGTCGTTCGA 5'The mutant mir-106a target oligos had nucleotides three through six of the seed region mutated (italicized):5' CTAGTAATCCCTGTTCATTGTAAGCGTCCTTGCTCAGCA 3'3' ATTAGGGACAAGTAACATTCGCAGGAACGAGTCGTTCGA 5'We utilized established methods [21] to clone these synthetic versions of putative miRNA target sites into a luciferase reporter gene (pMIR-REPORT; Ambion).Cells and Cell CultureNaïve human embryonic kidney (HEK)-293 cells were purchased from ATCC. Cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, 2 mM L-glutamine,100 units/ml penicillin and 100 μg/ml streptomycin.Transfections and Luciferase Assays10,000 Naïve HEK-293 were plated in 24 well plates. The next day, cells were transfected with a miRNA overexpression vector [14], reporter vectors bearing either the miRNA target sequence or the miRNA seed region mutant target sequence, and one tenth of the molar volume of pRL-SV40, a Renilla Luciferase control vector. We utilized Arrest-In transfection reagent (Open Biosystems Inc.); any differences in transfection efficiency were accounted for by measuring Renilla luciferase activity. 48 hours post-transfection, cell were lysed using 100 μL of GLB (Glo Lysis Buffer, Promega). Firefly and Renilla luciferase activities were measured using a dual luciferase reporter assay kit (Promega), per the manufacturer's protocol. Firefly luciferase activity was normalized to Renilla luciferase activity.Western Blot Analysis200,000 Naïve 293 cells were plated in 6 well plates. The next day, cells were transfected with a miRNA overexpression vector. Using previously described methods [22], quantitative Western blots were performed using equal amounts of total protein.AntibodiesA polyclonal antibody specific for the C-terminus of human APP (A8717; Sigma Aldrich, Inc) and a monoclonal antibody specific for human β-Actin (A5441; Sigma Aldrich) were used as primary antibodies. Secondary antibodies were HRP-conjugated goat anti-rabbit (GE Healthcare) and HRP-conjugated goat anti-mouse (GE Healthcare).RNA extraction and Quantitative PCR48 hours post-transfection, cells were washed with cold PBS and total RNA was isolated using RNeasy Mini Kit (Qiagen Inc.). To quantify APP mRNA levels, cDNA was synthesized using total RNA, N6 random primers and SuperScript II Reverse Transcriptase (Invitrogen). cDNA was then diluted 1:15 using RNase free water and mixed with APP or GAPDH primer/probe sets (Applied Biosystems, Inc.; APP Catalog # Hs00169098_m1; GAPDH Catalog # Hs99999905_m1), 2× PCR Universal Master Mix (Applied Biosystems, Inc.) and amplified using an ABI 7500 Real Time PCR system following the manufacturer's directions. GAPDH was used as an internal control. To determine differences in APP mRNA levels, we utilized the ΔΔCt method.To quantify miRNA levels, cDNA was reverse transcribed from total RNA samples using specific miRNA primers from the TaqMan MicroRNA Assays and reagents from the Taq Man MicroRNA Reverse Transcription kit (Applied Biosystems). The resulting cDNA was amplified by PCR using TaqMan MicroRNA Assay primers with the TaqMan Universal PCR Master Mix and analyzed with a 7500 ABI PRISM Sequence Detector System (Applied Biosystems) according to the manufacturer's instructions. The relative levels of miRNA expression were calculated from the relevant signals by the ΔΔCt method by normalization to the signal of RNU44 [23].Statistical AnalysisValues in the text and figures are presented as means ± standard deviations of experiments carried out in triplicate, at least. Each experiment was carried three times. Equal variance or separate variance two-sample student's t-test were used, as appropriate, to compare two groups. Where appropriate, Bonferroni analysis was used to correct for multiple comparisons within a single experiment.Competing interestsAJS declares that he is a share holder in TorreyPines Therapeutics. The remaining authors declare that they have no competing interests.Authors' contributionsAll authors have read and approved the final manuscript. NP designed the experiment, acquired, analyzed and interpreted the data and drafted the manuscript. DH and NM cloned the target and mutant sequences in the reporter vectors. SA helped to draft and edit the manuscript. JTR contributed towards experimental design. QH contributed towards experimental design and provided miRNA over-expression vectors.JCL performed the statistical analysis. AJS oversaw the experimental design, data analysis, data interpretation, and drafting/editing the manuscript.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2529306.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2529306",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2529306\nAUTHORS: Teresa Spanu, Donato Rigante, Gianpiero Tamburrini, Barbara Fiori, Tiziana D'Inzeo, Brunella Posteraro, Domenico Policicchio, Maurizio Sanguinetti, Giovanni Fadda\n\nABSTRACT:\nIntroductionStaphylococcus lugdunensis is an unusually virulent coagulase-negative staphylococcus that has rarely been implicated in central nervous system infections.Case presentationTwo children hospitalized in the Neurosurgery Unit developed ventriculitis caused by methicillin-resistant Staphylococcus lugdunensis following placement of external ventriculostomy drains. The causative organisms were identified by molecular studies. The patients recovered without significant sequelae after high doses of intrathecal vancomycin.ConclusionDistinguishing Staphylococcus lugdunensis from other coagulase-negative staphylococcus species is crucial because it carries a substantial risk for severe central nervous system infections displayed by patients with implanted cerebrospinal fluid devices. Clinicians should not underestimate the importance of the isolation of this species from cerebrospinal fluid specimens.\n\nBODY:\nIntroductionFirst described by Freney et al. in 1988 [1], Staphylococcus lugdunensis is an unusually virulent coagulase-negative staphylococcus (CoNS) known primarily as a cause of endocarditis, especially in immunocompromised patients [2]. It has also been associated with septic arthritis, osteomyelitis, peritonitis, brain abscesses, and infections of the skin and soft tissues, urinary tract, and prosthetic medical devices [3]. Its remarkable virulence has been attributed to the production of extracellular slime, which facilitates colonization and interferes with phagocytosis-associated activities of neutrophils [4]. Some strains also produce a synergistic hemolysin that resembles the δ-hemolysin of S. aureus, consisting of three very similar 43-residue peptides closely related to the gonococcal-growth-inhibitor bacteriocin secreted by S. haemolyticus. Nucleic acid sequences related to the accessory gene regulator (the major determinant of virulence in S. aureus) have also been demonstrated in S. lugdunensis [5].A MEDLINE search conducted with the keywords \"S. lugdunensis\" AND \"cerebrospinal fluid (CSF)-shunt infection\" OR \"central nervous system (CNS) infection\" yielded only four cases, which are summarized in Table 1. Cases #1 through #3 were S. lugdunensis ventriculo-peritoneal shunt (VPS) infections [6,7], and case #4 was a S. lugdunensis meningitis unrelated to implanted CSF drainage devices [8]. The other two cases of S. lugdunensis infections reported in Table 1 were recently diagnosed in the Pediatric Neurosurgery Unit of our Medical Center.Table 1Clinical data and outcome of ventriculitis or meningitis caused by Staphylococcus lugdunensisCase #Gender  – AgeUnderlying diseaseCSF shuntaTime to infectionbSigns and symptomsAntimicrobial therapyOutcomeReference (year)Case #1 M – 74 yearsVentriculomegalyVPS14 daysFever,abdominal  pain, sweatingVancomycinc +rifampicin +cefuroxime,thenvancomycinc +rifampicin, thenciprofloxacin +       rifampicinRecoveredSandoe (2001)Case #2F – 10  monthsObstructive hydrocephalusVPS3 daysFever,irritability, decreased  activityOxacillinRecoveredElliott (2001)Case #3F –  16 yearsAqueduct stenosisVPS2 yearsFever, lethargy,abdominal  complaintVancomycincthen oxacillinRecoveredElliott (2001)Case #4M –  12 yearsObstructive hydrocephalusNDd-Fever,headache,vomiting,   lethargyVancomycinc +cefotaxime,then oxacillin +   rifampicinRecoveredKaabia (2002)Case #5M – 7  yearsTumor within theposterior cranial  fossaEVD20 daysFever,headache,vomiting,   lethargyVancomycineRecoveredThis studyCase #6M – 2  monthsMalformation of Galen's veinEVD19 daysFever, seizures,impaired  consciousnessVancomycineRecoveredThis studyaCSF shunt, cerebrospinal fluid shunt included VPS or EVD.bFrom shunt placement to infection onset.cAdministered intravenously.dND = not described.eAdministered intrathecally.This report analyzes the management of these patients in light of the few previously reported cases of S. lugdunensis CNS infections and summarizes the molecular characteristics of the isolates recovered from CSF and ventricular drain cultures.Case presentationPatient 1A 7-year-old boy was hospitalized in the Pediatric Neurosurgery Unit for headache, vomiting, and right ocular dysmetria. Magnetic resonance imaging (MRI) revealed obstructive hydrocephalus caused by a posterior fossa tumor. The child was taken to the operating room for placement of an external ventriculostomy drain (EVD). CSF cultures yielded no growth. There was no improvement and, on the 10th day of hospitalization, he had a second operation for partial removal of the tumor (a medulloblastoma). A new EVD was inserted; cultures of the original EVD and CSF yielded no growth. Seven days later, the child developed severe headache, fever (39.5°C), vomiting, lethargy, and signs of EVD malfunction. Shunt puncture yielded cloudy CSF containing 900 leukocytes/mm3 (80% polymorphonuclear cells); 100 mg protein/dl); and 21 mg glucose/dl (blood glucose: 87 mg/dl). Cytocentrifuge Gram staining revealed Gram-positive cocci that were later identified by biochemical and molecular methods as Staphylococcus lugdunensis.Meanwhile, a presumptive diagnosis of ventriculitis was made [9], a new EVD was inserted, and intrathecal (IT) vancomycin (40 mg/day) was started. Cultures of the CSF and the ventricular tip of the second EVD grew methicillin-resistant S. lugdunensis. Three blood cultures yielded no growth. Swabs of both inguinal folds and the surgical incision were cultured, but none yielded S. lugdunensis. Defervescence occurred after 2 days of IT vancomycin. After 14 days of vancomycin, the composition of the CSF was normal, and cultures of CSF and EVD were negative. The boy was discharged after 62 days of hospitalization, and no sign of CNS infection was noted at the 6-month follow-up visit.Patient 2A 2-month-old male was admitted to the Pediatric Neurosurgery Unit for rapid head growth with a tense anterior fontanel. He was placed in the same room where Patient 1 was still being cared for. Triventricular hydrocephalus was evident on cranial ultrasonography, and MRI with angiographic sequences disclosed a malformation involving Galen's vein that caused arteriovenous shunting and obstruction of CSF flow through the aqueduct. The malformation was treated successfully with intravascular embolization, but 1 week later, the infant developed seizures. Cerebral angiography confirmed occlusion of the aneurysmal malformation, but computed tomography revealed an intraventricular/subarachnoid hemorrhage. An EVD was placed for continuous intracranial pressure monitoring and collection of CSF specimens. All CSF cultures were negative, including the one obtained during EVD placement. Three days later, the child was febrile (39.2°C) and drowsy. A cloudy CSF specimen collected from the EVD contained 129 mg/dl protein, 20 mg/dl glucose (blood glucose: 113 mg/dl), 800 leukocytes/mm3 (60% were polymorphonuclear cells), and 5 to 10 Gram-positive cocci per microscopic field that were identified in biochemical and molecular assays as S. lugdunensis.The clinical picture was compatible with ventriculitis [9], and IT vancomycin (40 mg/dose/day) was started immediately after EVD replacement and continued for 14 days. Cultures of the CSF and the tip of the original EVD grew methicillin-resistant S. lugdunensis. Three blood cultures were negative. Skin cultures (inguinal folds, surgical incision site) were all negative for S. lugdunensis. The infant's condition rapidly improved after vancomycin was started, and cultures of CSF collected 15 days later and of the second EVD showed no growth. The infant was discharged after 40 days of hospitalization. Six months later, he was well with no evidence of infection and no acute neurological signs.Microbiological diagnosisOur routine protocol for suspected CSF shunt infections includes Gram staining of cytocentrifuged CSF, aerobic culture at 35°C on MacConkey agar, microaerobic culture (35°C in room air with 5% CO2) on Columbia and chocolate agars, anaerobic culture (35°C) on Schaedler agar (all from bioMérieux, Marcy-L'Etoile, France), and 72 hours of aerobic and anaerobic cultures on brain-heart infusion broth supplemented with 5% NaCl.In these two patients, cultures of CSF and EVD tips on Columbia agar produced yellowish beta-hemolytic colonies (diameter: 0.8–2.5 mm) of Gram-positive cocci. Tube coagulase tests with rabbit plasma (Becton Dickinson Microbiology Systems, Sparks, MD) were negative. The isolates were positive for catalase, clumping factor, pyrrolidonyl arylamidase, and ornithine decarboxylase. They were identified as S. lugdunensis by the API ID32 STAPH (bioMérieux) and the VITEK 2 (bioMérieux, Inc, Hazelwood, MO) systems.Bacterial DNA was extracted from CSF specimens and culture isolates with the QIAmp DNA Mini kit (QIAGEN, Hilden, Germany). Species-level identification was confirmed by sequencing of the 16S rRNA gene (using the RIDOM entries) and the RNA polymerase B gene [10]. Sequence analysis of the 16S rRNA gene revealed 100% homology with the prototype strain sequence of S. lugdunensis ATCC 43809 (Z26899). A partial sequence of the rpoB gene of each isolate revealed 99.5% homology with the prototype strain sequence of S. lugdunensis ATCC 43809 (EF173667).Antimicrobial susceptibility testing with the E-test (AB Biodisk, Solna, Sweden) yielded the following MICs (μg/ml): oxacillin, 256; vancomycin, 0.5; erythromycin, 0.03; ciprofloxacin, 0.03; clindamycin, 0.03; rifampicin, 4.0; quinupristin-dalfopristin, 0.5; linezolid, 1.0. Resistance was defined by Clinical Laboratory Standards Institute breakpoints [11]. The mecA gene (reflecting methicillin resistance) was detected by PCR, as described by Geha and colleagues [12]. Pulsed-field gel electrophoresis (PFGE) analysis of SmaI- and ApaI-digested chromosomal DNA [13] revealed identical profiles for all isolates recovered from Patients 1 and 2 (data not shown).DiscussionTo date, there have been no reports of S. lugdunensis ventriculitis associated with EVDs, which are widely used in neurosurgery for continuous intracranial pressure monitoring, injection of therapeutic agents, and temporary external drainage of CSF [9]. EVD-associated CSF infections can be classified as ventriculostomy-related infections or ventriculitis [9]. The former are generally associated with few clinical symptoms. The CSF is characterized by culture- or Gram-stain-positivity with progressive decrease in glucose and progressive increase in proteins. Progression to ventriculitis is heralded by high-grade fever and signs of meningitis (for example, nuchal rigidity, photophobia, deteriorating mental status, seizures, moribund appearance). The latter pattern was seen in both of our patients. Fever and meningeal signs were also reported in cases #2, 3 and 4 (Table 1), whereas the VPS infection in case #1 was associated with severe non-neurological symptoms (intra-abdominal sepsis with purulent peritonitis) [7].Our patients and one of those reported by Elliot and colleagues [6] had hospital-acquired infections, which developed 3 to 7 days after shunt insertion. The other infections shown in Table 1 were evident at hospital admission and would thus seem to be community-acquired (although in one case [7] the admission occurred 14 days after a previous hospitalization during which a CSF shunt had been inserted).EVD use in critically ill neurosurgical patients is on the rise and reported rates of infection associated with these devices vary widely (from 0 to 45%). The prevailing opinion is that the infecting agent is usually introduced during EVD placement [9], which is consistent with the fact that most EVD-associated infections are caused by skin flora, particularly Staphylococcus epidermidis.Like other CoNS species, S. lugdunensis is considered part of the resident flora of the human skin and mucous membranes, although the preferred carriage site seems to be the perineum [14]. In both of our patients, infection onset occurred a few days after surgery, the patients were being cared for in the same hospital room, and all isolates displayed the same pulsotype. These findings are suggestive of a common source, which unfortunately has not been identified.We cannot exclude the possibility that the infections were transmitted during manipulation of the catheters. Healthcare-provider hand cultures (data not shown) and patient skin cultures yielded no growth of S. lugdunensis. However, the staff surveillance cultures were collected after the second child had been infected, not during the time of EVD insertion, so the infection reservoir might have been missed. Environmental cultures and inguinal cultures of providers were not performed.Hellbacher and colleagues [13] recently suggested that PFGE is unsuitable for analyzing outbreak situations involving S. lugdunensis. The homogeneity they observed among 39 isolates collected over 4 years suggests that S. lugdunensis is either a highly conserved species or that specific clones are more likely to cause invasive infections. Other investigators [14], however, have found that PFGE with SmaI macrorestriction analysis is inappropriate for epidemiological investigations of S. lugdunensis infections only when the strains are β-lactamase-producers, since these isolates usually display a high level of genetic homogeneity. Identification of adequate typing tools for this bacterial population will probably require multimodal molecular characterization of a larger collection of S. lugdunensis strains.The standard approach to CSF shunt infections caused by a methicillin-resistant Staphylococcus spp. includes shunt removal and intravenous vancomycin [9]. The previously reported CSF infections caused by S. lugdunensis were treated with intravenous vancomycin and/or oxacillin, alone or with rifampicin, and bacteriological and clinical cures were documented in all four [6-8]. CSF shunt infections caused by methicillin-resistant staphylococci are common in our hospital, so when Gram-stain data indicated the possibility of staphylococcal ventriculitis, our patients were both treated empirically with vancomycin. Later, the isolates' methicillin resistance was confirmed by susceptibility testing and PCR analysis. Our patients were treated with intrathecal vancomycin (40 mg/day), and bacteriological and clinical cures were achieved in both cases. Since then, guidelines have been published in which considerably lower doses are recommended for children (for example < 5 mg/24 hours) [15]. Although no adverse effects were experienced by our patients, we have modified our protocol, and staphylococcal CSF shunt infections are now treated with IT vancomycin at the dosage indicated above and with drug levels closely monitored.The frequency of S. lugdunensis infections may well be underestimated. The species is likely to escape detection by screening tests or certain automated microbiological systems. Its ability to produce colony variations is well known, described in three of the 11 strains included in the initial description of S. lugdunensis in 1988 [1]. Awareness of this risk has led to the development of genetic tools for identifying this species [7]. No colony variation was observed in either of our patients and all isolates consistently exhibited the normal S. lugdunensis phenotype in blood agar cultures.ConclusionOur experience confirms that, unlike other CoNS which usually display low virulence, S. lugdunensis can cause severe CNS infections in patients with implanted CSF drainage devices. Accurate species-level identification of isolates causing staphylococcal CSF shunt infections is clearly essential for their successful treatment, but it is also fundamental for epidemiological surveillance and for improving our understanding of the pathophysiological factors affecting the clinical outcome of these infections. With the increasing use of invasive medical devices for management of neurosurgical patients, CSF shunt infections are likely to become more common. Failure to identify their causative agents can be particularly disastrous when the infection is due to S. lugdunensis.Nucleotide sequence accession numbersThe sequences of the isolates evaluated in this study have been deposited in the GenBank database under accession nos. FM177467 and FM177468, respectively, for the rpoB gene and under accession nos. FM177469 and FM177470, respectively, for the 16 rRNA gene.AbbreviationsCNS: Central nervous system; CoNS: Coagulase-negative staphylococcus; CSF: Cerebrospinal fluid; EVD: External ventriculostomy drain; IT: Intrathecal; VPS: ventriculo-peritoneal shunt; MRI: Magnetic resonance imaging; PFGE: Pulsed-field gel electrophoresis.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsTS and DR participated in the conception and design of the study, acquisition of data, analysis and interpretation of data, drafting of the manuscript and its critical revision. GT and DP managed both children as neurosurgeons. BF, TDI, BP and MS carried out the laboratory studies of patients. GF revised the manuscript. All authors read and approved the final version of the manuscript.ConsentWritten consent was obtained from each patient's parents for the publication of this report. A copy of the written consent is available for review by the Editor-in-Chief of this journal.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2529313.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2529313",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2529313\nAUTHORS: Marie-Laure Samson\n\nABSTRACT:\nBackgroundThe Drosophila gene embryonic lethal abnormal visual system (elav) is the prototype of a gene family present in all metazoans. Its members encode structurally conserved neuronal proteins with three RNA Recognition Motifs (RRM) but they paradoxically act at diverse levels of post-transcriptional regulation. In an attempt to understand the history of this family, we searched for orthologs in eleven completely sequenced genomes, including those of humans, D. melanogaster and C. elegans, for which cDNAs are available.ResultsWe analyzed 23 orthologs/paralogs of elav, and found evidence of gain/loss of gene copy number. For one set of genes, including elav itself, the coding sequences are free of introns and their products most resemble ELAV. The remaining genes show remarkable conservation of their exon organization, and their products most resemble FNE and RBP9, proteins encoded by the two elav paralogs of Drosophila. Remarkably, three of the conserved exon junctions are both close to structural elements, involved respectively in protein-RNA interactions and in the regulation of sub-cellular localization, and in the vicinity of diverse sequence variations.ConclusionThe data indicate that the essential elav gene of Drosophila is newly emerged, restricted to dipterans and of retrotransposed origin. We propose that the conserved exon junctions constitute potential sites for sequence/function modifications, and that RRM binding proteins, whose function relies upon plastic RNA-protein interactions, may have played an important role in brain evolution.\n\nBODY:\nBackgroundThe elav (embryonic lethal abnormal visual system) gene of D. melanogaster was the the first identified member of a family of neuronal RNA binding proteins that is conserved in metazoans [1,2]. The proteins in this family contain three RNA Recognition Motifs (RRM), with a hinge region separating the second and third RRMs and an optional non-conserved N-terminal region. The hinge includes signals essential for nuclear export and subcellular localization [3].RRM are common protein domains found in all life kingdoms. In humans, there are 497 genes encoding RRM containing proteins, which represent 2% of the human gene products. Proteins containing one or several of these domains are capable of interacting in a sequence specific manner with single stranded RNA molecules and of directing the assembly of multiprotein complexes [4,5]. In spite of the remarkable sequence conservation of the RRM domains, RRM-containing proteins perform numerous functions, intervening at all the possible steps of RNA metabolism. The RRM domain is composed of about 90 amino acids and contains a conserved octapeptide termed RNP-1 (ribonucleoprotein motif) and a conserved hexapeptide termed RNP-2. Structural studies indicate that four antiparallel beta-sheets form the RNA interaction surface, with RNP-1 and RNP-2 on the two inner sheets (beta 1 and beta 3). In RNA-RRM complexes, nucleotides establish contacts with residues in the RNPs, with regions in the RRM beyond the RNP domains also involved in RNA recognition. The plasticity of RRMs in their sequence-specific recognition of topologically diverse RNA is likely to be correlated with their presence in a variety of proteins involved in the diverse steps of post-transcriptional regulation.There are three elav-related genes in D. melanogaster. The elav gene encodes a nuclear product present in all neurons throughout development and is required for the differentiation of postmitotic neurons and their maintenance [1]. The rbp9 (RNA binding protein 9) product is present in neuronal nuclei starting at the third larval instar and also in the cytoplasm of cystocytes during oogenesis. Although neuronal expression is predominant, rbp9 mutations reveal a role in cystocyte proliferation and differentiation, but no neuronal defects have been reported [6,7]. The expression of fne (found in neurons) resembles elav's, but with a slightly delayed onset. FNE is cytoplasmic, but the elav and fne genes interact, suggesting protein shuttling [8,9]. The products of elav family members are essentially present in the nervous system, in all of the neurons in the case of elav itself, but more generally in subsets of neurons and/or neuroblasts and glial cells. Expression has also been detected in other tissues, in particular in testes and ovaries, or found to be ubiquitous (for instance [10]). Diverse molecular functions in the control of RNA half life, nuclear export, RNA 3' end formation, alternative RNA processing, polyadenylation and translation have been proposed for these proteins [9,11-17]. Multiple functions, both cytoplasmic and nuclear have been demonstrated for HuR, an ubiquitously expressed member of the human family [11,16,17].The evolutionary relationship between members of the family are complex. For instance, the four human proteins share 74–91% identity, while the three Drosophila proteins share only 59–68% identity. The goal of the work reported here was to investigate these relationships. We found that the elav family has an eventful evolutionary history, somewhat masked by the high level of amino acid conservation of the gene products, but revealed by analysis of the gene structure of the different family members (11 species, 23 proteins). We attribute the rapid functional evolution of the family members, as opposed to the high level of sequence conservation, to the plasticity of the RRM domains, where small changes in critical positions have the potential to modify interactions with RNA.ResultsThe paralogs fne and rbp9 share a conserved organization of their coding regions but elav, the third family member, is distinctAll three Drosophila paralogs elav, rbp9 and fne are essentially expressed in neurons. elav null mutants are embryonic lethal, while the rbp9 null mutation is viable, but surprisingly confers a female sterility phenotype. fne null mutants, although not fully characterized, are also viable (Zanini and Samson, in preparation). In order to understand the evolutionary mechanisms responsible for the generation of these paralogs, we examined their gene structure. Although their organizations are apparently quite distinct, we found remarkable conservation in the correspondance between exons and specific protein regions in rbp9 and fne (Fig. 1A). There are two differences (1) the presence of new mini-exons respectively specific for each of the two genes and (2) the use of a single exon in fne but two in rbp9 to encode the third RRM. Strikingly, this organization is totally different in the elav gene, whose complete ORF, except for the A of the ATG initiation codon, is encoded by a single exon.Figure 1Correspondance between exons and protein regions in the elav family of D. melanogaster. A: RNA structures. RNA nomenclature as in FlyBase, with details in the Methods. Boxes represent exons. The black horizontal lines are introns, with dashes respectively replacing the 5.8 kb long intron in the rbp9-RA transcript and the 2.2 kb long intron in the elav-RA transcript. White: non coding, Vertical stripes: non-conserved, Crossed: gene-specific mini-exons, respectively a 15 nucleotide long region present in alternative forms of rbp9 and a 45 nucleotide long region present in fne. All others are color coded based upon sequence similarity and according to exon-exon boundaries. B: Schematic representation of the ELAV family protein products. The color coding corresponds to that used for the RNA representation. The regions encoded by gene specific sequences have been omitted. RRM: RNA Recogntion Motif. The pairs of white vertical bars represent conserved motifs (RNP-1 and RNP-2) diagnostic of RRMs.Conserved exon junctions are present in most elav orthologsWe took advantage of the recent sequencing of complete genomes [18-22] to survey the gene family in 11 species by (1) identifying all the family members and (2) comparing the organization of the ORF in exons. In humans, D. melanogaster and C. elegans, we extracted from the databases protein sequences deduced from cDNA analyses, and aligned genomic DNA with cDNA to determine the exon-intron structure. In other cases we used the predicted protein sequences, either published or computed for our purpose, as detailed in the Methods. We examined species from the chordata (1 species), arthropoda (9 species) and nematoda (1 species), for a total of 23 genes (Fig. 2).Figure 2Exon organisation of the elav-related genes in 11 metazoans. The analyzed species are listed on the left, with classical phylogenetic relationships represented. The number of elav-like genes is listed next to the species names. Percentages of identity between their protein products and the D. melanogaster proteins ELAV, FNE and RBP9 are listed on the right side of the figure. At the top, a typical ELAV-like protein is represented, with its three RRMs and the hinge region between RRM2 and 3. The vertical arrows below point at protein regions that are, depending upon each of the 23 analyzed proteins, either encoded by exon-junctions (Jx, x = 1 to 8, see text) or by an internal exon sequence. The presence of the junction-encoded region is indicated by a vertical bar for each protein.First, we found that the size of elav families varies (one to four members) among the 11 species that we studied, with no clear relationship between family size and brain/animal complexity (Fig. 2). For instance, dipterans possess three elav genes, while the hymenopteran Apis mellifora, with ten times as many neurons as Drosophila, possesses only one gene. Levels of identity between the proteins encoded by the 23 genes are high, with the lowest score (47%) obtained in the comparison of D. melanogaster ELAV with the unique C. elegans protein. Between humans and Drosophila, there is 54–64% amino acid identity in the ELAV-related proteins, 38% identity for the arginase proteins (ubiquitous metabolic enzymes, see below) and 33% identity for the engrailed proteins (conserved transcription factors, not shown). The levels of ELAV-related protein identity are thus remarkable. The crystal structure of the first two RRM of human HuD associated with cfos RNA, identifies 12 amino acids whose side chain is making direct RNA contacts [23]. These residues are conserved in all 23 ELAV-like proteins that we examined, except for the arginine in RNP1 of the second RRM, which appears to be specific to the human proteins and to one of the B. mori ELAV-like, Bm-2. In the other species there is a conserved substitution by a lysine.Second, we found remarkable conservation of exon structure. From vertebrates to invertebrates, we identified eight exon junctions in the RRMs/hinge region (Fig. 3). We named them J1 to J8, from the most upstream to the most downstream. All are present in several phyla, except for J1 and J4 which are specific for FNE and RBP9 from Drosophila and are implicated in the generation of mini-exons in the sequence coding (alternative forms of) these proteins (Fig. 1 &3). Overall, the J2 junctions (respectively J3, J5 and J8, Fig. 2 &3) are unambiguously homologous since (1) the level of protein sequence conservation is such that the amino acid positions where the junctions intervene are clearly aligned and conserved (Fig. 3) and (2) nucleotide sequence analysis shows that at a given exon junction, the splice is at the same position in the codons: specifically between the first and the second bases of the spliced codons (for J2 and J5, as well as for the species-specific J1 and J4) or exactly between codons (J3 and J8). There are two exceptions to this strict conservation. First, J5 is interrupted in rbp9 of D. melanogaster by the intronic insertion of an alternative mini-exon, without alteration of the J5 5' or 3' splice sites. Second, in fne, J2 is split by the intronic insertion of a mini-exon, the J2 donor splice site is additionally shifted downstream while the J2 acceptor splice site is maintained (Fig. 3). Interestingly, the junctions J2 and J5 occupy the same position relative to RNP-1 in RRM1 and RRM2.Figure 3Protein sequence comparison among 27 ELAV-like proteins forms. Alternative protein forms are included, specifically for Drosophila RBP9 (A and D) and three of the human proteins (HuB, HuC and HuD, where HuX-n refers to the n amino acid long form of the HuX protein). \"*\" indicate that amino acids are identical in all 27 sequences, \":\" and \".\" respectively indicate conserved and semi-conserved substitutions. The octamer RNP-1 and the hexamer RNP-2, diagnostic of RRMs, are underlined. Also underlined is a conserved octamer present in the region that is crucial for nuclear export and localization. The regions in light grey boxes have been mapped as necessary for these processes in D. melanogaster ELAV, human HuR and human HuD. We identified eight exon junctions labelled J1 to J8 (see text). Bold characters and dark grey boxes are used to identify amino acids encoded by exon junctions. When the splicing connects intact codons, two amino acids are bold (J3 and J8). The symbol//replaces 85 non-conserved amino acids in the C. elegans sequence.The junctions J6 and J7 map in a moderately conserved coding region, essential for nuclear export and proper subcellular localization (Reviewed in [2]), including only a conserved hexamer (R-SP----). Both J6 and J7 split the spliced codons between the second and the third bases. In this region, three types of events affecting the splicing seem to have occured independently: 1) the introduction of a mini-exon (in humans), that can be alternatively spliced (HuB), (2) the shift of the 5' splice site (example: N. vitripennis vs T. castaneum) (3) the shift of the 3' splice (example: the T. castaneum vs Ae-2 genes or the alternative human forms HuD-366 and HuD-380). Noticeably, the regions close respectively to J1/J2, J4/J5 and J6/J7 as well as the entire hinge region between RRM2 and RRM3 appear more variable than the rest of the protein.Intronless elav-like genes are present in Diptera and LepidopteraInterestingly, for six of the analyzed genes (Ag-1, Ae-1, Cp-1, elav in Diptera, and Bm-1, Bm-2 in Lepidoptera), the entire conserved region of the protein is encoded by a single exon. Based upon both their gene structure and the level of protein sequence identity, the dipteran intronless genes constitute a homogeneous elav-type group. In contrast, although intronless like elav, the B. mori genes encode proteins more similar to FNE/RBP9 than to ELAV. This observation suggests that distinct evolutionary forces shaped the B. mori genes and the dipteran elav-like intronless-genes, respectively. To evaluate this hypothesis, we performed a phylogenetic analysis of the 27 ELAV orthologs/paralogs, using the UPGMA algorithm, with bootstrap analysis (Fig. 4). This analysis shows with high confidence (bootstrap values greater or equal to 99%) that in dipterans, the proteins encoded by the intronless genes (Ag-1, Ae-1, Cp-1, elav) cluster together, while the two B. mori genes products cluster with the FNE/RBP9 sequences Similar results were obtained when performing sequence alignments using the neighbor joining method (not shown).Figure 4Phylogenetic tree of 27 ELAV-like proteins. Sequences were aligned and bootstrapped 500 times. Numbers near the branches are the bootstrap values, and the scale indicates the number of substitutions per site.Because the D. melanogaster elav gene is nested in the third intron of the arginase gene [24], we probed the gene environment of the intronless elav orthologs that we report here. We found that the nested elav/arg gene organization is unique to Drosophila, specifically D. melanogaster and 11 additional Drosophila species whose genomes have recently been sequenced [25] (Fig. 5). In the 10 other non-Drosophila species examined here, there is no close linkage between the arginase gene(s) and the elav gene family members. In particular, the mosquitos, similar to D. melanogaster, each have three elav-like genes, including one intronless version, but unlike D. melanogaster they have an intronless arg, which obviously rules out the possibility of a nested gene. In B. mori, although the two intronless Bm-1 and Bm-2 genes map at loci distinct from the arg locus, an intron putatively homologous to the third intron of the D. melanogaster arginase gene is present (Fig. 5).Figure 5A unique nested gene arrangement for the elav and arginase genes in D. melanogaster. A: The elav gene is nested inside the third intron of the arginase gene. Complementary strands are transcribed to generate the elav and arg RNAs with inverse polarities [28]. B: Examination of the relative arg-elav arrangement in 11 metazoans. There are two arginase genes in humans, only one in the other examined species. Column 1 documents the status of the arginase third intron. Column 2 specifies the nested (+) or independent (-) arrangement of the arginase/elav genes. N.A.: Not applicable. The third column indicates the percentage of amino-acid sequence identity of D. melanogaster compared with other species. *: N-terminally truncated arginase sequence for P. humanus corporis. See Additional file 3 for arginase alignments.DiscussionThe D. melanogaster gene elav is specific to the dipteran phylum and results from retrotranspositionThe elav gene from Drosophila was the first identified member of this family, is considered as its prototype [1], and most of the subsequently discovered orthologs are named after it. However, the present analysis highlights unique characteristics of this gene that suggest it is of recent evolutionary origin, after the separation of dipterans and lepidopterans. Aside from elav, only the dipteran genes Ae-1, Ag-1 and Cp-1 encode proteins that are more similar to ELAV than to FNE and RBP9. In addition to the intronless elav-likes, dipteran genomes carry two genes encoding proteins of the type FNE/RBP9, also found in the seven other genomes analyzed. Thus elav, Ae-1, Ag-1 and Cp-1 represent a newly evolved gene form specific to dipterans.In addition, the elav gene structure is suggestive of retrotransposition, a process considered significant in the evolution of genomes, including Drosophila [26]. The genes Ae-1, Ag-1 and Cp-1 from mosquitoes share with elav not only a higer level of similarity between their products, but also the property of having their ORF in a single exon. The absence of introns (restricted to dipterans and B. mori in this gene family) is atypical: we identified conserved exon junctions that are a landmark present in most of the elav-related genes. Furthermore, the elav gene of Drosophila is nested in the arginase gene. In humans, retrotransposition is an important contributor to the generation of nested genes [27]. We thus propose that elav originated from a recent retrotransposition event. It is possible that the same retrotransposition is at the origin of both the lepidopteran intronless fne/rbp9-like genes and the dipteran elav-like genes. A duplication of the retrotransposed gene in the ancestor to B. mori and different fates for the ancestral gene copies in the two groups would bring about the present situation. Alternatively, we do not exclude that independent retrotranspositions happened in lepidopteran and dipteran ancestral lineages.Interestingly, the nested arg/elav arrangement found in D. melanogaster is not conserved in the mosquitoes, where the host gene (arginase) became intronless. This parallels the nested arrangement of the intronless sina gene in an intron of the Rh4 gene, as found in mosquitoes and nine species of the Drosophila genus. The remaining three species of the genus have an intronless Rh4, with a loss of the ancestral Rh4 copy where sina was originally embedded [28]. These situations show the lability of nested gene arrangements.elav: the genesis of a new functionIt was unexpected to find that the copy number of elav family members varied from species to species. Given the maintenance of this gene family in all metazoans, we assume that there is a function for at least one, if not all, of the genes in each species. Mutants have been reported in only three species. The knockout of neuronal HuD in mice causes motor and sensory defects [29]. It is not excluded that the mild phenotype of this mutant is the consequence of gene redundancy. In C. elegans, cholinergic synaptic transmission is altered in mutants of the single elav ortholog EXC-7, which is expressed in a subset of neurons and other non-neuronal cells [30]. In both cases, viability and apparent morphology are normal. In Drosophila melanogaster, the vital gene elav is required in all neurons [1], whereas rbp9 is essential for female fertility [7] but does not affect viability. We recently generated null mutations of the fne gene (Zanini and Samson, in preparation), whose preliminary analysis indicates that they are viable in adults and lead to no apparent morphological defects. Aside from elav itself, characterized mutations of the elav gene family are viable, suggesting a non-vital function of the ancestral gene.Considering that elav appears to be a new member of the family, its vital function is quite striking. This situation is reminescent of that of Sex-lethal (Sxl), a gene fundamental to sex determination in Drosophila, but which does not act as a sex determining factor in non-Drosophilids. The Drosophilid genomes indeed contain two Sxl paralogs (79% identity in D. melanogaster), while non-Drosophilids have one. It has been proposed that there was a duplication of the ancestral gene in Drosophilids and acquisition of a new function by one of the copies [31]. We believe that a retrotransposition of the elav/fne/rbp9 ancestor gene at the time of the separation of dipterans/lepidopterans led to a gene duplication and the evolution of a new function for elav.Conserved RNA binding proteins: a reservoir for accelerated functional evolutionWe have pointed out that the ELAV-like proteins, including ELAV itself, have maintained a high level of sequence conservation between species, higher than that of engrailed, a conserved transcription factor with a homeodomain, or that of arginase, a ubiquitous metabolic enzyme that arose before the divergence of procaryotes and eucaryotes. This is intriguing in light of the extensively documented diversity of the properties of individual members of the family. First, although there is expression in the nervous system of at least one of the elav family members in every investigated metazoan (mammals, fishes, amphibians, birds, amphioxus, C. elegans, D. melanogaster), expression is also detected in other tissues and is even sometimes ubiquitous [2]. Second, the functions of these proteins are multiple, whether at the cellular level, where they include cell differentiation/survival [1,6,29,32] and cell proliferation/control of the cell cycle [7,33] or at the biological level, with impacts on motor/sensory activity, memory, fertility or viability [1,6,29,34]. Finally, the apparent subcellular localization of these proteins is diverse (nuclear, subnuclear, cytoplasmic or both), in agreement with diverse molecular functions [2,3].The data thus reveal a diversification of the functions and of the specificity of expression of ELAV family members and implies a diversification of the interactions with other macromolecules, most evidently the RNAs whose metabolism is regulated by the RRM containing proteins. The DNA duplications and retrotranspositions that occured in the elav gene families constitute a starting point for the diversification of gene function. Changes in cell or tissue specificity of expression are often linked to modifications of non-translated regulatory regions. However, changes affecting the sub-cellular localization, known to be dependent upon the hinge region between RRM2 and RRM3, or changes in the interactions with proteins or RNA must depend upon the protein product of the elav-like genes.Sequence alignments of the ELAV-like proteins shows that they are overall very conserved. But we were puzzeld by the fact some of the conserved exon junctions (J1/J2, J4/J5 and J6/J7) are adjacent to sequences that are among the most variable of the proteins. They include short insertions of amino acids, (alternative) exon addition and amino acid variations. The intron sequence indeed provides a potential source of sequence variability: it is conceivable that intron extremities become integrated into coding sequences by shifting of the exon boundaries. Alternatively, the intron can serve as the site of insertion of a new exon. An additional surprising point was the fact that these variable micro regions are almost directly upstream of important conserved motifs, specifically RNP-1 (in RRM1 and RRM2) and the octapeptide in the region essential for nuclear export and subcellular localization. The modification of residues outside of the RNP has the potential to alter the interactions between the RRM and an RNA [5]. Additionally, alterations of the region responsible for nuclear export/cellular localization modify this function (reviewed in [2]). We thus propose that the maintenance of the exon junctions is vital to the evolution of the ELAV family, in particular the generation of new functions. As a consequence, one would predict that RRM1, RRM2 and the hinge region have prominent roles in functional specificity. It may be significant in this respect that RRM3 replacements in ELAV by RRM3 from RBP9 or HUD are fully functional, while RRM1 or RRM2 replacements by corresponding RRMs from RBP9 or SXL are largely non-functional [35].More generally, it seems that RRM-containing proteins could serve as favorable targets for the rapid evolution of gene functions. Because of the structural versatility of the RRM domain, it can be adapted for sequence specific recognition of many different nucleic acid structures and different protein partners [5]. The SXL protein, a crucial regulator of sex determination in Drosophila contains 2 RRM, and appears to be the result of such a rapid adaptation of function. In the search for genetic changes that distinguish our brains from that of our ancestors, the focus has been on the identification of non-synonymous changes in coding regions and the modification of regulatory sequences [36]. Our work suggests that the very conserved RRM-containing proteins may have contributed to human brain evolution, especially when considering the fundamental importance of the regulation of RNA metabolism in neurons, where alternative splicing [37] and localized RNA translation and degradation [38,39] take place with impacts on cortex development, neuronal regeneration and plasticity.ConclusionThe elav gene family encodes proteins with three RNA Recognition Motifs (RRM) acting as neuronal post-transcriptional regulators in all metazoans. Since they show remarkable sequence conservation, the documented diversity of their molecular roles is unexpected. We report the occurence of elav-like gene duplications and deletions in metazoans, and show that the vital elav gene of Drosophila is newly emerged, specific to dipterans and of retrotransposed origin, challenging its status of prototype for the family. These findings, together with the plasticity of the interactions between RRM and RNA, suggests that the elav-like proteins may have played an important role in the evolution of the gene functions crucial in brain evolution.MethodscDNA sequences used for the analysis of coding sequence organization in the elav gene family of Drosophila melanogasterWe used the transcripts data from FlyBase [18] to assess the relationship between RNA and protein coding regions. Multiple RNA isoforms from one gene were taken into account if they were a source of polypeptide diversity. For instance, seven alternative RNA forms have been reported for rbp9, which are predicted to encode six distinct polypeptides. Only one level of variation was relevant to the present analysis, that is the alternative inclusion of a mini-exon that causes the addition of 15 nucleotides (five amino acids), hence the choice of using the rbp9-A and the rbp9-D RNA forms, that differ by the presence/absence of the mini exon. In the case of both fne and elav, several transcripts have been reported but they encode a single polypeptide.Identification of elav orthologs in completely sequenced genomes and prediction of ELAV-like protein sequencesWe used protein sequences from the data bases deduced from cDNA analysis whenever possible, with NCBI accession numbers as follows: in humans BAD92531 (HuB, 367 amino acids), AAH30692/Q12926-2 (HuB, 346 aa), AAA58677 (HuC, 359 AA), AAH14144/Q14576 (HuC, 367 AA), AAH36071/Q8IYD4 (HuD, 366 aa), AAK57541/AAK57541 (HuD, 380 aa), AAH03376/Q15717 (HuR, 326 aa), in D. melanogaster AAA28506 (ELAV, 483 aa), AAF43091 (FNE, 356 aa), AAF51179 (RBP9 isoform A, 647 aa) and AAN10401 (RBP9 isoform D, 642 aa), in Caenorhabditis elegans NP_496057 (EXC-7, 456 aa). UniProtKB/Swiss-Prot Accession numbers are also provided for further details on the proteins: Q12926 (HuB), Q14576 (HuC), Q8IYD4 (HuD), Q15717 (HuR), P16914 (ELAV,), Q9VYI0 (FNE), Q9VQJ0 (RBP9) and Q20084 (EXC-7).When no cDNA sequences were available, we performed searches of the entire genomes using the tblastn program [40] to identify orthologs of ELAV-related genes. We analyzed the genomic regions encoding these orthologs by performing a three frame translation of the genomic sequences, and using the gene prediction program genescan [41] as well as a splice site prediction program [42]. The predicted protein coding sequences were the result of integration and manual review of these data.Using the procedures detailed above to identify elav orthologs, we reviewed predicted protein sequences that have been proposed for Apis mellifora, Aedes aegypti and Anopheles gambiae [19]. Some of our conclusions were consistent with the automated predictions of genome projects (A. mellifora, XP_394166, 343 aa), but we edited sequences of A. aegypti, and A. gambiae ELAV orthologs. The decision of editing was based upon the identification of manifest errors in the automated predictions, such as the prediction of a four base pair intron 5'-CCCT-3', missing the consensus GT-AG sequences typically flanking introns for the Ag-3 predicted transcript (XM_309157). For those two species, as well as for those where no prediction had yet been proposed, we relied upon the above procedure to identify and propose predicted sequences of ELAV orthologs. They respectively derive from genomic sequences CH477489 (Ae-1), CH477672 (Ae-2), CH477401(Ae-3) in A. aegypti, from CM000357 (Ag-1), CM000360 (Ag-2), CM000359 (Ag-3) in A. Gambiae, DS231997 (Cp-1), DS232556 (Cp-2), DS231816 (Cp-3) in Culex pipiens, CM000276 in Tribolium castaneum, DS265619 in Nasonia vitripennis, AADK01020611 (Bm-1), CH391062 (Bm-2) in Bombyx mori and DS235033 in Pediculus humanus corporis.In our analysis we used only the approximately 325 amino acids region of the proteins including the three RRM and a hinge region that links RRM2 and RRM3, because the N-terminus, when present, is not conserved. The sequences used are listed in Additional file 1.Identification of arginase genes in completely sequenced genomes and prediction of arginase protein sequencesArginase sequences have been deduced from cDNA sequences for several species: human (ARG1: P05089, Arg2: P78540), D. melanogaster (Q9NHA5), C. elegans (Q22659). For the other species, we used the procedure described above to propose arginase sequences. The protein sequences derive from genomic sequences CH477248 in A. aegypti, from CM000359 in A. gambiae, DS232533 in C. pipiens, CM000280 in T. castaneum, DS265617 in N. vitripennis, CH389642 in B. mori and DS235286 in P. humanus corporis. We were not able to predict a complete P. humanus corporis arginase sequence, because of the lower level of conservation. See Additional file 2 for the arginase sequences.Protein sequence alignments and percentages of identityAlignments were performed with the ClustalW program using default parameters [43]. In the case of arginases, we focused on the region homologous to that including intron 3 in D. melanogaster. The values for percentages of identity were extracted from the ClustalW score tables.Phylogenetic analysisWe used the CLC combined workbench (CLC bio A/S) version 3.6.2 to align the 27 protein sequences with an unweighted pair group method using arithmetic averages (UPGMA) and to evaluate the reliability of the inferred tree with a bootstrap analysis (500 replicates).Authors' contributionsThe author takes full responsability for the work. She asked the question, devised the approach, performed it, analyzed the results and wrote the manuscript.Supplementary MaterialAdditional file 1Fasta sequences of the three RRMs and the hinge regions of ELAV-like proteins. 27 Fasta sequences.Click here for fileAdditional file 2Fasta sequences of the arginases. 12 Fasta sequences.Click here for fileAdditional file 3Protein sequence comparison among 12 arginases from 11 metazoans. Arginase sequences alignment with legend.Click here for file\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2529317.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2529317",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2529317\nAUTHORS: Yingfeng Luo, Xiaoli Xu, Zonghui Ding, Zhen Liu, Bing Zhang, Zhiyu Yan, Jie Sun, Songnian Hu, Xun Hu\n\nABSTRACT:\nBackgroundPhenylobacterium zucineum is a recently identified facultative intracellular species isolated from the human leukemia cell line K562. Unlike the known intracellular pathogens, P. zucineum maintains a stable association with its host cell without affecting the growth and morphology of the latter.ResultsHere, we report the whole genome sequence of the type strain HLK1T. The genome consists of a circular chromosome (3,996,255 bp) and a circular plasmid (382,976 bp). It encodes 3,861 putative proteins, 42 tRNAs, and a 16S-23S-5S rRNA operon. Comparative genomic analysis revealed that it is phylogenetically closest to Caulobacter crescentus, a model species for cell cycle research. Notably, P. zucineum has a gene that is strikingly similar, both structurally and functionally, to the cell cycle master regulator CtrA of C. crescentus, and most of the genes directly regulated by CtrA in the latter have orthologs in the former.ConclusionThis work presents the first complete bacterial genome in the genus Phenylobacterium. Comparative genomic analysis indicated that the CtrA regulon is well conserved between C. crescentus and P. zucineum.\n\nBODY:\nBackgroundPhenylobacterium zucineum strain HLK1T is a facultative intracellular microbe recently identified by us [1]. It is a rod-shaped Gram-negative bacterium 0.3–0.5 × 0.5–2 μm in size. It belongs to the genus Phenylobacterium [2], which presently comprises 5 species, P. lituiforme (FaiI3T) [3], P. falsum (AC49T) [4], P. immobile (ET) [2], P. koreense (Slu-01T) [5], and P. zucineum (HLK1T) [1]. They were isolated from subsurface aquifer, alkaline groundwater, soil, activated sludge from a wastewater treatment plant, and the human leukemia cell line K562, respectively. Except for P. zucineum, they are environmental bacteria, and there is no evidence that these microbes are associated with eukaryotic cells. The HLK1T strain, therefore, represents the only species so far in the genus Phenylobacterium that can infect and survive in human cells. Since most, if not all, of the known microbes that can invade human cells are pathogenic, we proposed that HLK1T may have pathogenic relevance to humans [1]. Unlike the known intracellular pathogens that undergo a cycle involving invasion, overgrowth, and disruption of the host cells, and repeating the cycle by invading new cells, HLK1T is able to establish a stable parasitic association with its host, i.e., the strain does not overgrow intracellularly to kill the host, and the host cells carry them to their progeny. One cell line (SW480) infected with P. zucineum has been stably maintained for nearly three years in our lab (data not shown).In this report, we present the complete genome sequence of P. zucineum.ResultsGenome anatomyThe genome is composed of a circular chromosome (3,996,255 bp) and a circular plasmid (382,976 bp) (Figure 1; Table 1). The G + C contents of chromosome and plasmid are 71.35% and 68.5%, respectively. There are 3,861 putative protein-coding genes (3,534 in the chromosome and 327 in the plasmid), of which 3,180 have significant matches in the non-redundant protein database. Of the matches, 585 are conserved hypothetical proteins and 2,595 are proteins with known or predicted functions. Forty-two tRNA genes and one 16S-23S-5S rRNA operon were identified in the chromosome.Table 1Genome summary of P. zucineum Strain HLK1TGenomic ElementChromosomeplasmidLength (bp)3,996,255382,976GC content (%)71.3568.54Proteins3, 534327Coding region of genome (%)88.85%81.94%Proteins with known or predicted function2,394(67.75%)201(61.47%)Conserved hypothetical proteins560(15.84%)25(7.65%)Hypothetical proteins580(16.41%)101(30.88%)rRNA operon10tRNAs420Proteins in each[J] Translation, ribosomal structure and biogenesis185 (5.24%)3 (1.21%)COG category[K] Transcription210 (5.94%)22 (8.91%)[L] Replication, recombination and repair139 (3.93%)23 (9.31%)[D] Cell cycle control, cell division, chromosome partitioning27 (0.76%)0[V] Defense mechanisms51 (1.44%)3 (1.21%)[T] Signal transduction mechanisms166 (4.7%)24 (9.72%)[M] Cell wall/membrane/envelope biogenesis195 (5.52%)15 (6.07%)[N] Cell motility62 (1.75%)4 (1.62%)[U] Intracellular trafficking, secretion, and vesicular transport96 (2.72%)13 (5.26%)[O] Posttranslational modification, protein turnover, chaperones151 (4.27%)32 (12.96%)[C] Energy production and conversion188 (5.32%)16 (6.48%)[G] Carbohydrate transport and metabolism161 (4.56%)15 (6.07%)[E] Amino acid transport and metabolism293 (8.29%)5 (2.02%)[F] Nucleotide transport and metabolism58 (1.64%)3 (1.21%)[H] Coenzyme transport and metabolism116 (3.28%)3 (1.21%)[I] Lipid transport and metabolism215 (6.09%)12 (4.86%)[P] Inorganic ion transport and metabolism223 (6.31%)24 (9.72%)[Q] Secondary metabolites biosynthesis, transport and catabolism152(4.3%)9 (3.64%)[R] General function prediction only444 (12.57%)28 (11.34%)[S] Function unknown307 (8.69%)20 (8.10%)Figure 1Circular representation of the P. zucineum strain HLK1T chromosome and plasmid (smaller circle). Circles indicate (from the outside): (1) Physical map scaled in megabases from base 1, the start of the putative replication origin. (2) Coding sequences transcribed in the clockwise direction are color-coded according to COG functional category. (3) Coding sequences transcribed in the counterclockwise direction are color-coded according to COG functional category. (4) Proteins involved in establishment of intracellular niche are TonB-dependent receptors (orange) and pilus genes (sienna). (5) Functional elements responsible for environmental transition are extracytoplasmic function sigma factors (royal blue), transcriptional regulators (violet red), two-component signal transduction proteins (deep sky blue), heat shock molecular chaperons (spring green), type IV secretion systems (plum), chemotaxis systems (green yellow) and flagellum proteins (gray). (6) G + C percent content (10-kb window and 1-kb incremental shift for chromosome; 300 bp window and 150 bp for incremental shift for plasmid); values larger than average (71.35% in chromosome and 68.5% in plasmid) are in red and smaller in medium blue. (7) GC skew (10-kb window and 1-kb incremental shift for chromosome; 300 bp window and 150 bp for incremental shift for plasmid); values greater than zero are in gold and smaller in purple. (8) Repeat families, repeats 01-08 are in dark salmon, dark red, wheat, tomato, light green, salmon, dark blue and gold, respectively.There are 7 families of protein-coding repetitive sequences and a family of noncoding repeats in the genome (Table 2). Notably, identical copies of repeats 02–04 were found in both the chromosome and the plasmid, suggesting their potential involvement in homologous recombination.Table 2Repetitive elements in the P. zucineum genomeRepeat IDLength bpDR1Number of copiesPosition of insertionIdentity (%)Coding informationComplete2PartialChromosomePlasmidRepeat0132,58773104>99TransposaseRepeat0241,26233122100TransposaseRepeat0351,392NA4242100TransposaseRepeat0461,257NA10073100TransposaseRepeat051,554NA2020>98Hypothetical proteinRepeat061,136NA2020>90Isovaleryl-CoA dehydrogenaseRepeat071,077NA2020>982-nitropropane dioxygenaseRepeat08≈130NA130130>90Noncoding repeats1Size in base pairs of the consensus and the direct repeat (DR) generated by insertion into the genome target site.2A copy is complete if the length of the repeat is ≥ 90% of the consensus, otherwise, the copy is partial.3One complete copy which harbors a 7 bp direct repeat (TCCTAAC) that disrupts the VirD4.4The partial copy is located in the plasmid.5Two partial copies are located in the chromosome, of which a \"partial\" copy with full length is inserted by a copy of repeat04.6repeats 01–04 are IS elementsOn the basis of COG (Cluster of Orthologous Groups) classification, the chromosome is enriched in genes for basic metabolism, such as categories E (amino acid transport and metabolism) and I (lipid transport and metabolism), accounting for 8.29% and 6.09% of the total genes in the chromosome, respectively. On the other hand, the plasmid is enriched for genes in categories O (posttranslational modification, protein turnover, chaperones) and T (signal transduction mechanisms), constituting 12.96% and 9.72% of the total genes in the plasmid, respectively.As to genes in the plasmid that cope with environmental stimuli, about half of the genes in category O are molecular chaperones (17/32), including 2 dnaJ-like molecular chaperones, 2 clusters of dnaK and its co-chaperonin grpE (PHZ_p0053-0054 and PHZ_p0121-122), a cluster of groEL and its co-chaperonin groES (PHZ_p0095-0096), and 9 heat shock proteins Hsp20. Of 23 genes in category T, there is one cluster (FixLJ, PHZ_p0187-0188), which is essential for the growth of C. crescentus under hypoxic conditions [6].General metabolismThe enzyme sets of glycolysis and the Entner-Doudoroff pathway are complete in the genome. All genes comprising the pentose phosphate pathway except gluconate kinase were identified, consistent with our previous experimental result that the strain cannot utilize gluconate [1]. The genome lacks two enzymes (kdh, alpha ketoglutarate dehydrogenase and kgd, alpha ketoglutarate decarboxylase), making the oxidative and reductive branches of the tricarboxylic acid cycle operate separately. The genome has all the genes for the synthesis of fatty acids, 20 amino acids, and corresponding tRNAs. Although full sets of genes for the biosynthesis of purine and pyrimidine were identified, enzymes for the salvage pathways of purine (apt, adenine phosphoribosyltransferase; ade, adenine deaminase) and pyrimidine (cdd, cytidine deaminase; codA, cytosine deaminase; tdk, thymidine kinase; deoA, thymidine phosphorylase; upp, uracil phosphoribosyltransferase; udk, uridine kinase; and udp, uridine phosphorylase) were absent. The plasmid encodes some metabolic enzymes, such as those participating in glycolysis, the pentose phosphate pathway, and the citric acid cycle. However, it is worth noting that the plasmid has a gene (6-phosphogluconate dehydrogenase) that is the only copy in the genome (PHZ_p0183).Like most other species in the genus Phenylobacterium, the strain is able to use L-phenylalanine as a sole carbon source under aerobic conditions [1]. A recent study revealed that phenylalanine can be completely degraded through the homogentisate pathway in Pseudomonas putida U [7]. P. zucineum may use the same strategy to utilize phenylalanine, because all the enzymes for the conversion of phenylalanine through intermediate homogentisate to the final products fumarate and acetoacetate are present in the chromosome (Table 3).Table 3Phenylalanine-degrading enzymes in the P. zucineum genomeGeneP. zucineum LocusLength (bp)Alignment coverage (%)ScoreAmino acid Identity (%)Gene nameP. putidaP. zucineumP. putidaP. zucineumphhAPHZ_c140926230883.5971.7521948.65phenylalanine-4-hydroxylasephhBPHZ_c00771189779.6693.8138.526.32carbinolamine dehydratasetryBPHZ_c164439840660.0557.3933.921.86tyrosine aminotransferasehpdPHZ_c283335837498.3293.5839857.984-hydroxyphenylpyruvate dioxygenasehmgAPHZ_c283143337760.2867.6453.522.3homogentisate 1,2-dioxygenasehmgBPHZ_c03134302269.7718.1427.739.53fumarylacetoacetate hydrolasehmgCPHZ_c031421021298.198.1121351.67maleylacetoacetate isomeraseFunctional elements responding to environmental transitionHLK1T is able to survive intracellularly and extracellularly. Consistently, the genome contains the fundamental elements to support the life cycle in different environments. The genome contains abundant two-component signal transduction proteins, transcriptional regulators, and heat shock response proteins, enabling the strain to respond to extra- and intra-cellular stimuli at transcriptional and post-translational levels. Among the total of 102 two-component signal transduction proteins (91 in the chromosome and 11 in the plasmid), there are 36 histidine kinases, 48 response regulators, and 18 hybrid proteins fused with histidine kinase and response regulator. Sixteen pairs of histidine kinase and response regulator (1 in the plasmid) are adjacently aligned and may act as functional operons. These tightly linked modules make two-component signal transduction systems respond to environmental changes efficiently. The genome encodes 170 transcriptional regulators (16 in the plasmid) (Table 4). Notably, we annotated the proteins of 93 bacteria (see methods – comparative genomics) with the same annotation criteria used for P. zucineum and found that the fraction of two-component signal transduction proteins and transcriptional regulators was positively correlated with the capacity for environmental adaptation (Figure 2). The genome contains 17 extracytoplasmic function (ECF) sigma factors (3 in the plasmid) (Table 5). ECFs are suggested to play a role in environmental adaptation for Pseudomonas putida KT2440, whose genome contains 19 ECFs [8]. P. zucineum has 3 heat shock sigma factors rpoH (2 in the plasmid) and 33 heat shock molecular chaperons (17 in the plasmid) (Table 6), which can cope with a variety of stresses, including cellular energy depletion, extreme concentrations of heavy metals, and various toxic substances. [9].Table 4Transcriptional regulators in the P. zucineum genomeFamily nameAction typeChromosomePlasmidProposed rolesAsnC familyActivator/repressor80Amino acid biosynthesisAraC familyActivator101Carbon metabolism, stress response and pathogenesisArsR familyRepressor80Metal resistanceBlaI familyRepressor20Penicillin resistanceCold shock familyActivator60Low-temperature resistanceCro/CI familyRepressor92Unknown2Crp/Fnr familyActivator/repressor72Global responses, catabolite repression and anaerobiosisGntR familyRepressor70General metabolismLacI familyRepressor40Carbon source utilizationLuxR familyActivator51Quorum sensing, biosynthesis and metabolism, etc.LysR familyActivator/repressor151Carbon and nitrogen metabolismMarR familyActivator/repressor60Multiple antibiotic resistanceMerR familyRepressor92Resistance and detoxificationTetR familyRepressor220Biosynthesis of antibiotics, efflux pumps, osmotic stress, etc.XRE familyRepressor22Unknown (initial function is lysogeny maintenance)Other types2-345-Total-15416-1Initial function is related to controlling the expression of phage gene2\"Other types\" include the transcriptional regulators with only one member in the P. zucineum genome or transcriptional regulators that could not be classified into any known family.Table 5Extracytoplasmic function (ECF) sigma factors in the P. zucineum genomeLocusLocation of proteinsCOG categoryGenomic element5'-end3'-endPHZ_p0151Plasmid171,032170,316COG15951PHZ_p0174Plasmid208,703208,053COG1595PHZ_p0192Plasmid229,133228,516COG1595PHZ_c0249Chromosome249,840250,553COG1595PHZ_c0301Chromosome296,299295,706COG1595PHZ_c1475Chromosome1,676,9201,677,492COG1595PHZ_c1529Chromosome1,730,7831,731,403COG1595PHZ_c1531Chromosome1,732,2191,732,800COG1595PHZ_c1907Chromosome2,134,9712,135,507COG1595PHZ_c2171Chromosome2,447,5812,448,396COG1595PHZ_c2233Chromosome2,526,8362,527,369COG1595PHZ_c2394Chromosome2,724,7592,725,307COG1595PHZ_c2577Chromosome2,965,2502,964,390COG1595PHZ_c2585Chromosome2,970,3682,969,811COG1595PHZ_c2684Chromosome3,077,2723,076,727COG1595PHZ_c0569Chromosome605,441604,233COG49412PHZ_c3154Chromosome3,582,0103,583,269COG49411COG1595, DNA-directed RNA polymerase specialized sigma subunit, sigma24 homolog;2COG4941, predicted RNA polymerase sigma factor containing a TPR repeat domainTable 6Distribution of heat shock related proteins in P. zucineum and representative alphaproteobacteria with different living habitatsContent\\SpeciesS. melilotiB. suisC. crescentusP. zucineumR. conoriiG. oxydansChromosomePlasmidrpoH, heat shock sigma factor12211211dnaK, molecular chaperone2 (Hsp70)1111211grpE, molecular chaperone (co-chaperonin of Hsp70)1111211dnaK-like molecular chaperone1111011dnaJ, molecular chaperone1111011dnaJ-like molecular chaperone4336213groEL, molecular chaperone (hsp60)5111111groES, molecular chaperone (Hsp10, co-chaperonin of Hsp60)3111111molecular chaperone Hsp205223903molecular chaperone Hsp3311110011rpoH may be responsible for the expression of some or all heat shock proteins2The function of molecular chaperones is to protect unfolded proteins induced by stress factors through renaturation or degradation in cooperation with protease.Figure 2Comparative analysis of transcriptional regulators and two-component signal transduction proteins in 6 groups of bacteria classified according to their habitats. (A): The mean number of transcriptional regulators in each megabase pair of the genomes. (B): The mean number of two-component signal transduction proteins in each megabase pair of the genomes. The fraction of transcriptional regulators and two-component signal transduction proteins (solid black circle) of P. zucineum were 41.56 genes/Mb and 23.30 genes/Mb, respectively. Error bars represent standard errors. O: Obligate (26 species), S: Specialized (5 species), AQ: Aquatic (4 species), F: Facultative (28 species), M: Multiple (27 species), T: Terrestrial (3 species).The genes for cell motility include 3 chemotaxis operons, 7 MCP (methyl-accepting chemotaxis) genes, 15 other genes related to chemotaxis (Table 7), and 43 genes for the biogenesis of the flagellum (Table 8).Table 7Chemotaxis proteins in the P. zucineum genomeLocus P. zucineum5'-end3'-endNameOrthologs C. crescentusOperonBest BLAST matchPHZ_c0690753,270753,812chemotaxis protein CheW-1M. magneticum AMB-1PHZ_c0691753,812755,218chemotaxis protein methyltransferase CheR-1M. magnetotacticum MS-1PHZ_c0692755,240755,836chemotaxis signal transduction protein-1Rhodospirillum centenumPHZ_c0693755,836757,488methyl-accepting chemotaxis protein-1M. magneticum AMB-1PHZ_c0694757,501759,642chemotaxis histidine kinase CheA-1M. magnetotacticum MS-1PHZ_c0695759,642760,709chemotaxis response regulator CheB-1Rhodospirillum centenumPHZ_c32303,661,5143,661,050CheE protein-2C. crescentus CB15PHZ_c32313,662,0993,661,527chemotaxis protein CheYIIICC04402C. crescentus CB15PHZ_c32333,662,8603,662,477chemotaxis protein CheYIICC05912R. palustris CGA009PHZ_c32343,663,1863,666,188chemotaxis histidine kinase CheACC05942Azospirillum brasilensePHZ_c32353,666,1883,666,733chemotaxis protein CheWCC05952Rhodospirillum centenumPHZ_c32363,666,7863,669,191methyl-accepting chemotaxis protein McpHCC33492R. palustris CGA009PHZ_c32373,670,1663,669,336chemotaxis protein methyltransferase CheRCC05982R. palustris HaA2PHZ_c32383,671,2423,670,166chemotaxis response regulator CheBCC05972M. magneticum AMB-1PHZ_c33713,820,1213,819,669CheE proteinCC04413C. crescentus CB15PHZ_c33723,820,7293,820,124chemotaxis protein CheYIII-3C. crescentus CB15PHZ_c33733,821,0343,820,729CheU proteinCC04393C. crescentus CB15PHZ_c33743,821,6513,821,082chemotaxis protein CheDCC04383C. crescentus CB15PHZ_c33753,822,0373,821,651chemotaxis protein CheYIICC04373C. crescentus CB15PHZ_c33763,823,0683,822,040chemotaxis response regulator CheBCC04363C. crescentus CB15PHZ_c33773,823,9553,823,068chemotaxis protein methyltransferase CheRCC04353A. cryptum JF-5PHZ_c33783,824,4103,823,946chemotaxis protein CheWCC04343Rhizobium etli CFN 42PHZ_c33793,826,6143,824,422chemotaxis histidine kinase CheACC04333A. cryptum JF-5PHZ_c33803,826,9973,826,635chemotaxis protein CheYICC04323Caulobacter vibrioidesPHZ_c33813,827,2993,826,997CheX proteinCC04313Sinorhizobium melilotiPHZ_c33823,829,2343,827,306methyl-accepting chemotaxis protein McpACC04303A. cryptum JF-5PHZ_c010194,22093,750CheE protein-scattedC. crescentus CB15PHZ_c010294,79594,220chemotaxis protein CheYIII-scattedC. crescentus CB15PHZ_c0297292,469292,864chemotaxis protein CheYIVCC3471scattedC. crescentus CB15PHZ_c0298292,867293,679chemotaxis protein methyltransferase CheRCC3472scattedC. crescentus CB15PHZ_c0732803,383804,876methyl-accepting chemotaxis protein McpBCC0428scattedC. crescentus CB15PHZ_c09611,057,1341,058,720methyl-accepting chemotaxis protein McpICC2847scattedR. palustris CGA009PHZ_c11981,380,8831,383,294methyl-accepting chemotaxis protein McpU-scattedA. cryptum JF-5PHZ_c11991,383,2971,383,758chemotaxis protein CheW1-scattedSinorhizobium melilotiPHZ_c16871,890,2741,891,176chemotaxis MotB proteinCC1573scattedC. crescentus CB15PHZ_c19362,169,6342,169,939chemotactic signal response protein CheLCC2583scattedC. crescentus CB15PHZ_c22112,499,7442,499,274chemotaxis protein CheYIII-scattedO. alexandrii HTCC2633PHZ_c23922,720,6112,720,144chemotaxis protein CheYIII-scattedC. crescentus CB15PHZ_c27413,142,7503,143,238chemotaxis protein CheYIIICC3155scattedC. crescentus CB15PHZ_c31233,549,1503,550,016chemotaxis MotA proteinCC0750scattedC. crescentus CB15PHZ_c34013,848,8113,850,766methyl-accepting chemotaxis protein McpA-scattedC. vibrioidesTable 8Flagella genes in the P. zucineum genomeLocus5'-end3'-endNameGene symbolProposed rolePHZ_c008075,41376,462flagellin modification protein FlmAflmAregulatorPHZ_c008176,46777,621flagellin modification protein FlmBflmBregulatorPHZ_c0745816,772818,034flagellar hook-length control protein FliKfliKflagellar structurePHZ_c0787868,051866,696flagellar hook protein FlgEflgEflagellar structurePHZ_c0788868,860868,171flagellar hook assembly protein FlgDflgDflagellar structurePHZ_c0789870,604868,865flagellar hook length determination proteinflageregulatorPHZ_c0790870,819872,918flagellar hook-associated proteinflaNflagellar structurePHZ_c0791872,933873,862flagellin and related hook-associated proteins-flagellar structurePHZ_c0853945,008946,354flagellum-specific ATP synthase FliIfliIprotein export ATPasePHZ_c0854946,354946,758fliJ proteinfliJflagellar structurePHZ_c0857950,714948,621flagellar biosynthesis protein FlhAflhAexport apparatusPHZ_c0859952,470952,138flagellar motor switch protein FliNfliNmotorPHZ_c0860953,126952,479flbE proteinflbEregulatorPHZ_c0861954,151953,126flagellar motor switch protein FliGfliGmotorPHZ_c0862955,794954,151flagellar M-ring protein FliFfliFflagellar structurePHZ_c09131,007,7531,006,992flagellar L-ring protein FlgHflgHflagellar structurePHZ_c09141,008,5081,007,753distal basal-body ring component protein FlaDflaDflagellar structurePHZ_c09151,009,3001,008,515flagellar basal-body rod protein FlgGflgGflagellar structurePHZ_c09161,010,0521,009,318flagellar basal-body rod protein FlgFflgFflagellar structurePHZ_c09171,010,2721,010,874flagellar basal body-associated protein FliLfliLflagellar structurePHZ_c09181,010,9101,011,983flagellar motor switch protein FliMfliMmotorPHZ_c09221,017,0851,016,351flagellar biosynthesis protein FliPfliPexport apparatusPHZ_c09231,017,4201,017,151flagellar protein FliOfliOexport apparatusPHZ_c09241,017,5021,017,918flagellar basal-body rod protein FlgBflgBflagellar structurePHZ_c09251,017,9421,018,355flagellar basal-body rod protein FlgCflgCflagellar structurePHZ_c09261,018,3701,018,678flagellar hook-basal body complex protein FliEfliEflagellar structurePHZ_c09301,021,7961,022,056flagellar biosynthesis protein FliQfliQexport apparatusPHZ_c09311,022,0791,022,837flagellar biosynthesis protein FliRfliRexport apparatusPHZ_c09321,022,8371,023,913flagellar biosynthesis protein FlhBflhBexport apparatusPHZ_c13801,563,2811,562,745putative flagella accessory protein FlaCEflaCEflagellar structurePHZ_c13811,565,1451,563,358flagellin modification protein FlmGflmGregulatorPHZ_c13821,565,3431,565,765flagellar repressor protein FlbTflbTregulatorPHZ_c13831,565,7821,566,093flagellar biosynthesis regulator FlaFflaFregulatorPHZ_c13841,566,3751,567,202flagellin FljMfljMflagellar structurePHZ_c13851,567,4691,568,314flagellin FljMfljMflagellar structurePHZ_c13861,568,4341,568,724flagellin FlaGflaGflagellar structurePHZ_c13871,568,8871,569,720flagellin FljLfljLflagellar structurePHZ_c19352,168,5222,169,634flagellar P-ring protein FglIfglIflagellar structurePHZ_c19372,169,9422,170,382flagellar basal-body protein FlbYflbYflagellar structurePHZ_c25952,982,5502,983,593flagellin modification protein FlmDflmDregulatorPHZ_c25972,984,8742,986,508flagellin modification protein FlmGflmGregulatorPHZ_c25992,989,3152,989,974flmC; flagellin modification protein FlmCflmCregulatorPHZ_c26002,990,5492,989,977flagellin modification protein FlmHflmHregulatorThe genome contains sec-dependent, sec-independent, typical type II (Table 9) and IV secretion systems (Table 10), which are known to play important roles in adapting to diverse conditions [10,11].Table 9Distributions of proteins involved in environmental adaptation in P. zucineum and representative alphaproteobacteria with different living habitatsSpeciesS. melilotiB. suisC. crescentusP. zucineumR. conoriiG. oxydansGenome size (Mb)6.693.324.024.381.272.92GC content (%)62.257.367.271.132.460.8HabitatMultiple1Facultative1Aquatic1Facultative2Obligate1Multiple3ECF, extracytoplasmic function sigma factor (/Mb)11 (1.6)2 (0.6)15 (3.7)17 (3.9)0 (0)2 (0.7)Transcriptional regulator (/Mb)433 (64.7)149(44.9)183 (45.5)170 (38.8)11 (8.7)89 (30.1)Two-component signal transduction protein (/Mb)113 (16.9)44 (13.3)111 (27.6)102 (23.3)7 (5.5)41 (14.1)molecular chaperone23121433814Flagellar protein413742431040Chemotaxis protein4244841011Pilus protein13491624Sec-dependent secretion system111111111112Sec-independent secretion system444434Type II secretory protein2081303Type IV secretory protein989311511The habitats of S. meliloti, B. suis, and R. conorii were indicated in a recent publication [42].2According to our recent publication [1], P. zucineum was classified as \"facultative\". 3Given that G. oxydans is often isolated from sugary niches (such as flowers and fruits) and associated soil (such as garden soil and baker's soil) [43], we classified G. oxydans as \"multiple\".Table 10Type IV secretion systems in the P. zucineum genomeLocusLocation of proteinNameGenomic element5'-end3'-endPHZ_p0007Plasmid6,7867,445type IV secretion protein, VirB1PHZ_p0008Plasmid7,4837,800type IV secretion protein, VirB2PHZ_p0009Plasmid7,8168,148type IV secretion protein, VirB3PHZ_p0010Plasmid8,14410,546type IV secretion protein, VirB4PHZ_p0011Plasmid10,54611,298type IV secretion protein, VirB5PHZ_p0012Plasmid11,55312,488type IV secretion protein, VirB6PHZ_p0013Plasmid12,81613,493type IV secretion protein, VirB8PHZ_p0014Plasmid13,49314,320type IV secretion protein, VirB9PHZ_p0015Plasmid14,32015,543type IV secretion protein, VirB10PHZ_p0016Plasmid15,54316,538type IV secretion protein, VirB11PHZ_c1506Chromosome1,709,4811,709,999type IV secretion protein, TraFPHZ_c1508Chromosome1,711,0581,712,773type IV secretion protein, VirD2PHZ_c1509Chromosome1,712,7901,714,763type IV secretion protein, VirD4PHZ_c1512Chromosome1,716,2621,717,242conjugal transfer protein, TrbBPHZ_c1513Chromosome1,717,2421,717,559conjugal transfer protein, TrbCPHZ_c1514Chromosome1,717,5621,717,828conjugal transfer protein, TrbDPHZ_c1515Chromosome1,717,8361,720,283conjugal transfer protein, TrbEPHZ_c1516Chromosome1,720,2831,721,014conjugal transfer protein, TrbJPHZ_c1517Chromosome1,721,2381,722,398conjugal transfer protein, TrbLPHZ_c1518Chromosome1,722,4011,723,084conjugal transfer protein, TrbFPHZ_c1519Chromosome1,723,0871,724,064conjugal transfer protein, TrbGPHZ_c1520Chromosome1,724,0701,725,212conjugal transfer protein, TrbIPHZ_c2348Chromosome2,660,5172,660,813type IV secretion protein, VirB2PHZ_c2349Chromosome2,660,8092,661,144type IV secretion protein, VirB3PHZ_c2350Chromosome2,661,1192,663,497type IV secretion protein, VirB4PHZ_c2352Chromosome2,664,3742,665,309type IV secretion protein, VirB6PHZ_c2353Chromosome2,665,4822,666,159type IV secretion protein, VirB8PHZ_c2354Chromosome2,666,1592,667,004type IV secretion protein, VirB9PHZ_c2355Chromosome2,667,0042,668,041type IV secretion protein, VirB10PHZ_c2356Chromosome2,668,0462,669,035type IV secretion protein, VirB11PHZ_c2357Chromosome2,669,0912,670,872type IV secretion protein, VirD4To better understand the roles of proteins responsible for environmental transition, we computed the distributions of those proteins in 5 representative alphaproteobacteria with typical habitats (see methods – comparative genomics). Like other multiple bacteria and facultative bacteria, which can survive in multiple niches, P. zucineum encodes a higher fraction of ECFs, transcriptional regulators and two-component signal transduction proteins than obligate bacteria (Table 9). Notably, P. zucineum has the largest number of heat shock related proteins (Table 6), in comparison to the 5 representative alphaproteobacteria and 93 bacteria (data not shown). Among the plasmid-encoded heat shock related proteins are 2 RpoH (PHZ_p0049 and PHZ_p0288) and 2 DnaK-GrpE clusters (PHZ_p0053-0054 and PHZ_p0121-0122). Further phylogenetic analysis suggested that the plasmid-encoded DnaK-GrpE clusters may have undergone a genus-specific gene duplication event (Figure 3C &3D).Figure 3Neighbor-joining trees of 5 representative alphaproteobacteria and P. zucineum, inferred from (A) 16S rRNA genes, (B) RpoH proteins, (C) DnaK proteins and (D) GrpE proteins. The node labels are bootstrap values (100 replicates). The plasmid-encoded DnaK and GrpE of P. zucineum may have undergone a genus-specific gene duplication event (C &Adaptation to an intracellular life cycleTo survive intracellularly, P. zucineum must succeed in adhering to and subsequently invading the host cell [12], defending against a hostile intracellular environment [13-16], and capturing iron at very low concentration [17].It is well known that the pilus takes part in adhering to and invading a host cell [12]. We identified one pili biosynthesis gene (pilA) and 2 operons for pili biosynthesis (Table 11).Table 11Pilus proteins in the P. zucineum genomeLocus5'-end3'-endNameGene symbolPHZ_c0356362,116362,289pilus subunit protein PilApilAPHZ_c29923,412,8003,413,318Flp pilus assembly protein TadGtadGPHZ_c29953,415,2203,415,468Flp pilus assembly protein, pilin Flp-PHZ_c29963,415,5323,416,023Flp pilus assembly protein, protease CpaAcpaAPHZ_c29973,416,0393,416,899pilus assembly protein CpaBcpaBPHZ_c29983,416,8993,418,350pilus assembly protein CpaCcpaCPHZ_c29993,418,3553,419,587pilus assembly protein CpaEcpaEPHZ_c30003,419,5943,420,991pilus assembly protein CpaFcpaFPHZ_c30013,421,0303,421,944Flp pilus assembly protein TadBtadBPHZ_c30023,421,9443,422,903Flp pilus assembly protein TadCtadCPHZ_c30273,451,6373,452,566Flp pilus assembly protein CpaBcpaBPHZ_c30283,452,5803,453,893Flp pilus assembly protein, secretin CpaCcpaCPHZ_c30293,453,8933,455,056Flp pilus assembly protein, ATPase CpaEcpaEPHZ_c30303,455,0593,456,489Flp pilus assembly protein ATPase CpaFcpaFPHZ_c30313,456,4893,457,445Flp pilus assembly protein TadBtadBPHZ_c30323,457,4923,458,391Flp pilus assembly protein TadCtadCThe genes involved in defense against oxidative stress include superoxide dismutase (PHZ_c0927, PHZ_c1092), catalase (PHZ_c2899), peroxiredoxin (PHZ_c1548), hydroperoxide reductase (ahpF, alkyl hydroperoxide reductase, subunit f, PHZ_c2725, ahpC, alkyl hydroperoxide reductase, subunit c, PHZ_c2724), and the glutathione redox cycle system (glutathione reductase [PHZ_c1740, PHZ_c1981], glutathione synthetase [PHZ_c3479], and γ-glutamylcysteine synthetase [PHZ_c0446, PHZ_c0523]).Since intracellular free Fe is not sufficient to support the life of bacteria, to survive intracellularly, they must use protein-bound iron, such as heme and transferrin, via transporters and/or the siderophore system. The P. zucineum genome has one ABC type siderophore transporter system (PHZ_c1893-1895), one ABC type heme transporter system (PHZ_c0136, PHZ_c0139, PHZ_c0140), and 60 TonB-dependent receptors which may uptake the iron-siderophore complex (Table 12).Table 12TonB-dependent receptors in the P. zucineum genomeAnnotationChromosomePlasmidCOG categoryTonB-dependent receptor512COG16291TonB-dependent receptor vitamin B1230COG42062TonB-dependent receptor40COG477131COG1629, Outer membrane receptor proteins, mostly Fe transport2COG4206, Outer membrane cobalamin receptor protein3COG4774, Outer membrane receptor for monomeric catecholsComparative genomics between P. zucineum and C. crescentusComparative genomic analysis demonstrated that P. zucineum is phylogenetically the closest to C. crescentus [18] (Figure 4), consistent with the phylogenetic analysis based on 16S RNA gene sequences (Figure 5).Figure 4List of top 10 complete sequenced bacteria closest to P. zucineum. All 10 are alphaproteobacteria. Among all the sequenced bacterial genomes, C. crescentus shares the greatest number of similar ORFs with P. zucineumFigure 5Neighbor-joining tree of the alphaproteobacteria, inferred from 16S rRNA genes. The node labels are bootstrap values (100 replicates). C. crescentus is phylogenetically the closest to P. zucineum.Though the genome size and protein number of P. zucineum (4.37 Mb, 3,861 proteins) are similar to those of C. crescentus (4.01 Mb, 3,767 proteins), no large-scale synteny was found between the genomes. The largest synteny region is only about 30 kb that encodes 24 proteins. The conservation region with the largest number of proteins is the operon encoding 27 ribosomal proteins. In addition, the species share only 57.8% (2,231/3,861) of orthologous proteins. Categories J (translation, ribosomal structure and biogenesis), F (nucleotide transport and metabolism), and L (replication, recombination and repair) are the top 3 conservative COG categories between the species, sharing 88.01%, 81.67%, and 80.65% of the orthologs, respectively.Comparison of cell cycle genes between P. zucineum and C. crescentusSince P. zucineum is phylogenetically closest to C. crescentus, and since the latter is a model organism for studies of the prokaryotic cell cycle [19,20], we compared the genes regulating the cell cycle between these species.The cell cycle of C. crescentus is controlled to a large extent by the master regulator CtrA, which controls the transcription of 95 genes involved in the cycle [19,20]. On the other hand, ctrA is regulated at the levels of transcription, phosphorylation, and proteolytic degradation by its target genes, e.g., DNA methyltransferase (CcrM) regulates the transcription of ctrA, histidine kinases (CckA, PleC, DivJ, DivL) regulate its activity, and ClpXP degrades it. These regulatory 'loops' enable CtrA to precisely control the progression of the cell cycle.P. zucineum has most of the orthologs mentioned above (Table 13). Among the 95 CtrA-regulated genes in C. crescentus, 75 have orthologs in the P. zucineum genome (Additional file 1). The fraction of CtrA-regulated genes with orthologs in P. zucineum (76.9%, 73/95) is significantly greater than the mean level of the whole genome (57.8%, 2,231/3,861), indicating that the CtrA regulatory system is highly conserved. Genes participating in regulating central events of the cell cycle, such as CcrM (CC0378), Clp protease (CC1963) and 14 regulatory proteins, except for one response regulator (CC3286), are present in the P. zucineum genome. The genes without counterparts in P. zucineum are mostly for functionally unknown proteins.Table 13Comparison of the signal transduction pathways regulating CtrA between the P. zucineum and the C. crescentusLocusLengthAmino acid Identity (%)AnnotationC. crescentusP. zucineumC. crescentusP. zucineumCC0378PHZ_c057735535980.00modification methylase CcrMCC1078PHZ_c093369166367.22cell cycle histidine kinase CckACC2482PHZ_c268184260663.78sensor histidine kinase PleCCC1063PHZ_c271259750453.83sensor histidine kinase DivJCC3484PHZ_c021876976967.66tyrosine kinase DivLCC2463PHZ_c130913012189.26polar differentiation response regulator DivKCC1963PHZ_c181720220580.19ATP-dependent protease, ClpP subunitCC1961PHZ_c181442042090.47ATP-dependent protease, ClpX subunitNotably, the sequence of CtrA is strikingly similar between P. zucineum and C. crescentus, with 93.07% identity of amino acid sequence and 89.88% identity of nucleotide sequence. In addition, they share identical promoters (p1 and p2) [21] and the motif (GAnTC) recognized by DNA methyltransferase (CcrM) (Figure 6) [22], suggesting that they probably share a similar regulatory loop of CtrA.Figure 6Nucleotide acid sequence alignment of the ctrA promoter regions (-200 to +21) of C. crescentus and P. zucineum. Blue background: identical nucleotides; \"-\": gaps; red and black box: P1 and P2 promoter; black underline: motif recognized by CcrM; red underline: first 21 nucleotides starting with initial codon \"ATG.\".Consistent with the results from in silico sequence analysis, the CtrA of P. zucineum can restore the growth of temperature-sensitive strain LC2195 (a CtrA mutant) of C. crescentus [23] at 37°C, indicating that the CtrA of P. zucineum can functionally compliment that of C. crescentus in our experimental conditions (data not shown).Taken together, the comparative genomics of P. zucineum and C. crescentus suggests that the cell cycle of the former is likely to be regulated similarly to that of the latter.Presence of ESTs of the strain in humanSince P. zucineum strain HLK1T can invade and persistently live in several human cell lines [1], we were curious about whether this microbe can infect humans. By blasting against the human EST database (dbEST release 041307 with 7,974,440 human ESTs) with the whole genome sequence of P. zucineum, we found 9 matched ESTs (Table 14), of which 3 were from a library constructed from tissue adjacent to a breast cancer, and 6 were from a library constructed from a cell line of lymphatic origin. The preliminary data suggest that P. zucineum may invade humans.Table 14Human ESTs matching the genome sequences of P. zucineumQuery GISample originQuery LengthQuery PositionChromosome PositionScoreE ValueSimilarity (%)BeginEndBeginEnd14251638Breast tissue1226411751,276,9141,277,0482042.00E-5394.078261474Breast tissue11611081,277,0421,276,9371672.00E-4296.3114251634Breast tissue142191341,277,0541,276,9372041.00E-5397.4633194938Lymphatic cell line244184411,029,5751,029,142749096.7733194696Lymphatic cell line65286521,029,5751,028,9311,166097.6733193754Lymphatic cell line65486541,029,5751,028,9291,191098.157117824Lymphatic cell line40574051,558,8311,558,433735098.2533194587Lymphatic cell line63876382,864,4702,863,8381,191098.897114909Lymphatic cell line34763473,498,6243,498,283654099.121All of three sequences come from the library BN0075 containing 182 ESTs; the original dataset was produced by a modification of the EST sequencing strategy ORESTES (open reading frame expressed sequences tags)[44,45]2All six sequences come from the library NIH_MGC_51 containing 2,381 ESTs; the original dataset was produced and released by the \"Mammalian Gene Collection\" project [46].ConclusionThis work presents the first complete bacterial genome in the genus Phenylobacterium. Genome analysis reveals the fundamental basis for this strain to invade and persistently survive in human cells. P. zucineum is phylogenetically closest to C. crescentus based on comparative genome analysis.MethodsBacterial growth and genomic library constructionP. zucineum strain HLK1Twas grown in LB (Luria-Bertani) broth at 37°C and then harvested for the preparation of genomic DNA[1]. Genomic DNA was prepared using a bacterial genomic DNA purification kit (V-Gene Biotech., Hangzhou, China) according to the manufacturer's instructions. Sheared DNA samples were fractionated to construct three different genomic libraries, containing average insert sizes of 2.0–2.5 kb, 2.5–3.0 kb and 3.5–4.0 kb. The resulting pUC18-derived library plasmids were extracted using the alkaline lysis method and subjected to direct DNA sequencing with automated capillary DNA sequencers (ABI3730 or MegaBACE1000).Sequencing and finishingThe genome of P. zucineum was sequenced by means of the whole genome shotgun method with the phred/phrap/consed software packages [24-27]. Sequencing and subsequent gene identification was carried out as described in our earlier publications [28-30]. Briefly, during the shotgun sequence phase, clones were picked randomly from three shotgun libraries and then sequenced from both ends. 44,667 successful sequence reads (>100 bp at Phred value Q13), accounting for 5.47× sequence coverage of the genome, were assembled into 563 sequence contigs representing 60 scaffolds connected by end-pairing information.The finishing phase involved iterative cycles of laboratory work and computational analysis. To reduce the numbers of scaffolds, reads were added into initial contig assembly by using failed universal primers as primers and by using plasmid clones that extended outwards from the scaffolds as sequence reaction templates. To resolve the low-quality regions, resequencing of the involved reads in low quality regions with universal primers and primer walking the plasmid clones were the first choice, otherwise, resequencing with alternate temperature conditions resolved the remaining low-quality regions. New sequence reads obtained from the above laboratory work were assembled into existing contigs, which yielded new contigs and new scaffolds connected by end-pairing information. Then, consed interface helped us to do nest round of laboratory work based on new arisen contig assembly. After about four iterative cycles of the above \"finish\" procedures to close gaps and to resolve the low-quality regions, the PCR product obtained by using total genomic DNA as template was sequenced from both ends to close the last physical gap. In addition, the overall sequence quality of the genome was further improved by using the following criteria: (1) two independent high-quality reads as minimal coverage, and (2) Phred quality value = Q40 for each given base. Collectively, 3,542 successful reads were incorporated into initial assembles during the finishing phase. The final assembly was composed of two circular \"contigs\", of which a smaller one with a protein cluster (including repA, repB, parA and parB) related to plasmid replication was assigned as the plasmid, and the larger one was the chromosome.AnnotationtRNA genes were predicted with tRNAscan-SE [31]. Repetitive sequences were detected by REPuter [32,33], coupled with intensive manual alignment. We identified and annotated the protein profiles of chromosome and plasmid with the same workstream. For the chromosome, the first set of potential CDSs in the chromosome was established with Glimmer 2.0 trained with a set of ORFs longer than 500 bp from its genomic sequence at default settings [34]. The resulting 5,029 predicted CDSs were BLAST searched against the NCBI non-redundant protein database to determine their homology [35]. 1,174 annotated proteins without the word \"hypothetical\" or \"unknown\" in their function description, and without frameshifts or in-frame stop codons, were selected as the second training set. The resulting second set of 4,018 predicted CDSs (assigned as \"predicted CDSs\") were searched against the NCBI non-redundant protein database. Predicted CDSs that accorded with the following BLAST search criteria were considered \"true proteins\": (1) 80% of the query sequence was aligned and (2) E-value ≤ 1e-10. Then, the ORFs extracted from the chromosome region among \"true proteins\" were searched against the NCBI non-redundant protein database. The ORFs satisfying the same criteria as true proteins were considered \"true ORFs\". Overlapping proteins were manually inspected and resolved, according to the principle we described previously [30]. The final version of the protein profile comprised three parts: true proteins, true ORFs, and predicted CDSs located in the rest of the genome. The translational start codon of each protein was identified by the widely used RBS script [36] and then refined by comparison with homologous proteins [30].To further investigate the function of each protein, we used InterProScan to search against the InterPro protein family database [37]. The up-to-date KEGG pathway database was used for pathway analysis [38]. All proteins were searched against the COG database which included 66 completed genomes [39,40]. The final annotation was manually inspected by comprehensively integrating the results from searching against the databases of nr, COG, KEGG, and InterPro.Phylogenetic tree construction16S rRNA genes were retrieved from 63 alphaproteobacteria, P. zucineum and Escherichia coli O157:H7 EDL933. A neighbor-joining tree with bootstrapping was built using MEGA [41]. The gammaproteobacterium E. coli was used as the outgroup to root the tree. To illustrate the evolutionary history of heat shock related proteins (RpoH, DnaK and GrpE), neighbor-joining trees based on the 16S rRNA genes and the above three proteins of 5 representative alphaproteobacteria (Sinorhizobium meliloti 1021, Brucella suis 1330, C. crescentus CB15, Rickettsia conorii str. Malish 7, Gluconobacter oxydans 621H), P. zucineum and E. coli O157:H7 EDL933 were constructed.Comparative genomicsSequence data for comparative analyses were obtained from the NCBI database . The database has 520 completely sequenced bacterial genomes (sequences downloaded on 2007/06/05). All P. zucineum ORFs were searched against the ORFs from all other bacterial genomes with BLASTP. The number of P. zucineum ORFs matched to each genome with significance (E value = 1e-10) was calculated.To illustrate the contribution of transcriptional regulators and two-component signal transduction proteins to environmental adaptation, we compared the mean fraction of these two types of proteins in bacteria living in 6 different habitats, as described by Merav Parter [42]. These are: (1) obligate bacteria that are necessarily associated with a host, (2) specialized bacteria that live in specific environments, such as marine thermal vents, (3) aquatic bacteria that live in fresh or seawater, (4) facultative bacteria, free-living bacteria that are often associated with a host, (5) multiple bacteria that live in many different environments, and (6) terrestrial bacteria that live in the soil. For bacteria with more than one sequenced strain, we chose only one strain for the comparative study. The numbers of bacterial species in each group were: 26 obligate, 5 specialized, 4 aquatic, 28 facultative, 27 multiple, and 3 terrestrial. We annotated the proteins of these 93 species with the same workflow used for P. zucineum and calculated the mean fraction of transcriptional regulators and two-component signal transduction proteins.In addition, we annotated the ORFs of 5 representative alphaproteobacteria with different habitats (multiple bacteria S. meliloti 1021 and G. oxydans 621H, facultative bacterium B. suis 1330, aquatic bacterium C. crescentus CB15, and obligate bacterium R. conorii str. Malish 7) using the same workflow and computed the distributions of proteins involved in environmental adaptation.Ortholog identificationAll proteins encoded by one genome were BLASTP searched against a database of proteins encoded by another genome [35], and vice versa. The threshold used in these comparisons was 1e-10. Orthology was identified if two proteins were each other's best BLASTP hit (best reciprocal match).Data accessibilityThe sequences reported in this paper have been deposited in the GenBank database. The accession numbers for chromosome and plasmid are CP000747 and CP000748, respectively.AbbreviationsEST: Expressed Sequence Tag; KEGG: Kyoto Encyclopedia of Genes and Genomes.Authors' contributionsXH and SH designed the project; YL, XX, ZD, ZL, ZY and JS performed the research; SH and BZ contributed new reagents\\analytical tools; YL, XX, and ZD analyzed the data; and XH, YL, and SH wrote the paper. All authors read and approved the final manuscript.Supplementary MaterialAdditional file 1Supplemental Table 1 Comparison of genes directly regulated by CtrA between P. zucineum and C. crescentus.Click here for file\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2529333.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2529333",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2529333\nAUTHORS: Jesus Vioque, Xavier Barber, Francisco Bolumar, Miquel Porta, Miguel Santibáñez, Manuela García de la Hera, Eduardo Moreno-Osset\n\nABSTRACT:\nBackgroundThe effect of tobacco smoking and alcohol drinking on esophageal cancer (EC) has never been explored in Spain where black tobacco and wine consumptions are quite prevalent. We estimated the independent effect of different alcoholic beverages and type of tobacco smoking on the risk of EC and its main histological cell type (squamous cell carcinoma) in a hospital-based case-control study in a Mediterranean area of Spain.MethodsWe only included incident cases with histologically confirmed EC (n = 202). Controls were frequency-matched to cases by age, sex and province (n = 455). Information on risk factors was elicited by trained interviewers using structured questionnaires. Multiple logistic regression was used to estimate adjusted odds ratios and 95% confidence intervals (CI).ResultsAlcohol drinking and tobacco smoking were strong and independent risk factors for esophageal cancer. Alcohol was a potent risk factor with a clear dose-response relationship, particularly for esophageal squamous-cell cancer. Compared to never-drinkers, the risk for heaviest drinkers (≥ 75 g/day of pure ethanol) was 7.65 (95%CI, 3.16–18.49); and compared with never-smokers, the risk for heaviest smokers (≥ 30 cigarettes/day) was 5.07 (95%CI, 2.06–12.47). A low consumption of only wine and/or beer (1–24 g/d) did not increase the risk whereas a strong positive trend was observed for all types of alcoholic beverages that included any combination of hard liquors with beer and/or wine (p-trend<0.00001). A significant increase in EC risk was only observed for black-tobacco smoking (2.5-fold increase), not for blond tobacco. The effects for alcohol drinking were much stronger when the analysis was limited to the esophageal squamous cell carcinoma (n = 160), whereas a lack of effect for adenocarcinoma was evidenced. Smoking cessation showed a beneficial effect within ten years whereas drinking cessation did not.ConclusionOur study shows that the risk of EC, and particularly the squamous cell type, is strongly associated with alcohol drinking. The consumption of any combination of hard liquors seems to be harmful whereas a low consumption of only wine may not. This may relates to the presence of certain antioxidant compounds found in wine but practically lacking in liquors. Tobacco smoking is also a clear risk factor, black more than blond.\n\nBODY:\nBackgroundMost epidemiological studies have identified tobacco smoking and alcohol drinking as the main risk factors for esophageal squamous-cell carcinoma or unspecified esophageal cancer [1-3], usually with a monotonic and strong dose-response relationship [4]. The evidence that high alcohol consumption increases the risk of cancer of the esophagus is quite convincing at present; most of it suggests that it is the amount of alcohol consumed, rather than the particular drink that determines the risk. However, a prospective cohort study in Denmark observed that low wine consumption, compared to beer and hard liquor, did not increase risk [5]. Two population-based case-control studies also presented data supporting this observation [6,7]. In an ecological study in Spain, trends of per capita consumption of beer, spirits, and total alcohol consumption were positively correlated with oesophageal cancer mortality in men whereas wine consumption showed no relationship with oesophageal cancer mortality either in men or women[8]. As the consumption of all types of alcoholic drinks is quite prevalent in Spain, it is relevant to assess the risk of esophageal cancer in relation to them.Tobacco smoking has also a clear role in the aetiology of esophageal cancer [1,2]. Despite the high prevalence of consumption, particularly of black tobacco, an association between the type of tobacco smoking and esophageal cancer has never been explored by epidemiological studies in Spain, only in South America or France[4,9].In addition to smoking and alcohol, other factors such as a low socioeconomic status, an infrequent consumption of fruits and vegetables and, in some areas of Asia, the betel nut chewing, have also been related to esophageal cancer [10-13]; these factors account for a high proportion of cases, particularly for the esophageal squamous cell type, the most frequent histological type [11,12,14].We conducted a case-control study in Valencia and Alicante, Spain, to estimate the independent effects of different alcoholic beverages (beer, wine and spirits) and type of tobacco smoking (black and blond) on the risk of esophageal cancer and its main histological cell type (squamous cell carcinoma).MethodsDesignThis research was part of the PANESOES project, a prospective hospital-based case-control study designed to explore the influence of major lifestyles and diet on the risk of three gastrointestinal cancers, pancreas, oesophagus and stomach. The PANESOES Study aimed to recruit approximately 200 cases for oesophagus, 200 for pancreas cancer, 400 cases for stomach cancer, and 400–450 controls. This sample size was planned to estimate as statistically significant (p < 0.05) a RR over 1.5 for stomach cancer and high prevalent exposures such as tobacco smoking and alcohol drinking, and a RR over 1.8 for oesophagus and pancreas cancer.Eligible subjects were Spanish-speaking men and women 30–80 years old, and hospitalized between January 1995 and March 1999 in any of nine participant hospitals in the provinces of Alicante and Valencia. These nine hospitals are among the ten main hospitals from the Health Care Service in Valencia and Alicante that were invited to participate (only one hospital in Valencia declined participation). The access to the Health Care System in Spain is free and universal; thus, our case series may be considered representative as the participant hospitals accounted for approximately 90% of cases in both provinces.All subjects were informed of the study objectives and gave their informed consent before the interview. Research protocols were approved by the local ethics and/or research committees of the participating Hospitals and the University.SubjectsCases were patients newly diagnosed with a primary invasive cancer of the oesophagus. Pathology reports were obtained for all case patients initially diagnosed as esophageal cancer, and final inclusion in the case series was based on histological confirmation. A total of 211 cases were initially identified as eligible for participation; two cases refused interview (0.9%), and seven cases were not finally confirmed (3.3%). The final analysis was thus based on 202 cases. The pathological diagnosis was confirmed as squamous cell carcinoma in 160 cases (79.2%) and adenocarcinoma in 42 cases (20.8%).Control subjects were selected from the same hospitals from which cases were identified and during the same period. Controls were frequency matched to the expected distribution of case subjects of the whole PANESOES study (i.e. cases of oesophagus, stomach and pancreas, and controls) by three age groups (<60 years, 60–69 and 70–80 years), sex, and province (Alicante and Valencia). A wide inclusion criterion was used to select controls from diseases not related a priori to the main exposures of interest (tobacco, alcohol and diet). The overall participation rate of the 457 eligible controls was 99.6%, leaving 455 controls subjects with completed interviews. The distribution of the main diagnoses for the control group was, in decreasing order: hernia (34.0%), degenerative osteoarthritis/arthritis (21.3%), fractures/injuries/orthopaedic processes (18.9%), appendicitis (6.4%) and other, less prevalent, conditions (19.3%).Exposure dataFace-to-face interviews were conducted in-hospital for all participants by trained interviewers, using a structured questionnaire. Interviews were administered directly to the study subjects, rather than to the next of kin, for 89.2% of the target case subjects and for 96% of the target control subjects. The average time required to complete the questionnaire was 40 minutes. While interviewers could not be blinded to the case/control status, they were unaware of the main study hypothesis and were trained to administer strictly the structured questionnaires in an equal manner to case and controls.Information was collected on demographic characteristics, tobacco and alcohol use, medical history and other lifestyle factors. The interview elicited details on usual tobacco use, including the product/tobacco type (black or blond cigarettes, cigars, pipes) as well as intensity of use, the age at which the habit started and stopped, the total duration of use excluding the years stopped, the years since last use for each type of product, and the level of inhalation when smoking (partially or totally). A never smoker was defined as someone having smoked fewer than 100 cigarettes ever or less than one cigarette per day for one year. A former smoker was defined as someone having stopped smoking 1 or more years before the interview. We computed the average number of cigarettes smoked per day including type of tobacco smoked, lifetime number of cigarettes smoked and pack-years of smoking (number of 20-cigarette packs per day multiplied by number of years smoking). Only 2 cases and 8 controls simultaneously smoked both blond and black tobacco, and they were considered as mainly blond tobacco smokers.Alcohol consumption patterns were assessed through inquiries into the usual intake for each type of beverage separately, i.e. beer, wine, or liquor. Since the intake of only wine and/or beer intake was uncommon among cases, these two alcoholic beverages were combined in the analysis. Since only 8 participants consumed liquors but not beer or wine, we estimated a category for the combined consumption of liquors with any combination of beer and/or wine (All types of beverages). The average relative contributions of beer, wine and liquor consumption in this category were 30%, 50% and 20% respectively. The three types of alcohol were also combined to give an overall estimate of the alcohol consumption. A never drinker was defined as having consumed less than one drink per month. One drink was defined as 200 cc of beer, 125 cc of wine, or 50 cc of hard liquor. The content of pure alcohol was calculated according to the following concentrations specific for Spain: 5% for beer; 12 percent for wine; and 40 percent for hard liquors. The resulting values were converted to grams multiplying 1 ml of pure ethanol per 789 mg [15]. The average in grams of pure ethanol consumed per day, type of alcoholic beverage, and life time duration of the habit, the age at starting and stopping the habit, and then the years since quitting drinking were also estimated. A former drinker was defined as having stopped drinking 1 or more years before the interview.The fruit and vegetable intake was assessed by a food frequency questionnaire (FFQ). For this study, we adapted and validated a FFQ of 93 food items similar to the Harvard questionnaire in order to assess the diet five years before the interview in the hospital [16-18]. Participants were asked to report their average consumption of 12 vegetables and 10 fruit items. The average daily intakes for each fruit and vegetable were summed to compute the total fruit and vegetable intake in grams, and they were adjusted for energy intake using the residual method [19]. We further computed tertiles of energy-adjusted intake of fruit and vegetable using the distribution of cases and controls in the whole PANESOES study (cases of oesophagus, stomach and pancreas, and controls).Statistical AnalysisWe used unconditional logistic regression to estimate odds ratios (OR), as an estimate of relative risk, and corresponding 95% confidence intervals (CIs) [20]. All regression models included as covariates the three frequency-matched factors, sex (men/women), age (<60 years, 60–69 years, 70–80 years) and hospital origin (Valencia/Alicante) entered as indicator variables; the educational level (<primary, primary completed, and ≥ secondary school entered as indicator variable); and the different variables of alcohol intake and tobacco smoking also using indicator variables. In addition, we adjusted for a potential confounding effect of fruit and vegetable intake (in tertiles and entered as indicator variables), and energy intake.Tests for trend in the ORs across exposure strata were calculated for ordinal variables by using logistic models that included categorical terms as continuous variables in a model with all the potential confounders and, where appropriate, omitting the never or former users/exposed. For trend-tests, we used the likelihood ratio test statistic with one degree of freedom. Although the study sample size did not allow us to estimate tests of interactions, we performed exploratory analyses for the effects of alcohol and tobacco by strata of never-ever smoking status and never-ever drinking status, respectively. All analyses were performed with STATA-8 [21]. Statistical significance was set at 0.05. All reported P values are from two-sided tests.ResultsTable 1 shows the distribution of cases and controls according to demographic characteristics and main exposure variables. The distribution of the frequency-matched variables age, sex and province were comparable between the control series and the overall case series of the PANESOES study (i.e., cancers of oesophagus, stomach and pancreas, data not shown). The educational level was comparable between cases and controls. Alcohol intake, tobacco smoking and a low intake of fruit and vegetable were more prevalent among cases than controls.Table 1Distribution of socio-demographic characteristics and other exposure variables among case and control subjectsVariablesCasesa%Controls%Sex Men18792.6128562.64 Women157.3917027.78Age <60 years old8642.3614932.75 60–69 years old7939.4116736.70 ≥ 70 years old3718.2313930.55Province Alicante4522.6613930.55 Valencia15777.3431669.45Educational level <Primary11456.6524654.07 Primary6733.0017237.80 ≥ Secondary2110.34378.13Alcohol intake Never167.8817137.36 Ever18692.1228462.64Tobacco smoking Never smoker2311.3921847.91 Ever smoker17988.6123752.09Fruit and Vegetable daily intake (in tertile) < 166 g/d12159.9010422.86 166–255 g/d4321.2914130.99 >255 g/d3818.8121046.15a All cases were histologically confirmed.Table 2 shows risk estimates according to various patterns of alcohol intake. We identified moderate to strong effects for all measures of amount and duration (years of alcohol drinking); risk was particularly strong for all types of beverages. Former drinkers and current drinkers experienced, respectively, a 4.3 and 2.1-fold increase of risk over never drinkers. An increasing risk of esophageal cancer was observed according to the daily amount of pure ethanol consumption (p-trend < 0.00001); risk was very high and significant among subjects consuming ≥ 75 g/d (OR = 7.65). Concerning the beverage type, drinkers of all types of beverages showed the highest risk (OR = 4.39) and ever drinkers of only beer also showed a significant effect (3.07); however, no significant effects were observed for drinkers of only wine or wine and/or beer. When the daily amount by type of beverage was considered, no significantly increased risks were observed for wine-beer combined consumption. For all types of alcoholic beverages combined, a non-significant risk was observed for a daily consumption of 1–24 g/day (OR = 1.53), whereas risk increased sharply to 3.9 and 10.6-fold for the upper categories (p-trend<0.0001).Table 2Adjusteda odds ratios (OR) for esophageal cancer, and esophageal squamous-cell carcinoma, according to alcohol consumptionNo. of controlsAll esophageal cancer casesEsophageal squamous-cell carcinoma casesCasesORa95% CICasesORa95% CIAlcohol Drinking Status Never<171161.0061.00 Ever2841862.41(1.24 – 4.71)1545.34(2.05 – 13.91) Never171161.0061.00 Former drinker49384.28(1.92 – 9.56)3111.03(3.73 – 32.62) Current drinker2351482.06(1.04 – 4.08)1234.48(1.69 – 11.83)Average of pure ethanol (g/day) Never171161.0061.00 Former drinker49385.40(2.43 – 12.00)3116.03(5.34 – 48.07) 1–24 g/d147271.16(0.54 – 2.49)121.71(0.56 – 5.20) 25–74 g/d62452.89(1.29 – 6.48)388.02(2.64 – 24.40) ≥ 75 g/d26757.65(3.16 – 18.49)7223.20(7.19 – 74.90)  p-value for linear trend<0.00001<0.00001Type of drink (g/day) Never171161.0061.00 Only wine65121.20(0.49 – 2.91)71.92(0.56 – 6.64) Only beer2683.07(1.06 – 8.90)67.98(2.13 – 29.92) Only wine and/or beer67151.44(0.59 – 3.47)92.49(0.73 – 88.46) All types of beverages1261514.39(2.08 – 9.22)13212.15(4.17 – 34.56) Never171161.0061.00 Former drinker49385.50(2.47 – 12.27)3116.74(5.53 – 50.67) Wine and/or Beer 1–24 g/d107141.04(0.45 – 2.41)61.48(0.43 – 5.05) Wine and/or Beer ≥ 25 g/d30132.04(0.76 – 5.46)115.48(1.52 – 19.71) All types 1–24 g/d40131.53(0.58 – 4.01)62.47(0.62 – 9.86) All types ≥ 25 g/d581085.76(2.59 – 12.83)10017.22(5.68 – 52.21)   All types 25–74 g/d40393.88(1.64 – 9.15)3210.47(3.26 – 33.64)   All types ≥ 75 g/d186910.62(4.14 – 27.24)6835.03(10.28 – 119.31)  p-value for linear trend<0.0001<0.0001Years of alcohol drinking 0171161.0061.00 1–1916154.06(1.37 – 12.00)118.21(2.08 – 32.32) 20–39121922.96(1.42 – 6.14)766.28(2.27 – 17.35) ≥ 40129711.71(0.80 – 3.61)623.99(1.41 – 11.24)  p-value for linear trend0.3160.036Age at starting <18 yr51431.00391.00 18–20 yr88501.94(0.88 – 4.27)400.63(0.29 – 1.37) 21–29 yr59341.04(0.53 – 2.02)310.83(0.34 – 2.02) ≥ 30 yr72521.71(0.86 – 3.40)400.91(0.36 – 2.29)Years since quitting Current Drinker2351481.001231.00 <5 yr12163.60(1.34 – 9.69)145.89(2.01 – 17.25) ≥ 5 yr37221.71(0.86 – 3.41)171.70(0.79 – 3.66)a Adjusted for sex, age, educational level, province and tobacco smoking (never, past, <20, 20–49 and ≥ 50 pack-years), the energy-adjusted intake of fruit and vegetable in tertiles, and energy intake.Regarding duration of drinking, statistically significant increased risks were observed for up to 40 years of drinking, after which no significant increase was observed (Table 2). Age at start drinking was not significantly associated to a higher risk of esophageal cancer. Drinking cessation in the last 5 years was associated with a strong risk excess compared with persistent drinkers (OR = 3.60, 95% CI 1.34–9.69); the risk decreased thereafter although still remained higher than in current drinkers.Table 2 also shows the estimated ORs according to various patterns of alcohol intake specifically for the esophageal squamous-cell carcinoma histological type (n = 160 cases). Most effects were approximately 3 times higher than risk estimates for the whole case series (n = 202 cases). The effects of all types of alcoholic beverages, were in general 2 to 4-fold higher than those observed for drinkers of only wine and/or beer. The highest risk of esophageal cancer was observed among cases with an average daily intake of pure ethanol ≥ 75 g/d (OR = 23.20, 95% CI: 7.19–74.90), and particularly for those consuming ≥ 75 g/d of all types of alcoholic beverages (OR = 35.03, 95% CI: 10.28–119.31). No significant effects were observed for only wine or beer drinkers 1–24 g/d.Table 3 shows adjusted risk estimates according to various patterns of tobacco smoking. The risk of esophageal cancer was more than doubled among ever smokers. Current smokers presented a 2.6-fold increase of risk with respect to never smokers. The adjusted ORs increased with the number of cigarettes smoked per day up to 5 among cases smoking ≥ 30 cigarettes/day (p-trend = 0.002), and with increasing years of smoking (p-trend = 0.030). Pack-years also showed a significant dose-response (p-trend = 0.005). No association was found with age at which subjects started smoking. After smoking cessation there was a reduction of risk of 35% for ten or more years and 45% for less than ten years that overall was statistically significant (LRS, p-value = 0.042). A statistically significant increased risk was observed for black tobacco but not for blond tobacco. Smokers who totally inhaled the tobacco experienced a statistically significant higher risk than non- or partially inhaling smokers. The risks for esophageal squamous-cell carcinoma were very similar to those observed for the whole case series.Table 3Adjusteda odds ratios (OR) for esophageal cancer, and esophageal squamous-cell carcinoma, according to tobacco smokingNo. of controlsAll esophageal cancer casesEsophageal squamous-cell carcinoma casesCasesORa95% CICasesORa95% CISmoking Status Never218231.00151.00 Ever2371792.12(1.06 – 4.23)1451.70(0.72 – 4.03) Never218231.00151.00 Former-Smoker117551.61(0.76 – 3.40)391.08(0.43 – 2.73) Current Smoker1201242.58(1.26 – 5.28)1062.28(0.95 – 5.51)Average No. of cigarettes/day Never218231.00151.00 Former smoker117551.68(0.79 – 3.57)391.16(0.45 – 2.99) <1527111.70(0.63 – 4.58)81.35(0.41 – 4.48) 15–2958582.45(1.11 – 5.44)472.05(0.78 – 5.42) ≥ 3023515.07(2.06 – 12.47)485.82(1.96 – 17.23) Pipe/Cigars1241.58(0.41 – 6.21)31.49(0.30 – 7.46)  p-value for linear trend0.0020.006Years of cigarette smoking Never218231.00151.00 <2039141.78(0.69 – 4.59)40.54(0.12 – 2.11) 20–2948321.94(0.81 – 4.63)261.62(0.57 – 4.65) ≥ 301501342.25(1.10 – 4.58)1152.03(0.84 – 4.92)  p-value for linear trend0.0300.029Pack-yearsb Never smoker (0 p-y)218231.00151.00 Past117551.67(0.79 – 3.55)391.15(0.45 – 2.95) <2038111.30(0.49 – 3.47)60.69(0.19 – 2.51) 20–4955592.81(1.27 – 6.19)482.49(0.96 – 6.51) ≥ 5027543.79(1.60 – 8.95)523.93(1.40 – 10.99)  p-value for linear trend0.0050.007Age at starting <1540411.00351.00 15–1778611.11(0.56 – 2.21)491.45(0.65 – 3.23) ≥ 18116700.91(0.47 – 1.74)561.21(0.57 – 2.58)Years since quitting Current Smoker1201241.001061.00 <1050260.55(0.29 – 1.06)200.44(0.20 – 0.96) ≥ 1067290.65(0.35 – 1.21)190.49(0.23 – 1.06)Type of tobacco Never218231.00151.00 Blond71311.19(0.50 – 2.82)240.80(0.28 – 2.30) Black1501442.50(1.23 – 5.09)1182.05(0.85 – 4.97)Inhalation of smoke Never218241.00151.00 Partially164931.50(0.74 – 3.07)771.32(0.54 – 3.25) Totally72852.88(1.37 – 6.06)672.53(1.00 – 6.46)a Adjusted for sex, age, educational level, province, and alcohol drinking (never, past, 1–24 g/d, 25–74 g/d, ≥ 75 g/d), the energy-adjusted intake of fruits and vegetables in tertiles, and energy intakeb Pack-years = the number of packs of 20 cigarettes per day multiplied by the number of years smoking.DiscussionOur results confirm that alcohol drinking and tobacco smoking are both strong and independent risk factors for esophageal cancer in Spain. We found that heavy drinkers had higher increased risk than heavy smokers, particularly for the esophageal squamous-cell carcinoma. These results are consistent with most of the epidemiological studies carried out in Western countries and some areas of Asia [4,11,12,22-25]. Other studies have found however, chewing or tobacco smoking to be a similar or even a stronger risk factor than alcohol drinking [13,26-28] although some of these studies did not include subjects with a high consumption of alcohol [13,26].Regarding alcohol drinking, we found that the intensity (i.e. the average daily alcohol intake in grams of pure ethanol) was a more relevant predictor of risk than the duration of the habit. The age at which subjects started drinking was not associated with risk and the cessation of drinking was not associated with any beneficial effect even ≥ 5y after cessation. The consequences of drinking cessation has been studied less frequently than smoking cessation and results are more controversial, probably because the number of people who quit drinking is lower than those who quit smoking. A beneficial effect has been found in some studies particularly 10 years after giving up drinking [25,29] although some studies have shown a rapid decline in risk after cessation of drinking [4,30,31]. On the contrary, other studies have shown either a non beneficial effect [23] or a higher risk among former drinkers [31,32]. In our study we observed a statistically significant increased risk among former drinkers who stopped drinking <5y. Unfortunately, we had an insufficient number of former drinkers to explore the drinking cessation after 10 years in more detail. Although we considered as former drinkers those who reported quitting at least one year before the interview, it is possible that some of the patients were in fact heavy drinkers quitting the habit because of their disease; alternatively, it may also be possible that some heavy drinker cases misreported their habit, declaring no consumption when in fact they were still drinking and consequently exhibited a higher risk than never drinkers.We were also interested in exploring the effect of the alcohol type since the consumption of wine, beer and spirits is quite common in Spain [33] and thus, it may have a great interest for public health. Most of the studies have supported the hypothesis that the amount of alcohol consumed is the determinant of the risk rather than the particular drink, however, a prospective cohort study in Denmark showed that a moderate intake of wine, compared to beer and hard liquor, did not increase the risk [5]. Previously, two population-based case-control studies in the United States did not find an association between wine drinking and the risk of esophageal cancer [34,35], and later, two other studies showed data supporting a lack of effect for wine drinking as well [6,7]. Conversely, in other studies where wine was by far the most common alcoholic beverage, wine drinkers showed higher risk than those of other beverages [23,36]. In our study, the consumption of all types of beverages that include hard liquors was a much stronger risk factor than wine and/or beer. We observed a statistically significant increased risk of EC among only beer drinkers but not among only wine or wine/beer drinkers. When the daily amount by type of beverage was considered, no increased risks were observed for a low-moderate wine-beer consumption (1–24 d/day), and a 2-fold increase risk among drinkers of ≥ 25g/d although it was not statistically significant. However, we observed a 5.76-fold increase risk among drinkers of ≥ 25 g/d of all types of beverages (risk for the combined categories of 25–74 g/d and ≥ 75 g/d of all types of beverages, table 2). A low-moderate wine-beer consumption of 1–24 g/d did not increase significantly the risk of esophageal squamous-cell carcinoma either. If the lack of effect of wine and/or beer is real, there may be some protective ingredients in wine such as resveratrol [37] and other antioxidants that may cause such a protective effect [38].It is not clear the exact mechanisms by which alcoholic beverages induce esophageal cancer risk. An ingredient common to all beverages is ethanol although it is possible that other components or contaminants such as N-nitrosamines and urethane with carcinogenic properties may increase cancer risk. It has been shown in practically all the studies that the risks are greatest for drinkers of hard liquors which is consistent with evidence that the concentration of ethanol plays an important role in alcohol-related tumours of the upper aero-digestive tract [13,39,40]. Although ethanol has not been shown as carcinogenic in laboratory animals, it may act through its major metabolite, acetaldehyde, a carcinogen in animal models [41,42]. Thus, it has been suggested that in addition to a systemic effect, ethanol can be converted to acetaldehyde in saliva and exert a promoting effect by either solubilising tobacco-specific carcinogens or enhancing their penetration into the esophageal mucosa, by nutritional deficiencies associated with heavy drinking or by other mechanisms (e.g., direct toxic or oxidative effect on the epithelial mucosa) [41,43]. In order to explain the differential effect of alcoholic beverages it should be considered that certain compounds such as N-nitrosamines are found in liquors and to lesser extent in beer, and urethane, a potential carcinogen in experimental studies, may be found in liquor but not in beer and wine. In addition, while most of the antioxidant compounds found in wine and to some extent in beer, are almost completely lacking in spirits [38].It has been also suggested that part of the effect observed for alcohol and/or tobacco smoking could be due to other factors or different lifestyles, such as a low fruit and vegetable intake [10]. However, our data were adjusted for the intake of fruit and vegetables showing that alcohol and tobacco were strong risk factors although the risk estimates were much higher when we did not adjust for fruits and vegetables intake (data not shown). We also observed that alcohol drinking was a much stronger risk factor for squamous-cell carcinoma than for the whole control series. This finding would indirectly support the hypothesis that the effect of alcohol for adenocarcinoma may be much weaker, if any. In fact, when we estimated the risk of adenocarcinoma for wine-beer drinkers and all types of alcohol beverages with respect to never drinkers, we did not find any increase of risk (OR = 0.83 and OR = 0.97, respectively); and for former and current drinker vs never drinkers, we did not find any significant risk either (OR = 1.83 and OR 0.74, respectively). Although some studies have found slight excesses of adenocarcinoma among drinkers [34,35,44], most of the previous case-control studies between alcohol drinking and the risk of oesophageal adenocarcinoma have generally found no association [6,7,45-47]. Unfortunately, the number of adenocarcinoma cases in our study was too small (n = 42) to allow us a separate analysis in great detail. The reasons for the differential effect of alcohol drinking on oesophageal squamous cell-carcinoma and adenocarcinoma should be further examined.Concerning tobacco smoking, the effect estimates were lower than the observed for alcohol intake although dose-response trends were still evident for the amount and the duration of smoking. Although the effect of tobacco smoking has been considered by practically all studies on esophageal cancer only a few have explored the effects by type of tobacco black/blond. The increased risk of black tobacco compared to blond tobacco is consistent with ecological studies showing a relatively high incidence of cancers of upper aero-digestive tract in southern Europe and Latin America, where this kind of tobacco is mainly consumed [48]. Our results are consistent with other studies carried out in high-risk areas of South America [4,49]. We observed a statistically significant effect for black tobacco but not for blond tobacco although the lack of effect for low-to-moderate blond tobacco smoking was based on a small number of cases (n = 31). It has been suggested that the higher concentrations of some carcinogenic compounds such as N-nitrosamines found in smoke of black tobacco could be a mechanism involved [43].Unlike our findings for alcohol cessation, we observed a beneficial effect of cessation of smoking. Similarly to others [4,6], we found that the risk of esophageal cancer declined within a decade of smoking cessation which may suggest that smoking could act during later stages as a promoter in the development of esophageal cancer, predominantly for the esophageal squamous-cell carcinoma.In order to explore the joint effects of alcohol and tobacco, unfortunately we could not perform interaction tests because of the small numbers. There was evidence that the effect of both exposures were nearly multiplicative. If we accept a multiplicative effect, we could assume that the risk of being simultaneously a heavy drinker (≥ 75 g/d) and heavy smoker (>50 p/y) could be as high as 40 times that of abstainers, or even 100 times when referring to esophageal squamous-cell carcinoma.As in other case-control studies, this study may present limitations. The sample size was small, and we could not explore interaction between tobacco and alcohol in more detail but the effects of each exposure became evident even among those non-exposed to the other. Another limitation may relate to the use of hospital controls. It is known that controls in hospital-based studies may be heavier smokers and heavier drinkers compared with general population; but we used a wide criterion to select controls from diagnosis a priori not related to the main risk factors of our study. In fact, the prevalence of drinking or smoking among control subjects was similar to that observed in the adult general population of the same sex and age composition of Alicante and Valencia, and thus it could be considered as representative of that population[33]. The presence of other types of bias were minimized by selecting controls from the same hospital as the cases, using the same interviews and procedures in both cases and controls to avoid any differential misclassification. The strength of the associations, the existence of a dose-response, the reduction of effect after cessation, the control for other potential confounder such as education and the intake of fruits and vegetables, and the consistency with other studies would support that the study results are real, and that alcohol drinking and tobacco smoking are likely to be causally related to esophageal cancer.ConclusionIn conclusion, this case-control study shows that the risk of esophagus cancer, and particularly the squamous cell type, is strongly associated with alcohol drinking. The consumption of any combination of hard liquors seems to be harmful whereas a low consumption of only wine may not. Tobacco smoking is also a strong risk factor, black more than blond. Smoking cessation was shown a beneficial effect within ten years whereas drinking cessation was not. A possible differential effect of alcohol drinking on oesophageal squamous cell-carcinoma (harmful) and adenocarcinoma (no effect) should be further examined.Competing interestsThe authors declare that they have no competing interestsAuthors' contributionsJV, Principal Investigator, conceived, designed and coordinated the study, performed the statistical analysis and drafted the manuscript; XB, performed the statistical analysis; FB, MP, MS and MGdlH made substantial contributions to the interpretation of data and drafting the manuscript; EM-O, made substantial contributions to acquisition of data, verifying the diagnoses for case series and drafting the manuscript. All the authors read and approved the final manuscript.Pre-publication historyThe pre-publication history for this paper can be accessed here:\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2529402.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2529402",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2529402\nAUTHORS: Rong-Xia Li, Hai-Bing Chen, Kang Tu, Shi-Lin Zhao, Hu Zhou, Su-Jun Li, Jie Dai, Qing-Run Li, Song Nie, Yi-Xue Li, Wei-Ping Jia, Rong Zeng, Jia-Rui Wu\n\nABSTRACT:\nBackgroundRecent advances in proteomics have shed light to discover serum proteins or peptides as biomarkers for tracking the progression of diabetes as well as understanding molecular mechanisms of the disease.ResultsIn this work, human serum of non-diabetic and diabetic cohorts was analyzed by proteomic approach. To analyze total 1377 high-confident serum-proteins, we developed a computing strategy called localized statistics of protein abundance distribution (LSPAD) to calculate a significant bias of a particular protein-abundance between these two cohorts. As a result, 68 proteins were found significantly over-represented in the diabetic serum (p<0.01). In addition, a pathway-associated analysis was developed to obtain the overall pathway bias associated with type 2 diabetes, from which the significant over-representation of complement system associated with type 2 diabetes was uncovered. Moreover, an up-stream activator of complement pathway, ficolin-3, was observed over-represented in the serum of type 2 diabetic patients, which was further validated with statistic significance (p = 0.012) with more clinical samples.ConclusionsThe developed LSPAD approach is well fit for analyzing proteomic data derived from biological complex systems such as plasma proteome. With LSPAD, we disclosed the comprehensive distribution of the proteins associated with diabetes in different abundance levels and the involvement of ficolin-related complement activation in diabetes.\n\nBODY:\nIntroductionDiabetes mellitus (DM) is one of the most common metabolic disorders in the world, in which more than 90% are grouped to type 2 diabetes mellitus (T2DM) [1]. Given the predicted explosion in the number of T2DM cases worldwide [2], the biomedical researchers face much stronger challenges, particularly on understanding the pathogenesis of disease and discovering biomarkers for tracking the disease process.T2DM is characterized by abnormal glucose homeostasis leading to hyperglycemia, and the serum glucose has been used as a golden standard for diabetes diagnosis. However, T2DM is a kind of disease involving defects of multiple organs, which cannot be distinguished through the measurement of the serum-glucose level. In addition, T2DM is a multiple-stage disease, which usually covers several decades from impaired plasma glucose to various complications. The serum-glucose level only reflects the consequence of multiple physiological disorders in the given stage. Therefore, many efforts have been made to identify genetic and protein markers to reveal the molecular/cellular details or progression of diabetes [3]–[9]. The genetic defects certainly render more probability to diabetes. On the other hand, the protein markers can track real-time status of diabetes. It has been found there are changes in the protein abundances of serum in diabetes progression [10], [11]. For instance, a number of studies suggest that the elevated circulating inflammatory biomolecules such as C-reactive protein and serum amyloid A can be used for predicting the development of T2DM [12]–[15]. However, since the traditional strategy of diabetic diagnosis only relies on the individual molecules as the biomarkers, the sensitivity and accuracy of the biomarkers might be fluctuated due to ethnic or personal variance [16]–[18]. Proteomic technology might provide the new solutions for solving this problem, which can identify large set of the proteins in cells or tissues through high-throughput methods, and provide a globe view of the protein changes associated with diabetes.It is well known that serum severs the optimal resource for discovery of disease biomarkers. Up to now, a few proteomic analyses of serum related to diabetes have been reported. For example, Dayal B et al. used the protein-chip to identify the high-density lipoproteins apoA-I and apoA-II and their glycosylated products in patients with diabetes and cardiovascular disease [19]. Zhang et al. found that the protease inhibitors including clade A and C, alpha 2-macroglobulin, fibrinogen, and the proteins involved in the classical complement pathway such as complement C3, and C4 exhibited the higher expression-levels in insulin resistance/type-2 diabetes [20]. Bergsten et al. analyzed the serum proteins in T2DM by SELDI-TOF-MS and peptide-mass fingerprinting (PMF), and found the expression levels of apolipoprotein, complement C3 and transthyretin were over-represented, whereas albumin and transferrin were under-represented in T2DM [21].However, none of these above works provided the real globe view for the protein profile of the diabetic serum, since the proteomic analysis of serum is a formidable challenge for its huge complexity and dynamic range [11], [22]. Recent advances in serum sample preparation such as a depletion of high abundance proteins can be coupled to 1D or 2D-LC-MS/MS analysis, which have provided the new ways for large-scale serum proteomic analysis [23]–[25]. However, the step of the depletion of the high abundance proteins might cause some artifacts. In the present study, we used a label-free proteomic method with LC-MS/MS to investigate the protein profiling between the non-diabetic and diabetic serum without removing the high abundant proteins. After analyzing the proteomics data according to the stringent criteria, a total of 3,010 proteins and 3,224 proteins were identified from the non-diabetic and diabetic serum, respectively. In-depth bioinformatic analysis was employed for these differential proteins between the non-diabetic and diabetic serum.ResultsSelection of non-diabetic subjects and diabetic patientsPrevious studies observed that T2DM might occurred at a greater frequency in adults who are younger than 65 years, suggesting that people who are old than 65 without diabetes mellitus usually do not anticipate the genetic susceptibility [26]. Therefore, we set age criteria for sample cohort that an adult in the present study must be old than 65 years (Non-diabetic subjects: age 67.6±1.67 years; type 2 diabetic patients: age 67±1.71 years) in order to reduce the genetic effects related to T2DM between non-diabetic and diabetic cohort. Furthermore, the careful selection of samples was performed based on the clinical parameters of non-diabetic and diabetic cohorts. Supplementary Table S1 summarized the clinical parameters of the selected non-diabetic subjects and diabetic patients, in which type 2 diabetic patients group had higher FPG, PG2H, WT, BMI, HOMA, HbA1c and C-peptide compared with control. To reduce the individual variance of serum proteins within the cohort, we pooled all the serum of each cohort for proteomic analysis, respectively.Semi-quantitative proteomic identification in non-diabetic and diabetic serumWe analyzed differential protein profile in two cohorts using shotgun proteomics and label-free quantitative strategy. In order to reduce sample complexity, proteins in non-diabetic and diabetic serum were first separated on SDS-PAGE gel and the gel bands were excised and subjected to in-gel tryptic digestion, respectively (Figure 1A). The proteins were identified with criteria corresponding to an estimated false dicovery rate of 2.5%. After combining the MS/MS data generated from our experiment, we were able to assign 1,212,256 MS/MS spectra to 150,881 peptide counts, leading to identification of 5,882 unique peptides corresponding to 3,010 protein groups in non-diabetic serum, and 1,211,006 MS/MS spectra to 189,792 peptide counts, resulting in 5,960 unique peptides corresponding to 3,224 protein groups in diabetic serum (all these identified protein groups are called proteins in the text below for clarity). Supplementary Figure S1 showed the quite similar distributions of the identified peptides and proteins between non-diabetic and diabetic serum, indicating non-bias of the identified MS/MS spectra between non-diabetic and diabetic serum.10.1371/journal.pone.0003224.g001Figure 1Overview of Idnetitication of proteins in non-diabetic and diabetic serum.(A) Scheme of label-free strategy to differential protein identification in non-diabetic and diabetic serum. Pooled serum samples from five non-diabetic and five diabetic sera were separated respectively by gel electrophoresis. Each gel lane was divided into 42 regions and each section was processed for mass spectrometry. (B) 1377 proteins were identified by at-least two peptide spectral counts in either serum. 888 overlapped proteins were identified both in non-diabetic and diabetic serum, whereas 223 proteins were identified uniquely from the non-diabetic serum and 266 proteins were found uniquely from the diabetic serum.Among the identified 3,010 proteins in non-diabetic serum and 3,224 proteins in diabetic serum, 942 (30.30%) and 1,046 (32.44%) proteins were selected respectively under the condition that each identified protein contained at least two peptide spectral counts. Totally 1,377 proteins were obtained according to these more stringent filter, resulting the false discovery rate of 1.6%. There were 888 identified proteins overlapped in non-diabetic and diabetic serum, whereas 223 proteins were identified uniquely from the non-diabetic serum and 266 proteins were found uniquely from the diabetic serum (Figure 1B, Supplementary Table S2).Localized statistics of protein abundance distribution (LSPAD)Since the peptide-spectral-count distributions of identified 1377 serum-proteins were widely spread out to the range of 105 (Supplementary Table S2), we developed M-A plotting referring to microarray analysis in order to display a relative protein-abundance distribution of each protein. First, for each protein, X1 representing its peptide spectral counts in diabetic serum was transformed into Y1 with formula f(X1) = log2(X1+1) as diabetic protein abundance, while the X2 in non-diabetic serum was transformed into Y2 with the same formula as a non-diabetic protein abundance. Then, we defined “M” as differential protein abundance between diabetic and non-diabetic serum by the formula of Y1−Y2, and “A” as an average protein abundance by the formula of (Y1+Y2)/2. Based on these formulas, total 1377 proteins were plotted as a scatter chart, in which the values of M were distributed on the Y-axis, and the values of A were distributed on the X-axis (Figure 2A).10.1371/journal.pone.0003224.g002Figure 2Quantitative strategy of proteins in non-diabetic and diabetic serum.(A) M-A plotting of 1377 identified proteins. “M” was defined as differential protein abundance ratios of each protein between diabetic and non-diabetic serum, and “A” was defined as protein-abundance of each protein. In addition, ret dots represented statistically significant over-represented proteins in diabetic serum, green dots represented statistically significant under-represented proteins in diabetic serum, and grey dots were proteins without statistically-significant change in diabetic serum and non-diabetic serum. (B) The distribution profiles of 1377 identified proteins (black line), identified proteins with M less than 5 (red line), between 5 and 10 (green line), and more than 10 (blue line).This scatter chart showed that the log2-ratio-range of the differential protein-abundances between non-diabetic and diabetic serum was considerably decreased along M-axis when the protein-abundances were increased along A-axis (Figure 2A). These observations indicated that the abundance ratio based on peptide spectral counts cannot be simply used as indicators for differential significance between diabetic and non-diabetic serum. For example, the significance of 2-fold change from 2 to 1 peptide spectral counts is not equal to the significance of 2-fold change from 20000 to 10000. In addition, we realized that the protein-distribution profiles at the low, middle and high level of protein abundance, respectively, were considerably different (Figure 2B), suggesting significance-calculation of particular differential proteins should be localized to a certain range of related abundance level. Therefore, we developed a computing method called Localized Statistics of Protein Abundance Distribution (LSPAD) to evaluate the statistical significance of protein-abundance bias between diabetic and non-diabetic serum, by which the differentia significance of a particular protein should be calculated through its local protein-abundance distribution-window rather than through whole distribution range from the lowest to highest protein-abundances. Since the whole distribution range of protein abundances could be generally subdivided into three parts (high, middle and low protein-abundances, see Figure 2 and Supplementary Table S2), we postulated a width of the local window for statistics as 33%, i.e. only neighbored proteins with A value located within the 33% A-axis around a particular protein should be used for calculation.In detail, for a particular protein, all the average peptide spectral counts of neighbored proteins whose A value were within the 33% abundance-window of the target protein were calculated as a background to evaluate the statistical significance (p value) of over-representation or under-representation of the target protein by performing fisher's exact test on a following four-fold table:\nD\n\nND\nPeptide spectral counts of a target proteinX1\nX2\nSum of counts of all the other proteins in the windowS1\nS2\nThe p-values derived from the fisher's exact test were linearly transformed into p′ in order to evaluate the bias of each protein-abundance between diabetic and non-diabetic serum. (sgn = 1 indicates that a protein is over-represented in diabetic sample, and sgn = −1 indicates that a protein is over-represented in non-diabetic sample)To evaluate the reliability of LSPAD, we carried out the MA-plotting analyses to two duplicates of diabetic serum sample. First, the duplicates of one pooled diabetic-serum sample were separated by SDS-PAGE, and the entire gel was cut into 12 gel slices for LC-MS/MS analysis (Supplementary Figure S2A). The results showed the consistent proteomic data from these two duplicates (Supplementary Figure S2B–E). Then these data were subjected to LSPAD analysis. The result showed few protein-variants by comparing the protein-abundances between two duplicates of one pooled diabetic-serum sample with LSPAD method (Supplementary Figure S3A). Furthermore, we analyzed the expression-differentiation significance of one diabetic-serum duplicate versus a non-diabetic serum (Supplementary Figure S3B), and the other diabetic-serum duplicate versus the same non-diabetic serum (Supplementary Figure S3C). The Supplementary Figure S3D showed the high correlation coefficient of the M values between the significantly differential proteins in Supplementary Figure S3B and S3C. Taken together, these results indicate that this LSPAD method is reliable for exploring the differentiation of the protein abundances between non-disease and disease serum.Accordingly, after 42 gel bands were analyzed in diabetic and non-diabetic serum respectively (Figure 1), 1377 identified proteins were analyzed by LSPAD approach. All the significant abundance-biases of 1377 proteins were calculated (Supplementary Table S2). Furthermore, we marked the proteins with p′<0.01 in red color as the significantly over-represented in diabetic serum, the proteins with p′>0.99 in green color as the significantly under-represented in diabetic serum, and the non-significantly differential proteins in grey color (Figure 2).The 68 significant over-represented proteins in diabetic serum were listed in Table 1. Many known risk factors of diabetes such as C-reactive protein, serum amyloid A and haptoglobin were over-represented in diabetic serum, in agreement with the observations by traditional approaches based on the analysis of individual proteins [27]. In addition, a number of other factors including the novel proteins associated with diabetes were detected by this large-scale survey (Table 1). On the other hand, 74 proteins were found under-represented in diabetic serum (Supplementary Table S2). As far as we know, some studies reported that Keratin and IgG were associated with diabetes [28], [29]. In addition, a lot of keratins were also involved in the pathway of cell communication (Supplementary Figure S4) in our results. According to our pathway-associated differential significance analysis, we found keratin associated pathway were significantly overall bias with diabetic serum, which might not result from the bias of sample preparation.10.1371/journal.pone.0003224.t001Table 1Characterization of proteins significantly over-represented in diabetic serum compared to non-diabetic serum based on LSPAD method. (P<0.01).IPI IDProtein nameDiabetic peptide spectral countNon-diabetic peptide spectral countP valueIPI00022434ALB protein61457470824.09E-91IPI00514824Complement component C4B8751831.44E-80IPI00555805Complement component 4A389621091.63E-69IPI00032258Complement C4 precursor384620773.06E-69IPI00453459Complement Component 4B preproprotein393321419.77E-69IPI00418163C4B1381120772.48E-66IPI00384697ALB protein47105373236.64E-37IPI00556148Complement factor H273216911.30E-30IPI00465313Alpha 2 macroglobulin variant17016130137.50E-26IPI00478003Alpha-2-macroglobulin precursor17344133353.06E-24IPI00385264Ig mu heavy chain disease protein16148804.42E-23IPI00164623Complement C3 precursor975472678.64E-22IPI00479708Immunoglobulin heavy constant mu (IGHM)200712041.02E-21IPI00549273Immunoglobulin heavy constant mu (IGHM)199511903.09E-21IPI00019943Afamin precursor5532211.57E-20IPI0047916965 kDa protein193211812.35E-18IPI00022488Hemopexin precursor195212682.99E-14IPI00426051Hypothetical protein DKFZp686C15213520338356.18E-14IPI00021727C4b-binding protein alpha chain precursor6383211.01E-13IPI00478493Haptoglobin precursor421431007.28E-12IPI00550991Alpha-1-antichymotrypsin precursor10886282.99E-11IPI00019591Splice Isoform 1 of Complement factor B precursor11836964.42E-11IPI00021842Apolipoprotein E precursor3941813.28E-10IPI00019399Serum amyloid A-4 protein precursor143439.21E-10IPI00021857Apolipoprotein C-III precursor144493.87E-08IPI00022392Complement C1q subcomponent, A chain precursor103301.25E-07IPI00021841Apolipoprotein A-I precursor406931122.14E-07IPI00010865Casein kinase II beta subunit2302.70E-07IPI00396929PREDICTED: similar to immunoglobulin M chain165681.55E-06IPI00410714Alpha 2 globin variant4332443.33E-06IPI00163446The Human Immunoglobulin Heavy Diversity (IGHD)134534.03E-06IPI00171834Keratin, type I cytoskeletal 19140571.29E-05IPI00399007Hypothetical protein DKFZp686I04196511440391.41E-05IPI00003590Quiescin Q61504.53E-05IPI00022389Splice Isoform 1 of C-reactive protein precursor1504.53E-05IPI00015309Keratin, type I cytoskeletal 1289337.63E-05IPI00290077Keratin, type I cytoskeletal 15142628.21E-05IPI00217963Keratin, type I cytoskeletal 162231170.000146102IPI00418422The Human Immunoglobulin Heavy Diversity (IGHD)69230.000152193IPI00423461Hypothetical protein DKFZp686C022208285480.000223242IPI00450768Keratin, type I cytoskeletal 17147690.000275352IPI00011261Complement component C8 gamma chain precursor2661520.000391696IPI00556567Ficolin-3 protein80330.000819734IPI00441196Hypothetical protein309024500.000949718IPI00386839Amyloid lambda 6 light chain variable region SAR180980.001229635IPI00017601Ceruloplasmin precursor226018160.001476612IPI00383953VH4 heavy chain variable region precursor132640.001483067IPI00009866Keratin, type I cytoskeletal 13107520.001918932IPI00470798Hypothetical protein DKFZp686E23209450836470.002098573IPI00017530Ficolin-2 precursor900.002266054IPI00021854Apolipoprotein A-II precursor8535820.002359293IPI00004550Hypothetical protein FLJ2026196450.00238822IPI00011252Complement component C8 alpha chain precursor81360.002415224IPI00293898Hepatocellular carcinoma associated protein TB61940.002727717IPI00384444Keratin, type I cytoskeletal 142071200.003122976IPI00021856Apolipoprotein C-II precursor32110.00418406IPI00219806S100 calcium-binding protein A7800.004391148IPI00446354Hypothetical protein FLJ41805800.004391148IPI00479762115 kDa protein800.004391148IPI00022446Platelet factor 4 precursor82390.00501026IPI00300725Keratin, type II cytoskeletal 6A158900.005161139IPI00242956Fc fragment of IgG binding protein2480.006549075IPI00384401Myosin-reactive immunoglobulin kappa chain variable region2580.006595492IPI00293665Keratin, type II cytoskeletal 6B141790.00706398IPI00299145Keratin, type II cytoskeletal 6E144830.007903541IPI00383603Anti-thyroglobulin light chain variable region700.008537501IPI00452748Serum amyloid A protein precursor700.008537501IPI00021304Keratin, type II cytoskeletal 2 epidermal8105750.009876282Pathway-associated differential significance analysisTo further reveal the significant bias of the protein abundances at the level of biological pathways in diabetic serum, we mapped those 1377 proteins into KEGG pathways [30]. Total 1377 identified proteins in the present study involved in 147 related pathways (Supplementary Table S3). Then, we calculated these proteins with their p-values at the pathway level in order to discover overall bias of pathways associated with diabetic-serum. The calculation procedure was as follows: Supposing all the proteins are non-differential expressed and independent of each other, their p-values, p, should follow a uniform distribution between[0,1]. Thus, z = qnorm(p), should follow a standard normal distribution (here qnorm is normal inverse distribution function). After the normal inverse transformation of pi to zi, the summarized Z score for a certain pathway j was generated by the formula, . Here nj was the number of the proteins involved in the pathway j in our experiments, and ix = {ixi} denoted the index of these proteins. Because the proteins involved in the pathway j were supposed to be non-differential expressed and independent of each other, the summarized score for pathway j, Zj, should also follow a standard normal distribution. In our case, for pathway j, the following hypothesis test was performed:H0: Zj follows the standard normal distribution, indicating that the pathway is not un-biased in diabetic serum.H1: Zj doesn't follow the standard normal distribution, indicating that the pathway is over-represented or under-represented in diabetic serumP value for pathway j, Pj, was transformed from Zj by a normal cumulative function, p = pnorm(z). Under a statistic significance threshold α, an over-represented pathway in diabetic serum was identified with  and under-represented pathway was identified with . If the P value is less than 0.01, it indicates that this pathway is significantly overall overrepresented in diabetic serum compared with non-diabetic serum. If the P value is more than 0.99, it means that this pathway is significantly overall overrepresented in non-diabetic serum.Among the 147 pathways, we selected 18 pathways, in which each pathway should have at least 5 identified proteins as well as more than 10% coverage of all the pathway-proteins in the database, to evaluate the pathway-bias between non-diabetic and diabetic serum. All the values of the protein-abundance biases in these 18 pathways were presented in Supplementary Figure S4. Particularly, the pathways of complement system, PPAR system, cell communication and Alzheimer's disease showed the significantly overall over-representation in diabetes serum (p<0.01), while insulin signaling, coagulation cascade, focal adhesion and long-term pathways presented significantly overall bias in non-diabetic serum (p>0.99) (Figure 3).10.1371/journal.pone.0003224.g003Figure 3The overall bias analysis of selected pathways found in non-diabetic and diabetic serum.Proteins identified in non-diabetic and diabetic serum were mapped to known pathways using KEGG. The p value of each pathway was digitized to the length of the bar diagram.These significant differential pathways could be subdivided into two major categories: one had many significant-differential components in one pathway, and the other had a few highly significant-differential components in one pathway. For example, on the PPAR pathway, three apolipoproteins were all over-represented significantly in diabetic serum (Figure 4A). In Alzheimer's disease pathway, the apoliprotein E over-presentation also contributed the overall bias of this pathway to diabetic serum. Therefore, apolipoproteins could be considered as a kind of the important biomarkers associated with diabetes. As previous reports, many apolipoproteins were involved in lipid metabolism [31]–[43]. These proteins were further selected to show their abundance biases between non-diabetic and diabetic serum. As shown in Figure 4B, 8 proteins including apolipoprotein A-I, AII, C-II and C-III were significantly over-represented in diabetic serum, whereas 6 proteins were significantly under-represented in diabetic serum, which covered some regulatory factors such as paraoxonase 1 (PON1) in lipid metabolism.10.1371/journal.pone.0003224.g004Figure 4The identified proteins and abundance biases in specific pathways.(A)PPAR system, (B) Apolipoproteins associated Lipid metabolism. The p value of identified protein was digitized to the length of the bar in each pathway.Over-representation of ficolin-related complement pathway in diabetic serumOur results showed that 12 proteins associated with complement system were significantly over-represented in diabetic serum (Figure 5A). It has been known that the complement system can be activated through three different ways, including lectin, classical and alternative pathways (Figure 5B) [44], [45]. The present work showed that two trigger factors of lectin-complement activation, ficolin-2 and ficolin-3, were both over-represented significantly in the diabetic serum (Table 1), while the ficolin-3 was detected with much higher abundance than ficolin-2. Another kind of lectin related to complement initiation, mannose biding lectins (MBL), was not detected. These results indicate that ficolin-3 might be the major trigger of lectin-complement activation in diabetic patients.10.1371/journal.pone.0003224.g005Figure 5Overview of proteins associated with complement system.(A) The identified proteins and the abundance biases in complement system. The p value of identified protein was digitized to the length of the bar in each pathway. (B) The three activation pathways of complement system: the classical, mannose-binding lectin, and alternative pathways. The three pathways converge at the point of cleavage of C3. Therefore, the C3 cleavage is the crucial step in activation of the three complement pathway. Molecules of C3 are cleaved to C3a and C3b by the C3 convertase. C3b binds covalently around the site of complement activation. Some of this C3b binds to the C4b and C3b in the convertase enzymes of the classical and alternative pathways, respectively, forming C5 convertase enzymes. This C3b acts as an acceptor site for C5, which is cleaved to form the anaphylatoxin C5a and C5b, which initiates the formation of the membrane-attack complex. Excitedly, ficolin-3 is a biologically active protein of the lectin-complement activation in association with MBL-associated serine protease (MASP). In this figure, significantly up-regulated proteins in diabetic serum were denoted with red color, slightly up-regulated proteins in diabetic serum were denoted with light red color, significantly up-regulated proteins in non-diabetic serum were denoted with blue color, and slightly up-regulated proteins in non-diabetic serum were denoted with light blue color. Not identified proteins or the fragment of the complement component were denoted with light grey color.Validation of ficolin-3 related complement activation in diabetic serumWhen the complement system is activated, the complement C3 is cleaved to C3a and C3b, which is the common and crucial step in all complement activation pathways (as shown in Figure 5B, [46]). To validate the level of C3 and its activation, Western blotting for C3, corresponding fragment C3a and C3b were performed in the non-diabetic and diabetic serum. It was confirmed that these proteins were over-represented in diabetic serum (Figure 6).10.1371/journal.pone.0003224.g006Figure 6Western blot confirmation of the serum level of C3 (∼187 kD), C3a (∼9 kD), C3b (alpha' chain, ∼104 kD) and Ficolin-3 (∼34 kD).The Non-diabetic serum: the mixture of equal amount of serum from five non-diabetic subjects in Table 1, Diabetic serum: the mixture of equal amount of serum from five diabetic patients in Table 1.It has been known that lectin is one of the trigger to complement activation [46], [47]. Our studies identified 33 and 80 spectral peptide counts of ficolin-3 from non-diabetic and diabetic serum, respectively (Table 1). Among these detected peptides, two particular peptides (VVLLPSCPGAPGSPGEK and YAVSEAAAHK) were detected exclusively from diabetic serum (Figure 7A and 7B). Taken together, these findings indicate that ficolin-3 in diabetic serum are over-represented in diabetic serum. We further confirmed this observation by Western blotting (Figure 6).10.1371/journal.pone.0003224.g007Figure 7MS/MS spectra of representative peptides from ficolin-3 and validation of ficolin-3 up-regulation in diabetic sera.(A) VVLLPSCPGAPGSPGEK (B) YAVSEAAAHK (C) Western blot validation of the serum ficolin-3 level in the non-diabetic subjects and diabetic patients (n = 24, respectively) were conducted. N: non-diabetic serum; D: diabetic serum.In order to evaluate the correlation of ficolin-3 with diabetes, the protein-abundance of ficolin-3 was validated by Western blotting in additional clinical sera from 24 non-diabetic subjects and 24 diabetic patients (Supplementary Table S4). As shown in Figure 7C and Supplementary Figure S5, the level of serum ficolin-3 was 0.90±0.43 in non-diabetic sera and 1.43±0.87 in diabetic sera (p = 0.012). Taken together, these results suggest a ficolin-3 related complement activation in diabetic serum.DiscussionThe strategy for analyzing the highly dynamical range of protein abundancesIn this study, LC-MS/MS coupled with a label-free quantitative strategy was applied to analyze the differential serum-protein abundance profile between non-diabetic and diabetic patients. The label-free quantitation based on peptide-spectral counts offers a high-coverage identification of proteins, and then gives a comprehensive and rapid comparison to the differential proteins, especially to the plasma proteins [48]. Since the distribution range of the peptide-spectral counts of the serum-proteins was up to 105 (Supplementary Table S2), we applied M-A plotting method referring to microarray data-analysis for analyzing the effects of the different abundance-levels as well as the count-ratio of a particular protein between non-diabetic and diabetic serum (Figure 2A). From the Figure 2B, we realized that the lower the abundance-level of the peptide-spectral counts, the higher the deviation of the count-ratio. In this regard, we cannot fix a count-ratio as a threshold covering low abundance-level to high abundance-level for evaluating the bias of the protein abundance of diabetic serum. In other words, the quantitative selection of differentia proteins based on the ratio of the particular protein-abundance, which is usually used in isotope-labeling proteomic methods, seems not suitable in the peptide-spectral counts quantification for the systems with the highly dynamic range of protein-abundances, i.e. serum proteome.Therefore, we developed a localized statistics of protein abundance distribution (LSPAD) for identifying the over- or under-represented proteins in diabetic serum. Based on this method, we can calculate the significance of the peptide-spectral-count bias for differentia proteins instead of using the count-ratio. Furthermore, we defined an abundance-window of 33% around a target protein as a localized background for calculating the statistical significance, by which we can evaluate the significant bias of a target protein-abundance compared to the abundance-distribution range of its neighbored proteins rather than to the abundance-distribution range of all identified proteins.Involvement of lipid metabolism and inflammation in type 2 diabetesIn this study, many individual proteins associated with T2DM reported in previous studies were also identified. In the group of apolipoproteins, for example, many components were over-represented in diabetic serum including Apolipoprotein E, CII, CIII and serum amyloid. Apo E content of postprandial TG-rich lipoproteins in subjects with both T2DM and coronary artery disease was increased [49]. Serum amyloid A, a major apoprotein (45%) in high-density lipoproteins (HDL), was increased due to inflammation [50]. Apolipoprotein C III (apo C III) plays a central role in regulating plasma metabolism of triglyceride-rich lipoprotein (TRL). Previous studies suggested that apo C III might be an independent risk factor for atherosclerotic diseases in Chinese type 2 diabetes [51]. On the other hand, we identified some under-represented regulatory factors in lipid metabolism such as paraoxonase1 (PON1). PON1 is an anti-inflammatory enzyme, which participates in the prevention of low density lipoprotein (LDL) oxidation [52], [53]. Recently, Mackness et. al reported that high C-reactive protein and low paraoxonase1 in diabetes might be used as risk factors of coronary heart disease [53].We also found certain proteins associated with acute-phase response were over-represented in diabetic serum such as C-reactive protein [54], [55], serum amyloid A [56], haptoglobin [57], α-1-acid glycoprotein [12], ceruloplasmin [58] and Von Willebrand factor [59]. Recently, abundant scientific evidences suggested the elevated circulating inflammatory markers such as C-reactive protein could be used for the prediction of the development of T2DM [12]–[15]. Moreover, C- reactive protein was also as a biomarker for inflammation in uremia [60]. Studies also showed that haptoglobin and C-reactive protein were increased significantly in both diabetes and glucose intolerance [57]. There has been an explosion of interests that the chronic low-grade inflammation and the activation of the innate immune system were closely involved in the pathogenesis of T2DM [61].Complement activation in type2 diabetesCross-sectional study have demonstrated strong correlation between complement C3 and insulin resistance, which showed that C3 was associated with a increased risk of developing diabetes [47]. In the present study, the serum levels of C3 and its fragments C3a were over-represented in diabetic serum by western blot analysis, indicating the activation of complement system. Adipsin/complement factor D is a serine protease that is secreted by adipocytes into the bloodstream. Adipsin is deficient in several animal models of obesity [62]. In our study, the expressing level of adipsin was under-represented in diabetic serum. Lectin is also a trigger for complement activation. This process begins due to the binding of mannose-binding lectin (MBL) or ficolins with MBL-associated serine protease-2 (MASP-2), and leads to the formation of a C3 convertase [63]–[66]. Up to now, only a few evidences showed that the increased level of MBL can provide prognostic information in patients with T2DM [67]. In the present work, MBL was not detected by mass spectrometry in serum, while both ficolin-2 and ficolin-3 were detected over-represented in diabetic serum. However, ficolin-2 was uniquely identified in diabetic serum with only 9 spectral counts while ficolin-3 was detected with much higher spectral counts. Therefore, it seems that ficolin-3 should be the major trigger and indicator of lectin-complement activation. The Western-blotting of serum ficolin-3 with a lager clinical population supports that serum ficolin-3 was significantly over-represented and positively correlated with T2DM. Thus, we argue that ficolin-3 triggers the lectin-complement pathway, which might play an important role in the chronic low-grade inflammation and activation of the innate immune system associated with T2DM.In summary, the LSPAD approach developed in this present work is well useful for analyzing proteomic data derived from biological complex systems such as plasma proteome, by which we disclosed the comprehensive distribution of the proteins associated with diabetes among high, medium and low abundant proteins. In addition, we found the involvement of the ficolin-related complement system in type 2 diabetes.Materials and MethodsClinical sample collection and preparationTen male adults were selected for this investigation, including five non-diabetic subjects (FPG 4.82±0.21 mmol/L; PG2H 4.78±1.54 mmol/L; BMI 21.67±0.81 kg/m2; HbA1c 5.68±0.54%; C-peptide 1.09±0.25 ng/mL; and homeostasis model assessment [HOMA] 1.04±0.67), and five type 2 diabetic patients (FPG 7.26±2.73 mmol/L; PG2H 12.2±1.21 mmol/L; BMI 27.03±4.23 kg/m2; HbA1c 7.14±0.42%; C-peptide 3.44±1.31 ng/mL; HOMA 5.67±3.96). The Homeostasis Model Assessment (HOMA) for insulin resistance and β-cell function was calculated from fasting plasma glucose and insulin concentrations. Informed consent was obtained from each person in written format and approved by Shanghai No. 6 People's Hospital Review Committee.Immediately after collection, fasting blood samples were allowed to clot at room temperature for four hours, and the serum were collected and centrifugated at 3000 rpm/min for 15 min. Before pooling the samples, the protein concentration of the serum samples was determined by Bradford assay on a Microplate Reader (Bio-Rad, Model 680). Five non-diabetic serum samples were mixed as control-pool sample, and five diabetic serum samples were also mixed as disease-pool sample. The two pooled serum samples were diluted respectively to ∼20 mg/mL with 100 mM phosphate buffer (pH 2.0, containing 5% ACN). Then, the pooled serum samples were filtered through 0.22 µm filters (Agilent technologies) by spinning at 10 000 g at 4°C for 30 min and dialyzed to 100 mM phosphate buffer (pH 2.0, containing 5% ACN).Gel electrophoresis and In-Gel DigestionThe serum sample containing 1.8 mg proteins was reduced by adding 2 µL of 1 M DTT to 10 mM and incubated at 37°C for 2.5 hours. The mixture then was added with 10 µL of 1 M IAA and incubated for 40 min in darkness at room temperature. After these treatments, the samples were subjected to SDS-PAGE on a 7.5–17.5% gradient gel. The gel lane stained with Coomassie Blue was excised into 42 sections. Each excised section was cut into approx. 1 mm3 pieces and destained using 30% acetonitrile/70% 100 mM ammonium bicarbonate solution, followed by dehydration in 100% acetonitrile for 5 min. Gel pieces were placed under vacuum centrifugation until completely dried. Each gel slice was then incubated in a 50 mM ammonium bicarbonate solution containing 10 ng/µL trypsin (Promega Biotech Co., Madison, WI, USA.) overnight. Peptides were extracted with 0.1% TFA/80% acetonitrile, dried by vacuum centrifugation, and stored at −80°C for further analysis with mass spectrometry.Label-free shotgun proteomic identificationEach gel slice containing peptides was dissolved in 60 µL 0.1% formic acid, and then the half of this peptide-solution was loaded into the RP column. RP-HPLC was performed using an Agilent 1100 Capillary system (Agilent technologies) with C18 column (150 µm i.d., 100 mm length, Column technology Inc., Fremont, CA). The pump flow rate was 1.6 µL/min. Mobile phase A was 0.1% formic acid in water, and mobile phase B was 0.1% formic acid in acetonitrile. The tryptic peptide mixtures were eluted using a gradient of 2–55% B over 135 min. The mass spectral data were acquired on a LTQ linear ion trap mass spectrometer (Thermo, San Jose, CA) equipped with an electrospray interface operated in positive ion mode. The temperature of heated capillary was set at 170°C. A voltage of 3.0 kV applied to the ESI needle. Normalized collision energy was 35.0. The number of ions stored in the ion trap was regulated by the automatic gain control. Voltages across the capillary and the quadrupole lenses were tuned by an automated procedure to maximize the signal for the ion of interest. The mass spectrometer was set as one full MS scan was followed by ten MS/MS scans on the ten most intense ions from the MS spectrum with the following Dynamic Exclusion™ settings: repeat count, 2, repeat duration, 0.5 min, exclusion duration, 1.5 min.Data analysisAll .dta files were created using Bioworks 3.1, with precursor mass tolerance of 1.4 Da, threshold of 100, and minimum ion count of 15. The acquired MS/MS spectra were searched against the Human International Protein Index protein sequence database (version 3.07, www.ebi.ac.uk/IPI) combined with sequences of real protein and reverse sequences of proteins, by using the TurboSEQUEST program in the BioWorks 3.1 software suite, with a mass tolerance of 3.0 Da. All cysteine residues were searched as carboxamidomethycystein (+57.02 Da). Up to one internal cleavage sites were allowed for tryptic searches. All output results were combined together using the in-house software named BuildSummary to delete the redundant data. Searches were conducted against the Human International Protein Index protein sequence database to control the false discovery rate at 2.5% and all spectral peptide count had a ΔCn score of at least 0.1. The proteins identified by two or more peptide counts in either non-diabetic or diabetic serum were used to the following bioinformatics analysis.Western bolt analysis of C3 and its fragmentsEach of 100 µg non-diabetic and diabetic serum-proteins was subjected to PAGE-gel electrophresis, and then proteins in the gel were transferred to a nitrocellulose membrane. The membranes were incubated first with the appropriate primary antibodies (C3b: ab11871, C3a: ab11872, purchased from Abcam Ltd, Cambridge, MA), respectively, and then incubated with HRP-conjugated secondary antibodies for 45 min. The proteins were detected by enhanced chemiluminescence (ECL-plus, Amersham Pharmacia Biotech).Validation of ficolin-3 over-representation in larger samples0.4 uL of each individual serum sample (non-diabetic and diabetic subjects, n = 24, respectively) diluted to 1/10 with 1.0 M Tris (pH 6.8) buffer was separated by SDS-PAGE, and electro-transferred to a nitrocellulose membrane (Whatman International Ltd., England.). The membrane was blotted with a mouse monoclonal antibody against human ficolin-3 (R&D Systems, Inc., 1∶500). Signal detection was achieved with ECL Plus chemiluminescence system (Amersham Biosciences). Signal of bands from Western blot were scanned with PDQUEST GS-710 a flat-bed scanner and digitized with Gel-PRO Analyzer software (Media Cybernetics, Inc., USA). To decrease the system discrepancy, we used the serum of the same patient as the reference. Relative level of serum ficolin-3 was calculated by the proportion of density ratio of sample bands to that of the reference band. These density ratios were used for statistical analyses of serum ficolin-3 level between non-diabetic and diabetic subjects.Statistical analysisData were expressed as means±standard deviation (SD) for normally distributed values. Differences between groups for normally distributed variables were tested using t-test (analysis of variance). All calculations were performed with GraphPad Prism software system (GraphPad San Diego, CA, USA) and SPSS13.0 statistical package (Statistical Software, Los Angeles, CA, USA). A P value below 0.05 was considered statistically significant.Supporting InformationFigure S1The distribution of proteins and peptides identified in 42 gel slices of non-diabetic serum and diabetic serum(0.02 MB PDF)Click here for additional data file.Figure S2Reproducibility of Gel-LC-MS/MS separations and identification.(0.21 MB PDF)Click here for additional data file.Figure S3Reproducibility and reliability of LSPAD method(0.15 MB PDF)Click here for additional data file.Figure S4The identified proteins and abundance biases in 18 pathways(0.23 MB PDF)Click here for additional data file.Figure S5Western blot analyses of the serum ficolin3 level in the non-diabetic subjects(n = 24)and diabetic patients(n = 24)(0.14 MB PDF)Click here for additional data file.Table S1Baseline characteristics of five non-diabetic subjects and five diabetic patients(0.02 MB PDF)Click here for additional data file.Table S2Proteins identified by two or more peptide spectral counts in non-diabetic and diabetic serum(0.43 MB PDF)Click here for additional data file.Table S3Pathway analysis by mapping 1377 proteins into KEGG pathways. Ratio (%): (100 Ã? Gene number found in pathway) / Totallygene number in pathway. P value: present overall bias of pathways associated with diabetic-serum or non-diabetic serum(0.08 MB PDF)Click here for additional data file.Table S4General and clinical parameters of non-diabetic subjects and type 2 diabetic patients(0.05 MB PDF)Click here for additional data file.\n\nREFERENCES:\n1. KorcM\n2003\nDiabetes mellitus in the era of proteomics.\nMol Cell Proteomics\n2\n399\n404\n12851465\n2. ZimmetP\n2003\nThe burden of type 2 diabetes: are we doing enough?\nDiabetes Metab\n29\n6S9\n18\n14502096\n3. HorikawaYYamasakiTNakajimaHShinguRYoshiuchiI\n2003\nIdentification of a novel variant in the phosphoenolpyruvate carboxykinase gene promoter in Japanese patients with type 2 diabetes.\nHorm Metab Res\n35\n308\n312\n12916001\n4. KimEYShinCHYangSW\n2003\nPolymorphisms of HLA class II predispose children and adolescents with type 1 diabetes mellitus to autoimmune thyroid disease.\nAutoimmunity\n36\n177\n181\n12911285\n5. VendrellJFernandez-RealJMGutierrezCZamoraASimonI\n2003\nA polymorphism in the promoter of the tumor necrosis factor-alpha gene (-308) is associated with coronary heart disease in type 2 diabetic patients.\nAtherosclerosis\n167\n257\n264\n12818408\n6. LindgrenCMWidenETuomiTLiHAlmgrenP\n2002\nContribution of known and unknown susceptibility genes to early-onset diabetes in scandinavia: evidence for heterogeneity.\nDiabetes\n51\n1609\n1617\n11978663\n7. RaoAASridharGRDasUN\n2007\nElevated butyrylcholinesterase and acetylcholinesterase may predict the development of type 2 diabetes mellitus and Alzheimer's disease.\nMed Hypotheses\n8. RaoAASridharGRSrinivasBDasUN\n2007\nBioinformatics analysis of functional protein sequences reveals a role for brain-derived neurotrophic factor in obesity and type 2 diabetes mellitus.\nMed Hypotheses\n9. OnYKParkHKHyonMSJeonES\n2007\nSerum resistin as a biological marker for coronary artery disease and restenosis in type 2 diabetic patients.\nCirc J\n71\n868\n873\n17526982\n10. ScottEMCarterAMFindlayJB\n2005\nThe application of proteomics to diabetes.\nDiab Vasc Dis Res\n2\n54\n60\n16305059\n11. AndersonNLAndersonNG\n2002\nThe human plasma proteome: history, character, and diagnostic prospects\n845\n867\n12. SchmidtMIDuncanBBSharrettARLindbergGSavagePJ\n1999\nMarkers of inflammation and prediction of diabetes mellitus in adults (Atherosclerosis Risk in Communities study): a cohort study.\nLancet\n353\n1649\n1652\n10335783\n13. DuncanBBSchmidtMIOffenbacherSWuKKSavagePJ\n1999\nFactor VIII and other hemostasis variables are related to incident diabetes in adults. The Atherosclerosis Risk in Communities (ARIC) Study.\nDiabetes Care\n22\n767\n772\n10332679\n14. PradhanADMansonJERifaiNBuringJERidkerPM\n2001\nC-reactive protein, interleukin 6, and risk of developing type 2 diabetes mellitus.\nJama\n286\n327\n334\n11466099\n15. FestaAD'AgostinoRJrTracyRPHaffnerSM\n2002\nElevated levels of acute-phase proteins and plasminogen activator inhibitor-1 predict the development of type 2 diabetes: the insulin resistance atherosclerosis study.\nDiabetes\n51\n1131\n1137\n11916936\n16. PeetersABeckersSVerrijkenARoevensPPeetersP\n2007\nVariants in the FTO gene are associated with common obesity in the Belgian population.\nMol Genet Metab\n17. YeXYuZLiHFrancoOHLiuY\n2007\nDistributions of C-reactive protein and its association with metabolic syndrome in middle-aged and older Chinese people.\nJ Am Coll Cardiol\n49\n1798\n1805\n17466231\n18. NedelkovDKiernanUANiederkoflerEETubbsKANelsonRW\n2005\nInvestigating diversity in human plasma proteins.\nProc Natl Acad Sci U S A\n102\n10852\n10857\n16043703\n19. ZhangRBarkerLPinchevDMarshallJRasamoelisoloM\n2004\nMining biomarkers in human sera using proteomic tools.\nProteomics\n4\n244\n256\n14730686\n20. DayalBErtelNH\n2002\nProteinChip technology: a new and facile method for the identification and measurement of high-density lipoproteins apoA-I and apoA-II and their glycosylated products in patients with diabetes and cardiovascular disease.\nJ Proteome Res\n1\n375\n380\n12645894\n21. SundstenTEberhardsonMGoranssonMBergstenP\n2006\nThe use of proteomics in identifying differentially expressed serum proteins in humans with type 2 diabetes.\nProteome Sci\n4\n22\n17163994\n22. GeorgiouHMRiceGEBakerMS\n2001\nProteomic analysis of human plasma: failure of centrifugal ultrafiltration to remove albumin and other high molecular weight proteins.\nProteomics\n1\n1503\n1506\n11747208\n23. HePHeHZDaiJWangYShengQH\n2005\nThe human plasma proteome: analysis of Chinese serum using shotgun strategy.\nProteomics\n5\n3442\n3453\n16047309\n24. TuCJDaiJLiSJShengQHDengWJ\n2005\nHigh-sensitivity analysis of human plasma proteome by immobilized isoelectric focusing fractionation coupled to mass spectrometry identification.\nJ Proteome Res\n4\n1265\n1273\n16083276\n25. JinWHDaiJLiSJXiaQCZouHF\n2005\nHuman plasma proteome analysis by multidimensional chromatography prefractionation and linear ion trap mass spectrometry identification.\nJ Proteome Res\n4\n613\n619\n15822942\n26. KoopmanRJMainousAG3rdDiazVAGeeseyME\n2005\nChanges in age at diagnosis of type 2 diabetes mellitus in the United States, 1988 to 2000.\nAnn Fam Med\n3\n60\n63\n15671192\n27. GentlemanRCVHuberWIrizarryRDudoitS\n2005\nBioinformatics and Computational Biology Solutions Using R and Bioconductor: Statistics for Biology and Health\n473\n28. VirellaGCarterRESaadACrosswellEGGameBA\n2008\nDistribution of IgM and IgG antibodies to oxidized LDL in immune complexes isolated from patients with type 1 diabetes and its relationship with nephropathy.\nClin Immunol\n127\n394\n400\n18533284\n29. MirzaeiHBaenaBBarbasCRegnierF\n2008\nIdentification of oxidized proteins in rat plasma using avidin chromatography and tandem mass spectrometry.\nProteomics\n8\n1516\n1527\n18383005\n30. KanehisaMGotoS\n2000\nKEGG: kyoto encyclopedia of genes and genomes.\nNucleic Acids Res\n28\n27\n30\n10592173\n31. DuchateauPNPullingerCROrellanaREKunitakeSTNaya-VigneJ\n1997\nApolipoprotein L, a new human high density lipoprotein apolipoprotein expressed by the pancreas. Identification, cloning, characterization, and plasma distribution of apolipoprotein L.\nJ Biol Chem\n272\n25576\n25582\n9325276\n32. GetzGSReardonCA\n2004\nParaoxonase, a cardioprotective enzyme: continuing issues.\nCurr Opin Lipidol\n15\n261\n267\n15166781\n33. HellerMStalderDSchlappritziEHaynGMatterU\n2005\nMass spectrometry-based analytical tools for the molecular protein characterization of human plasma lipoproteins.\nProteomics\n5\n2619\n2630\n15892164\n34. KarlssonHLeandersonPTagessonCLindahlM\n2005\nLipoproteomics II: mapping of proteins in high-density lipoprotein using two-dimensional gel electrophoresis and mass spectrometry.\nProteomics\n5\n1431\n1445\n15761960\n35. KotiteLZhangLHYuZBurlingameALHavelRJ\n2003\nHuman apoC-IV: isolation, characterization, and immunochemical quantification in plasma and plasma lipoproteins.\nJ Lipid Res\n44\n1387\n1394\n12700345\n36. KunitakeSTCarilliCTLauKProtterAANaya-VigneJ\n1994\nIdentification of proteins associated with apolipoprotein A-I-containing lipoproteins purified by selected-affinity immunosorption.\nBiochemistry\n33\n1988\n1993\n8117655\n37. McVicarJPKunitakeSTHamiltonRLKaneJP\n1984\nCharacteristics of human lipoproteins isolated by selected-affinity immunosorption of apolipoprotein A-I.\nProc Natl Acad Sci U S A\n81\n1356\n1360\n6424116\n38. SprecherDLTaamLGreggREFojoSSWilsonDM\n1988\nIdentification of an apoC-II variant (apoC-IIBethesda) in a kindred with apoC-II deficiency and type I hyperlipoproteinemia.\nJ Lipid Res\n29\n273\n278\n3379339\n39. RezaeeFCasettaBLevelsJHSpeijerDMeijersJC\n2006\nProteomic analysis of high-density lipoprotein.\nProteomics\n6\n721\n730\n16419016\n40. XuNDahlbackB\n1999\nA novel human apolipoprotein (apoM).\nJ Biol Chem\n274\n31286\n31290\n10531326\n41. ZhangLHKotiteLHavelRJ\n1996\nIdentification, characterization, cloning, and expression of apolipoprotein C-IV, a novel sialoglycoprotein of rabbit plasma lipoproteins.\nJ Biol Chem\n271\n1776\n1783\n8576182\n42. O'BrienPJAlbornWESloanJHUlmerMBoodhooA\n2005\nThe novel apolipoprotein A5 is present in human serum, is associated with VLDL, HDL, and chylomicrons, and circulates at very low concentrations compared with other apolipoproteins.\nClin Chem\n51\n351\n359\n15528295\n43. NavabMAnanthramaiahGMReddySTVan LentenBJAnsellBJ\n2004\nThe oxidation hypothesis of atherogenesis: the role of oxidized phospholipids and HDL.\nJ Lipid Res\n45\n993\n1007\n15060092\n44. WhaleyKSchwaebleW\n1997\nComplement and complement deficiencies.\nSemin Liver Dis\n17\n297\n310\n9408965\n45. PascualMFrenchLE\n1995\nComplement in human diseases: looking towards the 21st century.\nImmunol Today\n16\n58\n61\n7888067\n46. OstergaardJHansenTKThielSFlyvbjergA\n2005\nComplement activation and diabetic vascular complications.\nClin Chim Acta\n361\n10\n19\n15996650\n47. EngstromGHedbladBErikssonKFJanzonLLindgardeF\n2005\nComplement C3 is a risk factor for the development of diabetes: a population-based cohort study.\nDiabetes\n54\n570\n575\n15677517\n48. VaisarTPennathurSGreenPSGharibSAHoofnagleAN\n2007\nShotgun proteomics implicates protease inhibition and complement activation in the antiinflammatory properties of HDL.\nJ Clin Invest\n117\n746\n756\n17332893\n49. SyvanneMRosseneuMLabeurCHildenHTaskinenMR\n1994\nEnrichment with apolipoprotein E characterizes postprandial TG-rich lipoproteins in patients with non-insulin-dependent diabetes mellitus and coronary artery disease: a preliminary report.\nAtherosclerosis\n105\n25\n34\n8155085\n50. PatrickLUzickM\n2001\nCardiovascular disease: C-reactive protein and the inflammatory disease paradigm: HMG-CoA reductase inhibitors, alpha-tocopherol, red yeast rice, and olive oil polyphenols. A review of the literature.\nAltern Med Rev\n6\n248\n271\n11410071\n51. ChenXTianHLiuR\n2004\nAssociation of serum apolipoprotein C III levels and apolipoprotein C III gene Sst I polymorphism with carotid intima-media thickness in Chinese type 2 diabetic patients.\nDiabetes Res Clin Pract\n66\n41\n47\n15364160\n52. RothemLHartmanCDahanALachterJEliakimR\n2007\nParaoxonases are associated with intestinal inflammatory diseases and intracellularly localized to the endoplasmic reticulum.\nFree Radic Biol Med\n43\n730\n739\n17664137\n53. MacknessBHineDMcElduffPMacknessM\n2006\nHigh C-reactive protein and low paraoxonase1 in diabetes as risk factors for coronary heart disease.\nAtherosclerosis\n186\n396\n401\n16140307\n54. ThorandBLowelHSchneiderAKolbHMeisingerC\n2003\nC-reactive protein as a predictor for incident diabetes mellitus among middle-aged men: results from the MONICA Augsburg cohort study, 1984–1998.\nArch Intern Med\n163\n93\n99\n12523922\n55. FreemanDJNorrieJCaslakeMJGawAFordI\n2002\nC-reactive protein is an independent predictor of risk for the development of diabetes in the West of Scotland Coronary Prevention Study.\nDiabetes\n51\n1596\n1600\n11978661\n56. PickupJCMattockMBChusneyGDBurtD\n1997\nNIDDM as a disease of the innate immune system: association of acute-phase reactants and interleukin-6 with metabolic syndrome X.\nDiabetologia\n40\n1286\n1292\n9389420\n57. McMillanDE\n1989\nIncreased levels of acute-phase serum proteins in diabetes.\nMetabolism\n38\n1042\n1046\n2478861\n58. KimCHParkJYKimJYChoiCSKimYI\n2002\nElevated serum ceruloplasmin levels in subjects with metabolic syndrome: a population-based study.\nMetabolism\n51\n838\n842\n12077727\n59. BarravieraB\n1994\nAcute-phase response in snakebite.\nRev Inst Med Trop Sao Paulo\n36\n479\n7569618\n60. KaysenGA\n2001\nThe microinflammatory state in uremia: causes and potential consequences.\nJ Am Soc Nephrol\n12\n1549\n1557\n11423586\n61. HansenTKTarnowLThielSSteffensenRStehouwerCD\n2004\nAssociation between mannose-binding lectin and vascular complications in type 1 diabetes.\nDiabetes\n53\n1570\n1576\n15161763\n62. RosenBSCookKSYaglomJGrovesDLVolanakisJE\n1989\nAdipsin and complement factor D activity: an immune-related defect in obesity.\nScience\n244\n1483\n1487\n2734615\n63. TakahashiMMoriSShigetaSFujitaT\n2007\nRole of MBL-associated serine protease (MASP) on activation of the lectin complement pathway.\nAdv Exp Med Biol\n598\n93\n104\n17892207\n64. HansenTK\n2005\nMannose-binding lectin (MBL) and vascular complications in diabetes.\nHorm Metab Res\n37\nSuppl 1\n95\n98\n15918118\n65. ThielSVorup-JensenTStoverCMSchwaebleWLaursenSB\n1997\nA second serine protease associated with mannan-binding lectin that activates complement.\nNature\n386\n506\n510\n9087411\n66. Vorup-JensenTJenseniusJCThielS\n1998\nMASP-2, the C3 convertase generating protease of the MBLectin complement activating pathway.\nImmunobiology\n199\n348\n357\n9777418\n67. HansenTKGallMATarnowLThielSStehouwerCD\n2006\nMannose-binding lectin and mortality in type 2 diabetes.\nArch Intern Med\n166\n2007\n2013\n17030835"
+}

batch_8/PMC2530489.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2530489",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2530489\nAUTHORS: Camilo E. Villarroel, Cristina Villanueva-Mendoza, Lorena Orozco, Miguel Angel Alcántara-Ortigoza, Diana F. Jiménez, Juan C. Ordaz, Ariadna González-del Angel\n\nABSTRACT:\nPurposePaired box gene 6 (PAX6) heterozygous mutations are well known to cause congenital non-syndromic aniridia. These mutations produce primarily protein truncations and have been identified in approximately 40%–80% of all aniridia cases worldwide. In Mexico, there is only one previous report describing three intragenic deletions in five cases. In this study, we further analyze PAX6 variants in a group of Mexican aniridia patients and describe associated ocular findings.MethodsWe evaluated 30 nonrelated probands from two referral hospitals. Mutations were detected by single-strand conformation polymorphism (SSCP) and direct sequencing, and novel missense mutations and intronic changes were analyzed by in silico analysis. One intronic variation (IVS2+9G>A), which in silico analysis suggested had no pathological effects, was searched in 103 unaffected controls.ResultsAlmost all cases exhibited phenotypes that were at the severe end of the aniridia spectrum with associated ocular alterations such as nystagmus, macular hypoplasia, and congenital cataracts. The mutation detection rate was 30%. Eight different mutations were identified: four (c.184_188dupGAGAC, c.361T>C, c.879dupC, and c.277G>A) were novel, and four (c.969C>T, IVS6+1G>C, c.853delC, and IVS7–2A>G) have been previously reported. The substitution at position 969 was observed in two patients. None of the intragenic deletions previously reported in Mexican patients were found. Most of the mutations detected predict either truncation of the PAX6 protein or conservative amino acid changes in the paired domain. We also detected two intronic non-pathogenic variations, IVS9–12C>T and IVS2+9G>A, that had been previously reported. Because the latter variation was considered potentially pathogenic, it was analyzed in 103 healthy Mexican newborns where we found an allelic frequency of 0.1116 for the A allele.ConclusionsThis study adds four novel mutations to the worldwide PAX6 mutational spectrum, and reaffirms the finding that c.969C>T is one of the three more frequent causal mutations in aniridia cases. It also provides evidence that IVS2+9G>A is an intronic change without pathogenic effect.\n\nBODY:\nIntroductionAniridia is a congenital ocular disorder characterized by bilateral variable iris hypoplasia with an estimated occurrence of one in every 64,000–96,000 live births worldwide [1]. The manifestations of the aniridia phenotype are variable, ranging from thinning of the stroma and absent pupillary sphincter to complete aniridia [2,3]. In addition to iris hypoplasia, other ocular congenital defects may be present such as cataracts, foveal hypoplasia, nystagmus, corneal opacity, lens dislocation, and glaucoma with significant loss of vision [4]. Because of the wide spectrum of clinical manifestations associated with this ocular pathology, Gronskov et al. [5] proposed to categorize the phenotype into six different levels based on iris presentation. However, this classification is not widely used.Approximately two thirds of cases are familial with an autosomal dominant inheritance pattern, probably with complete penetrance [5,6]. Some sporadic aniridia cases have the WAGR syndrome (Wilms tumor, aniridia, genitourinary anomalies, and mental retardation; OMIM 194072). Several genes at 11p13 are deleted in the WAGR syndrome including WT1 and the evolutionarily conserved paired box gene 6 (PAX6) [7].The human PAX6 spans 26 kilobases (kb), contains 14 exons [8,9], and encodes the PAX6 transcription factor. PAX6 is considered the master control gene for ocular morphogenesis and contributes to central nervous system development [10]. Like other transcriptional activators of the PAX family, PAX6 contains two DNA-binding domains (a paired domain at the NH2-terminus and a middle homeodomain) and a proline-serine-threonine (PST)-rich transactivator domain at the COOH-terminus [8,9].Homozygous loss of PAX6 is thought to lead to early embryonic lethality [11]. Heterozygous mutations are found in approximately 40%–80% of all non-syndromic aniridia cases [9,12-15],  and most are searched by single strand conformation polymorphism (SSCP), which is considered one of the most useful molecular detection methods [12,16]. There are no clear gene hotspots, and the majority of mutations in PAX6 are predicted to introduce premature termination codons, most of which are assumed to be functionally null because of haploinsufficiency [15]. To date, more than 400 PAX6 mutations have been reported (Online Human PAX6 Allelic Database). The most frequent mutations are c.1080C>T (c.718C>T), c.969C>T (c. 607C>T), c.1311C>T (c.949C>T), and c.1629insT (c.1267dupT).The molecular basis of aniridia in Mexico is poorly characterized. In fact, there is only one report of three different intragenic deletions of PAX6 found in five unrelated cases in the Mexican population. Interestingly, the authors of this study suggested a founder effect for a four-base intragenic deletion (c.732_735delAACA) in exon 7 in Mexican aniridia patients because this mutation was found in three nonrelated cases [17]. In the present study, we further analyze PAX6 variants in a group of Mexican aniridia patients and describe associated ocular findings.MethodsWe evaluated 30 unrelated aniridia probands recruited from two referral hospitals in Mexico City, the National Institute of Pediatrics and the Dr. Luis Sanchez Bulnes Hospital. All individuals were of Mexican origin, showed no associated systemic abnormalities, and had normal psychomotor development. Patients were categorized according to Gronskov’s iris classification [5].This study was conducted in accordance with the World Medical Association Declaration of Helsinki and was approved by the respective local research and ethics committees. Written informed consent was obtained from all participants.Genomic DNA was extracted from peripheral blood leukocytes using the PureGene DNA purification kit (Gentra Systems, Minneapolis, MN). PAX6 mutation screening was performed by polymerase chain reaction (PCR) amplification of all 14 exons and immediate flanking sequences using the primers and conditions proposed by Love et al. [18] followed by SSCP analysis in 1X Mutation Detection Enhancement gels (BioWhittaker Molecular Applications, Rockland, ME). Gels were run under constant power (6 W) for 12 h at room temperature and visualized by silver nitrate staining (Silver Stain Kit, Bio-Rad Laboratories, Hercules, CA). Fragments displaying abnormal electrophoretic patterns were purified by the silica column method (QIAquick, Gel Extraction Kit; QIAGEN Inc. Valencia CA) and directly sequenced using a Big Dye Terminator Kit with an automated ABI PRISM Model 377 sequencer (Applied Biosystems, Foster City, CA) according to the manufacturer’s recommendations. The mutations identified in the probands were sought in parents that were available. The nomenclature used for describing novel genetic changes follows the recommendations of the Human Genome Variation Society [19], and nucleotides were numbered according to the consensus coding DNA sequence of PAX6 isoform a (CCDS31451.1). In silico analyses of novel missense mutations and intronic changes were performed using the SIFT program and the NetGene2 Server, respectively. The intronic nucleotide variation, IVS2+9G>A (c.-129+9G>A), reported previously as pathogenic [20], was sought in 103 nonrelated healthy Mexican newborns using the PCR restriction fragment length polymorphism (PCR-RFLP) method by amplifying the 3′ end of exon 2 according to Love et al. [18] and restricting with the AciI enzyme where the presence of the G allele eliminates the restriction site. The Hardy–Weinberg equilibrium conformance was evaluated using the SNPstats software.ResultsPhenotypic information was available from 28 of the 30 probands, and a summary of findings is given in Table 1. The median age of cases was 5.2 years, and 18 of the probands (62%) were female. Eighteen of the cases (62%) were sporadic cases, and 11 had at least one relative with aniridia. Absent or nearly absent irides were evident in 26 cases (93%), and these were categorized as Iris 5 or Iris 6 according to Gronskov’s classification [5]. Of the remaining two cases, one was classified as Iris 3 and 4 (one eye each) and the other was classified as Iris 4. At least two ocular-associated alterations were present in 21 patients (75%), and the most common alterations were nystagmus (75%), macular hypoplasia (57%), and congenital cataracts (53%). Other less frequent features were optic nerve hypoplasia and keratopathy. Six individuals had glaucoma, which was congenital in two cases. The iris defect was not associated with any other ocular abnormality in only one patient (case 13).Table 1Iris grade and ocular associated findings in 30 Mexican nonrelated aniridia cases.CaseSexAge (years)InheritanceIris gradeBest corrected visual acuityNystagmusCataractGlaucoma/
treatmentMacular hypoplasiaOther1F4SporadicIris 520/600+----2M18FamilialIris 420/200++-+Ptosis3F11SporadicIris 620/100-+--Ptosis4M0.5SporadicIris 5FF+--+-5F3SporadicIris 520/380+--+Ptosis, strabismus6F15FamilialIris 3 and 420/25-+---7M33FamilialIris 5FC 0.5 mt+++, SG, MD+Kerathopathy8F3SporadicIris 6FF+--+-9F9SporadicIris 520/200++-+Ptosis10F6SporadicIris 520/40+--+-11M10FamilialIris 6FC 4 mt+++, SG, MD+Ectopia lentis, ONH12M8SporadicIris 5FC 1 mt+-congenital, SG, MD+Corneal leucoma13F0.8SporadicIris 5FF-----14M14SporadicIris 520/40-+---15F47SporadicIris 6FC 1.5 mt+++, SG, MD+Kerathopathy, ONH16F0.5SporadicIris 5FF-+---17F2SporadicIris 5FF--+, MD+Ectopia lentis, microcornea18F0.5SporadicIris 5FF+--+Ectopia lentis, microcornea, ONH19M16SporadicIris 620/200++--Ptosis, strabismus20??????????21M1FamilialIris 5FF+--+-22F8FamilialIris 620/130++---23M16SporadicIris 520/160++-+-24M5Sporadic???????25F13FamilialIris 520/200++-+Ectopia lentis26F17SporadicIris 620/200++--Strabismus27F0.7FamilialIris 6FF-+congenital, SG, MD?Corneal leucoma28F1FamilialIris 5FF+--+-29F6FamilialIris 520/120+----30M0.4FamilialIris 6FF+----M: Male; F: Female; Iris 3: circumpupillary iris hypoplasia; Iris 4: atypical sector coloboma; Iris 5: subtotal aniridia; Iris 6: complete aniridia; FF: fix and follow; FC: finger count; +: present; -: absent; ?: information not available; SG: surgical; MD: medical; ONH: optic nerve hypoplasia.Molecular findings are summarized in Table 2. We detected 11 SSCP mobility shifts in PAX6 products, all of which were consistent with the presence of mutations or neutral polymorphisms after sequencing. Causal mutations of the aniridia phenotype were found in 9 of 30 cases, yielding a detection rate of 30%. All mutations were heterozygous and unique except for the recurrent mutation, c.969C>T, which was observed in two sporadic unrelated cases. Four mutations were novel, c.184_188dupGAGAC, c.361T>C, c.879dupC, and c.277G>A. The remaining four mutations identified (c.969C>T, IVS6+1G>C, c.853delC, and IVS7–2A>G) have been previously reported (Human PAX6 allelic database) . Additionally, we found two intronic, nonpathogenic variations, IVS9–12C>T and IVS2+9G>A, both of which have also been previously described [20,21]. Of the nine probands in whom pathological mutations were identified, only nine parents were available for molecular analysis (Table 2).Table 2PAX6 gene mutations and polymorphisms identified in nine non-related Mexican aniridia cases.CaseIris grade*Nucleotide change**Nucleotide change***mRNA/
protein effectExon/
DomainMother´s GenotypeFather´s GenotypeStatus/
Reference4Iris 5c.184_188dupGAGACp.T63fsX18Exon 6/
Paired boxWild-typeNot availableNovel6Iris 3 and 4c.361T>Cp.S121PExon 7/
Paired boxHeterozygous for c.361T>CNot availableNovel10Iris 5c.607C>Tc.969C>Tp.R203XExon 8/
Linker regionWild-typeNot availablePreviously reported (Human PAX6 allelic database)18Iris 5c.357+1G>Cc.IVS6+1G>CCryptic donor splice-site and in-frame deletion of 36 amino acids coded by exon 6Intron 6/
Paired boxWild-typeWild-typePreviously reported [22]  Heterozygous for
c.-129+9G>AHeterozygous for IVS2+9G>ANoneIntron 2Homozygous for G alleleHeterozygous for IVS2+9G>A
(c.-129+9G>A)Previously described as polymorphism (Human PAX6 allelic database), but also as a possible pathogenic variant [20]. Present study confirmed that it is a polymorphism20?c.491delCc.853delCp.P164fsX43Exon 7/
Linker regionNot availableNot availablePreviously reported [5,22]21Iris 5c.879dupCp.T293fsX47Exon 10/
PST domainHeterozygous for c.879dupCWild-typeNovel22Iris 6c.277G>Ap.E93KExon 6/
Paired boxNot availableNot availableNovel  c.766-12C>TIVS9-12C>TNoneIntron 9Not availableNot availablePolymorphism previously reported [21]24?c.607C>Tc.969C>Tp.R203XExon 8/
Linker regionWild-typeWild-typePreviously reported (Human PAX6 allelic database)26Iris 6c.524-2A>GIVS7-2A>GIn silico prediction: 3 cryptic acceptor splice-sites (2 out-of-frame and 1 in-frame) inside exon 8 or in-frame exon 8 skipping.Intron 7/
Linker regionNot availableNot availablePreviously reported [22]An asterisk indicates that the measurements were according to Gronskov’s classification [5]. A question mark means that an ophthalmic evaluation was not available. A double asterisk symbol indicates that the gene mutation nomenclature was according to den Dunnen and Antonarakis [19]. A triple asterisk symbol denotes that the gene mutation nomenclature was according to previously proposed nomenclature by Ton et al. [8].With respect to novel changes, case 4 showed an insertion of a GAGAC sequence at nucleotide position 184, causing a frameshift arising from tandem duplication of nucleotides 184–188 that is predicted to encode a protein truncated in the paired domain. At evaluation, the patient exhibited a phenotype characterized by nystagmus, macular hypoplasia, and subtotal aniridia defect (Iris 5 in Gronskov’s classification). His mother had a normal ocular phenotype and did not have the mutation. A DNA sample from his father was not available, but he was referred to as visually healthy.Case 6 was a female patient with a novel missense substitution. Her right eye exhibited an eccentric pupil, circumpupillary iris hypoplasia (Iris 3), and cortical cataract. In the left eye, she had an atypical sector nasal iris coloboma (Iris 4), stromal hypoplasia, and total cataract (Figure 1 and Figure 2). The missense mutation identified was c.361T>C in exon 7 that changes serine 121 to proline (p.S121P) in the paired domain. Her mother exhibited foveal hypoplasia and nystagmus with whole irides, and her sister had congenital cataracts, nystagmus, and macular hypoplasia. Both affected relatives had the mutated allele.Figure 1Right eye iris and pupil of aniridia case 6 who had a novel missense mutation (c.361T>C) located in the NH2-region of the paired domain of PAX6. This eye exhibited eccentric pupil and circumpupillary iris hypoplasia (Iris 3).Figure 2Left eye iris of aniridia case 6 who had a novel missense mutation (c.361T>C) located in the NH2-region of the paired domain of PAX6. This eye exhibited partial absence of iris, an atypical sector nasal iris coloboma (Iris 4), stromal hypoplasia, and a total cataract.A base duplication at position 879 in exon 10 was found in case 21 and his mother. This previously unreported duplication (c.879dupC) causes a frameshift and introduces a premature stop codon 47 nucleotides downstream in the PST domain. The patient had Iris 5 with the associated ocular abnormalities of macular hypoplasia and nystagmus. The clinical manifestations of his mother were not available.Female case 22 showed the novel missense substitution, c.277G>A, in exon 6, which encodes part of the extreme amino end of the paired domain. The mutation changes glutamate at position 93 to lysine. This case also had a previously reported intronic polymorphism (IVS9–12C>T) [21]. The patient presented with total aniridia (Iris 6), nystagmus, and congenital cataracts. Her mother was referred to as affected, but we could not accomplish family studies because the patient resided in an orphanage.With respect to previously reported mutations, we found the IVS6+1G>C splice-site mutation [22] in case 18 who had Iris 5, microcornea, nystagmus, ectopia lentis, and macular and optic nerve hypoplasia. Her unaffected parents did not show this splice site change. Additionally, the patient and her father showed the previously described intronic substitution, IVS2+9G>A [20]. We searched for this substitution in 103 Mexican healthy controls and observed 19 heterozygotes (G/A) and two newborns homozygous for the A allele.The only deletion that we observed was the previously reported loss of cytosine at position 853 (c.853delC) that introduces a premature stop codon 43 nucleotides downstream [5]. This deletion was found in case 20, but phenotypic information was not available.We found the c.969C>T nonsense substitution (Human PAX6 allelic database), which changes arginine 203 to a UGA stop codon in the linker region, in two unrelated probands (case 10 and case 24); both were sporadic aniridia cases. Unfortunately, phenotypic information on case 24 and his parents were unavailable, but the molecular study was normal in both parents. Case 10 was a female dizygotic twin who showed subtotal aniridia (Iris 5), nystagmus, and macular hypoplasia. Her male twin and mother were genotypically normal and had a normal ocular phenotype, but the father was not studied.Finally, we also observed a mutation that produces a substitution in the splice acceptor site of intron 7 (IVS7–2A>G). An in silico analysis of this mutation, which has been previously reported in another single study [22], revealed the possible use of different cryptic splice sites. The individual with this mutation (case 26) had Iris 6 with nystagmus, cataract, and strabismus. Other members of her family were referred to as having a normal ocular phenotype, but they were unavailable for study.DiscussionTo the best of our knowledge, this is the first work on aniridia, apart from the original report, that uses the Gronskov classification of iris hypoplasia. Gronskov originally reported that the proportion of patients with Iris grade 1 to 4 was approximately 40% [5]  whereas we found only two index cases (7%), one with Iris grade 3 and 4, another with Iris 4, and none with lesser severity. This discrepancy might be explained by ascertainment bias, reflecting the fact that first-contact ophthalmologists are more familiar with the classic or severe aniridia presentation than with milder phenotypes. Another reason might be that individuals with milder cases, which are generally asymptomatic, do not seek medical care. In our opinion, Gronskov’s classification [5] should be widely used as a way to improve diagnosis, detect potential complications, and provide genetic counseling in aniridia cases with milder phenotypes.To our knowledge, this work represents the third largest aniridia series (only smaller than those published by Gronskov et al. [14] and Vincent et al. [15]) that included a molecular study of PAX6. Although we analyzed the entire coding region of the PAX6 gene in this work, the mutation detection rate of 30% that we found was lower than the 80% and 55% rates reported by the groups of Gronskov et al. [14] and Vincent et al. [15], respectively, who used diverse techniques for detecting pathological mutations. In this work, we used the SSCP technique exclusively, which is a widely used and efficient method for detecting mutations in PAX6 [12,16]. However, a low rate of PAX6 mutation detection (40%) using the SSCP technique has also been reported in patients described by Glaser et al. who proposed the possibility of mutations in more distant cis regulatory sequences [9]. Our low detection rate might be consistent with this interpretation because contiguous regulatory or non-coding sequences were not analyzed in our study. However, it also could be because of limitations of the SSCP technique itself as large genomic rearrangements would not be identified by this methodology. The inclusion of other mutation detection techniques in future studies would be expected to improve our mutation detection rate.We identified eight different causal PAX6 mutations in nine unrelated cases with isolated aniridia. The nature of the mutations was very similar to that reported in other populations [5,13,15,21]. Interestingly, we did not find the intragenic deletions previously reported in five Mexican patients, suggesting that these deletions might not be as frequent in our population as thought by Ramirez-Miranda et al. [17]. In this same context, our findings do not provide support for a founder effect of a specific mutation in the Mexican population [17].The only intragenic deletion identified (c.853delC) produces a frameshift and introduces a premature stop signal 42 codons downstream in exon 8. If it were translated, the predicted truncated PAX6 product would retain the paired domain but lack the homeobox and PST transactivator domain. This mutation has been observed twice before, once in a male patient with aniridia (Iris 4), cataracts, and nystagmus [5] and once in a female in which only aniridia was mentioned [22]. Unfortunately, our case was unavailable for phenotype-genotype correlation.The duplications, c.184_188dupGAGAC and c.879dupC, are novel, and both give rise to frameshifts, introducing premature stop codons in the paired domain and PST region, respectively. Phenotypes observed in other cases with insertion mutations are severe [5,23]. Consistent with this, our cases with these mutations had Iris 5.The nonsense substitution, c.969C>T, which changes an arginine codon (CGA) to a stop codon (UGA), was detected in two unrelated, sporadic cases (cases 10 and 24). This mutation has been previously found in at least 20 patients worldwide including familial and sporadic cases, making it one of the three more frequent changes in PAX6 along with c.1080C>T (27 cases) and c.1311C>T (20 cases; Human PAX6 allelic database). The differences in the ethnic origins of patients bearing the c.969C>T change indicate that this mutation is recurrent in PAX6. The recurrence of these three mutations might be explained at least in part by the presence of CpG dinucleotides in PAX6 that tend to become methylated and might thereby create conditions favorable for C>T substitutions as a consequence of spontaneous deamination of cytosine residues [23]. Our two patients positive for c.969C>T might represent independent mutational events since they were unrelated.With respect to the phenotype of c.969C>T heterozygotes, there are only five cases described in the Human PAX6 allelic database. Interestingly, one had partial aniridia with foveal hypoplasia and nystagmus, and the other four had aniridia with the associated ocular manifestations of nystagmus, cataracts, glaucoma, or corneal erosion. Of our two patients positive for c.969C>T, clinical information was available for only case 10. This patient had a severe phenotype and was classified as Iris 5 with nystagmus and macular hypoplasia.Literature reports based on the haploinsufficiency model have suggested that frameshift and nonsense mutations predicted to result in a truncated protein such as those described above are likely to exert their pathological effects through a “nonsense-mediated-decay” process where translation to protein might not occur because the mRNA is degraded [21,23]. However, it has also been noted that truncating mutations located downstream of DNA-binding domains especially those in exons 12 and 13 might have a dominant-negative effect [23,24]. In the present work, we did not identify nonsense mutations in this extreme 3′ region of the PAX6 gene.On the other hand, both novel missense mutations observed in the present work–c.277G>A (p.E93K) and c.361T>C (p.S121P)–might affect the function of the paired-box domain of the PAX6 protein because the properties of the substituted amino acids are quite different. In one case (p.E93K), a negatively charged glutamate is replaced by a positively charged lysine. In the other (p.S121P), the polar serine residue is replaced by the non-polar amino acid, proline. Moreover, glutamate 93 and serine 121 are largely invariant among closely related PAX family members with glutamate 93 conserved in PAX3, PAX4, and PAX7 and serine 121 conserved in eight PAX family genes (Protein BLAST). An in silico analysis using the SIFT program predicted that protein function would be affected (p<0.01), providing support for a possible pathogenic effect of these mutations, but further functional analyses are needed to confirm this.Missense mutations, which account for roughly 17% of changes in PAX6 worldwide, potentially retain residual protein activity and have been associated with milder phenotypes [5,16,23]. Consistent with this, case 6 who had a c.361T>C mutation showed Iris 3 (circumpupillary iris hypoplasia) and Iris 4 (atypical sector coloboma), which were the mildest iris grades found in the probands of our series. In contrast, case 22 carrying a c.277G>A substitution had complete aniridia (Iris 6) as well as nystagmus and cataracts. Although both of these mutations affect the paired domain, the c.277G>A mutation is located in the NH2-region and would therefore be expected to have a more profound effect on paired domain structure and function than the COOH-terminally localized c.361T>C mutation. This difference in location may account for the observed phenotypic differences, but additional studies will be required to support this idea.In some cases, missense mutations in PAX6 have also been associated with neurodevelopmental abnormalities such as absence/hypoplasia of the anterior commissure, callosal area, or pineal gland; olfactory system anomalies; cerebellar coordination problems; mental retardation; and epilepsy [11,16,20,25-28]. In fact, Dansault et al. [20] suggested that these abnormalities should be systematically investigated in every patient with aniridia. In cases 6 (age 8 years) and 22 (age 15 years) who had the missense mutation, clinical neurological anomalies were not observed, but cerebral CT scan or MRI imaging were not performed. Further descriptions of aniridia cases with missense mutations and neurodevelopmental anomalies will be needed to improve genotype-phenotype correlations. In addition to the novel missense substitution, c.277G>A, female case 22 had the intronic polymorphism, IVS9–12C>T, which is thought to represent a neutral variant [21].With respect to the splice-site mutation, IVS7–2A>G [22], an in silico analysis performed with the NetGene2 Server predicted that this change would eliminate the activity of the natural acceptor site in intron 7 and activate different cryptic acceptor sites within the exon or intron 8. It could, however, result in the use of the natural acceptor site in intron 8 and thereby lead to an in-frame, exon-skipping event that deletes exon 8. This mutation has been previously observed in a single case [22] with aniridia, cataracts, nystagmus, and corneal dystrophy (Human PAX6 allelic database). Similarly, our patient with this mutation (case 26) had a complete iris defect (Iris 6), nystagmus, cataract, and strabismus but without the corneal anomalies that might be present at an older age.The previously reported IVS6+1G>C substitution [22] disrupts the conserved dinucleotide GT in the intron 6 splice-donor site and might lead to the use of an alternative in-frame donor site inside exon 6. The predicted protein would lack the last 36 amino acid residues encoded by this exon, and the resulting deletion of a portion of the paired domain would be expected to lead to a severe phenotype (Human PAX6 allelic database). Consistent with this, the ocular phenotype of our patient was Iris 5 with nystagmus, microcornea, ectopia lentis, and macular and optic nerve hypoplasia. Both parents were considered healthy and were negative for IVS6+1G>C. This mutation has been reported once before in an aniridia patient but without the description of other clinical data [22]. Remarkably, there have been at least nine previous reports of a substitution at guanine by either adenine or thymine in the +1 position in GT donor sites in aniridia patients [12,29,30].In addition, case 18 and her unaffected father showed the previously described IVS2+9G>A substitution [20]. Although this intronic change was assumed to be potentially pathogenic by Dansault et al. [20] who observed it in a sporadic case with microphthalmia and other ocular abnormalities but not in 200 normal healthy individuals, an in silico analysis of this variant predicted that the binding capacity of the natural donor site would be unchanged. In our own search of 103 healthy Mexican newborns, we found this variant in a heterozygous state in 19 individuals and in a homozygous state in two. Hence, our data indicate that IVS2+9G>A is a neutral polymorphism and is not responsible for a pathological phenotype. The allele frequencies obtained for this polymorphism were in Hardy–Weinberg equilibrium.In summary, most of the mutations detected in our analysis alter invariant amino acid residues in the paired domain or predict truncation of the PAX6 protein. Four of the PAX6 mutations identified in this study are novel. In addition, our results lend support to the notion that c.969C>T is one of the three more frequent causal mutations in isolated aniridia cases and provide evidence that the IVS2+9G>A (c.-129+9G>A) variant is a neutral polymorphism.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2530865.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2530865",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2530865\nAUTHORS: Adam Monier, Jean-Michel Claverie, Hiroyuki Ogata\n\nABSTRACT:\nPhylogenetic mapping of metagenomics data reveals the taxonomic distribution of large DNA viruses in the sea, including giant viruses of the Mimiviridae family.\n\nBODY:\nBackgroundViruses are ubiquitous and the most numerous microbes in marine environments. Previous analyses using electron microscopy, epifluorescence microscopy and flow cytometry revealed the existence of 106 to 109 virus-like particles per milliliter of sea water [1-3]. Infecting marine organisms from oxygen-producing phytoplankton to whales, viruses regulate the population of many sea organisms and are important effectors of global biogeochemical fluxes [4,5]. It is also becoming clear that viruses hold a great genetic diversity; comparative genomics [6,7] and virus-targeted metagenomics studies [8-10] revealed a large amount of viral sequences having no detectable homologs in the databases. As a reservoir of 'new' genes as well as vectors of 'old' genes, viruses may significantly contribute to the evolution of microorganisms in marine ecosystems.Despite this progress in characterizing the environmental significance of viruses, a quantitative description of the marine virosphere remains to be done. This includes the determination of the relative abundance of virus families and the assessment of the level of their genetic diversity. In this context, large viruses, whose particle sizes can exceed those of small bacteria [11], are of particular concern. Most of them, such as Acanthamoeba polyphaga [12], may be retained on the 0.16-0.2 μpore filters specifically used in virus-targeted metagenomic studies and may not be gathered in the fraction traditionally associated with viral sequences [11]. A recently released marine microbial metagenomic sequence data set, produced by the first phase of the Sorcerer II Global Ocean Sampling (GOS) Expedition [13], provides an opportunity to quantitatively investigate viral diversity in marine environments. The GOS data comprise a large environmental shotgun sequence collection, with 7.7 million sequencing reads assembled into 4.9 billion bp contigs. In the GOS expedition, microbial samples were collected mainly from surface sea waters, and some others were collected from non-marine aquatic environments. Most DNA samples were extracted from the 0.1-0.8 μsized fraction, which is dominated by bacteria. Williamson et al. [14] recently reported that at least 3% of the predicted proteins contained within the GOS data are of viral origin. Notably, a number of sequences most similar to the genome of the giant mimivirus have been found in the Sargasso Sea metagenomic data set [15], produced by a pilot study of the GOS expedition [16], as well as in the new GOS metagenomic data set [17].Determining taxonomic distribution, referred to as 'binning', is the first step to analyze microbial populations in metagenomic sequences [18]. One simple binning approach uses database search programs such as BLAST to find the best scoring sequence of known species. A majority rule can be used to assign a taxonomic group to a metagenomic sequence [14,19]. Similar to the best hit criterion used to define orthologous genes in complete genomes [20,21], two-way BLAST searches were used to detect 'mimivirus-like' sequences in metagenomic data [15,17]. Such a post-processing of homology search results can improve the accuracy of taxonomic assignment. However, the use of homology search programs has serious drawbacks [22]. For instance, BLAST scores are highly sensitive to alignment sizes and to insertions/deletions. Further, it is difficult to infer evolutionary distances among high scoring hits only from the BLAST scores.Phylogenetic analysis remains the most powerful way to determine taxonomic distribution of metagenomic sequences. Short and Suttle [23] used phylogenetic methods to classify PCR-amplified gene sequences and suggested the existence of previously unknown algal viruses in coastal waters. Similar phylogenetic studies were performed to assess the diversity of T4-type phages [24] or RNA viruses [25,26] in marine environments. In these studies, different markers, such as the major capsid genes or RNA-dependent RNA polymerase gene sequences, were amplified by PCR or RT-PCR and analyzed by phylogenetic methods. To examine taxonomic distribution of large DNA viruses in a metagenomic sequence collection, B-family DNA polymerase (PolB) is a useful marker [23,27,28]. PolB sequences are conserved in all known members of nucleocytoplasmic large DNA viruses (NCLDVs) [29], which include 'Mimiviridae' [30], Phycodnaviridae, Iridoviridae, Asfarviridae, and Poxviridae. PolB genes are also found in other eukaryotic viruses, such as herpesviruses, baculoviruses, ascoviruses and nimaviruses, in some bacteriophages (for example, T4-phage, cyanophage P-SSM2), and in some archaeal viruses (for example, Halovirus HF1). Eukaryotes have four PolB paralogs (catalytic subunits of α, δ, ε and ζ DNA polymerases). PolB genes are found in all of the main archaeal lineages (Nanoarchaeota, Crenarchaeota and Euryarchaeota). The presence of PolB homologs in bacteria (the prototype being Escherichia coli DNA polymerase II) is limited; PolBs are found in Proteobacteria, Acidobacteria, Firmicutes, Chlorobi and Bacteroidetes. PolB genes are suitable for the classification of large DNA viruses [31,32] thanks to their strong sequence conservation and an apparently low frequency of recent horizontal transfer [28,33].When applying phylogenetic methods to environmental shotgun sequences, the treatment of short sequences requires special attention. These sequences show large variation in size and possibly correspond to different parts of a selected marker gene. Piling up multiple short sequences on representative markers from known organisms does not provide an appropriate alignment (whatever software is used) with enough signals for the subsequent phylogenetic analysis. In this study we developed a new phylogeny-based method. The method called 'phylogenetic mapping' analyzes individual metagenomic sequences one by one and determines their phylogenetic positions using a reference multiple sequence alignment (MSA) and a reference tree. As an attempt to investigate the presence, the taxonomic richness and the relative abundance of different large DNA viruses in marine environments, we analyzed the GOS data set using PolB sequences as our reference. Our study does not address the abundances of small DNA viruses or RNA viruses [14,34].ResultsPhylogenetic mappingWe searched the GOS data set for PolB-like sequences using the Pfam hidden Markov profile (PF00136). This resulted in a set of 1,947 sequences (from 23-562 amino acid residues). These sequences are referred to as 'PolB fragments' in this study. We next built a reference MSA of PolB homologs from known organisms (Additional data file 1). The reference MSA (Additional data file 2) corresponds to the polymerase domains of PolB homologs and contains 101 sequences, which were selected to achieve the widest possible taxonomic/paralog coverage (but with a non-exhaustive sampling for closely related species) for the analysis of the GOS metagenomic data. The reference MSA was used to generate a maximum likelihood tree (that is, the reference tree; Figure 1). Although the phylogenetic reconstruction did not provide statistical support for most of the basal branches, many peripheral groupings (supported by bootstrap values ≥ 70%) were coherent with the current taxonomy of viruses and cellular organisms. In this tree, we identified eight viral groups: poxviruses; chloroviruses; phaeoviruses; mimivirus and related algal viruses (Pyramimonas orientalis virus PoV01, Chrysochromulina ericina virus CeV01 and Phaeocystis pouchetii virus PpV01); iridoviruses grouped with ascoviruses; herpesviruses; baculoviruses; and one phage group. The PolB homologs from African swine fever virus (ASFV, Asfarviridae), Emiliania huxleyi virus 86 (EhV-86, Phycodnaviridae), Heterosigma akashiwo virus 1 (HaV, Phycodnaviridae) and the phage RM378 did not show well supported clustering with other PolB sequences. We also identified eleven groups in the reference tree for cellular PolB homologs: seven archaeal groups, one bacterial group and three eukaryotic groups (α, δ and ζ subtypes). Each of the GOS PolB fragments was then examined for its phylogenetic position using the reference MSA and the reference tree. To reduce the computation time and to streamline tprocess of summarizing results, we reduced the size of the reference MSA. Specifically, we selected 51 representatives from the 101 reference sequences and removed the remaining sequences. The reference tree was also reduced so that the resulting tree contains only the selected 51 representatives, while we conserved the original topology of the full reference tree shown in Figure 1. The reduced reference tree has 99 branches (including internal branches). A constraint on this topology defines 99 possible branching positions for each of the GOS PolB fragments. We aligned, one by one, each of the PolB fragments on the reduced reference MSA using the T-Coffee profile method. Based on the resulting profile MSA containing 52 sequences, the likelihoods for all 99 possible branching positions (thus 99 different topologies) were computed by ProtML [35]. A statistical significance for the best tree among the 99 topologies was assessed by the RELL (resampling of estimated log likelihoods) bootstrap method [36,37]. We considered the branching position of a PolB fragment to be supported when the RELL bootstrap value for the best topology was ≥ 75%.Figure 1Maximum likelihood tree of 101 PolB sequences in the complete reference set. The phylogenetic tree was built using PhyML [73] (Jones-Taylor-Thornton substitution model [76], 100 bootstrap replicates) based on a multiple sequence alignment generated using M-Coffee [72]. This tree is unrooted per se. The phage group was arbitrarily chosen as an outgroup for presentation purposes. The lengths of branches do not represent sequence divergence. Bootstrap values lower than 70% are not shown. The selected 51 representatives for the phylogenetic mapping and the associated branches are highlighted in bold face and black lines, respectively. Different colors correspond to different taxa: viruses (blue), eukaryotes (orange), bacteria (green) and archaea (pink).Diversity of large DNA viruses in the GOS data setOur phylogenetic mapping method could assign the best branching position for 1,423 PolB fragments, of which 1,224 (86%) were mapped on viral branches. The best branching position was statistically supported by the RELL method for 869 PolB fragments, of which 811 (93%) were mapped on viral branches. Figure 2 and Additional data file 3 show the taxonomic distribution of the GOS PolB fragments. The largest fraction of the PolB fragments was mapped on the phage group. Of 866 cases of mapping within the phage group, 633 were supported. This appears consistent with the current estimate of the large number of phage-like particles and their genetic richness in marine environments [3]. The second largest number of supported mappings was found to fall into large eukaryotic viruses commonly found in aquatic environments. Among them, the 'Mimiviridae group' (mimivirus, PoV01 and CeV01 [17]) represented the largest fraction, with 115 supported cases. The chlorovirus group gathered 51 supported cases of mapping. The iridovirus/ascovirus group and the branch leading to HaV showed five supported mappings each. In contrast, no PolB fragment was mapped for the groups for baculoviruses or herpesviruses commonly found in terrestrial animals. Interestingly, we found two PolB fragments mapped with good support on the ASFV branch (JCVI SCAF 1101668126451, JCVI SCAF 1101668152950). When these two PolB fragments were compared to the NCBI non-redundant amino acid sequence database (NRDB) using BLASTP, they were most similar to the ASFV PolB sequence. ASFV is pathogenic to domestic pigs and is currently the sole representative of the Asfarviridae family [38]. Concerning cellular organisms, eukaryotic homologs gathered few mappings, as expected from the sample filtration threshold used in the GOS metagenomic study. Two archaeal groups - the group III containing crenarchaeotes (for example, Pyrobaculum aerophilum, Cenarchaeum symbiosum) and the group IV containing euryarchaeotes (for example, Thermoplasma acidophilum, an uncultured euryarchaeote Alv-FOS1) - had 23 and 17 supported cases of mapping, respectively. The bacterial group presented ten supported mappings.Figure 2Phylogenetic mapping results of the GOS PolB fragments. Results of the phylogenetic mapping are summarized and displayed for each group in the reference tree. Numbers in parentheses (X/Y) are the total number of mapped PolB fragments (Y) and the number of supported cases (X). The tree topology is the same as the one shown in Figure 1. Branches with bootstrap values ≥ 70% are marked with filled circles. The 99 branches examined by our phylogenetic mapping are shown with black lines; other peripheral branches are shown with gray lines. The length of the scale bar corresponds to 0.5 substitutions per site. colors correspond to different taxa: viruses (blue), eukaryotes (orange), bacteria (green) and archaea (pink).Validation of the mapping results using long PolB fragmentsWe examined the phylogenetic mapping result and the sequence diversity of the PolB fragments classified in large eukaryotic virus groups (that is, NCLDVs). From those mapped on NCLDV branches, we selected long PolB fragments that generated a profile MSA showing at least 150 non-gapped sites. We computed a single alignment of these long PolB fragments together with the reference PolB sequences from large eukaryotic virus groups. A maximum likelihood tree (Figure 3) based on the alignment was perfectly consistent with our one-by-one mapping result (Figure 2) in terms of taxonomic assignment. The Mimiviridae group contained 16 PolB fragments showing substantial sequence variations. Twelve of them were significantly closer (bootstrap 100%) to CeV01 or PpV01 (both viruses of haptophytes) than to mimivirus or PoV01 (a green algal virus). Three of the rest were grouped with either mimivirus (bootstrap 89%) or PoV01 (bootstrap 96%). The last one (JCVI SCAF 1096627348452) was placed at the basal position of the Mimiviridae group. Although this basal positioning was not statistically supported, it was consistent with our one-by-one phylogenetic mapping result. The mimivirus PolB shared 47% identical amino acid residues with its closest homolog (JCVI SCAF 1101668170038). A large and diverse group containing 27 PolB fragments (bootstrap 92%) was also found beside the chlorella virus group (Paramecium bursaria chlorella viruses 1, K2 and NY2A). The DNA polymerase gene from the recently released Ostreococcus virus OtV5 genome (GenBank: EU304328) [39] was found grouped together with these PolB fragments. The grouping of a PolB fragment with ASFV PolB was also confirmed (bootstrap 100%).Figure 3Maximum likelihood tree of PolB sequences belonging to NCLDVs. The phylogenetic tree was built using PhyML [73] (Jones-Taylor-Thornton substitution model [76], 100 bootstrap replicates) based on a multiple sequence alignment generated using MUSCLE [77]. Bootstrap values lower than 50% are not shown. GOS sequences are marked with filled circles and displayed in purple. The tree was mid-point rooted. The DNA polymerase gene from the recently released Ostreococcus virus OtV5 (GenBank: EU304328) was included in this tree. The OtV5 PolB was not included in our reference set as it was not available at the time of our phylogenetic mapping study. The length of the scale bar corresponds to 0.5 substitutions per site.Viral PolBs are more diverse than bacterial PolBsWe investigated the abundance of viral PolB genes relative to bacterial PolB genes in the GOS data set. Here, we used read coverage as a proxy to measure the abundance of the cognate DNA molecules in the samples. We computed the read coverage of each contig harboring a PolB fragment mapped on the reference tree with significant support, and then obtained the median of the read coverage values for each set of contigs mapped on the same branch (Additional data file 3). PolB sequences mapped on viral branches exhibited low median coverage values ranging from 1.31 for the ASFV branch to 2.00 for a phage branch. The median coverage value for the contigs mapped on the mimivirus branch (12 contigs) was 1.32. The viral contig with the largest read coverage (6.68) was the one mapped on the cyanophage P-SSM4 branch. In contrast, a higher median coverage value (8.40) was found for bacterial contigs mapped on the branch leading to Shewanella frigidimarina. One of the bacterial contigs exhibited a read coverage of 29.17. Viral branches were thus characterized by a large number of mapped contigs exhibiting a low coverage. This is consistent with numerous and very diverse viral populations [40]. On the other hand, the bacterial branches exhibited a lower number of mapped contigs with a larger read coverage. This is consistent with numerous but less diverse populations of bacterial species, although our results concern only bacteria having PolB homologs.Geographic distributions of viral PolBsGOS metadata provide physicochemical and biological parameters associated with each sampling site, such as water temperature, salinity, chlorophyll a concentration, and sample's water depth. These data offer additional dimensions to analyze the viral PolB fragments identified by our phylogenetic mapping. Here we compared the relative abundance of the predicted viral PolB fragments and the associated metadata across different GOS sampling sites (Figure 4a).Figure 4Geographic localization. (a) The different sampling sites of the Sorcerer II Global Sampling expedition. The samples 00 and 01 are part of the Sargasso Sea pilot study [16]. The inset shows samples 27 to 36, which were sampled in the Galapagos Islands. The sampling sites displayed in light gray were not analyzed in the GOS original study, nor in this study. This part of Figure 1 was reproduced from [13]. (b) Relative abundance of PolB fragments for virus groups across GOS sampling sites. The left-most panel shows the relative abundance of viral PolBs in difierent GOS samples. The mimivirus group clearly appears as the most ubiquitous after phages. Four area plots (second to fifth panels from the left) show water temperature, chlorophyll a concentration (no information was available for GS20, GS30, GS32, GS33, GS47 and GS51 sites), salinity (no information was available for GS06, GS11, GS13, GS14, GS28, GS30, GS31, GS32, GS34 and GS37 sites) and sample depth, respectively. Two far right histograms (sixth and seventh panels) show the proportion and the estimated number of reads associated with the viral PolB fragments among total reads for a given sample.Predicted viral PolB fragments were detected in all of 44 GOS sampling sites (Figure 4b). The relative abundance of different virus groups showed substantial variation across these samples. This is consistent with the diverse ecosystems covered by the GOS expedition.PolB fragments classified in the phage group were found in 42 (95%) of the 44 sample sites; the two samples without phage PolB fragments were GS08 (Newport Harbor, Richmond, USA) and GS32 (mangrove). In most samples (32 sites), putative phage PolBs exhibited a higher abundance relative to putative eukaryotic viral PolBs. On the other hand, the relative abundance of eukaryotic viral PolBs was higher than that of phage PolBs in 12 sampling sites. We found a significant positive correlation between the relative abundance of phage PolBs and water temperature (p = 0.001; Fischer's exact test with no correction for multiple testing): phage-type PolBs showed a higher relative abundance than eukaryotic viral PolBs in tropical waters (T ≥ 20°C), while a reversed tendency was observed in temperate water (T < 20°C). Interestingly, among eukaryotic viral PolBs, putative Mimiviridae PolBs showed the most widespread distribution, being detected in 38 (86%) of the total sites. One of these sampling sites (mangrove located on Isabella, Ecuador) exhibits only viral PolBs classified in the Mimiviridae group. This is the sole mangrove site of all the GOS sampling locations. Mimiviridae PolBs were also relatively abundant in two of the three samples from a hydrostation located in the Sargasso Sea. Three samples correspond to different size fractions: 3.0-20.0 μm for GS01a; 0.8-3.0 μm for GS01b; and 0.1-0.8 μm for GS01c. Putative Mimiviridae PolBs were identified in the GS01a and GS01c samples. The GS01a sample, which was targeted to small eukaryotes, might have contained host species infected by putative viruses of the Mimiviridae group. PolB fragments grouped with chloroviruses were also widely distributed. They were detected in 16 (36%) samples. The relative abundance of this putative eukaryotic virus group showed a significant positive correlation with chlorophyll a concentration, a measure of primary productivity in oceanic regions (p = 0.00002; Fisher's exact test with no correction for multiple testing).The sample exhibiting the broadest taxonomic richness of viral PolBs was from Chesapeake Bay (GS12, MD, USA), which is an estuary. The GOS metagenomic sequences from this site exhibited PolB fragments classified in phages, chloroviruses, Asfarviridae and Mimiviridae. Notably, this site is a highly eutrophic estuary with an extremely high chlorophyll a concentration. PolBs classified in Asfarviridae were also detected in another estuary site (GS11, Delaware Bay, NY, USA), which is close to Chesapeake Bay.Prediction of putative 'new' viral genesContigs harboring putative viral PolB homologs were relatively small, ranging from 0.4-12.5 kb (average 1,874 bp) for contigs mapped on eukaryotic viral branches and 0.5-8.8 kb (average 1,885 bp) for phages. To examine the presence of additional open reading frames (ORFs) in these contigs, these putative viral contigs were searched against NRDB using BLASTX. We detected several genes or gene fragments that are usually specific to viruses. For example, several contigs (for example, JCVI SCAF 1096626858151, JCVI SCAF 1096626920680) containing PolB fragments assigned to the chlorovirus group also harbor an ORF most similar to the OtV5 putative major capsid gene. Several putative phage-type contigs (for example, JCVI SCAF 1096628232224, JCVI SCAF 1096626847406) mapped on the cyanophage P-SSM4 branch exhibited ORFs similar to regA (translation repressor of early genes) or uvsX (recA-like recombination and DNA repair protein genes). The presence of such 'virus-specific' genes next to the 'virus-like' PolB homologs corroborates the validity of our phylogenetic mapping approach.During this search, we found an ORF similar to RimK, a protein involved in post-translational modification of the ribosomal protein S6, in a contig (JCVI SCAF 1096626956347) having a PolB fragment mapped on the cyanophage P-SSM4 branch. In this contig, the rimK homolog was flanked by a phage-specific regA homolog (Figure 5). rimK homologs are found in bacteria, archaea and eukaryotes [41]. To our knowledge, no rimK homolog has been found in a viral genome. Using this putative viral RimK homolog as a query of TBLASTN, we screened the entire GOS data set. We identified more than a hundred contigs harboring RimK homologs with higher similarities (BLAST score from 137 up to 732; E-value < 10-30) than those exhibited by cellular homologs (BLAST score < 132; E-value > 10-29) in NRDB. The sequences of those putative phage RimK homologs were readily aligned with Escherichia coli RimK along its entire length (not shown), and showed amino acid residues highly conserved in the ATP-graps domain of bacterial RimK [41]. Several GOS RimK sequences showed an additional domain of unknown function (DUF785, PF05618, E-value < 0.001) at the carboxy-terminal side of the ATP-graps domain. A DUF785 domain is present also in RimK of some bacteria (at the amino-terminal side of the ATP-graps domain) such as Synechococcus sp. (Q7U6F4) and euryarchaeotes (at the carboxy-terminal side of the ATP-graps domain) such as Halobacteria (for example, Q5V351). Furthermore, many of the GOS contigs encoding RimK homologs exhibited additional ORFs usually specific to phages such as T4-like clamp loader subunit genes, contractile tail sheath protein genes or T4-like DNA packaging large subunit terminase genes (Figure 5). Our phylogenetic analysis indicates that those RimK homologs are closely related to each other and distantly related to bacterial RimK (Figure 6). These results suggest the existence of phages carrying rimK homologs in marine environments.Figure 5Gene organization of GOS contigs with putative phage RimK sequences. Putative phage rimK genes are shown in red. Other predicted genes are color coded according to their best BLAST hit taxonomies in NRDB as shown in the inset panel. MT-A70 corresponds to the adenine-specific methyltransferase. gp17 is a T4-like DNA packaging large subunit terminase homolog. gp18 is a contractile tail sheath protein homolog. The crystal structure of a GOS homolog for the protein encoded by the hypothetical gene (gray) has been determined and is available in the Protein Data Bank (3BY7).Figure 6Maximum likelihood tree of RimK sequences. RimK sequences were retrieved from UniProt [78] and from the GOS metagenomic data set using BLASTP. The phylogenetic reconstruction was performed using PhyML [73] (Jones-Taylor-Thornton substitution model [76], 100 bootstrap replicates) based on a multiple sequence alignment generated with MUSCLE [77]. Bootstrap values lower than 50% are not shown. The tree was mid-point rooted. GOS sequences are marked with filled circles and displayed in purple. The length of the scale bar corresponds to 0.4 substitutions per site.DiscussionUntil recently, the marine virosphere was terra incognita. The increasing amount of environmental sequence data now provides unprecedented opportunities to explore the viral world. Previous studies characterized the abundance and the genetic richness of marine viruses using environmental sequencing approaches [8,14,19,23,24]. However, the extent of species diversity within individual viral groups is still unclear. This is especially the case for large DNA viruses. Large DNA viruses were often overlooked or were not the specific focus of marine metagenomic projects. In this study, we used a new phylogenetic mapping approach to identify viral PolB sequences contained in the GOS metagenomic data set and assessed their taxonomic distribution. This study does not concern small viruses, including RNA viruses. Beyond BLAST searches, our phylogenetic mapping approach provided a somewhat unexpected picture of the taxonomic distribution of viral sequences in the metagenomic data.In the GOS data we identified 811 PolB-like sequences closely related to known viral PolB sequences. This is consistent with the existence of a wide taxonomic spectrum of PolB-containing DNA viruses in marine environments [23]. As previously noted [14], phages are the main contributors to this diversity; our method predicted that 78% (633/811) of the viral PolB fragments were of phage origin. This proportion is likely an underestimate of the actual taxonomic diversity of double-stranded DNA phages in the GOS sampling areas as only a subset of DNA phages carry PolB genes.Interestingly, the mimivirus group was the second largest in terms of the number of assigned PolB fragments (that is, 115 cases of mapping). Previous studies revealed the existence of mimivirus-like sequences in the GOS metagenomic data set [15,17]. Our data now suggest that the species/strain richness contained in the GOS metagenomic samples for this viral group may be comparable to those exhibited by other groups of eukaryotic large DNA viruses, including most of the previously characterized phycodnaviruses. The amoeba infecting mimivirus has the largest known viral genome (1.2 Mb). Its particle size is approximately 0.7 m in diameter including its filamentous layer [11]. In addition, the mimivirus group contains two haptophyte viruses (CeV01 (510 kb), and PpV01 (485-kb)) and a virus infecting a green algal species (PoV01 (560 kb)) [17,42]. Their genomes are also larger than any other eukaryotic viruses sequenced so far [43,44]. The particle sizes of these three algal viruses are 0.16-0.22 μm, being compatible with the filter sizes used in the GOS sampling. Notably, their particle sizes are comparable to those of classic phycodnaviruses with a mean diameter of 0.16 ± 0.06 μm [45,46]. By counting overlapping PolB fragments mapped on the mimivirus group, we estimated that at least 85 distinct species/strains of Mimiviridae are present in the GOS metagenomic samples. Within the mimivirus group, two haptophyte viruses (PpV1 and CeV01) were clustered together with a high bootstrap value (Figure 3). Most (84%; 97/115) of the Mimiviridae-like PolB fragments were mapped within this subgroup. Haptophyte species may thus be the major hosts of putative viruses corresponding to the PolB subgroup. Overall, these data suggest that large DNA viruses composing the Mimiviridae group represent one of the main components of marine eukaryotic large DNA viruses.The branch leading to the chloroviruses presented 51 cases of GOS PolB fragment mapping. These GOS sequences were closely related to the recently determined PolB sequence from OtV5. OtV5 infects Ostreococcus tauri, a small green algal species of prasinophyte (approximately 1 μm in diameter) found in diverse geographic locations [47]. Short and Suttle identified a group of viral sequences closely related to prasinoviruses (Micromonas pusilla viruses) through sequencing PCR products targeted to algal virus PolBs [23]. We found that some of the sequences studied in their work were also highly similar to the OtV5 PolB sequence. For instance, the sequence named BSA99-5 (GenBank: AF405581) in their study exhibited 93% amino acid sequence identity to the OtV5 PolB sequence. This suggests that the major hosts for this putative viral group may be prasinophytes.Surprisingly, we identified two PolB fragments most closely related to the ASFV PolB. ASFV is currently the sole isolated member of the Asfarviridae family. The known natural hosts of ASFV are terrestrial animals, including warthogs, bush pigs and soft ticks [38]. ASFV causes a persistent but asymptomatic infection in these hosts. In domestic pigs, ASFV causes an acute hemorrhagic infection with mortality rates up to 100% depending on different viral isolates. We now predict the existence of additional Asfarviridae in marine environments, although the contamination from terrestrial origin cannot be excluded. In a recent metagenomic study, Marhaver et al. [48] analyzed the viral communities associated with healthy and bleaching corals. They showed that alphaherpesvirus-like and gammaherpesvirus-like sequences accounted for 4-8% of the analyzed environmental sequences. GOS sampling sites include a coral reef atoll site (GS51). No herpesvirus-type PolB fragment was detected in our study.Through the analysis of geographic distribution, we found that putative viral PolB fragments were identified in all of the 44 GOS samples. This suggests a wide presence of PolB-encoding viruses in diverse marine environments. Interestingly, phage PolB sequences were more abundant than eukaryotic viral PolB sequences in samples from tropical areas; conversely, many samples from temperate areas were enriched in eukaryotic viral PolBs. Further, most of the samples showing a great taxonomic richness of viral PolB sequences corresponded to those from temperate areas. This observation is consistent with the current understanding of the distribution of eukaryotic and bacterial phytoplankton in oceans. Gibb et al. [49] surveyed the spatial distributions of phytoplankton pigments across the Atlantic Ocean over 100° of latitude (from 50°N to 50°S). They showed a major transition in pigment characteristics from temperate to tropical/sub-tropical waters; temperate waters were characterized by larger phyto-biomass enriched in eukaryotic phytoplankton, while tropical/sub-tropical waters exhibited smaller phyto-biomass enriched in prokaryotic phytoplankton such as prochlorophytes [49].The relatively high abundance of eukaryotic viral PolBs in samples from temperate areas (showing high chlorophyll a concentrations) was mainly due to the abundance of the GOS PolB sequences grouped with chlorovirus PolBs. This again suggests that the hosts of these putative viruses are green algae (such as prasinophytes). In contrast, Mimiviridae-like PolB fragments showed a wider geographical distribution. They were identified in sequences from most of the GOS sampling sites, from northeast Atlantic Ocean to southwest Pacific Ocean. These sites correspond to a variety of habitat types, such as coastal seas, open oceans, fresh water sites (GS20, Lake Gatun, Panama; GS32, mangrove, Isabella, Ecuador) and even hypersaline waters (GS33, Punta Cormorant Lagoon, Floreana, Ecuador). The detection of Mimiviridae-like PolB fragments was not correlated with chlorophyll a concentration. Hence, the hosts of these putative Mimiviridae viruses are not limited in temperate/eutrophic waters. In fact, species of haptophyte have been found and known to occasionally form blooms in waters from sub-polar to (sub-)tropical latitudes, including oligotrophic areas [50-52]. Acanthamoeba, the host of mimivirus, also have the ability to survive in diverse environments [53].Finally, our study allowed the identification of putative phage rimK. In E. coli, RimK catalyzes the post-translational addition of glutamic acid residues to the amino terminus of ribosomal protein S6 [54]. A resistance to antibiotics was suggested for the E. coli mutant lacking the activity of the S6-modification [55]. Reeh and Pedersen [56] showed that the relative level of the S6-modification was not affected by the growth rate in culture. Besides these observations, however, much is unknown for the functional consequence of the S6 modification in E. coli. Bacteriophage T7 modifies ribosomal protein S6, S1 and translational initiation factors by phosphorylation upon infection of E. coli [57]. The modifications of host translational proteins are performed by a T7-encoded kinase, and enhance phage reproduction under sub-optimal growth conditions. It was suggested that the phosphorylation of these proteins serves to stimulate translation of the phage late mRNAs. The RimK homologs found in phage-like contigs may be involved in a similar process. Unexpected homologs of cellular genes are continuously identified in viral genome sequences [12,58,59]. We believe that our phylogenetic mapping approach will be useful to identify further occurrences of unexpected viral genes in environmental sequences.ConclusionThe use of a phylogenetic approach provided a comprehensive picture of the taxonomic distribution of large viruses enclosed in the GOS metagenomic data. As expected, the highest genetic richness corresponded to phages. Interestingly, our data suggest that Mimiviridae represent a major and ubiquitous component of large eukaryotic DNA viruses in diverse marine environments.Materials and methodsExtraction of PolB fragments from the GOS metagenomic data setWe retrieved the combined assemblies of the GOS metagenomic data through the CAMERA website [60]. The data set was composed of 3,081,849 scaffolds. We extracted all the stop-to-stop ORFs (≥ 60 amino acid residues) from the assembled sequences using EMBOSS/GETORF [61]. We obtained a set of 21,406,171 ORFs. Those ORFs were translated into corresponding amino acid sequences. To identify PolB-like fragments in this set, we used the Pfam profile (PF00136, both long and fragment search versions: 'ls' and 'fs') [62] and the HMMER software as a search engine [63] using an E-value threshold of 0.001. We then removed redundancy (due to the double use of 'ls' and 'fs' versions of the Pfam profile) and false positive detections (having the best hit against non-PolB sequences in the NRDB) by BLASTP [64] using an E-value threshold of 10-5). We extracted only the parts of metagenomic amino acid sequences that were aligned on the Pfam profile representing the polymerase domains of PolB. Thus, additional domains (such as endonuclease domains) were not included in our PolB sequence set. No contig was found to contain more than one PolB homolog. As a result of these processes, we obtained 1,947 distinct PolB-like sequences (from 23-562 amino acid residues); these sequences are referred to as PolB fragments in this study. We parsed the GOS PolB fragments to find intein insertions by the TIGRFAM profiles TIGR01445 (intein amino terminus) and TIGR01443 (intein carboxyl terminus) [65], but none of these fragments had a detectable intein domain. In this study, we did not include the protein priming subfamily of the B family DNA polymerase [28], which is represented by the Pfam profile PF03175. The members of this subfamily are found in eukaryotic linear plasmids of mitochondrion, phages and adenoviruses.PolB homologs from the NRDBWe retrieved PolB homologs from the NRDB, RefSeq [66] and KEGG [67] databases using BLAST using multiple query sequences (E-value < 10-5) and the PolB Pfam profile (E-value < 0.001). We removed species redundancy using BLASTCLUST [64] while keeping the widest possible taxonomic/paralog coverage (but with a non-exhaustive sampling for closely related species). This resulted in a set of 120 PolB homologs (Additional data file 1). We removed intein sequences in the PolBs of mimivirus [68], HaV [69] and CeV01 (GenBank: ABU23716).Construction of the reference alignment and the reference treeWe next constructed an alignment of PolB homologs from known organisms (that is, the reference MSA) and generated a phylogenetic tree of PolB homologs (that is, the reference tree). There is a tradeoff between the number of distant homologs included in the reference MSA (contributing to a wider taxonomic/paralog coverage) and the quality of the resulting MSA and tree (contributing to a reliable classification of metagenomic sequences). Among the 120 PolB homologs, we identified 19 highly divergent sequences that decrease the quality of the resulting PolB alignment and tree but that show no close homologs in the GOS PolB fragments. This process was performed through multiple trials of building alignments by T-Coffee [70] and phylogenetic trees by PhyML for the PolB homologs. These 19 sequences correspond to six groups of PolB homologs: eukaryotic DNA polymerase ε, a Trichomonas vaginalis DNA polymerase α-like paralog, PolBs of unclassified herpesviruses (Ostreid, Ictalurid and Ranid herpesviruses), Heliothis zea virus, a nimavirus (shrimp white spot syndrome virus), and PolBs of a group of bacteria related to Prosthecochloris vibrioformis and Chlorobium tepidum. There was no PolB-like fragment in the GOS data exhibiting a best BLAST hit against these groups of PolB homologs. Therefore, the removal of the six groups of PolB homologs from our reference data set does not affect the interpretation of the results described in this manuscript. After discarding these 19 sequences, the final PolB set was composed of 101 sequences. We aligned the 101 PolB sequences using M-Coffee accessible from a public server [71] with the use of default options. M-Coffee is a meta-method for assembling multiple sequence alignments [72]. We extracted only the polymerase domain sequences from the alignment (that is, the reference MSA; Additional data file 2). The reference alignment showed four conserved regions (numbered from I to IV) previously described as the signatures of the PolB polymerase domains [33]. We next built a maximum likelihood tree based on the reference MSA (that is, the reference tree) using PhyML after removing gap-containing sites [73] with JTT substitution model and a gamma low (four rate categories). Bootstrap values were obtained after 100 bootstrap replicates. We used the phylogeny.fr platform [74] to generated scalable vector graphics from newick formatted trees.Phylogenetic mappingEach of the metagenomic PolB fragments was taxonomically assigned by aligning it against the reference MSA and by examining its phylogenetic position in the reference tree. In order to reduce the computation time and to avoid unnecessary complications in summarizing results within too dense a tree, we reduced the size of the reference MSA and the reference tree. Specifically, we selected 51 PolBs from the 101 PolBs contained in the initial set. We kept the selected 51 PolBs in the reduced set, and deleted the remaining PolBs. The selection of the 51 representatives was carried out in the following way. First, we selected all the PolBs (that is, ASFV, EhV86, HaV, Phage RM378) that were not grouped with other PolBs with a statistical support (≥ 70% bootstrap value) in the initial reference tree (Figure 1). Second, we selected two or three representatives from each of the statistically supported monophyletic groups (≥ 70% bootstrap value). The choice of representatives from a monophyletic group was arbitrary. We simply selected two or three relatively distant sequences from the members of the monophyletic group. To obtain a reduced reference MSA composed of the selected 51 sequences, we extracted a part (that is, lines) of the initial reference MSA (containing gaps). The initial reference tree (composed of 199 branches including internal ones) was also reduced by pruning branches leading to the non-selected leaves using BAOBAB [75].The reduced reference tree has 99 branches (including internal branches); the constraint on the topology of the reduced reference tree thus defined 99 possible branching positions for each PolB-like fragment extracted from the metagenomic data set. The reduced reference MSA and the reduced reference tree are the basis for our phylogenetic mapping in this study. Each of the PolB fragments from the GOS data set was aligned on the reduced reference MSA (containing gaps) using T-Coffee [70] with a profile alignment option. For the T-Coffee profile alignment, we used the option '-profile comparison = full10'. If a GOS PolB fragment generates an alignment with less than 50 sites after removing gap-containing sites, we discarded the GOS PolB fragment from our analysis. Based on the resulting alignment (51 reference sequences and one GOS PolB fragment), the likelihoods of all 99 possible branching positions (thus 99 different topologies) for the PolB fragment were computed by ProtML [35]. A statistical significance for the best tree among the 99 topologies was assessed by the RELL method [36,37]. We considered the branching position of a PolB fragment to be supported when the RELL bootstrap value for the best topology was ≥ 75%.Read coverageRead coverage for a contig was defined by dividing the cumulated size of reads contributing to the contig by the size of the contig.Relative abundance of PolBsFor the analysis of the relative abundance of PolB sequences, we used the same approach used by Williamson et al. [14]. Briefly, we first estimated the average number of reads overlapping with a part of a contig where a PolB domain was encoded, by taking into account the length of the PolB domain (as defined by the Pfam hit) and the length of the contig. The abundance of the PolB-sequences for each viral group for a given sample site was then quantified by the total number of reads associated with the relevant set of PolB-sequences (that is, the sum of the estimated read numbers). For a given site, the viral PolB proportion was computed by dividing the total number of viral PolB reads (for all viral groups) by the total number of reads obtained from the site.AbbreviationsASFV, African swine fever virus; CeV, Chrysochromulina ericina virus; EhV86, Emiliania huxleyi virus 86; GOS, Global Ocean Sampling; HaV, Heterosigma akashiwo virus 1; MSA, multiple sequence alignment; NCLDV, nucleocytoplasmic large DNA virus; NRDB, NCBI non-redundant amino-acid sequence database; ORF, open reading frame; PolB, B-family DNA polymerase; PoV, Pyramimonas orientalis virus; PpV, Phaeocystis pouchetii virus; RELL, resampling of estimated log likelihoods.Authors' contributionsAM performed the analyses. HO designed the experiments. All authors analyzed the data and contributed to the writing of the manuscript.Additional data filesThe following additional data are available with the online version of this paper. Additional data file 1 is a table listing the PolB sequences used in the study. Additional data file 2 is a multiple sequence alignment of 101 PolB sequences retrieved from databases. Additional data file 3 is a figure summarizing the results of the phylogenetic mapping of the GOS PolB fragments, which are displayed for each of the 99 branches tested.Supplementary MaterialAdditional data file 1The IDs and species names of the PolB sequences retrieved from databases are given. Sequences used in the reference multiple alignment are in bold.Click here for fileAdditional data file 2Sequences used in the final reduced reference multiple alignment are displayed with an asterisk.Click here for fileAdditional data file 3The GOS PolB fragments are displayed for each of the 99 branches tested. Numbers in parentheses (V/W) are the total number of mapped PolB fragments (W) and the number of supported cases (V) (displayed in red). Read coverage values are presented as follows: [X-Y]-(Z) where X and Y are the read coverage value range (minimum/maximum) and Z the read coverage median value.Click here for file\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2530870.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2530870",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2530870\nAUTHORS: Yong-Hwee E Loh, Lee S Katz, Meryl C Mims, Thomas D Kocher, Soojin V Yi, J Todd Streelman\n\nABSTRACT:\nLow coverage survey sequencing shows that although Lake Malawi cichlids are phenotypically and behaviorally diverse, they appear genetically like a subdivided population.\n\nBODY:\nBackgroundCichlid fishes from the East African Rift lakes Victoria, Tanganyika, and Malawi represent a preeminent example of replicated and rapid evolutionary radiation [1]. This group of fishes is a significant model of the evolutionary process and the coding of genotype to phenotype, largely because tremendous diversity has evolved in a short period of time among lineages with similar genomes [2-4]. Recently evolved cichlid species segregate ancestral polymorphism [5,6] and may exchange genes [7,8]. Numerous genomic resources have been developed for East African cichlids (many of which are summarized by the Cichlid Genome Consortium [9]). These include the following: genetic linkage maps for tilapia [10-12] and Lake Malawi species [10,13]; fingerprinted bacterial artificial chromosome libraries [14]; expressed sequence tag sequences for Lake Tanganyika and Lake Victoria cichlids [15]; and first-generation microarrays [16,17]. Many studies have used these resources to study cichlid population genetics, molecular ecology, and phylogeny (for review [18,19]). Recent reports have capitalized on the diversity among East African cichlids to study the evolution and genetic basis of many traits, including behavior [20], olfaction [21], pigmentation [22-24], vision [25,26], sex determination [24,27], the brain [28], and craniofacial development [10,13,29].In 2006, under the auspices of the Community Sequencing Program, the Joint Genome Institute (JGI) completed low coverage survey sequencing of the genomes of five Lake Malawi species. Species were chosen to maximize the morphological, behavioral, and genetic diversity among the Malawi species flock. This represents a novel genome project. Low coverage sequencing is now a routine strategy to uncover functional or 'constrained' genomic elements [30]. The rationale is as follows; one compares genome sequences of distantly related organisms (for example, shark, diverse mammals) with that of a reference (for instance, human, mouse), and outliers of similarity will be observed against the background expectation of divergence [31-34]. Our interests in diversity suggest a conceptually similar but logically reversed research objective. When the background expectation is similarity, how does one use low coverage genome sequencing to detect that which makes organisms distinct?Here, we report computational and comparative analyses of survey sequence data to address the question of diversity. We had four major goals: to produce a low coverage assembly for each of the five Lake Malawi species; to identify orthologs of vertebrate genes in these data; to predict single nucleotide polymorphisms (SNPs) segregating between species; and to use SNPs to evaluate the degree of genomic polymorphism and divergence at different evolutionary scales. Consequently, we produced assemblies for the five species ranging in aggregate length from 68 to 79 megabases (Mb), identified putative orthologs for more than 12,000 human genes, and predicted more than 32,000 cross-species segregating sites (with about 2,700 located in genic regions). We genotyped a set of these SNPs within and between Lake Malawi cichlid lineages and demonstrate signatures of differentiation on the background of similarity and polymorphism. Our work should facilitate further understanding of evolutionary processes in the species flocks of East African cichlids. Moreover, the approach we outline should be broadly applicable in other lineages where phenotypic and behavioral diversity has evolved in a short window of evolutionary time.ResultsSequence assemblyTrace sequences of five Lake Malawi cichlid species, namely Mchenga conophorus (MC; formerly genus Copadichromis), Labeotropheus fuelleborni (LF), Melanochromis auratus (MA), Maylandia zebra (MZ; formerly genus Metriaclima) and Rhamphochromis esox (RE), were downloaded from the GenBank Trace Archive and assembled into contiguous (contig) sequences. The average cichlid genome is 1.1 × 109 bases [35], so the traces represent a sequence coverage of 12-17% for each of the five species (see Additional data file 1). Through several quality filtering and assembly steps (see Materials and methods [below]), the resultant genomic assemblies of the five cichlid species yielded an average of 60,862 contigs with a mean length of 1,193 bases per contig. The total first-pass assembly sequence length for each species ranged from 68,238,634 bases (MA) to 79,168,277 bases (MZ), or about 7% of an average cichlid genome. Assembly statistics are shown in Table 1.Table 1First-pass genomic assembly statistics for five Lake Malawi cichlid speciesMCLFMAMZRETotal number of contigs in assembly61,92358,24563,29765,09455,751Total length (bases)73,425,56470,858,38168,238,63479,168,27771,295,074Genome coveragea (%)6.686.446.207.206.48Mean trace length (bases)1,0551,0929911,1451,153Shortest contig length (bases)5050505050Longest contig length (bases)19,63217,43721,60115,37121,351Mean contig length (bases)1,1861,2171,0781,2161,279Q25 contig length (bases)759846783805934Q50 (median) contig length (bases)9661,0639491,1631,113Q75 contig length (bases)1,4031,3551,1021,4171,407Total genic length (bases)2,863,110 (3.9%)2,841,933 (4.0%)2,761,941 (4.0%)2,851,968 (3.6%)2,797,548 (3.9%)aUsing an average cichlid genome size of 1.1 × 109 bases. LF, Labeotropheus fuelleborni; MA, Melanochromis auratus; MC, Mchenga conophorus; MZ, Maylandia zebra; RE, Rhamphochromis esox; Q25, 25th percentile; Q50, median or 50th percentile; Q75, 75th percentile.We noted that these first-pass assemblies were 'over-assembled' by roughly a factor of 2 when compared with theoretical expectations [36]. Theory suggests that random shotgun sequencing of single copy DNA, at 15% coverage of a 1.1 gigabase genome, will result in an assembly length of about 153 Mb. We reasoned that our assemblies might be shorter than expected because multicopy elements were grouped as if they were single copy sequence. Given the theoretical expectation (again for 15% coverage of a 1.1 gigabase genome) that individual bases should only be sequenced a maximum of four to five times, we examined whether contigs were built from five or more trace sequences contributing overlapping bases. We observed that about 10 Mb of each first-pass assembly were derived from such contigs, and excluded these data from subsequent analyses (for example SNP prediction [see below]). Notably, individual sequences contributing to these 'high trace number' contigs were not identified by RepeatMasker but did sometimes have Basic Local Alignment Search Tool (BLAST) matches to putative repetitive elements (for example, pol polyprotein, reverse transcriptase). Because of the keen interest in repetitive DNA families in cichlids [37] and other organisms [38], we have retained alignments of these 'high trace number' contigs and have marked them as such (see Additional data files 3 and 4).Gene content and coverageTo establish the extent of gene content and coverage present in each assembly, we carried out BLASTX similarity searches (10-10 E value cutoff) for each of the five assemblies against a reference human proteome (RefSeq proteins). The average proportion of putative genic sequence amounted to 3.9% of the available genomes. The MZ assembly contained the highest gene coverage, possessing genic loci that were significantly similar to approximately 5,240 unique human proteins. The remaining four species yielded approximately similar numbers ranging from 5,020 to 5,170 genes. It must be noted, however, that most of these genes are highly fragmented and incomplete, because of low coverage of the assembly. In all, a total of 36% (12,211 genes out of 34,180; see Additional data file 2) of the reference human proteome could be identified in one or more of the cichlid species.Clustering and alignmentWe obtained 25,458 clusters of putatively orthologous sequences, which were individually assembled into multi-species alignments for subsequent comparative analyses. Genic regions, as identified by similarity searches to known human and fish genes, were marked onto each alignment. Figure 1 illustrates a typical example of one such alignment.Figure 1Alignment of a typical cluster of orthologous sequences. (a) Overall alignment of assembly contigs from three different cichlid species with alignment positions indicated. (b) Expanded detail of nucleotide alignment. Filled pink block shows the expanded alignment corresponding to dotted red box in panel a. Filled blue block shows the alignment of corresponding species' traces that made up the assembly sequences. Lower case nucleotides have base quality scores under 20. Dashes '-' represent sequence unavailability. Asterisks '*' represent gaps inserted into the sequences. Dots '·' represent identity in alignment. Cap '^' represents segregating site. Alignment positions shown after consensus sequence. Polymorphism quality score shown below A-G single nucleotide polymorphism site.Roughly 1% of the alignments (294 alignments) showed percentages of variable sites above 2% (about tenfold higher than the average). It is impossible to know, given the low coverage of the sequenced genomes, whether these represent orthologous but divergent regions of cichlid genomes or the alignment of paralogous sequence. We therefore retained these alignments, and included a calculation of polymorphism for each alignment (see Additional data file 3), for the consideration of researchers using these data. For example, alignment 108,866 contains sequence with similarity to asteroid homolog 1, with 8% of sites variable and a majority of replacement polymorphism. Given the lack of functional information about this novel signaling protein (first described in Drosophila [39]), this alignment provides useful information even if (and perhaps because) it includes paralogous loci. Another 12% of the alignments (2,119 total) contained individual species contigs that had consensus base positions derived from five or more trace sequences (see above).For all subsequent analyses, we excluded 2,413 alignments that exhibited a high percentage of variable sites and/or higher than expected coverage. More than 11.6 million bases of multiple species alignments remain, of which roughly 1.06 Mb were inferred as genic. This included 10,902,011 (986,506 genic) bases of two-species alignments, 721,049 (75,371 genic) bases of three-species alignments, 27,951 (2,898 genic) bases of four-species alignments, and 877 (193 genic) bases of alignments containing all five species.Segregating sitesFurther analysis of these 11.6 million bases of multiple alignments identified a total of 32,417 (0.28%) cross-species SNPs. In order to classify the quality of an identified variable site, a polymorphism quality score (PQS) was defined, corresponding to the first digit of the lowest Phrap quality score among the nucleotides of the different species present at the polymorphic site (for example, a polymorphic site between four species with base quality scores of 34, 45, 46, and 50 would be assigned a PQS of 3). In total, 4,468 (13.8%) variable sites had a PQS of 5 or higher, 7,952 (24.5%) had a PQS of 4, 8,236 (25.4%) a PQS of 3, and the remaining 11,761 (36.3%) had a PQS of 2. PQS for each variable site are provided on the alignments described in Additional data file 3 (also available online [40]). Nucleotide diversity (Watterson's θw) averaged over two-, three-, and four-species alignments was 0.00257. Roughly 8% of all polymorphic sites (2,709) were located within the putative genic regions identified earlier. Alignments with fish and human proteins provided us with the phase information required to further classify these into 1,066 synonymous and 1,643 nonsynonymous SNPs. Summaries of all alignments containing genic and nongenic polymorphisms are provided in Additional data files 3 and 4.In order to investigate the pair-wise differences between any two of the five species, all sequence alignment segments with two or more species were broken up into all possible pair-wise alignments; this resulted in 1.06 to 1.55 Mb of alignment per pair. We then calculated the Jukes-Cantor distance between species pairs. The three shortest distances were between LF and MZ (0.229%), followed by MA/MZ (0.232%) and LF/MA (0.241%), and the greatest was between LF and RE (0.288%). These genetic distances include both within-species polymorphism and the fixed differences between species. Currently, there is no exhaustive estimate of within-species polymorphism for Malawi cichlids. Unpublished data from our own group (Streelman JT) indicates that for LF and MZ, within-species diversity (π) may be as high as 0.2%. Thus, the percentage of fixed genetic differences is likely to be extremely small in this assemblage (see following sections).Finally, we calculated the ratio of replacement to synonymous substitutions (Ka/Ks) for concatenated genic alignments among all pairs of species. We used concatenated sequences because each segment represented only a small fraction of a gene, with only few nonsynonymous and synonymous sites. Ka/Ks ranged from 0.380 in MC/LF to 0.562 in LF/MA. These numbers are greater than the ratios found between Fugu and Tetraodon (0.127 to 0.144 [41]). Such high Ka/Ks values may indicate that positive selection, driven by adaptive radiation, is prevalent in cichlid fishes. However, given the expectation of few fixed differences between groups, this topic should be revisited with more data on the levels of segregating and fixed nucleotide substitutions among lineages.Validation and generality of SNPsWe genotyped 96 SNPs in 384 Lake Malawi cichlid samples using Beckman Coulter SNPstream™ technology (Beckman Coulter, Inc., Fullerton, CA). The SNPs were partitioned into three categories to help us evaluate the comparative success rate of automated SNP prediction. First, we included 13 positive controls: genes previously sequenced by others [3,25] and by us (Streelman JT, unpublished data), with expected variation in Malawi cichlids. Positive controls included genes involved in morphogenesis (otx1, otx2, and pax9), pigmentation (mitf, ednrb, and aim1), and visual sensitivity (opsins rh1, sws1, lws, sws2a, and sws2b). Next, we genotyped 59 SNPs identified using the automated procedure described in this report. We selected these SNPs to represent a range of PQS (from 2 to 5) and a variety of sequence types (genic, nongenic with a BLAST match < e-100 to Tetraodon, and nongenic with no BLAST match). Finally, we wished to compare our automated SNP selection to a manual approach. Therefore, we included an additional 24 SNPs identified by manual inspection of BLAST matches between single JGI traces and Tetraodon chromosome 11; we have previously shown Tetraodon 11 to share orthologs with cichlid chromosome 5 [13]. Note that these SNPs were most often not discovered by our automated procedure because they originated in single traces that did not meet percentage quality cutoffs and/or they did not align into comparative contigs because of overlap cutoffs.Our validation strategy sought to document the general use and segregation of these markers among Lake Malawi cichlids. Given recent divergence times among species (some as recent as 1,000 years [2]), we expected that SNPs might segregate throughout the assemblage. Therefore, Malawi samples comprised about ten individuals from each of ten populations of MZ and LF, as well as one to five individuals of 77 additional species (25 of which were rock-dwelling mbuna). Taxa were included to represent the morphological, functional, and behavioral diversity of the Malawi lineage, which may contain more than 800 species [42].Ten out of 13 (about 77%) positive controls gave reliable genotypes and were variable across the dataset. For the 59 SNPs predicted by our automated procedure, 11 were fixed (no variation) in all samples, indicating an error in sequencing (or genotyping), an error in prediction, or the presence of a low frequency allele in the sequenced samples. Six predicted SNPs did not produce data reliable enough for genotype calls. The remaining 42 loci from automated predictions (about 71%) were polymorphic across the dataset. For 24 SNPs predicted using manual similarity searches, four were fixed and four failed reliability for genotype calls, with the remaining 16 loci (about 67%) showing polymorphism (Table 2). Twelve out of 20 (60%) predicted SNPs with PQS of 3 or less were successful, whereas 30 out of 39 (76%) predictions with PQS of at least 4 yielded polymorphisms (Table 3). There is evidence of ascertainment bias in our genotypic data (see Additional data file 5). For example, three SNP loci (Aln100674, Aln114498, and Aln102321) exhibit alleles unique to Rhamphochromis. Similarly, SNPs predicted from comparisons of RE and mbuna (LF, MA, and MZ) are sometimes fixed in mbuna. Polymorphisms predicted from comparisons of mbuna taxa are more likely to vary within LF and MZ populations and across mbuna species.Table 2SNP genotyping success categorized by detection methodSNP detection methodControl genesAutomatedManual BLASTNumber of genotyped loci135924Number of polymorphic loci104216Number of fixed loci3114Number of failed loci064Successful SNP detection (%)76.971.266.7BLAST, Basic Local Alignment Search Tool; SNP, single nucleotide polymorphism.Table 3SNP genotyping success categorized by polymorphic quality scorePolymorphic quality score2345Number of genotyped loci5152811Number of polymorphic loci210246Number of fixed/failed loci3545Successful SNP detection (%)4066.785.754.5SNP, single nucleotide polymorphism.Genetic polymorphism and divergence at multiple scalesStrikingly, among all 68 loci showing polymorphism, no SNP locus was alternately fixed between LF and MZ, or between rock-dwelling mbuna and non-mbuna. We thus sought to investigate the degree of polymorphism versus divergence at multiple evolutionary scales.The data (Additional data file 5) support the previously reported population structures in MZ [43,44] and LF [45], as well as the genetic distinction between these species (MC Mims, unpublished data). For example, mean genetic differentiation (FST) in MZ is 0.148 and in LF is 0.271. Mean FST between LF and MZ was 0.215, and between mbuna (25 species) and non-mbuna (52 species) it was 0.224, demonstrating that most genetic variation segregates within and not between lineages, regardless of evolutionary scale. Nevertheless, these distributions of FST yielded statistical outliers, which exhibit greater than average genetic differentiation (Figure 2). Four loci were found to be statistical outliers for FST among MZ and LF populations. In MZ the opsin loci lws (FST = 0.514), sws1 (0.572) and rh1 (0.733), and in LF the opsin locus rh1 (0.853) exhibit differentiation between populations. Between LF and MZ, three loci were identified as outliers: a nonsynonymous polymorphism in csrp1 (FST = 0.893), a synonymous polymorphism in β-catenin (Aln101106_1089; FST = 0.904), and an intronic polymorphism in ptc2 (Aln100281_1741; FST = 0.863). Two statistical outliers were identified for FST between rock-dwelling mbuna and non-mbuna groups: a nonsynonymous polymorphism in irx1 (Aln102504_1609; FST = 0.984), and a nongenic polymorphism (Aln103534_280; FST = 0.919) in sequence with similarity to pufferfish and stickleback genomes between contactin 3 and ncam L1.Figure 2Box-and-whisker plots of FST values. FST values were calculated for the following: within MZ, within LF, LF versus MZ, and Mbuna versus non-Mbuna. Upper and lower box bounds represent 75th and 25th percentiles, respectively. The solid lines within boxes represent the median value. Whiskers mark the furthest points from the median that are not classified as outliers. Unfilled circles represent outliers that are more than 1.5 times the interquartile range higher than the upper box bound. FST, genetic differentiation; LF, Labeotropheus fuelleborni; MA, Melanochromis auratus; Mb, megabases; MC, Mchenga conophorus; MZ, Maylandia zebra.Genetic clustering and ancestryTo further visualize the segregation of SNPs across the Malawi cichlid flock, we utilized a Bayesian approach that assigns individuals to a predefined number of genetic clusters [46]. Specifically, we were interested in how species would be assigned to major Malawi cichlid lineages identified in previous studies [3,4,47]. There are three such groups supported by the majority of molecular data: the rock-dwelling mbuna; pelagic and sand-dwelling species; and a group comprised of Rhamphochromis, Diplotaxodon, and other deep-water taxa. Analysis of 68 SNP loci accurately classifies species to respective lineages (Figure 3). For instance, all species considered mbuna (blue) cluster with other mbuna, to the exclusion of other groups; species thought to represent the earliest divergence within the species flock (Rhamphochromis) clustered together as a separate group (green); all remaining non-mbuna species formed the third group (red). Notably, deepwater genera Diplotaxodon and Pallidochromis contain individuals with mosaic genomes (red and green) and Astatotilapia calliptera, a nonendemic species and possible Malawi ancestor [48] combines mbuna and non-mbuna genomes.Figure 3Bayesian assignment of Lake Malawi cichlids to different evolutionary lineages. We show the contribution to each individual genome (q, which ranges from 0% to 100%) from each of K = 3 predefined genetic clusters (blue, red, and green), for data derived from single nucleotide polymorphisms (SNPs) in Tables 2 and 3. Note that this method predefines the number but not the identity of genetic clusters. Species names are written once; multiple individuals from species are grouped together (for example, four individuals of Pseudotropheus crabro). Species considered mbuna (blue) cluster with other mbuna, to the exclusion of other groups; species thought to represent the earliest divergence within the species flock (Rhamphochromis) clustered together as a separate group (green); and all remaining non-mbuna species formed the third group (red).For comparison, additional analyses were performed setting the predefined number of genetic clusters to from two to five. When set to two genetic clusters, species were accurately classified as mbuna or non-mbuna. At settings of four or five, the program was unable to yield stable classification results between replicate runs. Thus, these latter three sets of analyses (data not shown) did not provide any further insights into the genetic lineages of Malawi cichlids.DiscussionAfrican cichlid fishes are important models of evolutionary diversification in form and function [44]. They are singularly remarkable for the extent of phenotypic and behavioral diversity on a backdrop of genomic similarity. Lake Malawi is home to the most species-rich assemblage of African cichlids; as many as 800 to 1,000 species are thought to have evolved from a common ancestor during the past 500,000 to 1 million years ago [42]. These recently formed species segregate ancestral polymorphism and exchange genes by hybridization [5,7,49]. Such circumstances present both opportunities and challenges for understanding evolutionary history and biological diversity. Opportunistically, researchers have used molecular markers across studies to interrogate the genetic basis of phenotypic differentiation [13,22,24,29]. This approach views Malawi cichlid species as natural mutants screened for function by natural selection, with essentially identical ancestral genomes honed by contrasting historical processes. By contrast, the task of reconstructing a phylogeny of species has been hindered by the very same phenomena of genomic similarity and mosaicism [2,3]; even the promising approach of Amplified Fragment Length Polymorphism (AFLP) does not provide strong resolution of the relationships among genera [23,48,50,51]. The data we present here should provide new resources and perspectives for cichlid evolutionary genomics.Cichlid species exhibit genomic polymorphismLake Malawi cichlid species sequenced by the JGI embody the phylogenetic, morphological, and behavioral diversity found within the assemblage. Rhamphochromis esox (RE) is a large (about 0.5 m) pelagic predator that represents one of the basal lineages of the species flock [3,4,47]. Mchenga conophorus (MC) is a sand-dwelling species that breeds on leks, where males construct 'bowers' to attract females. Melanochromis auratus (MA), Maylandia zebra (MZ), and Labeotropheus fuelleborni (LF) are rock-dwelling (mbuna) species that differ in color pattern, trophic ecology, body shape, and craniofacial morphology (pictures of these and others are available online [52]).Our data confirm the conclusions from previous genetic analyses on a smaller scale; Lake Malawi species are genetically similar. Nucleotide diversity observed among the five cichlid species (Watterson's θw = 0.26%) is less than that found among laboratory strains of the zebrafish Danio rerio (Watterson's θw = 0.48% [53]). Although overall nucleotide diversity is less than that observed in Danio, the ratio of replacement to silent change is nearly fivefold higher in the Lake Malawi genomes. Such a result might suggest that East African cichlid evolution is characterized by adaptive molecular evolution, as has been indicated in a few instances [25,54], or a relaxation of purifying selection attributable to small effective population size. However, we should view this estimate of Ka/Ks with caution because of one of the remarkable features of these data (see below). Variable sites identified from cross-species alignments are not substitutions fixed between species. The Ka/Ks approach to identifying selection may be largely inappropriate for such young species where ancestral alleles segregate as polymorphisms.The pattern of variation observed across the approximately 75 species genotyped in this study demonstrates that biallelic polymorphisms segregate widely throughout the Malawi species flock. SNPs segregate within and between MZ and LF populations, as well as within and among mbuna species and other lineages. No SNP locus surveyed is alternately fixed in LF versus MZ, nor between mbuna and non-mbuna. Remarkably, the degree of genetic differentiation (FST) within species is roughly equivalent to that between species and to that between major lineages. Lake Malawi cichlid species are mosaics of ancestrally polymorphic genomes. Add to this a propensity of recently diverged species to exchange genes [2], and Malawi cichlids present a case of complex and dynamic evolutionary diversification, where recombination and the sorting of ancestral polymorphism may be more important than new mutation as sources of genetic variation. Despite allele sharing, SNP frequencies contain a clear signal of ancestry for the entire flock. Rock-dwelling mbuna comprise a genetic cluster, as do pelagic and sand-dwelling species, in addition to Rhamphochromis. Notably, Astatotilapia calliptera, one of a few nonendemic haplochromines in Lake Malawi, appears to retain a reservoir of ancestral polymorphisms from which mbuna and non-mbuna genomes have emerged.Genomic polymorphism and the divergence of Malawi cichlidsOur hierarchical sampling design allows us to consider whether there are loci exhibiting extreme genetic differentiation against the background of shared polymorphism within species, between species, and between major lineages. Strikingly, regardless of the evolutionary scale, statistical outliers comprise approximately 3% to 5% of loci surveyed. Opsin loci lws, rh1, and sws1 are differentiated among populations of LF and MZ, adding to reports that opsin polymorphisms are associated with population-specific color patterns or visual environments [55].SNPs in csrp1, β-catenin, and ptc2 exhibit greater than expected differentiation between LF and MZ. Csrp1 (cysteine-rich protein) is a vertebrate LIM-domain family member acting in the noncanonical WNT pathway, expressed in gut, intestine, and cardiac mesoderm [56]. β-catenin acts to transduce signals in the canonical WNT pathway [57] and is expressed in developing cichlid fins, dentitions, brains, and lateral lines (Fraser GJ, Streelman JT, unpublished data). Patched is a receptor for sonic hedgehog [58]; both areexpressed in developing cichlid dentitions, jaws, and brains (Fraser GJ, Sylvester JB, Streelman JT, unpublished data). A SNP in irx1 nearly perfectly differentiates rock-dwelling mbuna from the remainder of the Malawi species flock. Irx1 acts to position the boundary between the telencephalon and the posterior forebrain [59]. Finally, a SNP located between contactin 3 and ncam L1 exhibits differentiation between mbuna and non-mbuna lineages; these genes are linked in other genomes and functionally interact to pattern dendritic branching in the neocortex [60]. Taken together, differentiated loci are interesting in the context of cichlid diversification because they affect the phenotypes that vary among lineages: color and vision [25,26], guts [61], dentitions [13,62], jaws [10,29], and brains [28].Discovery for evolutionary biologyThere are obvious challenges when attempting to extract information from low coverage genomic sequence, and also obvious payoffs [31-34]. Most previous studies have used this information for species-specific discovery (for example, dog breeds) or broad evolutionary comparisons with respect to a reference genome (for example, dog-human, shark-human, or cat-mammal). Our goals in the present analysis stem from the unique characteristics of Lake Malawi cichlids; these are biologic species that behave genetically like a single subdivided population. Therefore, our biggest challenge was to devise a strategy that retains information from these low coverage survey sequences (75% genomic coverage spread over five closely related species), but minimizes error and bias in assembly and cross-species alignment for SNP identification. For example, we excluded many contigs because they appeared to be over-assembled, and we excluded multi-species alignments if they exceeded a polymorphism threshold. The over-assembly problem limits the coverage of these genomes in relation to expectation; this phenomenon, observed in the cat genome and in simulation, has complex and varying causes and has yet to be fully resolved [63]. It is likely to be mitigated to some degree by comparison with a higher coverage reference sequence. The power of the data we present comes from the broad utility of the genic sequences and SNPs we have identified for many questions in genomic evolutionary biology.Our analyses identified about 12,000 Lake Malawi cichlid sequences with similarity to human and fish proteins. This is a significant advance in our understanding of cichlid genomic content. To put this in context, approximately 13,500 unique expressed sequence tags, from three different East African cichlids, represent the sum total of such publicly released sequences [15]. Our contribution roughly doubles the available data.The approximately 32,000 (2,700 genic) SNPs we identified should provide a wealth of molecular markers for studies of population genetics and molecular ecology, linkage and quantitative trait locus mapping, association mapping, and phylogeny. We convert about 70% of predicted SNPs to polymorphic markers; this percentage is comparable to that of other studies from white spruce (74% to 85%, depending on quality cutoffs [64]), zebrafish (65% [53]), and cow (43% [65]). We have shown these biallelic markers to be of general use, many segregating across the major cichlid lineages of Lake Malawi. We used the SNPs to assign Malawi species to ancestral genetic clusters, and this approach should hold promise for similar questions of genetic structure that span the population versus species continuum. It is important to note that early runs of this analysis, with fewer SNP loci, resulted in stable results with more individuals showing mosaic genomes. This suggests that careful consideration should be given to the number of polymorphic loci necessary to yield confidence in evolutionary interpretation. As more SNP loci (with known genome coordinates) are assayed, it will be possible to compute and compare ancestry proportions across scales (for example, genome versus chromosome versus gene cluster).Notably, we have used the background level of genomic similarity and polymorphism to identify loci that may have experienced a history of selection within species, between species and between major lineages. Because SNP markers are co-dominant, easy to genotype, reliable and reproducible from laboratory to laboratory, and readily mapped in silico (NHGRI will sequence a related cichlid, the tilapia, to 7-fold draft assembly coverage in 2008), they are likely to complement microsatellites and AFLP for most applications in cichlid evolutionary genomics. Given the unique mosaic structure of Lake Malawl cichlid genomes, it is exciting to envision experiments employing SNPs to identity genotype-phenotype associations, using the entire species flock as a mapping panel. Finally, as sequencing costs continue to drop, the approach we outline here should prove applicable to those studying evolutionary and phenotypic diversity among closely related species [44].Materials and methodsSamplesIndividuals of Mchenga conophorus (MC), Labeotropheus fuelleborni (LF), Melanochromis auratus (MA), Maylandia zebra (MZ), and Rhamphochromis esox (RE) were sampled from the wild during an expedition to Malawi in 2005. Specimens prepared for survey sequencing by the JGI were collected from Mazinzi Reef (MZ), Domwe Island (LF and MA), and Otter Point (MC and RE), all of which are locales in the southeastern portion of the lake. High-quality DNA was extracted and prepared in the laboratory of TDK.Trace sequencesTrace sequences generated by the JGI for MC, LF, MA, MZ, and RE, together with their sequence quality scores, were downloaded (6 May 2007) from the National Center for Biotechnology Information (NCBI) Trace Archive. The dataset for each species consisted of an average of about 152,000 individual trace reads with total read lengths ranging from 137 to 185 million bases. Detailed sequence statistics for each species are provided in Additional data file 1.Sequence preprocessing and assemblyThe trace and quality sequences were first pre-processed for assembly by masking out all possible vector sequences available from the NCBI UniVec vector sequence database (downloaded 6 May 2007). The vector masking was performed using the cross_match.pl perl script provided by the Phred-Phrap package [66]. In order to reduce the computational complexity and time required for the final assembly, repeat sequences were masked before assembly using RepeatMasker version 3.1.8 (Smit AFA, Hubley R and Green P, unpublished data) in conjunction with the latest repeatmasker libraries from RepBase Update [67]. Bases with sequencing quality score of less than 20 were also masked. The actual assembly of each species' trace sequences into contiguous sequences (contigs) was then performed using the Phrap version 0.990329 assembly program from the Phred-Phrap package. Contigs with more than 80% low quality bases (defined as <20 assembly quality score) were removed from the assembly. This whole genome shotgun project has been deposited at DDBJ/EMBL/GenBank under the project accessions ABPJ00000000 (MC), ABPK00000000 (LF), ABPL00000000 (MA), ABPM00000000 (MZ), and ABPN00000000 (RE). The versions described in this paper are the first versions: ABPJ01000000, ABPK01000000, ABPL01000000, ABPM01000000, and ABPN01000000.Similarity search and alignmentOrthologous genomic contig pairs were first identified using reciprocal BLASTN similarity searches with a strict E-value cutoff of 10-100, performed across the sequence contigs of all possible species pairs. To reduce spurious ortholog assignments, putative ortholog contig pairs were only retained if their regions of high sequence similarity formed good end-to-end overlaps (defined as within 100 bases of the 5' end or 30 bases from the 3' end of a sequence) or overlap more than 80% of the shorter contig. Although some of the filtered regions could represent biologically relevant loci where recombination or translocations might have occurred, we decided to remove them from this analysis. Contig pair assignments were then passed to an algorithm that created clusters of contigs whereby each contig within the cluster must be related to all other contigs in the cluster through one or more putatively orthologous relations.Each cluster of contigs was then individually aligned using Phrap, resulting in a continuous alignment tiling path where each alignment position may consist of a base from any one or up to all five cichlid species (Figure 1). Segregating sites were then identified from alignment positions with high quality bases (>20 score) from two or more species. A PQS was defined, corresponding to the first digit of the lowest Phrap quality score among the nucleotides of the different species present at the polymorphic site (for example, a polymorphic site between four species with base quality scores of 34, 45, 46, and 50 would be assigned a PQS of 3). To compare the extent of nucleotide diversity among the five cichlid species, we calculated Watterson's theta (θw [68]). This measure takes into account the number of variable positions and the sample size analyzed. Our data violate the assumption of an infinite, interbreeding population, but we chose this metric to in order to make direct comparisons to similar measures from study of other genomes (for example, zebrafish).Protein-coding sequence identificationCichlid protein coding sequences were inferred based on similarity searches to known protein databases of fishes and humans. BLASTX searches with E-value cutoff of 10-10 were performed for the each cichlid genomic assembly as well as the overall consensus sequence of the cluster alignments, against a protein database made up of all GenBank Actinopterygii (ray-finned fishes) sequences (downloaded 2 June 2007; 163,471 entries) and all human RefSeq proteins (downloaded 25 June 2007; 34,180 sequences). The alignment with the highest scoring hit for each genomic locus was then used as a reference to determine the coding strand and phase of the protein-coding cichlid locus.Evolutionary sequence divergence among JGI speciesAll cluster alignment segments with contributing bases from two or more species were split into pairwise alignments (each two, three, four, or five species alignment position can be split into one, three, six, or ten pair-wise alignments respectively). Pair-wise alignments within each of the ten possible species pair combinations (MC-LF, MC-MA, MC-MZ, MC-RE, LF-MA, LF-MZ, LF-RE, MA-MZ, MA-RE, and MZ-RE) were then concatenated and the number of substitutions counted. Jukes-Cantor correction for multiple substitutions was applied to these direct distance measurements [69]. Pair-wise alignments consisting of only genic sequences were obtained from multi-species cluster alignment segments in a manner similar to that described above. The DNAStatistics package of Bioperl [70] was then used to calculate the Ka/Ks values of pair-wise alignments.Genotyping and validation of SNPsWe genotyped 96 SNPs in 364 diverse Lake Malawi cichlid samples. These SNPs included 13 positive controls, 59 loci from the automated procedure described in this report, and an additional 24 loci chosen manually by BLAST of individual traces to the Tetraodon genome (see main text for further description). The GenomeLab SNPstream Genotyping System Software Suite v2.3 (Beckman Coulter, Inc.) was used for experimental setup, data uploading, image analysis, genotype calling and QC review, at Emory University's Center for Medical Genomics. In brief, marker panel data (multiplexed SNP panel designed by SNPstream's Primer Design Engine website [71]) were first uploaded to the SNPstream database using the PlateExplorer application software. Also uploaded was the Process Group Data containing all test sample information generated through a Laboratory Information Management System (Nautilus 2002; Thermo Fisher Scientific, Waltham, MA, USA). An on-board CCD camera of the SNPstream Imager took two snapshot images of each well of the 384-well tag array, one under a blue excitation laser and the other under a green excitation laser. Image application software was used to analyze the captured images to detect spots, overlay an alignment grid, and determine spot intensity. The fluorescent pixel intensity data for each SNP under the two channels, representing the relative abundance of the two alleles, were uploaded to the database. The GetGenos application software was used to calculate and generate a Log(B+G) versus B/(B+G) plot, where B and G were the pixel intensities under the blue and green channels, respectively, for each sample and each SNP. Next, automated genotype calling was accomplished using the QCReview application software based on a number of criteria (for instance, signal baseline, clustering pattern of the three genotypes, and Hardy-Weinberg score). A genotype summary was generated using the Report application software.Genetic differentiation within and among lineagesLocus-specific FST [72] was calculated using FSTAT version 2.9.3.2 [73] for three evolutionary scales: within LF and MZ; between LF and MZ; and between mbuna and non-mbuna. We determined that a SNP locus was a statistical outlier using the empirical distribution of FST values. FST outliers exceed the sum of the upper quartile value and 1.5 times the interquartile range.Genomic assignmentWe used a Bayesian method (STRUCTURE v.2.2 [46]) to determine how well our SNP genotypes assigned individuals to evolutionary lineages. We chose to define the number of K genetic clusters in accord with previous research showing about three major evolutionary groups of Lake Malawi cichlids [3-5,47]. Note that we do not intend this to mean that three is the best supported estimate of K in these data; our rationale is rather to demonstrate how individual genomes are composites (or not) of the major evolutionary lineages found in the lake. Thus, we used the admixture model to estimate q, the proportion of each genome derived from each of K genetic clusters. For comparison, we also ran analyses with K set to two, four, or five (not shown). Each run of the program included 50,000 cycles of burn-in and run length of 50,000 steps. Multiple runs were conducted to ensure reliability and consistency of results.AbbreviationsBLAST, Basic Local Alignment Search Tool; FST, genetic differentiation; JGI, Joint Genome Institute; Ka/Ks, ratio of replacement to synonymous substitutions; LF, Labeotropheus fuelleborni; MA, Melanochromis auratus; Mb, megabases; MC, Mchenga conophorus; MZ, Maylandia zebra; NCBI, National Center for Biotechnology Information; PQS, polymorphism quality score; RE, Rhamphochromis esox; SNP, single nucleotide polymorphism.Authors' contributionsYHL, JTS, SVY, and TDK conceived the idea and designed the study. YHL, LSK, and MCM performed the research. YHL and JTS analyzed the data and drafted the manuscript. All authors read and approved the final manuscript.Additional data filesThe following additional data are available with the online version of this paper. Additional data file 1 is a table of trace sequence statistics of five Lake Malawi cichlid species. Additional data file 2 is a list of human gene homologs found in the five cichlid species. Additional data file 3 is a list of alignments and polymorphic sites. Additional data file 4 is a list of alignments with BLAST hits to fish and humans. Additional data file 5 is a table of major allele frequencies for biallelic SNPs surveyed across Lake Malawi cichlid populations and species.Supplementary MaterialAdditional data file 1Presented is a table of trace sequence statistics of five Lake Malawi cichlid species.Click here for fileAdditional data file 2Presented is a list of human gene homologs found in the five cichlid species.Click here for fileAdditional data file 3Presented is a list of alignments and polymorphic sites.Click here for fileAdditional data file 4Presented is a list of alignments with BLAST hits to fish and humans.Click here for fileAdditional data file 5Presented is a table of major allele frequencies for biallelic SNPs surveyed across Lake Malawi cichlid populations and species.Click here for file\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2531074.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2531074",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2531074\nAUTHORS: Emanuel Tanne\n\nABSTRACT:\nThis report highlights a conference designed for patient education on elevated cerebrospinal fluid (CSF) pressure. The conference centered on chronic intracranial hypertension (IH) including the latest research and clinical information. It was sponsored by the Intracranial Hypertension Research Foundation and held at the University of Texas Medical School, Houston, on June 21–22nd, 2008.\n\nBODY:\nBackgroundThe Intracranial Hypertension Research Foundation (IHRF) in Vancouver, Washington, promotes research into the pathophysiological basis of chronic IH and the evolution of better treatment and, ultimately, a cure. IHRF is a multi-functional organization. It not only encourages research, but also facilitates understanding and management of chronic primary and secondary IH, through research, training and education programs worldwide.IHRF sponsors programs for researchers and clinicians, as well as educational conferences for patients and families. The Houston conference was attended by 108 patients and family members who heard presentations by nine clinicians and researchers.Patients asked many questions during extensive panel sessions. The conference also allowed patients to speak about personal experiences with the disorder. IHRF held the first conference on chronic IH in 2006 at the Oregon Health & Science University (OHSU). A wide variety of subjects were presented to enhance patient knowledge and awareness. Patient education importantly improves patient insight about the disorder. It also facilitates active cooperation with their physicians. Understanding limitations of accepted medical and surgical treatment leads to realistic goals in management.Additionally, patient education is vital in controlling health care costs.DiscussionThe opening presentation by Conrad Johanson, Professor of Clinical Neuroscience at Brown University dealt with CSF production in the choroid plexus and its possible role in IH. Covered in the lecture was the structure of the choroid plexus, the dynamic turnover of ions and water in CSF production by the choroidal epithelium, the transport of CSF solute for the brain, and homeostasis of CNS extracellular fluid. A discussion of translational research goals involving the choroid plexus and CSF and possible directions for research in IH concluded the presentation.John McGregor, Associate Professor of Neurosurgery at Ohio State University, spoke on \"Neural Hydrodynamics Disorders: Hydrocephalus, Intracranial Hypertension and the Chiari Malformation. Relationships, Similarities and Differences\" which was followed by two lectures covering CSF shunt and valve technology. The latter two presentations discussed in depth the available technology with details of the features of each device.The conclusion was that despite the numerous designs, the overall success rate for the various shunts and valve equipment are similar. Better designed shunts are greatly needed.The next lecture by Conrad Johanson: \"Current Theories on Causation and Reduction of Elevated CSF Pressures: Implications for Intracranial Hypertension\", began with a discussion of CSF fluid dynamics centering on CSF reabsorption including the controversial arachnoid and CSF lymphatic drainage. The role of neuropeptides such as atrial natriuretic peptide (ANP) in CSF production and the role of growth factors like basic fibroblast growth factor (FGF2) in CSF reabsorption, were identified. The intriguing possibility of peptides as pharmacological agents in controlling intracranial pressure (ICP) was thought provoking and stimulating, especially since no specific drug is available to control production and egress of CSF.Steven Katz, neuro-ophthalmologist and Associate Professor of Ophthalmology at Ohio State University, lectured on the symptoms and signs in idiopathic intracranial hypertension (IIH) and the medical management options. He discussed approaches that work best in his practice. Steven Katz had previously demonstrated the somatostatin receptors 1 and 2 in normal human choroid plexus and arachnoid granulations and thus surmised that somatostatin is involved in CSF production and egress. His preliminary clinical use of octreotide, a peptide that mimics somatostatin, was discussed, including the potential complications of somatosatin analogs to control IH. Steven Katz described his techniques for optic nerve sheath decompression and demonstrated findings from many of his procedures. He attributed excellent outcomes to reduced surgery duration, i.e. especially the time of optic nerve stretching. He emphasized that short exposure and optimal surgical approach lead to minimal diplopia, ptosis and most importantly, minimal vision loss.Marc Criden, neuro-ophthalmologist and Assistant Professor of Ophthalmology at the University of Texas, Houston, presented an updated overview of pediatric IH. In children under 10 years, gender and weight are not the factors they are in adults. He pointed out that many physicians consider IH in children under 10 to be a different disorder because of these characteristic differences. Headaches associated with IH and how best to manage them were discussed by neurologist and neuro-ophthalmologist, Leonard Hershkowitz of Baylor University. He indicated his management techniques that worked well for him.A reception followed in which patients again communicated with the speakers.The 2nd conference day opened with a discussion of the mission and goals of IHRF and a review of IHRF-sponsored research by Emanuel Tanne, Clinical Assistant Professor of Ophthalmology, OHSU and president of IHRF. He pointed out that IHRF works to remove significant obstacles to research: under-funding, under-coordination of effort, incomplete recognition of the life-altering effects of IIH, and low recruitment of researchers in this area. IHRF funded animal model development projects at the University of Chicago and the University of Arkansas. Under investigation is a knockout mouse that develops IH shortly after weaning. Other research areas included joint testing at the University of Utah of a NASA-developed, non-invasive, closed loop ultrasound device to measure ICP in microgravity and an investigation of vitamin A receptors in arachnoid granulations at Ohio State University. IHRF also partners with the Casey EyeInstitute, in regard to the Intracranial Hypertension Registry at the Oregon Health & Science University, Portland. The IH Registry is a relational database management system designed for medical research. Emanuel Tanne discussed ongoing Registry research, including studies in genetics, economics, weight gain and pregnancy. Jessica Tanne, IHRF Director, Communications & Development, discussed raising community awareness, fundraising and the importance of becoming an IH ambassador. Illustrating and discussing their fundraising creativity were Dori Clements and Jacque Tate, both parents of IH children.Clark Sitton, Assistant Professor of Radiology, University of Texas, Houston, discussed types of imaging studies and the goals of imaging in IH. He presented an extensive collection of studies demonstrating pathological findings associated with IH.Bariatric surgery as a possible option in the management of IH was presented by Erik Wilson, Assistant Professor of Surgery, University of Texas, Houston, who covered risks and benefits of bariatric surgery, using his extensive experience as a guide. He also discussed types of procedures available and his personal approach to follow-up and long term goals for patients. Marc Criden presented a new hypothesis to consider for clinical management of chronic IH: establishing target pressures for each patient with IH. The hypothesis arose out of his and Steven Katz's experience with patients at the last IHRF patient conference, where they were impressed by the variety of symptoms and patient reports contradictory to current teaching. They hypothesized that chronically elevated ICP is neurologically damaging, even in patients without significant visual dysfunction or intractable headache. They are presently investigating and evaluating this hypothesis. The final speaker, Kapil Kapoor, resident in ophthalmology at the University of Texas, Galveston, presented his research on hyposmia and IH. He found that patients with elevated ICP have a decreased sense of smell and concluded that other nerves such as the olfactory nerve may be functionally compromised in the setting of IH by a mechanism similar to that of optic nerve compression. Therefore, it may be appropriate to consider IH as a more global neuroanatomic insult of augmented CSF pressure than previously considered.ConclusionWhile the intent of educating patients was admirably accomplished by this conference, the conference served as a unique experience for researchers and clinicians to hear from a large group of patients about the nature of their disorders. Therefore, not only did physicians from a variety of sub-specialties have an opportunity to exchange ideas and explore future collaborative projects they also left with a different perspective of chronic IH and a new appreciation of patient-centered conferences as a mechanism to expedite translational CSF research.Competing interestsThe author declares he has no competing interests.Authors' contributionsI am the sole author and have read and approved the final version of this paper.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2531077.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2531077",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2531077\nAUTHORS: Roger Wong, Jaheed Khan, Temi Adewoyin, Sobha Sivaprasad, Geoffrey B Arden, Victor Chong\n\nABSTRACT:\nBackgroundTo assess the ability of the Chromatest in investigating diabetic maculopathy.MethodPatients with Type 2 diabetes and no concurrent ocular pathology or previous laser photocoagulation were recruited. Visual acuities were assessed followed by colour contrast sensitivity testing of each eye using Chromatest. Dilated fundoscopy with slit lamp biomicroscopy with 78 D lens was then performed to confirm the stage of diabetic retinopathy according to the Early Treatment Diabetic Retinopathy Study.Results150 eyes in 150 patients were recruited into this study. 35 eyes with no previous laser photocoagulation were shown to have clinically significant macular oedema (CSMO) and 115 eyes with untreated non-proliferative diabetic retinopathy (NPDR) on fundus biomicroscopy. Statistical significant difference was found between CSMO and NPDR eyes for protan colour contrast threshold (p = 0.01). Statistical significance was found between CSMO and NPDR eyes for tritan colour contrast threshold (p = 0.0002). Sensitivity and specificity for screening of CSMO using pass-fail criterion for age matched TCCT results achieved 71% (95% confidence interval: 53–85%) and 70% (95% confidence interval: 60–78%), respectively. However, threshold levels were derived using the same data set for both training and testing the effectiveness since this was the first study of NPDR using the ChromatestConclusionThe ChromaTest is a simple, cheap, easy to use, and quick test for colour contrast sensitivity. This study did not achieve results to justify use of the Chromatest for screening, but it reinforced the changes seen in tritan colour vision in diabetic retinopathy.\n\nBODY:\nBackgroundThe debilitating nature of untreated diabetic retinopathy promotes the need for cost-effective screening methods. Various studies have shown that cost effective screening can reduce blind registration due to diabetes [1-3]. Although seven field 30 degree stereo colour fundus photographs are the gold standard for diabetic screening, both remain relatively expensive and difficult to obtain [4,5]. In the UK, the National Screening Program for Diabetic Retinopathy utilises non-stereo digital photography as this meets the Diabetes UK standards for sensitivity and specificity.Non-stereo fundus imaging is easier to obtain but has limitations in establishing macular oedema [6]. There is evidence that tritan colour vision is diminished in patients with diabetic maculopathy, but testing with the FM100 hue and Farnsworth-Lanthony D-15 test are labour intensive and time consuming [7]. Colour vision testing with a computer graphics system is an effective alternative [8]. This study assesses the ability of an automated, digital colour contrast sensitivity program in investigating diabetic maculopathy.MethodsPatients from either the Diabetic Eye Screening Service or patients returning for their follow-up appointment in the Medical Retina Service were recruited for this study. Inclusion criteria included Type 2 diabetic patients with untreated non-proliferative diabetic retinopathy (NPDR) and untreated clinically significant macular oedema (CSMO). Exclusion criteria included Type 1 diabetes, proliferative diabetic retinopathy, previous laser photocoagulation, and concurrent ocular pathology including infection, trauma, amblyopia, glaucoma, and/or vascular occlusion.Medical history including duration of diabetes, hypertension, renal disease, recent HbA1c, and smoking were recorded. Concurrent eye disease and previous treatment were also recorded. Examination of best corrected logMar visual acuities (BCVA) was followed by colour contrast sensitivity testing of each eye by occluding the fellow eye and using the diabetic module of ChromaTest, a software program analyzing the age-corrected tritan (TCCT) and protan color contrast thresholds (PCCT). A brief explanation of what the patient is expected to see and their expected response was made prior to the test. The right eye was tested first followed by the left.For the Chromatest, the subject is seated at a fixed distance from the monitor so the alphabetical letter displayed on the computer screen subtends a constant angle on the retina. The letter size creates an image that tests the central 6.5 degrees of the retina. The letters are displayed on a background of equiluminance. The operator has no influence on the contrast of the test letter given. The computer finds the endpoint of the test by a Modified Binary Search method; if response is correct, on the next presentation the colour difference between letter and background is halved. If response is incorrect, the colour -contrast is doubled. Incorrect responses prolong the test, but do not influence the final threshold. This method of determining thresholds leads to finite steps which reach a plateau at the colour contrast sensitivity threshold. The reproducibility of this measurement is 1%, which is the sensitivity of the test. The Chromatest has been further described in various articles [8-10]. Control data was obtained from unpublished data collected by G.B. Arden from diabetic patients without any diabetic retinopathy prior to this study (Table 1). Test and training sets are both from the group studied in this report.Table 1Colour Contrast Sensitivity in Patients with Diabetes and No Clinical Retinopathy (N = 30)AgeTritanProtan3712.42.5449.43.1484.24.2484.12.9484.22.45111.35.9514.22.5515.94.7546.96.6544.14.8547.93.7576.82.5598.62.5599.42.46015.72.6606.25.46115.711.6627.12.7628.611.4647.93.7679.45.16713.65.46817.35.46811.75.71696.86.86913.94.77017.34.77012.45716.73.87221.75.4Control: Age, TCCT, PCCTDilated fundoscopy with slit lamp biomicroscopy and 78 D lens was performed by a specialist registrar (RW) to confirm the grading of CSMO according to the Early Treatment Diabetic Retinopathy Study extension of the modified Airlie House classification [11]. CSMO is defined as any retinal thickening within 500 microns of the centre of the fovea; hard, yellow exudates within 500 microns of the centre of the fovea with adjacent retinal thickening; or at least 1 disc area of retinal thickening, any part of which is within 1 disc are of the centre of the fovea.Each age group (eg. 30–49 years old, 50–69, 70–89) separated by 2 decades was assigned pass-fail criterion for TCCT as previous data suggests age related change in threshold for tritan colour. Since this is the first study of NPDR using the Chromatest, threshold levels were derived using the same data set for both training and testing the effectiveness. Pass-fail criterion for each age group was chosen piecewise and sensitivity/specificity calculations were made according to these arbitrarily assigned levels.Sensitivity, specificity, confidence intervals, and χ2 test were calculated by web-based statistical calculator made available by Professor Lowry at Vassar College, New York . Wilcoxon Rank Sum Test for non-parametric statistical analysis was performed using web software .Results150 eyes of 150 patients were included in this study. Of the 150 eyes, 115 eyes had untreated NPDR (Table 2) and 35 eyes had untreated CSMO (Table 3). Median age was 60 years. Median duration of diabetes was 16.0 years.Table 2Colour Contrast Sensitivity in Patients with NPDR (N = 115)AgeLog Mar VATritanProtan31013.63.43205.23.23206.72320.215.43.241016.115.44106.12.14106.22.141061.74108.43.942011.43440.29.64.8440.213.38.1450.216.14.2450.222.15.5450.419.95.84805.62.9480.520.63.8480.629.554907.43.44906.32.24908.43.94908.42.64909.43.14909.93.449010.32.949030.56.149034.54490.133.66490.79.22.6490.712.23.651013.64.4510.1185.8510.219.1752010.82.6520.282.49.354093.154022.14.6540.223.64.355014.43.155020.25.4550.218.43.5550.217.62.1550.319.64.4550.385.97.7550.422.17.75608.12.756011.12.5560.16.62.657010.33.6570.16.72.9570.17.22.1570.214.92.9580.113.93.8580.2113.3580.221.42.8580.2383.8590.26.82.1590.26.31.4590.210.12.7600.283.1600.212.24.46105.72.76107.52.5610.28.62.7610.213.42.862010.42.8620.398.778.2620.398.775.76309.94630.115.45630.125.36.564018.53.7640.220.24640.275.721.4650.315.46.3650.337.919.967018.37.767020.66.7670.119.94.6670.157.73.8670.28.12.5670.3206.5670.350.42.9670.552.48.4670.618.16.7680.132.76680.210.62.7680.231.53.969014.44.4690.149.66.2690.519.95.27109.213.371011.13.8710.17.213.7710.29.62.5720.221.55.7720.45.52.6720.460.36.1720.534.86.4720.618.63.375012.92.2750.119.94750.340.43.6760.327.64.4760.370.59.6770.111.93.6780245.2780.217.64780.220.97.1780.322.412.9790.552.621.7790.598.767.682013.55.2820.223.66.8NPDR patients: Age, VA, TCCT, PCCTTable 3Colour Contrast Sensitivity in Patients with CSMO (N = 35)AgeLogMar VATritanProtan3108.53.631011.14420.214.14.544071.944018.82.6510.28.82.652029.63.5520.372.310.7550.218.43.5560.318.42.9560.5365.6580.17.72.7580.378.213.7590.223.6362070.57.7620.149.911.4630.427.36.7650.185.914.4650.398.716.9670.116.13.2670.211.83670.380.812.4680.213.33.2690.123.35.3690.530.316.170021.56.870035.45.670032.75.570062.89700.598.720.871098.714.7710.264.820710.398.742.3720.76818.4720.957.716.9CSMO patients: Age, VA, TCCT, PCCTMedian LogMar BCVA for NPDR patients was 0.20 and for CSMO patients was 0.20. Interquartile range for VA NPDR and CSMO was 0.20 and 0.30, respectively. Median PCCT for NPDR was 3.9% and for CSMO patients was 5.6%. Wilcoxon Rank Sum Test analysis revealed statistical significant difference between CSMO and NPDR eyes for PCCT (p = 0.01). When compared to controls with sample size N = 30 (Table 1), PCCT for NPDR had no statistical significance (p = 0.15) whereas PCCT for CSMO was significant (p = 0.002). Median TCCT for NPDR was 15.4% and for CSME patients was 29.6%. Statistical significance was found between CSMO and NPDR eyes for TCCT (p = 0.0002). Both were also statistically significant when compared to controls (p < 0.001)The piecewise pass/fail criterion for TCCT for each age group was as follows: 11.0 (30–49 year old); 23.0 (50–69 year old); 32.0 (70–89 year old). Sensitivity and specificity for screening of CSMO using the above pass-fail criterion for age matched TCCT results achieved 71% (95% confidence interval: 53–85%) and 70% (95% confidence interval: 60–78%), respectively (Table 4).Table 4χ2 test for TCCT detection of CSMOTrue PositiveTrue NegativeTotalTest Positive253560Test Negative108090Total35115150Sensitivity = 71% (CI: 53–85%), Specificity: 70% (CI: 60–78%); χ2 test: p < 0.0001 comparing proportions of true positives among the test positive versus test negative subjectsWhen repeating the analysis in Table 4 for only subjects with logMar BCVA > = 0.1, sensitivity to detect CSMO improves to 75% (CI: 47–91%) and specificity to 85% (CI: 67–89%) p = 0.0002. Similarly, when repeating the analysis in Table 4 for only subjects with CSMO with central macular thickening, sensitivity to detect CSMO improves to 83.3% (CI: 58–96%) p < 0.0001.DiscussionCost effective screening for chronic and debilitating disorders such as diabetic retinopathy is not only important to the well being of the patient, but these healthy adults contribute to the economy of a nation. With the rise in type 2 diabetes in obese adolescents due to dietary and lifestyle changes, the need for an optimal method of screening for sight threatening diabetic retinopathy becomes a critical essential [12].Abnormal protan and especially tritan colour vision is associated with diabetic retinopathy [13]. Blue-yellow defect has also been described in both diabetic retinopathy and glaucoma [14]. In contrast to the optotype used for testing macular function, the Chromatest has a separate glaucoma module for which it is designed to measure peripheral colour sensitivity changes in an arcuate manner using a central fixation point. This study did not cross examine patients with glaucoma and diabetic retinopathy using both glaucoma and macular modules, but it is feasible that further testing may reveal an overlap in colour defect for these patients. Although the mechanism of altered colour vision is unknown, there is evidence that reduced retinal oxygen saturation is associated with impaired colour vision in diabetics [15]. Error scores in colour vision have been found to be directly correlated to severity of macular oedema [16]. This may be similar to the effects of retinal detachment where photoreceptors are shifted obliquely [16]. Correlation between selective loss of short wavelength pathway sensitivity and the severity of diabetic macular oedema has been demonstrated [17,18]. Therefore, we have concentrated on the study of untreated CSMO to ascertain the viability of such a screening method. The use of smaller letters (1.5 degree; Chromatest module for age related macular degeneration) might give better results for CSMO as it may test macular function better than the larger 6.5 degree optotype.This study included only patients with type 2 diabetes to reduce the possible variability in pathogenesis. Although the mechanism of diabetic retinopathy is likely to be identical in both type 1 and type 2 diabetes, previous studies such as the Early Treatment Diabetic Retinopathy Study and Diabetic Retinopathy Study have investigated each type of diabetes separately. Laser photocoagulation was an exclusion criterion as it affects tritan colour vision [19]. Cataract and pseudophakia were not excluded as both are more common in diabetics and exclusion would have limited the usefulness of the Chromatest in screening. It is understood that lens-yellowing effects due to cataract may cause pre-retinal absorption of short-wavelength light resulting in tritan deficits. This may have influenced the overall sensitivity and specificity of the study, but it was a representation of the realistic setting clinicians experience in their practice.In colour contrast testing, the higher the TCCT or PCCT score, the more abnormal the result compared to age-matched normal levels. 30% (35 of 115) patients with NPDR had TCCT above normal levels. 12 male patients were suspected to have congenital colour blindness as their PCCT were considerably worse than normal and not corresponding to their visual acuity or their fundus appearance. This was not confirmed with any other mode of investigation as the study was aimed at mimicking realistic clinical setting where high volume testing can be conducted without further time consuming tests. 16 cases had severe NPDR and may have contributed to the poor results whereas the remaining 7 had results not corresponding to their fundus appearance. We postulate that these 7 eyes may have had concurrent disease indistinguishable by indirect biomicroscopy such as more advanced ischaemia. Ultimately, fluorescein angiography may have further elucidated the true pathology.29% (10 of 35) CSMO patients had TCCT better than normal levels. 8 eyes had CSMO qualified as 1 disc area of retinal thickening within 1 disc area of the fovea. 2 eyes had exudates with associated retinal thickening within 500 microns of the fovea, but both were left eyes and it is possible that the patients were able to perform educated guesses because they had been conditioned following testing with their right eye.Unfortunately, we were forced to obtain normal threshold levels through the same dataset. These levels were obtained through analysis of cases without CSMO. Therefore, the results may be biased. However, because this device is relatively new and the limited availability of further data from diabetics, we are limited to using this dataset to obtain \"normal\" threshold values. Further data will strengthen our case of the power of this diagnostic tool.The Chromatest is unable to successfully screen those patients with congenital blindness and performs less well for patients without foveal pathology. Conditioning following testing with the right eye may also allow patients to perform better on their left eye. From anecdotal evidence, time for testing of the second eye was observed by the investigators to be shorter than the first eye. Repeated testing which was not done in our study may alleviate this problem. This study has studied more untreated CSMO eyes with colour vision than any other that have been published, but it requires more data to solidify our findings. Colour contrast analysis may become a useful tool for defining the need for laser treatment, but so far our experience fails the Exeter Standards of the British Diabetic Association (Diabetes UK), which established screening levels of at least 80% sensitivity and 95% specificity [20].Despite the limitations of the results, there was no discrimination for age and visual acuity due to the ease of the test. All patients were able to perform this test unlike the 1.5% of patients failing to perform another automated TCCT test [21]. Average test time was fast at 5 minutes and requires no mydriasis unlike fluorescein angiography and fundus photography. Conditioning after repeated testing is an issue for reliability, but this study was aimed at mimicking realistic clinical settings where patients have no experience of colour contrast testing. Further studies to distinguish repeatability and data for classifying normal results from abnormals are planned. The equipment required is relatively cheap and readily available compared to those required for optical coherence tomography or stereomacular photographs. It is also a non-invasive procedure and less labour intensive compared to fluorescein angiography.ConclusionNon-ophthalmic doctors can have a retinopathy detection rate of 49% compared to 96% for ophthalmologists [22]. Therefore, a cost effective method for screening is essential for diabetic retinopathy. Screening by digital photography proposed under the National Service Framework is offered to all patients with diabetes in the United Kingdom. It is supplemented by biomicroscopy by the ophthalmologists in monitoring and treating sight threatening disease. Furthermore, optical coherence tomography has become a powerful tool in screening and monitoring CSMO with sensitivity and specificity rates of near 80% and 90%, respectively [23]. Perhaps with further investigation, TCCT testing may become a supplement for detecting and monitoring sight threatening pathology without much equipment or trained technicians. However, with current data, all forms of TCCT testing including the Chromatest do not qualify for use in screening for CSMO.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsRW examined patients, conducted investigation, conceived, drafted the manuscript. TA performed the statistical analysis. JK compiled patient list and conducted investigation. SS compiled patient list and conducted investigation. GA performed the statistical analysis. VC conceived of the study, and participated in its design and coordination. All authors read and approved the final manuscript.Pre-publication historyThe pre-publication history for this paper can be accessed here:\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2531086.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2531086",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2531086\nAUTHORS: Shona Fielding, Peter M Fayers, Alison McDonald, Gladys McPherson, Marion K Campbell\n\nABSTRACT:\nObjectiveQoL data were routinely collected in a randomised controlled trial (RCT), which employed a reminder system, retrieving about 50% of data originally missing. The objective was to use this unique feature to evaluate possible missingness mechanisms and to assess the accuracy of simple imputation methods.MethodsThose patients responding after reminder were regarded as providing missing responses. A hypothesis test and a logistic regression approach were used to evaluate the missingness mechanism. Simple imputation procedures were carried out on these missing scores and the results compared to the actual observed scores.ResultsThe hypothesis test and logistic regression approaches suggested the reminder data were missing not at random (MNAR). Reminder-response data showed that simple imputation procedures utilising information collected close to the point of imputation (last value carried forward, next value carried backward and last-and-next), were the best methods in this setting. However, although these methods were the best of the simple imputation procedures considered, they were not sufficiently accurate to be confident of obtaining unbiased results under imputation.ConclusionThe use of the reminder data enabled the conclusion of possible MNAR data. Evaluating this mechanism was important in determining if imputation was useful. Simple imputation was shown to be inadequate if MNAR are likely and alternative strategies should be considered.\n\nBODY:\nBackgroundMissing data are a common occurrence in any area of research, and are especially problematic in quality of life (QoL) studies. Data may be missing for a variety of reasons. If these reasons relate to the QoL of the patient, the missingness is informative. Simply excluding those with missing data from the analysis (\"complete case analysis\"), will bias the results if those who did not respond had significantly lower (or higher) QoL scores than those who did respond.Rubin [1] defines three main mechanisms of missing data: missing completely at random (MCAR), missing at random (MAR) and missing not at random (MNAR). MCAR requires very strong assumptions. An observation is said to be MCAR if the missingness is independent of all observed and unobserved (i.e. previous, current and future) QoL assessments [2]. For example a patient may simply forget to post the questionnaire back. Observations can also be MCAR if the missingness only depends on values of fixed covariates that are measured prior to treatment assignment – often termed covariate-dependent dropout. For example, if elderly patients were less likely to respond, missingness would be dependent on age group.A more relaxed assumption about the missing data mechanism is missing at random (MAR), where missingness is independent of all unobserved (missing or future) QoL values, although it may be dependent on the observed values. The \"observed values\" may comprise a baseline measure of QoL or a previous assessment and any appropriate covariates.A process that is neither MCAR nor MAR is called missing not at random (MNAR). MNAR occurs if missingness depends not only on the observed data but also on the unobserved (missing) values. An example is that a person with reduced QoL due to side effects of treatment may be less likely to return the questionnaire. The missing value depends on the unobserved QoL scores and the missingness mechanism is informative.Many investigators have explored approaches to determine the mechanism of missingness. They have either generated artificial datasets using simulation techniques [3], or have made use of existing datasets in which missing data were then artificially created [4]. These procedures are potentially misleading: the missing patterns are predetermined and pre-specified, and usually the performance of the various tests can be anticipated through the known mechanism that was used to generate the samples.One approach to deal with missing data is simple imputation, which is the process whereby a single estimated value for the missing observation is obtained, thereby enabling standard statistical methods to be applied to the augmented data set. Various methods can be implemented to impute the missing data. However, the accuracy of imputation cannot normally be determined, as the true values are not known. Various authors have explored the potential accuracy of imputation methods by artificially removing data from a dataset and treating it as missing [3-5]. This is a circular argument, as noted above, because the data are either removed at random or according to some known and pre-specified pattern. In practice, the major analytical problem is that one does not know the exact missing mechanism.Engels and Diehr [6] noted the need to use data with real missing patterns, and attempted to overcome these problems by using a dataset where a value was observed after one or more missing values had occurred; the observed value was treated as the true value for the missing data at the preceding time points. Various imputation methods were applied for the missing values, and the results compared against the observed value to assess accuracy of the imputation methods. As Engels and Diehr [6] comment, \"this analysis hinges on the similarity of a known value following a string of missing values to other observations that are missing at that same time.\"Poor compliance with collecting QoL data is a well-recognised problem in clinical trials. In an attempt to minimise the level of missing data, the Health Services Research Unit (HSRU) at the University of Aberdeen makes strenuous efforts to recover QoL data. When QoL questionnaires are not returned, HSRU not only issues repeated reminders (including telephone contact), but in addition offers to interview patients by telephone. Therefore, a proportion of patients who initially had missing data – as would have been the case in most clinical trials – then have \"true\" values which were subsequently recovered. This provided a unique opportunity to investigate the performance of tests for identifying missing data mechanisms and methods of imputation, because the results could be evaluated against the data that was later recovered.MethodsThe datasetThe RECORD trial was a randomised placebo-controlled trial of daily oral vitamin D and calcium in the secondary prevention of osteoporosis-related fractures in older people [7]. Patients' QoL was assessed by postal questionnaire at 4, 12, 24, 36 and 48 months. The four month data were considered the \"baseline\" measure as QoL for many patients at entry to the trial would be artificially low while they were being treated in hospital for their primary fracture. The questionnaire included the five items of the EuroQoL EQ5D [8], and the 12-item SF12 questionnaire [9]. The EQ5D produces a single QoL score, and the SF12 gives two summary scores, the physical and mental component scores (PCS and MCS). The results for EQ5D data are presented here. At each occasion, if a participant did not return the questionnaire within two weeks, up to two reminders were issued (two weeks apart). Patients who returned the questionnaire without needing a reminder were considered 'immediate-responders', while those that returned a questionnaire after one or two reminders provided 'missing yet known' data, and were termed 'reminder-responders'. In the analyses that follow, the scores obtained for reminder-responders were regarded as missing – what they would have been in some clinical studies.Identifying the missing data mechanismHypothesis testsThe pattern of missing data can be described as either \"terminal\", when no further observations were made on a patient after a set of complete observations, or \"intermittent\", in which case one or more observations for a patient were missing before a subsequent observation was observed. It was possible for a patient to have a mixed pattern, with a period of intermittent dropout followed by terminal dropout.There are a number of hypothesis tests that can be carried out to test the assumption of MCAR. Little [10] developed a test based on the means of the variable of interest under the different missing data patterns (including intermittent and terminal missingness). Alternative hypothesis tests have been suggested by Diggle [11], Ridout [12] and Listing and Schlittgen [13], all requiring terminal missingness. Diggle [11] used an approach which tests whether the subset about to dropout are a random sample of the whole population. Ridout [12] adopted a similar approach to Diggle by utilising logistic regression. Listing and Schlittgen [13] proposed a test based on means. These alternatives to Little [10], will be less optimal in a situation where intermittent missingness is evident. Restricting the analysis to only those showing a terminal missingness pattern would cause a loss of information. Since RECORD contained intermittent missingness, Little's test was used to illustrate a hypothesis test for MCAR.Little's test of MCAR versus MAR [10] is based on the rationale that if the data are MCAR then at each time point the calculated means of the observed data should be the same irrespective of the pattern of missingness. For example, it should not matter whether the previous assessment was observed or not, nor whether the one before that was observed. If the data are not MCAR, the mean scores will vary across the patterns. Consider a study with J measurements of QoL. Let P be the number of distinct missing data patterns (Ri) where J{p} is the number of observed variables. n{p} is the number of cases with the pth pattern and ∑n{p} = N. Let M{p} be a J{p} x J matrix of indicators of the observed variables in pattern P. The matrix has one row for each measure present consisting of (J-1) zero's and one 1 identifying the observed measure.Y¯{p} is the J{p} x1 vector of means of the observed variables for pattern p, μˆ is the maximum likelihood (ML) estimate of the mean of Yi and ∑ˆ is the maximum likelihood estimate of the covariance of Yi. The ML estimates assume the missing data mechanism is ignorable. μˆ{p}=M{p}μˆ is the J{p} x1 vector of ML estimates corresponding to the pth pattern and Σ˜=NN−1M{p}ΣˆM{p}' is the corresponding J{p} x J{p} covariance, matrix with a correction for degrees of freedom. Little's proposed test statistic when Σ is unknown, takes the formX2=∑p=1Pn{p}(Y¯{p}−μˆ{p})'Σ˜{p}−1(Y¯{p}−μˆ{p}).This test statistic is asymptotically chi-squared with (Σ J{p} - J) degrees of freedom.Logistic regressionFairclough [14] described an approach to determine the missing data mechanism using logistic regression. The process investigates the missingness mechanism from a cross-sectional standpoint, each time point assessed in turn. Those people who did not respond were excluded from these analyses. An indicator variable was created to identify those patients who responded without the need for a reminder (immediate-responders) and those which were reminder-responders. The first step identified covariates that predict the occurrence of missing observations (reminder-response). Differences between the two groups with respect to a number of covariates were explored with t-tests and chi-squared tests. Logistic regression analyses were used to model the probability of missing an assessment. Identified covariates were forced into the model and the observed QoL scores tested as to whether they also contributed to the prediction of missingness [14], as indicated by a reduction in deviance (change in -2*log likelihood). The statistical significance of this reduction in deviance was assessed by comparing it to an appropriate chi-squared distribution (χ21).The advantage of this approach in our setting was the incorporation of the reminder data. A subset of data containing only responders was utilised. The data obtained by reminder was regarded as missing. Initially the process outlined above was carried out assessing whether the covariates and observed QoL were significant predictors of missingness (reminder response). Since the current QoL scores were known, the significance of these to predict missingness (reminder-response) could be assessed. If these scores were found to be statistically significant, the process suggests that data were potentially MNAR.Simple imputationMethods of imputationSimple imputation methods use information from other people (cross-sectional), or information pertaining to the person whose QoL data were missing (longitudinal) [15]. Longitudinal methods include last value carried forwards (LVCF), next value carried backwards (NVCB), last-and-next (LaN – average of last value and next value), average available (Avg), average of previous (prev) and average of future (post). Regression can also be carried out utilising other observed QoL scores (regP) or suitable covariates (RegC) or both together (regP2). Some of these methods cannot be utilised at every time point, e.g. LaN cannot be used to impute the 48 month scores since there is no 'next' value. Cross sectional methods include mean imputation, regression and hot-decking (random selection from those observed). A disadvantage of regression methods is that people with the same covariate set will have an identical imputed value. This can lead to the variance of the imputed data being artificially small, producing inappropriate standard errors, leading to inflated test statistics and falsely narrow confidence intervals and inappropriate p-values in any subsequent analysis [14,15].A newer method not considered here is that of multiple imputation [14]. This procedure imputes a number of values for the missing data incorporating both the variability of the QoL measure and the uncertainty surrounding the missing observation. Each dataset is then analysed and the results combined. The focus of this paper however, is the adequacy of simple imputation.Assessing accuracy of methodsThe reminder-responses were regarded as missing and imputed using the methods explained above. The accuracy of these methods was then assessed by comparing imputed scores to the actual observed scores (of the reminder-responders), using a bias measure and proportionate variance (PV):Bias=∑(y−yˆ)/mPV=var⁡(yˆ)/var⁡(y)Where yˆ is the imputed value, y is the actual value and m is the number of missing values. A positive Bias indicated that on average the imputed value underestimated the true QoL value. The PV is the ratio of the observed variance to the true variance and assesses the under-dispersion for each method. A PV of one indicates that the variance of the imputed values is equal to that of the true values. A PV of less than one implies underestimation of the true variance. The bias and PV were calculated for each patient and then an average was taken across all patients. To produce confidence intervals (CIs) for each of the accuracy estimates, the bootstrapping technique [16] was used within the statistical package STATA.ResultsDescription of datasetThe RECORD trial recruited 5,292 patients, with characteristics shown in Table 1. The majority were female (85%), and most lived in their own home prior to (88%) and after (86%) the index fracture. The recruiting fracture was less than 90 days before recruitment for 82%, and 94% could walk outdoors unaccompanied. Recruiting fractures were in the arm (62%) or leg/hip (38%). Patients aged over 70 were eligible and 13% of those recruited were 85 and over. At four months, the proportion of deaths was larger in the older age group (85+).Table 1Patient characteristic of study population (N = 5292)Percentage with score available at 4 mPercentage without score available at 4 mAll Patients Number (%)No reminderAfter reminderNot returnedAbsent or withdrawnDeadAge group70–741917 (36)403729291275–791665 (32)333131301880–841030 (19)171923243385+680 (13)1013171736SexMale811 (15)1613151331Female4480 (85)8487848769Type of recruiting fractureProximal femur904 (17)1617201847Other leg and pelvic1130 (21)2220212017Distal arm1846 (35)3635313620Other arm1403 (27)2628282517Other9 (<1)00000Locomotor ability (Walk unaccompanied)Yes4979 (94)9593939385No300 (6)577715Time since recruiting fracture≤ 90 days4331 (82)8184828673> 90 days961 (18)1916181427Residence type prior to recruiting fractureOwn home4628 (88)8986848778Sheltered housing538 (10)912131111Other126 (2)223211Residence type after recruiting fractureOwn home4555 (86)8885828574Sheltered housing531 (10)911121011Other206 (4)346515Marital statusSingle348 (7)76767Married2069 (40)4236323925Divorced222 (4)45522Widow(er)2634 (50)4752565265Table 2 shows the number of EQ5D assessments at each time point. The number of questionnaires sent at each assessment reduces for two reasons. Firstly, not all patients were followed up after two years. Only those which were recruited early on in the trial were followed up for longer. These patients continued to be followed up until those recruited later had reached the two year assessment. Once all recruited patients were followed up for at two years, follow up stopped and no further data were collected. At 36 months, only 3,663 patients were followed up and this reduced further to 1,629 patients at 48 months. Secondly, some patients withdrew from the trial or died. The proportion of those sent questionnaires that provided valid QoL scores with or without reminder varied from 79% at 4 months to 86% at 48 months. Of those completing forms, 20% to 26% were reminder-responders. Overall, more than half of the data initially missing were recovered by the reminder system.Table 2Number (%) of EQ5D scores at each follow up pointMonth of assessment412243648EQ5D score (no reminder)2908 (59)2648 (62)2511 (67)1670 (69)661 (69)EQ5D score (after reminder)999 (20)840 (20)693 (18)406 (17)162 (17)Not returned1042 (21)763 (18)561 (15)338 (14)138 (14)Total sent4949 (100)4251 (100)3765 (100)2414 (100)961 (100)Total available for follow up52925292529236631629Not Sent343 (6)1041 (20)1527 (29)1249 (34)668 (41)Proportion of responders who did so by reminder26%24%22%20%20%Identifying the missing data mechanismHypothesis tests of MCARConsidering data from the first three time points, Little's test statistic was X2 = 133.75 (9 df) with p < 0.001. The data were restricted to those patients who responded at each of the first three time points (N = 2606) and data collected by reminder was set to missing. In this situation Little's test statistic was X2 = 39.6 (9 df) with p < 0.001. Therefore, there was evidence against MCAR, suggesting that QoL impacted on whether or not a patient responded with or without the need for reminder.Logistic regressionThis section deals with responders only and the reminder-responders were regarded as missing. Using logistic regression at 12 months the covariates found to be significant predictors of missingness were gender, locomotor ability, residence type prior to fracture and marital status; at 24 months -gender, age group, locomotor ability and type of recruiting fracture; while at 36 months – age group and marital status; finally at 48 months – locomotor ability and time since recruiting fracture.The change in deviance was used to determine whether the previous QoL score was a significant predictor having adjusted for covariates (Table 3). Previous QoL was defined as the most recent known QoL score prior to the time point of interest. The change in deviance was significant at 12 and 24 months. This indicated that, after adjusting for covariates, previous QoL remained important in modelling the probability of missing assessment. The null hypothesis of MCAR was rejected at 12 and 24 months. At 36 and 48 months there was insufficient evidence to reject the possibility that missingness was MCAR.Table 3Log-likelihood's for models 1–4Month of assessment12243648Log-Likelihood L1: MCAR fixed covariates-1680.3-1411.6-846.3-350.8 L2: MAR fixed covariates + previous QoL-1673.9-1409.5-845.7-350.8 L3: MNAR fixed covariates + current QoL-1669.4-1406.1-843.2-350.7 L4: MAR fixed covariates + previous QoL + current QoL-1669-1406.1-843-350.7Change in log-likelihood -2*(L1 – L2) – test of MAR12.8*4.2*1.20 -2*(L1 – L3) – test of MNAR21.8*11.0*6.2*0.2 -2*(L2 – L4) – test of MNAR9.8*6.8*5.4*0.2* significant change, p < 0.05In normal circumstances the investigation would stop at this point, because in most trials the true current score, xc, is not available for the \"missing\" group. However, using data collected by reminder the process was continued. Table 3 shows the log-likelihoods for model 3 (covariates + current QoL) and model 4 (covariates + previous and current QoL). After adjusting for both covariates and previous QoL, at 12, 24 and 36 months the current QoL was significant in the model, suggesting there was evidence of MNAR data. At 48 months there was no evidence that current or previous QoL were important in the model – but, at this time our sample size was substantially depleted.Another question of interest was whether the non-responders were in any way different to the reminder-responders. A similar process was undertaken as above. The non-responders differed in one or two covariates at each time point but having adjusted for this, their previous score was not a significant predictor. Thus, there was no evidence that the previous QoL experience differed between the non-responders and the reminder-responders at a given assessment. This gave confidence that the reminder-responders were perhaps similar to the non-responders.Imputation of reminder-responder scoresResults for the imputed data were compared with the actual data and the 24 month data are presented in Figures 1 and 2. Figure 1 shows that at 24 months the smallest bias occurred with the post method (b = -0.002), while second smallest was NVCB (b = -0.014). The bias was significantly greater for the regression and cross-sectional approaches. At 4 and 12 months (data not shown), the average and NVCB were the best methods in terms of bias. At 36 months, none of the procedures provided a sufficiently accurate estimate and the bias was greater than -0.04. The number of procedures applicable at 48 months was reduced with the regression based on baseline characteristics showing the smallest bias (b = -0.004).Figure 1Bias results of EQ5D imputation at the 24 month follow up.Figure 2PV results of EQ5D imputation at the 24 month follow up.Figure 2 shows the best PV value for the 24 month data occurred with the hotdecking methods, which was perhaps expected since these methods impute using random selection from the immediate-responders. Hotdecking with stratification was the best of the two (PV = 0.979). The 'after' methods of post and NVCB had slightly lower PV, just under 0.8. The three regression procedures were very poor at preserving the variance. Since the same value is imputed for all missing values using the 'mean' methods, there was no variation in the imputed values, which would have a big impact on any subsequent tests and p-values.At other time points (data not shown), where applicable, NVCB showed reasonable PV. The regression methods were consistently poor at preserving the variance. The hotdeck with stratification procedure was reasonably good at maintaining the variance at all time points (PV ranged from 0.87 to 1.27). By nature of the hotdecking procedure it is expected that the variance of the imputed values would be the same as that of the true values. Although the observed PV was not equal to one, the 95% CI did include the desired value of one, suggesting that the sample being imputed was similar to that from which values are being selected.In general, for the RECORD trial methods involving QoL scores surrounding (and in particular those after) the point of imputation were the most accurate in terms of bias and at preserving the variance.DiscussionIdentification of the correct mode of missingness and most appropriate method of imputation can make a large impact on the analysis of clinical trials. The sensitivity of different analyses depends on the proportion of missing assessments and the strength of the underlying causes for missing data [17]. The undesirable effect of missingness on bias and power increases with the severity of non-randomness as well as the proportion of missingness [18].Little's test [10] for MCAR showed evidence against MCAR in favour of MAR between responders and non-responders and also between the immediate- and reminder-responders. The logistic regression approach showed on the whole, at each of 12, 24 and 36 months, after adjusting for the required covariates, both the previous and current QoL scores were significant predictors of missing assessment (response by reminder). This implied there was evidence of MNAR data at 12, 24 and 36 months. It is possible that the \"reminder-responders\" may differ from the persistent non-responders, but the analyses found no evidence of this in terms of previous QoL scores. This approach using data collected through reminders has provided an indication of MNAR, with the rationale that reminder-responders were more likely to be similar to the non-responders than the immediate-responders.It should be noted that data collected through reminders has been assumed to be equivalent to that collected immediately. However, data collected via reminder are actually reflecting a time two (or four) weeks later than the original assessment time. This may bias the recovered data, but for the purposes of this investigation we assumed it to be comparable to data collected without the need for reminder.The missingness mechanism was identified as potentially MNAR, but was simple imputation adequate? The results suggested that for the RECORD study the missing QoL scores could be imputed using assessments close to the point of imputation. In many QoL studies the assessments are taken at frequent intervals and the correlations between successive measurements may be high. Those imputation methods that focus on within-patient assessments close in time to the missing values are likely to be most effective. The population based methods assume the data are either MAR or MCAR. Since the data in this study were most likely MNAR, it is not surprising that these imputation methods were less accurate.Data that are MNAR may depend on current and future observations, thus methods that utilise this data are intuitively going to be more accurate than those based on previous measures. NVCB and post-average showed the smallest bias. Although, the methods involving previous scores are useful, they can never be entirely accurate in the presence of MNAR. The methods of NVCB and post-average may not be practical as they are dependent on future QoL scores being available, which will only happen when missingness is intermittent. Often, in trials, the final assessment is the main focus and no future data are available to inform the imputation. Only methods using 'before' data are available, and these methods have shown to provide greater bias, suggesting that simple imputation is inadequate in the presence of MNAR data.Limitations of this study are that the data are from a single trial, involving older people, and the studied disease is perhaps not typical of studies involving QoL assessments. However, our results agree with Engels and Diehr [6], despite being from a different disease, different country and for different QoL outcomes. We infer from this that the results may perhaps be generalisable.If imputation procedures are to be employed, researchers need to be confident of their accuracy. One apparent advantage of imputation is that, once missing values have been filled in, standard methods of analysis can be undertaken on this augmented dataset comprising the observed and the imputed values. However, imputed values cannot be regarded as the same as if the full data has been observed. Although some summary statistics such as means and medians may not be distorted, the corresponding standard deviations may be shrunk and this will have consequences for the subsequent calculation of the confidence intervals [15]. This consequence of simple imputation is present whatever the missingness mechanism and provides a major disadvantage against the use of simple imputation procedures, even if one can assume the unlikely scenario of MCAR data.Although the imputation may overestimate the true values in the reminder group, it may still bring the overall scores closer. What matters most is minimising the bias in treatment comparisons. An investigation into the effect of the different methods of imputation on the treatment effects forms the basis of future work.During RECORD the issuing of reminders substantially increased the number of included patients, with corresponding gains in statistical power and the assurance of reducing the bias by avoiding the need for imputation. The reminder system entails extra resources. However, in any study having as much data as possible for analysis is very important and if the use of reminders can generate a significant proportion of extra data then it is a useful procedure. The reminder process is a viable approach not only for use with postal questionnaires, but also in computer based testing and integrated voice response methods. It should be noted that the best way to prevent the problems of missing data is to simply avoid it, by employing good data collection techniques and making an effort to chase up missing information. When the proportion of missing data becomes too large, no statistical technique will provide the solution.ConclusionThe first step in the analysis of incomplete data should involve quantifying the extent of missingness, identifying which individuals have missing data and at which assessments. In usual situations none of the missing QoL data are retrieved, and thus it is not possible to test formally a hypothesis that missingness is MAR as opposed to MNAR. Our study provided an example in which it was possible to carry out a formal test, confirming that data were MNAR and that simple imputation was unsatisfactory in this situation.AbbreviationsCIs: confidence intervals; DF: degrees of freedom; HSRU: Health Service Research Unit; LaN: last-and-next; LVCF: last value carried forwards; MAR: missing at random; MCAR: missing completely at random; MCS: mental component score; MNAR: missing not at random; NVCB: next value carried backwards; PCS: physical component score; PV: proportionate variance; QoL: quality of life; RCT: randomised controlled trial.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsSF analysed and interpreted the data, drafted the manuscript and gave final approval to the submitted manuscript. PMF conceived the idea, assisted in interpretation of the results, commented on drafts and gave final approval to the submitted manuscript. AM and GM were involved in the design and running of the RECORD trial including data collection, commented on drafts and gave final approval to the submitted manuscript. MKC was involved in the design and running of the RECORD trial, commented on drafts and gave final approval to the submitted manuscript.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2531097.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2531097",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2531097\nAUTHORS: Arthur L Koch\n\nABSTRACT:\nThe great advantage of being a sexually transmitted disease is the ability to survive and specialize solely on a host species that is present in low numbers and widely distributed so that contact between infected and uninfected organisms by chance is rare.Pathogens of a sparse, but widely distributed host species, must either: i) have an alternative host; ii) be able to survive in a dormant state; or iii) be non-destructive to their host. For the pathogens of a diploid there is a particularly effective strategy, that of being sexually transmitted. Then the hosts' themselves transfer the pathogen.\n\nBODY:\nSexually Transmitted DiseasesThe advantages of this mode of pathogenic transmission are clear, but the great disadvantage is that evolution has now produced many very sophisticated mechanisms in the host to block pathogens of various origins and specialties from entering and growing effectively in the sexual apparatus of the female mammal.Thus, the biological problems that STD mechanisms face are because of the countermeasures against the pathogens that are due to evolution of the host. However, the mechanisms that have developed that protect the fetus from the host's immune system do help the STP.Problems with the STD approach from its point of viewThere are two major problems with the STD strategy: First, a sexually transmitted disease must be (at least relatively), a 'gentle pathogen' or a 'prudent predator' and remain as such (i.e., it may only evolve towards greater virulence slowly, or not at all). Secondly, such a pathogen must, to a large extent, be able to avoid destroying its mammalian host's offspring before the child grows to a sexual adult stage. Both factors would seemingly be countermanded by the immediate selective advantage of being more virulent and producing more propagules.The Vertebrate's Female Reproductive Tract (FRT)The immune system of the mammalian sexual system has many wonderful properties because of the many problems that the Female Reproductive Tract (FRT) faces. The immune system of the FRT is nonpareil for its diverse abilities to solve these problems. Possible pathogens include a large range of viruses, bacteria, and protozoa. Mostly these are destroyed or eliminated by the FRT. Additionally, when there is an embryo: this propagule is foreign because half its genes are not derived from the mother. The mechanisms of the FRT must not operate against it. Although, it is a foreign object like all the pathogens of the host, this particular entity is to be protected.The Fetus is SpecialConsider a Sexually Transmitted Disease (STD) from the pathogen's point of view: it must successfully pass into the female sexual tract, possibly into her body, and into the next adult generation of hosts. So the STD faces a quite different fate than other kinds of pathogens.Advantages of Vertical TransmissionTo reiterate the virtue from the pathogens' point of view, the STD strategy is favorable because the pathogen is propagated to new hosts without entering the environment. Thus, the infection of a new host is expedited by the old host's behavior in searching for mates, and thus many mechanisms used by other pathogens for transmission are not needed.Complications of the Many Roles in the Female Sexual TractThe FTR is a complex of organs: ovary, fallopian tubes, uterus, cervix, and vagina. They have many physiological aspects that are quite different from each other and each is essential for the propagation of the host species. They, of course, all do act in a way favorable for reproduction, but unfavorable, at various degrees, to pathogens. But the STD's are treated quite differently.Consequence of the Absence of an Immune System in the Fetus and NewbornThe lack of an effective immune system in the fetus (even in the presence of maternal passive immunity) and in the newborn (even given mother's milk with colostrum) implies that the STD pathogens might propagate in utero even though the mucosal immune system of the female sexual tract may be able to reduce the danger of infection and disease to many disease agents to the offspring.Importance of Protecting the Fetus, the Neonate, and the Pre-pubescence ChildIn the long run, with many ins-and-outs, the current situation is favorable for the STD pathogens because the long term interest of the host species is that the young members of the host population grow to become sexually mature and this provides hosts to become infected and propagate the pathogen.Stone age lifeWhen populations of humans arose and expanded and emigrated in the early Stone Age, the problem that pathogens would have of how to propagate and surviving in this sparse host population became critical. Cave-dwelling primitive humans, 30 thousand years ago that lived in small well-separated tribal populations would have had quite different diseases than humans have today. Workable disease possibilities are few but would include, sexual transmitted diseases. However, this required that the pathogens were 'gentle' to their host. Other possibilities for these pathogens are either: acquiring and using an ability to remain dormant for long periods of time or of being able to infect a ubiquitously occurring alternative species. These latter two possibilities will not be examined further here, but they do occur in nature. The option of being 'gentle' to the fetus is the main topic here with respect to STDs, but a further aspect of this will also be presented; this is that an STD of a mammal will face the special problems of preventing the destruction of the fetuses and neonates of its host because these are needed for its own propagation.The Advantage of being an STD in a Sparsely Populated WorldThere is an essential feature needed for a successful infectious disease of social animals that are distributed in dispersed state in individual groups with small numbers. This is to have a way to spread from social group to social group. In sundry diseases [1-4], this is done by forming long-lived spores, or passing through intermediate hosts or vectors, or being carried by various animal and insect vectors over significant distances. Another very effective and quite successful way, however, is to use a sexual mode of transmission, which depends on the host actively interacting with spatially remote populations of hosts. In general, many STD pathogens are not long lived within the environment outside the host's body. In addition, they are usually highly specific so that they do not have an alternate host (except only very rarely); consequently before the invention of the hypodermic needle, these STD diseases usually had only a sexual mode of transmission within the individuals of a species [5-9]. STDs must be extremely common since there are many viruses that are quite mild (Hoeprich et al. [1]) and I assume that there are many more such agents that are so mild that they have not been detected.Depending on the degree of interaction between social groups, this strategy would be unsuccessful if the pathogen was highly lethal and the host populations widely distributed, since they would be eliminated from a local population by destruction of subpopulations and usually not be successful in reaching neighboring communities. This is the way, I presume, that ebola was in rural Africa in the past: it arose from some animal into a human social group, destroying that group, then the viruses were eliminated, ending that particular episode. Consequently, any successful STD disease would need to be as 'gentle' as possible while still retaining its infectiousness, if it is to be effectively transmitted to other hosts during rare encounters between groups when population levels are low and well separated.A distinction needs to be made from a mechanism that differentiates survival of a 'gentle pathogen' from the evolution of the pathogen as the results of group selection. Group selection occurs, for instance, if there is the possibility that some pathogen may contain or acquire a gene that confers some advantage or level of resistance to the pathogen. This gene and its population would then be selected because pathogens with that gene would prosper and become dominant.The long-term persistence of a disease among sparsely distributed social animals would depend on the particulars of the social interaction between groups. Passage of a STD from group to group is aided by the social behavior of the hosts, which may institutionalize transfer. For example, the social groups of many kinds of mammals' centers around a single dominate male or a female of a bonded social group [9-13]. Consequently, other (usually younger) males or females of a variety of species are ejected from their natal small groups. Sometimes these may become the dominant in other groups. However, they may incidentally carry STDs with them.In chimpanzee troops and other primates living in groups there is often the exchange of young females between groups as the females reach sexual maturity. In many cases, these females are forced to leave their original group. Possibly this same strategy was practiced by early human groups. After the female chimpanzees emigrate, they are taken into neighboring colonies, spreading STDs. Not only is this mixing of populations known from direct observation of various non-human primate species, but it also can be deduced by the smaller degree of polymorphism for genetic markers exhibited in the male population of a clan of chimpanzees relative to the females in the same group.This process of exchange may serve the primate species very well by limiting the effects of inbreeding. The explanation of these behaviors by geneticists and sociobiologists is that consanguinity is bad for any species (except obligatorily selfing-organisms, like Mendel's garden peas that do practice incest habitually). Whatever the validity of this explanation and how the custom arose, sexual mixing occurs between social groups as an institutionalized process even in the absence of prostitution, rape, and war. Even if these violent transmission events between host groups may be fairly infrequent they, and the less violent, custom-justified, mixing between social groups are essential to the STD's way of life.Diseases in primitive humansA sexually transmitted disease gains most of the advantages of vertical transmission in not needing to be transmitted through the environment. Moreover and very importantly, it can survive in sparse populations because of its host's sexual proclivities. In these two sentences the essential elements for a parasite (especially a virus) to survive by infecting humans as its only host in the Stone Age have been spelled out. In such circumstances it was necessary for a pathogen to be able to compensate for the low population density and sporadic distribution of its host. During the tail end of the most recent Ice Age, the human hosts survived in small, mostly isolated, groups that figuratively chased mastodons and other big game as a group effort.Those pathogens that depend on humans as a resource faced a much different problem after domestication of plants and animals than during the hunter/gatherer Stone Age culture. Tuberculosis did occur in people in the ancient Egyptian empire, but by then the population was locally quite dense and transmission through the air from person to person in large populations became efficient. The problem of pathogens became still quite different during the Industrial Revolution and in the medically sophisticated world of today. Possibly, at the time of Christ the world population was a thousand-fold larger than during the Ice Age, and probably, the world population of humans has increased more than a thousand-fold since. Today, with the world human population increased, with transportation easier, with higher local population densities, with more rapid migration, and with mixing of humanity taking place at an unprecedented rate the situation is entirely altered. These changes must result in a great increase and alteration of the spectrum of diseases, and particularly of communicable diseases that are transmitted effectively between people that are crowded closely together. For this reason, it can be argued that the major diseases of Stone Age humans were largely STDs (or with those pathogens capable of remaining dormant or propagating in other hosts) and the typical major infectious epidemic diseases experienced in the early Christian era and afterwards were not. The aftermath of hypodermic needles, jet planes, and other modern inventions are that the spectrum of diseases has become much different still.The two keys of the matter are: first, that many of today's STDs have had a long association with primates, including humans, and have had an opportunity to modify their hosts; and in return, the pathogens have been modified by their host's biology. Second, the same group-group interactions, as in the hunter/gatherer cultures described above, apply to other primate populations living in the wild today. So we can assume that STDs specializing in particular species have been around for a very long time in human and other primates populations, and that they should generally be 'well-tuned' to their host species' particular sociality. Furthermore it can be assumed, that only occasionally will such adapted STDs be dangerous or lethal to their specific host species, or at least to a fraction of the individuals of that population. To the degree that HIV has recently entered the human population or more specifically into the population of a some times medically treated, modern, industrialized, jet age men and women, it is now in an unfamiliar milieu and is especially dangerous (Gilbert [2]).The strategy of cave man's diseases and those of modern primatesAll STD's of cave men (I presume) and today's wild primates face the same general problems. But here we will be more specific and consider the problems of the retroviral STDs of monkeys and man [10-14]. Now let us match the retroviruses of primates to the design criteria for a STD pathogen of sparse populations. Point by point these criteria are met by viruses abundant in non-human primate populations in Africa today, such as SIV (Simian Immunodeficiency Virus) that is resident in the African Green Monkey. Of course, SIV may be more virulent when transferred to different monkeys; e.g., to the geographically distant Asian ones, than the species from which it was initially isolated. Of course, SIV would be virulent in animals that happen to have a defective immune system. The general point is that the small deviation from its behavior that has been optimized for a retroviruses' life in its native host now may lead to catastrophic problems in the variant host, whether it is the human, the Asian monkey, or the compromised host.It is assumed that HIV emigrated to humans relatively recently; see [15-21] (and it is thought that 1930 was in time of transmission to man from primate in the Belgian Congo [20]. This virus and its new human hosts have not had a chance to adapt genetically to each other though there are some indications that there have been developments in this direction (Ewald [21]). One of the mismatches between the disease and its human host that most affects disease virulence is that the immune system after HIV infection of the human being deteriorates in 5 to 15 years; this same period would be unimportant to non-human primate populations in the wild because they have shorter mean life-spans and maturation periods and may generally die of other causes before they lose their effective immune system. On this basis, it may be that the changes needed to re-establish the 'gentle' parasite mode in the new human hosts of SIV are minimal: For example, just a shortening of the human life span to that of the turn of the nineteenth century level would do. A disturbing, but realistic, suggestion of a change that would certainly make HIV infection relatively more 'gentle' is a general re-emergence of life-shortening infectious diseases, such as tuberculosis. There are many other diseases that may erupt as the antibiotic era closes, and with the loss of efficacy of antibiotics, the human life span may decrease dramatically and the immune system may survive to the end of the life-span and outlasts the shortened length of life-span of its host as the result of deaths from other causes than just the effects of the AIDS. Then there would be a lower proportion of individuals with signs of ARC or AIDS. Changes of the other kinds will be suggested below that might increase the longevity of the immune system in an AIDS-infected individual in a long-lived population, but other possible changes might decrease it.The mucosal immune systemOf the many kinds of immunological responses, the ones that function at mucosal surfaces are most relevant to STDs during their transmission from individual to individual [22-40] The feature relevant to STDs epidemiology is that the female sexual tract is the right place for ways to prevent new infections.Immunology of the female reproductive tractIn order to be propagated, a sexually transmitted disease needs to grow and be transmitted through both the female and male reproductive tracts to new hosts. On the other hand, particularly the female productive tract has evolved mechanisms to overcome pathogens that enter it. However, an important additional component for the disease is that the STD's somehow needs to ensure that the embryo, fetus, and child of its host do not die, but grow to sexual adults so that they too can serve as habitats for STD to grow. \"Ensure\" is too strong a word as long as the pathogen is only successful in attacking a small portion of the host population.Evidently, there are mechanisms contributed from both the human and the virus to protect the juvenile against destruction by the HIV, beginning at conception and continuing on to near adulthood. This is because otherwise the fetus and neonate would be destroyed before leading to an adult that can receive and propagate the virus.The complex immunological processes within the pregnant female are indeed very sophisticated. This has been well studied because the interest in this question is high since this may be a possible basis for the prevention of the spread of AIDS. It was, and is hoped, that immunological ways could be developed. However, here the interest is to try to understand the biology to the host-parasite interaction from the point of view of the disease process. There must be some blockade to transmission extent now because even though HIV infection through sexual contact of adult-to-adult is efficient, there are some fetuses and children that are not infected or killed by the infection, but grow up to become adults to be reservoirs for growth and propagation of the virus.A strong indicator that HIV is usually prevented from infecting children perinatally is that while the majority of children born to HIV-positive mothers are not infected, they almost all carry maternal antibodies against HIV 41 (McWhinney, Pagano, and Thomas. [40]). Presumably they had them when in utero, but either had not become infected or had overcome it.Going even farther out on a limb, one can imagine that the mucosal immune response could protect the baby during the birthing process [41]. This might be accomplished by a strong IgA or T-cell response, or by an IgG response that is delivered through the placenta [42]. If this were possible in terms of the host's immunological repertoire modulated by some stimulatory action of the pathogen, this would increase the long-term fitness of the STD. Such a situation would lead to continued selection for such genetic variants of the host and/or in the STD. Then of course, through successive generations of virus and host in an environment with high levels of other pathogens, these developments would 'tune' the growth of the particular retrovirus to a particular primate species. These changes may lead to effective prevention of infection due to secondary pathogens or due to superinfection of the resident sexual disease by other copies of itself, and would have a long term (and incidental effect of maintaining the host population). Such selection would lead to corresponding changes so that the pathogen becomes milder to the host as a consequence of induction of greater immunogenicity against the resident pathogen and would be beneficial to the host because the first pathogen would prevent other, possibly more dangerous pathogens, from entering and destroying it.The converse possibility is that the pathogen causes deleterious effects due to stimulation of the immune system. In many cases these effects are significant, but generally not lethal.Human antimicrobial factorsDefensinsThere are a number antimicrobial factors made by the mammal that protect it against pathogens. These include many defensins. Of these, human β-definsin-1 is especially to be considered because of its location in urogenital tissues (Valora et al., [43]).Retrocyclins that are θ-defensins and their variations have important effect against anthrax [44]. Also, some forms of θ-defensin and α-defensin have protective effect against HIV-1 [45,46]. But the latter paper did not find an effect on the mother-to-child transmission. Quinones-Mateu et al. [47] show that human epithelial β-defensins 2 and 3 inhibit HIV-1 replication. Aono et al. [41](2006) find that in bovines the bovine β-defensin-1 acting against E. coli and is present in teat mucosa, vagina, ovary, and oviduct. These references, and also MasCasullo et al. [48], allows the possibility that I have not seen explicitly suggested that these substances in the (female reproductive tract) FRT act to prevent the embryo and fetus from infection by a virus that has infected the mother. Furthermore, that they, except for the α-defensins, have a role in the partial protection of the newborn from infection by way of the mother's milk.Secretory leukocyte protease inhibitorThe secretory leukocyte protease inhibitor (SLPI) may also be a factor that acts in prevention of HIV infection. It is known that salivary gland tissue may have a role in suppressing transmission by the oral route. Wahl et al. [49] were able to account for the rarity of oral transmission even though there is HIV in oral secretions. Once inside the oral cavity, HIV is exposed to antiviral levels of SLPI, and their data suggest that this may greatly impede infection. An extrapolation, for which there is no data, could be that HIV might stimulate the synthesis of SLPI. A further extension from known facts is that it may also takes place in the genital tract and prevents secondary infection of HIV viruses, or variants, or a broader class of STDs [50]. These suggestions could be tested.Variable stimulation of the immune system of primates under different conditionsThe long-term survival of the retrovirus pathogen in vertebrate hosts depends on proper balance of the pathogen's activity versus the hosts' immune systems. A possibly critical factor of difference between modern humans, the Stone Age man, and the modern non-human primate needs to be recognized here. Some of the possible differences would make the response greater or weaker, and consequently could support or conflict with the hypothesis proposed here. This balance could depend on the variety and variable extent of immune stimulation in general and is, in addition, to the responses of these host species to their particular varieties of retroviruses. The stimulation of the immune system, variety of stimulations, and the age dependency of individuals exposed to other antigens before they are challenged by viral antigen do vary (see for example, Esquerré M et al. [50]; Walker et al., [51]; Chirmule et al., [52]; Rabin et al., [53]; Miedema and Klein, [54]; Wolensky et al., [55]). There must be marked differences in the sensitivity to antigenic stimulation of the members of primate societies in the wild, of the humans in developing countries, and of the humans in developed countries. Either too much, too little, or the wrong kind of stimulation may mean that the immune system will be either over or under active, because of the effect on the mucosal part of the system and, this is significant for the theme of this paper. A slight variation may eliminate the pathogen or fail to prevent new pathogens from entering and supplanting the original one. The immune system may be secondarily modified by destruction of certain specialized T-cells in which HIV or SIV pathogen are present. As a speculative example, young female chimpanzees engage in some sexual activity for some time before they can become pregnant. If the female has been infected with SIV for several years before she becomes fecund it would not be surprising if she had slowly developed a range of mucosal immune responses that would have a protective effect on her and/or the developing fetus. This effect might not occur if she had been infected more immediately before pregnancy. Some aspects of the AIDS disease process in humans are certainly similar to autoimmune diseases and the destruction of cells of the immune system bearing viral or similar receptors is well known. All that is being suggested here is that such phenomena need be only slightly altered to give virulent disease instead of an innocuous one in the not quite well adapted AIDS-human pair from those of well-adapted long-term retrovirus-primate pairs. In the current worldwide AIDS epidemic these could affect the timing of symptom onset and the severity of debilitating disease.At an earlier time when our blood supplies were contaminated, hemophiliacs were very likely to receive the AIDS virus and become HIV seropositive. A significant point is that their time to conversion to ARC and to fulminating AIDS was 90% longer than other non-hemophiliac seropositive persons at that time Darby et al. [56] in 1995. One possible reason is that the medically treated hemophiliacs were being continuously challenged and immunized against a large variety of other substances that were present in the various blood transfusions that they regularly received. An additional factor concerning the AIDS disease at the beginning of the world epidemic is that when hemophilia patients were initially infected by transfusion, the disease would have been started with a much larger number of viruses than that transmitted either by usual or unusual sexual practices or by a drug addict's re-used needle. Consequently, their immune system was stimulated in a way that the immune systems of normal people infected by sexual contact are not, simply because the intensity of the immune challenge was greater. These facts raise the possibilities that African green monkeys are immunologically equivalent to human hemophiliacs receiving blood transfusions. As a result, an African green monkey from which SIV can be isolated might naturally have an especially effective immune system and be able to continuously destroy a much larger proportion of retroviruses and thus limit the viremia. This could mean that the animal would live for a long time before it might become immunologically deficient. In any case, it will be interesting to see the disease progression in fresh hemophiliac patients that were infected by sexual transmission and not by transfusion either while they are receiving recombinant clotting factor without these multiple sources of diverse immunogenic stimulation or with them.Immune reactivity can depend on the presence of other pathogens: Infection with mycoplasma, Herpes, Epstein-Barr, and several other viruses may affect how HIV infection leads to an HIV-immune deficiency state. With co-infection of such viruses, it is likely that the time at which AIDS erupts may be sped or slowed in either the human or monkey. This raises the possibility that a different spectrum of viruses and other possible diseases might influence the course of a retroviral disease. A sometimes-symptomless retrovirus that suppresses the development of debilitating immune deficiency might do so differently when infecting a different host species. On the contrary, other pathogens may trigger the emergence of the AIDS virus from the host's chromosomes. These assorted immunological events could greatly affect the life history of a retrovirus in any new primate host. Consequently, it can be argued that the AIDS virus may well have been adapted to be 'gentle' and un-obstructive in its old host, however, at present, those strategies do not work as well in humans because the human pathogens are different than the non-human primate. Because of our social organization, our life expectancy, and our antigenic environment, the outcome of this parasitism might well be quite different.A very striking observation was made fourteen years ago [57,58](Ho et al., 1995; Wei et al., 1995; but see Pang S, [59] and see Wain-Hobson, [60] and Miller, Antia and Levin, [61]. The major conclusion is that the HIV viruses grow very rapidly and are very rapidly destroyed by the apparently healthy, but HIV positive individuals. It leaves unanswered how much of this destruction is due directly to the action of the host immune system versus the action of the host when modified by the virus or responding in other ways.The possibility is that there are protective mechanisms implemented by the resident HIV. A finding that is relevant concerns an individual who became infected with only one strain even though two different varieties of HIV were transfused simultaneously into him [62].Epidemiology of AIDS and the 'rapid' change of HIVBy comparing the epidemiology of HIV-I and HIV-II in their respective parts of Africa, Ewald [21] has made the case that the diseases vary in the virulence/gentleness scale in a way that is correlated with the number of sexual partners and the group's societal norms. This would be predicted by a game theory type calculation on the bald assumption that the virus was omniscient without providing any suggestion of how it got to be so. Accepting his ideas and facts as true, we can only be amazed by the speed with which the virus can apparently change its strategy and optimize the course of infection to the new conditions present in a new host. I think that this means something more significant. As an alternative to de novo evolution, I suggest that this apparent rapid adaptation is consistent with the hypothesis that STD retroviruses have been exposed over long evolutionary times (in terms of millions of years) to fluctuations in the behaviors of a series of hosts (or a host under a series of very different conditions) and have developed and retain a genetic repertoire to be able to achieve response to variations in the host. From the molecular point of view this could allow them to track and respond to the behavior of their current host and conditions. Although the genetic sequences are now known, the roles of the gene products are not fully evident and some of the regulatory genes may have functions over a long time period of years and centuries.Why mammalian STD pathogens must protect their new generation of pathogens from their own lethal actionA careful strategy must be maintained for long term persistence between the STD virus spreading from adult-to-adult and in injuring or killing the children of an infected host. The biology of pediatric AIDS is reviewed in Pizzo and Wilfert [63], Remington [64], Roizman [3], Sweet and Gibbs [65], and Kaschula [66]. Teleologically, the virus must spare the host's young in order that they can be used as a resource in the future, but that necessary-truth provides no mechanistic way to implement such a situation. HTLV-I (see below) seems to have achieved this balance by infecting most offspring of infected mothers but having the virus grows so slowly that a child's life before puberty is virtually unaffected. Such slow viral growth permits propagation of the virus to virus-free individuals within a society in which sexual contact is frequent and avoids the usual destructive effects of vertical transmission from mother to child. The AIDS viruses, HIV-I and HIV-II, do not seem able to avoid injuring the children after infection occurs. But still a large proportion of the young of infected mothers do not become infected – more than 60% and in some estimates as much as 85% for HIV-1. Some infants, born infected, clear the HIV virus (Clerici et al. [67]; Roques et al. [68]).An infectious disease spreading to the next generation by vertical transmission has advantages over pathogens disseminated in other ways. The main advantage is that the disease organism never has to survive in the environment outside of the host. Survival of a pathogen outside of its host in air or water takes special mechanisms. Although the fetus is a convenient and a built-in susceptible host, the vertical process in vertebrates has a special complexity beyond that present in the cases when the host grows by binary fission as bacteria do. This is because the animal host gives birth to an immunologically ill-equipped neonate that may not be able to survive due to damage caused by the pathogen. This is a critical problem for an STD's survival strategy because they generally have no alternative propagation strategy such as survival in alternative hosts or persisting in the environment.Because the newborn offspring do not have a full immunity system the result is often catastrophic. This is what happens when some pathogens, such as Herpes Type II, Rubella, or Cytomegalovirus infect an embryo, a fetus, or a newborn child. In these cases, without a fully protective immune mechanism, the infection may be destructive or lethal to the offspring and is much more severe than if the pathogen infects an adult that has a responsive immune system with an immune repertoire already in place.Therefore, in many cases a would-be pathogen cannot effectively persist via pure vertical transmission simply because the pathogen cannot depends on its host immune response system to bridle its growth. Viruses like Herpes, Rubella, and Cytomegalovirus that go across uterine, vaginal, and placental tissue are frequently dangerous to the fetus, but of course these viruses survive because their primary means of spreading through the population is from an adult individual to an adult individual and are not dependent on vertical transmission. Destroying a few children apparently does not upset the growth success of these viruses. However, from the viewpoint of a disease that has elected to only use a non-virulent STD strategy, going directly from mother to baby could be expedient, although destructive. This must somehow be avoided, and it apparently is often curbed by successful pathogens.There may be a number of factors involved in the AIDS case. One is that the AIDS virus is just not very infectious or long-lived. Some suggestion of this comes from the fact that AIDS is not transmitted by way of bloodsucking mosquitoes or efficiently through punctures of health care workers with contaminated needles. Presumably, in the mosquito example, this is because the virus does not last long enough between successive blood meals of the female mosquito. Additionally, it may not be transmitted because infection requires large inoculums. Similarly for the second example, the fact that HIV is quite seldom transmitted through needle sticks to health care workers, is possibly because most particles are inactive and many viable particles are needed to cause infection of a healthy health care provider.While these could be trivial or based on molecular biological necessity, I suggest that the virus has been selected by evolution to extend these characters. Evidently, this is the characteristic of particular viruses, but it also may be due a human or primate characteristic. This can be if mankind has habitually, over the eons, been exposed to retrovirus type pathogens that are not AIDS. For both virus and host, their fitness is increased in not permitting infection of the still immunological-incompetent neonates. It may be that the low infectivity of neonates is because of some molecular biological or biochemical limitation or necessity, but in the retrovirus case it has been especially extended. This can be argued since there are diseases that are transmitted very efficiently by mosquito bites and by limited blood-to-blood contact (such as Hepatitis B). So I suggest that the HIV retrovirus is poorly infective due to innate biological mechanisms evolved to favor the long-term goals of its survival as an STD. This is a prediction of the model proposed here that would have aspects that could be experimentally pursued.Although I will discuss below how the retrovirus human T-cell lymphotrophic virus I (HTLV-1) avoids this difficulty by known mechanisms, just how HIV perinatally infects only a portion (15 – 30%) of children (Scott [69]; Cotton [70]) and not more is not clear. Women that acquire HIV after delivery have a higher transmission rate to their child by breast-feeding than do women previously infected. We have also a hint because reducing the viremia by zidovudine (AZT) treatment before and during the delivery process has been shown to reduce the transmission to the child (Connor et al., [71]; Spector et al. [72]: Rouse et al., [73]; Brossard et al., [74]) and now is routine. However, there are reports of several children that have been initially infected and apparently cleared themselves of the infection (Bryson [75]); this makes it difficult to compare different studies.The coping strategy of HIV, HTLV-I, and HTLV-II with their hostsHTLV-I and HTLV-II are both retroviruses of the subclass oncornaviruses that propagate primarily in human T lymphocytes. The former parasitizes the CD4-bearing helper T cells and the latter the CD8-bearing cytotoxic T cells (See Anderson [76]; Höllsberg and Hafler [77]; Wiktor and Blattner [78]; Blattner [79]; White and Fenner 1994 [5]). HIV is a retrovirus of the subclass, lentivirus, and both it and the oncornaviruses incorporate their reversely transcribed double-stranded DNA into a chromosome of a human cell as the heart of their survival strategy. Both classes of virus infect only a particular human cell-type and only a cell type that continues to divide. Nerve cells and kidney cells are not appropriate host cells because in the adult they divide only very occasionally and therefore the equivalent of the prophage state would never be adequately created or propagated. Also an inappropriate cell type would be a relatively rapidly growing epithelial cell because while the stem cell remains, the sister cell is sloughed from the skin or into the intestine and does not remain and be propagated within the body.The oncornaviruses, but not the lentiviruses, have a growth pattern that allows both effective vertical transmission and horizontal transmission to new hosts. The viruses are passed vertically to the neonate, although not with high efficiency and not in a way that causes childhood death. They are also passed as a STD between sexually active individuals. In addition, in the modern world these viruses can also be passed via needles. Before extensive movements of peoples of the world population, both HTLV viruses were geographically restricted in distribution. Although HTLV-I is now found world wide, it was and is highly abundant in southern Japan and the Caribbean. It is also present in South America, west and central Africa, India, Melanesia, and Iran. There is evidence that similar viruses inhabit non-human primates. HTLV-II is now highly prevalent in intravenous drug users, but is (and presumably was) abundant in the Guaymi tribe in Panama and, more generally, in Amer-Indians.The strategy of both HTLV viruses in the absence of IV drug usage and rapid movements of peoples was very close to the 'gentle' pathogen persuasion. Both viruses persist with little damage to their host by replicating very slowly while being passed vertically or between adults as an STD. Being poorly transmitted from cell-to-cell and individual-to-individual favors the 'gentle' character, but another feature is that they have mechanisms to cause the hosts T-cells to replicate. Thus a syndrome, similar to that of AIDS does not occur, resulting from depletion of T-cells. This allows more opportunities for the virus to be transmitted between humans. They are poorly transmissible and transmission from male to female is rare and the transmission from female to male is very slow (at least for HTLV-I). Though it is passed presumably in the same three ways that HIV is passed to the neonate (i.e., in utero, perinatally by blood-to-blood transfer, and (predominately) postnatally via mother's milk). There is little damage to the child because the virus grows so slowly. Thus the immune system has time to develop and respond to these viruses and is even aided by the increased growth of T-cells. Both viruses are almost perfect 'gentle' pathogens. However, the current longevity of the human host is sufficient to cause problems (see White and Fenner, [5]). For example, there is the occasional generation by HTLV-I of an adult T-cell leukemia (ATL). Less frequently T-cell chronic lymphocytic leukemia, non-Hodgkin's lymphoma, mycosis fungoides, and Sezary syndrome arise. HTLV-II has not been adequately studied, but appears to be even more 'gentle', and only very rarely causes certain diseases: i.e., glomerular nephritis and HTLV-associated myelopathy.Molecular biological information is mainly available for HTLV-I, but the studies of the rex and tax genes are relevant to the trick that this virus has of alternately stimulating T-cell growth and then subsequently switching to the HTLV-I replication mode (Yoshida et al. [80]; Hinrichs et al. [81]). Of particular importance to the major thesis of this speculative paper is that the rex gene of HTLV-I interacts with the rev gene of HIV and this may suggest that the human virus may act to protect its host against some other pathogen.How HIV is 'Gentle' to its hostSome of the methods that HIV uses to be lysogenic and spread from host to host are obvious from its generally known biology (Daniel et al. [82]). Most obvious is that a retrovirus, in principle, would only be able to grow as the provirus in concert with a reproducing cell. HIV for example, can enter a quiescent cell and give rise to DNA products, but the DNA is only integrated after activation of the cell and replication of its chromosome (Zack et al., [83]; Stevenson et al., [84]; Haase et al., [85]). But the sole use of CD4 as a target receptor limits its opportunity to grow and reproduce. Maintenance of the latent state is another essential factor that keeps the virus from rapidly destroying the immune system. This is in addition to the host's immune system acting as an essential safeguard for viral long-term propagation within the host and transmission into other hosts.Conclusion(a) The evolutionary mechanisms that maintain the non-virulent state of a pathogen are of high interest. Since there are a large number of mild pathogens in nature and in spite of the selective advantage that a virulent mutant would temporarily have, the 'gentle' pathogen state does arise and does persist. A proposed mechanism for this, not involving group or kin selection, has been made Koch [86]. This new model is that the pathogen protects itself (and incidentally its host) against other similar or identical pathogens arriving subsequently and that this generates and maintains the 'gentle' state.(b) It is argued that pathogens that depend on humans as a resource faced a much different problem during the hunter/gatherer Stone Age than now, and at this early time the spectrum of viruses could be expected to have been largely STDs because of the sparse availability of hosts. Almost of necessity many pathogens must have been 'gentle' because the interactions between social groups were infrequent. Before moving to humans from the primates, HIV must have been gentle to its nonhuman primate host, and possibly it has not had time to readjust to become gentle its new host.(c) Being a 'gentle' pathogen requires elaborate controls to self-limit growth of its host. Persistence over long times of a sexually transmitted disease depends, therefore, on growth inhibitory mechanisms in part coded by the virus, but frequently dependent on host function. The host immune system limits the viremia and viral encoded mechanisms and may act to modulate the immune response and act in other ways to control and limit viral growth. It can be assumed that the lentiviruses of nonhuman primates, such as SIV are adapted to a low rate of vertical transmission because of the devastating action of many viruses on neonates due to the latter's underdeveloped immune system. These may be avoided by slow growth. HIV is transmitted to offspring in utero, perinatally, or via breast-feeding, but the transmission is less efficient than for some other viral diseases. This suggests that host and viral mechanisms restrict vertical transmission or its effects for the case of HIV for the fetus and the neonate.(d) The HIV retrovirus strategy depends on selectively infecting a restricted class of cells, mainly the CD4+ or CD8+ T helper cells. These and other T and B cells happen to be nearly the only suitable cells, a priori, in an mammal; these uniquely continue to grow, replicate, and divide to form progeny that remaining inside the host. Such a type of cell is the necessary condition for the strategy of retroviral growth. Thus these host cells are different than other cells of the body that only divide rarely or of cells that are shed from the animal such as skin and intestinal epithelium.(e) It is proposed that the response of the mucosal part of the host immune system is not only the key factor to the prevention of many viral infections, but also it is to prevent or reduce the infection of the offspring of an infected female. This is because the selection for an ability to elicit an effective mucosal immune response by the products generated by the virus that would block secondary infection and is in the best interests of the virus, therefore. But it also makes the resident pathogen less destructive to its host and its host's progeny. I suggest that the role of mucosal immunological response and other defensive responses is critical for the biology of the STD retroviruses.(f) The important link in the STD lifestyle as typified by the HIV/human interaction is that the virus must work against itself under conditions in which the hosts are at a premium and must somehow protect the fetus and neonate. It is the fact that such protection is manifest even with HIV infection of humans. About 85% of the offspring of HIV-infected mother are not HIV infected and in simian viruses in wild monkeys the number may be closer to 100%. Compared with some other viral diseases, this suggests that specific immune protection of the fetus perinatally does occur.(g) The change from the forebears of HIV, presumed to be a 'gentle' virus of its non-human primates, to the devastating human virus of AIDS is mostly because the new host of the virus lives much longer than its previous host. Added to this are the factors due to the human host's social behavior, most notably the extensive movements of humans from place to place. Also HIV appears very ungentle now because of its spread to a greatly expanded habitat, and because of the sociology due to the homosexual involvement and intravenous drug use.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2531103.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2531103",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2531103\nAUTHORS: Ken-ichi Inoue, Takashi Shiga, Yoshiaki Ito\n\nABSTRACT:\nRunt-related (Runx) transcription factors control diverse aspects of embryonic development and are responsible for the pathogenesis of many human diseases. In recent years, the functions of this transcription factor family in the nervous system have just begun to be understood. In dorsal root ganglion neurons, Runx1 and Runx3 play pivotal roles in the development of nociceptive and proprioceptive sensory neurons, respectively. Runx appears to control the transcriptional regulation of neurotrophin receptors, numerous ion channels and neuropeptides. As a consequence, Runx contributes to diverse aspects of the sensory system in higher vertebrates. In this review, we summarize recent progress in determining the role of Runx in neuronal development.\n\nBODY:\nHistoryRunt related (Runx) genes are evolutionarily conserved developmental regulators in metazoa, where they play diverse roles in several different biological systems, including cell differentiation. One of the Drosophila pair-rule genes, Runt, controls segmentation, sex-determination and neuronal development [1]. The mammalian Runx gene was first identified as AML1, which is frequently involved in the chromosomal translocations associated with acute myeloid leukaemia (AML) [2]. Both Runt and AML1 encode a DNA binding subunit of the heterodimeric transcription factor PEBP2/CBF. Polyomavirus enhancer binding complex (PEBP2/PEA2) was identified during the characterization of the cellular mechanisms involved in differentiation using embryonal carcinoma cells [3]. CBF was first identified as a protein that binds to the core sequence of the murine retrovirus enhancer, which influences the tissue specificity of viral replication [4].There are three mammalian RUNX genes, RUNX1 (AML1), RUNX2 (CBFA1) and RUNX3 [5]. RUNX1 is essential for definitive hematopoiesis and frequently involved in human leukaemia [6]. Runx2 is a master regulator of bone development [7]. Moreover, haploinsufficiency of RUNX2 is one of the causes of the hereditary bone disease Cleidcranial displasia [8]. RUNX3, the third member of the RUNX gene family, was the least characterized until gene targeting studies opened up new avenues of investigation into Runx function. First of all, RUNX3 is involved in many types of human cancer as a tumour suppressor [9,10]. Hypermethylation of the RUNX3 promoter and deletion of the RUNX3 gene are frequently observed in several cancers, and RUNX3 protein is now best considered as an apoptosis inducer [11,12]. Second, RUNX3 controls the generation of the T-cell sub-lineage [13-15]. In particular, transcriptional regulation of CD4 silencer and Th-POK have been described in detail [13,15]. Finally, Runx3 controls the development of proprioceptive dorsal root ganglion (DRG) neurons [16,17]. The last discovery was particularly relevant to developmental neurobiology and, since then, several groups have characterized not only Runx3, but also Runx1 as a crucial regulator of DRG neurogenesis [18,19].Expression of Runx1 and Runx3 in the nervous systemEarlier in situ hybridization studies indicated strong expression of Runx1 mRNA in spinal motor neurons, DRG, cranial ganglia and specialized sensory epithelial structures such as olfactory and gustatory mucosa, and follicles of the vibrissae [20]. Subsequently, the generation of specific antibodies against Runx1 and Runx3 and the utilization of Runx1β-gal or Runx3β-gal mice revealed the expression of Runx1 and Runx3 in the nervous system in more detail [16,21,22]. Runx1 is synthesized in both the central and peripheral nervous systems of mouse embryos. In the central nervous system, Runx1 is synthesized in selective populations of somatic motor neurons in the spinal cord and in cholinergic branchial and visceral motor neurons in the hindbrain, such as dorsal vagal nucleus and nucleus ambiguus [21,22]. In the peripheral nervous system, Runx1 is localized to DRG and selective cranial ganglia, including trigeminal (V) and vestibulocochlear (VIII) ganglia and the glossopharyngeal-vagal (IX-X) ganglia complex [21,22]. In contrast to Runx1, Runx3 is confined to the peripheral nervous system, specifically to DRG and cranial ganglia [16,21]. Although Runx1 and Runx3 are almost exclusively found in postmitotic neurons in the central nervous system and peripheral ganglia [16,21,22], a rare exception is the expression of Runx1 in proliferating progenitors of the olfactory epithelium [23]. These observations suggest Runx1 and Runx3 have extensive functions in the mammalian nervous system.Roles of Runx3 in the development of DRG neuronsDRG neurons convey peripheral somatosensory stimuli to the spinal cord. There are three major subpopulations of DRG neurons – nociceptive, mechanoreceptive, and proprioceptive – which differ in their cell size, dependency on neurotrophins, and distinct axonal terminal fields in the spinal cord and peripheral tissues. Runx1 and Runx3 are synthesized initially in TrkA+ nociceptive and TrkC+ proprioceptive neurons, respectively (Figure 1) [17,24,25]. This complementary expression pattern suggests specific roles for Runx1 and Runx3 in subtypes of DRG neurons. Indeed, the phenotype of Runx3 knockout mice is similar to that of NT3 and TrkC knockout mice [16,17,26-29]. Namely, Ia/Ib type DRG neurons fail to form a stretch reflex circuit with motor neurons in the spinal cord, resulting in severe motor discoordination [16,17]. What is the molecular basis of the phenotype? Several elegant studies have been performed to answer this question.Figure 1Runx proteins control the diversification of sensory neurons.(a) Proprioceptive (TrkC+) and mechanoreceptive (TrkB+) DRG neurons are derived from the common precursors (TrkB+, TrkC+). During segregation of two complementary sensory populations, Runx3 represses trkB expression in TrkC+ neurons. (b) During early postnatal periods, TrkA+ DRG neurons differentiate into two nociceptive subpopulations; TrkA+ peptidergic neurons, and Ret+ non-peptidergic neurons that repress trkA. In Ret+ non-peptidergic neurons, Runx1 represses trkA and neuropeptide CGRP. Runx1 also activates a number of nociceptor-specific G protein coupled receptors, ATP channels, and TRPV channels. (c) G protein coupled receptor MrgA, B and C are under dynamic transcriptional regulation in DRG neurons. A carboxy-terminal VWRPY motif of Runx proteins is critical for binding to Groucho corepressor. Runx1, which lacks VWRPY, fails to repress MrgA, B and C in DRG neurons.First, the role of Runx3 in the neurotrophin receptor phenotype was shown by Arber and her colleagues [25], who thoroughly compared neurotrophin receptor synthesis in mouse strains in which Runx3 had been disrupted or expressed ectopically. In DRG neurogenesis, dynamic changes are observed during the synthesis of neurotrophin receptors (TrkB, TrkC) [25]. At early developmental stages, most DRG neurons synthesize TrkC protein first before the onset of TrkB synthesis. Thus, some TrkC+ DRG neurons co-synthesize TrkB (Figure 1a). Subsequently, the ratio of TrkB/TrkC-hybrid neurons declines to produce DRG neurons that synthesize either TrkC or TrkB (Figure 1a). During this segregation, Runx3 is observed in most TrkC+ neurons but not in TrkB+ neurons [25]. One of the functions of Runx3 is to repress TrkB when DRG neurons acquire TrkC+ identity (Figure 1a) [25].Second, the axonal outgrowth and/or axonal guidance of propiroceptive DRG neurons are also regulated by Runx3. Two different interpretations were proposed for the phenotype of the Runx3-/- DRG. One group proposed that Runx3 controls the appropriate axon targeting of trkC-expressing proprioceptive DRG neurons to motor neurons [16]. However, another group observed massive cell death of TrkC+ neurons in Runx3-/- DRG in apparent contradiction to the previous proposition [17]. A recent study with Runx3 and Bax-double knockout mouse revealed clearly that the axonal projection of propioceptive DRG neurons to motor neurons is still lost in the Runx3 mutant even in the absence of apoptosis [30]. The study further clarified that the initial model 'Runx3 → TrkC and Runx1 → TrkA' might not apply to later developmental stages [30]. They observed that Runx3 co-localizes not only with TrkC, but also TrkA and TrkB at postnatal day 0 (P0) [30]. Of note, Runx1+ and Runx3+ neurons were clearly segregated at embryonic day 16.5 (E16.5) but almost all Runx3+ neurons co-synthesize Runx1 at E18.5 and P0 [30]. It is possible that Runx3 has some functions not only in proprioceptive neurons, but also in nociceptive neurons [30]. Overall, the evidence obtained from Runx3 and Bax compound mutants support a role for Runx3 in the control of axonal projection, although the molecular mechanisms remain unknown [30]. Prior studies showed that DRG explants from Runx3-knockout mouse embryos extended short neurites in the presence of NT3, a ligand for TrkC, but not in the presence of NGF, a ligand for TrkA [16]. This suggests that Runx3 may regulate the axonal outgrowth of specific DRG neurons independently of the target tissue. On the other hand, Chen et al. [24] revealed, using a tour de force method, that Runx3 activity determines the dorso-ventral position of axonal termination of DRG neurons in the spinal cord. DRG neurons with high Runx3 activity extended their axons far into the ventral spinal cord like proprioceptive neurons, whereas those neurons with low Runx3 activity extended their axons into the dorsal spinal cord. Ectopic expression of Runx3 is sufficient to drive axons from the dorsal to the ventral spinal cord, indicating that Runx3 per se has instructive roles in central axon targeting in DRG neurons.Thus, Runx3 controls the neurotrophin receptor phenotype as well as the axonal projection of proprioceptive DRG neurons. The two functions may not be mutually exclusive but closely related to each other. For example, NGF/TrkA signalling and NT3/TrkC signalling are required for proper axonal projection [31,32].Roles of Runx1 in the development of DRG neuronsIn contrast to Runx3, the study of Runx1 function in DRG development was delayed owing to the early embryonic lethality of the targeting mouse [22,23,33]. Thus, Runx1 knockout mice die due to a lack of definitive hematopoiesis by E12.5, which is before the onset of major events in the development of TrkA+ DRG neurons. However, recent studies have investigated the roles of Runx1 in DRG neurons using different experimental models.First of all, Runx1 controls the lineage diversification of nociceptive neurons [25,33,34]. During late embryonic and early postnatal periods, trkA-expressing neurons differentiate into two subpopulations of nociceptive neurons; trkA-retaining peptidergic neurons, and non-peptidergic neurons that repress trkA and instead activate Ret, a receptor for glial-derived neurotrophic factor (GDNF; Figure 1b). During the late embryonic stages, most trkA-expressing DRG neurons coexpress Runx1 (Figure 1b). Postnatally, Runx1 disappears in trkA-retaining peptidergic neurons but continues to exist in Ret-inducing non-peptidergic neurons (Figure 1b). Using the Runx1-conditional knockout mouse, it was shown that Runx1 is dispensable for the de novo induction of TrkA [34]. This was confirmed by Shiga and his colleagues [33], who used a different gene-targeting method that relied on the rescuing of Runx1 expression in hematopoietic cells. However, Marmigere et al. [35] showed that virally expressed Runx1 induced de novo synthesis of TrkA in the DRG and spinal cord of chick embryos. One possible explanation is that the minimal enhancer of trkA, which Runx1 regulates [35], may not be required for the de novo induction of trkA expression [36]. On the other hand, Runx1 is essential for the late repression of trkA and induction of Ret when TrkA+ and Ret+ neurons segregate (Figure 1b) [34]. In addition to trkA, Runx1 also represses the neuropeptide, calcitonin-gene-related peptide (CGRP; Figure 1b) [25,33,34]. More surprisingly, nearly all the known marker genes for nociception are under the control of Runx1. In the conditional Runx1 mutant DRG, expression of a number of nociceptor-specific G protein coupled receptors, ATP channels, and TRPV channels is attenuated (Figure 1b) [34].Similar to Runx3, Runx1 also regulates the axonal outgrowth and guidance of nociceptive neurons. Marmigere et al. [35] revealed that the transfection of Runx1 into boundary cap-derived neural crest stem cells increased neurite length and branching. In Runx1-knockout mice, the axonal projection to laminae IIi of the dorsal spinal cord was perturbed [33,34]. In the wild type, peptidergic nociceptive axons project to layer I/IIo in the superficial dorsal horn, whereas non-peptidergic nociceptive axons project to deeper layer IIi. In Runx1-knockout mouse, non-peptidergic axonal projection displays dorsal shift to layer I/IIo [34].Thus, Runx1 controls a battery of genes that are associated with the generation of non-peptidergic nociceptive neurons. The findings that both Runx3 and Runx1 play critical roles in distinct sensory neurons suggest that Runx factors are involved in the evolution of sophisticated sensory systems in higher vertebrates.Upstream/downstream genesThe upstream signals and transcriptional regulation of RUNX genes have been studied in non-neuronal tissues [37]. However, only limited studies have addressed this issue in the nervous system. Both Runx3 and Runx1 genes contain Brn-3a binding sites in their 5'-upstream regions, suggesting that Runx3 and Runx1 are candidate downstream targets of Brn-3a, a well characterized transcription factor in sensory neurons [38,39]. Microarray studies have shown decreased levels of Runx1 and Runx3 transcripts in the sensory neurons of Brn-3a-knokout mice [40,41]. Kramer et al. [25] investigated the putative upstream signal of Runx1/Runx3 in DRG neurons. Plausible candidates are TrkC/TrkA signalling and the basic helix-loop-helix transcription factors Ngn2/Ngn1; however, a genetic study has excluded these possibilities [25]. Ginty and colleagues [42] investigated the roles of NGF and the Ret receptor in DRG neurons. In Ngf-Bax compound knockout DRG, TrkA neurons are hypotrophic although de novo Runx1 expression is unaffected [42]. However, Runx1 expression is not maintained to the neonate stage and the expression of all putative Runx1 target genes is altered [42]. Thus, NGF signalling is essential for sustained expression of Runx1. In Ret conditional knockout DRG, Runx1 expression is normal but a part of Runx1 target genes are affected, suggesting the GFR/Ret dependent transcriptional regulation by Runx1 in DRG neurons [42]. Although this study placed Runx in a pivotal position in developmental signalling cascades, the upstream signalling event(s) still remains elusive.On the other hand, how does Runx1/Runx3 regulate downstream transcriptional cascades? In DRG neurons, TrkC is a critical signalling receptor involved not only in the control of cell survival, but also in axon path-finding and fate determination of proprioceptive DRG neurons [32,43,44]. Therefore, it is natural to infer that trkC is a transcriptional target of Runx3 [17]. However, unbiased computational analysis suggested that a cis-regulatory element exists in the gene locus of TrkB, rather than in the gene locus of TrkC [45]. This was unexpected because trkB is expressed in neurons of an alternative sensory fate, TrkB+TrkC- neurons [43]. The strategy \"to repress alternative traits\" appears to be a common feature in neuronal lineage commitment [46]. At the molecular level, trkB possesses a conserved cluster of Runx binding sites that function as a silencer of the trkB promoter in cultured DRG neurons [45]. In Runx3 knockout DRG, derepression of trkB seems to be a crucial event, influencing lineage commitment [25,45], and, eventually, resulting in drastic behavioural consequences [16,17].Runx protein works both as an activator and repressor, depending on the molecular context [47]. The finding that Runx3 represses trkB raises a question as to the identities of its partner molecules in the transcriptional repressor complex. The function of Runx1 as a transcriptional repressor has been widely studied [48,49]. A plausible candidate in the context of DRG is the Groucho corepressor. In motoneuron fate specification, Groucho-mediated repression is a common mechanism for homeodomain proteins containing the EH1 domain [46]. Runx proteins have the evolutionarily conserved VWRPY carboxy-terminal motif, which is considered to be critical for Groucho binding/function [50,51]. Yarmus et al. [52] generated mice in which Runx3 lacks these amino acids. Surprisingly, VWRPY knockout mice displayed the normal development in DRG neurons, though they showed the phenocopy to Runx3 knockout mice in dendritic cells [52]. The results suggest that Runx3 represses trkB through a Groucho independent mechanism. Recently, Ma and his colleagues [53] investigated the significance of the VWRPY motif of Runx1 in DRG neurons. Runx1 cDNA, which lacks the VWRPY coding sequence, was knocked into the native Runx1 locus in delta446 mice [54]. In the delta446 mice, derepression of Mrg class G-protein-coupled receptor genes was observed, suggesting that Mrg genes are repressed by a Groucho-dependent mechanism (Figure 1c) [53]. Interestingly, two putative target genes that are repressed by Runx1, trkA and CGRP, were unaffected in the delta446 mice [53]. These results suggest that Runx1 represses target genes through either a Groucho-dependent or an independent mechanism in DRG neurons.Chen et al. [34] indicated that Runx1 controls nearly all known marker genes critical for nociceptive functions. Such global control by Runx over the transcription landscape is also observed in other physiological functions, such as hematopoietic stem cell formation (Runx1) and osteoblast maturation (Runx2). How this unique transcription factor has such a huge influence on many different transcriptional cascades remains a challenging question.Other neurological phenotypes of Runx1/Runx3 knockout miceStifani and his colleagues [22,23] have worked on the neurological phenotypes of the Runx1 knockout mouse other than those arising from defects in DRG neurons. They analysed the cranial sensory neurons as well as cholinergic branchial and visceral motor neurons of hindbrain at an early embryonic stage [22]. The expression of Runx1 was restricted to post-mitotic neurons, and disruption of Runx1 resulted in massive neuronal apoptosis [22]. In contrast to this finding, Runx1 is expressed in the proliferating neuronal progenitors/precursors of olfactory receptor neurons (ORNs) [23]. Runx1 drives the cell cycle in ORN progenitors through transcription repression of the cyclin dependent kinase inhibitor p21 [23]. Unlike DRG, they did not observe any changes in the lineage markers in the neurons examined (cranial, hindbrain and olfactory), indicating that Runx1 has distinct functions in different types of neurons [22,23].The study of the neurological function of Runx3 other than in DRG is very limited. Levanon et al. [17] reported that TrkC+ neurons in the trigeminal ganglion survive in contrast to DRG neurons in Runx3-/- mouse. Most Runx3 knockout mice of the C57/B6 strain die within one day after birth [9,16]. The main cause of death may be starvation, as little milk is found in the stomachs of these mice [9]. As this is probably related to the pups being unable to swallow milk, it is interesting to note that Runx3 is strongly expressed in cranial ganglia, including the glossopharyngeal ganglion [16,17]. It is possible that Runx3 is essential for the functional glossopharyngeal system (swallowing), suggesting the critical roles in developmental cranial neurons.ConclusionAlthough the roles of Runx in neural development have just begun to be investigated, studies in gene knockout mice indicate that the roles of Runx in the nervous system are as important as its roles in other, non-neuronal tissues. However, a number of open questions should be addressed in the future. First, upstream signalling cascades remain elusive. The mRNA expression and protein synthesis for Runx1/Runx3 are tightly regulated and DRG is one of the tissues in which Runx1/Runx3 display their highest protein levels among the entire body; how do DRG neurons achieve such a high protein level for Runx1/Runx3? Second, the molecular bases of tissue specificity are largely unknown. Runx1 and Runx3 are highly homologous but they control the development of distinct subpopulations of sensory neurons. In particular, Runx1+ neurons and Runx3+ neurons project axons into totally different target tissues; how is this specificity achieved? Third, transcriptional regulation is not the only determinant of DRG neurogenesis. Ectopic synthesis of TrkC receptor per se influences the lineage commitment of DRG neurons [44], while Runx3 plays a crucial role in TrkB/TrkC status [25,45]. It is likely that Runx and neurotrophin status are closely related to each other. How this cross-regulation is carried out is a challenging question. Finally, since all three Runx proteins have common features, some of the knowledge about Runx function in oncology, haematology, immunology and bone biology is likely to be applicable to neuroscience as well, particularly at the molecular level [53].Competing interestsThe authors declare that they have no competing interests.Authors' contributionsThe first draft of this review was written by KI together with TS, which was then complemented by YI. The figure was composed by KI.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2531107.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2531107",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2531107\nAUTHORS: Mark C Perry, Leon M Straker, Peter B O'Sullivan, Anne J Smith, Beth Hands\n\nABSTRACT:\nBackgroundAdolescent neck/shoulder pain (NSP) is a common and sometimes debilitating problem. Several risk factors for this condition have been investigated, but no studies have previously evaluated associations between fitness, motor competence, body composition and adolescent NSP.Methods1608 males and females of mean age 14 years answered questions on their history of NSP (4 measures), and were tested for aerobic fitness, upper and lower limb power, trunk endurance, grip strength, shoulder flexibility, motor competence and anthropometric factors. Univariate and multivariate logistic regressions were used to test for associations between NSP and physical variables.ResultsThere were significant gender differences for most physical and pain variables. After multivariate analysis, males had lower odds of NSP if they had reduced back endurance [OR: 0.66 (95% CI: 0.46–0.97)], reduced persistent control [0.42 (0.19–0.95], and increased muscle power [0.33 (0.12–0.94)], and higher odds of NSP if they had a higher basketball throw [2.47 (1.22–5.00)] and jump performance [3.47 (1.55–7.74)]. Females had lower odds for NSP if they had a reduced jump performance [0.61(0.41–0.92)], a better basketball throw [0.60(0.40–0.90)], lower shoulder flexibility [0.54 (0.30–0.98)] and a higher aerobic capacity [0.61 (0.40–0.93)], and higher odds for NSP if they had greater abdominal endurance [1.57(1.07–2.31)] and greater bimanual dexterity [1.77(1.18–2.65)]. Females showed a U shaped relationship between NSP and back endurance [low: 2.12 (1.20–3.74); high 2.12 (1.18–3.83)].ConclusionAdolescent NSP was associated with fitness and motor competence, although the associations varied with gender, and their strength was limited.\n\nBODY:\nBackgroundNeck/shoulder pain (NSP) may affect up to half of adolescents [1], leading to significant loss of function [2]. Up to 25% of adolescents with NSP experience some degree of disability [3] and 11% may require prescription drugs to manage pain [4]. Some risk factors for adolescent NSP have been identified, including high levels of computer use [5], employment [6], negative psychosocial factors [6-8], female gender [8], and sustained postures [1]. Very low and high levels of physical activity [7] are also associated with adolescent NSP. Activity levels may influence NSP directly, or via other factors such as physical characteristics.Physical characteristics such as muscle strength, flexibility, endurance or motor competence may be associated with spinal posture [1,7,9] or spinal stability [10], both of which may have an association with spinal pain [1,4,11]. There is some evidence that physical characteristics and adult NSP are related. Adult studies have reported that decreases in neck flexibility [12], neck endurance [12], neck muscle motor control [13], grip strength [14] and high body mass index (BMI) [15] are associated with NSP.However the evidence for a link in adolescents is less clear. It has been noted [16] that low levels of flexibility in male adolescents and low levels of trunk endurance in female adolescents have been associated with a greater risk of \"tension neck syndrome\" 25 years later. Similarly, lower arm endurance in males during adolescence has been associated with more NSP in adulthood [14]. With respect to body composition, one study noted no associations between BMI at age 14 and NSP in early adulthood [17]. However, only one study to our knowledge has investigated the association of adolescent physical characteristics with NSP experienced during adolescence [6] and it reported no association between NSP and BMI. Salminen [18] investigated the relationship between flexibility and adolescent NSP, but this was in conjunction with LBP, and so a specific relationship with NSP was not defined. No adolescent studies have investigated the links between NSP and aerobic capacity or motor competence.There is therefore a need for an initial exploratory study to examine the suspected links between adolescent NSP and certain physical characteristics. If it can be shown that any of these physical characteristics are related to adolescent NSP, this will provide the basis for further longitudinal work, which may in turn inform the development of specific strategies to prevent this common problem. The research question was whether lower and/or higher levels of fitness, motor competence and body composition were related to increased risk of NSP in adolescents. The physical variables used in this study reflect generalized motor performance and characteristics, rather than specific neck muscle performance, as the more general variables bear a greater relation to performance measures customarily used in schools and clinics.MethodsParticipantsData were collected from 1608 adolescents (783 females, 825 males) of mean (SD) age 14.06 (0.20) yrs, who were participating in the Western Australian Pregnancy Cohort \"Raine\" Study . This project began with a cohort of women attending antenatal clinics at King Edward Memorial Hospital for Women, Perth, Australia between 1989 and 1991. The children have been followed at birth, 1, 2, 3, 5, 8, 10, and now 14 years of age. Inclusion criteria for the women were gestational age of 16–20 weeks, adequate English language to understand the implications of participation, and an intention to remain in Western Australia throughout follow-up. 2337 adolescents were eligible for the 14 year follow-up, and 1704 (72.9%) of these agreed to participate in some aspect of the follow-up. 1608 (68.8%) completed the data collection requirements for the analysis reported in this paper. There were no exclusion criteria for this part of the cohort. A comparison of the cohort with the Western Australian general population showed a higher proportion of high risk births, as would be expected for a major specialist hospital [19].ProcedureWith the assistance of a research assistant, participants completed a laptop questionnaire at an assessment centre. The questionnaire contained 130 questions concerning a broad range of issues, many of which were not relevant to this study. Adolescents were asked about their experience of NSP, described as pain in the area of the posterior neck and upper trapezius, as diagrammatically defined by Kourinka et al. [20]. The relevant NSP questions were: Have you ever had neck/shoulder pain? (\"yes\" or \"no\"), Has your neck/shoulder been painful in the last month? (\"yes\" or \"no\"), and Did your neck/shoulder pain last for more than 3 months? (\"yes\" or \"no\"). The full questionnaire took about 1 hour to complete, and the NSP questions occurred in the first half. The life prevalence question is very similar to that used by Chiu and Leung [21], which was shown to be reliable and valid.Information on diagnosed neck pain was obtained from a paper questionnaire given to the primary carer, which included the question, \"Does your child have now, or has your child had in the past, any of the following health professional diagnosed medical conditions or health problems?\". The primary carer had to indicate which medical diagnoses their child had experienced from a short list of general medical problems, which included \"neck pain\". This question was part of a questionnaire given to the primary carer, covering many other factors that are not relevant to this study.A physical assessment of the child was carried out after the laptop questionnaire, and parts relevant to this study are described below. Height (m), body mass (kg), waist girth (cm) and arm girth (cm) were measured without shoes. Several physical performance tests were then carried out. Maximal aerobic capacity was estimated from heart rate recordings during sub-maximal cycle ergometry using the Physical Work Capacity 170 protocol [22]. The sustained back extension test [23] and the number of abdominal curls performed in 3 minutes [24,25] were used to measure trunk muscle performance. Limb muscle performance was evaluated by standing long jump [24,26], seated basketball throw [24,27] and grip strength [26,27]. Flexibility was tested using the shoulder stretch [28]. Motor competence was evaluated using the McCarron Assessment of Neuromuscular Development (MAND) [26]. The Neurodevelopmental Index (NDI) was derived from summing gender and age corrected scores from 10 tests of fine and gross motor skills, and then converting to a scale with 100 as the mean and a SD of 15. Four sub indices (muscle power, kinaesthetic integration, bimanual dexterity and persistent control) [26], each comprising two of the items, were calculated. Details of the 10 items used to generate the overall NDI and 4 sub indices are shown in table 1.Table 1Summary of MAND testsTestMeasurementSub-indicesRod slideSmoothness and slowness of moving handle along a metre long rod, repeated both hands.PCFinger/nose fingerAccuracy and smoothness of index finger from nose to opposite hand's index finger, repeated both sidesPCHand strengthHand grip strength with a hand dynamometer, repeated both sidesMPStanding long jumpDistance and quality of two footed jumpMPHeel toe walkQuality of walking forwards and backwards along a 10 foot lineKIStanding one legTime of balance on each leg with eyes open, then eyes closed.KIBeads on rodNumber of beads placed on rod held in non-dominant hand in 30 seconds, repeated with eyes open and closedBDNut and boltTime to turn a large bolt, held in the dominant hand, fully onto a nut, repeated with a small bolt.BDFinger tappingNumber and quality of taps of index finger in 10 seconds, repeated both hands-Beads in boxNumber of beads moved from one box to an adjacent box in 30 seconds, repeated both hands.-PC = Persistent Control, MP = Muscle Power, KI = Kinaesthetic Integration, BD = Bimanual DexterityPC = persistent control, MP = muscle power, KI = kinaesthetic integration, BD = bimanual dexterityAll of these physical performance tests have been previously validated in very similar forms [25-27,29-31] except the shoulder stretch, which has acceptable face validity. Reliability of the same or similar versions of the tests is also good [26,30-32], although there are no reports on the shoulder stretch or the basketball throw.Data analysisGender differences were analysed using independent t tests for each of the continuous variables, and Chi squared tests for the categorical variables. To facilitate the interpretation of non-linear relationships, continuous variables were banded into the bottom 25%, inter-quartile range and top 25%, and the proportion of subjects with neck pain in each segment were compared. Univariate logistic regression models predicting lifetime, last month, chronic (lasting more than 3 months) and diagnosed neck pain from each physical characteristic were calculated separately for males and females, with statistical significance set at p < 0.05, and the interquartile range of each continuous variable defined as the reference category. For the only binary variable (shoulder stretch), being able to perform the stretch was the reference category. Corrections for multiple univariate tests were not carried out as the multivariate results were the end point of the study.Backwards stepwise likelihood ratio multivariate logistic regression models were used to evaluate the combined associations of performance factors for males and for females, with the probability for entry and removal of the likelihood ratio score statistic being p = 0.05 and 0.10 respectively. For each gender, 4 multivariate analyses for each of neck pain ever, last month, chronic and diagnosed were performed. Height and weight were included in an initial step, with abdominal curls, basketball throw, jump, back muscle endurance, PWC170, hand strength and shoulder stretch included in a second step, along with one of the body composition variables and either the NDI score or the 4 motor competence factor scores. The choice of body composition variables or motor competence variables was determined by the strength of univariate relationships with pain, and was determined for each of the eight multivariate analyses separately. The strength of the predictive ability of the model was estimated by Nagelkerke R2. All statistical analysis was performed using SPSS version 13.EthicsAdolescents provided written informed assent and their parent/guardian provided written informed consent prior to participation. The study was approved by the Human Research Ethics Committees of Curtin University of Technology and Princess Margaret Hospital.ResultsNeck/shoulder painNSP ever was experienced by 46.8% of the participants, NSP in the past month by 28.7%, 'chronic' (lasting more than 3 months) NSP by 8.2% and diagnosed NSP by 7.1%. Females had a significantly higher prevalence of NSP ever, month and chronic, but not diagnosed NSP (see table 2). 64.9% of those with chronic NSP also had experienced NSP within the last month and 20.9% of those with chronic NSP also had diagnosed NSP according to parental report.Table 2Pain prevalence and physical test performance for males and femalesPain variableAll Participants % (count) with history of painMale % (count) with history of painFemale % (count) with history of painGender differenceχ2PNSP ever46.841.952.016.3<0.001NSP in last month28.722.934.727.1<0.001Chronic NSP8.26.89.84.80.029Diagnosed NSP7.16.97.20.050.828Physical variableAll Participants mean (sd)Males mean (sd)Females mean (sd)Gender differencetdf(unless stated otherwise)PHeight1.64 (0.08)1.66 (0.09)1.62 (0.06)-0.421598<0.001Weight57.7 (13.2)58.6 (14.1)56.7 (12.1)-1.9215990.004BMI21.29 (4.15)21.05 (4.14)21.53 (4.16)2.3015980.022Waist girth (cm)75.5 (10.8)76.3 (11.4)74.6 (10.1)-3.1515790.002Arm circumference (cm)25.2 (3.3)25.3 (3.4)25.1 (3.3)-1.1715990.244PWC 170 score (W)111.2 (29.9)124.3 (31.7)97.2 (19.9)-19.601501<0.001Back muscle endurance (seconds)80.9 (60.4)77.8 (60.1)84.2 (60.5)2.1215740.034Abdominal muscle endurance (number of curls in 3 min)21.4 (17.4)25.4 (18.8)17.2 (14.6)-9.601569<0.001Standing long jump distance (metres)1.46 (0.29)1.59 (0.28)1.32 (0.23)-20.901588<0.001Basketball throw (metres)5.3 (1.0)5.7 (1.0)4.8 (0.7)-21.721583<0.001Total Hand strength – right and left combined (kg)51.8 (13.5)57.0 (14.8)46.3 (9.1)-17.201597<0.001Shoulder stretch (L) (%able)88.9%84.7%92.8%χ2 = 26.11 <0.001NDI score97.2 (17.4)97.3 (18.1)97.0 (16.6)-0.3115760.741Persistent control factor score103.3 (25.4)99.9 (26.4)106.8 (23.7)5.441594<0.001Muscle Power factor score95.9 (20.2)102.4 (19.8)89.2 (18.5)-13.791583<0.001Kinaesthetic Integration factor score96.9 (15.2)96.7 (15.7)97.2 (14.7)0.6815960.501Bimanual Dexterity factor score97.1 (19.3)95.1 (19.3)99.1 (19.1)4.091598<0.001Females had more NSP ever (P < 0.001, χ2 = 16.26), more NSP in the past month (P < 0.001, χ2 = 27.11) and more chronic NSP (P = 0.029, χ2 = 4.75). There were significant gender differences for height and weight and all physical characteristics (P < 0.05) except arm circumference, kinaesthetic integration and the NDI score.*P < 0.1, **P < 0.05, ***P < 0.01. OR = Odds Ratio, 95%CI = 95% confidence interval, IQR = interquartile range, NDI = neurodevelopmental index, MP = muscle power, PC = persistent control, BD = bimanual dexterity, PWC = Physical Work Capacity.1 group unable to do stretch is compared to group able to do stretch.Physical characteristicsDescriptive statistics for physical characteristics are given in table 2. Females obtained significantly higher mean scores for BMI, back endurance and the motor competence factors of Persistent Control and Bimanual Dexterity. A greater proportion of females could perform the shoulder stretch test. Males obtained significantly higher mean scores for waist girth, aerobic capacity, abdominal curl number, standing long jump, basketball throw, grip strength and the motor competence factor of muscle power. Males were also taller and heavier, with a lower BMI. There were no gender differences in arm circumference, NDI or Kinaesthetic integration.Associations between NSP and physical performanceMalesOn univariate analysis, there was an increase in risk of NSP in the past month for male subjects with greatest height, an increase in risk of chronic NSP for male subjects with greatest basketball throw and NDI, and a decreased risk for chronic NSP for male subjects with the lowest weight and arm circumference. There were no significant effects on the risk of diagnosed NSP (Table 3).Table 3Univariate relationship between physical characteristics and neck/shoulder pain in males.Variable groupPhysical variable% with pain in each groupLog Regression Lowest 25% relative to IQRLog Regression Highest 25% relative to IQRlow QIQRhigh QpORCIpORCINSP everAnthropomheight42%40%45%0.6351.080.781.520.2351.240.871.76weight42%41%44%0.8701.030.731.450.5671.140.811.60Body compBMI45%40%43%0.1981.250.891.760.4721.130.811.60waist circ.43%42%42%0.8081.040.741.460.9101.020.721.44arm circ.42%41%44%0.7921.050.751.460.4091.160.821.63AerobicPWC 17042%42%42%0.8931.020.721.450.9721.010.711.43Muscle performanceback end.38%45%41%0.1270.770.541.080.3900.860.611.22curls43%41%42%0.5131.120.801.580.6901.070.761.52jump44%41%41%0.5091.120.801.580.9351.020.721.44throw40%43%40%0.4970.890.631.250.3880.860.611.21hand strength39%43%43%0.3700.860.611.200.9251.020.721.44Flexibilitysh stretch 134%43%0.0830.700.471.04Motor competenceNDI45%43%36%0.5231.120.801.570.1480.770.541.10PC43%42%40%0.7591.050.751.480.6610.930.651.31MP41%44%37%0.4270.870.621.220.1790.760.511.13KI45%40%42%0.2081.230.891.710.7511.060.731.56BD44%42%38%0.6031.080.801.470.5040.870.581.31NSP monthAnthropomheight21%21%30%0.8941.030.691.540.010**1.691.132.50weight19%23%27%0.2170.770.501.170.2511.250.851.85Body compBMI22%23%24%0.8500.960.651.440.7090.920.581.46waist circ.20%23%25%0.3560.830.551.240.6561.090.741.63arm circ.19%25%24%0.1000.710.481.070.8050.950.641.42AerobicPWC 17021%23%24%0.7470.930.611.420.7511.070.711.61Muscle performanceback end.20%25%21%0.1040.710.471.070.2250.780.511.17curls25%20%25%0.1391.360.910.200.1501.350.902.02jump23%23%23%0.8821.030.691.540.9351.020.681.53throw21%24%22%0.3350.820.541.230.6020.900.601.34hand strength18%24%26%0.1120.720.471.080.6381.100.741.63Flexibilitysh stretch 119%23%0.2750.760.471.23Motor competenceNDI25%23%20%0.5981.110.751.650.4470.850.561.29PC21%25%20%0.2680.800.531.190.1520.740.491.12MP20%25%20%0.1630.750.501.120.2540.760.471.22KI24%23%22%0.7651.060.721.560.7530.930.591.46BD24%23%22%0.8351.040.731.490.8050.940.581.52NSP chronicAnthropomheight6%6%9%0.9370.970.491.930.2541.460.762.79weight3%8%8%0.050*0.430.191.000.7241.120.602.07Body compBMI6%7%7%0.7400.890.451.760.7990.920.471.80waist circ.7%7%7%0.9581.020.521.990.7981.090.562.13arm circ.4%9%7%0.028*0.430.200.910.3570.730.381.42Muscle performancePWC 1706%8%6%0.3690.720.351.470.3610.720.351.46back end.7%6%9%0.4531.290.662.540.1831.560.812.99curls7%6%8%0.7591.120.562.240.2971.420.742.74jump7%7%7%0.8930.960.491.880.9521.020.522.01throw6%5%10%0.3941.370.662.840.010**2.331.224.44hand strength5%6%10%0.3980.720.341.530.0841.720.933.18Flexibilitysh stretch 13%8%0.0540.310.091.02Motor competenceNDI8%5%10%0.1171.730.873.450.025*2.131.104.12PC6%8%6%0.3840.740.371.470.4910.790.391.56MP6%6%10%0.8770.950.481.890.1441.660.843.29KI7%6%8%0.9561.020.531.970.6141.200.592.48BD8%6%4%0.3171.340.752.390.2550.570.211.51NSP DiagAnthropomheight8%7%7%0.6901.140.602.170.9911.000.502.03weight4%7%9%0.1100.520.231.160.5871.190.642.21Body compBMI6%7%8%0.6280.840.421.690.7651.100.582.12waist7%6%9%0.8701.060.532.120.2101.510.792.86arm5%7%9%0.4330.750.371.530.4401.290.682.43Muscle performancePWC 1709%6%6%0.3001.410.742.690.7930.910.431.89BE7%7%7%0.9851.010.521.970.9200.970.491.92Curls6%6%8%0.9340.970.481.980.4621.280.662.50Jump8%6%9%0.3381.390.712.730.0871.770.923.40BT5%6%9%0.6470.840.411.750.3141.390.732.64HS6%8%5%0.3190.710.371.390.2260.640.311.32Flexibilitysh stretch 18%7%0.4891.290.632.63Motor competenceNDI9%6%7%0.2301.490.782.860.4971.270.642.53PC5%8%5%0.1500.600.301.210.1740.600.291.25MP9%7%5%0.4441.270.692.350.4460.710.291.73KI7%6%9%0.6631.160.602.220.3181.440.712.91BD8%7%3%0.5721.180.672.090.1100.420.141.22OR = Odds Ratio, 95%CI = 95% confidence interval, IQR = interquartile range, BMI = body mass index, PWC = physical work capacity, SS = shoulder stretch (L), NDI = neurodevelopmental index, PC = persistent control, MP = muscle power, KI = kinesthetic integration, BD = bimanual dexterity. *P < 0.05, **P < 0.01*P < 0.1, **P < 0.05, ***P < 0.01. OR = Odds Ratio, 95%CI = 95% confidence interval, IQR = interquartile range, NDI = neurodevelopmental index, MP = muscle power, PC = persistent control, BD = bimanual dexterity, PWC = Physical Work Capacity.1 group unable to do stretch is compared to group able to do stretch.For multivariate analysis, the common variables described in the methods section were entered, together with BMI and NDI for the NSP ever analysis, arm circumference and the four MAND factors for the NSP month analysis, arm circumference and NDI for the NSP chronic analysis, and waist circumference and the four MAND factors for the NSP diagnosed analysis, according to the strength of univariate relationships. Males in the lowest quartile of back muscle endurance were less likely to have NSP ever, and there was a similar trend for those in the highest quartile of NDI. There were no multivariate associations between physical characteristics and male NSP in the past month. Males in the highest quartile of basketball throw distance were more likely to have chronic NSP and there was a trend for those unable to do the shoulder stretch to be less likely to have chronic NSP. Males in the highest quartile of jump distance were more likely to have diagnosed NSP, but those in the highest quartile of muscle power were less likely to have diagnosed NSP. Those in the lowest quartile of persistent control were less likely to have NSP. Nagelkerke R2 of logistic regression models ranged from 0.019 to 0.085 (Table 4).Table 4Multivariate relationships between physical characteristics and each type of neck/shoulder pain in males and females.GenderNSP variablePhysical characteristics associating with NSP (at P < 0.1)Lowest 25% relative to IQROR, (95% CI)Highest 25% relative to IQR,OR (95% CI)MaleNSP everNDI1.26 (0.87–1.83)0.73 (0.50–1.1)*back endurance0.66 (0.46–0.97)**0.82 (0.57–1.18)NSP month---NSP chronicthrow1.96 (0.87–4.45)2.47 (1.22–5.00)**shoulder stretch10.30 (0.09–1.01)*NANSP diagnosedjump0.75 ((0.27–2.07)3.47 (1.55–7.74)***MP1.92 (0.70–5.30)0.33 (0.12–0.94)**PC0.42 (0.19–0.95)**0.69 (0.33–1.46)FemaleNSP everAbdominal curls1.36 (0.94–1.97)1.57 (1.07–2.311)**throw0.97 (0.66–1.42)0.60 (0.40–0.90)**shoulder stretch10.54 (0.30–0.98)**NANSP monththrow1.27 (0.84–1.90)0.53 (0.34–0.84)***jump0.61 (0.41–0.92)**0.70 (0.46–1.06)*BD0.86 (0.58–1.26)1.77 (1.18–2.65)***PWC0.751 (0.50–1.13)0.61 (0.40–0.93)**NSP chronic---NSP diagnosedback endurance2.12 (1.20–3.74)**2.12 (1.18–3.83)***P < 0.1, **P < 0.05, ***P < 0.01. OR = Odds Ratio, 95%CI = 95% confidence interval, IQR = interquartile range, NDI = neurodevelopmental index, MP = muscle power, PC = persistent control, BD = bimanual dexterity, PWC = Physical Work Capacity.1 group unable to do stretch is compared to group able to do stretch.FemalesOn univariate analysis there was an increase in risk for NSP in the past month for females with the highest bimanual dexterity, and an increase in risk of chronic NSP for females with the lowest hand strength. There was a decreased risk of NSP ever for females with the highest basketball throw, and a decreased risk of NSP in the past month for females with highest PWC170, and lowest and highest jump distance. There were no significant effects on the risk of diagnosed NSP (Table 5).Table 5Univariate relationship between physical characteristics and neck/shoulder pain in females.Variable groupPhysical variable% with pain in each groupLog Regression Lowest 25% relative to IQRLog Regression Highest 25% relative to IQRlow QIQRhigh QpORCIpORCINSP everAnthropomheight50%52%52%0.6580.930.661.30.9550.990.71.41weight51%53%51%0.6810.930.661.310.5640.90.641.28Body compBMI54%52%49%0.6171.090.771.540.5010.890.631.26waist circ.51%52%52%0.7790.950.681.340.97910.71.41arm circ.52%53%48%0.8000.960.691.330.3020.820.571.19AerobicPWC 17050%55%48%0.3140.830.581.190.1360.760.531.09Muscle performanceback end.51%51%55%0.8780.970.691.380.3891.170.821.65curls54%48%56%0.1811.260.91.780.0911.360.951.94jump47%55%51%0.0690.730.521.030.3510.850.61.2throw54%54%44%0.98810.711.410.039*0.680.470.98hand strength51%51%55%0.9751.010.721.410.3121.20.841.71Flexibilitysh stretch 140%53%0.0740.600.341.05Motor competenceNDI48%54%53%0.1370.770.551.090.8480.970.681.37PC47%55%52%0.0790.750.541.030.6170.90.61.35MP49%56%48%0.1180.770.551.070.0800.730.511.04KI51%52%55%0.8400.970.691.350.5831.120.751.65BD50%51%57%0.97510.721.380.1631.290.91.85NSP monthAnthropomheight35%35%34%0.9510.990.691.410.7230.940.651.35weight36%34%34%0.7011.070.751.530.9230.980.681.41Body compBMI37%35%31%0.5671.110.781.590.4310.860.61.25waist circ.37%34%32%0.4191.160.811.660.7230.940.651.36Arm circ.38%25%31%0.4761.130.811.590.3470.830.561.23AerobicPWC 17034%38%29%0.3310.830.571.210.031*0.650.440.96Muscle performanceback end.34%35%34%0.7980.950.661.370.7110.930.651.35curls33%34%37%0.8330.960.671.380.4641.150.791.66jump31%40%30%0.033*0.680.470.970.024*0.650.450.95throw38%36%28%0.6051.10.771.560.0520.670.451hand strength33%36%35%0.4450.870.611.240.8030.950.661.38FlexibilitySh stretch 131%35%0.5740.850.471.66Motor competenceNDI30%35%39%0.2850.820.571.180.2721.220.851.76PC30%36%38%0.0750.730.521.030.7251.080.711.63MP35%36%31%0.6890.930.661.320.2400.80.541.17KI31%36%38%0.2750.820.571.170.6551.10.731.64BD31%33%44%0.5940.910.641.290.008**1.641.142.37NSP chronicAnthropomheight11%9%9%0.5171.20.692.10.9410.980.531.8weight12%9%9%0.1831.460.842.530.7721.090.61.99Body compBMI12%8%10%0.0911.630.932.860.3231.350.752.43waist circ.11%9%11%0.3501.320.742.330.2821.370.772.44arm circ.12%9%10%0.2371.380.812.370.5531.210.652.26Muscle performancePWC 1709%10%11%0.5760.840.451.550.9291.030.571.84back end.10%9%10%0.6041.170.652.090.8001.080.591.97curls12%8%9%0.1771.480.842.610.7851.090.582.05jump9%11%9%0.4100.780.441.40.4150.780.421.43throw10%10%10%0.9431.020.571.820.8421.060.581.97hand strength13%7%11%0.032*1.851.053.250.1141.620.892.95Flexibilitysh stretch 17%10%0.5720.740.262.13Motor competenceNDI11%10%8%0.6171.150.662.020.6540.870.471.61PC8%10%11%0.3620.770.431.360.6651.150.612.18MP11%10%8%0.7701.090.631.880.4770.790.421.51KI9%11%7%0.3890.780.431.380.1860.610.291.27BD11%9%9%0.3981.270.732.180.8401.070.581.98NSP diagnosedAnthropomheight6%8%8%0.4290.760.381.520.9941.000.511.94weight7%7%7%0.8000.920.461.810.9800.990.511.93Body compBMI8%7%7%0.5511.220.632.380.7251.130.572.23waist circ.8%7%7%0.8541.070.552.070.7850.910.451.83arm circ.8%7%7%0.5161.230.662.300.9011.050.502.18Muscle performancePWC 1705%8%7%0.2260.620.291.340.7910.910.461.81back end.9%6%8%0.1091.710.893.290.2751.470.742.92curls5%6%10%0.5200.780.371.670.1671.580.833.03jump5%8%7%0.1960.630.311.270.4330.760.381.51throw7%7%8%0.6951.140.582.250.4601.300.652.60hand strength6%7%8%0.4670.770.381.560.7541.110.582.15Flexibilitysh stretch 111%7%0.2201.780.714.32Motor competenceNDI7%8%6%0.5470.810.411.600.4000.740.361.51PC6%7%10%0.5550.820.421.600.2721.490.733.01MP6%7%8%0.6750.860.431.720.6561.170.592.29KI6%8%6%0.4410.770.391.510.5100.770.351.70BD8%7%7%0.5361.220.652.320.6951.150.572.33OR = Odds Ratio, 95%CI = 95% confidence interval, IQR = interquartile range, BMI = body mass index, PWC = physical work capacity, SS = shoulder stretch (L), NDI = neurodevelopmental index, PC = persistent control, MP = muscle power, KI = kinesthetic integration, BD = bimanual dexterity.*P < 0.05, **P < 0.01 *P < 0.1, **P < 0.05, ***P < 0.01. OR = Odds Ratio, 95%CI = 95% confidence interval, IQR = interquartile range, NDI = neurodevelopmental index, MP = muscle power, PC = persistent control, BD = bimanual dexterity, PWC = Physical Work Capacity.1 group unable to do stretch is compared to group able to do stretch.For multivariate analysis, the common variables described in the methods section were entered, together with arm circumference and the four MAND factors for the NSP ever analysis and the NSP month analysis, and BMI and the four MAND factors for the NSP chronic and diagnosed analysis, according to the strength of univariate relationships. Females in the highest quartile of basketball throw were less likely to have NSP ever, and females in the highest quartile of abdominal curls were more likely to have NSP. Females unable to do the shoulder stretch were less likely to have NSP. Females in the lowest quartile of jump distance were less likely to have NSP in the past month. Females in the highest quartile of basketball throw, bimanual dexterity and PWC170 were less likely to have NSP in the past month. There was also a trend for those in the highest quartile of jump distance to be less likely to have NSP in the past month. There were no multivariate associations between physical characteristics and chronic NSP in females. Females in the lowest quartile of back endurance were more likely to have diagnosed NSP, and those in the highest quartile of back endurance were more likely to have diagnosed NSP. The Nagelkerke R2 of logistic regression models ranged from 0.001 to 0.064 (Table 4)DiscussionNSP is clearly a common problem in adolescents, with this study showing a prevalence of pain similar to that reported in other adolescent studies [1,2]. That almost one in ten adolescents have experienced NSP of at least 3 months duration is a strong indicator that adolescent NSP is a significant problem. The search for adolescent risk factors is therefore of great importance, so that effective prevention and management can be implemented. This study is the first to suggest that some physical characteristics are associated with adolescent NSP, although the strength of these associations was weaker than anticipated.Cross-sectional dataThis study analysed cross-sectional data only, so relationships identified could be the result of causality in both directions: NSP could be influenced by physical characteristics or vice versa. There is evidence that adults with back pain may experience a 'deconditioning' effect associated with pain inhibiting and restricting participation in work, leisure and household activities [33]. In contrast, there is evidence that poor back muscle endurance increases the risk of back pain episodes in manual workers [34].Longitudinal data (currently being collected on this cohort) is required to elucidate the direction of any relationship. The remainder of this section discusses the cross-sectional results and suggests potential mechanisms for observed relationships.Body compositionAlthough a weak univariate relationship between low arm circumference and a lower risk of chronic NSP was observed in males, body composition was not associated with any form of NSP in either gender after multivariate analysis. This concurs with previous adolescent findings [35] and underlines the importance of multivariate analysis with a comprehensive range of covariates.Aerobic capacityHigher aerobic capacity, after correction for other variables including body weight, was associated with a lower risk of NSP in the last month for females only, with a similar trend in NSP ever. The lower risk of NSP with improved aerobic capacity for NSP in females may be associated with increased levels of physical activity which is known to sometimes have a beneficial effect on spinal pain disorders [7]. This may also relate to a deconditioning mechanism, where females with NSP reduce their participation in physical activity and lose aerobic capacity. The lack of any relationships for males may indicate a differing mechanism or response to neck pain based on gender.Muscle performanceThere were inconsistent associations between arm muscle performance and NSP after multivariate analysis. Greater upper body power, as measured by the basketball throw, was protective in females for both NSP ever and in the past month, but a risk factor for chronic NSP in males. The reason for this gender difference is unclear although other factors such as specific sport participation may influence these findings. Females (but not males) engaging in more dynamic arm activities have less pain [8,36], and given that greater amounts of dynamic arm activities may increase upper body power, this may explain the pattern in females. The opposite pattern in males, with increased risk of chronic NSP in the most powerful quartile, may partly relate to greater arm activity not having a protective effect in males [8,33], and also because their high arm power may be a proxy for greater overall physical activity levels (not just upper limb activity), which relates to greater NSP in males [37]. In contrast, Barnekow-Bergvist et al. [14] reported that greater arm endurance in adolescent males was related to a reduced risk of NSP in adulthood, which may relate to a deconditioning effect secondary to NSP.Multivariate associations between NSP and leg power were very different to those with arm power. In females, a low jump performance decreased risk of NSP in the past month, effectively the opposite effect seen with upper body power. Aurvinen et al. [38] reported that higher overall activity levels may increase NSP risk in females. Since it is possible that higher overall activity may be associated with greater leg power, this may explain our finding of low leg power reducing risk. Although differing effects on NSP from arm activity levels and overall activity levels may initially appear paradoxical, it is possible that the relationship between overall activity levels and NSP is not direct but mediated by performance in sports that may increase risk of NSP. Similarly, diagnosed neck pain was associated with greater jump distance in males, although this was not observed for the other pain variables. This result may indicate a similar mechanism to that described in females.A very similar pattern was observed between abdominal endurance and NSP ever after multivariate analysis, with better performance associated with greater risk of pain in females only. Mechanisms may be similar to those described for leg power. In contrast, Mikkelson et al. [16] reported that poorer female adolescent abdominal endurance was a risk factor. However, Mikkelson et al. [16] reported these outcomes in adulthood.Less back muscle endurance was associated with a decreased risk of NSP ever in males after multivariate analysis, which was analogous to the findings for leg power and abdominal endurance in females, and may again relate to the males being involved in more vigorous physical activity [37]. Similarly, females with a diagnosis of NSP were more likely to have high back endurance, and this could relate to greater overall activity levels, as previously described. However, females with low back endurance also had a higher risk of diagnosed NSP. It is possible that these females were below a threshold of endurance at which any effects on spinal stability became important, or alternatively were experiencing a deconditioning effect as a result of the pain. However, this effect was not seen in males, who had lower back endurance overall.FlexibilityOne surprising multivariate finding was that poorer shoulder girdle flexibility, as measured by the shoulder stretch, was related to a significantly decreased risk of NSP in the past month in females. There was also a strong trend for the same effect on chronic NSP in males. Though counter-intuitive, there are reports of a relationship between lower shoulder rotational flexibility and greater upper limb activity levels in elite adult water polo [39] and volleyball players [40]. Greater amounts of dynamic upper limb activity have also been shown to reduce the risk of female adolescent NSP [8,36] and so these separate findings may explain the overall association of reduced flexibility and lower risk of NSP observed in this study.Motor competenceMales with higher levels of the motor competence factor of muscle power had a reduced risk of diagnosed NSP, and there was a trend for higher overall motor competence (NDI) to be associated with lower risk of NSP ever in males after multivariate analysis. This was expected, given that higher motor competence might have a protective effective on the musculoskeletal system [41]. However this relationship may be weakened by males with better motor competence being more involved in vigorous activities, as suggested by evidence that pre-pubescent children with higher motor competence engage in more vigorous play [42], and thus more likely to develop NSP [37]. This potential confounding may possibly explain the contradictory finding of lower persistent control being associated with a lower risk of diagnosed NSP. In contrast, poorer coordination may be a result of reduced motor practice as part of a reduction of activities associated with NSP.In females, higher bimanual dexterity significantly increased risk of NSP in the past month. Bimanual dexterity relates to the co-ordination of fine motor skills across both arms, and might be developed by activities such as playing musical instruments, needlework, computer work or craftwork, which are known risk factors for female adolescent NSP [36].Strength of associationsEvidence from longitudinal studies [14,16] demonstrates that physical performance in adolescence can influence the development of NSP in adulthood, although a deconditioning response to the presence of NSP is also possible. The predictive utility of the models in the current study was very low however, with Nagelkerke R2 ranging from only 0.001 to 0.085. The lack of stronger relationships was not due to missed curvilinear relationships as these were accounted for in the analysis, and the study was not underpowered as weak relationships were detected.This may indicate that either physical performance is not strongly related to NSP during adolescence or that the direction and mechanisms are more complex and other factors need to be considered. One of the strengths of our study was the broad range of physical variables adjusted for in the analyses, but certain possible confounders such as activity levels and sport participation were not included in the current analyses. Consideration of these may have either reduced or strengthened observed relationships, and should be attempted in further work. In addition, NSP was treated as a homogenous entity, but in reality it may have several sub-groups with different aetiologies. Real, but differing, associations between physical performance and each sub-group may thus have been lost in the current analysis. Further work towards subgroup identification is intended.NSP is not a simple construct and thus four measures were used, including a parental report of health professional diagnosed NSP. Whilst parental report of diagnosed neck pain has limited detail and accuracy, it reinforced the self-report measure of NSP. Further, strong relationships could be expected to be more consistent across the different measures.Associations were very different across genders, with no common effects seen. These differing gender effects may be the result of differences in the type and vigour of sporting activities [43], as well as anthropometric differences, and possible variation in underlying pain mechanisms and psychosocial effects. Whatever the cause, these differences emphasise the need to continue to consider gender in future work, as gender will be a possible confounder of many pain/physical characteristics relationships.ConclusionNSP is clearly an important health issue for adolescents. Some aspects of physical performance are associated with adolescent NSP. Interestingly, better performance sometimes increased rather than decreased risk, suggesting that the direction and mechanisms are complex. Associations differed between genders, suggesting that NSP in males and females may have different mechanisms, or that these factors may interact differently. However, despite the large sample and examination of curvilinear relationships, multiple physical factors and gender specific effects, the associations were weak suggesting complex mechanisms for NSP development.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsMCP analysed data, drafted the manuscript and assisted with final approval. LMS designed the study, designed and revised the manuscript and assisted with final approval. PBO designed the study, revised the manuscript and assisted with final approval. AJS analysed data, revised the manuscript and assisted with final approval. BH designed the study, revised the manuscript and assisted with final approval.Pre-publication historyThe pre-publication history for this paper can be accessed here:\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2531131.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2531131",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2531131\nAUTHORS: Andrea Matucci, Paola Rossolillo, Miriam Baroni, Antonio G Siccardi, Alberto Beretta, Donato Zipeto\n\nABSTRACT:\nBackgroundA recently identified genetic polymorphism located in the 5' region of the HLA-C gene is associated with individual variations in HIV-1 viral load and with differences in HLA-C expression levels. HLA-C has the potential to restrict HIV-1 by presenting epitopes to cytotoxic T cells but it is also a potent inhibitor of NK cells. In addition, HLA-C molecules incorporated within the HIV-1 envelope have been shown to bind to the envelope glycoprotein gp120 and enhance viral infectivity. We investigated this last property in cell fusion assays where the expression of HLA-C was silenced by small interfering RNA sequences. Syncytia formation was analyzed by co-cultivating cell lines expressing HIV-1 gp120/gp41 from different laboratory and primary isolates with target cells expressing different HIV-1 co-receptors. Virus infectivity was analyzed using pseudoviruses. Molecular complexes generated during cell fusion (fusion complexes) were purified and analyzed for their HLA-C content.ResultsHLA-C positive cells co-expressing HIV-1 gp120/gp41 fused more rapidly and produced larger syncytia than HLA-C negative cells. Transient transfection of gp120/gp41 from different primary isolates in HLA-C positive cells resulted in a significant cell fusion increase. Fusion efficiency was reduced in HLA-C silenced cells compared to non-silenced cells when co-cultivated with different target cell lines expressing HIV-1 co-receptors. Similarly, pseudoviruses produced from HLA-C silenced cells were significantly less infectious. HLA-C was co-purified with gp120 from cells before and after fusion and was associated with the fusion complex.ConclusionVirionic HLA-C molecules associate to Env and increase the infectivity of both R5 and X4 viruses. Genetic polymorphisms associated to variations in HLA-C expression levels may therefore influence the individual viral set point not only by means of a regulation of the virus-specific immune response but also via a direct effect on the virus replicative capacity. These findings have implications for the understanding of the HIV-1 entry mechanism and of the role of Env conformational modifications induced by virion-associated host proteins.\n\nBODY:\nBackgroundA whole-genome association study of major genetic determinants for host control of HIV-1 has identified two polymorphisms that explain nearly 15% of the variation among individuals in viral load during the asymptomatic set-point period of infection. One of these polymorphisms is located in the 5' region of the HLA-C gene, 35 kb away from transcription initiation and has been reported to be associated with differences in HLA-C expression levels [1]. As a classical MHC class I gene, HLA-C has the potential to restrict HIV-1 by presenting epitopes to cytotoxic T cells (CTLs) [2,3], resulting in the destruction of infected cells. However, the potential ability of HLA-C to present epitopes to CTLs is severely limited by its poor expression at the cell surface (10-fold lower than either HLA-A or -B) [4] and its tendency to accumulate as free heavy chains or heavy chains associated with β2-microglobulin but free of peptides as a result of poor assembly [5]. HLA-C has also the least diversity of the three classical MHC class I loci. Accordingly, an analysis of the class I restricted CD8+T cell responses against HIV-1 revealed that variation in viral set-point and absolute T cell count is strongly associated with particular HLA-B, but not HLA-A or HLA-C allele expression [6]. In addition, HLA-Cw4/+ heterozygosity promotes rapid progression to AIDS illness, as does HLA-Cw4/Cw4 homozygosity [7]. Interestingly, the virus has evolved a strategy to selectively down-regulate HLA-A and -B but not HLA-C, via the regulatory protein Nef [8]. The immunity of HLA-C to Nef-mediated down modulation confers to the virus the capacity to escape NK cell attack since HLA-C is a dominant inhibitory ligand of NK cells [9]. Thus, the overall trade-off of high HLA-C expression might be favourable to the virus, and not to the host. The relative importance of CTLs and NK cells in vivo is still unclear and the interpretation of genetic studies showing association to viral set-point is particularly complex.Like other MHC class I and II molecules, HLA-C is selectively incorporated into the HIV-1 envelope [10,11]. A study previously reported by our group [12] demonstrated that virion-associated HLA-C molecules have a profound influence on the infectivity of HIV-1. MHC class I negative cell lines were non permissive for the replication of primary HIV-1 isolates and only partially permissive for the replication of T cell line adapted viruses. Transfection of HLA-Cw4 into these cell lines restored their capacity to support viral replication. The increased infectivity of viruses grown in the presence of HLA-Cw4 was associated with changes in viral envelope protein conformation, which included an enhanced expression of epitopes not normally exposed upon CD4 binding.Here we further investigate this phenomenon in a different experimental system where the expression of HLA-C was selectively silenced by small interfering RNA sequences (siRNA) and the infectivity-enhancement effect evaluated in fusion assays with cells expressing CCR5 and/or CXCR4 co-receptors. To overcome unknown effects of other viral gene products on viral infectivity, pseudotyped viruses expressing the same viral genome backbone, but different env, were used. The association of HLA-C with Env was tested using our previously reported technique for the detection of molecular complexes formed at the surface of cells during the fusion process (fusion complexes) [13].ResultsEffects of HLA-C on the HIV-driven fusion processTo assess the role of HLA-C in the fusion process we used a cell fusion assay between CHO cells expressing gp120/gp41, either alone or in combination with HLA-C and CHO cells expressing CD4-CCR5 (Table 1) [13]. When CHO-gp120-HLA-C cells were co-cultivated with CHO-CD4-CCR5 cells, a dramatic increase (p < 0.05) in the number and size of syncytia, as compared to those obtained with the same cells not expressing HLA-C, was observed (Fig. 1A). The increased fusion efficiency was not due to a higher expression level of gp120/gp41 in CHO-gp120-HLA-C cells, since they express on average 27% less gp120/gp41 than CHO-gp120/gp41 cells, when analyzed in ELISA using HIV-1 positive human sera (Fig. 1B).Figure 1Fusion efficiency of CHO cells expressing HLA-C and HIV-1 Env. Panel A: Syncytia formation after co-cultivation of effector CHO cells expressing gp120/gp41 and HLA-C, or CHO cells expressing only gp120/gp41, with target CHO-CD4-CCR5 cells. The number and the extent of syncytia is significantly higher (p < 0.05) when effector cells express HLA-C. Panel B: ELISA analysis of Env expression. CHO, negative control; CHO-gp120, cells stably expressing the Env gene of the R5 tropic HIV-1 isolate 91US005; CHO-gp120-HLA-C: CHO-gp120 cells stably expressing HLA-Cw4; gp120: positive control, consisting of a mixture of five different gp120s. The higher fusion efficiency of CHO-gp120-HLA-C cells is not due to an increased level of Env expression, since they express 27% less gp120 than CHO-gp120 cells.Table 1Summary of the HIV-1 envelopes tested in the different experimental models.Experimental modelHost cell (HLA-C allele)Env (tropism/subtype)HLA-C siRNA silencing and cell fusion assaysHeLa (Cw12)ADA (R5/B)LAI (X4/B)NDK (X4/D)gp120/gp41 transient transfection and cell fusion assaysCHO-HLA-C (Cw4)93MW965 (R5/C)91US005 (R5/B)92UG024 (X4/D)NDK (X4/D)J500 (X4/B)Pseudovirus transductions293T (Cw7)pRHPA4259.7 (R5/B)6535.3 (R5/B)NDK (X4/D)m7NDK (X4/D)HLA-C silencing was conducted on human cells (HeLa-derived) physiologically expressing HLA-C and stably expressing Env of different strains (ADA, LAI, NDK).Transient transfections experiments with plasmids encoding different Envs were conducted on non-human CHO cells stably expressing HLA-C to directly compare the effect of HLA-C in the absence of other human MHC class I molecules.Pseudoviruses were produced in HLA-C silenced 293T cells since this human cell line is the election host for efficient and quantitative production of pseudotyped virus particle. The Envs tested belong to a standard reference panel (NIBSC EVA CFAR ARP2066) except NDK.Similar results were obtained in a different cell fusion assay where CHO and CHO-HLA-C cells, transiently transfected with gp120/gp41 from different primary and laboratory HIV-1 isolates, were fused with TZM-bl cells and fusion quantified by luciferase transactivation. All gp120/gp41 tested (93MW965, 91US005, 92UG024) showed higher fusion efficiency when co-cultivated with TZM-bl cells if co-expressed with human HLA-C (Fig. 2). Only two X4-tropic isolates (J500 and NDK) failed to show a statistically significant fusion increase.Figure 2Transient transfections of CHO cells expressing human HLA-C with different env sequences. CHO (-, grey bars) and CHO-HLA-C (+, black bars) cells transiently transfected with plasmids encoding Tat, Rev and Env from different primary and laboratory HIV-1 isolates and co-cultivated for 6 hours with TZM-bl target cells. After Tat driven transactivation of firefly luciferase expression, fusion efficiency was quantified and expressed as counts per second (CPS). Each value represents the average of four replicates. The gp120/gp41 of primary isolates 93MW965 (R5), 91US005 (R5) and 92UG024 (X4) are HLA-C sensitive (p < 0.05) while isolates J500 (X4) and NDK (X4) are less sensitive to the presence of HLA-C (p not significant).HLA-C silencing of cells expressing gp120/gp41HeLa cells constitutively express HLA-C and HLA-A and, at lower levels, HLA-B [14]. Various HeLa-derived cell lines, constitutively expressing HIV-1 Env, were silenced by HLA-C specific siRNAs (Table 1). The expression of gp120 in HeLa-ADA, -LAI and -NDK, as well as that of β2-microglobulin and GAPDH genes was not affected.There was no unwanted off-target silencing of non HLA-C genes (Fig. 3A). The expression of HLA-C protein on HeLa-ADA and 293T cells was undetectable at 72 hours from siRNA transfection (Fig. 3B). Fusion efficiency, determined by counting the number of syncytia formed, was significantly lower (p < 0.01) when HLA-C silenced cells expressing HIV-1 gp120/gp41 of the LAI strain were co-cultivated with HeLa P4.2 cells as target cells (Fig. 4). Fusion efficiency of HeLa-NDK cells was less affected by HLA-C silencing, confirming that the NDK gp120/gp41 has a lower sensitivity to the presence of HLA-C [12]. When silencing was performed with siRNAs specific for HLA-C or with a pool of siRNAs silencing also HLA-A and -B, similar levels of reduction in fusion efficiency were observed.Figure 3Specific silencing of HLA-C in human cell lines. Panel A: off-target effect analysis by RT-PCR in HLA-C silenced (+) and non-silenced (-) HeLa cells expressing HIV-1 gp120/gp41 (ADA). PCR was performed with primers specific for HLA (A, B, C), gp120, β2-microglobulin and GAPDH. M: molecular weight marker. No off-target effect due to HLA-C mRNA silencing is affecting the mRNA levels of the other MHC class I genes, as well as β2-microglobulin, HIV-1 gp120 or the housekeeping control gene GAPDH. Panel B: western-blot analysis of HLA-C protein expression. After 72 hours from siRNAs transfection, HLA-C is undetectable both in HeLa-ADA and in 293T cells.Figure 4Cell fusion of HLA-C silenced HeLa-Env cells with HeLa-P4.2 target cells. Analysis of syncytia formation by co-cultivating HLA-C silenced (+) and non-silenced (-) HeLa-LAI and HeLa-NDK cells with target HeLa-P4.2 cells, expressing CD4 and CXCR4. The number of syncytia formed is lower (p < 0.01) using HLA-C silenced HeLa-LAI cells. Fusion efficiency of HeLa-NDK cells is not significantly affected by HLA-C silencing.Syncytia formation using CCR5 or CXCR4 co-receptorsTo test the role of HLA-C in the fusion process with cells expressing CCR5 or CXCR4 co-receptors, we measured the fusion index in co-cultures of HeLa-ADA and 3T3.T4.CCR5 cells or HeLa-LAI and 3T3.T4.CXCR4 cells with or without siRNA silencing of HLA-C. In both cultures, the fusion index was significantly lower (p < 0.01) in HLA-C-silenced cells than in the corresponding non-silenced controls (Fig. 5) showing that HLA-C increases the fusion efficiency of both CCR5 and CXCR4 tropic viruses.Figure 5Comparison of the fusion efficiency of HLA-C silenced HeLa-Env cells with 3T3.T4.CCR5 and 3T3.T4.CXCR4 cells. HLA-C silenced (+, grey bars) and non-silenced (-, black bars) HeLa cells expressing gp120/gp41 of different HIV-1 isolates (ADA, LAI, NDK) co-cultivated with NIH 3T3.T4.CXCR4 and NIH 3T3.T4.CCR5 cells. Fusion efficiency of X4 tropic gp120 LAI is significantly lower (p < 0.01) in HLA-C silenced cells when fusing with CXCR4 target cells. Similarly, fusion efficiency of the R5 tropic gp120 ADA is lower (p < 0.01) in HLA-C silenced cells when fusing with CCR5 target cells. The fusion of ADA gp120 in HLA-C silenced cells with cells expressing CXCR4 is significantly (p < 0.01) less efficient, while that of LAI gp120 with cells expressing CCR5 is similar, irrespective of HLA-C silencing. The NDK gp120 is HLA-C insensitive, when using either the CXCR4 or the CCR5 co-receptor.3T3.T4.CXCR4 cells express 2–3 times more CXCR4 than HeLa-P4.2 and TZM-bl cells. Similarly, 3T3.T4.CCR5 cells express about 10 times more CCR5 as compared to TZM-bl cells (data not shown). We observed that these cells allowed the fusion with cells expressing Envs with a different co-receptor tropism, although at lower level. The use of the heterologous co-receptor, already evident [15] using pseudotyped viruses, is increased in fusion assays with Env-expressing cell lines, in particular for longer co-cultivation times. Under these experimental conditions, we investigated the role of HLA-C in modulating fusion efficiency in the presence of the heterologous co-receptor. We observed that the R5-tropic gp120/gp41 ADA was sensitive to HLA-C presence when fusing with 3T3.T4.CXCR4 cells whereas the X4-tropic LAI was not affected by HLA-C presence when fusing with 3T3.T4.CCR5 cells (Fig. 5). Also in these experiments, the NDK gp120/gp41 was found to fuse with the same efficiency with 3T3.T4.CXCR4 and, at lower levels, with 3T3.T4.CCR5 cells, when using HLA-C silenced or non-silenced HeLa-NDK cells (Fig. 5).Pseudovirus infection assayPseudoviruses produced on normal and HLA-C silenced 293T cells were quantified for p24 content and used in transduction assays (Table 1). Pseudoviruses bearing subtype B 6535.3 and pRHPA4259.7 HIV-1 env genes showed a statistically significant reduction in infectivity when produced in HLA-C silenced 293T cells. Conversely, no significant differences were observed with either NDK subtype D env gene or control virus pseudotyped with the VSV-G protein (Fig. 6A).Figure 6Transduction efficiency of pseudoviruses produced in HLA-C silenced cells. Panel A: luciferase reporter gene assay analysis after transduction with pseudoviruses expressing subtype B HIV-1 env (6535.3 and pRHPA4259.7) or subtype D HIV-1 env (NDK), produced in HLA-C silenced (dashed line, open circles) and non silenced (continuous line, close squares) 293T cells. Each point (expressed as counts per second, CPS) represents average and standard deviation of four replicates. HLA-C sensitive pseudoviruses 6535.3 and pRHPA4259.7 show a significant lower infectivity (p < 0.0001) when produced on HLA-C silenced cells. The NDK pseudovirus as well as a virus pseudotyped with the VSV-G envelope protein, do not show significant differences in infectivity when produced in HLA-C silenced or non silenced 293T cells. Panel B: analysis of the relation between pseudovirus infectious dose and HLA-C sensitivity. 1×, pseudovirus infectious titer giving a luciferase signal (expressed as counts per second, CPS) of 1000 at 16 hours post infection. When the HLA-C insensitive NDK pseudovirus was analyzed at lower infectious titers (0.3× and 0.1×), its infectivity was significantly increased by HLA-C. When the HLA-C sensitive pseudovirus pRHPA4259.7 was analyzed at higher infectious doses (3.3×, 10×), it remained sensitive to HLA-C presence.When the HLA-C insensitive NDK-pseudovirus was used at infectious doses that were 1/3 and 1/10 of the original inoculum, a significant infectivity difference between pseudoviruses produced in HLA-C silenced and non-silenced cells was noted. The HLA-C sensitive pseudovirus pRHPA4259.7 maintained its sensitivity to HLA-C also at lower m.o.i. (1/10 of the original inoculum, data not shown). When the m.o.i. of the pRHPA4259.7 pseudovirus was increased, the infectivity levels of pseudoviruses produced on normal and HLA-C silenced 293T cells was kept significantly different (Fig. 6B).HLA-C/gp120 association on cells before and after fusionIn the previous study we provided evidence of a specific association between virionic HLA-C molecules and gp120 by co-immunoprecipitating the two molecules with the HLA-C-specific monoclonal antibody L31 and a gp120-specific antibody [12]. In this work we looked for additional evidence of HLA-C-gp120 association occurring on cells taken after fusion using a previously described method that allows the isolation of CD4-CCR5-gp120/gp41 fusion complexes after fixation with paraformaldehyde or DTSSP and purification with Galanthus nivalis (GN) lectin, which specifically binds to gp120 [13]. The presence of HLA-C molecules within the fusion complexes could be tested by dot blot with the antibody L31 which also recognizes the denatured protein [16]. Fig. 7 panel A shows a dot-blot with antibody L31 of total cell lysates or proteins eluted from GN lectin columns. L31-reactive molecules were detected in total cell lysates of CHO-HLA-C (lane c) and CHO-gp120-HLA-C cells (lane d) but not in the HLA-C negative CHO cell line (lane a) and the CHO-CD4-CCR5 fusion partner (lane b). The eluate of GN lectin columns loaded with a mixed extract of CHO-gp120-HLA-C and CHO-CD4-CCR5 cells which had been fixed before fusion, displayed a significant amount of L31 reactive molecules (lane g), showing that a specific association between HLA-C and gp120 occurred in cells co-expressing the two molecules, as previously described in LAI-infected 221-Cw4 cells [12]. When the same cells were allowed to fuse before being fixed, the eluate of GN lectin purified cell extract displayed an increased amount of L31-reactive molecules (lane h) indicating that during the process of cell fusion additional HLA-C molecules are recruited within the fusion complexes. The lack of L31-reactive molecules in the eluate of GN lectin purified CHO-HLA-C cells (lane f) demonstrates that in this experimental setting HLA-C molecules are purified via their specific binding to gp120.Figure 7Co-purification of fusion complexes containing HLA-C molecules. Panel A: dot-blot analysis of purified fusion complexes for the presence of HLA-C. Lanes a, b, c and d: cell lysates before purification. Lanes e, f, g and h: cell lysates purified on Galanthus nivalis (GN) lectin columns. Panel B: western blot analysis to detect the presence of HLA-C in purified fusion complexes. Cells were treated with DTSSP, which fixes only proteins present on the cell membrane, and lysates purified on GN lectin columns. PC: positive control (HeLa cells expressing HLA-C); the arrow indicates HLA-C.To gain further evidence of the association between HLA-C and gp120, the same protein samples, after fixation with DTSSP and purification on GN lectin columns, were chemically reduced, separated on SDS-PAGE and blotted with L31 antibody which revealed a 45 kDa band corresponding to the HLA-C heavy chain. Also in this experiment a relatively higher amount of HLA-C was co-purified from cells which were allowed to fuse before fixation (fusion complex), as compared to non-fused cells (no fusion complex) (Fig. 7B). These results provide further evidence that HLA-C is associated to gp120 on the cell membrane and suggest that additional HLA-C is recruited within the fusion complex during cell fusion.Sequence analysis of HLA-insensitive EnvsThe sequence of the env gene of the HLA-C insensitive primary isolate J500 (clade B) was determined. When this was compared to the sequence of the other HLA-C insensitive isolate NDK (clade D), and to the sequences of the HLA-C sensitive Envs tested (93MW965, 91US005, 92UG024, ADA, LAI, 6535.3 and pRHPA4259.7), three identical aminoacid substitutions (N297K, N298Y and I318T, relative to the LAI env sequence) were identified in the V3 loop. Env sequence analysis of the Los Alamos HIV Reference Database showed that the I318T mutation is relatively uncommon, occurring in 92 out of the 1603 env sequences available (5.7%). Mutations N297K and N298Y are extremely rare, occurring only in 2 isolates reported in the database. In addition, the combination of these 3 mutations was found only in a single env sequence (isolate D.TZ.87.87TZ4622). Position 297 is associated with a potential N-glycosilation site [17].DiscussionThis work demonstrates that virion-associated HLA-C molecules, when present on cells expressing gp120/gp41, significantly enhance fusion efficiency and pseudovirus transduction. Our conclusions are supported by the following findings: a) CHO cells co-expressing HIV-1 gp120/gp41 and human HLA-C fuse more rapidly and produce larger syncytia than the original CHO-gp120/gp41 cells from which they are derived; b) transient transfection of gp120/gp41 from different primary isolates in CHO cells co-expressing HLA-C results in a significant increase in fusion; c) silencing of HLA-C in human cell lines expressing HIV-1 gp120/gp41 of R5 and X4 tropic strains, significantly suppresses fusion, d) pseudoviruses produced in HLA-C silenced 293T cells display a significant reduction of infectivity; e) the fusion enhancement property of HLA-C is specific for HIV-1 Env, since a virus pseudotyped with the G envelope protein of VSV is not influenced by the presence of HLA-C.The effect of HLA-C on fusion was observed with both exogenous HLA-C transfected into CHO cells and endogenous HLA-C after its silencing with siRNA in human cells.Some of the data point to the existence of HLA-C \"insensitive\" or \"less-sensitive\" variants since the fusogenic capacity of gp120/gp41 from two isolates, NDK and J500, was not different in HLA-C-silenced and non-silenced cells. However, we observed that HLA-C insensitivity is not an absolute feature, since there was a small difference, although not statistically significant, in the fusion efficiency of NDK and J500 in silenced and non-silenced cells. In addition, a relationship between the infectious dose and the HLA-C sensitivity of pseudoviruses was observed since when infections were performed at low infectivity ratios, the HLA-C insensitive NDK pseudovirus became HLA-C sensitive. Conversely, when high titers of an HLA-C sensitive pseudovirus were used, its infectivity remained dependant on the presence of HLA-C. The relative insensitivity of NDK to the presence of HLA-C could contribute to its reported higher cytopathicity and infectivity [18] and could be the result of a variable infectivity degree of Env [12] or of a lower level of incorporation of HLA-C [10].The comparison of the env sequences of the two unrelated, HLA-C insensitive gp120/gp41s identified, NDK (clade D) and J500 (clade B), with the sequences of the HLA-C sensitive Envs, revealed 3 identical aminoacid substitutions in the V3 loop, which were absent in all other HLA-C sensitive Envs analyzed. This would suggest an involvement of these mutations in the V3 loop in the acquisition of the HLA-C insensitive phenotype. We analyzed an NDK-derived Env mutant, NDKm7 [19], in which the KY mutations in position 297–298 reverted to NN. Virus particles pseudotyped with the NDKm7 env remained HLA-C insensitive as the original NDK env (data not shown), thus excluding the involvement of these mutations in reducing sensitivity to HLA-C presence. It is possible that other mutations, or their combinations, might directly affect the sensitivity to HLA-C by changing the pattern of interaction between HLA-C and gp120, as reported by other authors who studied mutations related to the acquisition of a CD4-independent tropism within gp120 [19,20].The data reported in this study confirm the physical association between HIV-1 gp120/gp41 and HLA-C, that was originally observed in experiments in which HLA-C and gp120 were co-immunoprecipitated from HIV-1 infected cells [12]. HLA-C molecules could be co-purified and detected in fusion complexes in association with gp120/gp41, CD4 and the co-receptor. Such an association may induce conformational changes of gp120 favouring the exposure of cryptic functional epitopes [12]. It has also been recently reported that viral particles carry more HLA molecules than gp120/gp41 trimers [21]. The association between a gp120/gp41 trimer and multiple HLA-C molecules might reduce gp120 shedding, thus keeping more functional the trimeric gp120/gp41 complexes on the viral envelope and resulting in increased fusion efficiency.The increase in fusion and viral infectivity was observed using CHO cells transfected with HLA-Cw4, as well as HeLa cells which express constitutively HLA-Cw12 and pseudoviruses originating from 293T cells which express HLA-Cw7 (Table 1). Similar results were obtained with the HLA-Cw3 allele (L. Lopalco, DIBIT-San Raffaele, Milan, personal communication). Altogether, the Cw3, Cw4, Cw7, and Cw12 serological alleles include members of both groups of the known HLA-C dimorphism [22] and account for almost 80% of all the common HLA-C serotypes. Due to the more limited polymorphism of HLA-C as compared to HLA-A and -B, this limited panel is inclusive enough to allow us to sample all the HLA-C-distinctive substitutions and most of the common allelic variations. Remarkably, most of these cluster around the binding groove, but the co-immunoprecipitation of env with HLA-C [12] was observed by immunoprecipitating the complex with antibody L31, that binds on the alpha 1 domain alpha helix, e. g. in proximity to the sites at which essentially all the polymorphic HLA-C positions cluster. This suggests that HLA-C polymorphism is unlikely to influence this association, and that the residues important for co-immunoprecipitation reside within the relatively invariant HLA-C backbone. In line with this finding, we have observed the infectivity-enhancement effect with all the alleles tested so far, suggesting that most HLA-C alleles bind Env. We cannot however exclude the possibility that some HLA-C allelic variants may be more efficient than others in binding Env and enhancing viral infectivity.An implication of these findings is that HLA-C may be selectively involved in protective immunity. A protective effect was observed in HIV serodiscordant couples with unmatched HLA-C alleles [23] and anti-HLA antibodies are frequent in exposed, but seronegative subjects [24,25]. It has also been reported that MHC class I concordance is associated with an increased risk of mother to child HIV-1 transmission [26,27]. Since early studies in primates were suggestive of anti-MHC antibodies being protective [28], the possibility of using HLA molecules for a HIV-1 vaccine has long been debated [29,30]. Our data point to an association between HLA-C and Env in mature virions which may induce the expression of critical conformational epitopes [12]. Since the few Env that showed lower sensitivity to HLA-C are X4 tropic, the inclusion of HLA-C in new immunogenic formulations may help eliciting broadly neutralizing antibodies that would be important for the in vivo host control of R5 tropic strains of HIV-1.ConclusionHLA-C influences viral replication by at least three distinct and opposite mechanisms: induction of cytotoxic T cells (suppression), inhibition of NK cells (enhancement) and enhancement of virus infectivity. This last effect is associated to a specific association of virionic HLA-C molecules to Env. The immunity of HLA-C to the Nef-induced down-regulation confers to the virus not only the capacity to escape NK cells control but also a higher replicative capacity suggesting that high HLA-C expression is advantageous to the virus and not the host.MethodsAntibodiesW6/32 is a mouse monoclonal antibody specific for HLA-A, -B and -C trimeric complex [31]. The L31 monoclonal antibody is specific for the α domain of HLA-C heavy chain [32-34], not associated to β2-microglobulin. Anti-gp120 human sera from HIV-positive patients were kindly provided by Dr. Lucia Lopalco, DIBIT-HSR, Milan, Italy. IgG were purified using Protein G Sepharose 4 Fast Flow (GE Healthcare Lifescience, Chalfont St. Giles, UK) following manufacturer's instructions.CellsHeLa (HLA-Cw12, [35]) and HEK-293T (HLA-Cw07, [35]) cells were obtained from the American Type Culture Collection (ATCC).HeLa-derived effector cell lines expressing the HIV-1 env gene of strains ADA, LAI [36] and NDK [37] and the indicator target cell line HeLa P4.2 [38] were kindly provided by Dr. Mark Alizon and Dr. Uriel Hazan, Institut Cochin, Paris, France.NIH 3T3 cells expressing the HIV-1 receptor CD4 and the chemokine receptor CCR5 (3T3.T4.CCR5) or CXCR4 (3T3.T4.CXCR4) were obtained from the NIH AIDS Research & Reference Reagent Program, division of AIDS, NIAID, Dr. Dan R. Littman [15].The TZM-bl cell line [39] was from the EU programme EVA/MRC, CFAR NIBSC, UK. This cell line expresses CD4, CCR5 and CXCR4 and contains HIV-1 LTR-driven E. coli β-galactosidase and firefly luciferase reporter cassette that are activated by HIV-1 Tat expression.CHO and CHO-gp120/gp41 [13] cells were stably transfected with the vector pZeoSV2(+) (Invitrogen, Carlsbad, CA, USA) bearing the HLA-Cw4 gene, and the cell lines obtained were named CHO-HLA-C and CHO-gp120-HLA-C, respectively.CHO and CHO-HLA-C cell lines were transiently transfected with HIV-1 env genes from primary and laboratory isolates NDK, J500 (a primary X4 tropic isolate [40]), 92UG024, 93MW965 and 91US005 [41] cloned in the expression vector pCDNA3.1 (Invitrogen, Carlsbad, CA, USA).RNA silencing of HLA-CThe HLA-C mRNA [GenBank: NM_002117] target sites for siRNA were determined by using the Dharmacon siGENOME software and synthesized by Dharmacon (Lafayette, CO, USA). The siRNAs targeted different regions of the HLA-C mRNA.In particular, siRNAs J-017513-06 (5'P-UAAUCCAUCAACGCUUCAUUU-3') and J-017513-08 (5'P-UUUGGAAGGUUCUCAGGUCUU-3') were found to be specific for HLA-C silencing, while siRNAs J-017513-05 (5'P-AUAGCGGUGACCACAGCUCUU-3') and J-017513-07 (5'P-ACUUCUAGGAAUUGACUUAUU-3') also silenced HLA-A and -B mRNAs.HeLa cells expressing env genes were transfected with 100 nmol/well of siRNA following manufacturer's instructions, using DharmaFECT 1 reagent (Dharmacon, Lafayette, CO, USA). The silencing of HLA-C protein expression was verified by Western blot after 72 hours.The absence of off-target effects was verified both by RT-PCR of HLA-A, -B, -C, β2-microglobulin, HIV-1 env and GAPDH, and by ELISA analysis of gp120/gp41 expression using HIV-1 positive human sera.TZM-bl reporter gene assaysThe fusion process between gp120/gp41 effector cells (HeLa-ADA, HeLa-LAI, HeLa-NDK) and TZM-bl cells was assessed by measuring luciferase activity and by X-gal cell staining.TZM-bl cells (50.000 per well) were plated in 96 microtiter wells (Corning, NY, USA) to an equivalent number of effectors cells for 3 to 6 hours at 37°C. The luciferase activity resulting from fusion and transactivation was analyzed using the Brite Lite reagent following manufacturer's instructions and quantified by using a Victor 3 apparatus (Perkin Elmer, Waltham, MA, USA). All the assays were performed in triplicate.In situ staining of fusing cells for β-galactosidase gene activation was performed in a 24-well plate format (Corning Life Sciences, Lowell, MA, USA) as reported [42]. Blue-stained syncytia were photographed using a Nikon Eclipse 80 i microscope, counted and fusion efficiency determined by calculating the fusion index [43].Cell fusion assaysNIH 3T3.T4.CCR5 and 3T3.T4.CXCR4 were stained with the fluorescent lipophilic dye Vybrant DiI (Invitrogen, Carlsbad, CA, USA) following manufacturer's instructions. Cells were plated at 400,000 per well on a six-well plate (Corning Life Sciences, Lowell, MA, USA) and, 72 hours post siRNA transfection, co-cultivated at 1:1 ratio with HLA-C silenced and non-silenced HeLa-gp120/gp41 cells labeled with the fluorescent lipophilic dye Vybrant DiO (Invitrogen). After 6 hours, syncytia formation was analyzed using a fluorescence microscope (Nikon Eclipse 80i) for green and red fluorescence and the double positive yellow syncytia counted [44,45].gp120 ELISA detection assayNinety-six well plates (Nunc, Roskilde, Denmark) were coated with 50 μl/well of a solution of 2 μg/ml of the D7324 gp120 antibody (Aalto Bioreagents, Dublin, Ireland), and a 3 mg/ml solution of total protein from cell lysate samples was added as described [13]. Positive controls consisted of a 100 ng/ml pool of 5 different gp120s obtained from EVA/MRC Centralised Facility for AIDS Reagents, NIBSC, UK (CN54, IIIB, MN, SF2 and W61D). Plates were washed and incubated with 1:200 diluted purified human IgG from sera of HIV-1 positive patients (25 mg/ml), washed and incubated with 1:500 diluted goat anti-human horseradish peroxidase conjugate (BioRad). Optical signal was developed with SigmaFast OPD solution (Sigma, St. Louis, MO, USA).RT-PCRTotal RNA was extracted from 24 hours silenced and non-silenced cultured cells using the RNeasy Plus mini kit (Qiagen, Germantown, MD, USA) and treated with RNase-free DNase I (Sigma). Reverse transcription (RT) of 1 μg of total RNA was performed using the Quantitect Reverse Transcription kit (Qiagen) and random primers. For PCR amplification of HLA-A, -B and -C, primers and conditions were used as previously reported [14]. Primers used to amplify HIV-1 env gene were: 5'-GGGCCACACATGCCTGTGTA-3' forward and 5'-CTAATTCCATGTGTACATTGTACTGTG-3' reverse; for β2-microglobulin amplification 5'-GATGAGTATGCCTGCCGTGTG-3' forward and 5'-CAATCCAAATGCGGCATCT-3' reverse; for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) amplification 5'-GCATCCTGGGCTACACTGA-3' forward and 5'-TGACAAAGTGGTCGTTGAGG-3' reverse. PCR was performed for 32 cycles at 94°C, 60°C and 72°C for 1 min in each step. PCR products were analyzed on a 1% agarose gel and stained with Sybr Safe (Molecular Probes, Eugene, OR, USA). Images were acquired with an AutoChemi System apparatus (UVP, Cambridge, UK). Controls for genomic DNA contaminations consisted in RT reactions in which the polymerase was omitted.Western blot analysisSeventy-two hours after HLA-C siRNA transfection, cells were lysed, the total protein content of supernantant was measured using a colorimetric assay (DC protein assay, BioRad, Hercules, CA, USA) and used for western blot analysis.Equal amounts (30 μg/lane) of cell lysates were separated on 3 to 8% NuPAGE Tris-acetate acrylamide gels (Invitrogen, Carlsbad, CA, USA) and transferred onto polyvinylidene difluoride membranes (GE Healthcare Lifescience, Chalfont St. Giles, UK). Membranes were blocked in a Tris-buffered saline solution containing 5% non-fat dry milk and incubated with the L31 monoclonal antibody (1:200 dilution). Anti-mouse horseradish peroxidase-conjugated antibody (Dako, Carpinteria, CA, USA) was used as secondary antibody at 1:2,000 dilution and immunoreactive bands were visualized with the Opti-4CN detection kit (BioRad, Hercules, CA, USA).FACS analysisCells were analyzed by immunofluorescent staining and cytofluorimetry on a FACScanto apparatus (Becton Dickinson, San Jose, CA, USA). After incubating 500,000 cells with the primary anti-HLA monoclonal antibodies W6/32 or L31, these were washed and incubated with a 1:200 dilution of the goat-anti mouse IgG-FITC secondary antibody (Becton Dickinson). The analysis was conducted using the FACSDiva software (Becton Dickinson). For L31 epitope unmasking through β2-microglobulin stripping, cells were pre-treated with acidified medium as described [46] and immediately analysed.Infectivity of pseudoviruses produced on HLA-C silenced cellsHLA-C mRNA was silenced in 293T cells as previously described for HeLa cells. After 24 hours, silenced and non-silenced 293T cells were co-transfected with backbone (pSGΔenv) and env plasmids (subtype B isolates 6535.3 and pRHPA4259.7, subtype D isolate NDK, and Vescicular Stomatitis Virus (VSV) envelope protein G), as described [47]. Pseudoviruses were collected after 48 hours and quantified for p24 content using a standard ELISA Kit following manufacturer's instructions (Innotest-HIV antigen mAb, Innogenetics, Gent, Belgium). Both normal and HLA-C silenced pseudoviruses were used at a p24 concentration of 150 pg/ml. Infections of TZM-bl cells were done in quadruplicate and luminescence measured after 4, 8, 24 and 48 hours of incubation using a Victor 3 luminometer (Perkin Elmer) as previously described.Fusion complex analysisFusion complexes were fixed with paraformaldehyde, purified and analyzed as described [13]. Briefly, CHO-gp120-HLA-C and CHO-CD4-CCR5 cells were co-cultivated 4 hours at 37°C, fixed and lysed. Cell lysates were passed over a snowdrop Galanthus nivalis lectin column and eluted in 1 M methyl α-D-mannopyranoside (Sigma). Fusing cells were also fixed with DTSSP (Pierce Biotechnology, Rockford, IL, USA), following manufacturer's instructions. Fusion complexes were purified and dissociated using 5% β-mercaptoethanol in SDS-PAGE sample buffer. Effector and target cells were also separately fixed prior purification. Paraformaldehyde fixed fusion complexes were analysed for HLA-C co-purification by dot-blot and DTSSP fixed complexes by Western blot with HLA-C specific mAb L31.Statistical analysisData were analyzed by ANOVA and unpaired Student's t-test with Welch's correction, using the software GraphPad Prism 4.0c (GraphPad Software, Inc., CA, USA).Sequence analysisHIV-1 Env sequences (NDK [GenBank: A34828], LAI [GenBank: AF004394]; ADA [GenBank: AY426119]; 92UG024 [GenBank: U43386]; 93MW965 [GenBank: U08455]: 91US005 [GenBank: U27434]) were aligned and compared using CLC Sequence Viewer 4.6.2, developed by CLC bio A/S  for Apple Mac OSX.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsAM carried out siRNA silencing, RT-PCR, cell transfections, ELISA, Western blot and FACS analysis, cellular fusions and pseudovirus infections experiments. PR carried out sequencing, pseudovirus preparation and titration and fusion complexes preparation and analysis. MB isolated and cloned the HLA-C insensitive env sequence from the J500 primary isolate. AGS participated in the design and coordination of the study and drafted the manuscript. AB participated to study design, data analysis and gave a significant contribution in drafting and revising the manuscript. DZ produced Env-coding plasmids and stably transfected cell lines, did fusion complexes preparation and analysis, conceived the study and carried out its design, and, as corresponding author, carried out the drafting of the manuscript. All authors read and approved the final manuscript.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2531182.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2531182",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2531182\nAUTHORS: Jin-Yan Wang, Han-Ti Zhang, Jing-Yu Chang, Donald J Woodward, Luiz A Baccalá, Fei Luo\n\nABSTRACT:\nBackgroundExpectation is a very potent pain modulator in both humans and animals. There is evidence that pain transmission neurons are modulated by expectation preceding painful stimuli. Nonetheless, few studies have examined the influence of pain expectation on the pain-related neuronal activity and the functional connectivity within the central nociceptive network.ResultsThis study used a tone-laser conditioning paradigm to establish the pain expectation in rats, and simultaneously recorded the anterior cingulate cortex (ACC), the medial dorsal thalamus (MD), and the primary somatosensory cortex (SI) to investigate the effect of pain expectation on laser-induced neuronal responses. Cross-correlation and partial directed coherence analysis were used to determine the functional interactions within and between the recorded areas during nociceptive transmission. The results showed that under anticipation condition, the neuronal activity to the auditory cue was significantly increased in the ACC area, whereas those to actual noxious stimuli were enhanced in all the recorded areas. Furthermore, neuronal correlations within and between these areas were significantly increased under conditions of expectation compared to those under non-expectation conditions, indicating an enhanced synchronization of neural activity within the pain network. In addition, information flow from the medial (ACC and MD) to the lateral (SI cortex) pain pathway increased, suggesting that the emotion-related neural circuits may modulate the neuronal activity in the somatosensory pathway during nociceptive transmission.ConclusionThese results demonstrate that the nociceptive processing in both medial and lateral pain systems is modulated by the expectation of pain.\n\nBODY:\nBackgroundPain is a personal and subjective experience. The psychological factors therefore play important roles in shaping pain perception. One of these factors is expectation. In clinical situations, when pain is anticipated, patients often report the worsening of pain [1-3]. Conversely, expectation of pain relief is considered to be effective means for producing placebo analgesia [4-6]. Recently, Keltner et al. demonstrated that the level of expected pain intensity significantly altered perceived pain when the comparison was made between two noxious thermal stimuli of almost equal intensity [7].Interests in pain anticipation-related brain activity has increased in recent years. EEG recording revealed significantly enhanced signals during anticipation of painful stimuli or priming with pain-related adjectives [8,9]. Functional imaging studies suggested that expectation of pain could cause alteration in both pain perception and forebrain pain transmission [10-12], even amplify brain responses to nonpainful somatosensory stimulation [13]. More interestingly, the pain anticipation-related areas are largely overlapped with pain-related areas, such as the primary somatosensory cortex (SI), anterior cingulate cortex (ACC), periaqueductal grey (PAG), insular cortex (IC), prefrontal cortex (PFC), and cerebellum [10-12,14,15].Although expectation of pain has been extensively studied in humans using neuroimaging techniques, neural mechanisms underlying the modulating effect of expectation are far from clear. Functional imaging studies are able to detect expectation-related signal changes, but can not directly measure neuronal spike activity. In fact, the correlation between imaging signals and action potential firing of neurons is still unclear [16]. Furthermore, activation of brain sites revealed by imaging studies can not resolve issues as to how the information is transferred among the brain regions. The issues can only be addressed in animal experiments [17].We reported previously that the medial and lateral pain pathways were activated in parallel manner by cutaneous noxious radiant heat in awake rats, using a multiple-channel single-unit recording method [18]. In that study, we incidentally observed anticipatory responses in the medial pain system, including the ACC and medial dorsal thalamus (MD). To further explore this issue, we investigated the effect of pain anticipation on nociceptive behavior and neural activity in awake rats. To establish anticipation in rats, we employed a Pavlovian conditioning paradigm, in which a neutral conditioned stimulus (tone) was paired with a noxious unconditioned stimulus (laser). The establishment of pain expectation was determined by the acquisition of conditioned responses (tone-induced avoidance). The aim of this study was to determine whether and how pain expectation could alter nociceptive processes and functional connectivity in the nociceptive neural networks.ResultsBehavioral responseRats responded to auditory stimuli with high level exploratory activity in the early stage of Session 1 as evidenced by rearing, head movement, and short-lasting freezing. After repeated tone presentation, these exploring behaviors substantially subsided, as Fig. 1A (Session 1) illustrated. By contrast, noxious laser stimulation caused marked escaping responses such as foot jumping, lifting and licking. Such nociceptive behaviors always occurred throughout the first session (Fig. 1B).Figure 1Tone or laser-elicited behavior in the first two sessions. (A) The learning effect demonstrated by tone alone-elicited behavior. The behavioral score was accumulated every 5 successive trials. One-way ANOVA followed by the Dunnett test for multiple comparisons were used to compare the difference of behavioral scores between Session 2 and 1 (* p < 0.05). As can be seen, rats learned to escape immediately after the tone after about 25 tone-laser pairing trials. (B) Laser-induced nociceptive behavior in the first session. (C) The acquisition and extinction of the conditioned response demonstrated by tone alone-elicited behavior. *, *** p < 0.05, p < 0.001, respectively, compared with \"Baseline\"; ###, p < 0.001, compared with \"Trained\", one-way ANOVA followed by the Newman-Keuls Multiple Comparison Test. \"Baseline\" and \"Trained\" are the averaged behavioral scores in the first and second sessions (trials 1–30 in Session 1 and trials 51–55 in Session 2, see above Fig. 1A), respectively.In the second session, rats received tone-laser conditioning training. As can be seen in Fig. 1A (Session 2), rats gradually learned to escape after tone but prior to the delivery of noxious stimulation. Following 25 tone-laser pairing trials, the auditory cue became a reliable predictor for the forthcoming painful stimuli. In the testing phase, it was observed that the tone alone was able to elicited escaping behavior (Fig. 1C), indicating the acquisition of the conditioned response. Interestingly, at the same time when rats are sensitive to the warning signal, the nociceptive behavior induced by actual pain stimulation was significantly reduced compared to Session 1 (9.18 ± 1.05 vs. 12.92 ± 0.16, p < 0.05).Laser- and tone-induced neuronal activityA total of 216 – 224 single units were simultaneously recorded (72 – 73 from the ACC, 61 – 64 MD, and 83 – 87 SI, varied between sessions due to neuron drifting). Noxious laser induced predominantly excitatory responses, displayed in sharp or sustained manner, as shown in Fig. 2A. The neurons exhibiting excitatory responses accounted for 32%, 51%, and 60% in ACC, MD thalamus, and SI cortex, respectively. Inhibitory neuronal responses were occasionally encountered and less than 5%. Auditory stimuli also elicited discharge of a small proportion of neuron within the recorded areas, with 7% in ACC, 16% in MD, and 11% in SI, as shown in Fig. 2B. Fig. 2C illustrated the typical neuronal response produced by tone-laser pairing.Figure 2The typical neuronal response elicited by simple laser (A), simple tone (B), and paired tone and laser (C). PSTHs illustrated the average firing rate of a neuron around a stimulus. Time = 0 on the x-axis corresponded to the time of noxious (A, C) or tone (B) stimulus onset.We compared the laser-induced neuronal responses during post-stimulus time across the three sessions. As shown in Fig. 3A, a significant difference was detected across sessions (ACC, F(2,2130) = 59.74, p < 0.0001; MD, F(2,1850) = 63.57, p < 0.0001; SI, F(2,2520) = 109.2, p < 0.0001). A post hoc Bonferroni test for multiple comparisons showed that the pain-related responses in Session 2 were significantly higher than in Session 1 (p < 0.05) for all the recorded areas, suggesting that anticipation of pain may enhance the nociceptive transmission in the brain. No significant difference was found between Sessions 1 and 3, indicating that no sensitization or tolerance was developed throughout Session 1–3. Tone-related responses were relatively weak with respect to the laser-induced responses. As illustrated in Fig. 3B, comparing with tone presentation alone, paring the tone cue with nociceptive stimulation significantly increased tone-related neural activity in the ACC but not in the MD and SI. These results suggest that the ACC may play a significant role in neural processing involved in pain anticipation.Figure 3Laser (A) and tone-induced (B) responses in each session. The magnitude of neuronal discharge was assessed by Z-scores. The laser-induced response was presented as a time course post laser stimulus and averaged every 1 sec. The tone-elicited response was calculated 0 – 1 s following the onset of the tone. *, # p < 0.05 indicate significantly different from Session 1 and Session 3, respectively.Functional connectivity within and between the recorded areas during noxious stimulationCorrelations between the neurons within the same region were observed more often than those between different regions. For the within-area cross-correlations, Chi-square tests showed that the correlated activity in Session 2 was significantly higher than that in Session 1 (p < 0.05) for all the recorded areas (Fig. 4A). For the between-area cross-correlations, the significantly enhanced correlated activity was observed between the MD and the other two regions (ACC and SI) in session 2 in comparison with session 1 (Fig. 4B).Figure 4The percent of significantly correlated neuronal pairs within (A) and between (B) recorded brain areas during noxious stimulation. * p < 0.05 indicates significantly different from Session 1.Information flow between the recorded areas during noxious stimulationPDC analysis was used to determine the direction of information flow from one region to others under different experimental conditions. A two-way ANOVA was performed to measure the difference in the normalized PDC between the Session 1 and 2. There was no significant Session × Direction interaction for all the regional pairs. However, significant effect was found in directions for ACC-SI (F(1,31) = 13.86, p = 0.0008) and MD-SI (F(1,30) = 5.184, p = 0.0301), which indicated that the information flow from the ACC to SI, and the MD to SI was significantly larger than that in the opposite direction in both sessions (Fig. 5A and 5B).Figure 5Result of partial directed coherence analysis during noxious stimulation. (A) Two-way ANOVA showed that the information flow from ACC to SI, and MD to SI was significantly larger than that in the opposite direction in both Session 1 and 2. 'ACC-MD' indicates directed coherence from ACC to MD, and the same as the other regional pairs. Values are normalized to the pre-stimulation baseline level. * p < 0.05, *** p < 0.001, compared with PDC in the opposite direction. (B) An example of the amount of partial directed coherence observed between recorded areas. These PDC values were normalized to z-scores relative to the mean and variance of baseline (pre-stimulation) PDC. The normalized PDCs exceeding 95% confident interval of the baseline were displayed in pseudo colour. Warm and cool colours indicate the increase and decrease in PDC, respectively. The direction of information flow are from the column area to the row area.DiscussionIn the present study, simultaneous single unit recording was performed in the ACC, MD and SI to study the neural mechanism underlying the effect of pain expectation using tone-laser conditioning model in rats. There were three main findings. First, under anticipation condition, neuronal responses to the auditory cue were significantly increased only in the ACC whereas those to nociceptive stimuli were enhanced in all the recorded areas. Second, expectation of impending pain enhanced correlated neural activity within and between recording areas following noxious stimuli and third, there were larger amounts of information flow from the medial (ACC and MD) to the lateral (SI cortex) pathway during pain processing.Neuroimaging studies have identified anticipation-related activation in many cortical areas including the SI, ACC, IC, and the PFC [10-12,14,15]. In the present study, we found significant increase in the neuronal response in the ACC but not in the SI during pain expectation. There are two possible explanations for the inconsistency between our and others' results. First, the animal model used in the present study is the type of 'certain' expectation, in which the neutral conditioned stimulus (tone) reliably predict the noxious unconditioned stimulus (laser). In contrast, most imaging studies on human expectation employed an uncertain paradigm. Certain and uncertain expectations have been demonstrated to be mediated by different neural pathways; the former is associated with activity in the ACC [10,15,19-21], whereas the latter involves changes in the SI [11,14,22]. Thus, our results provide evidence that the ACC, rather than the SI, is a structure critically involved in the neural process underlying certain expectation of pain. Another possible explanation is derived from the attention-related focusing mechanism. Previous studies in rats have found that the ACC is involved in tasks required visual or audio attention and preferentially activated during presentation of the conditional stimulus [23-27]. The increased activity in the ACC observed in the present study could also be due to the conditioning experiment employed.As previously described, expectation of an aversive event (painful stimulation) can modify subsequent behavior (pain reactivity). Conditioned expectation (certain expectation) is associated with the emotional state of fear, which produces hypoalgesia [28-30]. In contrast, unconditioned expectation (uncertain expectation) is related to anxiety, which has the opposite effect on pain, i.e., hyperalgesia [31-33]. In our study, we found that the nociceptive behaviors (paw lifting and licking) induced by actual pain stimulation were reduced during conditioning, which is consistent with prior studies that the conditioned fear leads to decreased behavioral reactivity [33]. Unexpectedly, the analysis of neuronal activity suggested that the laser-elicited responses in all recorded areas were enhanced under the expectation conditions. This seemed in contradiction with the behavioral findings that pain was decreased. It should be noted that noxious stimulation-elicited fear itself is a negative emotion. The emotional component of pain has been known to involve pathways through the medial thalamus to the ACC [34,35]. In addition, emotional states are found to be closely related to attentional states [12]. There is evidence in humans and animals for the involvement of the ACC as well as the primary somatosensory cortices in the attentional modulation of pain [36-38]. Thus, the neuronal responses in the recorded areas may reflect a mixed effect exerted by expectation. On the other hand, the increased neuronal activity and cross-correlations in ACC and MD during noxious stimulation may represent an endogenous antinociceptive activation instead of signalling nociceptive information. Early studies indicate that both the medial thalamus and ACC are involved in the activation of descending pain suppression mechanisms. Projections from the midline thalamic nuclei and ACC to the PAG have been described [39,40]. A high density of opioid receptors and activation induced by fentanyl within ACC support the participation of it in the down-regulation of pain perception [41]. Therefore, the inconsistence between the behavioral findings and neural activity change suggest a more complex role of medial system in pain processing, including both mediation and suppression.Converging evidence indicates that pain is a multi-dimensional experience that involves distributed brain regions comprising lateral and medial systems [42-45]. The medial pain system consists of the ACC and the medial thalamic nuclei and is believed to process the emotional-motivational component of pain [19,35]. The functional relationship and anatomical connection between the ACC and the medial thalamus have been demonstrated by numerous studies [46-50]. The present study simultaneously recorded neurons in these areas, and found a significant increase in the number of correlated neuronal pairs within the same and between different brain areas under anticipation conditions, compared to those under non-anticipation conditions. An increased synchronized activity observed in the present study suggests temporal coding may play a significant role in processing pain perception under the conditioning state. Together with previous discussion, these results indicated an enhanced network processing in the pain neuromatrix under the expectation of pain. Further studies will be required to elucidate the implication of this finding.Another interesting finding of this study is larger amount of information flow from the medial (ACC and MD) to the lateral (SI cortex) pathway as compared to those in the opposite direction. PDC analysis reveals causality of coherent neural activity of two regions. In this view, our result indicates that the emotion-related neural circuits may modulate the neuronal activity in the somatosensory pathway during nociceptive transmission. Our previous study on tonic pain has demonstrated an increase in the information flow from the medial to the lateral pain pathway during the first hour after formalin injection [51]. Although little available evidence supports direct linkage between the medial and the lateral pain systems [34], our prior and current results both suggest the medial system may modulate lateral system during nociceptive processing, Based on the fact that the medial system is composed of the medial and intralaminar thalamic nuclei and limbic cortical areas which have descending projections to nociception regulating centres such as PAG, it is possible that the medial system modulates somatosensory nociceptive transmission through the brainstem structures that control both spinal and trigeminal dorsal horn pain transmission neurons. Thus, clarifying the anatomical and functional interaction between the parallel systems can provide deeper insight into the neural mechanism of expectation related pain modulation and may help us to improve the treatment of clinical pain.ConclusionThe present study demonstrated that anticipation of pain enhanced the neuronal discharges and correlated neural activity within and between brain regions in the medial and lateral pain pathways induced by the following noxious stimuli, indicating that the nociceptive processing in both medial and lateral pain systems is modulated by the expectation of pain.MethodsAnimalsAll experiments were performed on nine male Sprague-Dawley rats (250–300 g) individually housed in a room maintained with a 12-h light/dark cycle. Food and water were available ad libitum. The experimental protocols were approved by the Animal Care and Use Committee at the Chinese Academy of Sciences and in accordance with the IASP guidelines for animal study. Every effort was made to minimize both animal suffering and the number of animals used.Surgery for microelectrode implantationAfter a 7-day period of habituation, animals received the surgery of microelectrode implantation. As in the previously described procedure [18], rats were anesthetized with ketamine (100 mg/kg, i.p.) and xylazine (10 mg/kg i.m.) and mounted on a stereotaxic frame. Following the retraction of skin and soft tissue, small holes corresponding to the recording sites were drilled on the skull. Then arrays of eight stainless steel Teflon-coated microwires (50 μm diameter, Biographics, Inc. Winston-Salem, NC, USA) were implanted unilaterally into the target brain areas. The coordinates were as follows: 3.2 mm anterior (A) to bregma, 0.8 mm lateral (L) to midline, 2.5 mm ventral (V) from the skull surface for the ACC; 2.3 mm posterior (P), 0.8 L, 5.5 V for the MD; and 1.0 P, 2.0 L, 2.0 V for the SI cortex. The microarrays were fixed in place with dental cement. Animals were housed individually and allowed to recover from surgery for at least 7 days before being subjected to the experiment.Apparatus and laser stimulationThe experimental chamber was 44 × 22 × 44 cm in dimension and made of transparent acrylic plastics. The floor of the chamber contained an array of holes (diameter 6 mm) spaced 10 mm apart (centre to centre). A computer-controlled CO2 laser stimulator (Model DM-300, Changchun Institute of Optics, Fine Mechanics and Physics, Chinese Academy of Science) was used to deliver pain stimuli. The laser radiation was 10.6 μm in wavelength and 2.5 mm spot diameter. The output power and duration of the stimulation were set at 8 W and 20 ms, which was designed to activate the primary nociceptive afferents without damaging the skin or the subcutaneous tissue [52]. The CO2 laser was equipped with a He-Ne aiming beam, which was visualized prior to firing the laser. The laser beam was emitted through the holes at the floor of the chamber to the plantar surface of rat hindpaw contralateral to the recording brain regions. A tone generator (Coulbourn Inc. USA) was mounted on the side wall of the chamber for audio stimuli (80 dB, 800 Hz, 100 ms). In tone-laser pairing trials, a laser pulse was delivered 900 msec after the tone. The onsets and durations of the tone and laser were controlled by a PC running program Magnet (Biographics, Inc.). A video camera was positioned in front of the chamber to record behavioral activity.Experimental protocolThe experiments were conducted under normal lighting. In the first 3 days, rats were placed individually in the experimental chamber each day for 1 h. The experimental sessions started from day 4. All rats underwent 3 experimental sessions in 3 successive days (Fig. 6). The behavior of rats was continuously videotaped throughout the sessions for later analysis.Figure 6Experimental procedure. Session 1 contains two blocks of stimuli, one consisted of 40 laser stimuli and the other consisted of 30 tone stimuli. Session 2 involves 60 tone-laser pairing stimuli trials. Session 3 comprises another 40 trials of laser stimuli. The inter-trial interval was no less than 60 sec.In Session 1, two blocks of stimuli were delivered. One consisted of 40 painful laser stimuli and the other consisted of 30 tone stimuli, with inter-trial interval of no less than 60 sec. The order of the stimuli was balanced across animals. This session serves as a normal control for the following two sessions.In the second session, animals were subjected to 60 tone-laser pairing stimuli trials. This conditioning training formed a stable linkage between auditory cues and animals' nociceptive behaviors, i.e. the auditory cue will forecast the impending painful stimulus. This learning phase was followed by a testing phase (30 trials), in which the tone was administered without paired with laser, to determine the acquisition and extinction of the learned responses. The Session 2 was used to assess the effects of anticipation on neural activity and functional connectivity within central pain networks.In Session 3, animals received another 40 trials of laser stimuli. This session tests the possibility of sensitization or tolerance for noxious stimuli.Unit recordingThe simultaneous extracellular recording of the three selected brain areas was performed throughout the experimental sessions. Neural electric signals were obtained from the stainless steel microwires and passed from the headset assemblies to a preamplifier via two lightweight cables and a commutator. The commutator was free to turn as necessary, permitting unrestricted movement of the rat. The signals were band-pass filtered between 0.5 and 5 kHz (6 dB cutoff) before being sent to a spike-sorting device (Biographics, Inc.). Valid spikes were selected using amplitude and duration thresholds and recorded into a database file with PC-based software (Magnet, Biographics, Inc.). The identity of clearly sorted single neurons was verified by graphical capture of waveforms. Onset of tone and stimulation events was recorded into the data file with a resolution of 1 ms.HistologyAnimals were sacrificed with overdosed pentobarbital at the end of experiment. Recording sites were marked by electrophoretically deposited iron (20 μA, 10 – 20 s) at the tips of selected wires. Animals were then perfused with 4% paraformaldehyde and the brains were post-fixed, frozen, and cut coronally into 40-μm sections. The iron deposits could be visualized as blue dots under light microscope. Data obtained from the microwires outside the target regions were not included in the analysis.Behavioral assessment and analysisBehavioral responses to nociceptive stimuli were assessed by off line video analysis. According to the method of Fan et al., the laser-induced nociceptive responses in rats can be classified as eight categories: head movement (Hm), body movement (Bm), foot jumping (Fj), foot elevation (Fe), foot movement (Fm), licking (Li), rearing (Re), and grooming (Gr) [52]. The frequency and duration of each response were then used to quantitatively evaluate nociceptive behaviors. Here we modified the method by focusing on five of the eight categories listed above, i.e., Hm, Bm, Fj, Fe, and Fm. A score of 0 was assigned if the rat stayed quietly; a score of 1 if the rat displayed Hm (exploring); a score of 2 if Bm or Fm (motivation of avoidance) was observed; a score of 3 if Fj or Fe (successful escape) occurred. The behavioral response was measured with cumulative scores every 5 successive trials. One-way ANOVA followed by the Dunnett test for multiple comparisons was used to compare behavior scores between sessions. The behavioral assessment was used to examine whether and when the tone-laser association was steadily formed.Data analysisThe neuronal firing rate was quantified for each neuron using peri-event time histograms (PSTHs). The bin size was 0.1 s for the computation of PSTHs. Bin counts for each trial were calculated using the analysis program NeuroExplorer (Plexon, Dallas, TX) and the results were exported to Matlab (The MathWorks, Inc.) in spreadsheet form. Neural responses to auditory or noxious stimulation were evaluated using a sliding window averaging technique, in which a 1-s time window was slid through the entire period of a trial at 0.1-s step. The bin counts of each window were compared with those of a preset 3-s control window 10 s before the stimulation event by Student's t-test. The differences were considered significant only when it reached a significance level of p < 0.005 in three consecutive steps, thus to achieve a global significance of p < 0.05. Units that significantly increased their activities after tone or laser stimuli were defined as excitatory; those that decreased their activities were considered inhibitory. To compare the neural responses between different sessions, the neuronal firing rates were transferred into Z scores using MatLab program: Z = (X-M)/S, where X is the actual firing rate obtained from PSTH, M and S are mean and standard deviation of the baseline discharging (-5 – -2 s), respectively.To identify the functional interactions between the recorded areas, cross-correlation and partial directed coherence (PDC) analyses were performed. In the cross-correlation analysis, one neuron was selected as the reference neuron and all other neurons were defined as partner neurons. The peri-spike histogram of a partner neuron within -0.5~0.5 sec around the reference neuron were calculated with a 5-ms bin size and a 3-bin Gaussian smooth. The significance level of the cross-correlograms was defined by 95% confidence level in the Nex program. Data falling into the 10-s period after laser stimulation were calculated.PDC is a frequency domain representation of the key concept of Granger causality. Briefly, if knowledge of x(n)'s past significantly improves prediction of y(n), we could then states that an observed time-series x(n) Granger-causes series y(n). This relation between time-series is by no means reciprocal. Absence of PDC between two structures at a given frequency means the lack of a direct link between them. Thus, PDC allows the detection of coactivations among simultaneous neuronal activities by highlighting one neuronal group that possibly drives another. The detailed methodology of PDC has been described elsewhere [53-57]. In the present study, principal component analysis (PCA) for neurons in each brain area was first performed in Nex. Then the first principal component (PC1) of a given brain area that had the largest response were exported into MatLab, and the value of PDC across 1 – 50 Hz for each 2.5-s analysis time window were calculated. These values were then averaged around the laser stimulation events (0–10 s post-stimulus) and normalized to Z scores relative to the baseline (before stimulation) data.ANOVAs and non-parametric Chi-square tests were performed to determine differences in Z scores and percentage of correlated neurons between sessions, respectively. Bonferroni's test was used for post hoc test and p < 0.05 was considered to be level of significance for all the statistics.AbbreviationsACC: Anterior cingulate cortex; MD: Medial dorsal thalamus; SI: primary somatosensory cortex; PAG: Periaqueductal grey; IC: Insular cortex; PFC: Prefrontal cortex; PSTH: Peri-event time histogram; PDC: Partial directed coherence; PCA: Principal component analysis.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsJYW participated in the design of the study and drafted the manuscript. HTZ carried out all the experiment and performed the statistical analysis. JYC and DJW assisted with the electrophysiological recordings and data analysis. LAB contributed to the data analysis and interpretation. FL conceived of the study, and participated in its design and helped to draft the manuscript. All authors read and approved the final manuscript.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2532682.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2532682",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2532682\nAUTHORS: Regina L Faulkner, Lawrence K Low, Xiao-Bo Liu, Jeffrey Coble, Edward G Jones, Hwai-Jong Cheng\n\nABSTRACT:\nBackgroundThe development of the corticospinal tract (CST) in higher vertebrates relies on a series of axon guidance decisions along its long projection pathway. Several guidance molecules are known to be involved at various decision points to regulate the projection of CST axons. However, previous analyses of the CST guidance defects in mutant mice lacking these molecules have suggested that there are other molecules involved in CST axon guidance that are yet to be identified. In this study, we investigate the role of plexin signaling in the guidance of motor CST axons in vivo.ResultsExpression pattern studies show that plexin-A3, plexin-A4, and neuropilin-1 are expressed in the developing cerebral cortex when the motor CST axons originating from layer V cortical neurons are guided down to the spinal cord. By analyzing mutant mice, we show that motor CST axons that turn dorsally to cross the midline at the pyramidal decussation require plexin-A3 and plexin-A4 signaling. Although other CST guidance defects are found in neuropilin-1 mutants, this dorsal turning defect is not observed in either neuropilin-1 or neuropilin-2 mutants, suggesting that the local cues that activate plexin signaling at the dorsal turning point are membrane-bound semaphorins. Further expression pattern study and mutant analysis indicate that Sema6A is one of the local cues for motor CST axon turning at the pyramidal decussation.ConclusionDorsal turning and midline crossing at the pyramidal decussation is a crucial step to properly direct CST axons into the dorsal spinal cord. We show that the signaling of plexin-A3, plexin-A4, and Sema6A is at least partially required for dorsal turning of the CST axons, while neuropilin-1 is required for proper fasciculation of the tract at midline crossing. Together with previous reports, these results demonstrate that several guidance cues are specifically utilized to regulate the dorsal turning and midline crossing of developing CST axons.\n\nBODY:\nBackgroundThe formation of functional neural circuits within the central nervous system (CNS) requires proper guidance of axonal projections to specific target regions. The guidance of axons to distant targets within the CNS relies on the presence of signals at different choice points to guide axons along a correct pathway [1-3]. The corticospinal tract (CST) represents the longest projection pathway in the CNS of higher vertebrates [4-8]. In developing rodents, the CST axons originate from layer V cortical pyramidal neurons [7]. They exit the neocortex through the internal capsule and cerebral peduncle. In the brainstem, they are guided along the pyramidal tract and turn dorsally at the pyramidal decussation to cross the midline and reach the contralateral side of the spinal cord (Figure 1a). The targeting of primary CST axons to the spinal cord is followed by axon collateral branching to several target areas and then by pruning of specific collateral branches [7,9].Figure 1Expression of PLXA3, PLXA4, NPN-1, and NPN-2 in the neocortex during corticospinal tract targeting.(a) Diagram of sagittal view of the brain and cross-section of the brainstem and spinal cord representing axon targeting of the corticospinal tract at P0. (b-e) In situ hybridization of PLXA3, PLXA4, NPN-1, and NPN-2. Radioactive (b, c) and non-radioactive (b', b\", c', c\") in situ hybridization demonstrates that PLXA3 and PLXA4 mRNA is expressed throughout the neocortex at P0. NPN-1 mRNA (d-d\") is expressed in deeper layers of the neocortex at P0. Insets in (b'-d') show cortical neurons (arrows) that co-express PLXA3, PLXA4, or NPN-1 with the layer V neuronal marker Ctip2. NPN-2 mRNA (e-e\") is not expressed in cortex at P0. (f, g) L1 immunohistochemistry (IH) of the sagittal brain demonstrating the normal course of subcortical projections through the internal capsule of P1 WT and PLXA3/PLXA4-/- mice. (h, i) Sagittal sections of the brain showing the normal course of BDA-labeled subcortical projections from the motor cortex of P25 WT and PLXA3/PLXA4-/- mice. Black arrows indicate BDA-labeled axons descending through the internal capsule. C, caudal; CP, cortical plate; D, dorsal; IC, inferior colliculus; IZ, intermediate zone; MC, motor cortex; Pn, pons; Pyr Dec, pyramidal decussation; R, rostral; SC, superior colliculus; SpC, spinal cord; V, ventral; VC, visual cortex; VZ, ventricular zone. Scale bars: 1,000 μm (b-e); 400 μm (b'-e'); 25 μm (insets in b'-d'); 100 μm (b\"-e\"); 500 μm (f-i).Recent evidence has demonstrated that molecules involved in axon guidance elsewhere in the CNS are also involved in regulating axon guidance decisions made by the CST [10]. Guidance of initial corticofugal projections to the cerebral peduncles is dependent on Slit function [11]. When CST axons approach the pyramidal decussation at the caudal medulla, intact netrin signaling via DCC and Unc5h3 receptors is required to prevent axon mistargeting [12]. The immunoglobulin (Ig) superfamily molecules L1 and NCAM have been implicated in maintaining the fidelity of the CST bundle as it turns and crosses at the pyramidal decussation [13,14]. As CST axons travel caudally from the decussation, repulsive cues by Wnt morphogens seem to determine the rostro-caudal positioning of the axons in the dorsal columns of the spinal cord [15]. Finally, when CST axons collateralize within the contralateral gray matter of the spinal cord, ephrin signaling is required to prevent axon branches from re-crossing the midline [16,17]. Together, the evidence demonstrates that the guidance choices of CST axons are highly dependent on the presence of local cues in their CNS environment. However, since loss of these molecules only results in partial defects in CST targeting, additional axon guidance signaling pathways might be involved in regulating CST axon targeting.Plexins belong to families of axon guidance molecules that act as receptors for semaphorin ligands. Together, they are by far the largest family of axon guidance molecules. Membrane-bound semaphorins (classes 4–7) directly interact with and signal through plexins, whereas most secreted semaphorins (class 3) signal through a receptor complex composed of plexins and their co-receptors, neuropilin (NPN)-1 or NPN-2 [18,19]. Semaphorin signaling through plexins is known to play roles in multiple aspects of neuronal development, and axon guidance is its most classical role [18-23]. Although several semaphorins have been shown to repel or attract neurites from cortical cultures in vitro [24-28], their roles in regulating the guidance of CST axons in vivo are still largely uncharacterized. Here we report that plexin (PLX)A3, PLXA4, and one of the membrane-bound semaphorins, Sema6A, are required for the dorsal turning of CST axons at the pyramidal decussation.ResultsThe expression of plexin-A3, plexin-A4, and neuropilin-1 in cortical neurons coincides with the guidance of motor CST axonsTo address whether semaphorin signaling through plexins regulates the guidance of CST axons, we focused on PLXA3 and PLXA4, as well as neuropilins, NPN-1 and NPN-2, and analyzed their expression patterns in the developing neocortex. The mRNAs of PLXA3 and PLXA4 were broadly expressed throughout the cortex from embryonic day (E) 18 to postnatal day (P) 0, immediately after layer V pyramidal neurons are born and migrate to their appropriate layer in the neocortex (Figure 1b–c\", and data not shown). NPN-1 was also expressed in the developing neocortex at P0, but its expression was more restricted (Figure 1d–d\"). By P3, once most CST axons have reached their targets in the spinal cord, NPN-1 expression in the cortex was reduced while PLXA3 and PLXA4 expression levels were maintained (data not shown). By contrast, NPN-2 transcripts were not expressed in the cortex during this time window (Figure 1e–e\"). CST axons arise predominantly from type I layer V neurons [7,29], which specifically express a transcription factor, Ctip2 [30]. We found that a majority of Ctip2 immuno-positive pyramidal neurons co-expressed mRNA for PLXA3, PLXA4, and NPN-1 at P0 (Figure 1b'–d'). These results suggest that PLXA3, PLXA4, and NPN-1 play roles in guiding the developing motor CST axons to the spinal cord. To confirm their roles in vivo, we investigated whether the guidance of motor CST axons is affected in mutant mice lacking these genes.Plexin-A3 and plexin-A4 are required for dorsal turning of motor CST axons at the pyramidal decussationA recent analysis of PLXA3 (PLXA3-/-) and PLXA4 mutant (PLXA4-/-) mice using NPN-1 expression as a marker suggested that NPN-1-positive axons projecting subcortically through the internal capsule and cerebral peduncles were defective in neonates [31]. To examine whether the initial guidance of CST axons through these structures is normal in PLXA3/PLXA4 double mutant (PLXA3/PLXA4-/-) mice, we studied the CST projections by using both L1-immunostaining at P1 [32] and biotinylated dextran amine (BDA) anterograde tracing of the motor CST axons at P25. Although subtle defects cannot be completely ruled out, targeting as well as fasciculation of these axons as they entered the internal capsule and cerebral peduncles appeared normal in P1 PLXA3/PLXA4-/- mice (n = 3) compared to wild-type (WT) mice (n = 3) (Figure 1f, g). When these initial projections from motor cortex were examined at P25 by BDA tracing, again the patterns were similar in WT (n = 3) and PLXA3/PLXA4-/- mice (n = 3) (Figure 1h, i), even though we could not exclude the possibility that subtle defects early on were corrected over time. Our results suggest that the CST projection through the internal capsule appears normal in PLXA3/PLXA4-/- mice.We next examined the CST axons within the pyramidal tracts of the brainstem and the spinal cord by anterograde tracing. DiI (1,1'-dioctadecyl-3,3,3',3'-tetramethylindocarbocyanine perchlorate) or BDA tracers were bilaterally injected into the WT or mutant motor cortex to label the CST axonal projection down to the spinal cord. Although the guidance of motor CST axons through the brainstem structures was unaffected, we found a large DiI-labeled bundle of axons that diverged toward the ventrolateral aspect of the spinal cord at the pyramidal decussation in P3 PLXA3/PLXA4-/- mice (n = 4; Figure 2c–d'). By contrast, with this labeling technique, no abnormal ventral CST axons were observed in WT mice at P3 (n = 3; Figure 2a–b'). The abnormal ventrolateral CST persisted into adulthood in all PLXA3/PLXA4-/- mice (n = 14; Figure 2e–f\"). Ultrastructural analysis of these mistargeted BDA-labeled axons demonstrated that they were myelinated and their perimeters were normal in size when compared with CST axons in WT mice (Figure 2g–i).Figure 2Motor corticospinal axon pathfinding is abnormal in mice lacking PLXA3 and PLXA4.(a-b') Diagram and sagittal view of the brain showing motor CST axons bilaterally labeled with DiI traveling within the pyramidal decussation (Pyr Dec; white arrow in (b')) and dorsal spinal cord (dSpC; white arrows in (b)) of P3 WT mice. (c-d') Diagram and sagittal view of the brain showing motor CST axons bilaterally labeled with DiI traveling past the pyramidal decussation and into the dorsal (white arrows in (d)) and ventral spinal cord (vSpC; yellow arrows in (d')) of P3 PLXA3/PLXA4-/- mice. Note that the white dashed lines in (b-b', d-d') indicate meninges surrounding the dorsal and ventral edges of the spinal cord and do not represent positive DiI labeling. (e) Diagram showing bilaterally labeled CST axons in P30 PLXA3/PLXA4-/- mice. In all diagrams (a, c, e), the normal motor CST axonal projection is indicated in blue and the abnormal ventral CST projection is indicated in yellow. (f-f\") Course of BDA-labeled motor CST axons along the pyramidal tract in the brainstem (black arrowheads) of P30 PLXA3/PLXA4-/- mice. BDA-labeled axons were observed in the dorsal (dSpC; red arrow) and ventral (vSpC; black arrow) spinal cord. Higher power views of arrowed areas are shown in the insets of (f', f\"). (g, h) Electron micrographs illustrating examples of BDA-labeled motor CST axons in the dorsal (g) and the ventrolateral (h) aspect of the cervical spinal cord in P30 PLXA3/PLXA4-/- mice. All labeled axons are myelinated (red arrows). (i) Average perimeters (mean ± standard error of the mean) of BDA-labeled axons are similar within the dorsal CST of P30 WT mice (n = 36 sections from 2 mice) and the dorsal (n = 13 sections from 2 mice) and ventrolateral (n = 26 sections from 2 mice) CST of P30 PLXA3/PLXA4-/- mice (dKO). p > 0.05, ANOVA, Neuman-Keuls test. (j) Average densities of axons (mean ± standard error of the mean of axons per 100 μm2) are similar in the dorsal CST of P30 WT (n = 36 sections from 2 mice) and PLXA3/PLXA4-/- mice (dKO; n = 13 sections from 2 mice). p > 0.05, Student's t-test. Each data set was averaged from randomly selected CST areas on all the electron micrographs taken from two animals. Scale bars: 500 μm (b-b', d-d'); 1,000 μm (f); 200 μm (f', f\"); 1 μm (g, h).In the PLXA3/PLXA4-/- mutants, approximately one-half of the motor CST axons could still turn dorsally at the pyramidal decussation. These axons were properly guided across the midline and entered the dorsal funiculus in the spinal cord. This finding suggests that the turning is partially compensated for by other signaling in vivo. Alternatively, the partial defects in the plexin mutants may be due to the expression of PLXA3 and PLXA4 in a subset of CST axons. Given the broad expression patterns of these two genes (Figure 1b–c), it is more likely that additional molecules are required in the process. The properly guided CST axons in the spinal cord of PLXA3/PLXA4-/- mice remained well fasciculated as their axon densities were normal compared to WT mice (Figure 2j). In addition, we did not observe any errors in the targeting of transient motor CST axon collaterals to the superior colliculus in PLXA3/PLXA4-/- mice (n = 3) at P9 (data not shown). Taken together, these results indicate that signaling through PLXA3 and PLXA4 is utilized at the pyramidal decussation to control the dorsal turning of the motor CST axons in vivo.The abnormally guided CST axons in plexin-A3/plexin-A4 mutants do not cross the midline at the pyramidal decussationWe also injected BDA unilaterally in the motor cortex to determine whether the abnormal ventrolateral spinal CST axons had crossed the midline. We examined the labeled CST axons in serial transverse sections and found that the abnormal CST axons maintained their course ipsilaterally at the pyramidal decussation and occupied a unique position in the ventrolateral region of the spinal cord in all PLXA3/PLXA4-/- mice (n = 6; Figure 3c–d\"). Again, the ventrolateral axons were not seen in WT mice (n = 5; Figure 3a–b'). Consistent with the bilateral labeling results, some of the unilaterally labeled axons were found in the contralateral dorsal funiculus of PLXA3/PLXA4-/- mice, but the number was significantly reduced compared to WT mice (Figure 3b', d'). We also noticed that the ipsilateral ventrolateral CST axons in the mutant mice did not travel beyond the upper thoracic spinal cord. In these sections, many mutant axons could be seen branching from the ventrolateral CST and crossing to the gray matter of the contralateral dorsal spinal cord (Figure 3d', d\"). This somewhat surprising observation suggests that at least some of the aberrant motor CST axons in PLXA3/PLXA4-/- mice can be directed to the appropriate final target area in the spinal cord.Figure 3Aberrant motor corticospinal axons in PLXA3 and PLXA4 double mutants are located in the ipsilateral, ventrolateral spinal cord. All panels represent cross sections of the pyramidal decussation and spinal cord. Unilateral BDA motor CST axon tracing was performed in (a-d\"). (a, b) Crossing of motor CST axons at the pyramidal decussation (Pyr Dec; black arrow in (a)) and into the dorsal funiculus of the spinal cord (SpC; black arrow in (b)) of P45 WT mice. (c, d) Normal crossed (black arrows) and aberrant uncrossed (red arrows) motor CST axons at the pyramidal decussation and spinal cord of P45 PLXA3/PLXA4-/- mice. (a'-d\") High power views of arrowed areas in (a-d). In the upper cervical spinal cord, many mutant axons branched out from the uncrossed CST (green arrowheads in (d\")). Some of these axons crossed to the contralateral gray matter (green arrowheads in (d')). (e-h\") CamKIIα immunohistochemistry of the pyramidal decussation and spinal cord of P25 WT and P25 PLXA3/PLXA4-/- mice. Crossed (black arrowheads) and uncrossed (red arrows) CamKIIα-immunolabeled CST axons are observed at the pyramidal decussation and spinal cord in PLXA3/PLXA4-/-'s. High power views of the spinal cord in (f, h) are shown in (f', h', h\"). Black dashed lines in (f', h') outline positive CamKIIα immunostaining of the dorsal CST in the dorsal funiculus of the spinal cord. (i) Comparison of percentages of 4- to 6-week old WT, PLXA3-/-, PLXA4-/-, and PLXA3/PLXA4-/- mice with an abnormal ventral CST apparent with BDA tracing. Numbers in parentheses indicate the number of mice analyzed. (j) Average normalized areas (see Materials and methods) of CamKIIα-labeled dorsal CST axons in WT, PLXA3-/-, PLXA4-/-, and PLXA3/PLXA4-/- mice. The dorsal CST area in each animal (mean ± standard error of the mean) is indicated by a black circle. The overall average dorsal CST areas (black lines) are decreased in the cervical spinal cords of PLXA3-/- (n = 6 mice), PLXA4-/- (n = 5 mice), and PLXA3/PLXA4-/- (n = 8 mice) versus WT (n = 6 mice) mice. **p < 0.01, Student's t-test. cc, central canal; DF, dorsal funiculus. Scale bars: 500 μm (a-h); 100 μm (a'-d\", f'-h\").To confirm that the misguided CST axons in mutant mice were motor axons, we labeled CST axons at P25 with an antibody against alpha calcium/calmodulin-dependent protein kinase type II (CamKIIα), which is specifically upregulated in motor CST axons in mice older than three weeks of age [33]. The results showed that the aberrant ventrolateral CST axons were labeled in bilateral regions of the medulla and spinal cord in PLXA3/PLXA4-/- (n = 4) but not WT (n = 3) mice (Figure 3e–h\"), indicating that they are indeed CST motor axons. Since this marker stained all the motor axons, we also confirmed that the area of the dorsal funiculus occupied by CST axons in PLXA3/PLXA4-/- mice was considerably reduced compared to WT mice (Figure 3f', h', j).We also assessed the individual contributions of PLXA3 and PLXA4 to the defect in single mutants. Although the phenotype was not present in all PLXA3-/- and PLXA4-/- mice, roughly equal numbers of PLXA3-/- and PLXA4-/- mice contained a ventrolateral CST (Figure 3i, and data not shown), suggesting that these two plexins partially compensate for each other's functions. Furthermore, we found that the area of the dorsal funiculus occupied by CST axons in PLXA3-/- and PLXA4-/- mice was smaller than in the WT, though the defect in the single mutants was less severe than in the PLXA3/PLXA4-/- animals (Figure 3j).Misguided ventrolateral CST axons are not observed in neuropilin mutantsOur expression pattern studies predict that NPN-1, but not NPN-2, is required for the guidance of developing CST axons. To address whether neuropilins are required for targeting motor CST axons towards the contralateral dorsal spinal cord in vivo, we analyzed NPN-1sema-/- (mutant mice expressing NPN-1 that lacks a semaphorin binding domain; n = 4) and NPN-2-/- (NPN-2 mutants; n = 3) for axon guidance defects in the CST axon projection [34,35]. As expected, no defects in motor CST axon guidance were observed in NPN-2-/- mice (Figure 4c–d', g–h). We did observe CST axon guidance defects in NPN-1sema-/- mice, but the abnormality was qualitatively different from that seen in PLXA3/PLXA4-/- mice. All the CST axons from NPN-1sema-/- mice turned dorsally and crossed the midline at the pyramidal decussation (Figure 4a, g). However, they were defasciculated when they crossed the midline and this resulted in a wider pyramidal decussation in NPN-1sema-/- mice than in WT mice (Figure 4a', h). Some of these defasciculated axons formed ectopic tracts in the contralateral half of the dorsal spinal cord (Figure 4b'). As expected, we found that PLXA3/PLXA4-/- mice had a pyramidal decussation that was smaller in width than WT since only a subset of their axons crossed at the pyramidal decussation (Figure 4h). These data show that PLXA3/PLXA4 and NPN-1 are differentially required for CST axon guidance, and suggest that neuropilins and secreted (class 3) semaphorins are not involved in turning CST axons away from the ventral side of the pyramidal decussation.Figure 4Motor corticospinal axon turning at the pyramidal decussation is independent of neuropilins.(a-f) Unilateral BDA motor CST axon tracing was performed in (a, b, e, f), and bilateral BDA motor CST axon tracing was performed in (c, d). (a'-f') Higher power views of arrowed areas in (a-f), respectively. The abnormal ventrolateral CST is not observed at the pyramidal decussation (a) and cervical spinal cord (SpC) (b) of P45 NPN-1sema-/- mice. However, crossing motor CST fibers are noticeably defasciculated at the pyramidal decussation (black arrows in (a, a')) and the defasciculated axons form ectopic tracts in the contralateral spinal cord (black arrows in (b, b')). Motor CST axons in P45 NPN-2-/- (c, d) and Sema3A-/- (e, f) mice travel normally at the pyramidal decussation (black arrows in (c, c', e, e')) and cervical spinal cord (black arrows in (d, d', f, f')). (g) Comparison of percentages of WT and mutant mice with an abnormal ventral CST apparent with BDA tracing. Numbers in parentheses indicate the number of mice analyzed. This result indicates that in contrast to PLXA3/PLXA4-/- mice, there is no ventral CST in NPN-1-/-, NPN-2-/-, or Sema3A-/- mice. (h) Average width of the pyramidal decussation (mean ± standard error of the mean) in WT, PLXA3/PLXA4-/-, NPN-1-/-, NPN-2-/-, and Sema3A-/- mice. As expected, the width of the pyramidal decussation in PLXA3/PLXA4-/- mice (n = 6 mice) was smaller than WT mice (n = 5 mice). **p < 0.01, Student's t-test. In addition, the width of the pyramidal decussation was larger in NPN-1-/- mice (n = 4 mice) than WT, suggesting that CST axons are defasciculated in NPN-1-/- mice as they cross at the pyramidal decussation. *p < 0.05, Student's t-test. The width of the pyramidal decussation in NPN-2-/- mice (n = 2 mice) and Sema3A-/- mice (n = 4 mice) was similar to WT. cc, central canal; DF, dorsal funiculus. Scale bars: 500 μm (a-f); 100 μm (a'-f').To further support the conclusion that secreted semaphorins are not involved in CST axon turning at the pyramidal decussation, we examined the projections of motor CST axons in Sema3A (Sema3A-/-) and Sema3E mutant (Sema3E-/-) mice. Sema3A is expressed in the ventral spinal cord during development and has been thought to play a role in CST guidance by interacting with NPN-1 and L1 based on in vitro analyses [25,26]. In agreement with a recent report [36], we observed that the dorsal turning and midline crossing of motor CST axons at the pyramidal decussation was normal in Sema3A-/- mice (n = 3 BDA tracing, n = 2 CamKIIα immunostaining; Figure 4e–g). Further, in contrast to NPN-1sema-/- mice, we found that the fasciculation of axons crossing at the pyramidal decussation was normal in Sema3A-/- mice (Figure 4h). Sema3E has recently been shown to bind directly to plexin [37]. We analyzed the expression pattern of Sema3E and found that Sema3E was not expressed in the ventral spinal cord. In accordance with this finding, we also found that the Sema3E-/- mice did not have a defect in CST axon guidance (n = 4; data not shown). Thus, our data support the role of PLXA3 and PLXA4 in CST axon turning at the pyramidal decussation that is independent of neuropilins and secreted semaphorins.Sema6A is required for proper guidance of motor CST axonsTo explore the possible semaphorin cue(s) that activate PLXA3/PLXA4 signaling to guide the CST axons dorsally at the pyramidal decussation, we turned to membrane-bound semaphorins. Since PLXA4 is known to interact with Sema6A in a neuropilin-independent manner [38,39], we studied the expression pattern of Sema6A and analyzed the targeting of motor CST axons in Sema6A mutant (Sema6A-/-) mice. We found that Sema6A was expressed ventrally along the posterior pyramidal tract and pyramidal decussation between E16 and E18 (Figure 5a–b; and data not shown). By P0, when the majority of the motor CST axons have crossed the pyramidal decussation into the dorsal spinal cord, Sema6A expression was restricted to the inferior olive and the pyramidal decussation, though the latter appeared to be less prominent than at earlier stages (Figure 5c). This expression pattern suggested that Sema6A could be the ligand responsible for the plexin-mediated dorsal turning of motor axons at the pyramidal decussation. In Sema6A-/- mice (n = 4), we observed mistargeted axons in the ventrolateral spinal cord using anterograde BDA tracing similar to what was seen in PLXA3/PLXA4-/- mice (Figure 5e, g). However, the defect appeared to be more severe because relatively fewer labeled Sema6A-/- axons turned dorsally at the pyramidal decussation (Figure 5e'). In addition, the variation of defects from animal to animal was relatively broad such that each animal had fairly varied numbers of axons that crossed at the pyramidal decussation, but the majority of these animals appeared to have a more severe defect than the PLXA3/PLXA4-/- mice (Figures 3c–d\" and 5e–f\", and data not shown). We further assessed the severity of the defect in Sema6A-/- mice with CamKIIα staining (n = 2) and found that the defect in these animals was very similar to that of the PLXA3/PLXA4-/- mice (Figure 5h). As noted in the PLXA3/PLXA4-/- mice, the misguided ventrolateral CST axons branched out and targeted to the contralateral gray matter at the level of the cervical spinal cord (Figure 5f–f\"). These analyses indicate that membrane-bound Sema6A is one of the local cues that induces proper turning of motor CST axons dorsally at the pyramidal decussation.Figure 5Motor corticospinal axon turning at the pyramidal decussation requires Sema6A.(a-d) Sema6A mRNA expression along the ventral pyramidal tract and pyramidal decussation (black arrows) during early motor CST axon guidance. Red box in (a) indicates the region of Sema6A expression shown in sagittal views of E16 (b) and P0 (c) WT mice, which is absent in the sense control (d). (e, f) Unilateral BDA motor CST axon tracing of P20 Sema6A-/- mice. (e', f\") Higher power views of arrowed areas in (e, f), respectively. A boxed area in (f) is enlarged in (f'). Very few axons cross at the pyramidal decussation in Sema6A-/- mice (black arrows in (e, e')). Instead, axons form aberrant tracts (red arrows in (e, f, f\")) in the ventrolateral spinal cord (SpC). Note that the aberrant tract moves out laterally as it traces down to the ispilateral spinal cord. The slightly different locations of the ectopic ventrolateral tracts seen here as compared to those seen in the PLXA3/PLXA4-/- mice in Figure 3 are due to different rostrocaudal locations of the sections. Similar to that seen in PLXA3/PLXA4-/- mice, many of these ventrolateral axons branch back toward the contralateral dorsal cervical spinal cord, though they are mistargeted below the dorsal funiculus (green arrowheads in (f', f\")). (g) Comparison of percentages of WT, PLXA3/PLXA4-/-, and Sema6A-/- mice with an abnormal ventral CST apparent with BDA tracing. Numbers in parentheses indicate the number of mice analyzed. (h) Average normalized areas (see Materials and methods) of CamKIIα-labeled dorsal CST axons in WT, PLXA3/PLXA4-/-, and Sema6A-/- mice. The dorsal CST area in each animal (mean ± standard error of the mean) is indicated by a black circle. The overall average dorsal CST area (black lines) is decreased in the cervical spinal cords of Sema6A-/- (n = 2 mice) versus WT (n = 6 mice) mice. **p < 0.01, Student's t-test. cc, central canal; DF, dorsal funiculus, IO, inferior olive. Scale bars: 500 μm (b-f); 100 μm (e'f\"); 25 μm (f').DiscussionThe development of the CST has served as a classic example for studying the guidance of long-range axons [7,9]. In the CNS, midline-crossing is an important phenomenon for the guidance of long axons [40,41]. During development, the ventrally positioned CST axons make dorsal turns to cross the midline at the pyramidal decussation. Previous reports have indicated that multiple signaling systems are utilized to ensure the dorsal turning and midline crossing of CST axons at the pyramidal decussation [10]. These include the netrin/DCC/Unc5h signaling system and the Ig superfamily signaling system. We report here that the semaphorin/plexin signaling system is also involved in guiding CST axons dorsally at the pyramidal decussation.By comparing the reported CST defects in mutant mice from these signaling systems, we find that they may function in a cooperative fashion to regulate the guidance of CST axons at the pyramidal decussation. However, major phenotypical differences are also noted between different systems. In the netrin/DCC/Unc5h signaling system [12], netrin is expressed at the midline beneath the central canal at the point at which CST axons decussate. DCC and Unc5h are netrin receptors responsible for axon attraction and repulsion, respectively. In DCC mutants, CST axons are not attracted by the midline netrin signal so the axons do not make the dorsal turn at the decussation and all the CST axons remain within the ventral spinal cord. In Unc5h3 mutants, some CST axons stay ventrolaterally, whereas others can turn dorsally and cross the midline. However, in contrast to what we have observed in PLXA3/PLXA4-/- mice, those Unc5h3 mutant axons that cross the midline do not target the dorsal funiculus, but enter the dorsal gray matter instead. Thus, the netrin/DCC/Unc5h signaling system seems to mainly control the dorsal turning of CST axons at the pyramidal decussation and the proper targeting of CST axons to the dorsal funiculus.The roles of the Ig superfamily signaling system in regulating the dorsal turning and midline crossing of CST axons are diverse [42]. In young NCAM mutant mice [14], many CST axons fail to turn dorsally and remain in the ventrolateral spinal cord. Among the mutant axons that make the dorsal turn at the pyramidal decussation, many fail to cross the midline and instead project to the ipsilateral dorsal funiculus. However, the abnormal CST axons are absent in adult NCAM mice, suggesting either a correction or loss of aberrant fibers. In adult L1 mutants [13], all CST axons turn dorsally at the pyramidal decussation, but many of them stay ipsilateral as they project to the dorsal funiculus. Therefore, the Ig superfamily signaling system seems to control both dorsal turning and midline crossing of the CST axons. It is interesting to note that the L1 subfamily of Ig molecules, including L1, NrCAM, and CHL1, also interact with neuropilins to mediate the signals from secreted semaphorins [25,26,43,44]. In vitro evidence has suggested that Sema3A signaling through an L1/NPN-1 complex contributes to midline crossing of CST axons at the pyramidal decussation [25]. However, in vivo analysis of the Sema3A-/- mouse by our lab and others [36] indicates no defects in dorsal turning or midline crossing of the CST in this mutant. We also show that, in contrast to L1 mutant mice, all the CST axons cross the midline in NPN-1-/- mice even though they are defasciculated. These results suggest NPN-1 and L1 function independently in regulating CST guidance at the pyramidal decussation. Recently, CHL1 has been shown to function together with NPN-1 to mediate the guidance of thalamocortical axons in vivo [44]. It would be interesting to test whether CHL1 is also involved in CST axon guidance.Our analysis has revealed the contributions of semaphorin/plexin signaling in the dorsal turning of motor CST axons at the pyramidal decussation (Figure 6). Specifically, we demonstrate that in the absence of PLXA3 and PLXA4, up to 50% of the motor CST axons are guided to the ventral spinal cord, resulting in an abnormal ipsilateral ventrolateral tract. The plexin-mediated CST turning defect appears to be neuropilin-independent as NPN-1-/- and NPN-2-/- mice do not display ventrolateral CST guidance defects. We also found that neither Sema3A-/- nor Sema3E-/- mice had such defects. These results indicate that the local environmental cues that act at the pyramidal decussation to direct plexin-mediated dorsal turning of motor CST axons are membrane-bound semaphorins. In support of this, we found that Sema6A-/- mice had a similar motor CST guidance defect to PLXA3/PLXA4-/- mice in which the majority of axons stayed ipsilateral and formed a ventrolateral tract.Figure 6Summary of CST axon guidance in WT and knockout animals and a model for plexin signaling in CST axon turning. (a) Diagrams of cross sections of brainstem and spinal cord summarizing CST axon guidance defects observed in adult mutant mice. Ventrolateral CST axons were only observed in plexin and Sema6A mutants. The Sema6A-/- phenotype was relatively diverse between animals. Summarized in this diagram are the major defects. (b) Model for motor CST axon turning at the pyramidal decussation (refer to Discussion). C, caudal; D, dorsal; Pyr Dec, pyramidal decussation; R, rostral; SpC, spinal cord; V, ventral.Several recent reports have nicely addressed the interactions between class 6 semaphorins and plexin-A family members. Specifically, it has been shown that PLXA4 directly binds Sema6A, and their interactions in vivo are important for the lamina-specific projection of mossy fibers in the hippocampus [39] and for the short-range repulsion of developing sympathetic axons [38]. In addition, it has been shown that Sema6A binds PLXA2 [45], which is also expressed in the motor cortex during CST axon guidance (data not shown). However, analysis of the PLXA2 mutant mouse revealed no defects in CST axon guidance (KJ Mitchell, personal communication). We have previously shown that PLXA3 and PLXA4 are co-expressed in neuronal tissues to mediate axon repulsion, axon pruning and neuronal migration [46-49], but these functions are mostly activated by secreted semaphorins. Here, our phenotypic analysis in mutant mice suggests that PLXA3 and PLXA4 may function with membrane-bound Sema6A in vivo. However, it is still unclear whether PLXA3 can directly bind to Sema6A, and how PLXA3 and PLXA4 interact to mediate Sema6A signals. It is also important to note that the CST guidance defects in Sema6A-/- mice are apparently more diverse and more severe than in PLXA3/PLXA4-/- mice. Although no apparent CST guidance defects were noted before axons reached the hindbrain in PLXA3/PLXA4-/- mice, guidance defects have been noted at the midbrain-hindbrain boundary in Sema6A-/- mice (KJ Mitchell, personal communication). This defect higher up in the CST projection pathway may account for the more severe defect in CST guidance across the pyramidal decussation seen in some Sema6A-/- mice. It is now apparent that CST axons are guided by specific signals at different choice points to reach their distant targets. The phenotypic differences between Sema6A-/- and PLXA3/PLXA4-/- mice indicate that other plexin or non-plexin receptors may also be used to mediate Sema6A signals in the guidance of motor CST axons.ConclusionWe have characterized the roles of PLXA3, PLXA4, NPN-1, NPN-2, Sema3A, Sema3E, and Sema6A in regulating the guidance of motor CST axons to the dorsal spinal cord in vivo (summarized in Figure 6). We find that PLXA3, PLXA4, and Sema6A are required for the proper dorsal turning of motor CST axons at the pyramidal decussation. As motor CST axons are crossing the midline, we find that NPN-1 is required for CST axons to remain fasciculated so they may target the dorsal funiculus appropriately. However, PLXA3 and PLXA4 are either compensated for by other receptors in this process or not required. We also find that the dorsal turning and midline crossing of motor CST axons are normal in NPN-2, Sema3A, and Sema3E mutants. Although many questions remain, it is evident that semaphorin signaling is one of several signaling systems that coordinate at specific points along the pathway to properly guide the long CST axons from the cerebral cortex to the spinal cord.Materials and methodsMouse breedingAnimal protocols were approved by the Institutional Animal Care and Use Committee at UC Davis. Genotyping on knockout mice was carried out as described previously [34,35,49-52]. NPN-1sema-/- mice were obtained from Jackson Laboratories (Bar Harbor, ME, USA). Sema3A-/- mice were a generous gift from Mark Fishman and Marc Tessier-Lavigne. Sema3E-/- and Sema6A-/- mice were a generous gift from Marc Tessier-Lavigne.Mouse tracer injectionsWild-type and mutant mice were injected with various tracers at different postnatal ages (P0 to P45). DiI (Molecular Probes, Carlsbad, CA, USA) and BDA (Molecular Probes) anterograde tracing was performed as described previously [53,54]. Mice were injected blindly prior to determining genotype. Briefly, DiI (20% in N,N-dimethylformamide) or BDA (10–20% in phosphate buffered saline) were injected focally in the motor cortex of WT and mutant mice in vivo and allowed to trace for a minimum of three days. Locations of the injection sites were confirmed in sagittal sections of the cortex to ensure tracers were injected in the appropriate regions of the cortex.Immunohistochemistry, in situ hybridization, and EM processingImmunohistochemistry was performed on floating sections as described previously [55]. Antibodies and concentrations used in the study were: CamKIIα (1:1,000; Chemicon, Temecula, CA, USA), Ctip2 (1:1,000; Abcam, Cambridge, MA, USA), and L1 (1:1,000; Chemicon). The plexin and neuropilin probes for in situ hybridization and the procedures for radioactive α-33P in situ hybridization were as described previously [51,56]. The procedure for non-radioactive in situ hybridization was as described previously [51]. Sections that contained BDA-labeled CST axons were preserved for ultrastructural analysis with EM as described [55].Analysis of CamKIIα immunostained spinal cord sectionsTransverse sections of CamKIIα-immunostained at the level of the pyramidal decussation or cervical spinal cord were selected for analysis. Raw images of the sections were digitally captured with a CCD camera (Zeiss, Thornwood, NY, USA) and imported into PhotoShop (Adobe Systems, San Jose, CA, USA). For quantification of CST area in the dorsal funiciulus, images were cropped and only the dorsal funiculus area was preserved for further analysis. Grayscaled images were thresholded to 30% above background levels as described [57]. Pixels that were above threshold were considered as positive labeling and these areas were measured using Image J (NIH, Bethesda, MD, USA). Positively labeled areas were subsequently normalized to the total area of the dorsal funiculus. For quantification of fasciculation at the pyramidal decussation, the width of the pyramidal decussation was measured in all available brainstem sections containing it.Statistics for all data were obtained from Statistica 6.0 (Statsoft, Tulsa, OK, USA) or Microsoft Excel with a Benjamini and Hochberg correction for multiple comparisons.AbbreviationsBDA: Biotinylated dextran amine; CamKIIα: alpha calcium/calmodulin-dependent protein kinase type II; CNS: Central nervous system; CST: Corticospinal tract; DiI: 1,1'-dioctadecyl-3,3,3',3'-tetramethylindocarbocyanine perchlorate; E: Embryonic day; Ig: Immunoglobulin; NPN: Neuropilin; P: Postnatal day; PLX: plexin; WT: Wild type.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsLKL and HJC initiated the project. RLF, LKL, XBL, EGJ, and HJC discussed and designed the experiments. RLF, LKL, XBL, and JC performed the experiments and analyzed the data. RLF, LKL, and HJC wrote the paper.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2532993.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2532993",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2532993\nAUTHORS: Evandro SF Coutinho, Astrid Fletcher, Katia V Bloch, Laura C Rodrigues\n\nABSTRACT:\nBackgroundFracture after falling has been identified as an important problem in public health. Most studies of risk factors for fractures due to falls have been carried out in developed countries, although the size of the elderly population is increasing fast in middle income countries. The objective of this paper is to identify risk factors for fall related to severe fractures in those aged 60 or more in a middle-income country.MethodsA case-control study was carried out in Rio de Janeiro-Brazil based general hospitals between 2002–2003. Two hundred-fifty hospitalised cases of fracture were matched with 250 community controls by sex, age group and living area. Data were collected for socio-demographic variables, health status and drugs used before the fall. A conditional logistic regression model was fitted to identify variables associated with the risk of fall related severe fracture.ResultsLow body mass index, cognitive impairment, stroke and lack of urine control were associated with increased risk of severe fall related fractures. Benzodiazepines and muscle relaxants were also related to an increased risk of severe fractures while moderate use of alcohol was associated with reduced risk.ConclusionAlthough the association between benzodiazepines and fractures due to fall has been consistently demonstrated for old people, this has not been the case for muscle relaxant drugs. The decision to prescribe muscle relaxants for elderly people should take into account the risk of severe fracture associated with these drugs.\n\nBODY:\nBackgroundAbout one third of the population aged 65 or more suffers at least one fall a year, of which 5 to 10% result in severe injuries [1-4]. More than 90% of hip fractures are the result of a fall [5,6]. Falling and the frequency of falls increases exponentially with age [1,3-5,7]. Injuries resulting from falls incur costs for health providers, social services, patients and their families [6-8].An active research agenda exists with a focus on prevention [2]. Cognitive impairment, low body mass index and certain medications such as benzodiazepines have been consistently associated with severe injuries from falls [5,9-12]. Data on the proportion of fall-related injuries attributable to each of these factors is sparse and it is likely that their relative contribution varies from one setting to another.Most studies of risk factors for fractures due to falls have been carried out in developed countries, although the size of the elderly population is increasing fast in middle income countries. In Brazil the proportion of people over 60 doubled from 4.1% in 1940 to 8.6% in the year 2000, and it is expected to reach 14% in 2025 [13-15]. Little is known about the frequency, circumstances, risk factors and consequences of falls in Brazil. Perracini & Ramos [16] examined risk factors for any kind of fall in a cohort of elderly community residents and Rozenfeld et al [17] did a cross-sectional investigation in a recreational facility for the elderly. Coutinho & Silva [18] carried out a case-control study with hospital controls investigating the association between drugs used in the previous 24 hours and the risk of severe fracture after falling.We conducted a study to investigate a range of health related factors associated with falls leading to hospitalisation due to fractures among elderly people.MethodsDesignA case-control study was carried out in the city of Rio de Janeiro, Brazil.ParticipantsTwo hundred and fifty cases were selected from patients aged 60 or more, admitted to five state hospitals (two are university hospitals and three are funded by the local government) with a severe fracture following a fall, between 2002–2003. A fall was broadly defined as an episode in which a person came to rest on the ground or floor and severe fracture was the one leading to hospital admittance. These hospitals admitted about 50% of all cases of severe fracture in the people aged 60 or over in Rio de Janeiro and were in different geographic areas of the city. The State Health System in the city covers about 70% of the adult population.Two hundred and fifty controls were individually matched for sex, age (± 2 years) and neighborhood (residence of cases and controls in the same block). No elected case refused to participate in the study. Selection of controls was carried out using a systematic procedure starting from the case address, with the direction around the block pre-defined by chance. We did not find more than one eligible control per household. Twenty-one people who filled the criteria for being a control did not want to take part in the study and were replaced.Data collectionThe interviewers visited the hospitals everyday to look for new cases. All individuals aged 60 or more admitted to the hospital to treat a fracture were approached and those reporting a fall as the cause of the injury were asked to participate in the study. Interviews took place at hospital (cases) and home (controls) through standard questionnaires applied by trained interviewers of both sexes. All interviewers had a university degree. The questionnaires consisted of a common set of questions and information was obtained from cases, controls and relatives. These included socio-demographic characteristics, circumstances of the fall, self-reported health status before the fall, information on drug use 24 hours and 15 days before the fall (for cases) or before the home visit (for controls), height, weight, current diseases (self reported) and cognitive impairment (evaluated by an adapted translation of the \"Short Care\" [19] which was validated in Brazil by Veras et al [20]); and history of falls and fractures in the previous 12 months. The present study used the information of the drugs taken in the previous 24 hours, categorised in 21 groups [see Additional file 1].Sample sizeWithout considering the matching, a sample size of 500 individuals would allow to identify an odds ratio of 2 for an exposure of 13% among controls (confidence level = 0.95 and power = 0.80).Statistical analysisFirst, an unadjusted analysis (except for the matching variables) was carried out for all socio-demographic and health related variables using conditional logistic regression. In this level all variables with p-value less than 0.25 were selected for multivariate analysis [21]. Second, a multivariate conditional logistic regression model including those variables was fitted to the data. At this stage, variables with p-value equal or less than 0.05 were maintained in the model. Third, variables with p-value larger than 0.25 in the first stage (univariate analysis) were entered in the model and retained if their p-value were equal or less than 0.05. This last stage was carried out aiming to reduce the chance of excluding important predictors for severe fall related fractures.Variations in the magnitude of the odds ratios after removing those variables, and multiplicative interactions between drugs and clinical variables were also investigated. The statistical significance of the interaction terms was investigated comparing the models with and without the interaction term through the likelihood ratio test.We used literature-based categories for BMI and cognitive impairment for ease of interpretation.Complementary analysis comparing means used Kruskal-Wallis test as data were either asymmetrical or variances were not homogeneous.All interviewed people signed an informed consent term. The study was approved by the ethical committee of the National School of Public Health – Oswaldo Cruz Foundation.ResultsAll 250 cases and 250 controls recruited took part in the study. The interval between the fall and the interview did not exceed 48 hours, although some additional information could have been obtained latter. The great majority were women and about half the individuals were aged between 70–79 years old (Table 1). Due to matching, cases and controls had a similar distribution for age, with mean age for cases 75.5 years old (sd: 8.2) and 75.3 years old (sd: 7.7) for controls. Widowhood was the most frequent marital status for both groups, but the proportion of divorced was higher among cases than controls. The large majority of those interviewed were not living alone. More than 40% did not complete elementary education. Less than 15% were working before the fall.Table 1Distribution of socio-demographic variables among cases (n = 250) and controls (n = 250).VariableCases n (%)Controls n (%)Sex Female55 (78.0)55 (78.0)Age group (years) 60–6959 (23.6)56 (22.4) 70–79118 (47.2)127 (50.8) 80–8961 (24.4)57 (22.8) 90 and more12 (4.8)10 (4.0)Marital status married73 (29.2)82 (32.8) widowed116 (46.4)127 (50.8) divorced23 (9.2)8 (3.2) never married38 (15.2)33 (13.2)Living alone no201 (80.4)196 (78.4) yes39 (15.6)47 (18.8) Institution10 (4.0)7 (2.8)Educational level none + elementary incomplete elementary104 (41.6)110 (44.0)  .level one (about 5 years)83 (33.2)84 (33.6)  .level two (about 4 years)37 (14.8)30 (12.0) secondary (about 3 years)20 (8.0)15 (6.0) university6 (2.4)11 (4.4)Working before the fall Yes34 (13.6)28 (11.2)Seventy percent of the falls resulting in severe fracture occurred between 6:00 am and 6:00 pm, with a similar proportion in the morning and in the afternoon. Most falls took place at home (67%), and this proportion increased with age. The most commonly fractured bone was the femur (72%) followed by arm/forearm (19%). Two cases had vertebral fracture (2.7%) and eleven (4.4%) had more than one bone fractured. Ninety nine per cent of the cases had to undergo surgical procedures.Health related factors with a level of significance less than 0.25 in their univariate association with fracture (matched analysis) are presented in table 2. Increased odds ratios were observed for low BMI, low blood pressure, dizziness, diabetes, cognitive impairment, history of stroke, lack of urine control, poor vision, limit in carrying activities of daily living (ADL), fall in the previous 12 months, use of antidepressants, benzodiazepines, muscle relaxants and cerebral vasodilators while reduced odds ratios were observed for poor health status, regular use of alcohol, calcium supplement and calcium channel blockers. Most benzodiazepines were long acting, and the most frequently prescribed was bromazepan. Almost all prescribed muscle relaxants were carisoprodol.Table 2Distribution of health related variables1 among cases and controls, odds ratios2 (OR), 95% confidence intervals (CI) and p-values (cases = 250, controls = 250).VariablesCases n (%)Controls n (%)OR (95% CI)P value1BMI – kg/m2 25 or more85 (34.1)118 (47.2)reference< 0.01 20–24.9107 (43.0)104 (41.6)1.23 (0.72–2.11) less than 2057 (22.9)28 (11.2)3.31 (1.49–7.37)Health status excellent27 (10.8)43 (17.2)reference0.01 good121 (48.4)126 (50.4)1.79 (0.85–3.76) fair90 (36.0)61 (24.4)1.82 (0.76–4.37) poor12 (4.8)20 (8.0)0.31 (0.08–1.27) Dizziness70 (28.0)57 (22.8)1.33 (0.60–2.21)0.17 Low blood pressure15 (6.0)3 (1.2)5.00 (1.45–17.27)0.01 Diabetes52(20.8)39 (15.6)1.42 (0.90–2.25)0.14 Cognitive impairment364 (28.7)27 (10.8)3.64(2.02–6.58)<0.01 Stroke28 (11.2)8 (3.2)4.33 (1.78–10.53)<0.01 Lack of urine control69 (27.6)30 (12.0)3.05 (1.82–5.12)<0.01 Poor vision421 (8.8)9 (3.6)2.63 (1.16–5.88)0.02 Limited in carrying ADL5146 (58.4)113 (45.2)1.85 (1.24–2.70)<0.01Current use of alcohol not used198 (79.2)161 (64.4)reference< 0.01 less than once a week35 (14.0)44 (17.6)0.79 (0.40–1.55) at least once a week17 (6.8)45 (18.0)0.42 (0.17–1.02) Fall in the previous 12 months94 (37.8)79 (31.6)1.34 (0.91–1.98)0.14 Antidepressant68 (3.2)3 (1.2)2.67 (0.71–10.05)0.15 Benzodiazepine644 (17.6)19 (7.6)2.56 (1.44–4.57)<0.01 Ca channel blocker522 (8.8)41 (16.4)0.47 (0.27–0.84)0.01 Ca supplement5 (2.0%)12 (4.8)0.42 (0.15–1.18)0.10 Diuretics635 (14.0)48 (19.2)0.68 (0.42–1.10)0.12 Muscle relaxant621 (8.4)8 (3.2)5.33 (1.55–18.30)<0.01 Cerebral Vasodilators627 (10.8)14 (5.6)2.08 (1.05–4.15)0.041. Selection criteria for this table was univariate p-value < 0.25. P-values for the 2 × n comparisons2. matched by sex, age group and neighbourhood. OR estimated using conditional logistic regression3. adapted translation of the \"Short Care\"4. unable to identify someone at the opposite side of the room5. unable to perform at least one of the following activities on his(her) own: use of public transport, drive, walk short distances, eat own meals, dress up, take medication, comb, go up and downstairs, take shower, cut nails, control urine.6. in the previous 24 hoursOsteoporosis, Parkinson disease, epilepsy, high blood pressure, use of angiotensin converting enzyme (ACE) inhibitors, antihistamines, analgesics, antiacids, alpha and beta-adrenergic blockers, nitrates, non steroidal anti-inflamatory drugs, digoxin, calcium and vitamin D supplements did not reached the pre-defined level of 0.25 significance [see Additional file 2].The average total number of drugs were 2.2 (sd = 1.47) in cases and 2.1 (sd = 1.48) among the controls. The difference was not statistically significant (p = 0.43).When variables presented in table 2 were entered in a multivariate conditional logistic model, diabetes, high blood pressure, rheumatism, poor vision and use of diuretics had significance levels over 0.05 and were dropped from the multivariate model. Table 3 presents the final model. Body mass index equal or less than 20 kg/m2, cognitive impairment, previous stroke and lack of urine control were associated with increased incidence of severe fall related fractures while use of alcohol at least once a week was associated with reduced incidence. Concerning the use of drugs in the previous 24 hours, benzodiazepines and muscle relaxants were related to an increased risk of severe fractures while calcium channel blockers were associated with a reduced risk. The highest odds ratio (approximately 5 fold) was observed for use of muscle relaxants and history of stroke, although confidence intervals were large. No effect modification was observed for the variables included in the final model.Table 3Association of health related variables and severe fall related fractures. Adjusted1 odds ratios (OR), 95% confidence intervals (CI) and p-values (cases = 250, controls = 250)VariablesOR (95% CI)P value2BMI 25 or morereference 20–24.91.18 (0.71–1.96)0.63 less than 203.43 (1.64–7.17)<0.01 Cognitive impairment2.19 (1.09–4.41)0.03 Stroke5.27 (1.31–21.20)0.02 Lack of urine control3.16 (1.42–7.03)<0.01Current use of alcohol not usedreference less than once a week0.71 (0.38–1.33)0.29 at least once a week0.40 (0.18–0.89)0.02 Benzodiazepine2.22 (1.07–4.58)0.03 Ca channel blocker0.40 (0.19–0.86)0.02 Muscle relaxant4.42 (1.02–19.21)0.041. Each variable is controlled for the other ones in the table plus self-reported health status and activities in daily living using conditional logistic regression.2. P-values for each variable category compared to the reference level.Model fit statistics: McFadden's R2 = 0.297, McFadden's Adj R2 = 0.206, Count R2 = 0.777, Likelihood-ratio χ2 (14 df) = 91.485, p < 0.001DiscussionMost of the cases of severe fracture due to falling in the present study were female. Most fell at home between 6:00 am and 6:00 pm and the great majority of the fractures affected the femur and the arm/forearm. Risk factors identified were low body index mass, cognitive impairment, stroke, lack of urine control, use of benzodiazepine and muscle relaxants.Our study in a middle-income country setting agrees with some previously reported associations with fracture due to fall observed in developed countries: low body index mass [9,22,23], cognitive impairment [5,22], stroke [9,10], lack of urine control [24-26].In the present study regular users of alcohol (\"at least once a week\") were at a reduced risk of severe fractures. Findings for alcohol have not been consistent. Peel et al [27] found a reduced risk of fall-related fracture for moderate alcohol intake while other authors have found the reverse: higher risk of falls leading to fracture associated with higher use of alcohol [10,28]. \"J\" shape patterns in alcohol use have been observed for outcomes such as cardiovascular disease, with non drinkers or those with very low intakes having higher risks compared to those with mild or moderate drinking patterns. In our study, those who reported using alcohol at least once a week rarely made use of alcohol more than twice a week. It is possible that people drinking at least once a week were healthier than those not drinking. Although we attempted to control for this by the inclusion of other health related variables in the model, we cannot exclude the possibility of residual confounding.As in our investigation, previous studies have identified an effect of benzodiazepines on the risk of fall-related fractures [12]. Hartikainen et al [29] carried out a systematic review of 29 studies that reported the association between the use of medicines and the risk of falls or fall-related fractures among people aged 60 or more. Nine of them were case-control studies matched by sex and age like our investigation. The authors concluded that central nervous system drugs, mainly psychotropic drugs, were associated with an increased risk of these accidents,Many drugs commonly used by elderly people have not been systematically studied as risk factor for falls [29]. An important and novel result from our study was the association between the use of muscle relaxants in the last 24 hours and severe fracture due to falling. The odds ratio was very high (OR = 4.42) although the 95% confidence interval was wide (1.02–19.21). To the best of our knowledge, there is only one study that reported this empirical association among the elderly. French et al [30] used database information to investigate the relationship between registered primary diagnosis of fracture and previous use of some drugs. The authors found that those registered with fracture were prescribed muscle relaxants 1.4 times more than controls (those with non-specific chest pain). This value is much lower than the one we found, but it is difficult to compare these findings as the study designs were quite different.That muscle relaxants can cause falls is biologically plausible: these drugs are recognized to cause weakness, drowsiness, sedation and anticholinergic effects [31]. Data on the use of muscle relaxants by elderly people, especially for extended periods, are limited but such studies that have been done reported usage by: 3% of the 60 and over population in Rio de Janeiro-Brazil [32], 0.77% of the 60 and over population in the USA [33], and 1.2% of the 75 and over age group in Finland [34]. Although muscle relaxants are recommended for short-term treatment of back pain, Dillon et al [33] reported a mean length of use of 2.1 years in the USA; 44.5% of users referred use for more than a year. Although it is generally acknowledged that the use of muscle relaxant may be inappropriate and hazardous in the elderly [31,35], the figures quoted above show that their use and their long term use remains a problem. It is likely that usage figures will be higher in places where there is easy access to medications over the counter, commonly in low and medium income countries such as Brazil. The 2002 criteria for potentially inappropriate medication use in older adults [31] does not mention explicitly the risk of falling and suffering a fracture in its evaluation of miorelaxants; the only group of drugs for which concern with falls is mentioned is long acting benzodiazepinesIn contrast to previous studies we did not find a significant association between visual impairment [9,16,36,37] and diabetes [28,38] with fall related fracture. We did not measure visual acuity but relied on self report and there may have been under-reporting leading to dilution of effect. In the case of diabetes, finding an association with falling may be influenced by the proportion of those with neurological and foot problems. Ottenbacher et al [38] found that the association between diabetes and hip fracture particularly for those taking insulin.Our study showed an unexpected inverse association between the use of calcium channel blockers (CCB) and the occurrence of severe fall related fracture. Two systematic reviews [29,39] did not find any association between these variables. We cannot exclude the possibility that this finding was due to residual confounding of self-reported health status. CCB and angiotensin converting enzyme inhibitors (ACEI) were the most reported antihypertensives. Although the proportion of controls taking CCB and ACE was the same (16%), the average total number of drugs referred by the first group was 3.0 while by the second group it was 3.5 (p = 0.02). This suggests that users of CCB could be healthier than those to which ACEI were prescribed.The study had some limitations. Most variables were self reported and, in some of the interviews, information was provided or added by relatives that were in the hospital (for cases) or at home (for controls). This could lead to an unknown degree of misclassification of exposures. Moreover, cognitive impairment was evaluated after the fall, and we cannot be sure about the influence of the accident on mental state.On the other hand, our study has some strengths. There were no refuses among cases and only few among controls and ascertainment of cases was likely to be high as severe fracture will be hospitalised. Controls were selected from same population as cases. Moreover, the study was done in a low income population from a middle-income country, a setting rarely reported for studies on fall related fracturesConclusionWhat causes falls and fractures in the elderly is an important question as the size of the elderly population is increasing fast in middle income countries. Our study identifies some similar factors and a few differences, including an important role for miorelaxant drugs which are prescribed over the counter in many countries. We studied fractures leading to hospitalisation and it is possible that muscle relaxants are also associated with less severe falls. Urgent and immediate attention is necessary to confirm and quantify the risks as to inform increased control of these drugs in elderly people.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsESFC had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: ESFC, LCR and AF. Acquisition of data: ESFC. Analysis and interpretation of data: ESFC, LCR and AF. Drafting the manuscript: ESFC, LCR, AF and KVB. Statistical analysis: ESFC and LCR. Study supervision: ESFC. All authors read and approved the manuscript.Pre-publication historyThe pre-publication history for this paper can be accessed here:Supplementary MaterialAdditional file 1Categories and ATC codes for the drugs used in the last 24 hours. The table provided presents the Anatomical Therapeutic Chemical Code for the drugs investigated.Click here for fileAdditional file 2Variables with p-value greater than 0.25 (univariate analysis). The table provided presents the odds ratios and 95% confidence intervals for the variables that did not reach statistical criteria in the preliminary analysis for inclusion in the multivariate analysis.Click here for file\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2533007.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2533007",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2533007\nAUTHORS: Maria De Luca, Michelle Moses Chambers, Krista Casazza, Kerry H Lok, Gary R Hunter, Barbara A Gower, José R Fernández\n\nABSTRACT:\nBackgroundThe objective of the present study was to map candidate loci influencing naturally occurring variation in triacylglycerol (TAG) storage using quantitative complementation procedures in Drosophila melanogaster. Based on our results from Drosophila, we performed a human population-based association study to investigate the effect of natural variation in LAMA5 gene on body composition in humans.ResultsWe identified four candidate genes that contributed to differences in TAG storage between two strains of D. melanogaster, including Laminin A (LanA), which is a member of the α subfamily of laminin chains. We confirmed the effects of this gene using a viable LanA mutant and showed that female flies homozygous for the mutation had significantly lower TAG storage, body weight, and total protein content than control flies. Drosophila LanA is closely related to human LAMA5 gene, which maps to the well-replicated obesity-linkage region on chromosome 20q13.2-q13.3. We tested for association between three common single nucleotide polymorphisms (SNPs) in the human LAMA5 gene and variation in body composition and lipid profile traits in a cohort of unrelated women of European American (EA) and African American (AA) descent. In both ethnic groups, we found that SNP rs659822 was associated with weight (EA: P = 0.008; AA: P = 0.05) and lean mass (EA: P= 0.003; AA: P = 0.03). We also found this SNP to be associated with height (P = 0.01), total fat mass (P = 0.01), and HDL-cholesterol (P = 0.003) but only in EA women. Finally, significant associations of SNP rs944895 with serum TAG levels (P = 0.02) and HDL-cholesterol (P = 0.03) were observed in AA women.ConclusionOur results suggest an evolutionarily conserved role of a member of the laminin gene family in contributing to variation in weight and body composition.\n\nBODY:\nBackgroundAs the prevalence of obesity and its related co-morbidities continue to increase worldwide [1], there is considerable effort being devoted to identify genetic pathways and mechanisms that control fat storage. To gain insights into the genetic basis of natural variation in fat storage, we have used D. melanogaster as a model system. Like mammals, insects store fat as TAG in neutral lipid droplets that are accumulated in the fat body, the functional equivalent of both mammalian liver and white adipose tissue. Drosophila shares many of the components of TAG biosynthesis, degradation, and regulation with mammals, including many of those implicated in human lipodystrophies, diabetes, and obesity [2]. D. melanogaster has proven to be an important model system to identify genetic loci that contribute to variation in quantitative traits, including lipid metabolism [3,4]. In contrast to rodent and human mapping efforts where high-resolution mapping is constrained by intensive labor demands and expense [5], in Drosophila the transition from chromosomal regions [quantitative trait loci (QTL)] identified by recombination mapping to candidate genes [quantitative trait genes (QTGs)] is made possible through the use of quantitative complementation (QC) tests with deficiency and mutant stocks [4,6]. This approach has been highly effective for identifying genetic loci within QTL that contribute to variation in several Drosophila traits, including low heritability traits such as olfactory behavior and life-span [4,7]. The QC test has been also used in mice to investigate the effect of a mutation of the Rgs2 gene on anxiety behaviors [8]. Recently, Drosophila deficiency mapping has been greatly enhanced by the release and availability of the DrosDel and Exelixis deficiency stocks in which all deficiencies occur in the same genetic background and have molecularly defined breakpoints [9,10]. The availability of Exelixis P and piggyBac stocks with single gene insertions all in the same co-isogenic background [11] has also significantly improved our ability to identify positional candidate genes within refined QTL regions.We previously mapped multiple QTL responsible for natural variation in TAG storage using a population of recombinant inbred (RI) lines derived from two unrelated Drosophila strains, Oregon R (ORE) and Russian 2b (2b) [3]. In this study we used quantitative deficiency mapping to fine-map two of the TAG QTL, one encompassing the cytological region 27B-30D on chromosome 2 and the other encompassing 63A-65A on chromosome 3. Subsequently, we performed QC tests with single gene mutant stocks to identify four candidate genes influencing TAG levels. One of the genes identified is uncoupling protein 4c (Ucp4c), which encodes a product involved in uncoupling of oxidative phosphorylation in mitochondria [12]. Notably, two mammalian homologues of Ucp4c, ubiquitous UCP2 and skeletal-muscle-specific UCP3, have already been shown to regulate mammalian fatty acid metabolism [13]. In addition, several human population studies have reported a strong association between polymorphic variants in UCPs genes and BMI [14]. The remaining three genes are novel candidate genes affecting fat storage: CG9135, CG1399, and Laminin A (LanA). CG9135 and CG1399 belong to a family of genes of unknown function [12]. LanA encodes a protein belonging to the α subfamily of laminin chains [12]. Laminins are heterotrimeric glycoproteins present in the basement membrane matrix where they play a role in cell-matrix adhesion, migration, growth, and differentiation of various cell types [15]. While in mammals different combinations of five α, four β and three γ chains can assemble into at least 15 diverse laminins [15], Drosophila appears to use only one β, oneγ, and two α chains [12]. The Drosophila laminin A chain has significant sequence homology with mammalian laminin α5 chain [16]. In humans, laminin α5 is encoded by the LAMA5 gene, which spans approximately 78 kb on chromosome 20q13.2-q13.3 [17]. Several genome-wide linkage scans have linked this chromosomal region 20q13.2-q13.3 to variation in body mass index (BMI) and percentage body fat [18]. In addition, QTL affecting body weight and adiposity have been mapped to a region on mouse chromosome 2 that is syntenic with chromosome 20q13.2-q13.3 in humans [18]. Taken together with our results from Drosophila, these observations suggested that polymorphisms in LAMA5 contribute to natural variation in body weight and adiposity in humans. To explore this hypothesis we examined the association between genetic variants in the human LAMA5 gene and phenotypic variation in several anthropometric traits, including those reflecting body composition and lipid profile in a in a cohort of 228 unrelated EA and AA pre-menopausal women. We selected three haplotype-tagging SNPs from the International HapMap project : rs659822 (T > C) in intron 1, rs2297588 (G > A) in intron 51, and rs944895 (T > C) a non-synonymous SNP in exon 68. Our results imply that genetic variation in the LAMA5 gene affects variation in human body composition and lipid profile. However, additional genetic work and functional studies will be necessary to identify causal associations.ResultsFine mapping and identification of positional candidate genes for TAG storage in DrosophilaWe tested the effects of 38 deficiencies that span the QTL intervals at cytological regions 27B-30D and 63A-65A. After Bonferroni corrections for multiple comparisons, seven deficiencies significantly failed to complement the TAG storage phenotypes of ORE and 2b in the 27B-30D QTL region and four deficiencies in the 63A-65A QTL region (Figure 1). The combined data therefore revealed multiple sub-QTL regions each containing at least one gene affecting variation in TAG storage (Figure 1).Figure 1Quantitative deficiency mapping of D. melanogaster TAG QTL. Long ticks mark sections and short ticks mark subsections of physical maps in the cytological interval 25F5;32B3 on the left (L) arm of chromosome 2 and in the cytological interval 63A6;65E8 on the left arm of chromosome 3. Gray bars represent non-significant deficiencies and red bars correspond to deficiencies with significant failure to complement ORE and 2b QTL for TAG storage. Yellow frames indicate regions where a QTL affecting TAG content between ORE and 2b maps.To identify candidate genes affecting TAG levels we then performed QC tests using crosses of the ORE and 2b parental strains to mutants of 21 of the genes that map in the refined sub-QTL regions (see Additional File 1: Summary of quantitative complementation tests with mutants of positional candidate genes in D. melanogaster). Four of the genes tested showed a quantitative failure to complement, indicating that allelic differences between ORE and 2b strains at these loci contribute to the differences in TAG storage between the two strains: Ucp4c, CG9135, CG13993, and LanA (see Additional File 1: Summary of quantitative complementation tests with mutants of positional candidate genes in D. melanogaster).Drosophila LanA influences TAG storage, live weight, and total protein contentTo independently verify the effect of the LanA gene on TAG storage, we measured this trait in flies that were homozygous for the insertional mutation LanABG02469 and non-mutant flies from the co-isogenic control line. We also investigated the effects of the insertion on live body weight and total protein content. For male flies, we found no significant difference between mutant and controls for TAG and live body weight (Figure 2a and 2b). However, LanABG02469 male flies had slightly reduced total protein content (P = 0.0486) compared to controls (Figure 2c). On average, total protein content of LanABG02469 males was 10% lower than that of controls. While the effect on males was minimal, the LanABG02469 mutation had a dramatic effect on female traits. Females with the LanABG02469 mutation had significantly lower TAG storage (P = 0.0068), live body weight (P = 0.0092), and total protein content (P = 0.0352) compared to control flies (Figure 2a–c). The reductions in female TAG storage, live weight, and total protein content relative to control flies were 10%, 16%, and 29%, respectively.Figure 2Effects of Drosophila LanA allele on TAG storage, body weight, and total protein content. Values represent mean ± SEM of TAG storage (panel a), live body weight (panel b), and total protein content (panel c) for n = 9 independent replicates of homozygous LanABG02469 and wild-type male and female flies.Human LAMA5 variants contribute to variation in anthropometric traits, body composition, and serum lipidsTable 1 summarizes the baseline characteristics of the human cohort stratified by ethnicity. Significant differences in total fat mass (TFM), serum TAG levels, and high density lipoprotein cholesterol (HDL-C) were observed between EA and AA. EA had higher mean TFM and serum TAGs and AA had higher mean HDL-C (Table 1).Table 1Characteristics of the human subjects by ethnicityPhenotype (unit of measurement)European AmericansAfrican Americans(n = 101)(n = 127)Age (yr)34.66 ± 6.033.4 ± 5.7BMI27.56 ± 2.1827.50 ± 2.44Height (cm)165.6 ± 0.6163.9 ± 0.6Weight (kg)75.67 ± 9.173.67 ± 9.0Total fat mass (kg)32.14 ± 6.9*30.29 ± 6.7Lean tissue mass (kg)40.22 ± 3.639.78 ± 4.4Triacylglycerol (mg/dl)115.06 ± 57.3***67.3 ± 25.6Total cholesterol (mg/dl)160.24 ± 31.2155.86 ± 34.1HDL-cholesterol (mg/dl)36.14 ± 9.2***43.21 ± 10.8LDL-cholesterol (mg/dl)101.08 ± 26.799.18 ± 31.9Values are means ± SD. P values for difference between European Americans and African Americans are obtained using Student's t-test. *p < 0.05 and ***p < 0.001.The allele and genotype frequencies for each SNP are shown in Table 2. All genotype groups were in Hardy-Weinberg equilibrium. There was no difference in allele or genotype frequencies between EA and AA for SNP rs2297588. However, there was a difference in genotype frequencies between EA and AA for SNP rs944895 and rs659822 (Table 2). These differences remained when the two populations were tested for pair-wise linkage disequilibrium (LD) between SNPs. In both populations rs2297588 was in weak LD with rs659822 (EA: r2 = 0.16; AA: r2 = 0.14) and was more highly lined with rs944895 (EA: r2 = 0.38; AA: r2 = 0.35), whereas rs944895 was associated with rs659822 in the EA population (r2 = 0.37), but not in the AA population.Table 2Allele and genotype frequencies of LAMA5 rs659822, rs2297588, and rs944895 polymorphisms in the study sampleEuropean AmericansAfrican Americans(n = 101)(n = 127)rs659822rs2297588rs944895rs659822rs2297588rs944895Genotype frequencyTT0.455b***GG0.524TT0.416b**TT0.229GG0.551TT0.260TC0.416GA0.446TC0.495TC0.543GA0.417TC0.520CC0.129AA0.030CC0.089CC0.228AA0.032CC0.220Allele frequencyT0.663b*G0.748T0.664T0.501G0.760T0.520C0.337A0.252C0.336C0.499A0.240C0.480HWEa0.6580.1170.3820.4750.1490.727aP values of Hardy Weinberg equilibrium (HWE) tests. bComparison between racial groups; *p < 0.05, **p < 0.01, ***p < 0.001.We next examined whether the SNPs were independently associated with each trait. As significant differences in genotype frequencies were observed for SNP rs944895 and rs659822 between the two populations, we tested the association between SNPs and each trait using the data separately by ethnicity. Age, genetic admixture, and appropriate potential confounding variables were included in the analysis as covariates. While no association was found between SNP rs2297588 and any of the traits (data not shown), we did find significant associations between SNP rs659822 and height, body weight, TFM, and lean tissue mass (LTM) in EA (Table 3) assuming a model of additive effects. When we fit the data to a recessive genetic model we also found a significant association between SNP rs659822 and HDL-C in EA (Table 3). On average, EA women that are homozygous for the C allele had short stature, lower mean body weight, TFM, and LTM than those homozygous for the T allele (Table 3; P < 0.05). EA women that were homozygotes for the C allele also had higher levels of HDL-C than those carrying at least one T allele (Table 3; P < 0.05). The association between SNP rs659822 and variation in weight and LTM was also observed in AA women (Table 3). In this case, however, AA women homozygous for the SNP rs659822 C allele had higher mean weight and LTM than those homozygous for the T allele or heterozygous (Table 3; P < 0.05). Finally, significant associations were observed in AA women between alternative alleles at rs944895 and variation in serum TAG levels and HDL-C assuming the additive and dominant models, respectively (Table 3). On average, AA women homozygous for the T allele had lower TAG levels than those homozygous for the C allele and lower HDL-C than those carrying at least one C allele (Table 3; P < 0.05).Table 3Mean ± SEM for anthropometric measures, body composition, and serum lipid profile of study subjects stratified according to LAMA5 rs659822 or rs944895 genotype and ethnicity.European AmericanAfrican Americanrs659822C/CC/TT/TP*C/CC/TT/TP n134246296929 BMI26.8 ± 0.527.6 ± 0.327.7 ± 0.30.128.2 ± 0.427.2 ± 0.327.4 ± 0.40.3 Height (cm)162.6 ± 1.2164.6 ± 0.8167.3 ± 1.10.02164.9 ± 1.2164.1 ± 0.8162.5 ± 1.40.2 Weight (kg)69.8 ± 2.374.7 ± 1.378.2 ± 1.40.00876.5 ± 1.773.2 ± 1.171.9 ± 1.50.05 Total fat mass (kg)28.8 ± 1.931.5 ± 1.033.7 ± 1.00.0131.3 ± 1.230.3 ± 0.929.4 ± 1.10.3 Lean tissue mass (kg)37.6 ± 0.839.9 ± 0.541.2 ± 0.50.00341.4 ± 0.839.5 ± 0.538.7 ± 0.80.03 Triacylglycerol (mg/dl)104.9 ± 11.5120.5 ± 11.4112.9 ± 6.30.868.1 ± 4.765.8 ± 3.270.3 ± 4.30.6 Total cholesterol (mg/dl)153.4 ± 6.3163.0 ± 4.9159.6 ± 4.90.8162.3 ± 6.5150.1 ± 3.8163.1 ± 6.90.6 HDL-cholesterol (mg/dl)42.4 ± 2.436.0 ± 1.634.5 ± 1.10.003a46.1 ± 1.841.2 ± 1.345.2 ± 2.10.07a LDL-cholesterol (mg/dl)90.0 ± 5.8102.9 ± 4.1102.5 ± 4.10.5102.7 ± 6.295.8 ± 3.6103.8 ± 6.40.9rs944895 n95042286633 BMI26.6 ± 0.727.5 ± 0.327.8 ± 0.30.327.8 ± 0.427.6 ± 0.327.1 ± 0.40.5 Height (cm)163.1 ± 1.2165.6 ± 1.0166.0 ± 0.90.3163.9 ± 1.4164.4 ± 0.8163.1 ± 1.20.6 Weight (kg)70.9 ± 2.875.5 ± 1.376.9 ± 1.30.0973.5 ± 1.474.5 ± 1.272.1 ± 1.60.7 Total fat mass (kg)28.2 ± 2.131.7 ± 1.033.3 ± 1.00.129.9 ± 1.130.7 ± 0.929.7 ± 1.20.5 Lean tissue mass (kg)38.5 ± 1.140.5 ± 0.640.2 ± 0.50.740.0 ± 0.840.2 ± 0.538.8 ± 0.90.4 Triacylglycerol (mg/dl)101.8 ± 8.6117.3 ± 9.2115.2 ± 8.10.978.9 ± 5.165.7 ± 3.160.9 ± 4.00.02 Total cholesterol (mg/dl)159.8 ± 7.5161.2 ± 4.6159.2 ± 4.90.7156.5 ± 5.6159.4 ± 4.2148.3 ± 6.50.7 HDL-cholesterol (mg/dl)41.2 ± 3.135.1 ± 1.436.3 ± 1.30.5b43.4 ± 2.145.2 ± 1.439.1 ± 1.50.03b LDL-cholesterol (mg/dl)98.2 ± 8.6102.7 ± 3.899.8 ± 4.10.797.4 ± 5.1101.1 ± 4.097.0 ± 6.11* P values represent the significance of the comparison among genotypes. P values without subscript were calculated assuming additive models. P values with subscripts (a) and (b) were calculated assuming recessive and dominant models, respectively. P values significant after permutations tests to correct for multiple comparisons are highlighted in bold caseExcept for the association between SNP rs659822 and weight in AA women, all the single-marker associations remained significant at an experiment-wise P = 0.05 after allowing for multiple testing by permutation analysis [19]. Pair-wise haplotype-based association analyses did not increase the power of these associations (data not shown).DiscussionWe performed quantitative deficiency mapping to dissect two previously identified QTL regions influencing variation in TAG storage among a set of RI lines established from two strains of D. melanogaster, ORE and 2b [3]. The fine mapping revealed that the two QTL broke down into multiple sub-QTL regions (Fig. 1). This indicates that the number of loci influencing variation in TAG levels among these RI lines is much greater than the number suggested by the initial QTL mapping. This finding is consistent with other studies that have fine-mapped QTL for several quantitative traits in Drosophila and mice [4,20,21] and corroborates the complexity of the genetic architecture of quantitative traits.Several QTL have been associated with BMI, body weight, fat mass, and fat-free mass in human linkage studies [18]. If the complexity observed in model systems turns out to be a common phenomenon also in human traits, then the identification of the genes underlying variation in these traits will remain a challenge. There is increasing evidence that genome-wide association (GWA) studies are a powerful method for identifying genes involved in human complex traits [22]. Taking advantage of the block-like patterns of LD that characterize the human genome [23], these studies rely on the use of hundreds of thousands of \"tagging\" markers that can capture a significant proportion of the genetic variation and provide power to detect associations. However, one limitation of this approach is that linkage of the markers with variants in a number of genes in the block can make it difficult if not impossible to identify the casual variant affecting the trait. In addition, because of the well-known context dependency of allelic effects of QTL on quantitative traits (e.g. epistasis and genotype by environment interaction) [4], association studies in controlled environments and defined genetic background will potentially allow a more detailed picture of the complexity of the genetic architecture of quantitative traits than that provided by human studies. Studies using D. melanogaster and other model systems will continue to play an important role in pinpointing potential candidate loci affecting quantitative traits.Using QC tests to mutants of positional genes, we identified Ucp4c, CG9135, CG13993, and LanA as candidate loci that influence variation in TAG storage between ORE and 2b. Notably, three of the implicated loci, Ucp4c, CG9135, and CG13993, are tightly linked, with CG9135 and CG13993 being only 9 kb apart [12]. These however represent only a fraction of the genes underlying the QTL effects identified in this study. Together, there are 286 genes currently mapped in the refined QTL regions. Mutant stocks for 99 of these genes are available from the Drosophila stock center. Many of these mutants are in different genetic backgrounds making it difficult to distinguish allelic effects on a trait at the tested locus from epistatic effects with genetic background. In this study we therefore chose to focus only on loci with mutations in the same genetic background of their controls. QC tests to all available mutations of the genes mapping within the refined regions are underway.Our studies in flies further suggest that the effect of Drosophila LanA is not limited to TAG storage, but it extends to body weight and whole-body protein content. The observed result on body weight is interesting since three-week old mice homozygous for a hypomorphic mutation in the LAMA5 gene, the mammalian homolog of LanA, have smaller size than their controls [24]. Here we report that natural variation in this gene may contribute to the underlying variation in these traits in human populations. We identified a significant association between a T/C variant in the human LAMA 5 intron 1 (rs659822) and height, body weight, TFM, LTM, and HDL-C in EA women. EA women homozygous for the less frequent variant (CC) on average had lower body weight, TFM, and LTM than those homozygous for the T allele. The effect of SNP rs659822 on weight and LTM was also observed in AA women. In this case, however, AA women that were homozygous for CC at this SNP had higher weight and LTM than women homozygous TT. The opposite effect of rs659822 genotypes on body weight and LTM in the two ethnic groups is intriguing and might be explained by the complexity of the processes that determine variation in these traits, including allelic epistatic interactions within the LAMA5 gene and interactions with other genes and with the environment. In this regard it is important to point out that the genotype frequencies of SNP rs659822 were significantly different across these two groups, with the frequency of the genotype CC being significantly lower in EA women than in AA (Table 3). This sensitivity of the allelic effects of SNPs on phenotypic traits has also been observed in disease-marker studies [25] and is implied in QTL studies in plants [26], Drosophila [27], and mice [28] that show significant differences in allelic effects on phenotypes depending on the genetic background in which they occur. Lin et al. used theoretical modeling to demonstrate that such \"flip-flop\" associations can occur because the lack of consideration of other genetic loci or environmental factors that influence complex traits [25]. They argue that this is particularly important when a non-causal genetic variant that is linked with the causal polymorphism is investigated [25]. Because genotypes of all polymorphic sites in the LAMA5 gene were not determined and SNP rs659822 is located in an intron, it is possible that rs659822 is not itself the causal polymorphism, but is in LD with the true causal polymorphism somewhere else in this gene. In our results SNP rs659822 was in weak LD with both rs2297588 (r2 = 0.16) and rs944895 (r2 = 0.37) in EA and both SNPs were not associated with any of the traits in this ethnic group. This observation suggests that SNP rs659822 is the site with the largest association with the true causal polymorphism. An overview of the pattern of linkage disequilibrium across the LAMA5 gene established by the HapMap Project in the CEU population of northern and western European ancestry from Utah showed that SNP rs659822 is in strong LD (r2 = 0.73) with a non-synonymous A to G variant in exon 47 (rs2274934) that could be the responsible polymorphism. Notably, this SNP converts the neutral amino-acid asparagine to the negatively charged amino-acid aspartate in one of the laminin EGF-like domains, which have been suggested to act as signals for cellular growth and differentiation [29]. A change in the amino acid structure of this laminin EGF-like domain might explain our finding that variation in LAMA5 associates with a pleiotropic effect on both anthropometric traits and body composition.We also identified a significant association between a non-synonymous T to C variant in the exon 68 (rs944895) that converts a tryptophan to an arginine in the laminin G (LG)-like 2 domain and variation in serum TAG levels and HDL-C and in AA subjects. AA women homozygous for the less frequent variant (CC) on average had lower serum TAG and HDL-C levels than those homozygous for the T allele. Laminin LG modules have been implicated in interactions with cellular receptors and other extracellular ligands, such as heparan sulfate proteoglycans (HSPGs) [30]. Interestingly, consistent evidence exists that HSPGs play a role in the turnover of lipoproteins, including the uptake of HDL-C in liver [31]. Moreover, cell surface HSPGs contribute to intracellular TAG accumulation in adipocytes [32]. Studies examining the functional effect of rs944895 polymorphism in lipoprotein metabolism will be necessary to understand the mechanisms underlying our findings.One limitation of our study is that the human association component involved a fairly small sample size and was restricted only to women. This cohort was chosen because measurements of genetic admixture were available for each individual, which allowed us to adjust for ancestry within ethnic groups and, therefore, limit false-positive results [33]. The human data set was also chosen because it provided detailed measurements of body composition for each individual. Clearly, replication of the results in other human cohorts is necessary [34], but the consistency in genetic effects of this member of the laminin gene family in both the fly and humans supports a generally conserved role for this gene in regulating traits reflecting body composition. This is particularly evident for the effect of the gene on lean tissue mass, which was not only observed in both EA and AA women, but also in male and female Drosophila.ConclusionOver the past few years, the number of chromosomal regions that contain one or more genes affecting obesity traits in humans and in mammalian models has dramatically increased [18]. Results from our study indicate that D. melanogaster may be a good model to pinpoint those genes with evolutionarily conserved effects on body composition that fall within the large chromosomal regions identified in mammalian QTL studies. Our cross-disciplinary genetic study implicates a member of the laminin gene family as a novel candidate gene affecting variation in body composition traits in natural populations. These observations motivate future studies in independent human populations to verify the effects of this gene.MethodsDrosophila deficiency and mutant complementation mappingDrosophila stocksDeficiency stocks used for the deficiency complementation mapping were obtained from the Bloomington Drosophila Stock Center . All deficiencies used in our study are from the Exelixis and DrosDel collections that have been generated in co-isogenic w1118 backgrounds [9,10].Mutant stocks were obtained from the Bloomington Drosophila Stock Center and from Trudy Mackay at NC State. Except for LanABG02469, all the mutations are DrosDel and Exelixis P and piggyBac insertions in the w1118co-isogenic background. LanABG02469 is a hypomorphic mutation generated by the insertion in the w1118;Canton S strain of a P-element that is located 339 bp upstream the coding region of the LanA gene. Flies were maintained in vials containing 10 ml of standard cornmeal, agar, sugar, and yeast medium at 25°C.Experimental design and phenotypic measurementsWe conducted the QC tests with deficiencies and mutations using ORE and 2b, the parental lines used to establish the mapping population for the recombination mapping study [3]. In our experiments, we crossed virgin females from ORE and 2b to males from each deficiency stock and to males from the w1118 strain. These crosses produce four possible genotypes: ORE/Deficiency, ORE/w1118 and 2b/Deficiency, 2b/w1118. We measured four replicate trait values of each genotypic class for each sex using the same experimental design described in [3], with the only exception being the way TAG content was measured. Briefly, we kept each genotypic class of flies in four replicate vials, each containing a group of 10 single-sexed individuals. After 4–5 days, we anesthetized each group of flies and measured live weight to 0.01 mg accuracy with an analytical balance. We then homogenized the flies using the protocol described in. We assayed TAG content spectrophotometrically using a commercially available kit (Sigma-Triglyceride Assay Kit) following the manufacturer's suggested protocol. To account for difference in body weight and total protein content, we used the live weight and total protein content as covariates in the analysis of the data. We measured total proteins for each homogenate using a standard Lowry protein assay.We conducted the QC test with mutant stocks using the same experimental design described for deficiencies, with the only exception being that nine replicate trait values of each genotypic class were measured for each sex.Statistical analysesAll quantitative complementation tests to deficiencies were carried out simultaneously. A quantitative failure of ORE and 2b QTL alleles to complement a deficiency was inferred if the difference in the mean trait value between the ORE and 2b alleles over the deficiency was significantly greater than the difference in the mean trait value of the ORE and 2b alleles over w1118 [35]. In a three-way factorial analysis of covariance (ANCOVA), these differences are indicated by a line-by-genotype (L × G) or line-by-genotype-by-sex (L × G × S) interaction terms according to the model: y = μ + L + G + S + LW + PRO + L × G + L × S + L × G × S + E, in which μ is the overall mean, L is the fixed main effect of line (ORE or 2b), S is the fixed main effect of sex, G is the fixed main effect of genotype (Def or w1118), LW and PRO are the covariates live body weight and total protein content, and E is the error term. A significant L × G × S interaction term is indicative of a sex-specific failure to complement. The deficiencies that showed a significant failure to complement were confirmed by re-testing nine replicate trait values of each genotypic class for each sex. Bonferroni corrections were performed to control for the effect of multiple comparisons.The data from QC tests with mutations were analyzed for each sex separately using the two-way factorial model of ANCOVA: y = μ + L + G + LW + PRO + L × G + E, in which μ is the overall mean, L is the fixed main effect of line (ORE or 2b), G is the fixed main effect of genotype (Def or w1118), LW and PRO are the covariates live body weight and total protein content, and E is the error term. We inferred a quantitative failure of ORE and 2b QTL alleles to complement the mutant allele if the L × G interaction term was significant. In addition, as no significant difference in TAG content was observed between the parental lines ORE and 2b[3], we also considered a significant L term as a failure to complement, if the difference between the parental strains was significant in the mutant background but not in the w1118 chromosome background [3]. Bonferroni corrections were performed to control for the effect of multiple comparisons.The statistical analyses were carried out using the SAS GLM procedures (Version 9.0; SAS Institute, 2002, Cary, NC, USA).Human studySubjectsA total of 228 European-American (n = 101) and African-American (n = 127) women were evaluated for the human association study. Subjects were participants of two ongoing longitudinal studies on the role of metabolism in the etiology of obesity conducted in AA and EA pre-menopausal women at the University of Alabama at Birmingham. Prior to testing, subjects were maintained in a weight-maintenance state for 4 weeks. During the final 2 weeks, meals were provided through the General Clinical Research Center at UAB to ensure weight stability of less than 1% variation and to maintain daily macronutrient intake in the range of 20–23% fat, 16–23% protein, and 55–64% carbohydrate. Subjects were then admitted as inpatients to the GCRC for 4 days, during the follicular phase of the menstrual cycle. All metabolic testing took place during this inpatient period. At the time of testing, subjects were sedentary (no previous history of exercise training), had a BMI range between 24 – 30 kg/m2, were nonsmokers, and were not taking any medication known to alter body composition (including hormones). Race was determined by self-reported African-American or Caucasian ancestry in both parents and grandparents.The study protocol was approved by the Institutional Review Board for human studies at the University of Alabama at Birmingham. A written informed consent was obtained from all study participants before enrolling in the study.Anthropometrical and serum lipid measurementsHeight and body weight were measured in light indoor clothes and without shoes. Blood samples were withdrawn after 12-h overnight fast. Analyses for serum lipids were performed in the Core Laboratory of the General Clinical Research Center and the Clinical Nutrition Research Center (CNRU) at UAB. Total cholesterol, HDL-C, and TAGs were measured with the Ektachem DT II System. With this system, HDL-C is measured after precipitation of low-density lipoprotein cholesterol (LDL-C) and very-low-density lipoprotein cholesterol with dextran sulfate and magnesium chloride. Control sera of low and high substrate concentration are analyzed with each group of samples, and values for these controls must fall within accepted ranges before samples are analyzed. The DT II is calibrated every six months with reagents supplied by the manufacturer. LDL-C was estimated using the Friedewald formula [36].Body compositionBody composition [TFM and LTM] was measured by dual energy X-ray absorptiometry using either a Lunar DPX-L densitometer (LUNAR Radiation Corp., Madison, WI) or a LUNAR Prodigy densitometer in the Department of Nutrition Sciences at UAB. Body composition assessed by these instruments generally differs by a coefficient of variation of 4% or less [37]. Subjects were scanned in light clothing while lying flat on their backs with arms at their sides.GenotypingTo test for associations between genetic variants in human LAMA5 and phenotypic traits, we selected three of the human LAMA5 SNPs identified by the International HapMap project. Using HapMap data release #21a, we estimated that these three SNPs captured 95.6% of common variation (Minor Allele Frequency >0.05, n = 23) at an r2 > 0.8 across LAMA5 gene. The genotypes of these polymorphisms were determined by Pyrosequencing technology [38] at the CNRU Genetics Core at UAB.To account for the confounding effects of population stratification, we used estimates of genetic admixture as a covariate in statistical models. The genetic admixture estimates were obtained from the genotyping of ancestry informative markers (AIMs) across the human genome. These AIMs are informative for parental ancestry, defined as those long-separated populations that intermixed during historical periods to produce new admixed populations. Genotyping for the measures of genetic admixture was performed at Prevention Genetics  using the McSNP method and agarose gel electrophoresis, as previously described [39]. Molecular techniques for the allelic identification and methodology for genetic admixture application have been described elsewhere[40,41]. Approximately 100 ancestry informative markers were utilized for the study. Information regarding marker sequences, experimental details, and parental population allele frequencies has been submitted to dbSNP  under the handle PSU-ANTH.Data AnalysesWe assessed Hardy-Weinberg equilibrium, estimated haplotype frequencies, and r2 linkage disequilibrium coefficients by methods implemented in Arlequin program 3.01 [42]. Allele and genotype frequency comparisons between EA and AA samples were performed by the χ2 test. To test the effect of each genotyped SNP on trait variation, we performed genotypic associations for dominant, additive, and recessive models using linear regression analysis. Age and genetic admixture were used as covariates in all the analyses. As strong correlations have been shown between fat mass and lipid profile [43], lipoprotein levels were additionally adjusted for TFM and TAGs. TAG levels were also adjusted for TFM. Dummy variables were assigned to code the three genotypes in each model. In the additive model, we used 0, 1 and 2 to code for individuals homozygous for the major allele, heterozygous, and homozygous for the minor allele, respectively. In the dominant and recessive models, we used 0 to code for individuals homozygous for the major and minor alleles, respectively, and 1 to code for individuals carrying at least one copy of the other allele. Pair-wise haplotype-based association analyses were also performed. For all regression models, studentized residuals were evaluated for normality and logarithmic transformations of the dependent variable was performed to improve normality. When normality of the residuals was not obtained after transformations, the observations that were above and below three standard deviations were removed from the analyses. To test for significant differences among means according to genotype, data from the final regression model was analyzed by analysis of variances and mean differences assessed by post-hoc Duncan tests at P < 0.05. To control for the effect of multiple comparisons, we performed permutation tests (1000 simulations) to generate empirical P values under the null hypotheses of no association between genotypes and traits [19]. All the analyses were performed using SAS (Version 9.0; SAS Institute, 2002, Cary, NC, USA).Authors' contributionsMDL conceived the study, participated in its design and coordination, carried out the Drosophila data analysis, and wrote the manuscript. MMC carried out the Drosophila complementation tests. KC carried out the human statistical analyses. JRF participated in the design and coordination of the study and the human statistical analyses. KHL carried out the human genotyping. JRF, BAG, and GRH contributed to design and acquisition of human data. BAG and JRF revised critically the manuscript. All authors read and approved the final manuscript.Supplementary MaterialAdditional file 1Summary of quantitative complementation tests with mutants of positional candidate genes in D. melanogaster. The table contains a list of all the positional candidate genes and the corresponding mutant alleles analyzed by quantitative complementation tests. In the table are also reported the cytological positions of the candidate genes and the P values for Line and Line × Genotype effects of two-way factorial ANOVAs (see text for further explanation).Click here for file\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2533008.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2533008",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2533008\nAUTHORS: Yao Huang, Zhiwei Li, Ningli Wang, Nico van Rooijen, Qi Cui\n\nABSTRACT:\nBackgroundWe recently showed that whereas inhibition of PI3K/akt or JAK/STAT pathway promoted retinal ganglion cell (RGC) survival after optic nerve (ON) injury in Fischer 344 (F344) rats, the same inhibition resulted in aggravated RGC loss after acute intraocular pressure (IOP) elevation in Sprague Dawley (SPD) rats. In addition, the responses of macrophages to ON injury and acute IOP elevation were different between F344 and Lewis rats, i.e., different autoimmune profiles. Using an acute IOP elevation paradigm in this study, we investigated 1) whether autoimmune background influences PI3K/akt and JAK/STAT functions by examining the effect of PI3K/akt and JAK/STAT pathway inhibition on RGC survival in F344 and Lewis rats, and 2) whether differential actions of macrophages occur in PI3K/akt and JAK/STAT pathways-dependent modulation of RGC survival. IOP elevation was performed at 110 mmHg for 2 hours. PI3K/akt and JAK/STAT pathway inhibitors were applied intravitreally to block their respective pathway signaling transduction. Because macrophage invasion was seen in the eye after the pathway inhibition, to examine the role of these pathways independent of macrophages, macrophages in the retina were removed by intravitreal application of clodronate liposomes. Viable RGCs were retrogradely labelled by FluoroGold 40 hours before animal sacrifice.ResultsSimilar to what was previously observed, significantly more RGCs were lost in Lewis than F344 rats 3 weeks after acute IOP elevation. As in SPD rats, inhibition of the PI3K/akt or JAK/STAT pathway increased the loss of RGCs in both F344 and Lewis rats. Removal of macrophages in the eye by clodronate liposomes reduced RGC loss due to pathway inhibition in both strains.ConclusionThis study demonstrates that following acute IOP elevation 1) PI3K/akt and JAK/STAT pathways mediate RGC survival in both F344 and Lewis rats, 2) autoimmune responses do not influence the functions of these two pathways, and 3) PI3K/akt and JAK/STAT pathway inhibition-dependent activation of macrophages is detrimental to RGCs.\n\nBODY:\nBackgroundLoss of retinal ganglion cells (RGCs) occurs in many pathological situations, and glaucoma is one of the common diseases that lead to RGC loss. A common feature of glaucoma is elevation of intraocular pressure (IOP) that causes progressive axonal degeneration and loss of RGCs. As acute glaucoma is characterised by rapid increase in IOP, acute IOP elevation paradigm in rodents has often been used to study acute IOP elevation-induced retinal ischemic/reperfusion injury and the possible mechanisms underlying acute glaucoma-associated RGC injuries [1-5].It is known that immune responses can influence neuronal survival after injury. Whereas ample evidence pointed to a damaging effect of inflammatory responses of macrophages and T-cells after CNS injury and in autoimmune diseases such as multiple sclerosis and experimental autoimmune encephalomyelitis (EAE) [6-13], T-cell dependent RGC protection in response to IOP elevation has also been reported in EAE-resistant Fischer 344 (F344) but not EAE-vulnerable Lewis rats [14]. In our earlier study, we found that macrophages, which are another major component of the autoimmune system, responded differently to IOP elevation or optic nerve (ON) injury between F344 and Lewis rats, and such differences led to different extents of RGC survival after IOP elevation or ON injury [15,16]. Macrophage activation in the eye by intravitreal injection of zymosan, a yeast wall preparation, protected RGCs after ON injury whereas the same activation aggravated RGC loss following acute IOP elevation [15-17]. In addition, under the same condition of either ON injury or acute IOP elevation the same macrophage activation resulted in different extents of RGC survival or loss between F344 and Lewis rats [16,17].Phosphatidylinositol 3-kinase (PI3K)/akt and janus kinase (JAK/STAT3) signal pathways are well known to mediate neuronal survival [18-25]. Recently we showed that ON injury or acute IOP elevation activates both pathways in the ganglion cell layer (GCL) [26-28]. However, our recent work also points to paradoxic actions of these two pathways in RGC survival under different pathological conditions. Though both ON injury and acute IOP elevation activate PI3K/akt and JAK/STAT pathways in the GCL, including RGCs, inhibition of these signalling pathways activates macrophages in the eye and contributes to RGC survival after ON injury in F344 rats [28]. However, the same pathway inhibition leads to RGC loss following acute IOP elevation in Sprague Dawley (SPD) rats [26,27]. In addition, as mentioned above, we also showed that macrophage activation in the eye by intravitreal injection of zymosan played a protective role in RGCs after ON injury [15,28]. But the same macrophage activation resulted in aggravated RGC loss in an IOP elevation model of the same F344 rat strain [17]. Macrophages thus appear to play an important role in the differences in RGC viability under the two pathological conditions.F344 and Lewis rats are inbred strains. They differ in susceptibility to EAE. It is known that the hypothalamic-pituitary-adrenal (HPA) axis modulates autoimmune response and vulnerability to EAE. In contrast to F344 rats, Lewis rats have abnormalities in HPA function. Variation in genotypes that are associated with different disease phenotypes between F344 and Lewis rats have been reported [29]. Furthermore, the greater frequency of CD8+ regulatory T cells, which functionally inhibit myelin basic protein-reactive T-cells, in F344 than in Lewis rats might contribute to the differing susceptibility to EAE between them [30]. It is currently unknown whether, after acute IOP elevation, 1) inhibition of the PI3K/akt or JAK/STAT pathway aggravates RGC loss in F344 and Lewis rats, 2) autoimmune background influences PI3K/akt and JAK/STAT signaling pathways in RGC survival, and 3) differential actions of macrophages occur in PI3K/akt and JAK/STAT pathway-dependent modulation of RGC survival. PI3K/akt and JAK/STAT pathways play an important role in mediating RGC survival following acute IOP elevation in SPD rats [26,27]. Autoimmune background is also known to modulate neuronal viability [16,17]. It is therefore important to clarify the issues as mentioned above. This study, using F344 and Lewis rats, was designed to accomplish these tasks.ResultsEffect of inhibition of PI3/akt and JAK/STAT pathways on RGC viability in F344 ratsRepresentative photomicrographs showing the appearances of FG-labelled viable RGCs (left column) and ED1+ macrophages (right column) in normal F344 and Lewis rats or 3 weeks after acute IOP elevation plus various experimental interventions are shown in Figure 1. The appearances of FG-labelled viable RGCs 3 weeks after acute IOP elevation alone in both F344 and Lewis rats were previously shown [17]. Compared with normal intact F344 (A) and Lewis (G) rats, significant loss of RGCs was seen 3 weeks after acute IOP elevation and pathway inhibition of PI3K/akt in both strains (C and I, respectively). Clodronate liposomes significantly reduced the number of macrophages in the eye of both strains and, compared with the same strain of rats not receiving clodornate liposomes, significantly enhanced the number of surviving RGCs. However, the extent of surviving RGCs after macrophage removal was still below the intact control, especially in Lewis rats. Similar observations were seen after PI3K/akt pathway inhibition by LY294002 or JAK/STAT pathway inhibition by AG490 and Jak Inhibitor I (images not shown). More surviving RGCs were seen in the central than the peripheral region. No clear change in the retinal thickness was observed before and after the acute IOP elevation [27].Figure 1Representative fluorescent photomicrographs of retinal wholemounts showing characteristics of retrogradely FG-labeled surviving RGCs (left column) and ED1+ macrophages (right column). A and B, normal F344 rat; C and D, after application of pathway inhibitor KY12420 in IOP-elevated F344 rat; E and F, after application of pathway inhibitor KY12420 and clodonate liposomes in IOP-elevated F344 rat; G and H, normal Lewis rat; I and J, after application of pathway inhibitor KY12420 in IOP-elevated Lewis rat; K and L, after application of pathway inhibitor KY12420 and clodonate liposomes in IOP-elevated Lewis rat. More surviving RGCs were seen in F344 than Lewis rats after IOP elevation and pathway inhibition (C versus I). After removal of macrophages by clodronate liposomes, the numbers of surviving RGCs increased in both F344 (E versus C) and Lewis (K versus I) rats. Similar results were obtained after inhibition of JAK/STAT pathway (images not shown). Scale bar = 50 μm.The average numbers ± SEM of surviving RGCs and ED1+ macrophages were 2356 ± 125/retina and 11 ± 2/retina, respectively, in intact F344 rats (n = 4). The average numbers ( ± SEM) of surviving RGCs in the retinas of the IOP elevation only group (n = 6) and the inhibitor carrier DMSO group (n = 5) were not significantly different from each other (1514 ± 78/retina versus 1314 ± 103/retina; Fig. 2A). Intravitreal application of DMSO marginally increased the number of macrophages (81 ± 10/retina versus 137 ± 9/retina) in the eye (Fig. 2A). LY303511, the negative control of LY294002 and which contains a single atom substitution in the morpholine ring compared to LY294002 and does not inhibit PI3K even at high concentrations, did not affect RGC survival, but substantially increased the number of macrophages in the eye (n = 5; Fig. 2A). Compared with the negative control LY303511 and DMSO groups, significant decreases in RGC survival were observed after intravitreal application of PI3K/akt pathway inhibitor LY294002 or KY12420 (n = 5 in each group; Fig. 2A). Concomitant with the decrease in RGC survival was a significant increase in the number of macrophages in the retina (Fig. 2A). Significant decrease in the number of surviving RGCs was also seen after intravitreal application of JAK/STAT pathway inhibitor AG490 or Jak Inhibitor I (n = 5 and 6, respectively; Fig. 2A). Accompanying the decrease in RGC survival was a significant increase in the number of macrophages in the retina (Fig. 2A). These observations are thus similar to what was seen after inhibition of PI3K/akt and JAK/STAT pathways in SPD rats [26,27]. Our earlier studies using Western blotting showed that the pathway inhibitors used at these concentrations effectively, though not completely, blocked collective signal transduction of the pathways [26-28].Figure 2Average densities (± SEM) of FG-labeled surviving RGCs and ED1+ macrophages after inhibition of PI3K/akt and JAK/STAT pathways in F344 rats. Significant decrease in RGC viability and concomitant increase in the number of macrophages were seen following PI3k/akt and JAK/STAT pathway inhibition (A). After macrophage removal (B), pathway inhibition-induced RGC loss was significantly but not completely prevented, indicating that both macrophage and PI3k/akt and JAK/STAT pathways were involved in RGC viability in this strain of rat. *p < 0.05, **p < 0.01 and ***p < 0.001 against DMSO group unless specified.Effects of macrophage removal on RGC viability in F344 ratsWe applied a macrophage remover, clodronate liposomes, intravitreally to deplete macrophages in the eye [26-28,31]. As expected, clodronate liposomes applied alone or in combination with KY12420 (n = 5 each group) significantly reduced the number of macrophages in the eye to a level slightly higher than that in DMSO group (Fig. 2B). Concomitant with this reduction of macrophages in the eye was significant improvement but not complete recovery in RGC survival in KY12420-treated F344 rats (Fig. 2B). Similarly, an increased number of surviving RGCs was also seen after co-application of clodronate liposomes with JAK/STAT pathway inhibitor AG490 (n = 5; Fig. 2B). The improvement but not complete recovery of RGC survival after macrophage removal in PI3K/akt or Jak/STAT pathway-inhibited eyes suggested that macrophages and the pathways modulated, in opposite directions, RGC survival after acute IOP elevation in F344 rats.Effect of inhibition of PI3K/akt and JAK/STAT pathways on RGC viability in Lewis ratsThe average numbers ± SEM of surviving RGCs and ED1+ macrophages were 2583 ± 82/retina and 8 ± 1/retina, respectively, in intact Lewis rats (n = 4). The average numbers ( ± SEM) of surviving RGCs and ED1+macrophages in the retinas of IOP elevation only (n = 5) and DMSO (n = 4) groups were not significantly different from each other in Lewis rats (Fig. 3A). However, the number of surviving RGCs 3 weeks after acute IOP elevation was significantly (6-fold) lower in Lewis than F344 rats (Fig. 2A and 3A). LY303511 also did not affect RGC survival although it marginally increased the number of macrophages in Lewis rats (n = 5; Fig. 3). Even though the levels of RGC survival were already low in the negative control and DMSO groups (Fig. 3A), further decrease in RGC survival was still seen after intravitreal application of PI3K/akt pathway inhibitor LY294002 or KY12420 (n = 5 each group). Concomitant with the decrease in RGC survival was a significant increase in the number of macrophages in the retina (Fig. 3A). Similarly, intravitreal applications of JAK/STAT pathway inhibitor AG490 or Jak Inhibitor I (n = 5 each group) also resulted in a significant decrease in the number of surviving RGCs and a significant increase in the number of macrophages in the retina (Fig. 3).Figure 3Average densities (± SEM) of FG-labeled surviving RGCs and ED1+ macrophages after inhibition of PI3K/akt and JAK/STAT pathways in Lewis rats. Significant decrease in RGC viability and concomitant increase in the number of macrophages were seen following PI3k/akt and JAK/STAT pathway inhibition (A). After macrophage removal (B), pathway inhibition-induced RGC loss was significantly but not completely prevented, indicating that both macrophage and the pathways were also involved in RGC viability in Lewis rats. Note that the numbers of surviving RGCs were significantly lower in Lewis than F344 rats. *p < 0.05, **p < 0.01 and ***p < 0.001 against DMSO group unless specified.Effects of macrophage removal on RGC viability in Lewis ratsSimilar to what occurred in F344 rats, clodronate liposomes applied alone (n = 5) or co-applied with KY12420 (n = 4) or AG490 (n = 5) significantly reduced the number of macrophages in the retina to a level lower than that in the DMSO group (Fig. 3B). Concomitant with the reductions of macrophages in the eye was significant improvement but not complete recovery in RGC survival in KY12420- and AG490-treated Lewis rats (Fig. 3B). These findings suggested that PI3K/akt and JAK/STAT pathways and macrophages are involved in RGC viability in similar fashion to that in F344 rats following acute IOP elevation.To clarify whether the observed reduction of RGC counts after inhibitor application resulted from interfered transport of FG, we carried out another experiment in which both FG and immunohistochemical approaches were used to label RGCs of the same retinas and the numbers of RGCs between the 2 approaches were compared. Compared with the number of TUJ1-immunostained RGCs of the same retinas, there were an average of 15% (n = 3) decrease and 12% (n = 3) increase in the numbers of FG-labeled RGCs in KY12420 and AG490 treatment groups, repsectively. The small differences between the 2 labelling approaches are not statistically significant, indicating that the intravitreal application of the pathway inhibitors does not influence the efficacy of the FG transport. Note that FG was applied 20 hours whereas the data presented below were obtained from rats that received FG 5 days after the administration of the pathway inhibitors. The further delayed application of FG rendered the inhibitors less likely to affect the retrograde transport of FG.DiscussionIn this study we investigated the roles of PI3K/akt and JAK/STAT pathways and macrophages in RGC viability after acute IOP elevation in F344 and Lewis rats, which are known to have different autoimmune profiles. We show that in both rat strains, inhibition of PI3K/akt or JAK/STAT pathway reduces RGC survival and activates macrophages in the eye. In addition, both macrophage activation and PI3K/akt and JAK/STAT pathways mediate RGC viability, in opposite directions, after acute IOP elevation.PI3K/akt and JAK/STAT are known to be the major signalling executors for neuronal survival [19,20,24,25,32-36]. Whereas PI3K/akt is the common signal transduction pathway underlying neurotrophin-induced biological actions, JAK/STAT is well-documented to be responsible for cytokine-induced effects [37]. Previously we showed that the pathway inhibitors applied at dosages as in this study significantly inhibited signal transduction of the respective pathways [26-28,38]. The clear differences in RGC survival or in macrophage recruitment between eyes treated with LY294002 and with its negative control LY303511 following IOP elevation in both F344 and Lewis rats are similar to what was seen in SPD rats [26,27]. These results confirmed that the actions of LY294002 on RGC viability and macrophage recruitment are dependent on PI3k/akt pathway-inhibition. The persistent loss of RGCs after PI3K/akt or JAK/STAT pathway inhibition in the absence of ocular macrophages (following clodronate liposome application) in both strains further verify that these signal transduction pathways mediate RGC survival following acute IOP elevation independent of influence of autoimmune background.Previously T-cells were shown to play a part in differential protection of RGCs following episcleral and limbal vein cauterization-induced IOP elevation in F344 and Lewis rats [14]. In our earlier study, we showed that macrophage reactions to acute IOP elevation were also different in rats with different autoimmune backgrounds [17]. In the present study we demonstrated that in contrast to the actions on macrophages, autoimmune background did not modulate signal transduction pathways of PI3K/akt and JAK/STAT in RGC survival.ConclusionPI3K/akt and JAK/STAT pathway mediate RGC survival after acute IOP elevation and autoimmune background does not influence the functional roles of these pathways. In addition, PI3K/akt and JAK/STAT pathway inhibition-induced macrophage activation in the eye is detrimental to RGCs following acute IOP elevation.MethodsA total number of 108 young adult (8–10 weeks old) F344 and Lewis rats were used, and each experimental group consisted of 4–6 rats. All experiments conformed to The Chinese University of Hong Kong Animal Experimentation Ethic Committee (AEEC) guidelines and were approved by the AEEC. All possible measures were taken to minimize suffering and limit the number of rats used in this study. All surgery was carried out under anaesthesia with a 1:1 mixture (1.5 ml/kg) of ketamine (100 mg/ml) and xylazine (20 mg/ml).Acute IOP elevation and the experimental groupsThe acute IOP elevation procedure has previously been reported [39]. Briefly, a 27-gauge needle was placed in the anterior chamber of the left eye. The needle was connected to a container carrying 500 ml sterile normal saline. The container was raised to a height of 1496 mm above the eye to elevate the IOP to 110 mmHg for 2 hrs. Each strain of rats was allocated to different experimental groups after acute IOP elevation. The first group received no intravitreal injections and served as intact controls. The second group received intravitreal injections of DMSO (3 μl each injection), which is the inhibitor carrier. The third group received intravitreal injections of LY303511 (Calbiochem, San Diego, USA; 2 mM × 3 μl), which was the negative control of PI3k/akt pathway inhibitor LY294002. The fourth and fifth groups received intravitreal injections of PI3k/akt pathway inhibitors LY294002 [2-(4-Morpholinyl)-8-phenyl-1(4H)-benzopyran-4-one hydrochloride, Sigma; 2 mM × 3 μl and KY12420 (Calbiochem, San Diego, USA; 2 mM × 3 μl), respectively. The sixth and seventh groups received intravitreal injections of JAK/STAT pathway inhibitors AG490 [α-Cyano-(3,4-dihydroxy)-N-benzylcinnamide, Calbiochem; USA; 2 mM × 3 μl] and Jak Inhibitor I [2-(1,1-Dimethylethyl)-9-fluoro-3,6-dihydro-7H-benz[h]-imidaz [4,5-f]isoquinolin-7-one; Calbiochem; 2 mM × 3 μl], respectively. Two inhibitors of each pathway, coupled with the available negative control, LY303511, were used in aim to clarify or exclude the possibility of chemical effects of these inhibitors on RGC viability.As macrophage activation has been seen to accompany acute IOP elevation and inhibition of PI3k/akt and JAK/STAT pathways, we used clodronate liposomes to remove macrophages in the eye and examined what roles macrophages and the two pathways played in RGC viability. In this part, 1 group received intravitreal injection (3 μl, once only) of clodronate liposomes immediately after acute IOP elevation, 2 groups received immediate intravitreal injection (3 μl, once only) of clodronate liposomes that were followed by 3 intravitreal injections of KY12420 and AG490, respectively, 3, 9 and 15 days later. Both clodronate and liposomes (if prepared of phosphatidylcholine and cholesterol) are not toxic. Liposome-encapsulated clodronate and liposomes containing PBS only (control liposomes) were prepared as previously described [40]. Clodronate was a gift of Roche Diagnostics GmbH (Mannheim, Germany). Phosphatidylcholine was obtained from Lipoid GmbH (Ludwigshafen, Germany), and cholesterol purchased from Sigma. Recently we showed that both PBS liposomes and clodronate liposomes applied at this dosage into the eye were not detrimental to RGCs [28,31]. All rats survived for 3 weeks after acute IOP elevation.For intravitreal injections of DMSO or the inhibitors, each rat received three posterior chamber eye injections on days 3, 9 and 15 after IOP elevation. Intravitreal injection of 3 μl clodronate liposomes was carried out once only on the day of IOP elevation, which was 3 days prior to pathway inhibitor injection. For the eye injection, the micropipette was deliberately angled to avoid damage to the lens [41].Retrograde labeling of viable RGCsTo retrogradely label viable RGCs, a small piece of gelfoam soaked with 4% FluoroGold (FG, Fluorochrome Inc, Denver, USA) was applied at the newly cut stump of the proximal ON to retrogradely label viable RGCs. Animals survived for another 40 hours to maximize retrograde transport of the dye. Note that RGCs only start to die 5 days after ON axotomy in adult rats [42]. While deeply anaesthetized, the rats were perfused with cold 4% paraformaldehyde in phosphate buffer (0.1 M, pH 7.4). After dissection from the eye-cups, the retinas were post-fixed in 4% paraformaldehyde for 45 min, flat-mounted and temporarily coverslipped in anti-fading medium (Dako Corporation, Carpinteria, CA, USA). The number of FG-labelled RGCs in each field (0.25 × 0.25 mm2), sampled at a fixed distance from one another and in a pattern of grid intersections, was counted throughout the whole retina. A total of 70–80 fields, about 8–10% of the total retinal area, were sampled per retina. The average density of viable RGCs was obtained. This approach avoided problems associated with uneven distribution of RGCs in the retina.In a separate experiment, we examined whether intravitreal application of the pathway inhibitors influenced the retrograde transport of FG. One inhibitor of each pathway (AG490 for JAK/STAT pathway and KY12420 for PI3K/akt pathway) was injected intravitreally into eyes of normal rats (n = 3 each group). FG was applied in the same way as above 20 hours after the inhibitor application, and the rats were killed 40 hours after FG application. After counting the number of FG-labeled RGCs, the retinas were immunostained for RGCs using the TUJ1 monoclonal antibody (against βIII-tubulin, BabCO, Richmond, CA, USA). βIII-tubulin has been shown to be an RGC-specific marker in retinal wholemounts [15,16,28,31,38,43,44].Immunohistochemical staining of macrophagesAfter counting the number of FG-labeled RGCs, retinas were used for immunostaining of macrophages. The retinas were thoroughly washed with PBS and blocked with 10% normal goat serum (NGS), 1% bovine serum albumin (BSA) and 0.2% Triton for 1 hour. They were then immunostained overnight at 4°C with ED1 antibody (1:200, Serotec, Oxford, UK) [15,28], were rinsed with PBS and incubated with conjugated Cy3 (Jackson ImmunoResearch Laboratories, West Grove, PA, USA; 1:400) secondary antibody overnight at 4°C. After 3 washes each for 5 minutes, the retinas were mounted with anti-fading fluorescence mounting medium and examined under the fluorescent microscope. ED1 positive (+) cells were counted in the same way as for FG-labeled RGCs.Statistical analysisData on RGCs from the different groups were pooled and statistically analysed using Bonferroni test following one-way analysis of variance (ANOVA). Bonferroni test was used to compare mean values among all intra-groups [15,38].AbbreviationsEAE: experimental autoimmune encephalomyelitis; F344: Fischer 344; HPA: hypothalamic-pituitary-adrenal; IOP: intraocular pressure; GCL: ganglion cell layer; JAK: janus kinases; ON: optic nerve; PI3K: phosphatidylinositol 3-kinase; RGC: retinal ganglion cell; SPD: Sprague Dawley; STAT: signal transducers and activators of transcription.Authors' contributionsYH carried out experiments and collected data. ZL carried out the double-labeling experiment. NW participated in the study design. NvR provided reagents (clodronate liposomes). QC designed the study, analyzed the data and drafted the manuscript. All authors read and approved the final manuscript.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2533009.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2533009",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2533009\nAUTHORS: Trine B Rounge, Thomas Rohrlack, Tom Kristensen, Kjetill S Jakobsen\n\nABSTRACT:\nBackgroundCyanopeptolins are nonribosomally produced heptapetides showing a highly variable composition. The cyanopeptolin synthetase operon has previously been investigated in three strains from the genera Microcystis, Planktothrix and Anabaena. Cyanopeptolins are displaying protease inhibitor activity, but the biological function(s) is (are) unknown. Cyanopeptolin gene cluster variability and biological functions of the peptide variants are likely to be interconnected.ResultsWe have investigated two cyanopeptolin gene clusters from highly similar, but geographically remote strains of the same genus. Sequencing of a nonribosomal peptide synthetase (NRPS) cyanopeptolin gene cluster from the Japanese strain Planktothrix NIES 205 (205-oci), showed the 30 kb gene cluster to be highly similar to the oci gene cluster previously described in Planktothrix NIVA CYA 116, isolated in Norway. Both operons contained seven NRPS modules, a sulfotransferase (S) and a glyceric acid loading (GA)-domain. Sequence analyses showed a high degree of conservation, except for the presence of an epimerase domain in NIES 205 and the regions around the epimerase, showing high substitution rates and Ka/Ks values above 1. The two strains produce almost identical cyanopeptolins, cyanopeptolin-1138 and oscillapeptin E respectively, but with slight differences regarding the production of minor cyanopeptolin variants. These variants may be the result of relaxed adenylation (A)-domain specificity in the nonribosomal enzyme complex. Other genetic markers (16S rRNA, ntcA and the phycocyanin cpcBA spacer) were identical, supporting that these geographically separated Planktothrix strains are closely related.ConclusionA horizontal gene transfer event resulting in exchange of a whole module-encoding region was observed. Nucleotide statistics indicate that both purifying selection and positive selection forces are operating on the gene cluster. The positive selection forces are acting within and around the epimerase insertion while purifying selection conserves the remaining (major) part of the gene cluster. The presence of an epimerase in the gene cluster is in line with the D-configuration of Htyr, determined experimentally in oscillapeptin E in a previous study.\n\nBODY:\nBackgroundCyanopeptolins are nonribosomally produced peptides with highly variable composition. The general structure of the cyanopeptolin peptide family encompasses 7 amino acids, including the residue 3-amino-6-hydroxy-2-piperidone (Ahp), where the six C-terminal amino acids form a ring [1,2] and the N-terminal amino acid frequently is N-modified. The N-terminal amino acid and all positions in the ring except position 2 (threonine) and position 4 (Ahp) can be occupied by variable amino acids, giving rise to a large number of cyanopeptolin variants [3].The succession of the modules [4,5] and specificity of A-domain binding pockets in nonribosomal peptide synthetases (NRPSs) [6,7] can give a good prediction of peptide composition and structure. NRPSs do not always perform stringent substrate selection and incorporation [7], thus, relaxed substrate specificity is common in NRPS [6,8,9]. In addition to the common module domains including the adenylation (A)-, condensation (C)- and thiolation (T)-domains, several tailoring domains have been found associated with cyanopeptolin synthetases. Methyltransferases are present in three cyanopeptolin gene clusters from Anabaena, Microcystis and Planktothrix (apd, mcn and oci). Halogenases are found in apd and mcn, while the tailoring domains responsible for side chain modification of the N-terminal amino acid are unique for each strain (i.e.; formyl transferase in apd, sulfotransferase and glyceric acid (GA) transferase in oci, absent in mcn).So far, only cyanopeptolin gene clusters derived from the genera Anabaena [10], Microcystis [11] and Planktothrix [12] have been characterized. They share the same basic domain structure but possess unique tailoring genes and A-domain substrate binding pockets, indicating independent evolution of cyanopeptolin genes within each lineage. Sequence identity is high (approximately 80% in the NRPS module coding regions) between Microcystis (mcn) and Planktothrix (oci) cyanopeptolin gene clusters. The more thoroughly investigated microcystin gene clusters show higher sequence identity within a genus than between genera. The same is likely to be the case also for the cyanopeptolin genes.Sequence variation in microcystin synthetase clusters has been investigated within strains of the genera Microcystis [13,14] and Planktothrix [15]. Modifications and reorganizations due to several recombination events have been reported [14-16], and together with differences in substrate specificity between equivalent A-domains [17-19] are the reason for the different peptide variants.Planktothrix NIVA CYA 116 (NIVA CYA 116), isolated from a Norwegian lake, produces cyanopeptolin-1138 [12] for which the amino acids configurations are unknown. This peptide was found to be highly similar to oscillapeptin E produced by Planktothrix NIES 205 (NIES 205), isolated in Japan [20]. Both peptides have the same molecular mass, but slightly different polarities [12]. A different content of L-/D-amino acids in the peptides was suggested as a possible reason for the observed difference [12]. To investigate the genetic basis of the differences between the peptides, we have cloned and sequenced the NIES 205 cyanopeptolin gene cluster and compared it to the previously characterized NIVA CYA 116 gene cluster. This has allowed us to explore NRPS evolution and genetic variations in closely related strains and to investigate to what extent selectional forces operate on these gene clusters.ResultsNIVA CYA 116 and NIES 205 have similar but not identical peptide profilesThe major peptides in the two strains consist of HO3SO-CH2-CH(OMe)-CO-HTyr-Thr-HTyr-Ahp-Ile-Phe(Me)-Ile ([Table 1, Additional file 1 figure 1] and Rounge et al [12]). However, spiking experiments (data not shown) revealed a slight difference in polarity between cyanopeptolin-1138 from NIVA CYA 116 [12] and oscillapeptin E from NIES 205 [20]. In contrast to NIVA CYA 116 producing only cyanopeptolins, screening of NIES 205 shows production of additional peptide from other peptide-classes (data not shown).Several cyanopeptolin variants were also detected in both strains. LC-MS-MS data identified minute amounts of seven cyanopeptolins in NIVA CYA 116, with variation in the first, third, fifth and/or seventh positions compared to cyanopeptolin-1138/oscillapeptin E [Additional file 1 figure 2]. An earlier study has shown that NIES 205 produce oscillapeptin C, D and E, based on spectroscopic analyses including 2D NMR [20]. Our LC-MS-MS analysis of NIES 205 confirmed the production of oscillapeptin D and E, but also identified a cyanopeptolin with the mass 1074, which is found in NIVA CYA 116 as well [Table 1 and Additional file 1 figure 3].NIVA CYA 116 and NIES 205 produced similar – but not identical – cyanopeptolin variants. The identified NIVA CYA 116 cyanopeptolins were mainly combinations of Hty/Ile/Leu in positions AA1 and AA3 and Ile/Leu/Val in positions AA5 and AA7. Other unidentified apolar amino acid-like residues were detected in position AA3. In contrast, the only variations observed in the NIES 205 peptides were Hty, Ile/Leu and HcAla in position AA3 (Table 1).Table 1Oci A-domains binding pockets and peptide profilesBinding pocketsOciA-A1OciA-A2OciB-A3OciB-A4OciB-A5OciB-A6OciC-A7NIVA CYA 116DLGFTGAVCKDFWNIGMVHKDAQSMGAIIKDVENAGVVTKDAFFLGVTFKDAWTIAGVCKDAFFLGVTFKNIES 205DLGFTGAVCKDFWNIGMVHKDAEGMGAIIKDVENAGVVTKDAFFLGVTFKDAWTIAGVCKDAFFLGVTFKNIVA CYA 116Mass DaSide chainAA 1AA2AA3AA4AA5AA6AA71138HO3-SO-CH2-CH(OMe)-COHHTyrThrHTyrAhpIlePhe(Me)Ile1124HO3-SO-CH2-CH(OMe)-COHHTyrThrHTyrAhpValPhe(Me)Ile1124HO3-SO-CH2-CH(OMe)-COHHTyrThrHTyrAhpIlePhe(Me)Val1074HO3-SO-CH2-CH(OMe)-COHHTyrThrIle/LeuAhpIlePhe(Me)Ile1010HO3-SO-CH2-CH(OMe)-COHIle/LeuThrIle/LeuAhpIlePhe(Me)Ile1088HO3-SO-CH2-CH(OMe)-COHHTyrThrXAhpIlePhe(Me)Ile1122HO3-SO-CH2-CH(OMe)-COHHTyrThrYAhpIlePhe(Me)IleNIES 205Mass DaSide chainAA 1AA2AA3AA4AA5AA6AA71138*HO3-SO-CH2-CH(OMe)-COHHtyrThrHtyrAhpIlePhe(Me)Ile1074HO3-SO-CH2-CH(OMe)-COHHtyrThrIle/LeuAhpIlePhe(Me)Ile1128**HO3-SO-CH2-CH(OMe)-COHHtyrThrHcAlaAhpIlePhe(Me)IleThe binding pocket residues of the NIVA CYA 116 and NIES 205 A-domains were identified by comparison to the GrsA-Phe A-domain (Residue 235, 236, 239, 278, 299, 301, 322, 330, 331, 517). The Oci-A3 binding pockets are different (in grey) between the two strains, and the divergent amino acids are shown in bold. The composition of cyanopeptolins produced by NIVA CYA 116 and NIES 205 and their molecular weights (M+H+) are shown with amino acids correlated with the putative binding pockets. HcAla is 3-(4'-hydroxy-2'-cyclohexenyl) alanine, and X and Y is unidentified amino acid derivates. Mutual peptides in NIVA CYA 116 and NIES 205 are highlighted in dark grey and light grey. See [Additional file 1 figure 1, 2 and 3 for the peptide structure and more details on MS data.* peptide named oscillapeptin E and **peptide named oscillapeptin DComparison of the 205-oci and 116-oci gene clustersAnticipating that two strains producing almost identical cyanopeptolins also should contain similar gene clusters, we sequenced a cyanopeptolin gene cluster in NIES 205 (205-oci) using primers designed for the cyanopeptolin (oci) gene cluster in NIVA CYA 116 (116-oci) [12]. The two gene clusters, including the ABC transporter genes and the intergenic spacers, were highly similar (93% identity between the nucleotide sequences), and the domain structures of the encoded synthetases were almost identical; except that 205-oci contained an epimerase encoding (E)-domain between T2 and C2 (Figure 1). The position of the E-domain corresponds to the Htyr in D-configuration in oscillapeptin E determined by Itou et al [20]. Both gene clusters included a GA-domain and a sulfotransferase domain. Comparison with cyanopeptolin gene clusters characterized in Microcystis (mcn) [11] and Anabaena (apd) [10] (Figure 1) showed a higher degree of similarity within the Planktothrix genus than between genera (70% identity between OciB and AdpB with the additional methyltransferase excluded and 77% identity between OciC and AdpD). A-domains and A-domain binding pockets signatures were identified from the gene clusters and aligned. The binding pocket signatures in 116-Oci and the corresponding 205-oci signatures were identical, except for 116-OciB-A3 (DAQSMGAIIK) and 205-OciB-A3 (DAEGMGAIIK) (Table 1). Corresponding pairs of 205-Oci and 116-Oci A-domains clustered together in phylogenetic analyses that included A-domains from cyanopeptolin [10-12], microcystin [17,21,22] nostocyclopeptide [23] and nostopeptolide [24] synthetases [Additional file 1 figure 4].Figure 1Comparison of the known cyanopeptolin operons. The overall structure of cyanopeptolin operons oci from Planktothrix NIES 205 [GenBank: EU109504] and NIVA CYA 116 [GenBank: DQ837301], mcn from Microcystis [GenBank: DQ075244] and apd from Anabaena [GenBank: AJ269505]. Gene names, transcription directions and approximate sizes are indicated above each gene cluster. Adenylation (red), condensation (green), thiolation (yellow), epimerization (turquoise), methyltransferase (blue) sulfotransferase (pink), halogenisation (purple) and termination domains (grey) are shown with their abbreviations. The putative activated amino acids are indicated for each A-domain. Amino acids detected in smaller amounts are beneath the major amino acid. Equivalent modules are depicted in light blue and light orange. The ABC transporter is transcribed in the opposite direction in the oci and mcn operons, and an ABC transporter is predicted downstream of the apd operon.E-domains are common in cyanobacterial NRPS, found in microcystin, aeruginosin and nostocyclopeptide synthetases, notably, E-domains have until now not been found in cyanopeptolin synthetases. The E-domain produces the D-isomer of the amino acid activated by the upstream A-domain and is also involved in the stereospecific selection of the D-isomer for incorporation in the peptide product. Most E-domains are flanked by T (TE)- and C-domains with special motives [25,26], and this was the case also in 205-Oci – as shown by the phylogenetic analyses (see Figure 2).Figure 2Phylogenetic analyses of E-domains. The E-domain phylogenetic tree was constructed utilizing MrBayes 3.1., Wag protein substitution model and gamma-shaped distribution. In addition, the bootstrap obtained for NJ (MEGA 3.1) at default settings and ML (RAxML) trees are indicated. Only posterior probability values and bootstrap replica values above 50% (out of 1000 (NJ) and 100 (ML) trees) are shown.The NIES 205-E-domain is localized downstream of 205-A1 and T2. A phylogenetic analysis of E-domains (Figure 2), including E-domains from microcystin synthetase (McyA-E) and nostocyclopeptide synthetase (NcpA-E), showed a close relationship between NcpA-E and 205-OciA-E (72% identity on the DNA and 67% similarity on the protein level).Phylogenetic analyses of the C-domains (Figure 3), including domains from cyanopeptolin [10-12], microcystin [17,21,22] nostocyclopeptide [23] and nostopeptolide [24] synthetases, clustered according to presence or absence of an upstream E-domain. The 205-Oci-C2-domain grouped with D-amino acid-specific C-domains, while the other 205-Oci-C domains formed a clade with the corresponding 116-Oci-C-domains.Figure 3Phylogenetic analyses of C-domains showing groups according to gene cluster and position/function. The C-domain phylogeny was constructed using Bayesian inference with gamma distribution, 4 mill generations tree sampling every 100 generations and removal of the first 3000. The topologies generated using NJ (MEGA 3.1) and ML (RAxML) analyses show near identical branching patterns-only minor differences are seen within the Apd group. Bayesian posterior probability, NJ (1000 bootstrap values) and ML (100 trees) above 50% are shown. CpRev protein substitution model was used in the Bayesian and ML analyses. Genus origin is shown with first letter abbreviations (P = Planktothrix, M = Microcystis, A = Anabaena and N = Nostoc), and the C-domains are labeled in numerical order according to direction of transcription (i.e. seven oci, seven mcn and six apd C-domains). Corresponding Oci C domains group together, except for C2 situated downstream of the 205-E-domain. C1–C4 apd, nos and ncp C-domains do not group according to functionThe specialized TE-domains associated with E-domains, show major differences within the core T motif compared to standard T-domains [25]. Comparisons of regular T-domains and TE-domains, including 205-Oci-TE2, McnA-T1 and NcpA-TE1-domain, showed an H/D and L/I difference in addition to a gap in the TE-domain motif [Additional file 1 figure 5]. N-terminal T-domains, including both 116-T1 and 205-T1, could also be distinguished from TE-domains and regular T-domains [Additional file 1 figure 5].Other genomic regions confirm a close relationship between the Planktothrix strainsSeveral markers were sequenced to further study the relationship between Planktothrix NIVA CYA 116 and NIES 205. The DNA sequences (16S rDNA (1357 bp), a part of ntcA (384 bp), a global transcriptional regulator of nitrogen assimilation in cyanobacteria, and the phycocyanin spacer cpcBA) displayed 100% identity between the two strains.Variation in substitution rates throughout the cyanopeptolin gene clustersInvestigation of the substitution rates within the 30 kb 116- and 205-oci gene cluster alignment can identify both putative recombination events and regions under specific selection pressure. The region containing the epimerization domain (T2, E, C2) was excluded due to too large overall differences to produce a reliable alignment. Figure 4 shows segregating sites (black lines) and nonsynonymous vs. synonymous substitution rates (red lines) in a sliding window analysis of the alignment. Only a few scattered substitutions can be seen in the first part, containing the ABC transporter, GA, T1, S and C1 domains, and in the last part, containing A6, M, T7, C7, A7, T8 and TE domains. However, the C3 and A3 domains contained several substitutions and the rate of mutations in nonsynonymous sites compared with synonymous sites (Ka/Ks) exceeded 1 – a putative sign of positive selection. A high substitution rate was also observed in a small region in C6 and the last part of A1, but the Ka/Ks ratios did not exceed 1.Figure 4Distribution of segregating sites and Ka/Ks ratios in the oci gene cluster. The ratios are displayed using the program DnaSP and sliding windows analysis on the alignment of 205-oci and 116-oci. Window length was 50 bp and step size 10 bp. The distribution of segregation sites (red) and Ka/Ks (black) ratios are shown in correlation with the domain alignment. Module 2 (T2-(E)-C2) has been excluded from the analyses.DiscussionCorrelation between cyanopeptolin gene clusters and peptidesThe presence of two highly similar NRPS gene clusters (oci) in NIVA CYA 116 and NIES 205, and the production of nearly identical peptides by the two strains corroborate the association between the oci gene cluster and cyanopeptolin-1138 proposed by Rounge et al [12]. This association is further substantiated by high degree of similarity to the cyanopeptolin gene cluster in Anabaena (apd), where the functional relationship between genes and peptides has been confirmed by a gene knock-out study [10] – as well as similarity to the Microcystis cyanopeptolin gene cluster (mcn) [11].Global dispersal and distribution of cyanopeptolin genesBased on the genomic regions studied here, two Planktothrix strains, NIVA CYA 116 and NIES 205, appear to be closely related despite the geographical separation. This is in accordance with the sequence comparison of 16S rDNA [27] identifying identical 16S rDNA sequences in Japan, China, The Netherlands, UK, Finland, Sweden and Norway, and thus may indicate a global distribution of closely related Planktothrix strains. Since Lake Årungen in Norway host international rowing competitions, a co-transport of this Planktothrix genotype with rowing equipment may be feasible. The data presented here do not allow any conclusions about global distribution without a more thorough analysis. The highly specific differences observed in the oci gene clusters are, independently of geographic distributions, intriguing. Our analyses indicate that the differences to some extent are due to positive selection at specific amino acid positions.Variation in peptide content due to lack of specificity in the A-domains?Previous studies have shown that lack of specificity in A-domains leads to activation of several amino acids with similar properties, thus giving rise to the synthesis of a series of related peptides from a single NRPS system [28]. Ile/Leu/Val activating A-domains have been reported in lichenysin biosynthesis [8], and fengycin synthetase [29] among others. It is likely that the 116-Oci-A5- and A7-domains can activate Leu, Ile and Val and that the 116-Oci-A1- and A3-domains, that mainly activates Htyr, also can activate Ile and Leu. Consequently, 116-Oci is responsible for production of all seven cyanopeptolin detected in NIVA CYA 116 in this study. Likewise, 205-Oci probably is responsible for all oscillapeptin variants. The biological significance of a single NRPS complex giving rise to several peptide variants is yet to be determined.Six of the seven binding pockets signatures of corresponding A-domains in NIES 205 and NIVA CYA 116 are identical (Table 1). If the different peptide profiles observed in the two strains are due to genetic differences in the NRPS genes, they are likely to be due to differences not involving the amino acids constituting the binding pocket signatures. LC-MS-MS-analyses were performed on strains cultivated on the same media, but we cannot completely exclude substrate availability as a contributory cause of variable peptide amount and peptide profile in the strain.Module exchange and amino acid configurationOver a stretch of total of 30 kb including the ABC transporter, the 116-oci and 205-oci gene clusters are remarkable similar, except for the modules encoding the T2-(E)-C2 domains. Too low sequence similarity is found between the whole T2-(E)-C2 modules in NIVA CYA 116 and NIES 205 to make a reliable alignment, suggesting that in one of these strains an entire module may have been exchanged through recombination. The E-domain trees (Figure 2) show a close relationship between cyanobacterial E-domains.Sequence similarity to other E-domains and the distinctive flanking C (Figure 3) and T [Additional file 1 figure 5] domains observed by phylogenetic analysis indicate that the Oci-E-domain is an active epimerase, and are responsible for epimerization of Htyr to D-configuration. The configuration of the amino acids in cyanopeptolin-1138 were not determined however, a D-Htyr in oscillapeptin E and a putative L-Htyr in cyanopeptolin-1138 might explain the small difference between the oligopeptides with regard to polarity observed by HPLC analysis, as reported by Rounge et al. [12].Interestingly, in the corresponding region of the Mcn cyanopeptolin synthetase in Microcystis the McnA-T1 and McnB-C2 include motifs suggesting association with an E-domain [11]. In this case, however, no E-domain is present.Sequence conservation and selection within cyanopeptolin modulesThe two cyanopeptolin gene clusters (205-oci and 116-oci) are highly similar also at the third codon position. The first part (ABC-transporter, the spacer, GA-, T1-, S-, and C1-domains) and last part (C4-, A4-, T5-, C5-, A5-, T6-, C6-, M-, T7-, C7-, A7-, T8- and TE domains) of the Planktothrix cyanopeptolin gene cluster are nearly identical, despite the geographical distance separating the strains. Mechanisms for such sequence conservation may be frequent homology-driven genetic exchange within a genotype, leading to homogenization – in line with the general models suggested by Rudi et al. [30], Gogarten et al.[31] and Papke et al. [32]. Or alternatively sequence conservation may be due to low evolutionary rates caused by purifying selection or very short time of independent evolution.Analysis of segregating sites and rates of nonsynonymous and synonymous nucleotide substitutions (Ka/Ks) indicate that module 3 (T3-, C3- and A3-domains) is different from the remaining domains by displaying higher substitution rates and signs of positive selection at several sites (Ka/Ks higher than 1). This is the module responsible for incorporation of the amino acid at position AA3 in the peptide.According to data from Itou et al [20], a single amino acid replacement in the AA3 position of oscillapeptin E and F alters the protease inhibitory profile, indicating that this position could be pivotal for the inhibitory activity of cyanopeptolins. Positive selection in the third module could thus be expected to increase the adaptability of the inhibitory- or other putative functions of cyanopeptolin.ConclusionThe Planktothrix strains of Japan and Norway harbor almost identical cyanopeptolin gene clusters and display very similar (but not identical) cyanopeptolin profiles. The notable gene cluster difference is the presence of an epimerase in NIES 205 corresponding to a D-Htyr in ocillapeptin E. Within a single gene cluster we have demonstrated both positive selection and purifying selection, the first promoting new gene cluster variants following recombination, the latter maintaining a high degree of conservation of the major parts of the gene cluster.MethodsBacterial culturesPlanktothrix agardhii NIVA CYA 116 was isolated in 1983 from Lake Årungen, Norway, and maintained in the NIVA culture collection of Algae. Planktothrix agardhii NIES 205 was isolated from Lake Kasumigaura/Ibaraki, Japan in 1982, and maintained in the NIES culture collection [20]. Both strains were cultured in Z8 [33] media at ~20°C with 12 hour illumination at about 15 μmol m-2 s-1 in Sanyo versatile environmental test chamber (FG-4P 36–40).PCR and sequencingDNA from NIES 205 was isolated utilizing Dynabeads (Invitrogen, Carlsbad, USA) [34]. Combinations of PCR primers designed for the cyanopeptolin (oci) gene cluster in NIVA CYA 116 [12] were used to amplify regions of a cyanopeptolin gene cluster in NIES 205. These PCR products were sequenced using primer walking. Additional PCR primers were designed to amplify regions between already obtained PCR products. BD Advantage 2 (BD Biosciences, Mountain View, USA) was utilized as polymerase in all PCR amplifications. The PCR products were sequenced using an ABI 3730 sequencer and v3.1 Big Dye solution.Sequence analysis and phylogenyOpen reading frames were identified and translated using Vector NTI (Invitrogen, Carlsbad, USA). Domains and their boundaries were identified using the NRPS database [35], A-domain binding pocket residues identified by aligning the sequences with the GrsA-Phe A-domain [6] and substrate specificity predicted utilizing the NRPS database  and phylogenetic analysis. A-, C-, T- and E- domain protein sequences were aligned using MEGA 3.1 and Neighbor-Joining (NJ) trees were constructed using MEGA 3.1 at default settings (Poisson correction as the amino acid substitution model) [36]. Optimal protein evolution model was found by ProtTest [37]. Trees were constructed utilizing MrBayes [38] 3.0 and 3.1 [39] on the UiO Bioportal [40] with an optimal protein substitution model. Variable substitution rates across sites were accounted for by gamma distribution. The MCMC chains were carried out for 4 million generations and trees were sampled every 100 generations, removing 3000 trees before the MCMC chain reached convergence. In addition, maximum likelihood inferences with RAxML [41] were performed on the E- and C- domain alignments. Similarity calculations were done in Vector NTI. DnaSP [42] was used to calculate Ka/Ks ratio and segregating sites with a sliding window with window length of 50 bp and step size 10 bp.Mass spectrometryFreeze-dried material of NIVA-CYA116 and NIES 205 was extracted with 50% MeOH (MeOH:water, v/v) and the extracts were subjected to a screening for cyanopeptolins by LC-MS. The instrument included a Waters Acquity UPLC system equipped with an Atlantis column (C18 2.1 × 150 mm, 5 μm particle size) and set to run a linear gradient starting with 80% solvent A (10 mM ammonium acetate, 0.1% acetic acid) and ending with 60% solvent A after 15 min. Solvent B was MeOH with 0.1% acetic acid. The flow rate was 0.2 ml min-1. The LC system was connected to a Waters Quattro Premier XE tandem quadropole mass spectrometer equipped with an electrospray probe. The detector was run in the positive ion mode at a cone voltage of 50 V. A total ion scan from 600 to 1400 Da was performed during the entire length of the LC gradient.The structures of putative cyanopeptolins were analyzed by MS fragmentation studies. MS fragments hold valuable structural information and have been successfully used before to identify and structurally elucidate cyanobacterial oligopeptides including cyanopeptolins [43-45]. Fragmentation experiments were carried out with the hardware configuration described above. The mass spectrometer was run in daughter ion scanning mode and all settings were automatically optimized for fragmentation at 30 eV. Fragments were recorded during the entire length of the LC gradient. The identification of fragments was assisted by the HighChemMassFrontier software version 3.0. This software predicts MS fragmentation patterns on the basis of a putative structure. Comparing predicted and actual fragmentation patterns was used to assess the accuracy of a putative structure. Further hints to the structure were obtained from the occurrence of typical diagnostic ions such as immonium ions and from predictions on the amino acid occurrence made by the genetic analyses.Authors' contributionsThis work was performed as part of the PhD thesis for TBR. TBR and TR carried out all experimentation and all authors have contributed to the experimental and analytical design. TBR performed the bioinformatics and phylogenetic analysis under supervision of KSJ and TK. TR carried out the peptide analyses. TBR, KSJ (thesis advisor) TK and TR wrote the ms. All authors have read and approved the final manuscript.Supplementary MaterialAdditional file 1Peptide structure and NRPS phylogeny. Figure 1: peptide structure of cyanopeptolin 1138. Figure 2: mass spectrometric fragmentation experiments data of NIVA CYA 116. Figure 3: mass spectrometric fragmentation experiments data of NIES 205. Figure 4: NRPS A-domain phylogeny. Figure 5: Sequence analyses of T-domains. Table 1: Accession numbersClick here for file\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2533023.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2533023",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2533023\nAUTHORS: Joan M Hevel, Laura C Olson-Buelow, Balasubramanian Ganesan, John R Stevens, Jared P Hardman, Ann E Aust\n\nABSTRACT:\nBackgroundAlthough exposure to asbestos is now regulated, patients continue to be diagnosed with mesothelioma, asbestosis, fibrosis and lung carcinoma because of the long latent period between exposure and clinical disease. Asbestosis is observed in approximately 200,000 patients annually and asbestos-related deaths are estimated at 4,000 annually[1]. Although advances have been made using single gene/gene product or pathway studies, the complexity of the response to asbestos and the many unanswered questions suggested the need for a systems biology approach. The objective of this study was to generate a comprehensive view of the transcriptional changes induced by crocidolite asbestos in A549 human lung epithelial cells.ResultsA statistically robust, comprehensive data set documenting the crocidolite-induced changes in the A549 transcriptome was collected. A systems biology approach involving global observations from gene ontological analyses coupled with functional network analyses was used to explore the effects of crocidolite in the context of known molecular interactions. The analyses uniquely document a transcriptome with function-based networks in cell death, cancer, cell cycle, cellular growth, proliferation, and gene expression. These functional modules show signs of a complex interplay between signaling pathways consisting of both novel and previously described asbestos-related genes/gene products. These networks allowed for the identification of novel, putative crocidolite-related genes, leading to several new hypotheses regarding genes that are important for the asbestos response. The global analysis revealed a transcriptome that bears signatures of both apoptosis/cell death and cell survival/proliferation.ConclusionOur analyses demonstrate the power of combining a statistically robust, comprehensive dataset and a functional network genomics approach to 1) identify and explore relationships between genes of known importance 2) identify novel candidate genes, and 3) observe the complex interplay between genes/gene products that function in seemingly different processes. This study represents the first function-based global approach toward understanding the response of human lung epithelial cells to the carcinogen crocidolite. Importantly, our investigation paints a much broader landscape for the crocidolite response than was previously appreciated and reveals novel paths to study. Our graphical representations of the function-based global network will be a valuable resource to model new research findings.\n\nBODY:\nBackgroundAsbestos is a family of naturally occurring silicate minerals that was once used extensively in a variety of building materials and industries and is still found in older structures. Exposure to certain forms of asbestos, such as crocidolite and amosite, have been shown to cause mesothelioma, asbestosis, fibrosis and carcinoma of the lungs, esophagus and stomach [2-4]. Many developing countries continue to mine and use asbestos, presenting a continued risk to individuals.The biodurability and chemical reactivity of crocidolite asbestos, taken together, create a formidable carcinogen for the human lung to handle. Crocidolite can induce DNA strand breaks and base alterations. One expected response to this damage is apoptosis/cell death. But under certain conditions, cell replication can occur before the DNA damage is repaired, resulting in the formation of mutations. Events which promote survival of the cell with DNA damage and stimulate replication may lead to cancer. An unfortunate consequence of apoptosis is the stimulation of surrounding cells to replicate in an effort to repair the integrity of the damaged tissue. If the surrounding cells have experienced DNA damage, the result could be mutations, which may lead to cancer. What sets crocidolite apart from most other carcinogens is the persistent nature of the inhaled fibers, allowing for continued damage to surviving cells throughout the lifetime of the individual. Therefore, knowledge of the delicate balance between pathways that lead to proliferation or survival and those which lead to apoptosis or cell death are crucial for understanding the etiologies behind several asbestos-induced lung disorders and diseases.Much of the deleterious effects of asbestos can be attributed to the sustained synthesis of reactive oxygen species (ROS) which in turn results in DNA damage [5-7] and oxidative stress within the cell. Iron associated with the fibers (up to 27% by weight in crocidolite) can participate in Fenton and Haber-Weiss chemistry and therefore plays an intimate role in ROS generation (reviewed by [8]). Signals which reduce glutathione synthesis and increase efflux of reduced glutathione result in the reduction of intracellular glutathione concentrations [9], thus, exacerbating the situation. At the crux of the decision to initiate apoptosis is a p53-dependent transcription response. Although the events upstream of p53 activation and the importance of p53 targets are not well characterized, the result of p53 activation is mitochondrial dysfunction leading to apoptosis [10]. Apoptosis prevents continued proliferation of the damaged cell, but factors released from the damaged cell can also affect nearby cells causing inflammation and proliferation. In mapping the signal cascades which are activated/deactivated by asbestos, both human and non-human cell lines of epithelial and/or mesothelial cells, and rodent animal models have been useful. Studies have identified MAPkinase [11-13], cytokines, Akt [11,14], PKC [15], p53 [10,16] and NF-κB [17,18] signaling pathways as important players. Chromosomal translocations, promoter silencing, point mutations, deletions, and/or familial genetic susceptibility are also likely to contribute to a cell's inability to respond appropriately to the genotoxic insult of crocidolite [19,20]. The number and complexity of the signaling components identified thus far mirrors the idea that the asbestos response is not only related to the presence of iron and the production of ROS, but also involves receptor-mediated processes. An unavoidable caveat of trying to combine all the current experimental data lies in the inherent differences that exist between species and within cell lines, primary cells, and tumor specimens (reviewed in [21]), and the involvement of neighboring tissue. The complexity of asbestos response and the multiple pathologies associated with the fiber would suggest the need for more systems biology approaches to the problem.This study was conducted to begin to elucidate how the A549 human lung epithelial cell transcriptome is altered when cells are exposed to crocidolite asbestos at 6 μg/cm2. An experimental strategy was developed to ensure a statistically robust, comprehensive data set from which global observations and analyses of specific pathways could be made. This study extends a previous microarray experiment [22] performed at 2 μg/cm2 where replicates were not available. Using the more than 2,500 genes that were differentially regulated by crocidolite in A549 cells, we were able to: 1) Statistically classify the data set based on gene ontologies. This analysis revealed significant representation by transcriptional corepressor and repressor activities. Additionally, a significant unique representation by DNA modification ontologies within a less significant DNA repair/response ontology was found; 2) Identify specific novel genes that may play a role in experimentally observed asbestos-induced responses; and 3) Use a knowledge-based network approach to reveal a highly integrated series of networks related to cell death, cancer, cell cycle, cellular growth, proliferation, and gene expression. Network analysis identified several functional modules in which previously unidentified genes may play a central role in the response of cells to asbestos, including participation from an extensive extracellular set of growth factors and cytokines. Importantly, by combining genome-wide transcript changes and functional network analysis, we have documented a novel global view of the crocidolite-treated A549 transcriptome, which bears signatures of both proliferation/cell survival and apoptosis/cell death.Results and DiscussionExperimental DesignGiven the difficulties in obtaining sufficient primary human lung epithelial cells to study the complex response to asbestos, many studies have instead employed the A549 human adenocarcinoma cell line. In doing so one must keep in mind that the transformed cell line may not be entirely applicable to normal human lung cells. However, all processes noted below are recapitulated in primary human lung epithelial cells, suggesting that A549 cells represent a valid model system. Importantly, use of a human cell line avoids the inherent differences that are seen between individual patients. Alveolar epithelial type II cells are key participants in inflammation, fibrogenesis, and carcinogenesis [23] and have been described as the targets of asbestos-associated lung carcinomas [24].The conditions used for obtaining the gene expression profile for crocidolite treated A549 cells were chosen to mimic conditions where a variety of biochemical observations have previously been made in our laboratories. Exposure of A549 cells to 6 μg/cm2 crocidolite for 24 hours had previously been shown to result in the αvβ5 integrin receptor-mediated endocytosis of asbestos fibers [25], mobilization of iron from the fibers within the cell [26], upregulation of ferritin protein to combat the iron overload [27], production of reactive oxygen intermediates [28], efflux of reduced glutathione [9], DNA damage [7], PARP cleavage and activation of initiator caspases [29]. Thus, at 6 μg/cm2 of crocidolite the microarray results could be related to an extensive body of biochemical information already available in A549 cells. In addition, this cell line has been used by other investigators for several studies to explore the mechanisms of DNA damage [30], apoptosis [10], and invasiveness [31]. Use of the A549 human lung epithelial cell line complements studies done in mesothelial cells [32] and by the less carcinogenic chrysotile in bronchial epithelial cells [33]. Our goal was to extend our results beyond lists of genes and ontological classifications to discover new pathways based on functional interactions. Rather than diffusing statistical power by examining differences among multiple time points and asbestos concentrations, this experiment focused on the use of chemically defined crocidolite asbestos, a single concentration of crocidolite versus a control condition, and a single exposure time. This focus allowed for three replicates of mRNA from control and crocidolite-treated cells to be analyzed using the Affymetrix HGU133 Plus 2.0 GeneChip to provide the most comprehensive whole genome expression profile.Analysis of the Microarray DataGraphical checks of the gene expression data revealed a high-quality data set, with no spatial artifacts in the chip images, and a high degree of reproducibility within both the control and treated replicates (Additional File 1). Using tools described in the Methods Section and a false discovery rate [34] of 0.05, 2,546 genes with q-values (FDR-adjusted p-values) less than 0.05 were called statistically significant. These results are represented in a volcano plot (Figure 1).Figure 1Volcano plot showing the magnitude of differential expression (log2 fold-change) compared to the measure of statistical significance (-log10 q-value). Color is on the density scale, so darker colors indicate over-plotting of points. Statistically significant genes are observed above the horizontal line, which corresponds to a q-value of 0.05. The low variability in the data caused a large number of genes (2,546) to be classified as statistically significant.Crocidolite Induces Large Changes in the Transcriptome of A549 Human Lung Epithelial CellsAmong the 54,120 probe sets on the GeneChip, 1808 were significantly up-regulated (492 of which were two-fold or greater), and 738 were significantly down-regulated (27 of which were two-fold or greater) when A549 cells were exposed to 6 μg/cm2 crocidolite asbestos for 24 h. Of the probe-sets that changed two-fold or greater, 234 correspond to known genes for the up-regulated probe sets, and 16 of the down-regulated probe sets correspond to known genes (some genes were represented more than once on the chip). Genes that increased in expression five-fold or greater or decreased two-fold or greater and were associated with q-values less than 0.05 are shown in Table 1. A complete table of all significant genes can be found as Additional File 2. The observation that most of the expression changes are upregulated is contradictory to a previous microarray study in A549 cells [22] which used a smaller dose of crocidolite (2 μg/cm2). A direct comparison of the two data sets was difficult due to differences in experimental design. Namely, replicates at a single time point were used in the current study versus single chips over a time course, and a lower crocidolite concentration, which is a likely cause for some differences. Cells exposed to asbestos may demonstrate a hierarchical oxidative stress response [35]. Additionally, small amounts of asbestos have been shown to result in proliferation [24]. This may be attributed to a transient response to an increase in iron, which is limiting in cells in culture. However, comparison of individual expression changes in our data set to other known experimental results (discussed below) demonstrated that our data set is consistent with the literature.Table 1List of genes for which expression increased five-fold or greater or decreased two-fold or greaterFold ChangeGene SymbolGene Name115.7EGR1Early growth response 138.3FOSv-Fos FBJ murine osteosarcoma viral oncogene homolog25.3ATF3Activating transcription factor 319.5GEMGTP binding protein overexpressed in skeletal muscle17.1NR4A2Nuclear receptor subfamily 4, group A, member 216.6IL8Interleukin 815.9FSTFollistatin15.6PPP1R15AProtein phosphatase 1, regulatory (inhibitor) subunit 15A13.7STC1Stanniocalcin 112.1MAFFv-Maf musculoaponeurotic fibrosarcoma oncogene homolog F (avian)11.2CXCL2Chemokine (C-X-C motif) ligand 210.4TNFAIP3Tumor necrosis factor, alpha-induced protein 39.4MXD1MAX dimerization protein 19.1ARID5BAT rich interactive domain 5B (MRF1-like)9.0NR4A3Nuclear receptor subfamily 4, group A, member 38.6JUNv-Jun sarcoma virus 17 oncogene homolog (avian)7.8LOC153222adult retina protein7.4BREbrain and reproductive organ-expressed (TNFRSF1A modulator)7.2MCTP1Multiple C2 domains, transmembrane 17.2DDIT3DNA-damage-inducible transcript 37.2FOSBFBJ murine osteosarcoma viral oncogene homolog B7.1GDF15Growth differentiation factor 156.9CITED2Cbp/p300-interacting transactivator, with Glu/Asp-rich carboxy-terminal domain, 26.9DUSP6Dual specificity phosphatase 66.6CXCL1Chemokine (C-X-C motif) ligand 1 (melanoma growth stimulating activity, alpha)6.5IL24Interleukin 246.3SPRY2Sprouty homolog 2 (Drosophila)6.1PTGS2Prostaglandin-endoperoxide synthase 2 (prostaglandin G/H synthase and cyclooxygenase)6.0KLF6Kruppel-like factor 66.0IL11Interleukin 115.6IRAK2Interleukin-1 receptor-associated kinase 25.5IL6Interleukin 6 (interferon, beta 2)5.5HIST1H4HHistone 1, H4h5.4FNIP1folliculin interacting protein 15.3HIST2H2BEHistone 2, H2be5.2MCL1Myeloid cell leukemia sequence 1 (BCL2-related)5.1DUSP10Dual specificity phosphatase 105.1DHRS2Dehydrogenase/reductase (SDR family) member 25.1SYNE1Spectrin repeat containing, nuclear envelope 15.1HAS2Hyaluronan synthase 25.1PER1period homolog 1 (Drosophila)5.1ZBTB10zinc finger and BTB domain containing 10-2.0FN1Fibronectin 1-2.0BCL2L11BCL2-like 11 (apoptosis facilitator)-2.0COL5A1collagen, type V, alpha 1-2.0VAV3vav 3 oncogene-2.0CSRP2BPCSRP2 binding protein-2.0PRSS23protease, serine, 23-2.1NRP2Neuropilin 2-2.1DAPK1Death-associated protein kinase 1-2.1CDH1Cadherin 1, type 1, E-cadherin (epithelial)-2.2ST8SIA4ST8 alpha-N-acetyl-neuraminide alpha-2,8-sialyltransferase 4-2.2RNASE4Ribonuclease, RNase A family, 4-2.2FGGFibrinogen gamma chain-2.1STAT4Signal transducer and activator of transcription 4-2.1NTRK3Neurotrophic tyrosine kinase, receptor, type 3-2.3PCDH9Protocadherin 9-2.2CYP4F3Cytochrome P450, family 4, subfamily F, polypeptide 3-2.2ST8SIA4ST8 alpha-N-acetyl-neuraminide alpha-2,8-sialyltransferase 4-2.3MAP2K6Mitogen-activated protein kinase kinase 6-2.3LXNlatexin-2.3TM4SF20transmembrane 4 L six family member 20-2.4AKR1C1///AKR1C2Aldo-keto reductase family 1, member C1 (dihydrodiol dehydrogenase 1; 20-alpha (3-alpha)-hydroxysteroid dehydrogenase)///aldo-keto reductase family 1, member C2 (dihydrodiol dehydrogenase 2; bile acid binding protein; 3-alpha hydroxysteroid dehydrogenase, type III)-2.7SLC40A1Solute carrier family 40 (iron-regulated transporter), member 1A549 human lung epithelial cells were exposed to crocidolite asbestos at 6 g/cm2 for 24 h. Genes for which expression levels increased 5.0 fold or greater, or decreased 2.0 fold or greater, having q values less than 0.05 are shown. Probe sets associated with products of unknown function were not included. Bold-faced genes were not observed in the statistically upregulated expression profile of hydrogen-peroxide-treated A549 cells [36]. A list of all genes for which expression levels changed 2.0 fold or greater can be found as Additional File 2.Hierarchical clustering analysis (Figure 2A) of our data set showed that the majority of the up-regulated probe sets clustered together in four different hierarchical clusters. The down-regulated probe sets were mainly in one hierarchical cluster, but had several individual genes spread throughout the dendrogram. Within the four up-regulated clusters, 24 genes that were upregulated 5-fold could be mapped by functional String analysis shown in Fig 2B. Most of the gene products are connected; i.e., could be linked functionally by way of known molecular interactions. Three major pathways (MAPK, JNK/SAPK and cytokine-cytokine receptor interactions) are represented in the cluster. This result is consistent with what is currently known about the effect of asbestos in epithelial and mesothelial cells. When the down-regulated genes were functionally clustered, only two (fyr and uvo) out of 16 had a connection (data not shown).Figure 2Unsupervised hierarchical clustering analysis of A549 probe sets in which the expression was altered by crocidolite.(A) The main dendogram represents the intensity of each probe set in relation to the entire data set with green being low and red being high. Four major clusters of upregulated genes and two clusters of downregulated genes were observed and are shown as bracketed dendograms where the color represents the intensity of each probe set in relation to each specific cluster. (B) STRING analysis using the genes within the upregulated clusters demonstrated a functional relationship between 24 of the genes which encompassed cytokine, MAPkinase and JNK/SAPK signaling pathways.Validation of the Microarray DataValidation of the microarray data was confirmed by quantitative RT-PCR on four of the genes (Egr1, ATF3, c-Jun, and JunB) and was further corroborated with previously noted results from experiments using human cells/cell lines (Table 2). Quantitative RT-PCR also demonstrated that increases in Egr-1, ATF3, c-Jun, and JunB message occurred after 6 hours and continued to increase through 24 hours (data not shown). Increased mRNA expression from ATF3, c-Jun, and JunB was also observed by quantitative RT-PCR in primary human epithelial cells from small airways (SAEC) treated with crocidolite (data not shown). Increased protein expression in A549 cells exposed to crocidolite was observed for ATF3, c-Jun, and JunB (data not shown). Finally, previous results using asbestos-treated A549 cells have demonstrated increased mRNA and/or protein levels for IL8, TP53, HMOX1, CDKN1A, SOD2 and c-myc (see Table 2 for additional references). The reproducibility of the microarray replicates, the significance level of the expression changes, and the validation by quantitative RT-PCR suggest that a highly useful and validated data set was obtained in this study. Furthermore, the large number of significant genes allowed us to perform a comprehensive analysis of the crocidolite-induced transcriptome in human lung epithelial cells.Table 2Validation of the microarray data from crocidolite-treated human A549 cellsGeneAlias for Protein ProductAffymetrix* (fold change) this studyQRT-PCR (fold change) this studyChanges in mRNA levels in human cells observed by other groupsEGR1Egr1115.7281.7 ± 3.8ATF3ATF325.323.7 ± 0.1c-Junc-Jun8.610.5 ± 0.7JunBJunB2.23.6 ± 0.2IL8Interleukin 816.6N.D.Increased [73]TP53p531.8N.D.Increased [16,74]SOD2Mn-SOD2.2N.D.Increased [75]‡HMOX1Heme oxygenase 12.1N.D.Increased [75,76]†c-mycmyc2.0N.D.Increased [77]CDKN1Ap21, Cip11.9N.D.Increased [16,74]*represents significantly greater than control, q less than 0.05; N.D., not determined† chrysotile-treated A549 human lung epithelial cells‡ crocidolite-treated human pleural mesothelial cellsSpecific Changes in the Crocidolite-Treated TranscriptomeAlthough it is clearly established that asbestos induces DNA damage in both primary cells and cell culture lines, the literature illustrates that the molecular mechanism underlying this process and how the fate of the cell is dictated are multifaceted. Although ROS are intimately tied to the mechanism of asbestos-induced fibrosis and carcinogenesis, ROS alone do not offer a complete understanding of the asbestos response. In order to identify genes for which expression level changes are specific for crocidolite compared to oxidative stress, we compared our data set to expression changes that occur in A549 cells when exposed to hydrogen peroxide. In a previous study that documented changes in the A549 cell transcriptome, Cotgreave and co-workers [36] showed that treatment of A549 cells with hydrogen peroxide results in DNA damage and apoptosis. Both hydrogen peroxide and crocidolite asbestos induced the upregulation of TNFRS10B, PPP1R15A, GADD34, CDKN1A, BTG2, DUSP1, DUSP5, DUSP14, SDC4, GDF15, IL8, ADM, FST, IER3, FOS, HMOX1, ATF3, and ZFP36. Therefore, expression changes in these genes may represent a response to an oxidative stress. Several unique genes that are differentially regulated in the crocidolite data set are noted in Table 1 and include some genes which are associated with carcinogenesis such as STC1, IL6, FN1, BRE, and PTGS2. Differences between the hydrogen peroxide-treated and crocidolite-treated transcriptomes may be due to the additional iron released from fibers, present in cells treated with crocidolite, or reactive nitrogen species, changes in glutathione content and/or the involvement of processes initiated at the cell surface by the fiber. It will be very interesting to investigate if these apparent crocidolite-specific gene regulations play a role in crocidolite-induced cytobiological endpoints.Ontological Analysis of the Crocidolite-Treated Lung TranscriptomeIn order to identify themes in the global expression pattern of crocidolite-treated A549 cells, we used Gene Ontology (GO) classification. The differentially expressed genes (up-regulated/down-regulated ± 2-fold with p-values less than 0.2) were analyzed using Onto-Express . The Biological Process tree was expanded to the fourth tier in the hierarchy as a balance between specificity and coverage. Processes affected by crocidolite are shown in Figure 3A. Within the Cellular Metabolism classification, processes involved in transcription and phosphorus metabolism were highly represented. Expansion of Molecular Functions to the third tier revealed a large number of genes with functions related to transcription (over-represented in Protein binding (Figure 3B)). This observation was expanded upon by performing a global test [37-39] to identify specific gene ontologies related to transcription, where the expression levels of all the genes in the ontology give useful information for predicting the clinical outcome (or in this case, a cellular outcome of crocidolite exposure). Ontologies of particular interest related to transcription factor activity (GO:0003700) are summarized in Figure 3C, with nodes colored to correspond to global test p-values; the lighter nodes have lower p-values. To reduce visual clutter, only the last four digits of each ontology's identifier are reported in Figure 3C. The ontological analysis illustrates that transcriptional corepressor activity and repressor activity are highly represented in the data set. Overall the transcriptome of crocidolite-treated human lung epithelial (A549) cells is heavily populated by gene products associated with transcription.Figure 3Gene ontology analysis of the (A) Biological Processes and (B) Molecular Functions most affected by crocidolite in A549 cells. In (C) a global test to identify ontologies related to transcription was used to pinpoint significant functions. The p-values for each of the GO terms (abbreviated as the last four digits of each ontology's identifier) has been overlaid onto the hierarchical tree where the darkest gray node represents a p = 0.034 and the lightest represents a p = 0.017.We also noted that DNA repair terms were not highly represented in the data set. Although this result was initially surprising, we noted that the transcriptome of A549 cells exposed to hydrogen peroxide, which also induces DNA damage, was also underrepresented in DNA repair terms [36]. In order to gain a better understanding of how the transcriptome of crocidolite-treated A549 cells is reorganized in response to the resulting DNA damage, the results of the global test discussed above were also applied to ontologies related to DNA damage/repair. Although this analysis (see Additional File 3) confirms the overall observation that many of the processes related to DNA repair/damage are not highly represented, the analysis also highlights specific nodes in the GO tree related to DNA damage-induced phosphorylation (GO:6975), DNA dealkylation (GO:6307), DNA methylation (GO:6306), and DNA alkylation (GO:6305). In light of the several studies directed at linking promoter methylation status and carcinogenesis (reviewed in [40]), it is interesting to point out that in as little as 24 hours of crocidolite exposure, the A549 transcriptome may be poised to affect the DNA methylation status that is associated with lung cancer pathogenesis.Identification of Novel, Putative Crocidolite-Related GenesEven though this study was initiated to understand the broad changes that occur in the lung transcriptome upon crocidolite exposure, we also sought to identify novel genes that had no previous association with crocidolite and to identify genes that may be downstream targets of well-known crocidolite-related players. Given the complex effects that asbestos has on the lung system, we expected that a systems biology approach may provide novel avenues to study. NR4A1, NR4A2, and NR4A3 belong to the steroid nuclear hormone receptor superfamily of immediate early genes that are induced by serum, growth factors and receptor engagement and are thus implicated in cell mitogenic responses. Although previously characterized as pro-survival, studies have also suggested an important role for these receptors in cell transformation and tumorigenicity via their anti-apoptotic and pro-apoptotic functions [41]. Thus, depending on cellular context, these gene products may serve as switches in determining cell fate. All three members of this family show increased expression in our data set (Additional File 2) but have not previously been implicated in crocidolite-induced pulmonary toxicity.Modulation of apoptosis can also be affected by BRE (brain and reproductive organ-expressed protein), a stress-modulating protein also known as TNFRSF1A modulator. BRE expression was upregulated >7-fold in crocidolite-treated A549 cells. Exogenous overexpression of BRE can attenuate intrinsic apoptosis and promote growth of the transfected Lewis lung carcinoma line in mice [42] which is consistent with the recent finding that BRE protein is overexpressed in human hepatocellular carcinomas [43]. Given the ability of BRE to interact with both Fas [44] and the TNF receptor 1 [45], and the observation that TNFα can attenuate asbestos cytotoxicity in mesothelial cells [17], it will be very interesting to investigate possible roles of BRE in crocidolite-treated human lung cells. Also noteworthy is the upregulation of several \"early response\" NF-κB targets [46] in our dataset (Additional File 2) including TNFAIP3, IL8, IL6, CXCL1, CXCL2, CXCL3, PTGS2, and PLAU. Activation of the NF-κB pathway is thought to play a critical role in cell survival in asbestos-treated cells [17], but only a few of the downstream targets of NF-κB have been identified.The relationship between asbestos and calcium has received little attention in recent years, but the initial studies suggest that calcium may have an important role to play (reviewed in [47]). Perhaps most compelling is the ability of the calcium-chelator Quin-2 to prevent crocidolite-induced DNA breaks [48]. Additionally, several players in the asbestos response are regulated by calcium levels, e.g., protein kinase C [15]. We were therefore interested in determining if the crocidolite-treated transcriptome demonstrated any clues regarding the regulation of calcium homeostasis. We observed that expression of MCTP1 and STC1 were both upregulated in our data set by 7.3- and 12.9-fold, respectively. MCTP1 is a transmembrane protein that binds calcium ions via C2 domains. Unique properties of this protein suggest that these proteins function in Ca2+ signaling at the membrane [49]. Stanniocalcin 1 (STC1) has long been studied as a regulator of both phosphate and calcium homeostasis in bony fish, but has recently received attention in mammalian systems. STC1 is a glycoprotein present in a variety of mammalian tissues where it can function as a regulator of gene expression and modulator of transendothelial cell migration [50], and can also affect cellular metabolism by perturbing mitochondrial electron transport chain and mitochondrial calcium transport [51]. STC1 affects calcium homeostasis in the heart [52,53] and the brain [54]. Growing evidence also points to a correlation between STC1 expression and the development of human cancers [55,56]. Quantitative RT-PCR also demonstrated a 30.7 ± 5.1-fold increase in STC1 message in primary human SAEC exposed to crocidolite.Finally, in recent years a significant number of studies have been directed at understanding how the disruption of dynamic chromatin remodeling is linked to carcinogenesis. Mechanisms including the previously mentioned DNA methylation status and the use of histone modifications have led to the discovery of prognostic biomarkers [57] and the use of HDAC inhibitors as cancer therapeutics [58]. It is of little surprise then to find several differentially regulated genes in crocidolite-treated A549 cells that could participate in chromatin remodeling including the Jumonji domain histone demethylases JMJD1C, JMJD3 and JMJD1A, all of which showed increased expression (Supplementary Table 1). Other genes of interest are discussed below.Pathway Analysis Provides Unique View of Function-Based Networks in Crocidolite-Treated CellsIn order to extract novel biological insight from the large number of genes upregulated/downregulated in our study, we employed a structured network knowledge-based approach to analyze genome-wide transcriptional responses in the context of known functional interrelationships among proteins, small molecules and phenotypes . Networks are generated not only based on functional interactions but also statistical likelihood. Genes/gene products are represented as nodes and are color-coded to represent fold-change in expression level. Interactions between nodes are designated as edges, or lines, and represent physical, transcriptional, and enzymatic interactions.Statistically Significant Function-based Networks and PathwaysUsing expression changes that were differentially regulated by ± 2-fold and having p values less than or equal to 0.01, Ingenuity Pathway® Analysis demonstrated a highly complex set of 16 interconnected networks. These networks were related to cell death, cancer, cell cycle, cellular growth and proliferation, and gene expression. All networks had at least one gene in common and many had two genes in common, underscoring the interplay and complexity of the crocidolite response. The top scoring network (Figure 4A) was composed of genes related to cell death, organism survival and cancer, and highlighted MYC, PDGF BB, and EPAS1/HIF2 as prominent nodes. In particular, EPAS1 was a node for four genes whose expression levels changed 6-fold or greater. EPAS1 (endothelial PAS domain protein-1, or HIF2α) is one of the transcription factors which belong to the basic helix-loop-helix PAS family. It shares sequence similarity to HIF-1α and analogously to HIF-1α, regulates transcription of VEGF and is observed to be upregulated at both the message and protein levels in A549 cells as a result of hypoxia [59]. Prognostic significance of increased levels of EPAS1 mRNA and/or protein has been observed in liver [60] and colon [61] cancers. Interestingly, EPAS1 -/- mice show impaired reactive oxygen species homeostasis [62], which may be linked to the role of EPAS1 in maintaining mitochondrial homeostasis [63], creating a hypothetical link between EPAS1 function and response to crocidolite.Figure 4Spatial depiction of the top scoring network and canonical pathways detected in crocidolite-treated A549 cells by Ingenuity Pathway® Analysis. The network was algorithmically generated based on the connectivity of each of the transcripts and the molecular interaction knowledge base. Each node represents a gene or gene product for which mRNA expression was upregulated (red) or downregulated (blue) in crocidolite-treated A549 cells. Edges/lines connecting the nodes represent molecular interactions between genes and/or gene products and are supported by at least one reference from the literature, a textbook, or from canonical pathway information stored in the Ingenuity Pathways Knowledge Base. The Nrf2-mediated oxidative stress canonical pathway identified by Pathway Analysis software shows differential upregulation of select genes within the cytoprotective arsenal.This first network (Figure 4A) is linked to the Nrf2-mediated oxidative stress pathway (one of the top scoring canonical pathways in our data set shown in Figure 4B) via the MAFF node. Cellular use of this canonical pathway has been linked to tumor cell survival by maintaining cellular redox homeostasis and protection against oxidative insult. Heterozygous Nrf2 (+/-) mice exposed to crocidolite fibers exhibit accelerated development of malignant mesotheliomas compared to wild-type littermates [64]. Several of the known Nrf2 targets did not show significant changes in expression levels in our data set (data not shown). Although not all of the downstream targets of NRF2 were differentially upregulated upon exposure to crocidolite at 6 μg/cm2, message levels for many of the Nrf2 targets were present in above average amounts on both the control and treated chips. Furthermore, data suggests that this pathway may be activated in a hierarchical fashion [35], dependent on exposure. Selective activation of the Nrf2 pathway may contribute to carcinogenicity of the crocidolite fibers, while dysfunctional constitutive activation of Nrf2 [65] has been observed in non-small-cell lung cancer.The topscoring network illustrates the interplay between processes involved in both cell survival/proliferation and cell death/apoptosis when A549 cells are exposed to crocidolite. Activation of the Nrf2 cytoprotective pathway in the transcriptome of crocidolite-treated A549 cells is imperfect. The observed selective Nrf2 target expression would define an environment that is rich in hydrogen peroxide and incapable of redox homeostasis.A Global View of the Functional Networks in A549 Cells in Response to CrocidoliteAnother approach to gain insight about the functional significance of the global changes in crocidolite-induced gene expression is to merge the individual networks and then identify nodes that are used frequently. Over 500 nodes were observed within the 16 networks, making presentation of the global cellular network difficult. Instead, the top five networks were merged using all known interactions in the knowledge-based database and the nodes arranged according to subcellular location resulting in Figure 5A. This representation comprises ~150 genes and their interactions and illustrates a transcriptome encoding an array of extracellular growth factors and cytokines. Even though a population of exposed cells undergoes apoptosis, some cells may survive and be stimulated to proliferate based on factors released from neighboring cells. Several high impact nodes including MYC, JUN, Akt, p38, and PDGF BB, all of which are established players in the response to asbestos, are also present. Other genes that formed prominent nodes in the prototypical cell or adjacent networks included PTGS2, SMARCA4, PTEN, and E2F1. They are shown as separate networks for clarity in Figures 5B, C and 5D. The PTGS2 product, Cox-2, PTEN and E2F1 have previously been implicated in carcinogenesis, but their roles in the epithelial cell response to asbestos have not been studied. Brg1, the protein product of SMARCA4, is a SWI/SNF related chromatin remodeling factor which recognizes acetylated lysine groups through bromo domains and is involved in cell growth arrest and apoptosis. Oxidative stress and TNF-α induce histone acetylation and NF-κB/AP-1 activation in alveolar epithelial cells [66] suggesting a potential mechanism to alter gene transcription in lung inflammation using Brg1. Inspection of the network surrounding the E2F1 node identified several genes whose expression was differentially regulated by asbestos that also demonstrated molecular interactions with TNF. Given the relationship between TNF and NF-κB activation observed in mesothelial cells exposed to crocidolite [17], and the recent identification of E2F1 as a transcriptional activator recruited by NF-κB [67], investigation into the role that E2F1 plays in human lung epithelial cells exposed to crocidolite should be forthcoming.Figure 5Pathway analysis of representative genes involved in the response of A549 cells exposed to crocidolite. In (A) the top five scoring networks were merged to create a cellular model consisting of ~150 genes for which expression levels changed ± 2-fold and demonstrated p-values less than 0.01. Nodes representing genes are colored red for upregulation or blue for downregulation and the intensity of the color reflects the degree of up- or downregulation. Lines connecting the nodes are indicators of interactions found in the knowledge database or current literature. In (B) and (C) the networks surrounding PTGS2, SMARCA4, and PTEN have been expanded for clarity. In (D) the relationship between E2F1 and TNF is observed.This analysis has provided the first function-based global view of the crocidolite-treated A549 transcriptome. Several new candidate crocidolite-related genes were identified in the context of experimentally observed findings. Apparent from the global analysis is a transcriptome bearing signatures of both apoptosis/cell death and cell survival/proliferation.Using Pathway Analysis to Probe the Role of p53Pathway Analysis also detected significance in p53-mediated processes. This result is consistent with the observation that both amosite and crocidolite induce p53 activation. Specifically, Kamp and co-workers who showed that p53 mediates amosite-induced apoptosis through mitochondrial dysfunction in A549 cells [10]. Although the mechanism is not clear, ROS generated by the mitochondria, p53-mediated transcription, and translocation of p53/BAX appear to be necessary components for apoptosis to occur. The authors suggested that other pro-apoptotic pathways may also be activated and a more thorough study of p53-targeted transcription pathways was needed. Figure 6 shows a scheme representing these findings overlaid with the expression changes from our data set and molecular interactions identified through network analysis. Since it is unknown whether ROS detection and p53 activation require new protein synthesis, we did not restrict this part of the analysis to differentially regulated genes. Several genes whose expression depends on p53 could be identified that were associated with both apoptosis and proliferation or survival. These results may reflect the expected heterogeneity in the population of cells after exposure to crocidolite, as discussed previously. Although crocidolite induces apoptosis in many of the cells, the bulk transcriptome shows evidence that transcripts are present which could tip the balance in favor of cell cycle arrest and/or cell survival or proliferation. Message levels of the p53-induced molecular switch p21CIP/WAF1 and KLF4 were upregulated in crocidolite-treated cells. Together these two products can have an anti-apoptotic effect [68].Figure 6Asbestos induces a p53-mediated response in A549 lung cells. Experimental data is indicated by the thick blue edges and is discussed in more detail in the text. Known p53 targets which may have a role in the asbestos response and their biological effects are shown using thin gray lines. Nodes are color coded shades of red when expression levels increased and shades of blue when expression levels decreased. Nodes surrounded by dashed lines represent genes/gene products which were not represented on the gene chip but which displayed a functional relationship to queried nodes.The importance of p53 in mediating apoptosis in asbestos-treated A549 cells has been documented [10]. Analysis of p53 targets upregulated in the crocidolite-treated A549 transcriptome identified several candidate genes that may function in the observed apoptosis. Interestingly, we also identified several p53 targets associated with cell survival/proliferation. This suggests that p53 may be an important molecular turning point in the decision of crocidolite-treated cells to undergo apoptosis or to proliferate, even in the presence of damaged DNA.ConclusionIn this study, we have provided a statistically robust and comprehensive global gene expression profile of A549 human lung epithelial cells exposed to crocidolite asbestos. Our data reveal a much altered transcriptome in which a large number of genes show upregulated expression. Crocidolite-treated A549 cells are rich in transcripts encoding extracellular growth factors and cytokines and intracellular regulators/mediators of transcription. A global view based on functional molecular interactions illustrated an intricate network of paths associated with both apoptosis and proliferation/cell survival. This network allowed for the identification of novel, putative crocidolite-related genes, leading to several new hypotheses regarding genes which are important in the response of human lung epithelial cells to crocidolite.Our analysis demonstrates the power of a functional network genomics approach to 1) identify and explore relationships between genes of known importance 2) identify novel candidate genes, and 3) observe the complex interplay between genes/gene products that function in seemingly different processes. This study represents the first function-based global approach toward understanding the response of human lung epithelial cells to the carcinogen crocidolite. We have provided graphical representations of the highly interconnected networks that will be instrumental in modeling the impact of new research findings. Our global function-based approach introduces new insights and novel avenues which can now be investigated in more detail to understand the effects of crocidolite on the human lung.MethodsReagents and AntibodiesCrocidolite was obtained from Dr. Richard Griesemer, NIEHS/NTP. The chemical formula for crocidolite is Na2FeIII2(FeII, Mg)3Si8O22(OH)2. It has a mean length of 10 μm, a density of 3.2–3.3 g/cm3 and contains 27% iron by weight [69]. A custom prepared Ham's F-12 cell culture medium free of iron salts and 0.5% trypsin with 0.2% EDTA were obtained from Invitrogen Inc. (Carlsbad, CA). Fetal bovine serum (FBS) was purchased from Hyclone (Logan, UT). Antibodies for Egr-1 (sc-189) and ATF3 (sc-188) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA) and the antibody for β-actin was purchased from Sigma (St. Louis, MO). Anti-rabbit and anti-mouse enzyme-conjugated secondary antibodies were purchased from Jackson Immunoresearch Laboratories Inc. (West Grove, PA). Primers were purchased from IdtDNA (Coralville, IA). All remaining reagents were purchased in the highest purity possible.Cell Culture and TreatmentA549 cells were purchased from American Type Culture Collection and grown in F12 medium without iron, with 10% FBS and 50 μg/mL gentamicin (BioWhittaker, Walkersville, MD). Cells were plated for treatments at a concentration of 20,000 cells/cm2 and incubated 24 h before treatment. For all treatments, medium was removed and replaced with either fresh medium (for control samples) or medium containing the appropriate stimulus (treated samples). For crocidolite treatments, fibers were suspended in sodium bicarbonate at a concentration of 1 mg/mL and immediately diluted to a final concentration of 6 μg/cm2 in complete medium. Sodium bicarbonate is used specifically to avoid iron mobilization in the media before the fibers can by endocytosed and maintains the pH [26]. These are the exact same conditions and methodologies used previously in several other studies [9,11,25-27].RNA Extraction and Use of DNA MicroarraysmRNA was isolated from cells at passage 9 using TRIzol® reagent from Invitrogen (Carlsbad, CA) following manufactures directions. This procedure yielded RNA of both high quantity and quality, as verified by both the A260/A280 ratio (> 1.8) and the Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA). The extracted RNA was processed by the Affymetrix core service at the Center for Integrated BioSystems, Utah State University (CIB-USU; Logan, UT). RNA was processed as per the manufacturer's instructions (Affymetrix Inc., Santa Clara, CA) for the first-strand cDNA synthesis and amplification, followed by cDNA synthesis and labeling. The cDNA was hybridized overnight to the HGU133-Plus2.0 Human Genome 2.0 genechips and post-hybridization washing, detection and data collection were performed as per manufacturer's instructions. Three biological replicates were performed. Each replicate consisted of a control plate and a crocidolite-treated plate. All experiments were done at passage 9.Statistical AnalysesUsing tools from the Bioconductor project [70], the data were pre-processed using the RMA algorithm [71], and a test for differential expression between control and treated (crocidolite-exposed) conditions was performed using the moderated t-statistic from the limma/eBayes approach [72]. The raw p-values from this test were adjusted to control the false discovery rate [34] at 0.05. Dendrograms (Figure 2) were drawn by Hierarchical Clustering Explorer (HCE; ) software version 3.0 to summarize RMA-normalized data after averaging across replicates. The cutoff value for minimum levels of gene expression was based on the whole genome expression profiles generated by HCE. The median cutoff value was set at the software-calculated median of gene expression and was colored black. The data were further subdivided into gene functional classes that had been annotated by the Human Genome Database. Each class was then clustered separately and the determined cutoff values were applied to generate colored expression maps for each individual class for ease of visualization. In addition, a global test [37-39] was used to identify gene ontologies of interest, where the expression levels of the genes in the ontology were statistically significant (p-value less than 0.05) for predicting the clinical outcome (crocidolite exposure). STRING analysis was performed at  using default parameters.Quantitative RT-PCRmRNA was isolated using TRIzol® reagent from Invitrogen (Carlsbad, CA) following manufactures directions. DNase from the RNAqueous®-4PCR kit (Ambion Inc., Austin, TX) was added to eliminate genomic DNA and removed following the manufacture protocol. cDNA was transcribed using SuperScript™ II Reverse Transcriptase (Invitrogen Inc., Carlsbad, CA). Specificity was confirmed by performing a melting curve on the final amplicons and running the amplicons on a 2% agarose gel. All qRT-PCR samples were also run on samples without reverse transcriptase to confirm the product was from mRNA, not DNA, and all of these samples showed no amplicon. Three different dilutions of cDNA were used to confirm that the samples were in the linear range.Western blot analysisCells were harvested with trypsin and then lysed using RIPA buffer (0.15 M NaCl, 50 mM Tris, pH 7.2, 1% deoxycholate, 1% Triton X-100, 0.1% SDS) containing protease inhibitors (30 μL/mL aprotinin, 4 μg/mL leupeptin, 4 μg/mL soybean trypsin inhibitor, 0.1 mM PMSF and 1 μM benzamidine). After incubation on ice for 30 min, cells were centrifuged for 10 min at 8200 × g to remove cellular debris. Protein concentrations were determined using the Bradford method (Bio-Rad Laboratories, Hercules, CA). Cellular protein (50 μg) was analyzed using SDS-PAGE (10–15%) and transferred to polyvinyl difluoride (PVDF) membranes. Primary antibodies were diluted 1:200 in 5% dried milk in TTBS and incubated for 1 h at room temperature. Secondary antibodies were diluted 1:5000 and incubated for 1 h at room temperature. Blots were visualized using ECL Plus reagent from GE Healthcare Life Sciences (Piscataway, NJ).Pathway AnalysisData were analyzed through the use of Ingenuity Pathways (Ingenuity® Systems, ). The asbestos data set with gene identifiers and corresponding expression values was uploaded into the application. Each gene identifier was mapped to its corresponding gene object in the Ingenuity Pathways Knowledge Base. A cutoff of p < 0.01 was set to identify genes whose expression was significantly differentially regulated by ± 2-fold. These genes, called focus genes, were overlaid onto a global network developed from information contained in the Ingenuity Pathways Knowledge Base. Networks of these focus genes were then algorithmically generated based on their connectivity. Biological functions associated with these focus genes were identified using the Functional Analysis using the Ingenuity Pathways Knowledge Base. Fischer's exact test was used to calculate a p-value determining the probability that each biological function assigned to the data set was due to chance alone. Networks are graphical representations of the molecular relationships between genes/gene products. Genes or gene products are represented as nodes, and the biological relationship between two nodes is represented as an edge (line). All edges are supported by at least one reference from the literature, a textbook, or from canonical pathway information stored in the Ingenuity Pathways Knowledge Base.Authors' contributionsLO-B carried out the cell culture and RNA preparations. BG carried out the initial analysis of the raw microarray data and aided LO-B in the clustering of the data. JRS carried out further statistical analyses in order to compare the data set to others and performed the gene ontology analyses. JPH carried out RT-PCR confirmations. AEA and JMH conceived of the study. JMH performed the Pathway Analysis, coordinated all other analyses and wrote the manuscript. All authors read and approved the final manuscript.Supplementary MaterialAdditional File 1Low variability in the dataset. A comparison of the RMA expression values for the 54,675 probe sets on each of the six arrays. The tighter relationships within the Control and Treated groups are indicative of a data set with low variability. This low variability allowed the identification of a large number (2,546) of statistically significantly differentially expressed genes.Click here for fileAdditional File 2List of genes for which expression was increased or decreased two-fold or greater.Click here for fileAdditional File 3Global test to identify specific gene ontologies related to DNA repair and damage. A global test (Goeman et al., 2004; Goeman et al., 2005; Goeman and Oosting 2007) was used to identify specific gene ontologies related to DNA repair and damage. The p-values for each of the GO terms (abbreviated as the last four digits of each ontology's identifier) has been overlaid onto the hierarchical tree where the darkest blue node represents a p = 0.15 and the lightest represents a p = 0.016.Click here for file\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2533038.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2533038",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2533038\nAUTHORS: Sandie Le Guédard, Valérie Faugère, Sue Malcolm, Mireille Claustres, Anne-Françoise Roux\n\nABSTRACT:\nPurposeProtocadherin-15 (PCDH15) is one of the five genes currently identified as being mutated in Usher 1 syndrome and defines Usher syndrome type 1F (USH1F). When PCDH15 was systematically analyzed for mutations in a cohort of USH1 patients, a number of deletions were found. Here we characterize these deletions as to extent, position, and breakpoints.MethodsMicrosatellite and single nucleotide polymorphism (SNP) analyses, used in a preliminary survey of an Usher cohort of 31 patients, revealed large deletions in three patients. These deletions were further characterized by semiquantitative PCR assays to narrow down the breakpoints.ResultsThe analysis of the three large deletions revealed that all six breakpoints are different. The breakpoint junction was identified in one patient and the four other breakpoints were mapped to 4 kb. There were no specific distinguishing features of the isolated breakpoints.ConclusionsA complete screen of PCDH15 should include a search for large deletions. Failure to screen for gross genomic rearrangements is likely to significantly lower the mutation detection rate. A likely explanation for the high rate of such deletions is the unusual gene structure. PCDH15 gene spans nearly 1 Mb for a corresponding open reading frame (ORF) of 7,021 bp. The intron sizes of PCDH15 are up to 150 kb, and the first three exons of the gene cover 0.42 Mb. The genomic structure of any gene should be taken into consideration when designing a mutation screening strategy.\n\nBODY:\nIntroductionUsher Syndrome type 1 (USH1) is the most severe form of Usher syndrome [1] and is characterized by congenital profound deafness, vestibular areflexia, and (generally) early onset of retinitis pigmentosa (RP). Six loci have been mapped and five genes have been identified: myosin VIIa (MYO7A), cadherin-23 (CDH23), protocadherin-15 (PCDH15), harmonin (USH1C), and SANS (USH1G) [2,3]. MYO7A appears to be the most frequently involved, and mutations have been reported in 29-54 [4,5] percent of cases but there have been few systematic studies on a cohort of patients [6,7].The USH1F locus was mapped about ten years ago to chromosome 10q21-22, and the PCDH15 gene was cloned in 2001 [8]. Several USH1F transcripts have been identified in humans, and the longest isoform (isoform A), consisting of 1 noncoding and 32 protein-coding exons, encodes a 1955 amino acid transmembrane protein that is predicted to contain 11 cadherin repeats, one transmembrane domain, and a cytoplasmic domain containing two proline-rich regions [8-10]. Recently, multiple alternative protocadherin-15 transcripts were characterized in the mouse inner ear. These transcripts define four major isoform classes alternatively spliced, and two of them encode new cytoplasmic domains, raising the number of exons to 39. Three of these isoforms have different spatiotemporal expression patterns in developing and mature hair cells, suggesting a specific role for each protocadherin-15 isoform in the sensory hair bundle [11]. These alternatively spliced exons encoding the two novel cytoplasmic domains were also detected in human retina, indicating that the organization of the human gene could be more complex than was initially thought [11]. Together with other USH1 proteins protocadherin-15 ensures hair bundle morphogenesis [12] via its binding to harmonin [13,14] and myosin VIIa [15].Around 25 mutations have been documented, nearly all predicted to lead to premature termination of the proteins (6-10). Ouyang et al. [6] studied PCDH15 together with other USH1 genes in a cohort of patients and found PCDH15 mutations in five patients but identified both causative mutations in only one of them.We present in this study an exhaustive analysis of the deletions that were detected in three different families [7]. We show that not only are all deletions different, they also account for a significant proportion of PCDH15 mutations, probably because of the genomic structure of this gene. We suggest that deletion screening should be part of the molecular analysis for PCDH15 and any other genes that have such an unusual genomic structure.MethodsPatientsThe project was approved by the local ethics committee. Consent to genetic testing was obtained from adult probands or parents of minors. Patients meeting the diagnostic criteria for USH1 were previously described [7]. USH1 was diagnosed on the basis of congenital profound sensorineural deafness, vestibular dysfunction, and retinal degeneration.U153 and U297 were sporadic cases whereas two affected siblings were available for family U382. All patients underwent audiological examination and all presented with profound deafness. The age of walking was delayed and ranged from 18 months (U153) to 36 months (U297). Electroretinograms (ERG) and fundus examinations were altered in all cases when diagnoses were made at 9 years old (U153 and U297). The ERG was already extinguished at 4 years old in both siblings in family U382.Sequencing analysis of PCDH15PCR amplification and sequencing of the PCDH15 gene, corresponding to isoform A as described by Ahmed et al. [10] (NM_033056), has already been reported [7]. PCR parameters and primers have already been published in the study from Roux et al. [7].HaplotypesHaplotypes were constructed from a combination of intragenic single nucleotide polymorphisms (SNPs) and seven microsatellite markers: D10S1124-D10S2522-PCDH15(IVS3-(CA)-D10S2536-D10S546)-D10S1643- D10S1762. The location of the markers is reported in Figure 1A. Sequences of the microsatellite primers are available on gdb with the exception of IVS3-(CA)-IVS3-F: 5'-GTA TGT ACA GTT AAT TGG TAG-3'; IVS3-R: 5'-GAT GCA GGT ATG GTT TCA G-3'.Figure 1Pedigrees at the USH1F locus. A: Genomic organization of the PCDH15 isoform A with numbering of the exons as described by Ahmed et al. [10] NM_033056. The noncoding exon 1 is represented in grey. Positions of the BAC clones and the microsatellite markers are indicated. B: Representation of the three pedigrees with haplotypes. (-) represents lack of amplification; N represents normal. In the first pedigree \"Ins\" stands for the c.423_430dup mutation (insertion of 7 bp). The different haplotypes are indicated by rectangles with various fillings.Microsatellites were analyzed on an ABI 3100 Avant genetic analyzer (Applied Biosystems, Applera France, France) whereas the SNPs were analyzed by direct sequencing.Semiquantitative assaysTwo semiquantitative approaches were used in parallel: the quantitative multiplex PCR of short fluorescent fragments (QMPSF) and semiquantitative nonfluorescent multiplex PCR. QMPSF containing multiplex PCR of 3-9 amplicons were analyzed on an ABI310 (Applied Biosystems). We applied to the PCDH15 gene the stategy used by Audrezet et al. for the CFTR gene [16]. Semiquantitative nonfluorescent multiplex PCR products were separated under nondenaturing conditions on a liquid chromatography system (3500 Wave HS system coupled to an HSD system, Transgenomic, Elancourt, France) then quantified by fluorescent detection using a post column intercalation dye, based on guidelines described by Dehainault et al. [17]. One advantage of the semiquantitative nonfluorescent multiplex PCR analyzed on the 3500 Wave HS system is that the primers used for routine sequencing can also be used to determine if a particular exon has been deleted.To narrow down the deletion breakpoints, we used PCR walking methods that included laboratory-designed amplicons localized in a first step every 50 kb both upstream and downstream of the identified deletions. The primers were chosen according to the sequence of the bacterial artificial chromosome (BAC) clones (their accession number is given in Figure 1A). Once a breakpoint was localized between two adjacent amplicons, further primers were designed for new amplicons until this initial 50 kb distance was reduced to a maximum 4 kb interval. Each breakpoint interval thus characterized by PCR walking is positioned on the BAC clones (Figure 2A).Figure 2Schematic localization of the deletion breakpoints on the PCDH15 gene and their analysis. A: Localization of the six deletion breakpoints on the bacterial artificial chromosome (BAC) clones. B: The breakpoint junction fragment identified in patient U297 is aligned with the wild-type sequences spanning the 5' and 3' breakpoints. The deleted sequences are crossed out.Identification of the junction fragment in patient U297Once proximal and distal amplicons were identified within 4 kb intervals, a junction fragment of 1.3 kb was obtained using the forward primer of the proximal amplicon with the reverse primer of the distal amplicon. Further internal primers (U297-bkp-prox-F: 5'-TGA AGA AAC CAC TAA GAC TGA G-3' and U297-bkp-dist-R: 5'-GTA GCC ATT GCA GGC ACA G-3') enabled the sequencing of a 360 bp junction fragment.Analysis of control DNAGuthrie cards were obtained from the neonatal screening center GREPAM in Montpellier. All samples were anonymously referenced and neither phenotypic nor ethnic origin data were available. DNA was extracted using standard procedures. A total of 172 control DNAs were amplified for the noncoding exon 1 and exon 2. We tested 88 DNAs using primers U297-bkp-prox-F and U297-bkp-dist-R.ResultsHaplotype analyses and evidence of the deletionsWe have previously reported the screening for mutations in MYO7A, CDH23, PCDH15, USH1C, and SANS in a cohort of 31 USH1 families [7]. While conducting a preliminary linkage analysis using microsatellites surrounding each USH1 gene, we detected apparent noninheritance of some markers or failure of amplification among the USH1F panel (D10S1124; D10S2522; IVS3-(CA); D10S2536; D10S546; D10S1643 and D10S1762) in two families (U153, and U382) as shown on Figure 1B. These results were confirmed by amplification in neighboring regions. Microsatellite analysis was not informative in patient U297 but a deletion was suggested after an apparently homozygous p.Arg290X mutation, localized in exon 8, was identified. This novel mutation appeared to be carried on different haplotypes as revealed by heterozygous intronic SNPs in the 3' end of the gene (Figure 1B).The resulting haplotypes of the three pedigrees are presented in Figure 1B together with SNPs analyses when informative. Sequencing of the entire coding region of PCDH15 revealed that the patients were compound heterozygotes for premature truncating mutations p.Ser144LeufsX15 (c.423_430dup) and p.Arg290X (c.868A>T) in trans to the two deletions identified in U153 and U297. A third homozygous deletion was identified in family U382 with a known history of consanguinity.Narrowing of the breakpointsTo narrow down the deletion breakpoints of the two compound heterozygous patients, we used two semiquantitative PCR walking methods. The narrowing of the deletion breakpoint in patient U382 was performed by simple PCR walking, looking at the amplification or nonamplification of each amplicon.When each breakpoint was localized within an interval below 4 kb, PCR was performed to identify the precise deletion breakpoints. One breakpoint was identified (U297; see Figure 2B). Unfortunately, several other attempts using different long-range PCR kits failed to identify a junction fragment in the other two patients, suggesting that the deletions may be more complex than anticipated.PCR walking in patient U382 narrowed the deletion to a proximal breakpoint localized within a 1.6 kb interval in the 5' region of the PCDH15 gene and to a distal breakpoint lying in IVS1 within an 1.4 kb interval. The size of the deletion is estimated as 190 kb (Figure 1 and Figure 2). None of the 172 control DNAs showed an absence of amplification of exon 1 excluding a similar homozygous deletion in these controls.The deletion in patient U153 was originally characterized as spanning exons 3-5 by means of the intragenic marker D10S2536 and SNP. However this deletion was further characterized as extending from a proximal breakpoint within IVS1, within an interval of 3.5 kb, to a distal breakpoint within an 0.8 kb interval of IVS5 (Figure 1 and Figure 2). The size of the deletion is about 341 kb. None of the 172 control DNAs showed an absence of amplification of exon 2, excluding an homozygous deletion spanning at least this exon in these controls.The deletion in patient U297 was narrowed down by semiquantitative PCR then further characterized by amplification of a junction fragment (Figure 2B). The deletion spans about 684 kb, includes exon 8, and is in trans to the p.Arg290X mutation (Figure 1 and Figure 2). The proximal breakpoint, lies within 2.4 kb of U382 breakpoint, but is not identical. This junction fragment was not detected in 176 alleles (i.e., 88 control DNAs).DiscussionNone of the deletions described in this paper have been found in normal controls, and they would all result in nonfunctional protein. Recent data has shown that multiple isoforms exist. No exon is present in all identified so far alternative transcripts [11]. However, no isoform has been observed to containing deletions extending from exon 2 to exon 5 or 8. In addition, an Arg3X mutation in the second exon has been observed in two patients, providing evidence that the presence of the first few exons is necessary [8,9]. Recently, an alternatively spliced isoform (lacking exons 3-15) was found to circumvent the effect of the mutant allele IVS14-2A>G in the homozygous Pcdh15av-5J mice [18]. Such a mechanism is not likely to be involved here as the patients described in this study are affected with typical USH1.The six intervals surrounding the breakpoints were placed on the BAC clones (NCBI Accession numbers given in Figure 1A and Figure 2A) and analyzed through repeat masker (repeatmasker). Although short interspersed elements (SINE), long interspersed elements (LINE), and long terminal repeats (LTR) were found in all cases (Figure 2A), there is no evidence for direct repeats or duplicons as found in some other cases of recurrent deletion or duplication. In patient U297 the only obvious feature at the breakpoint is a repetition of GAA (Figure 2B). This is in line with the findings in studies of deletions of the dystrophin gene in DMD [19] and duplications of PLP1 in Pelizaeus-Merzbacher disease [20].A detailed analysis of the gene structure provides a likely explanation for the high rate of such deletions. The gene spans nearly 1 Mb for a corresponding ORF of 7,021 bp. The intron size of PCDH15 is up to 150 kb, and the first three exons of the gene cover 0.42 Mb. The six breakpoint intervals lie in introns ranging from 22 to 140 kb in size localized in the first third of the gene.Because of the large size of PCDH15 and, in particular, the low proportion which is coding, predominantly in the 5' end of the gene, it is not surprising that large deletions, with differing breakpoints, form a significant proportion of PCDH15 mutations (30% in our cohort) which represents nearly 10% of all USH1 patients [7]. The situation is reminiscent of the high frequency of deletions within the dystrophin gene found in patients with Duchenne Muscular Dystrophy. The dystrophin gene coding region of 11 kb is encoded over 2.4 Mb of genomic DNA. Around 60% of mutations are large deletions [21], and many occur within the two large introns 7 and 44 [22].This observation has several implications. First, although an initial linkage analysis approach was incorporated to target which gene was the best candidate for mutation screening, it may also identify apparent noninheritance of markers. Second, these results show that restricting the molecular analysis of PCDH15 in USH1F to sequencing is not sufficient, and testing for large genomic rearrangements is recommended. A previous study has described only one PCDH15 mutation in patients [6]. It is possible that either a large genomic deletion or a mutation lying in the additional exons [11] accounts for the second pathogenic mutation in these patients. Detection of large genomic rearrangements is becoming easier and more routine with the development of methods such as multiplex ligation-dependent probe amplification (MLPA) and multiplex amplifiable probe hybridization (MAPH) [23] and should be considered, particularly for genes with an extended genomic structure. Third, large genomic rearrangement analysis cannot be routinely achieved by PCR of junction fragments as each deletion appears to be different and is likely to be ineffective across breakpoints involving complex rearrangements. PCR may still be customized by semiquantitative PCR, or other methods, such as MLPA, may be developed.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2533039.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2533039",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2533039\nAUTHORS: Juan Wang, Jinling Liu, Qingjiong Zhang\n\nABSTRACT:\nPurposeBlepharophimosis-ptosis-epicanthus inversus syndrome (BPES) is an autosomal dominant disorder where eyelid malformation associated with (type I) or without (type II) premature ovarian failure (POF). It is ascribed to mutations in the forkhead transcriptional factor2 (FOXL2) gene. The purpose of this study is to identify mutations in FOXL2 of Chinese patients with BPES.MethodsGenomic DNA was prepared from leucocytes of peripheral venous blood. The coding regions and nearby intron sequences of FOXL2 were analyzed by cycle and cloning sequencing.ResultsFour mutations in FOXL2 were identified in six families, including c.241T>C, c.650C>G, c.804dupC, and c.672_701dup. Of the four, the c.241T>C and c.650C>G were novel and would result in missense changes of the encoded proteins, i.e., p.Tyr81His and p.Ser217Cys, respectively. The c.672_701dup (p.Ala224_Ala234dup) was detected in three families, indicating a mutation hotspot. The c.804dupC (p.Gly269ArgfsX265) mutation was found in one family.ConclusionsOur results expand the spectrum of FOXL2 mutations and confirm the mutation hotspot in FOXL2.\n\nBODY:\nIntroductionBlepharophimosis-ptosis-epicanthus inversus syndrome (BPES, OMIM 110100) is a rare autosomal dominant disease with a prevalence of about 1 in 50,000 [1]. Clinically, BPES has been divided into two subsets depending on the association of ocular malformation with (type I) or without (type II) premature ovarian failure (POF) [2]. Genetically, however, both types are caused by mutations in FOXL2, and a genotype-phenotype correlation has been described in some cases [3,4].The human FOXL2 gene (OMIM 605597), located at 3q23, is a member of winged/forkhead transcription factor gene family [5]. This single-exon gene codes a protein with 376 residues, which consists of a DNA-binding forkhead domain (resudes 52-152) and a polyalanine domain (residues 221-234) [3,6,7]. A number of mutations in FOXL2 have been identified [8], including six novel mutations in the Chinese population [9-11].Here, we report four mutations identified in six Chinese families with BPES. Two novel missense mutations were associated with BPES type II.MethodsPatientsThirteen probands with BPES from unrelated families were collected from the Zhongshan Ophthalmic Center. Informed consent conforming to the tenets of the Declaration of Helsinki and following the Guidance of Sample Collection of Human Genetic Diseases (863-Plan) by the Ministry of Public Health of China was obtained from all participated individuals or their guardians prior to the study. The diagnosis of BPES was based on criteria previously established [12] with exclusion of microphthalmia.Mutation AnalysisGenomic DNA was prepared from leucocytes of peripheral venous blood [13]. Amplification of the genomic fragments encompassing FOXL2 coding regions (NCBI human genome build 35.1, NC_000003 for gDNA, NM_023067 for mRNA, and NP_075555 for protein) was carried out by PCR using primers as follows: AF: 5'-CAG CGC CTG GAG CGG AGA G-3', AR: 5'-CTT GCC GGG CTG GAA GTG C-3', BF: 5'-GAC CCG GCC TGC GAA GAC A-3', BR: 5'-GGC CGC GTG CAG ATG GTG T-3', CF: 5'-CGC GGC CGC TGT GGT CAA G-3', CR: 5'-GCT GGC GGC GGC GTC GTC-3'. The sizes of the amplified DNA fragments are 545 bp, 517 bp, and 500 bp, respectively.PCR amplification was carried out initially at 95 °C for 8 min, followed by 5 cycles at 94 °C for 40 s, at 68 °C for 40 s, at 72 °C for 40 s, then 5 cycles at 94 °C for 40 s, at 66 °C for 40 s, at 72 °C for 40 s, and a further 30 cycles at 94 °C for 40 s, at 64 °C for 40 s, at 72 °C for 40 s, and finally an elongation step at 72 °C for 5 min. Due to the high GC-rich nature of FOXL2, an additional 10% dimethylsulfoxide and 10% glycerol were added to the PCR mixture in order to successfully amplify the genomic fragments.Direct sequencing of the PCR products was performed with an ABI BigDye Terminator Cycle Sequencing Kit v3.1 (ABI Applied Biosystem, Foster City, CA), using an ABI 3100 Genetic Analyzer. Sequencing results from patients as well as the FOXL2 consensus sequences from the NCBI Human Genome Database (NC_000003) were imported into the SeqManII program of the Lasergene package (DNAStar Inc., Madison, WI) and then aligned to identify variations. Each mutation was confirmed by bidirectional sequencing. Mutation description followed the nomenclature recommended by the Human Genomic Variation Society (HGVS).Any variation detected in FOXL2 was further evaluated in available family members as well as in 100 normal controls by heteroduplex-SSCP analysis as described previously [14]. Two additional pairs of primers were used for heteroduplex-SSCP analysis. The sequence of these primers were: DF: 5'- CCG TAA GCG GAC TCG TGC-3', DR: 5'- AGT AGT TGC CCT TGC GCT C-3', EF: 5'- CGC ACT TCC AGC CCG GCA A-3', and ER: 5'- TGT GTA CGG CCC GTA CGA-3'.In addition, one variation of insertions with multiple nucleotides that was found in three families was further analyzed by cloning sequencing. PCR products harboring this mutation were subcloned to pMD18-T Simple Vector (TaKaRa BIO, Japan) according to the manufacture's instructions. Clones with the mutant allele as well as the normal allele were selected by using heteroduplex-SSCP analysis. Sequence of the cloned fragment was identified by cycle sequencing as described above. Mutations were confirmed by sequencing three positive clones from each family. One mutation, c.241T>C, was further analyzed by PCR-RFLP analysis since the mutation creates a new FOKI site.ResultsAll patients demonstrated typical features of BPES, including small palpebral fissures, ptosis of the eyelids, and epicanthus inversus (Figure 1). Upon complete sequencing analysis of FOXL2 for 13 probands with BPES, four heterozygous mutations were found in six probands, including c.241T>C, c.650C>G, c.804dupC, and c.672_701dup (Figure 2; Table 1). Of the four, c.241T>C and c.650C>G are novel. All four heterozygous mutations were further detected by heteroduplex-SSCP analysis, and one (c.241T>C) was further detected by FOKI digestion (Figure 3). These mutations were also present in affected patients from corresponding families but neither in unaffected individuals nor in 100 controls.Figure 1Clinical phenotype of a Chinese patient with BPES. A photo of the eyelid area from a 40 years old man (III:1 from family E in Figure 3) demonstrated typical phenotype of BPES. The horizontal diameter of his cornea was 11.5 mm for both eyes. Ultrasound A-scan recorded an axial length of 24.19 mm/OD and 24.43 mm/OS. His refractive measurement was -3.00DCX180°/OD and -4.50DCX180°/OS.Figure 2Sequencing results of the two novel mutations in FOXL2. Sequence chromatograms of Family A (A) and family B (B), and their corresponding normal sequences. The underline below each sequence highlights the codon containing the mutation.Table 1FOXL2 mutations detected in Chinese patients with BPES.FamilyDNA changeMutation typeLocationProtein changeBPESAc.241T>CMissenseForkheadTyr81HisType IIBc.650C>GMissenseImmediately upstream polyalanineSer21CysType IICc.804dupCInsertionDownstream of polyalanineGly269ArgfsX265UnknownDc.672_701dupDuplicationPolyalanine domainAla224_Ala234dupUnknownEc.672_701dupDuplicationPolyalanine domainAla224_Ala234dupType IIFc.672_701dupDuplicationPolyalanine domainAla224_Ala234dupUnknownSubtypes of BPES in families C, D, and F are unknown, as there were no female patients (families C and F) or the female patients were too young (family D, where the two female patients were only 4 or 2 years old, respectively).Figure 3Pedigrees and heteroduplex-SSCP analysis. Pedigrees of the different families (A, B, C, D, E, and F) are shown. Black filled symbols indicated patients affected with BPES in each family. The \"+/+\" or \"+/-\" sign indicated individuals analyzed with normal sequences or heterozygous mutation in FOXL2, respectively. The \"M\" under each lane indicated mutation, the \"N\" represented normal individuals, and \"N+M\" represented a mixture of PCR products resulted from normal and mutant clones. In addition, results of FOKI digestion for family A were present at the bottom of Figure 1A. The c.241T>C mutation in family A creates an additional FOKI site. By FOKI digestion of the 545 bp PCR product, normal allele yielded two fragments (340 bp and 205 bp), but the mutant allele yielded three fragments (275 bp, 205 bp, and 65 bp). Patients with the heterozygous c.241T>C mutation had four bands (only three bands were shown in the figure) as compared to normal individuals with two bands.The c.241T>C (p.Tyr81His) mutation results in substitution of a charge-free tyrosine with a charge-positive basic hydrophilic histidine within the forkhead domain. The c.650C>G (p.Ser217Cys) mutation is located immediately upstream of the polyalanine domain. The tyrosine at position 81 and the serine at position 217 are well conserved in FOXL2 by ClustalW analysis of 11 orthologs from related vertebrate species (Figure 4).Figure 4Multiple alignment of 11 FOXL2 orthologs. This demonstrated high conservation of residues involved by the p.Tyr81His and p.Ser217Cys mutation. The resource for the 11 FOXL2 orthologs was as follows: human (Homo sapiens, NP_075555), mouse (Mus musculus, NP_036150), pig (Sus scrofa, AAQ91845), rabbit (Oryctolagus cuniculus, AAQ91846), rat (Rattus norvegicus, XP_345976), mole vole (Ellobius lutescens, AAV30684), cow (Bos taurus, NP_001026920), goat (Capra hircus, AAM52099), chicken (Gallus gallus, NP_001012630), zebrafish (Danio rerio, XP_698915), and salmon trout (Oncorhynchus mykiss, AAS87040).DiscussionFOXL2 encodes a forkhead transcription factor containing a forkhead domain for DNA-binding and a polyalanine domain of uncertain function. Strong expression of FOXL2 has been found in eyelids [3,15], developing periocular muscles, and surrounding tissues [16,17]. Of the four mutations identified in this study, the c.241T>C affected the forkhead domain, while the other three (c.650C>G, c.804dupC, and c.672_701dup) were located upstream, within, and downstream of the polyalanine domain, respectively.Missense mutations in FOXL2 reported so far usually occurred at the forkhead domain [9,17-19], except two, such as c.650C>T in a Belgian family [4] and c.644A>G in a five-generation family from south-India [20]. The clinical subtypes of the patients with the c.650C>T and c.644A>G mutations were unknown. The novel c.650C>G (p. Ser217Cys) mutation identified in Chinese family B occurred at the same site as that found in the Belgian family, which is located immediately upstream of the polyalanine domain. The serine at position 217 is well conserved in 11 orthologs (Figure 4). It has been shown that mutations affecting the polyalanine domain induce extensive nuclear and cytoplasmic protein aggregation [21,22]. Missense changes have been suggested to act as null allele leading to BPES phenotype due to haploinsufficiency [4] or dominant-negative effect [20,23].It has been suggested that FOXL2 mutations truncating the protein led to BPES type I while those extending the mutant protein were associated with type II [3,4]. However, intra- and inter-family phenotypic variations have been found [3,4,19,24,25] so that this genotype-phenotype correlation might not be general [18,19,26]. The c.804dupC mutation has been shown to cause both types of BPES [4,19,25], and the c.672_701dup causing polyalanine expansion most likely leads to BPES type II [19]. Missense mutations have been associated with both BPES type I [17] and II [3,19]. The patients from families A and B in this study, with novel c.241T>C and c.650C>G mutations, respectively, had type II BPES. The c.650C>G mutation is the first mutation described that occurs immediately upstream of the polyalanine domain and associated with type II BPES. This may raise a possibility that the region containing the c.650C>G mutation is of importance for FOXL2 function.The c.672_701dup (p.Ala224_Ala234dup) was found in families D, E, and F (Table 1), consistent with a mutation hotspot. To check the origin of the c.672_701dup mutation in three families (families D, E, and F in Figure 3), six SNPs (including rs13325788, rs2291252, rs28937885, rs7432551, rs28937884, and rs11924939) were analyzed (Table 2). The SNP at rs2291252 is different between patient II:1 from family D and patient III:1 from family E, which may suggest a different origin of the mutant allele. The mutation in family F is most likely a de novo event as BPES was not present in the patients' parents although the SNPs in the patient II:1 in family F were the same as that of II:1 in family D. It has been reported that 30% of the FOXL2 mutations lead to polyalanine expansion [19]. The c.672_701dup has been found in BPES families of Caucasian [4,19,27,28] and Asian origin [10,29].Table 2Results of SNPs analysis of the three families with the c.672_701dup mutation.FamilyPatientUpstream of FOXL2Inside FOXL2Downstream of FOXL2rs11924939rs28937885rs7432551rs28937884rs13325788rs2291252dII:1CTCTGCeIII:1CTCTGTfII:1CTCTGCThe origin of the c.672_701dup mutation in family D is different from family E as they have different SNP at rs2291252. This mutation in family F is a de novo event, although patient II:1 in family F shares the same haplogroup of the six SNPs with that of family D.In summary, we identified two novel and two known mutations in FOXL2 of six Chinese families with BPES. The two novel mutations are the first reported instances that were associated with BPES type II. Our results expanded the spectrum of FOXL2 mutations and confirmed the mutation hotspot in FOXL2.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2533297.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2533297",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2533297\nAUTHORS: Christopher Keller, Ronit Katz, Mary Cushman, Linda F Fried, Michael Shlipak\n\nABSTRACT:\nBackgroundPrior studies using creatinine-based estimated glomerular filtration rate (eGFR) have found limited associations between kidney function and markers of inflammation. Using eGFR and cystatin C, a novel marker of kidney function, the authors investigated the association of kidney function with multiple biomarkers in a diverse cohort.MethodsThe Multi-Ethnic Study of Atherosclerosis consists of 6,814 participants of white, African-American, Hispanic, and Chinese descent, enrolled from 2000–2002 from six U.S. communities. Measurements at the enrollment visit included serum creatinine, cystatin C, and six inflammatory and procoagulant biomarkers. Creatinine-based eGFR was estimated using the four-variable Modification of Diet in Renal Disease equation, and chronic kidney disease was defined by an eGFR < 60 mL/min/1.73 m2.ResultsAdjusted partial correlations between cystatin C and all biomarkers were statistically significant: C-reactive protein (r = 0.08), interleukin-6 (r = 0.16), tumor necrosis factor-α soluble receptor 1 (TNF-αR1; r = 0.75), intercellular adhesion molecule-1 (r = 0.21), fibrinogen (r = 0.14), and factor VIII (r = 0.11; two-sided p < 0.01 for all). In participants without chronic kidney disease, higher creatinine-based eGFR was associated only with higher TNF-αR1 levels.ConclusionIn a cohort characterized by ethnic diversity, cystatin C was directly associated with multiple procoagulant and inflammatory markers. Creatinine-based eGFR had similar associations with these biomarkers among subjects with chronic kidney disease.\n\nBODY:\nBackgroundHigher levels of markers of inflammation, such as C-reactive protein (CRP) and interleukin 6 (IL-6), have been associated with cardiovascular disease in healthy populations [1-3]. In subjects with end stage renal disease (ESRD), inflammatory biomarkers are significantly elevated and predict poor outcomes [4-7]. In subjects with kidney disease not on hemodialysis, kidney function has been associated with markers of inflammation for creatinine-based estimated glomerular filtration rates (eGFR) below 60 mL/min/1.73 m2. Above that threshold, two studies did not find an association between eGFR and markers of inflammation [8,9].Cystatin C, a cysteine protease inhibitor secreted by all nucleated cells, is a novel serum marker for kidney disease that may better detect small changes in kidney function [10-12]. Since creatinine-based eGFR is not reliable above 60 mL/min/1.73 m2, cystatin C may be superior in detecting an association with inflammation in subjects with mild to moderate kidney disease [13,14]. Using cystatin C as a marker for kidney function in an ambulatory elderly cohort, all with creatinine-based eGFR ≥ 60, we reported linear associations between cystatin C and five inflammatory markers: CRP, IL-6, tumor necrosis factor alpha (TNF-α), and two soluble TNF-α receptors [15]. In a cohort with known coronary artery disease, cystatin C was associated with both CRP and fibrinogen across the entire cohort, while creatinine-based eGFR was significantly associated with CRP and fibrinogen only for eGFR < 60 [16].The current study investigated the association of both creatinine-based eGFR and cystatin C with six inflammatory and procoagulant biomarkers in the Multi-Ethnic Study of Atherosclerosis (MESA), a large cohort characterized by racial and ethnic diversity. Compared with prior studies on kidney function and markers of inflammation, this study featured a population with four racial/ethnic groups, a younger mean age (62 years), no clinical cardiovascular disease, and more extensive measurements of both inflammatory and procoagulant biomarkers. In addition, a second objective of this study was to test whether the association between kidney function and multiple inflammatory and procoagulant biomarkers differed by race/ethnicity.MethodsThe MESA cohort consists of 6,814 men and women who identified themselves as white, African-American, Hispanic, or Asian (of Chinese descent). At the time of enrollment, the subjects were 45 to 84 years of age and free of clinical cardiovascular disease (CVD). Participants were enrolled from July 2000 to August 2002, and recruited from six US communities. Subjects were initially recruited using residential and telephone listings; towards the end of the recruitment period, lists of Medicare beneficiaries and participant referrals were also used to ensure an adequate number of participants. The study was approved by institutional review boards at each center, and all study participants gave informed consent.All biochemistry assays were performed on plasma or serum drawn in the morning after an overnight fast during the initial visit and stored at -70°C. Cystatin C was measured using a BNII nephelometer on plasma specimens (N Latex Cystatin C; Dade Behring Inc., Deerfield, IL) [17]. The assay range is 0.195 to 7.330 mg/L, with the reference range for young, healthy individuals reported as 0.53 – 0.95 mg/L. Intra-assay coefficients of variation (CVs) range from 2.0 – 2.8% and inter-assay CVs range from 2.3 – 3.1%. Serum creatinine was measured using colorimetry with a Johnson & Johnson Vitros 950 analyzer (Johnson & Johnson Clinical Diagnostics Inc., Rochester, NY). The CVs for serum creatinine were ≤ 2%. Estimated GFR was calculated using the creatinine-based four-variable Modification of Diet in Renal Disease (MDRD) equation, which features adjustments for age, gender, serum creatinine, and black race [18]. Creatinine levels were calibrated to the Cleveland Clinic standard (0.9954*Cr + 0.0208) [19]. Lipid measurements were performed using the Roche COBAS FARA centrifugal analyzer (Roche Diagnostics, Indianapolis, IN). Low-density lipoprotein cholesterol (LDL) was calculated using the Friedewald equation [20]. Serum glucose was measured using the Vitros 950 analyzer. Urine samples for measuring creatinine and albumin were taken from a spot urine collection at the initial visit. Urinary creatinine was measured using the Vitros 950. Urinary albumin was measured using an Array 360 CE Protein Analyzer (Beckman Instruments Inc., Fullerton, CA). Chronic kidney disease was defined as eGFR < 60 mL/min/1.73 m2[13].CRP was measured using a BNII nephelometer (N high sensitivity CRP; Dade Behring Inc.), with intra-assay CVs of 2.3% – 4.4%, inter-assay CVs of 2.1% – 5.7%, and a detection level of 0.18 mg/L [21]. IL-6 was measured by ultra-sensitive ELISA (Quantikine HS Human IL-6 Immunoassay; R&D Systems, Minneapolis, MN) with an analytical CV of 6.3% and a detection level of 0.04 pg/mL. Tumor necrosis factor alpha receptor 1 (TNF-αR1) was measured by ultra-sensitive ELISA (Quantikine Human sTNF RI Immunoassay; R&D Systems) with an analytical CV of 5.0% and a detection level of 0.77 pg/mL. Intercellular adhesion molecule 1 (ICAM-1) was measured by ELISA (Parameter Human sICAM-1 Immunoassay; R&D Systems), with a CV of 5.0%. Fibrinogen was measured using a BNII nephelometer (N Antiserum to Human Fibrinogen; Dade Behring Inc.) with intra-assay and inter-assay analytical CVs of 2.7% and 2.6%, respectively. Factor VIII coagulant activity was determined by measuring the clotting time of a sample in factor VIII deficient plasma in the presence of activators utilizing the Sta-R analyzer (STA-Deficient VIII; Diagnostica Stago, Parsippany, NJ), with a reported normal plasma range in the adult population of 60–150%. CRP, IL-6, fibrinogen, and factor VIII were measured in the entire cohort. Intercellular adhesion molecule-1 (ICAM-1) was measured in a total of 2,614 subjects, including all participants who enrolled before February 2003 and a subset of 1,000 participants randomly selected from the first 5,030 participants enrolled. The baseline characteristics of the random subset of participants were not significantly different from those of the entire MESA cohort. TNF-αR1 (995 samples) was measured in a subset of the cohort chosen randomly after 75% of the participants had been enrolled.Participant characteristics, including demographics (age, sex, race/ethnicity); comorbid conditions (diabetes, defined as a fasting glucose ≥ 126 mg/dL (6.9 mmol/L) or by the use of insulin or oral hypoglycemic medications, and hypertension, defined as an average systolic blood pressure ≥ 140 mm Hg, an average diastolic blood pressure of ≥ 90 mm Hg, or by the use of antihypertensive medications), smoking history, defined as ever [current or former] or never; and statin use were obtained at the enrollment visit using data from standardized questionnaires. Resting blood pressure was determined by taking three measurements with the participant in the seated position; systolic and diastolic blood pressures were recorded as the average value of the last two measurements from both the first and second study examinations. Body mass index (BMI) was calculated at the initial visit using weight (kg) divided by height squared (m2).Statistical AnalysisBaseline characteristics were evaluated for statistical significance across quintiles of cystatin C using ANOVA or chi-square tests for trend. Histograms and q-q plots revealed that the urinary albumin to creatinine ratio and the levels of CRP and IL-6 were skewed; therefore, logarithmic transformations of these variables were used for analysis. First, partial correlations with each of the six biomarkers were determined separately for cystatin C and creatinine-based eGFR, adjusting for age, gender, race/ethnicity, and BMI. Two-tailed t-tests were used for significance testing of all partial correlations. We also compared the partial correlation coefficients of eGFR and cystatin C with each biomarker after stratification by presence or absence of CKD. The two correlation coefficients were transformed using the Fisher Z-transform and the difference and a p-value were computed.We created separate linear regression models for each of the six biomarkers. The primary predictors were either cystatin C (per SD) or eGFR by the four-variable MDRD equation, dichotomized as < and ≥ 60 ml/min/1.73 m2. Covariates from Table 1 were entered into the models based on their potential role as confounders due to their associations with both kidney disease and inflammation. We also determined the adjusted mean levels of each biomarker, stratified by the presence of CKD (eGFR < 60 mL/min/1.73 m2) and by quintile of cystatin C. The mean biomarker levels were also plotted across quintiles of cystatin C, with the y-axis scale standardized at ± 1 SD of the overall mean of the cohort for each biomarker. For TNF-αR1, the y-axis scale was set at ± 2 SD given the larger differences for that biomarker across quintiles of cystatin C. T-tests were used to evaluate significant differences in mean levels in subjects with and without CKD, and ANOVA across quintiles of cystatin C. We also tested for interactions of race/ethnicity with both cystatin C in each regression model. S-Plus (release 6.1, Insightful Inc, Seattle, WA) and SPSS statistical software (release 13.0.1, SPSS Inc, Chicago, IL) were used for the analyses. P < 0.01 was used for statistical significance because multiple comparisons were made and because a high level of power was available.Table 1Baseline characteristics of MESA population by quintiles of cystatin CCystatin C (mg/L)Quintile 1 Quintile 2Quintile 3Quintile 4Quintile 5≤ 0.740.75–0.820.83–0.890.91–1.02≥ 1.03(n = 1446)(n = 1377)(n = 1283)(n = 1359)(n = 1285)Mean ± SD, Median [IQR], or %p-value for trendAge57 ± 959 ± 962 ± 965 ± 1069 ± 10<0.001Female, %6553494749<0.001Race/Ethnicity, % White3334414144 Chinese171310118 African-American3030262428 Hispanic1924232420Smoking History, %4449505353<0.001BMI (kg/m2)27 ± 528 ± 528 ± 529 ± 630 ± 6<0.001Diabetes, %1413121221<0.001Hypertension, %3436435064<0.001LDL (mg/dL)117 ± 32119 ± 31118 ± 31117 ± 32114 ± 320.006HDL (mg/dL)56 ± 1651 ± 1551 ± 1449 ± 1448 ± 14<0.001Glucose (mg/dL)105 ± 37104 ± 32102 ± 22104 ± 28107 ± 310.128Medication use, % ACE-inhibitors0.60.90.91.51.90.005 Beta-blockers0.30.40.90.70.50.427 Statin use1114151619<0.001Urinary albumin/creatinine ratio (mg/g)5.1 [3.4, 10.3]5.1 [3.2, 9.5]5.1 [3.1, 9.0]5.3 [3.3, 10.7]7.4 [3.9, 20.0]<0.001Serum cystatin C (mg/L)0.68 ± 0.060.77 ± 0.020.86 ± 0.020.96 ± 0.031.22 ± 0.36<0.001Creatinine-based eGFR (ml/min/1.73 m2)93 ± 1687 ± 1582 ± 1977 ± 1365 ± 16<0.001To convert LDL or HDL from mg/dL to mmol/L, divide by 39; to convert glucose, divide by 18; to convert urinary albumin/creatinine ratio from mg/g to mg/mmol, multiply by 0.114; to convert serum creatinine from mg/dL to μmol/L, multiply by 88.ResultsThe participants of the MESA study had an average age of 62 years. In the cohort, 53% of the participants were female; 39% were white, 28% African-American, 22% Hispanic, and 12% Chinese. The mean cystatin C level in the cohort was 0.89 ± 0.24 mg/L, and the mean creatinine-based eGFR was 81 ± 18 mL/min/1.73 m2. On average, the participants in the highest quintile of cystatin C were older, and more likely white and male (Table 1). Those in the highest quintile were also more likely to smoke, to have hypertension and diabetes, to have a higher BMI, to have a higher urine albumin to creatinine ratio, and to have lower levels of both LDL and HDL. Overall, the total number of subjects with CKD in the cohort was 672 (10%).After adjustment for age, sex, ethnicity, and BMI, cystatin C had statistically significant partial correlations with all six biomarkers in participants both with and without CKD (Table 2, p < 0.01 for all). However, the associations with TNF-αR1 and fibrinogen were significantly stronger among participants with CKD. Similarly, creatinine-based eGFR had significant correlations with all biomarkers except ICAM-1 in subjects with CKD (p < 0.01). In subjects without CKD, eGFR was associated only with TNF-αR1. Of note, the strongest correlations for both cystatin C and creatinine-based eGFR were with TNF-αR1.Table 2Partial correlations of cystatin C and creatinine-based eGFR with inflammatory and procoagulant markers at the baseline visit of the MESA study, stratified by presence of chronic kidney diseaseMarker of Kidney FunctionInflammatory MarkerCRP (N = 6750)IL-6 (N = 6622)TNF-αR1 (N = 995)ICAM-1 (N = 2611)Fibrinogen (N = 6750)Factor VIII (N = 6750)Cystatin C0.075*0.156*0.748*0.209*0.137*0.112*Creatinine-based eGFR-0.019-0.003-0.285*0.025-0.059*-0.034*Persons with CKD Cystatin C0.116*0.230*0.903*0.160*0.248*0.140* Creatinine-based eGFR-0.101*-0.139*-0.763*-0.035-0.286*-0.146*Persons without CKD Cystatin C0.085*0.180*0.557*†0.261*0.111*†0.085* Creatinine-based eGFR-0.004‡0.025‡-0.168*‡0.027-0.021‡0.017‡Partial correlations adjusted for age, gender, race/ethnicity, and BMIChronic kidney disease (CKD) defined as eGFR < 60 mL/min/1.73 m2* Correlation is significant at the 0.01 level (2-tailed).† Correlation between cystatin C and the biomarker is significantly stronger among participants with CKD‡ Correlation between creatinine-based eGFR and the biomarker is significantly stronger among participants with CKDAdjusted mean levels of all biomarkers except ICAM-1 were significantly higher in participants with CKD (eGFR < 60) than in those without CKD (Table 3; p < 0.001 for all). The mean levels of each biomarker in subjects without CKD were unchanged by adjustment for albumin to creatinine ratio. Adjusted mean levels for all six biomarkers increased significantly across the five quintiles of cystatin C (Figures 1a–c). The p-values for trend of the mean biomarker levels across all quintiles were less than 0.01. For comparison, we found that the mean biomarker levels among persons with CKD were similar to the highest quintile of cystatin C in the entire cohort. Finally, adjusted linear regression models using cystatin C to predict biomarker levels were also examined for interactions by race/ethnicity. Statistically significant race/ethnicity interactions were found for the associations of cystatin C with CRP, IL-6, ICAM-1, and factor VIII (Table 4).Figure 1Adjusted mean biomarkers by quintiles of cystatin C at the baseline visit of the MESA study. Models adjusted for age, gender, race/ethnicity, smoking status, body mass index, diabetes, hypertension, low-density lipoprotein, high-density lipoprotein, use of statins, and log transformed urinary albumin to creatinine ratio. Y-axis scales are standardized to ± 2 standard deviations of the overall mean of the cohort for TNF-αR1 and ± 1 standard deviation of the overall mean of the cohort for all other biomarkers.Table 3Adjusted mean biomarker levels at the baseline visit of the MESA study, in participants with and without chronic kidney diseaseBiomarker (units)CKD Mean levels (95% CI)No CKD Mean levels (95% CI)P-valueCRP (log mg/L) N = 67500.81 (0.73, 0.89)0.63 (0.61, 0.66)<0.001IL-6 (log pg/mL) N = 66220.37 (0.33, 0.41)0.19 (0.18, 0.21)<0.001TNF-αR1 (pg/mL) N = 9951.90 (1.76, 2.03)1.23 (1.22, 1.25)<0.001ICAM-1 (ng/mL) N = 2611281 (271, 290)274 (271, 277)0.19Fibrinogen (mg/dL) N = 6750363 (360, 372)345 (343, 346)<0.001Factor VIII (%) N = 6750187 (181, 193)161 (160, 163)<0.001All biomarker models adjusted for age, gender, race/ethnicity, smoking status, BMI, diabetes, hypertension, LDL, HDL, and use of statinsChronic kidney disease (CKD) defined as eGFR < 60 mL/min/1.73 m2Table 4Associations of cystatin C and biomarker levels at the baseline visit of the MESA study using adjusted linear regression, stratified by race/ethnicityWhite β coefficient (95% CI)African-American β coefficient (95% CI)Chinese β coefficient (95% CI)Hispanic β coefficient (95% CI)p-value for interactionCRP (log mg/L)0.61 (0.37, 0.84)0.34 (0.14, 0.53)-0.12 (-0.52, 0.28)0.13 (-0.06, 0.31)<0.001IL-6 (log pg/mL)0.62 (0.49, 0.75)0.36 (0.25, 0.47)0.26 (0.01, 0.51)0.26 (0.15, 0.36)<0.001ICAM-1 (ng/mL)100.4 (78.8, 122.0)66.2 (18.7, 113.7)30.8 (-9.9, 71.6)48.6 (11.0, 86.3)0.015Factor VIII (%)52.0 (38.3, 65.9)46.8 (23.5, 70.0)22.4 (9.0, 35.8)32.5 (21.3, 43.8)0.003All biomarker models adjusted for age, gender, race/ethnicity, smoking status, BMI, diabetes, hypertension, LDL, HDL, and use of statins.Beta coefficients are given per standard deviation of cystatin CDiscussionOur analysis demonstrated that cystatin C was significantly correlated with all six procoagulant and inflammatory biomarkers across a broad range of kidney function, even after adjustment for age, gender, race/ethnicity, and BMI. While creatinine-based eGFR had significant correlations with all biomarkers except ICAM-1 in subjects with CKD, it was only associated with TNF-αR1 in participants without CKD. Adjusted mean levels of all biomarkers increased significantly across each quintile of cystatin C, and all biomarkers except ICAM-1 were elevated in persons with eGFR < 60 compared with eGFR ≥ 60. In general, TNF-αR1 had the strongest correlations with both cystatin C and eGFR in all groups.Both creatinine-based eGFR and cystatin C correlated with most inflammatory markers in subjects with chronic kidney disease. However, in patients without CKD, only cystatin C had significant correlations with all markers of inflammation. There are several possible explanations for the association of cystatin C with procoagulant and inflammatory biomarkers in patients without chronic kidney disease. One possibility is that GFR is linearly associated with inflammation, and using cystatin C, a marker of renal function that is less dependent on muscle mass or age, reveals the true association between kidney function and markers of inflammation when GFR is greater than 60 mL/min/1.73 m2[22]. Prior studies have shown that inflammatory markers are not associated with creatinine-based eGFR above 60 [8,9,16]. However, this absence of association may be due to imprecision of eGFR in the normal range. A second explanation is that cystatin C is associated with inflammation independent of kidney function [23,24]. One study found that cystatin C was associated with CRP independent of creatinine clearance; however, that study did not have the gold standard of measured GFR [25]. Another study, also without a gold standard for GFR, found that the association between cystatin C and CRP disappeared after adjustment for 24-hour urine creatinine clearance [16]. Our observation in this current study, that creatinine-based eGFR and cystatin C have similar associations with inflammatory markers among persons with CKD, makes it seem less likely that cystatin C has a direct association with inflammation that is independent of kidney function. However, our study also lacks a gold standard measurement of GFR, and therefore cannot be conclusive.Both cystatin C and eGFR had substantial correlations with TNF-αR1, while the associations with the other inflammatory and procoagulant markers were more modest. In a murine model, one study demonstrated that 125I-labeled soluble TNF receptors were primarily cleared by the mouse kidney [26]. In contrast, other inflammatory markers, such as CRP and IL-6, are primarily cleared by the liver [27,28]. While the strong association between kidney function and TNF-αR1 levels may be simply attributable to renal clearance of TNF-αR1, TNF-α itself may also play a more complex role in the mediation of kidney damage. A future study should evaluate whether TNF-α and its soluble receptors predict the longitudinal progression of kidney disease.A secondary goal of our analysis was to assess the role of race/ethnicity in the association between cystatin C and markers of inflammation. The association of four biomarkers – CRP, IL-6, ICAM-1, and factor VIII – with cystatin C had statistically significant interactions by race/ethnicity. Specifically, these four biomarkers had higher beta coefficients in whites compared with other races/ethnicities. One important issue is that, according to recent data, the assay used in MESA is only able to detect certain polymorphisms of ICAM-1 [29]. One allele in particular, ICAM-1 RS5491-T, is more common in African-Americans and not detected by the MESA assay. Further research is ongoing to evaluate the importance of polymorphisms on detection of circulating markers of inflammation. In the MESA study, we are not aware of other assay issues that would affect the interpretation of race/ethnicity interactions with biomarkers.One possible explanation for the interactions by race/ethnicity in cystatin C models is that regulation of inflammatory cytokines is more dependent on intact kidney function in whites than in other races/ethnicities. Prior studies have shown that CRP and fibrinogen levels are lower in whites than in African-Americans, and that whites at a given baseline level of CRP seem to have slower rises in serum creatinine over time compared with African-Americans [30-32]. These findings would suggest that whites may have a propensity for greater renal excretion of cytokines, although additional studies using urine measurements of cytokines may be helpful to evaluate these associations more effectively. Another explanation for these findings is that cystatin C is a better marker of GFR in whites versus other races/ethnicities. Overall, the association of cystatin C with GFR in non-white groups has not been well studied.Our analysis has several limitations. First, while cystatin C has been conclusively demonstrated to be a reliable marker of kidney function, it may have associations with inflammation that are dependent of kidney function. Such associations would not be supported by this study, however, since we found that cystatin C and creatinine-based eGFR had equally strong associations with inflammatory markers for subjects when eGFR < 60. Second, as a cross-sectional study, we are unable to determine temporality in the association between cystatin C and multiple biomarkers. For example, inflammation may lead to declining kidney function, or reduced kidney function may lead to elevated inflammatory biomarkers. As stated above, we did not have a gold standard measurement for kidney function, such as iothalamate clearance. We also assumed that eGFR < 60 was the appropriate cutpoint for chronic kidney disease for all subjects, although some data suggests that the established MDRD equation may need to be modified to more accurately characterize CKD race/ethnicity groups other than whites and African-Americans [33].ConclusionWe report significant associations between kidney dysfunction and markers of inflammation and procoagulation in a diverse population. Using cystatin C, associations were present in those with and without chronic kidney disease. Creatinine-based eGFR was similarly associated with these biomarkers primarily among subjects with CKD. These results suggest that markers of inflammation are progressively elevated as kidney function declines, even in subjects without chronic kidney disease.List of abbreviationsANOVA: analysis of variance; BMI: body mass index; CKD: chronic kidney disease; CRP: C-reactive protein; CV: coefficients of variation; CVD: cardiovascular disease; ELISA: enzyme-linked immunosorbent assay; ESRD: end stage renal disease; eGFR: creatinine-based estimated glomerular filtration rate; GFR: glomerular filtration rate; ICAM-1: intercellular adhesion molecule-1; IL-6: interleukin-6; LDL: low-density lipoprotein; MDRD: Modification of Diet in Renal Disease; MESA: Multi-Ethnic Study of Atherosclerosis; SD: standard deviation; TNF-α: tumor necrosis factor alpha; TNF-αR1: tumor necrosis factor alpha receptor 1.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsCK was the primary manuscript author and participated in the analysis design and overall data analysis. RK was the primary data analyst and also participated in analysis design and manuscript revision. MC participated in analysis design and manuscript revision. LFF participated in analysis design and manuscript revision. MS participated in the analysis design, overall data analysis, and manuscript revision. All authors have read and approved the final manuscript.Pre-publication historyThe pre-publication history for this paper can be accessed here:\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2533300.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2533300",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2533300\nAUTHORS: Armen R Kherlopian, Ting Song, Qi Duan, Mathew A Neimark, Ming J Po, John K Gohagan, Andrew F Laine\n\nABSTRACT:\nThis paper presents a review of imaging techniques and of their utility in system biology. During the last decade systems biology has matured into a distinct field and imaging has been increasingly used to enable the interplay of experimental and theoretical biology. In this review, we describe and compare the roles of microscopy, ultrasound, CT (Computed Tomography), MRI (Magnetic Resonance Imaging), PET (Positron Emission Tomography), and molecular probes such as quantum dots and nanoshells in systems biology. As a unified application area among these different imaging techniques, examples in cancer targeting are highlighted.\n\nBODY:\nSystems biologySystems biology [1-8] attempts to model the dynamics and structure of complete biological systems. To accomplish this goal, it enlists concepts and expertise from a wide array of fields such as mathematics, physics, engineering, and computer science in addition to the biological sciences. The \"building blocks\" of systems biology models are knowledge and data produced within experimental biology, and mathematical modeling provides the \"cement\" that links these \"building blocks.\" Systems biology extensively uses computational technology and numerical techniques to simulate complex biological networks. The goal is not only to describe biology on a single component level, but also to understand system processes, mechanisms, and principles. The insight gained from simulation results can then be used to design in vivo and in vitro experiments, and in turn further develop models in an ever more refined description of physical and biological reality.As seen in Figure 1, experimental biology can be aided by data mining, and thus statistical analysis, which can be used to extract hidden patterns from large quantities of data to form hypotheses. Hypothesis-driven models can then describe system dynamics. In this regard, systems biology includes in silico simulations in addition to in vitro and in vivo experiments. With adequate models of biological function it is possible to use control methods, as in incorporating feedback and regulatory loops into models and system understanding. Imaging plays a unique role in that it can both provide insight during experiments and also be used to gather data in a high throughput fashion for later analysis.Figure 1Components of systems biology. Systems biology supplements experimental biology by providing methods to both interpret and validate new findings. Data mining provides a way to gain insight from large data sets, while control methods facilitate the interplay of modeling and experimental biology. Imaging can be used for qualitative assessments during experiments and also provide a large amount of data amenable for data mining.There are two approaches or \"avenues\" describing the interplay between experimental and theoretical biology. The traditional approach has been for experimental results to drive model creation. An alternative approach is to generate models based on first principles and then test model-inspired hypotheses by new experiments. The ideal situation is to traverse both \"avenues,\" and so central to the methodology of systems biology is the notion of an iterative and strategic interplay between experimentation and modeling [1,3,5,6,8-11].Role of imagingSince the time of Galileo, imaging has been the \"eyes of science.\" Modern imaging technologies allow for visualization of multi-dimensional and multi-parameter data. Imaging is increasingly used to measure physical parameters such as concentration, tissue properties, and surface area [12] and to glean temporal insight on biological function. Molecular probes can also be employed to allow for both therapeutic and diagnostic applications [13-16]. As the spatial resolution and acquisition frequency of imaging techniques increase, using imaging to monitor substrate and protein dynamics in real time may be more readily achieved. Data acquired by imaging can provide the basis for mathematical modeling of protein kinetics and biochemical signaling networks [12,17]. Imaging can also be a suitable means to test computational models already developed.Digital image processing techniques such as segmentation and registration contribute to model creation and validation strategy. Segmentation can help outline and identify particular regions in an imaged volume where there is biological activity of interest taking place. Registration can assist in the alignment of imaged volumes and areas acquired at different times. Segmentation and registration used together can generate time series data for validating systems biology models. After segmentation and registration, volume and surface rendering can be employed for data visualization [12]. Implementing systems biology models in conjunction with imaging provides a way to refine understanding of biological systems [18]. Eventually, as imaging tools become more widely used, and as more biological processes are understood, systems biology models can be developed that will have true predictive capabilities. To reach this end biology will be propelled by computational models, and imaging science will guide their formulation and validation.Cancer applicationsMajor efforts are underway to apply systems biology methods to oncology [19,20]. Increasingly sophisticated and accessible genomics, proteomics, and metabolomics high throughput experiments provide a basis for new types of oncology research [21]. The number of published results based on gene expression microarray data alone has increased by a factor of 1700% over the last decade [22]. These advances in experimental systems biology coupled with new analysis techniques and quantitative imaging software tools are helping to generate a more complete picture of many cancer related signaling pathways [21-23].The actual development of cancer is a complex process, requiring the accumulation of multiple independent mutations each governing different pathways of cell growth and the cell cycle [21,24]. Genome-wide experiments have shown many signaling pathways to be interrelated and with many transcription factors serving as co-regulators in other signaling pathways [21,24,25]. This integrated nature of cancer pathways leads to difficulty in targeting specific pathway components. Efforts are underway to create comprehensive models of the cell cycle that can be used to better understand both the dynamics of cancer and to enable the design of targeted therapeutics [21,24,26].Advances in molecular imaging can help to satisfy the post-genomic era need for the study of complete biological pathways, and this can potentially accelerate the achievement of a systems level understanding of biological complexity [27,28]. Molecular imaging enables the determination of both the temporal and the spatial distributions of biological processes throughout an intact living subject. With this approach, it is possible to obtain more meaningful results than can be achieved by comparable in vitro methods [29].With the advance of molecular imaging techniques, properly tagged molecules can be visualized leading to insights on cell function, membrane binding sites, and the effectiveness of particular therapies [30-33]. For example, integrating imaging and modeling has led to successful monitoring of immune system functionality via T cell activity [10] and the development of bacteriophages for cancer targeting [34]. It is this type of integration of imaging and modeling that can enable new advances in oncology and other fields in the biomedical sciences.Imaging on multiple scalesThe next generation of imaging tools will include innovative microscopy methods, ultrasound, CT (Computed Tomography), MRI (Magnetic Resonance Imaging), and PET (Positron Emission Tomography). In the coming years, improvements in temporal sampling and spatial resolution will certainly continue. With the advent of molecular probes, imaging can be conducted not only to visualize gross anatomical structures, but also to visualize substructures of cells and monitor molecule dynamics. Thus, the imaging modalities of microCT, microMRI, fMRI, MRS, microPET will also play important roles. A comparison of these imaging technologies is summarized in Table 1. As for the organization of this review paper, each imaging technique is profiled with its respective underlying principle, a description of selected current applications, and a discussion of advantages and known limitations. As a common application area, topics in cancer targeting are highlighted.Table 1Comparison of imaging technology for systems biologyImaging TechniqueResolution ReferencesSpatial ResolutionScan TimeContrast Agents and Molecular ProbesKey UseMulti-photon Microscopy[29,38]15 – 1000 nmSecsFluorescent proteins, dyes, rhodamine amide, quantum dotsVisualization of cell structuresAtomic Force Microscopy[104]10 – 20 nmMinsIntermolecular forcesMapping cell surfaceElectron Microscopy[41]~5 nmSecsCyrofixationDiscerning protein structureUltrasound[29]50 μmSecsMicrobubbles, nanoparticlesVascular imagingCT/MicroCT[29,70]12 – 50 μmMinsIodineLung and bone tumor imagingMRI/MicroMRI[29,76]4 – 100 μmMins – HrsGadolinium, dysprosium, iron oxide particlesAnatomical imagingfMRI[105]~1 mmSecs – MinsOxygenated hemoglobin (HbO2) deoxygenated hemoglobin (Hb)Functional imaging of brain activityMRS[106,107]~2 mmSecsN-acetylaspartate (NAA), creatine, choline, citrateDetection of metabolitesPET/MicroPET[29,108]1 – 2 mmMinsFluorodeoxyglucose (FDG), 18F, 11C, 15OMetabolic imagingThe various micro versions of the imaging modalities (MicroCT, MicroMRI, MicroPET) as well as the microscopy techniques (Fluorescence, Multi-photon, Atomic, Electron) are primarily used in either cellular or animal studies. The remaining modalities (Ultrasound, CT, MRI, MRS, PET) are more widely used clinically.MicroscopyBasic principlesThe advent of fluorescence microscopy has been a major step forward in the study of living cells. Leveraging the characteristic emissions of excited biological fluorophores, such as fluorescent proteins, it is possible to gain insight on cell structure and function (Figure 2). Following traditional fluorescence microscopy has been the development of multi-photon methods, where fluorophores are excited by two or more photons [35]. Multi-photon absorption is achieved with a single pulsed laser focused to a diffraction-limited spot on the specimen. With higher peak power, there is an increase in probability for multi-photon absorption leading to fluorophore excitation. Two-photon fluorescence is depicted in Figure 3a. To meet the excitation energy in this case, two 800 nm photons are used. One 400 nm photon is of equivalent energy, as can be used in single photon excitation, but with multi-photon methods only the area of the laser focus on the specimen is excited. Due to more focused excitation, there is a lower overall phototoxic effect. Also, as scattering of longer wavelength photons is less, multi-photon methods have deeper penetration when compared to single photon excitation.Figure 2Fluorescent protein applications. (a) Three Madin-Darby canine kidney epithelial cells with GFP-rac1 and dsRed-E-cadherin. Rac1 is a pleiotropic signaling molecule that is closely associated with cell-cell adhesion and cell motility. E-cadherin is a cell-cell adhesion protein responsible for facilitating communication between two contacting cells. Scale bar: 10 μm. Contributed by Lance Kam (Columbia University, New York). (b) Membranes of human umbilical cord endothelial cells visualized using EYFP. Scale bar: 40 μm. (c) GFP-actin labeled human umbilical cord endothelial cell undergoing mitosis, with actin filaments aligned toward the centrioles. Scale bar: 30 μm. Contributed by Samuel Sia (Columbia University, New York).Figure 3Two-photon microscopy of in vivo brain function. (a) Basic mechanism of two-photon fluorescence. (b) Schematic of surgical preparation of exposed cortex, with sealed glass window and microscope objective positioning. Green dot shows location of two-photon fluorescence. (c) Examples of two-photon maps of the vasculature following intravenous injection of dextran-conjugated fluorescein. Black dots and stripes show red blood cell motion. (d) Dual-channel imaging of neuronal (green) and vascular (red) signals: (left) Oregon Green 488 BAPTA-1 AM calcium sensitive dye stained neurons and (right) transgenic mouse expressing green fluorescent protein (GFP) in a subpopulation of neurons (mouse supplied by Jeffrey M. Friedman, Rockefeller University, New York) [101]. Texas dextran red is the intravascular tracer in both cases. (e) Three channel imaging of Tg2576 APP Alzheimer's disease mouse model with amyloid-targeting dye (blue), GFP expressing neurons and dendrites (green) and vasculature (red). Adapted from [52] and contributed by Elizabeth Hillman (Columbia University, New York).In STED (Stimulated Emission Depletion) microscopy [36,37], two pulsed lasers are used in tandem to break the diffraction barrier. The first laser pulse has a wavelength that excites fluorophores, and is immediately followed by a second laser pulse that depletes fluorescence. The fluorescence depletion is achieved as the wavelength of the second laser is tuned to be longer than the fluorescence emission. Absorption of a photon from the second laser induces electrons to drop to a lower energy level (stimulated emission) preventing typical fluorescence. The difference in area of the two focused beams leaves only a very small area from where fluorescence is detected. This area is smaller than a diffraction-limited spot. Using STED, images have been captured with a resolution of ~30 nm [35]. In another study using rhomadine amide and a photoswitching technique, a resolution of 15 nm was achieved [38]. In [39,40], an optical trapping system was used to make angstrom resolution measurements of pair based stepping of RNA polymerase, and thus has established an important resolution benchmark in molecular biology.Electron microscopy has offered a resolution of ~5 nm for imaging biological tissue [41]. However, to prepare a sample to be imaged by an electron microscope is a rigorous process that does not allow for imaging of live samples [42]. One common sample preparation technique is cryofixation, which is a high pressure and deep freezing technique that results in contrast in electron microscopy [43]. Even though the samples are no longer viable, electron microscopy has provided invaluable insights on the structural details of organelles and membranes [44].Atomic force microscopes do not acquire information optically, but rather by recording intermolecular forces between a probe tip and a surface. The primary information acquisition component of an atomic force microscope is a cantilever with a nanometer-scale silicon tip. The tip is brought in close proximity with the sample and the deflection of the cantilever due to Van der Waals forces is recorded to generate a contour map of the sample surface [45,46]. In comparison with an electron microscope, the sample does not need any special treatment that would actually destroy the sample and prevent its reuse. However, using contact or tapping mode of an atomic force microscope, which impinges on the sample surface to acquire measurements such as strain, can mechanically damage cells and tissue. Other probe microscopy techniques include scanning tunneling microscopy and near-field scanning optical microscopy [47].Current applicationsFluorescence microscopy is often used in systems biology and there is a strong push for the development of high throughput methods. In the application of genome-wide RNAi screens to document the phenotype for each suppressed gene [48,49], there can be millions of images from a single screen which can amount to several terabytes of data [50]. It is systems biology modeling that relieves the bottleneck of processing this large amount of RNAi screen image data by providing an efficient means of classification. With high throughput microscopy there is much more data generated than can be annotated or evaluated manually, and so developing a fine-tuned and efficient classification model is paramount for unlocking the potential of high throughput methods.As seen in Figure 2, fluorescent proteins can be used to visualize many functional and structural aspects of cells. Using multi-photon methods as well can provide insight on cell structural and biochemical changes [51]. Multi-photon methods have a wide array of applications including in vivo brain imaging in animals (Figure 3) [52-54], where cortical micro-architecture has been investigated with single cell resolution [55]. Electron microscopes have been used to elucidate macromolecule structure [41]. As for cancer applications, atomic force microscopes have been used to monitor the super-coiled state of DNA, which is preferential to the binding of the tumor suppressing protein p53 [56]. Also in regard to detecting levels of p53 in cells, fluorescence microscopy has been used to determine the effectiveness of oncolytic adenoviruses. Specially designed oncolytic adenoviruses target cancerous tissues and are programmed to replicate if the cellular p53 level is low. Viral oncolytic therapy is an intense research area for cancer treatment and as microscopy techniques advance so will the ability to assess the effectiveness of viral vectors for tumor ablation [57,58].Advantages and limitationsIt has been long held that the wave nature of light imposes a seemingly fundamental limit on the resolving power of a microscope. The limitation was approximately half the wavelength of visible light or 200 nm. Recently, there has been over a 10-fold resolution improvement with advances in microscopy [35,38]. However, optical techniques have limited penetration as light readily scatters in tissue. This can be partially ameliorated by using more powerful lasers, but this in turn can lead to increased photobleaching effects which can limit the amount of time that an experiment can run.Microscopy, as with other modern imaging techniques, has become ever more dependent on software for image acquisition and analysis. Imaging technology can be enhanced or limited by the software it is coupled with. Table 2 contains an overview of current microscopy image analysis software. With advances in acquisition algorithms and optics holographic microscopy has been achieved, by which full three-dimensional information can be acquired in a single image [59,60]. As a result, volumetric time series data can be collected without the need of changing focus and scanning multiple z-planes.Table 2Overview of microscopy image analysis softwareVendorPackage NameImage SupportSupported DevicesWebsiteMVIAImage Analysis Software2D/3DAClemexClemex Vision PE2D/3DAMISPax-It PI-M300A2DEMedia CyberneticsImage-Pro Bundled SolutionsImage-Pro AMSImage-Pro MDAImage-Pro MCImage-Pro 3D Suite2D/3DA, DiMTtechnologyiSolution DT2D/3Dn/aDewinter OpticalDewinter Caliper ProDewinter BiowizardDewinter Material PlusDewinter Foundry PlusDewinter Micro Measurement Pro2DEMBF BioScience MicroBrightFieldAutoNeuronConfocal SDImageStackModuleNeuroLucidaSolidModelingModuleSteroInvestigatorVirtualSliceModule2D/3DC, D, ENascent TechnologyMedicalPlusMeasureProCapturePro2DAIntelligent PerceptionPixcavator Image Analyzer2Dn/aGSA Bansemer & Scheel GbRGSA Image Analyser2Dn/aBroad InstituteCellProfiler*CellVisualizer*2Dn/aIMASCellObserverEliSpotProcessAnalysis2DA, D, EWadsworth CenterSpider*2D/3DEMCIDMCID Core2D/3DA, EImageJImageJ for Microscopy*2D/3DC, DScion CorporationScion Imaging Software*n/an/a(*) Open source/freeware software packagesA = Automated microscopeB = Planar microscopeC = Confocal microscopeD = Functional microscopeE = Digital microscopeGaining insight from microscopy images typically requires some level of processing and analysis. Both commercial and open source software packages are available that can supplement or totally drive microscope usage.UltrasoundBasic principlesUltrasound imaging entails moving a hand held probe over the patient and using a water-based gel to ensure good acoustic coupling. The probe contains one or more acoustic transducers and sends pulses of sound into the patient. Whenever a sound wave encounters a material with different acoustical impedance, part of the sound wave is reflected which the probe detects as an echo. The time it takes for the echo to travel back to the probe is measured and used to calculate the depth of the tissue interface causing the echo. The greater the difference between acoustic impedances, the larger the echo is. A computer is then used to interpret these echo waveforms to construct an image [61].Current applicationsUltrasound has had a tremendous impact in cardiology. As seen in Figure 4, the use of ultrasound can enable the coupling of anatomical and strain information of the heart [62]. Going from the organ level to the molecular level has been made possible by advances in microbubble manufacturing. Microbubbles themselves are several micrometers in diameter and are intravascular tracers [63]. Ligands can be attached to microbubbles to make them target specific [64]. Many clinical applications of contrast enhanced ultrasound, such as monitoring angiogenesis and inflammatory response, rely on ultrasound detection of microbubbles that contain gas. Since microbubbles are confined to the vascular space, they are useful for targeting antigens expressed on endothelial and blood cell surfaces [63]. Smaller nanoparticle based contrast agents are also available that are capable of extravascular migration in regions of vascular injury or regions where vascular permeability is abnormally high. Ultrasound can also cause a mechanical interaction with microbubbles, leading to their destruction and the subsequent release of therapeutic compounds [65].Figure 4Transthoracic echocardiography and elastography of a healthy human left ventricle. (a), (b), (c), and (d) are the lateral, axial, radial, and circumferential systolic strains from myocardial elastography between end diastole and end systole, respectively. Strains are displayed on a scale of ± 50%. All the images were acquired approximately at the papillary muscle level and shown at end systole. Contributed by Elisa Konofagou (Columbia University, New York).In tumor and angiogenesis models, surface expression of αvβ3 has been demonstrated to be a strong ligand for targeting endothelial cells in angiogenic vessels [63]. In order to image this surface expression, microbubbles were conjugated with peptides that bind to αvβ3. These microbubbles have been shown to have a binding preference to the endothelial surface of Fibroblast Growth Factor (FGF) stimulated neovessels. The extent of neovascularization in a matrigel model matched the image enhancement in ultrasound images to a large extent. Thus, ultrasound imaging served to help validate this experimental model for angiogenesis [63].Advantages and limitationsThe signal-to-noise ratio for ultrasound images is much lower with nanoparticles than those using microbubbles. Although microbubbles are restricted to the vascular space, this can be an advantage since it minimizes potential signal interference from nonvascular cells [63]. As with other molecular imaging techniques, there is an inverse relationship between sensitivity and resolution for contrast enhanced ultrasound. The relative rate of unbound tracer clearance is also an important issue that determines temporal resolution. In this regard, with clearance time within minutes, microbubble tracers are ideal [63].CT/MicroCTBasic principlesIntrinsic differences in X-ray absorption among water, bone, fat, and air provide contrast in Computed Tomography (CT). In CT, a low energy X-ray source and a detector rotate around the subject, acquiring volumetric data. The detectors are typically Charged Coupled Devices (CCD) and act to phototransduce incoming X-rays [66]. For animal studies, microCT machines can be used which typically operate with higher energy X-rays when compared to human scanners. The increase in energy improves resolution, but exposes the specimen to more ionizing radiation which has adverse health effects.Current applicationsCT has relatively low soft tissue contrast for tumors and surrounding tissue, but with iodinated contrast agents organs and tumors can be detected [29]. As a result, incorporating iodine into new probes for CT imaging may be necessary. Furthermore, to detect a tumor or other target there must be sufficient site-specific accumulation of probes to result in attenuation of X-rays. With differential attenuation of X-rays, the target can be more readily delineated [7].CT can be used to image lung tumors and bone metastasis, given its fast imaging time and high spatial resolution. High throughput techniques using microCT have been used for phenotyping large numbers of transgenic mice and detecting macroscopic abnormalities [64]. In [67], the co-registration of microCT images containing tumor structural details with bioluminescence images allowed for the study of cell trafficking, tumor growth, and response to therapy in vivo. This image analysis method could potentially be used for assessing hematological reconstitution following bone marrow transplantation.As seen in Figure 5, microCT imaging and volumetric decomposition were used to provide insight on trabecular bone microarchitecture [68]. The bone samples were decomposed into individual plates and rods, and this imaging and processing scheme has been successfully applied to anatomic sites such as the proximal femur, proximal tibia, and spine. Several key morphological features of trabecular bone architecture were studied: plate and rod size, thickness, number density, and orientation. With this level of detail, it was determined that trabecular plates play an essential role in determining the elastic properties of trabecular bone [68]. Assessing such properties can be important for gauging bone health in conditions such as osteoporosis, and for designing viable replacement tissue in tissue engineering applications [69].Figure 5Complete volumetric decomposition procedure on a vertebral trabecular bone sample. (a) Example microCT bone volumetric data. (b) Closer view of plate and rod microstructures. (c) MicroCT image of a trabecular bone sample. (d) Completely decomposed trabecular bone structures with individual trabeculae labeled by color for each voxel. Image volume: 5 mm3. Contributed by X. Edward Guo (Columbia University, New York).Advantages and limitationsA key advantage of CT is its high spatial resolution, 12 – 50 μm [29,70], which is needed to visualize fine anatomical details. CT can also be combined with functional imaging technologies that provide dynamic and metabolic information. The radiation dose of CT, however, is not negligible and this limits repeated imaging in human studies due to health risks [64].MRI/MicroMRI, fMRI, and MRSBasic principlesMagnetic Resonance Imaging (MRI) is achieved by placing a subject in a strong magnetic field, typically 1.5 or 3 Tesla for human scanners, which aligns the hydrogen nuclei spins in a direction parallel to the field. A Radio Frequency (RF) pulse is applied to the sample which causes the spins to acquire enough energy to tilt and precess, where an RF receiver can record the resulting signal [71]. After the removal of the RF pulse, the spins realign parallel to the main magnetic field with a time constant of T1 which is tissue dependent. Signal strength decreases in time with a loss of phase coherence of the spins. This decrease occurs at a time constant T2 which is always less than T1. Magnetic gradients are used to localize spins in space, enabling an image to be formed. The difference in spin density among different tissues in a heterogeneous specimen enables the excellent tissue contrast of MRI [71]. MicroMRI follows the same principles, but a much higher magnetic field strength is used for animal studies. Increasing magnetic field strength improves resolution, but can disturb the visual system and lead to peripheral nerve stimulation.Functional Magnetic Resonance Imaging (fMRI) is a modality used to image brain activity in response to specified stimuli. When a stimulus solicits a response from a certain area of the brain, metabolism in that region increases. Metabolic demand leads to an increase in blood flow and more oxygenated hemoglobin in the region. As the supply of oxygenated hemoglobin exceeds the metabolic demand, the concentration of oxygenated hemoglobin increases. The balance between oxygenated and deoxygenated hemoglobin is altered leading to a change in image contrast. To detect a change, the image is compared with baseline measurements. Typical cortical activation leads to a 1 – 5% increase in image intensity [72].Magnetic Resonance Spectroscopy (MRS) is an emerging imaging and biochemical analysis technique in biomedical science. It combines the analytical ability of Nuclear Magnetic Resonance (NMR) to identify biochemical species with the capabilities of MRI to isolate individual voxels which are three-dimensional pixels. MRS employs chemical shift imaging to localize spectra for individual voxels [73]. This is achieved by phase modulated RF pulses which eliminate signal contamination into neighboring voxels. When MRS is combined with MRI, concurrent anatomical and biochemical information is obtained (Figure 6).Figure 6In vivo point resolved (single voxel) MRI spectroscropy. (a) Axial and (b) sagital views of human brain and outlined voxel for MRS. (c) 1H spectrum with readily visible N-acetylaspartate (NAA) peak. An aberrant NAA peak can be an indicator of brain injury or disease.Current applicationsThe range of microMRI applications spans from purely experimental to preclinical. MicroMRI technology has been used to track stem cells, monitor immune cell proliferation, and describe embryological development [74]. It has also been used to obtain three-dimensional high resolution representations of bone structure [75]. MicroMRI has advanced to the point at which individual cells, and their organelles, can be imaged with spatial resolution of <4 microns. Images of a paramecium and a spirogyra alga were acquired utilizing a magnetic field of 9 Tesla, phase encoding in all three axes (which improves signal to noise), and Carr-Purcell echo refocusing (incorporation of multiple 90 degree spin echo pulses into the sequence to minimize signal loss due to sample inhomogeneity) [76].Contrast agents have been developed with greater affinity for cellular and molecular targets. These include iron oxide particles (which have been used to label individual T cells), manganese ions (which act as a paramagnetic surrogate of calcium), and caged compounds. The latter involves chelated gadolinium surrounded by an enzyme substrate, which physically obstructs water molecules from approaching the gadolinium. When an enzyme cleaves the substrate, water is able to approach the gadolinium. This in turn reduces T1 and increases contrast. The caged-compound technique has been used to demonstrate regionalized in vivo gene expression in frog embryos whereas manganese ions have been used to trace neuronal pathways [74].fMRI is used to study the functions of the living brain in a non-invasive manner. It has been shown with fMRI that different cognitive functions, such as attention, perception, imagery, language, and memory, elicit specific cognitive activation patterns in different regions of the brain. One common clinical use of fMRI is in the treatment of patients with brain tumors, and a primary treatment goal is to preserve functional brain tissue. fMRI is used to determine the functionality of brain tissue surrounding the tumor so that potentially harmful therapy can be directed away from critical areas [77].Due to the ability of MRS to identify the presence of molecules within voxels, many studies have been devoted to using it to help diagnose cancer and characterize neoplastic tissue. Currently, MRS has been successfully employed in regard to brain, breast, and prostate cancer through identification of various biochemical markers of neoplasm in the imaged volume [78,79]. 1H has been the element of choice because of its large abundance, but studies involving 31P and 13C appear promising. The latter has been used as an effective dynamic marker of metabolic processes through a hyperpolarization technique [80].Advantages and limitationsThe two chief advantages of MRI are its excellent tissue contrast and lack of ionizing radiation [74]. Improved signal-to-noise ratio and resolution can be obtained via a small receiver coil radius and high magnetic field strength. However, high magnetic field strength is problematic in human applications because of arising physiological effects such as nausea and visual abnormalities. Also, higher field strength leads to other technical challenges including an increase in the operating frequency, which potentially generates artifacts.The main advantage of fMRI is its ability to non-invasively image brain. Since image contrast is achieved through the levels of oxygenated and deoxygenated hemoglobin, no external contrast agent is needed. However, due to the faster temporal resolution needed to acquire images of dynamic brain activity, spatial resolution is reduced.Due to the ability of MRS to reveal the presence of particular biomedical molecules and compounds within an in vivo sample, it seems ideally poised for use in systems biology research. However, certain challenges must be overcome such as large voxel size, long sampling times, and questionable quantitative accuracy of assessing molecular concentrations [81].PET/MicroPETBasic principlesIn Positron Emission Tomography (PET), radioactive tracers are incorporated into metabolically active molecules and then injected intravenously. There is a waiting period while the metabolically active molecules are concentrated in the target tissue. The molecule most commonly used in PET is fluorodeoxyglucose (FDG), which has radioactive fluorine and is readily taken up by tumors. The radioactive tracer decays and produces two 511 keV gamma-rays, which result from the annihilation of a positron and an electron. The two resultant gamma-rays are emitted nearly 180 degrees apart and observed by detector rings. Figure 7 contains several sample PET images. The sensitivity of PET at detecting molecular species is relatively high, in the range of 10-11 – 10-12 M. For animal studies, microPET has a volumetric resolution of 8 mm3, while next generation scanners have over an 8-fold increase in resolution and a field of view that encompasses the whole body of a mouse [29,82].Figure 7Phantom and anatomical PET images. (a) 11C PET image of a rod phantom. (b) FDG PET image of a brain. (c) Coronal view of thoracic area from a whole body PET scan.Current applicationsThere are many radioactive tracers for PET that are used in different preclinical and clinical applications [7]. The tracers that target specific tumors are essential for systems biology studies due to the information provided regarding metabolic activity [83,84]. Examples of small targeting ligands include 11C-labelled N-methylspiperone and 18F-labelled spiperone for targeting dopamine receptors on pituitary adenomas [83].PET is useful in systems biology studies related to bone metabolism [85] and metastasis. Bone metastasis is common for several cancers, including prostate, breast, and lung [86]. 15O-labelled water can be extracted from the blood and used to assess tumor blood perfusion. Tumors are in constant need of nutrients from the blood, and tumor neovascularization provides a crucial lifeline for rapidly dividing tumor cells. The uptake of tracer into tissues is proportional to delivery, and so is a measure of perfusion [87].PET can be used for measuring therapeutic effects on disease processes. Specific metabolic enzymes that are selectively expressed in prostate cancer cells constitute such a target. In [11], genes that were differentially expressed between early stage and late stage prostate cancer were studied. L-lactate dehydrogenase-A catalyzes the formation of pyruvate from S-lactase and was expressed at a high level in the late stage cancer cells. PET tracers based on this process would serve to validate this finding and may allow for the identification of prostate cancer metastasis [11].Advantages and limitationsPET is a highly sensitive, minimally-invasive technology that is ideally suited for pre-clinical and clinical imaging of cancer biology. By using radioactive tracers, three-dimensional images can be reconstructed to show the concentration and locations of metabolic molecules of interest [2]. Since the study of cancer cells in their normal environment within intact living subjects is essential, PET is ideally suited for monitoring molecular events early in the course of a disease, as well as during pharmacological or radiation therapy. Furthermore, it can be used to acquire prognostic information and to image for disease recurrence [2,82].PET spatial resolution is comparatively poor, and is limited by pixel sampling rate, the source size, and blurring in the phosphor screens of the detector rings. Another limitation of PET is that radioisotopes with very short half lives must be immediately injected after production. Due to the same decay type of the different radioactive tracers, it is only possible to trace one molecular species in a given imaging experiment or clinical scan [64].Molecular probesAchieving contrast is essential to imaging technology and is often made possible by contrast agents or molecular probes. As mentioned above, fluorescent proteins have played a key role in microscopy studies providing insight on cell structure. Microbubbles have greatly enhanced the use of ultrasound both in imaging and therapeutic applications. For CT, iodine has been instrumental in differentiating tissue types. In MRI based technologies, manipulation of hydrogen spins has allowed for excellent soft tissue contrast and functional imaging of the brain. FDG and other radioactively labeled tracers have enabled targeting of cancer and imaging of metabolic activity with PET. Below, two promising molecular probes are profiled, quantum dots and nanoshells, which may yield a new array of imaging applications.Quantum dotsBasic principlesQuantum dots (QD) are a class of polymer-encapsulated and bioconjugated probes that can fluoresce at multiple wavelengths spanning the visible spectrum. Larger quantum dots emit red light while smaller ones emit blue light. Quantum dots themselves are comprised of a semiconductor core, encased in another semiconductor material that has a larger spectral band gap. This construction enables fluorescence upon excitation. Quantum dots can be packaged in amphiphilic polymers and conjugated with targeting ligands for imaging applications [88]. Under harsh conditions such as wide pH range (1–14), varied salt conditions (0.01 to 1 M), and a strong corrosive environment (1.0 M hydrochloric acid), quantum dots demonstrate extraordinary resiliency and sustained functionality [14].Current applicationsFigure 8 shows the proliferation of human Mesenchymal Stem Cells (hMSC) that are labeled with quantum dots. After 22 days the quantum dots remained incorporated in the hMSCs. This study suggests that bioconjugated quantum dots are a viable probe for long-term labeling of stem cells [89].Figure 8Quantum dot labeled human mesenchymal stem cells undergoing proliferation. hMSCs were incubated for 16 hrs in a 30 nM solution of bioconjugated QDs (a-a2). Following the removal of extracellular QDs, QD-labeled hMSCs and unlabeled hMSCs of the same subpopulation were continuously cultured for 4, 7 and 22 days (b-b2, c-c2, d-d2, respectively). Scale bar: 30 μm. QDs were internalized in the cytoplasm, even after 22 days of culture-expansion (e-e2), apparently endocytosed as aggregates. Scale bar: 5 μm. Reproduced from [89] and contributed by Jeremy Mao (Columbia University, New York).Ligands on quantum dots can be tailored to target specific cancer lines. Quantum dots fashioned to target prostate cancer, QD-PSMA (Prostate Specific Membrane Antigen), showed active emission in the presence of C4-2 prostate cancer cells while other quantum dots did not [14]. Quantum dots can also be used to passively target tumors since leaky tumor vasculatures retain more quantum dots than surrounding healthy tissue. Thus, by both active binding and passive diffusion, more quantum dots will be present near cancerous tissue [14]. With the targeting capabilities of quantum dots there is potential for use as a delivery vehicle for therapeutic compounds. Delivery schemes can be based on the release of a therapeutic compound triggered by ligand binding [13,14,16,90]. As an example of a drug delivery application, in [91] quantum dots with cadmium sulfide were used as chemically removable caps inside mesoporous silica nanospheres to prevent the premature release of drug molecules. Targeted release of drug molecules was mediated by disulfide bond-reducing agents. Quantum dots could also be used in photodynamic therapy by which there is an energy transfer from the quantum dots to target cells, leading to the generation of reactive oxygen species, and thus potentially inducing apoptosis [92,93]. One limitation of such a therapy in vivo would be reliable and localized energy transfer to ensure the destruction of specific cells.As applied to bacteriophage development, quantum dots can be multi-purpose by validating the design model as well as showing the effectiveness of tumor targeting [34]. One design model for bacteriophages with quantum dots is based upon the characteristics of quantum dots themselves. These are namely durability due to the co-polymer shell and flexibility due to the possibility of several different ligands. Experiments have been conducted with quantum dot embedded bacteriophages in both in vitro and in vivo with the goal of destroying cancerous tissues. Iteratively designing and creating bacteriophages is an example where quantum dots provide both the effective targeting means, but also the validation of the design model due to visualization of ligand binding [34].Advantages and limitationsInformation acquired by using quantum dots are constrained by the physical limits of fluorescence microscopy, since that is the imaging technique typically used when detecting emissions from quantum dots. There have been some studies using quantum dots in electron microscopy, which has an order of magnitude higher resolution than light microscopy [90]. The quantum dots themselves experience \"blinking,\" as in each quantum dot randomly switches from on to off. Fortunately, the fluorescence of a bound quantum dot is stronger than of an unbound quantum dot. Still, the randomness of \"blinking\" imposes some limitations on applications requiring single molecule detection as well as on applications requiring quantification of total fluorescence [13]. Using two different color quantum dots, single molecule imaging has been achieved by co-localization on target molecules [94].NanoshellsBasic principlesNanoshells are a class of metal nanostructures consisting of a dielectric silica core surrounded by a very thin metallic shell. By varying the core to shell ratio and the overall size of the nanoshells, strong scattering properties can be achieved that result in resonance wavelengths generating heat [15]. See Figure 9 for cross sectional views of a nanoshell. Fabricating nanoshells with specific antibodies provide a means for scattering based molecular imaging [15,95], which provides molecule specific contrast on the nanometer scale.Figure 9Near field images of an Ag nanoshell. Nanoshell exposed to (a) 721 nm, (b) 492 nm, and (c) 336 nm laser beam and consequential dipole, quadrupole, and dark plasmons, respectively. Surface plasmon oscillations are collective electron motion resultant of optical illumination, and subsequent modes are shown. Adapted from [102].Current applicationsAs seen in Figure 10, nanoshells can facilitate tumor ablation. In another cancer related study [15], nanoshells were used as contrast enhancers to image HER2 expression, a clinically relevant marker in human breast adenocarcinoma cells. Gold nanoshells were fabricated and tuned for Near Infrared (NIR) imaging. Then, the nanoshells were exposed to HER2 (specific) or IgG PEG-ylated (non-specific) antibodies to facilitate targeting of cultured human mammary adenocarcinoma cells. A microscope equipped with a bright field and dark field were used to evaluate cell viability. In vitro photothermal nanoshell therapy was performed and silver staining demonstrated the tissue targeting specificity of HER2 nanoshells as well as the non-specificity of IgG nanoshells. In addition to mediating photothermal destruction of breast cancer cells in vitro, it was demonstrated that NIR absorbing nanoshell bioconjugates can provide molecular specific optical contrast enhancement without cytotoxicity [15].Figure 10Ablation of two tumors in a mouse. With exposure to an external infrared laser source, the nanoshells resonate and thermally destroy tumor cells and their respective vasculature. Adapted from [103].Advantages and limitationsOptical imaging with nanoshells offers the potential for non-invasive, high resolution in vivo imaging at relatively low cost [15]. Scattering based optical imaging technologies rely on inherent changes in indices of refraction. Strategies that depend only on the intrinsic optical contrast within tissue have proved clinically valuable in some screening applications. However, such techniques are not sensitive enough to resolve an image based on disease biomarkers [15]. In cancer, when early detection is critical to reducing morbidity and mortality, the use of molecule specific contrast agents provides the ability to optically sense and image abnormalities long before pathologic changes occur at the anatomic level [15]. In the future, nanoshells may provide excellent contrast for other imaging modalities such as CT [96].ConclusionIn this review we have assessed a range of imaging techniques in systems biology spanning from microscopy to clinical imaging. In addition to the techniques reviewed, there are multiple other technologies that have lead to significant contributions to a systems level understanding of biological processes. Two such techniques are optical coherence tomography [97,98] and hyperspectral imaging [99,100]. With the refinement of current technologies and the development of new techniques, additional information will be available to help dissect biological systems.As seen in Figure 11, there is a resolution gap between microscopy and anatomical imaging. This gap also represents the divide between experimental and clinical imaging applications. In contrast, acquiring anatomical and metabolic information with clinical scanners has been achieved by coupling imaging technologies. For example, it is now commonplace for usage of combination PET/CT scanners. This allows for metabolic information acquired in PET to be readily registered with higher resolution anatomical CT images. Also, with fMRI a slower MRI scan is also conducted to form a detailed brain atlas, to which functional images are later registered. As a result there is a resolution continuum between anatomical and metabolic imaging. To reach the same end for microscopy and anatomical imaging, molecular probes such as quantum dots and nanoshells may find more clinical applications and thus improve the resolution that can be achieved with clinical scanners.Figure 11Resolution spectrum of imaging techniques. The schematic shows the resolution gap between microscopy and anatomical imaging. Metabolic imaging has successfully been linked to anatomical imaging despite having lower resolution. The schematic axis is linear.Beyond improvements in resolution, a grand challenge remains for the imaging technology development community: to enable dynamic imaging of both biological system components and of their respective connections. For example, the ability to resolve and monitor an entire mammalian cortical circuit in vivo has yet to be realized. Electrophysiology has been increasing complemented by fMRI over the last 15 years, but with fMRI information on neural activity is provided as an indirect measure and on the scale of hundreds of thousands or millions of neurons. Two-photon imaging has provided for single cell resolution, but functionally visualizing hundreds of synapses performing computations is limited by axonal labeling of neuronal populations and also by overall temporal acquisition frequency. As a result, innovations in methods for visualizing neural circuitry and for deciphering spike times will be necessary to further advance systems neuroscience with imaging. In a broader set of application areas, using imaging to simultaneously monitor components of a molecular network will be useful in further understanding cellular processes, such as apoptosis which is critical for the development of new cancer treatments.The further development of imaging technologies will continue to be important in the advancement of systems biology. Imaging can provide a wide array of data that can be used to build and validate models. The information acquired with imaging can be readily incorporated into models as biochemical concentrations, functional activity, and anatomical coordinates. In addition, imaging provides data for new discoveries and diagnostic information. Oncology and other areas in the biomedical sciences will benefit greatly from imaging and systems biology approaches.Authors' contributionsThe authors collectively wrote this review.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2533319.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2533319",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2533319\nAUTHORS: Chikako Suzuki, Michael R Torkzad, Soichi Tanaka, Gabriella Palmer, Johan Lindholm, Torbjörn Holm, Lennart Blomqvist\n\nABSTRACT:\nBackgroundMagnetic resonance imaging (MRI) is used for preoperative local staging in patients with rectal cancer. Our aim was to retrospectively study the effects of the imaging protocol on the staging accuracy.Patients and methodsMR-examinations of 37 patients with locally advanced disease were divided into two groups; compliant and noncompliant, based on the imaging protocol, without knowledge of the histopathological results. A compliant rectal cancer imaging protocol was defined as including T2-weighted imaging in the sagittal and axial planes with supplementary coronal in low rectal tumors, alongside a high-resolution plane perpendicular to the rectum at the level of the primary tumor. Protocols not complying with these criteria were defined as noncompliant. Histopathological results were used as gold standard.ResultsCompliant rectal imaging protocols showed significantly better correlation with histopathological results regarding assessment of anterior organ involvement (sensitivity and specificity rates in compliant group were 86% and 94%, respectively vs. 50% and 33% in the noncompliant group). Compliant imaging protocols also used statistically significantly smaller voxel sizes and fewer number of MR sequences than the noncompliant protocolsConclusionAppropriate MR imaging protocols enable more accurate local staging of locally advanced rectal tumors with less number of sequences and without intravenous gadolinium contrast agents.\n\nBODY:\nBackgroundTotal mesorectal excision (TME) is the standard surgical treatment used for patients with primary rectal cancer. TME involves removal of a distinct anatomic compartment, the mesorectum, containing the rectal tumor, all local draining nodes and the mesorectal fat, by means of sharp dissection along the mesorectal fascia [1-3]. There is substantial evidence for efficacy of neoadjuvant therapy in combination with TME as being important to reduce local tumor recurrence rates [4-7]. When performing TME, knowledge of the relationship of the tumor to the circumferential resection margin (CRM) is of importance. When CRM is involved by the tumor, the risk of local recurrence is high [8-16]. The local prognostic factors assessed at preoperative magnetic resonance imaging (MRI) of rectal cancer include the extent of extramural tumor spread, involvement of the lateral resection margin, involvement of neighboring organs in the pelvis, presence of local lymph node metastases, extramural lymphovascular infiltration and peritoneal involvement [15,17]. This information helps select patients who should receive neoadjuvant treatment. This applies especially to cases with locally advanced rectal cancer, in order to maximize the chances of a complete resection and survival [18,19], and at the same time, to minimize morbidity and loss of quality of life. It is therefore of paramount interest to provide detailed anatomic knowledge of tumor and tumor invasion toward neighboring organs before treatment.Although evaluated in several studies during the past two decades, it is only during recent years that MRI gained wide acceptance as a valuable method for assessment in patients with rectal cancer [20-33].As a tertiary referral center responsible for patients with advanced rectal cancer, we assess magnetic resonance (MR) examinations from other institutions and hospitals at multidisciplinary team (MDT) meetings. When demonstrating these examinations at MDT meetings, variations in imaging sequences among different centers are noted. These differences may be related to both different equipments and level of dedicated experience in pelvic MRI.To our knowledge, no study has reported the importance of the imaging protocol for assessment of tumor involvement of neighboring organs in locally advanced rectal cancer. The aim of the present study was to compare the equivalence between MRI and histopathology in patients with locally advanced rectal cancer based on the effects of using different MRI protocols.Patients and methodsForty-one patients assessed as clinically suspicious for locally advanced primary rectal cancer by surgeons from 2000 to 2005, were included. 37 patients, 27 male and 10 female, with a mean age of 60.1 ± 9.8 (mean ± SD, range 28–79) who had available MRI of the pelvis were studied further. The surgeon's decision that a cancer might be advanced was based on findings at diagnostic laparotomy and/or by means of digital rectal examination.Radiological assessmentAll examinations were provided from ten different hospitals or institutions (two of which were university hospitals). Each MR examination (all done on 1.5 T) was assessed by two or three radiologists (C.T., M.R.T. and L.B.) in consensus without knowledge of the clinical and histopathological results prior to this study according to a standard evaluation looking specifically at which organs and/or structures had been involved. However, the radiologists were aware of the high suspicion for locally advanced tumors by the clinicians. Radiologists had evaluated the morphological characteristics of the primary tumor, local prognostic factors including threatening or involvement of the mesorectal fascia, and adjacent organs in each patient.For the part of this study, anterior organs were defined as those positioned ventral to the rectum and included the seminal vesicles, the prostate gland, the perineal body, uterus, vagina, ovaries, the small and large intestines, and the urinary bladder. Inferior and posterior organs had been defined as those that were located inferior and dorsal to the rectum, respectively, and included the levator ani muscles, obturator muscles, piriformis muscles and the sacral bone. Involvement of the abovementioned organs was defined as T4-tumor stage.The imaging protocol of each MR-examination was recorded by one author (C.T.). Those examinations that showed the following prerequisites were defined as compliant rectal imaging protocol vs. those that did not demonstrate the same sequences (called henceforth noncompliant):1. Sagittal and axial T2-weighted images of the pelvis performed,2. T2-weighted images with equal to or less than 3 mm slice thickness perpendicular to the rectal length at the level of the tumor with a 16–20 cm field of view and at least a 256 × 256 matrix, otherwise called 'high resolution imaging' [20,21,25,34].3. For low rectal tumors, coronal imaging obtained.If the patients underwent MR examinations twice but at two different institutions, with different protocols, one compliant and the other non compliant; these were noted separately as combination protocol but categorized with the compliant group regarding some aspects. The number of other sequences and different types of artifacts (if distinguishable) were also noted.The common denominators of all MR examinations, whether compliant or otherwise, were that they had to be performed on the request of a surgeon or oncologist for assessment of local extension of the rectal tumor preoperatively, and that the radiologist at the primary institution had not called the examination incomplete.Histopathological examinationAll evaluations were performed according to the protocol of Quirke, et al [16,35], by one pathologist (J.L.) with more than 10 years of experience in gastrointestinal pathology. The pathologist was blinded to the MRI study protocol. The tumor site was sliced transversely at 0.5–1.0-cm intervals. The extent of tumor spread into mesorectal fascia and other structures or organs was assessed both macroscopically and with high magnification. Tumor extension into the surrounding structures and organs at microscopical examination were used as the standard of reference against which MRI findings were compared. The extension of tumor cells into mesorectal fascia and other structures or organs was assessed from inspection of the histological macrosection by light microscopy at 20× – 200× magnification.Statistical analysisAll MRI findings including the size of tumor, the name and number of involved fascia(e) and organ(s), the pattern of tumor involvement according to MRI and histopathology as well as the MR imaging protocol were recorded using Microsoft Excel 2003 and Microsoft Access 2000. Sensitivity and specificity of MRI between different groups were compared and 95% confidence interval (CI) was calculated with P-value < 0.05 considered significant using Stat View J-5.0 (SAS Institute. Inc., Cary, NC).Ethical considerationsThe study was approved by the local ethical committee. No separate informed consent was obtained for this retrospective study.ResultsTumor staging according to MRINineteen patients were evaluated as T4 rectal tumors based on MRI. The remaining 18 were evaluated as T3 tumors without obvious invasion of neighboring organ.Assessment of imaging qualityEleven patients were assessed as having compliant (D) protocols and 13 patients as combination protocols (C) and 13 patients a noncompliant imaging (N).Regarding imaging parameters, compliant imaging protocols were used with smaller field of view (FOV) (D, 201.7 ± 77.0 mm; N, 263.5 ± 129.8 mm; mean ± SD, p = 0.03), thinner slice thickness (D, 3.8 ± 1.4 mm; N, 5.3 ± 1.9 mm; mean ± SD, p < 0.01), smaller slice gap (D, 0.2 ± 0.9 mm; N 2.0 ± 2.4 mm; mean ± SD, p < 0.01) and smaller voxel size (D, 1.3 ± 1.5 mm3; N, 6.7 ± 6.0 mm3; mean ± SD, p < 0.01). The total number of MR sequences performed in each patient was also larger in the N group (N, 9.2 ± 3.2 sequences vs. D, 5.2 ± 0.7 sequences; mean ± SD, p < 0.01 (table 1). One patient from the noncompliant group had some motion artifacts.Table 1Comparison of various MR imaging parameters, average number of sequences in each group and imaging protocols.Compliant protocol (D)Noncompliant protocol (N)P-valueParameters on T2-WI*Field of view Mean ± SD (mm)201.7 ± 77.0263.5 ± 129.80.03Slice thickness Mean ± SD (mm)3.8 ± 1.45.3 ± 1.9< 0.01Gap Mean ± SD (mm)0.2 ± 0.92.0 ± 2.4< 0.01Matrix size Mean (mm × mm)0.5 × 0.50.9 × 1.10.02Voxel size Mean ± SD (mm3)1.3 ± 1.56.7 ± 6.0< 0.01No. of sequence Mean ± SD (mm)5.2 ± 0.79.2 ± 3.2< 0.01*T2 weighted image;Involvement of the anterior organsIn the group with compliant protocols and the group with combination protocol, preoperative MRI indicated tumor involvement of anterior pelvic organs in seven out of the 24 patients. Compared to pathological examination, six cases were true positives and one was false positive. Among the remaining 17 patients without organ involvement on MRI, pathological examination revealed one false negative case and 16 true negatives (table 2). Figure 1 demonstrates the false-negative case. In this case, there appears to be no continuity between the tumor and the uterus. However, histopathological examination showed tumor invasion along the fascia, reaching the posterior wall of the uterus and the left adnexa. The radiologist failed to ascertain the anterior extension of the tumor correctly.Figure 1MR images of the 'false negative' case in the group with a compliant protocol. A-63-year-old female with rectal cancer involving the mesorectal fascia, peritoneal reflection and the parietal pelvic fascia. Imaging parameters: TR; 4056, TE; 130, NEX; 2, Thickness; 5 mm, Gap; 0 mm, FOV; 240 mm. (a) Sagittal T2-w image of the pelvis. Primary lesion is located at the rectosigmoid junction with an extramural component, extending dorsally toward the presacral fascia (arrowhead). The tumor seems to be very distant from the inner genitalia (arrow). b-e) Axial T2-w images demonstrated in a craniocaudal direction with b being the uppermost image. In b, the extramural component reaches and thickens the peritoneal fold (arrow), and more inferiorly even the pelvic side wall fascia (arrowheads in c). This fascial thickening continues (arrowheads in d, 15 mm below b), until it sweeps forward (arrow in e, 25 mm below b) and at this point the inner genitalia were involved. At the first glance, there appears to be no continuity between the tumor and the mesorectal fascia, however, histopathological examination proved tumor cells inside the fibrotic tissue and infiltrating the uterine parenchyma and the left adenxa (arrowhead in e).Table 2Comparison of various MR protocols in terms of diagnostic accuracies regarding involvement anterior to rectum.Compliant andcombination  protocol (D and C)Noncompliant protocol (N)Imaging accuraciesTrue positive62True negative163False positive16False negative12Sensitivity (%) (95% CI)85.7 (42–99)50.0 (6–93)Specificity (%) (95% CI)94.1 (71–99)33.3 (7–70)Positive Predictive Value (%) (95% CI)85.7 (42–99)25.0 (3–65)Negative Predictive Value (%) (95% CI)94.1 (71–99)60.0 (14–94)In the noncompliant imaging group, preoperative MRI was indicative of organ involvement in eight cases. Pathological examination revealed two as true positives and six as false positives (Figure 2). Among the remaining five patients without organ involvement, pathological examination revealed two false negatives and three true negatives.Figure 2MRI of the false positive case in the group with a noncompliant protocol. A 76-year-old male with rectal cancer suspected of invasion to the urinary bladder. Imaging parameters: TR 7000; TE 132; NEX 2; thickness 5 mm; gap 1.5 mm; FOV 400 mm. (a) Sagittal T2-WI of the pelvis. The large primary lesion (asterisk) originating from the upper part of rectum with accompanying desmoplastic and edematous changes seems to be invading the muscular wall of the bladder dorsally (white arrows). The tumor appears to penetrate into the muscular layer of the urinary bladder which shows higher signal intensity compared to the normal part. (b) Sagittal contrast-enhanced T1-WI of the pelvis with fat-suppression. The posterior bladder wall is not distinguishable, yet the tumor is seen enriching ventrally (white arrowheads) and therefore, it is suspicious for penetrating into the bladder wall. (c-f) Corresponding axial images. c, e, and f are T2-WI and d is T1WI with contrast-enhancement and fat-suppression. T1-w images after Gadolinium contrast enhancement with fat saturation give the impression of the tumor (asterisk) growing into the dorsal wall of the urinary bladder (arrowheads). However, histopathological examination revealed no tumor involvement of the urinary bladder.Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) in the compliant and combination protocol group were 85.7%, 94.1%, 85.7%, and 94.1%, respectively. On the other hand, in the group with non-compliant protocol, the sensitivity, specificity, PPV and NPV were 50.0%, 33.3%, 25.0%, and 60%, respectively. Statistically significant difference (p < 0.05) was observed regarding measured specificity (95% CI; 7–70 for group N vs. 95% CI; 71–99 for the other two groups, D and C). The difference in sensitivity in the two groups did not reach statistical significance levels (Table 2).Posterior or inferior organ involvementOnly three out of the present 19 patients with locally advanced tumor, showed involvement of an inferior organ (levator ani muscle, piriformis muscle) or a posterior organ (Os sacrum) by the tumor, without simultaneous involvement of any anterior organ. Two of these patients used compliant imaging, and pathological examination revealed both to be true positives. In one patient with noncompliant imaging an inferior organ involvement was suspected but pathological examination proved no obvious tumor infiltration or fibrosis in that organ (false-positive). The number of cases was too few to make any meaningful statistical analysis.DiscussionThe results of this study indicate considerable differences in correlation between preoperative imaging and histopathology depending on the imaging protocol. Using compliant imaging, despite fewer imaging sequences, a considerably better prediction of tumor invasion towards anterior pelvic organs is seen. On the contrary, this study also indicates that MRI performed with noncompliant imaging protocol does not allow accurate prediction. One other observation is that the radiologist tends to over-stage when the imaging protocol is not optimal. This could be due to the fear of positive resection margins caused by a false negative assessment and partial volume effect observed with thick slices not obtained in the appropriate planes. This could of course be due to nature of the study as well. The radiologists assessing the MR exams were aware of the selection criteria and might have felt compelled to over-stage.The lack of compliant imaging, and as we suspect the lack of high resolution T2-weighted imaging, probably forced the radiologists to rely on images with considerable volume averaging. Compared to the compliant imaging, both slice thickness including gap and voxel size were significantly larger in the noncompliant imaging group (P < 0.05). Larger slice thickness and gap yield more partial volume effect, thus leading the radiologists to make overestimation of tumor extent. In areas of the pelvis where there are small interfaces between tissues, such as in the anterior and low part of the rectum, this is probably of particular importance. In the compliant and combination groups, there was one false positive and one false negative finding of anterior organ involvement out of 24 cases.In the noncompliant imaging group, there were six false positive and two false negative cases out of 13 cases. This means that one patient out of 24 from D and C groups and six patients out of 13 from the N group might receive unnecessary extensive surgery and prolonged, preoperative chemoradiotherapy. Anterior pelvic organs are closely related to urinary and sexual function, and anterior organ surgery has great impact on the patient's quality of life after surgery. By contrast at least partially because of false negative assessments by radiologists, one out of 24 cases from D and C groups, and two out of 13 cases from the N group had involved resection margins.Although the low number of cases prohibits any meaningful analysis to be done regarding accuracy of MRI for assessment of organs inferior or dorsal to rectum, our findings suggest that compliant imaging might be superior to noncompliant imaging also for these patients. This low frequency could be due to less likelihood of involvement of posterior organs compared to anterior organs due to more distance between rectum and these neighboring organs [36].The number of MR sequences was different between various groups with larger numbers observed in the noncompliant imaging group. It seems that whenever the compliant sequences were not employed, there was a tendency to conduct several other sequences. One of the most widely used sequences in the N group was the one with usage of gadolinium intravenous contrast. Recently, Vliegen and others have shown that gadolinium-enhanced MRI does not improve the diagnostic accuracy in local staging of rectal cancer [37]. Unnecessary use of contrast agents might only lead to increased rate of adverse events and increased costs and time needed for examination, without any proven benefit for the patients.There are a number of other limitations in this study. First, we did not compare the same patients using different imaging protocols.Second, there was a difference in the sensitivity of MR examinations using different protocols when assessing detection of anterior organ involvement, however, the difference did not reach statistical significance which is probably due to the low power of the study and perhaps the nature of the study (i.e. the radiologists knew that these cases were more likely to be advanced cases).However, even with these limitations, the compliant imaging improves accuracy, especially in advanced and complicated cases. It is therefore of utmost importance that radiologists are made aware of pitfalls and the problems, and that radiologist are made up-to-date about recent developments in imaging. This current study reveals that there is a need for continued education in this field.ConclusionFor local staging of locally advanced rectal cancer, the correlation between MRI and histopathology was better when a predefined compliant rectal imaging protocol was used. It is possible that this also holds true for all patients assessed with rectal cancer and not only for anterior structures in the pelvis. However, this has to be assessed in further studies. Furthermore, this study indicates that continuous training of radiologists and radiology technicians, including work-shops and seminars seems to be an appropriate way to improve accuracy of MRI in patients with rectal cancer.AbbreviationsMR(I): Magnetic resonance (imaging); TME: Total mesorectal excision; CRM: Circumferential resection margin; T2-w (image): T2 weighted (image); FOV: Field of view; MDT: Multidisciplinary team; PPV: Positive predictive value; NPV: Negative predictive value; TR: Repetition Time; TE: Echo Time; NEX: number of excitations.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsCS idea, data collection, radiological assessment, manuscript preparation. MT idea, data collection, radiological assessment, manuscript preparation. ST idea, data collection, surgical and clinical assessment, histopathological evaluation, manuscript preparation. GP idea, data collection, surgical and clinical assessment, manuscript preparation. TH idea, data collection, surgical and clinical assessment, histopathological evaluation, manuscript preparation. JL idea, data collection, histopathological evaluation, manuscript preparation. LB idea, supervision, manuscript preparation. All authors read and approved the final version\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2533324.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2533324",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2533324\nAUTHORS: Gauthier Bouche, Virginie Migeot\n\nABSTRACT:\nBackgroundUsing the Internet to seek health information is becoming more common. Its consequences on health care utilisation are hardly known in the general population, in particular among children whose parents seek health information on the Internet. Our objective was to investigate the relationship between parental use of the Internet to seek health information and primary care utilisation for their child.MethodsThis cross-sectional survey has been carried out in a population of parents of pre-school children in France. The main outcome measure was the self-reported number of primary care consultations for the child, according to parental use of the Internet to seek health information, adjusted for the characteristics of the parents and their child respectively, and parental use of other health information sources.ResultsA total of 1 068 out of 2 197 questionnaires were returned (response rate of 49%). No association was found between parental use of the Internet to seek health information and the number of consultations within the last 12 months for their child. Variables related to the number of primary care consultations were characteristics of the child (age, medical conditions, homeopathic treatment), parental characteristics (occupation, income, stress level) and consultation of other health information sources (advice from pharmacist, relatives).ConclusionWe did not find any relationship between parental use of the Internet to seek health information and primary care utilisation for children. The Internet seems to be used as a supplement to health services rather than as a replacement.\n\nBODY:\nBackgroundUsing the Internet to seek health information is becoming more common in Europe [1] as well as in the USA [2]. In France little data is available on the proportion of people who have ever used the Internet to seek health information [3] but this proportion seems to be increasing rapidly (from 15% in 2002 [1] to 37% in 2005 [4]).People use the Internet to seek health information because of its advantages. The Internet is widely available (home, work, libraries), convenient (24 h a day) and anonymous. A recent review highlighted the main reasons of using the Internet to seek health information: to gather additional information after a consultation, to access more complex information about a symptom, a disease or a treatment, to look for information about healthy lifestyles or healthcare services, to participate in an online support group and to be aware of other treatment alternatives [5]. Specific surveys have been carried out in samples of parents looking for health information. Many studies asked parents attending a paediatric hospital whether and why they used the Internet to seek heath information. A high proportion of parents attending outpatient departments seek health information online (from 53% to 64% in the most recent articles) [6-8]. This proportion is even higher for parents of children with a chronic disease or condition (from 58% to 89%) [9-11]. The authors highlighted that health professionals should advise a few selected websites to parents. Population-based surveys pointed out that mothers are high information seekers [12], especially during pregnancy and during the first few years following delivery look to find parenting advice and online clinical health information [12-14].The widespread utilisation of the Internet raises some questions about its impact on health behaviour, health services utilisation and finally on health outcomes. Although some characteristics of Internet users who seek health information have been well identified [3,15-17], no sufficient data is available to answer the above questions, in particular on the relation between seeking health information on the Internet and health care utilisation.Some observers have suggested that use of the Internet might actually decrease the cost of primary care services in systems with universal health care [18]. In that case, one might expect a negative relationship between use of the Internet and primary care utilisation. Others might speculate that Internet usage represents just another channel for activated, information-seeking behaviour, in which case the prediction might be for a positive relationship with primary care utilisation. Results from the studies about the impact of the Internet on health care utilisation are heterogeneous, showing a positive relationship between Internet use and service utilisation [19], a negative relationship [20,21] or no relationship at all [22,23]. All these studies but one [21] were carried out within an adult population and none have been conducted elsewhere than in the USA.Since the health of young children is of particular concern to parents, we assume that parental use of the Internet to seek health information would be related to primary care utilisation for their children. As we previously stated, results from studies available in the literature were heterogeneous [19-23]. Thus, we did not initially presume a positive or negative association.To test this association, we designed a cross-sectional study in a population of parents of pre-school children. In this paper, we examined the relationship between parental use of the Internet to seek health information and self-reported primary care consultation frequency for their children.MethodsStudy design, study population and sample sizeWe designed a cross-sectional survey carried out within a population of parents of pre-school children in the department of Vienne, France. In France, even if pre-school attendance is not compulsory, almost 100% of children aged 3 to 6 years attend pre-school (\"écoles maternelles\"). We defined 7 pre-school strata according to their private or public status and to their rural, semi-urban, urban or ZEP location [a ZEP school (Zone d'Education Prioritaire) is a school located in an underprivileged area. It benefits from additional resources to cope with academic and social problems. No ZEP private school exists in the Vienne area]. We selected certain schools in each of the 7 strata to ensure the representativeness of our sample according to the characteristics of the schools.To show a difference of at least one consultation in the last 12 months between those who use the Internet to seek health information (seekers) and those who do not (non-seekers), we needed 750 questionnaires (250 of seekers and 500 of non-seekers), expecting a proportion of one seeker for two non-seekers [4], to meet our objective (two-tailed hypothesis, 80% power, 5% alpha). Expecting a response rate of 35%, we sent out more than 2100 questionnaires. To meet the required sample size we selected 35 schools which represented a population of 2 197 children.Data collectionIn June 2007, with the help of pre-school principals, the parents of pre-school children were given a letter including an anonymous questionnaire, an explanatory note, and an envelope to return the questionnaire. Parents were asked to answer for their youngest schoolgoing child.The data collected was: parental characteristics, characteristics of their child, primary care consultation frequency for their child, and health information sources and methods used to seek health information on the Internet. The parental characteristics noted included age, household occupation according to the most advantaged occupation of either of the parents (disadvantaged (workers and unemployed), moderately advantaged (self-employed and employees) or advantaged (managers and executives)), parental education level according to the highest education level accomplished by either of the parents (elementary-secondary, high school diploma, lower tertiary or higher tertiary), annual family income (<14 000 €, from 14 000 to 19 999 €, from 20 000 to 29 999 €, from 30 000 to 39 999 € and ≥ 40 000 €), single parent family, place of residence (urban, rural), stress level assessed with a 10-unit visual analogue rating scale (10 indicative of higher stress) and Internet access (home, work only, none). The characteristics of the child collected were school attended, age, birth order, gender, medical conditions if any (preterm infant, hospitalisation after birth, asthma, wearing of glasses, auditory disorders, allergy, behavioural disorder, surgical intervention and others), long-term use of medication if any, whether under homeopathic treatment, frequency of administration of over the counter drugs to their child without medical advice and whether the parents had taken advice from a pharmacist for their child within the last 12 months. We asked parents to self-report the number of primary care consultations for their child within the last 12 months (general practitioner, paediatrician or accidents and emergency department). For consultations and child's medical conditions, we asked parents to refer to their child's health record booklet if necessary. Parents were also asked about other health information sources that they had already used amongst the Internet, medical books, medical dictionaries, television, press, relatives working in the medical sector, and other relatives.The study as a whole had been previously approved by the consultative committee on the processing of information in medical research of CNIL, the French national commission on individual privacy (approval AR071193).Statistical analysisThe dependant variable was the number of primary care consultations. The explanatory variables were child's age, child's birth order, child's medical condition and treatment, parental age, socio-economic position of the family, single parent family, parental stress level and health information sources. The main variable of interest was parental use of the Internet to seek health information.We assumed that the number of primary care consultations followed a negative binomial distribution – an extension of the Poisson distribution in case of over-dispersion [24,25]. We therefore used a negative binomial regression model to explain the variability of the number of consultations. We took the design effect (cluster effect) into account to avoid errors in the estimation of the parameters of the model [26].To perform our analysis, we used the negbin function of STATA [27] along with the svy option to take the design effect into account. The variable \"parental use of the Internet to seek health information\" was forced into the model. Other variables were included in the initial regression model if they were associated with the number of consultations with a p-value < 0.20 (using bivariate negative binomial regression). We then performed a multivariate analysis. From the initial regression model, variables were selected using a stepwise descending process. We tested the first-order interactions in the final model. Association between the number of consultations and variables of interest are rate ratios.ResultsResponse rate and characteristics of the study populationOf the 2 197 questionnaires distributed, 1068 questionnaires were returned (49%). Characteristics of the population are presented in Table 1. The mean number of primary care consultations for a pre-school child within the last 12 months was 5.9 ± 4.6. Distribution of the self-reported number of primary care consultations is shown in Figure 1, which confirms the assumption of a negative binomial distribution. Data on the number of consultations were missing for 39 questionnaires (4%).Table 1Characteristics of the 1068 pre-school children and their parentsCharacteristicsCharacteristics of the childrenN%Age (n = 1067)- 2 or 3 years old22721%- 4 years old33531%- 5 years old32831%- 6 years old17717%Girls (n = 1068)53750%First in birth order (n = 1039)49848%Medical conditions (n = 1068)- Preterm infant969%- Hospitalisation after birth818%- Asthma15615%- Wearing of glasses14614%- Auditory disorders636%- Allergy21420%- Behavioural disorder444%- At least one surgical intervention22521%- Other medical conditions707%Long term medication use (n = 1068)737%Frequency of administrating over the counter drugs without medical advice (n = 1062)- Never16816%- Rarely31129%- Often50948%- Almost every time my child is ill747%Under homeopathic treatment (n = 1043)54953%Advice from a pharmacist within the last 12 months (n = 1051)44542%MeanSDN. of primary care consultations within the last 12 months (n = 1029)5.94.6Parental characteristicsN%Single parent family (n = 1066)13613%Rural place of residence (n = 1064)21720%Occupation (n = 1056)- Disadvantaged (workers, unemployed)15114%- Moderately advantaged (self-employed, employees)56353%- Advantaged (managers and executives)34233%Education level (n = 1037)- Elementary and secondary28527%- High school diploma (\"Baccalauréat\")23823%- Lower tertiary23523%- Higher tertiary27927%Annual family income (n = 949)- < 14 000 €16818%- 14 000 to 19 999 €20321%- 20 000 to 29 999 €28630%- 30 000 to 39 999 €17018%- ≥ 40 000 €12213%MeanSDAge of the respondent in year (n = 1058)34.05.0Parental stress level (n = 1040)4.92.3Figure 1Distribution of the self-reported number of primary care consultations for the child within the last 12 months.The Internet was the most used health information source with 556 families (52%) who at least once had used it to seek health information. Relatives working in the medical sector and television were the second and third most common health information sources, with 518 (49%) and 397 (37%) families respectively having at least once used these sources (Figure 2).Figure 2Frequency (and 95% confidence intervals) of the different health information sources used by the 1 068 parents.Use of the Internet to seek health information and number of primary care consultations for their childMean numbers of consultations according to population characteristics together with results of the bivariate analysis are presented in Table 2. The multivariate analysis has been carried out on the 886 questionnaires (83%) for which no data was missing. Four variables were dropped due to the stepwise regression analysis (parental age, single parent family, allergy and television as health information source). None of the first-order interactions between explanatory variables of the final model were significant. Results of the multivariate analysis are presented in Table 3. No association was found between use of the Internet to seek health information and the number of consultations within the last 12 months (adjusted rate ratio 0.97; 95% CI 0.86 to 1.09). Seeking health information from relatives (whether they were from the medical sector or not) was associated with a slight increase in the number of consultations. The main variables related to the number of primary care consultations were child's age and medical condition. Number of consultations within the last 12 months decreased with child's age (with a decrease of 16%, 23% and 33% for children aged 4, 5 and 6 years respectively compared to children aged 2 or 3 years). Most of the children's medical conditions were positively related to the number of consultations. Some parental characteristics were related to a lower number of consultations: moderately advantaged occupation and annual income from 20 000€ and over. Parental stress was related to a higher number of consultations with a 3% increase in the number of consultations for every one-unit increase in the visual analogue rating scale.Table 2Relation between mean number of primary care consultations for the child and population characteristics – bivariate negative regression analysis.Explanatory variables (n = 1029 unless otherwise indicated)MeanSDPHealth information sourcesMedical dictionaryNon = 8265.84.20.38Yesn = 2036.35.8InternetNon = 4905.94.70.92Yesn = 5395.94.5PressNon = 7535.84.30.20Yesn = 2766.35.2Medical booksNon = 8095.84.50.21Yesn = 2206.34.9TelevisionNon = 6465.74.20.05Yesn = 3836.25.1Relatives working in the medical sectorNon = 5285.64.20.004Yesn = 5016.35.0Other relativesNon = 7795.74.40.006Yesn = 2506.65.0Characteristics of the childAge (n = 1028)2–3 years oldn = 2217.35.60.00014 yearsn = 3196.03.85 yearsn = 3175.65.06 yearsn = 1714.73.1First in birth order (n = 1002)Non = 5235.84.80.41Yesn = 4796.14.4Child genderBoyn = 5146.04.90.71Girln = 5155.94.2Medical conditionsPreterm infant (n = 1025)Non = 9335.94.50.21Yesn = 926.65.4Hospitalisation after birth (n = 1023)Non = 9485.94.50.34Yesn = 756.65.7AsthmaNon = 8795.64.4<0.0001Yesn = 1507.95.1Wearing of glassesNon = 8885.84.30.05Yesn = 1416.76.2Auditory disordersNon = 9675.84.60.006Yesn = 627.44.6AllergyNon = 8255.64.6<0.0001Yesn = 4047.34.4Behavioural disorderNon = 9875.94.60.88Yesn = 425.84.2At least one surgical interventionNon = 8175.74.60.0004Yesn = 2126.84.4Other medical conditionsNon = 9605.94.50.10Yesn = 696.75.0Long term medication useNon = 9615.74.5<0.0001Yesn = 688.95.2Frequency of administrating over the counter drugs without medical advice (n = 1025)Nevern = 1585.85.00.63Rarelyn = 3005.84.3Oftenn = 4946.14.7Almost every time my child is illn = 735.54.4Under homeopathic treatment (n = 1008)Non = 4745.64.30.08Yesn = 5346.24.8Advice from a pharmacist within the last 12 months (n = 1027)Non = 5925.44.20.0003Yesn = 4356.65.0Parental characteristicsSingle parent family (n = 1028)Non = 8995.84.30.02Yesn = 1296.76.0Occupation (n = 1019)Disadvantagedn = 1426.86.30.04Moderately advantagedn = 5425.94.6Advantagedn = 3355.63.6Education level (n = 1002)Elementary and secondaryn = 2686.66.20.20High school diploman = 2335.94.1Lower tertiaryn = 2325.63.6Higher tertiaryn = 2695.73.9Annual family income (n = 921)< 14 000 €n = 1597.27.40.0314 000 to 19 999 €n = 1955.83.920 000 to 29 999 €n = 2795.53.530 000 to 39 999 €n = 1675.93.9≥ 40 000 €n = 1215.43.9Rural place of residence (n = 1025)Non = 8155.94.70.88Yesn = 2105.94.3Rate Ratio95% CIAge of the respondent (n = 1020)0.98[0.97 – 0.99]Parental stress level (n = 1005)1.03[1.01 – 1.06]Table 3Relation between number of primary care consultations for the child and population characteristics – multivariate analysis.Explanatory variablesAdjusted Rate Ratio*95% CIHealth information sourcesInternetNo1Yes0.97[0.86–1.09]Relatives working in the medical sectorNo1Yes1.08[1.01–1.16]Other relativesNo1Yes1.12[1.01–1.25]Characteristics of the childAge2–3 years old14 years0.84[0.73–0.96]5 years0.77[0.65–0.90]6 years0.67[0.56–0.80]Medical conditionsAsthmaNo1Yes1.28[1.13–1.46]Wearing of glassesNo1Yes1.19[1.03–1.45]Auditory disordersNo1Yes1.23[1.03–1.46]At least one surgical interventionNo1Yes1.12[1.02–1.22]Long term medication useNo1Yes1.26[1.06–1.49]Under homeopathic treatmentNo1Yes1.13[1.03–1.23]Advice from a pharmacist within the last 12 monthsNo1Yes1.14[1.05–1.24]Parental characteristicsOccupationDisadvantaged1Moderately advantaged0.85[0.75–0.96]Advantaged0.87[0.75–1.02]Annual family income< 14 000 €114 000 to 19 999 €0.86[0.73–1.01]20 000 to 29 999 €0.80[0.68–0.94]30 000 to 39 999 €0.87[0.76–0.99]≥ 40 000 €0.80[0.64–0.99]Parental stress level1.03[1.01–1.05]* Adjusted on other health information sources (dictionary, press, books and television), characteristics of the child (birth order, gender, medical conditions – preterm infant, hospitalisation after birth, allergy, behavioral disorder – frequency of administrating over the counter drugs without medical advice) and parental characteristics (age, single parent, education level and place of residence).We also fitted a model without income on the 974 families (91%) for whom no data other than income was missing (data not shown). Results were very similar though associations of auditory disorders and homeopathic treatment with the number of consultations were not significant anymore.DiscussionWe did not find any relationship between parental use of the Internet to seek health information and the number of self-reported consultations for their child. This finding runs counter to our initial assumption that parental use of the Internet to seek health information would be related to primary care utilisation, either in a positive or in a negative way.To our knowledge, this is the first study carried out elsewhere than in the USA to assess this relationship. Our results are consistent with an interventional study on mental health services utilisation, which did not find any significant difference in the number of mental health visits between a group that had Web site access and the control group [22]. A quasi-experimental study carried out by Wagner et al. found a null association for parents [23] and a negative association for children [21]. Data from this study are not fully comparable with our study because the intervention was complex and not entirely Internet-based (self-care books, telephone advice nurses and computers). Within a population of Internet users, Eastin and Guinsler found an interaction between anxiety and Internet use to seek health information [20]. Anxious individuals who used the Internet to seek health information had fewer consultations than anxious non-users, whereas such a difference was not found for less anxious individuals. Our findings do not corroborate this interaction, since we did not find any interaction between stress levels and parental use of the Internet to seek health information. Another American study conducted in 1999–2000 found a positive association with an increase of 1.6 consultations for women using computer-based resources [19]. Differences in time period (1999–2000), geographical area (Baltimore metropolitan area, USA) and potential selection bias in both studies are likely to explain the differences between these results and our findings. Finally, findings from a cross-sectional study in seven European countries that investigated patterns of health-related Internet use and its consequences support our results. Only 6% of the sample claimed that they had made, cancelled or changed a doctor's appointment based on health related Internet activity [28]. Even if the number of consultations was not collected in this study, only 6% of the sample claimed that they have made, cancelled or changed a doctor's appointment based on health related Internet activity.The association that we found between primary care utilisation and the child's age [29-31], child's medical condition [29], familial socio-economic position [32] and psychological factors [33] are consistent with previous findings. However, in most studies performed in the USA [29,30,34], lower socio-economic position was associated with less frequent primary care utilisation, which is contrary to our results. Explanations are likely to come from the differences between the French and the American health insurance coverage for children. In France, health insurance coverage is nearly universal [35]. In the USA, most of the studies have been carried out in the 1990's or before, when the problem of uninsured children was raised [29,30,34]. At that time, children from the poorest families were more likely to be uninsured resulting in a lower number of primary care consultations. A few new factors associated with the number of primary care consultations have been identified in our study. Taking advice from a pharmacist, using relatives or friends as a health information source or using homeopathy for one's child could be explained by increased parental consciousness of health issues. These findings might reflect the \"familial context\" mentioned by Cardol [36] which would explain about 20% of the variability of the number of consultations in primary care.The first limitation of our study was the overall response rate of 49%. However, details of response rates of each school gave us information to identify the bias due to non-responses. We found that response rates were lower in schools with a higher proportion of families of low socio-economic position and with a higher proportion of non-French speaking families. We therefore probably over-estimated the proportion of families who used the Internet to seek health information, and possibly under-estimated the mean number of consultations. The second limitation was that data on race/ethnicity was not asked in the questionnaire. According to the French population statistics, less than five percent of the inhabitants of the department of Vienne are non native French. In this context, race/ethnicity is not so important even if it is well established in the American context that disparities in Internet use for health information exist according to race/ethnicity [37,38].Another limitation was that the number of consultations was self-reported by the parents. Many studies have shown a tendency for underestimation when people were asked to report the frequency of their health care utilisation. Since we found a mean of 5.9 consultations per child within the last 12 months, which is consistent with the data of health care utilisation from the provider [39], the bias may be small. Missing data for parental income was another limitation, with 119 (11%) parents who did not report their annual family income. This omission in reporting annual family incomes is information probably not missing at random because it is more likely to occur when the income level is relatively high [40].ConclusionEven if our study had some limitations, we demonstrated that there was no relationship between parental use of the Internet to seek health information and primary care consultation of their children. The Internet seems to be used as a supplement to health services for some rather than as a replacement. Individuals are increasingly involved in the management of their own health and using the Internet to seek health information is one way to be involved actively. Some authors suggested that this would lead to saving in health costs [18]. According to our findings, this may not be true from a short-term perspective. In our opinion, what is more likely to occur is an improvement in the health of those who use the Internet to seek health information, which, from a long-term perspective, would eventually lead to saving in health costs.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsGB designed the study, collected the data and had primary responsibility for data analysis and manuscript preparation. VM helped with the implementation of the study, validated the methodology and contributed to data analysis and manuscript preparation. All authors read and approved the final manuscript.Pre-publication historyThe pre-publication history for this paper can be accessed here:\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2533326.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2533326",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2533326\nAUTHORS: Forouzan Rezapur-Shahkolai, Mohsen Naghavi, Mohammadreza Shokouhi, Lucie Laflamme\n\nABSTRACT:\nBackgroundKnowledge is sparse concerning injuries affecting rural populations in low and middle-income countries in general and in Iran in particular. This study documents the incidence and characteristics of severe injuries affecting rural people in the Iranian district of Twiserkan and it investigates these people's suggestions for injury prevention and control.MethodsAn interview-based investigation was undertaken that comprised all unintentional injuries leading to hospitalization (more than 6 hours) or death that had occurred within a twelve month period and that were identified in the files of the 62 \"health houses\" of the Twiserkan district. For each case, semi-structured interviews were conducted at the households of the injured people (134 injuries affecting 117 households were identified).ResultsThe incidence rates of fatal and non-fatal injuries were respectively 4.1 and 17.2 per 10 000 person-years and, as expected, men were more affected than women (77.6% of all injury cases). Traffic injuries (in particular among motorcyclists) were as common as home-related injuries but they were far more fatal. Among common suggestions for prevention, people mentioned that the authorities could work on the design and engineering of the infrastructure in and around the village, that the rural health workers could contribute more with local information and education and that the people themselves could consider behaving in a safer manner.ConclusionNot only domestic injuries but also those in traffic are an important cause of severe and fatal injury among rural people. Health workers may play an important role in injury surveillance and in identifying context-relevant means of prevention that they or other actors may then implement.\n\nBODY:\nBackgroundInjuries constitute an important health problem worldwide and they are one of the major causes of death among people under 45 years old [1,2]. The majority of all injury-related deaths occur in low and middle-income countries [1-4] where knowledge is scarce regarding injury distribution, pattern and prevention [5]. Epidemiological studies have been conducted in some low and middle-income countries but, most often, traffic-related injuries and those occurring in urban settings have been in focus [6-8]. Yet, studies in rural areas have been conducted in countries in various continents, including Asia (Pakistan [9], Bangladesh [10,11], India [12], and Vietnam [5,13]), Africa (Kenya [14], Ghana [15], Uganda [16] and Tanzania [17]) and South America (Nicaragua [18]). Those studies reveal that injuries constitute an important health problem in the rural areas.In Iran, where this research has been conducted, injury-related years of life lost are higher than for the worldwide average [19,20]. Studies on injury epidemiology and prevention are limited and are mainly urban [21-23]. Also people's experiences and opinions about injury prevention and control have rarely been addressed. Currently, approximately 33% of the total population live in rural areas [24], and people benefit from a well-established health network, consisting of village-based local \"health houses\", from which health workers (known as Behvarzes) work. The main function of the Behvarz is to offer primary health care services to the local population and to gather health information. Usually the Behvarzes are selected from their local community and can therefore establish a very close relationship with community members. This, in turn, can help to gather accurate data. Health house workers also contribute to the simple but well-integrated health information system [25].In 2004, the Ministry of Health and Medical Education launched a program ultimately aiming at the reduction of injuries in rural areas. As a first step, an injury registration and surveillance system has been developed that forms part of the health information system and implies that the Behvarzes are responsible for the registration of all injuries leading to hospitalization (at least 6 hours as a standard criterion) or death. This data should provide information on the frequency of occurrence and characteristics of severe injuries and allow for the follow-up of future national and local interventions. Twiserkan district, where this study was conducted, is one of the districts selected for the pilot phase of the implementation of the surveillance system.In the current study, we take advantage of the reports gathered by the Behvarzes over a one year period to assess the incidence of rural injuries and, through interviews with injured people or their relatives, characterize those injuries' epidemiology and document the suggestions of people from affected families concerning injury prevention and control.MethodsThe Twiserkan district is located in the Hamadan province (over 19 000 square kilometers), in western Iran. In 2002, Hamadan had over 1.7 million inhabitants, of whom 44% lived in rural areas [26,27]. In 2006, when this study was conducted, Twiserkan district had a population of about 110 000 inhabitants, of which 58% was rural. The number of rural households was 14,789 and the number of inhabitants amounted to 62,857.All unintentional injuries leading to hospitalization (more than 6 hours) or death, occurring over a one-year period (June 1, 2005 – May 31, 2006), were considered. These were first identified in the files compiled at the local health houses (n = 62) in the Twiserkan district (see below). Thereafter the household of each injured person was visited by a trained and experienced interviewer. Before any interview, the aim of the study and main content of the interview were explained to the interviewee who was also guaranteed confidentiality. Once verbal consent was given, a short face-to-face interview was conducted (in June 2006; 134 injuries in 117 households) with the family member identified as responsible of caring for the household (response rate 100%). To ensure as complete answers as possible, the injured family member took part in the interview any time she or he was present at the time of visit. This was the most common situation, except of course for fatal injuries. It can be underlined that the recall period ranged from a few days to one year post injury. About two-thirds of the injuries had occurred between 6 to 12 months before the interview and the remaining occurred either 3 to 6 (13.4%) or less than 3 months (18.7%).Prior to data collection, a structured questionnaire including both closed and open questions was developed and pre-tested by the research team. It included information about both the injured person (and his/her household) and the injury (type, nature and circumstances of occurrence). Open-ended questions were included so as to find out how people regarded the role of the community, the Behvarzes, and the authorities (health and others) with regard to injury prevention. For each actor, people were asked to give their opinion as to what more could be done to help reduce injury for the village residents. In households with more than one injury during the reference period, all injuries sustained were considered at once when addressing the questions about household opinions concerning the roles of different actors in injury prevention and control.The study was approved by the Iranian National Ethics Committee in Medical Research, Ministry of Health and Medical Education of Iran.Data were entered, processed and analyzed in Excel (version 2003). Injury incidence rates were estimated globally and for fatal and non-fatal injuries respectively. Injury characteristics were coded and categorized according to the WHO guidelines for injury survey and surveillance [28,29]. Thereafter, using descriptive statistics, the characteristics of the injured people (sex, age group, education and occupation), of the injury events (place of injury and injury mechanism) and of their consequences (nature, body region and injury severity/recovery) were highlighted. When a person sustained several injuries during the same injurious event, the most severe one was considered. This was made possible as space was provided in the questionnaire to identify the most severe injury, following the WHO guidelines and after discussion and consensus between two members of the research team.People's opinions about injury prevention were first entered as free text. Thereafter, answers were read by two members of the research team and key ideas/phrases were identified and discussed at different sessions. Meaningful categories were identified that represented specific and homogeneous domains of potential intervention or action. Since some people had several suggestions and as they were not asked to prioritize or rank them, all opinions expressed by each respondent were taken into account.ResultsInjury incidence of non-fatal and fatal injuriesA total of 134 injuries were reported by the Behvarzes during the study period. These were identified among 117 households (of 14,789 in total) and consisted of 26 fatal and 108 non-fatal injuries. The corresponding incidence rates of injuries per 10 000 person-years are therefore 21.4 injuries in total (95% CI 17.7–24.9), 4.1 fatal injuries per 10 000 person-years (95% CI 2.5–5.7) and 17.2 non-fatal injuries (95% CI 13.9–20.4).Injury deaths occurred most often in traffic crashes (n = 10), followed by burns (n = 6), poisoning (n = 5), falls (n = 4), and electrocution (n = 1). Of the 134 injury cases identified, 22 were attributable to seven injurious events: three traffic-related and four in the home (see below). In the remainder of the text, all injured people, even those injured in the same event, are considered as individual cases.Injured people's characteristicsTable 1 presents the characteristics of the injured people. In total, about three-quarters were males; 21.6% were aged 15 or less and an additional 22.4%, 56 and over. The majority had not completed high school (85.8%) and 26.1% were farmers.Table 1Characteristics of the fatal and non-fatal unintentional injured people (June 1, 2005-May 31, 2006)CharacteristicsNumber%SexMale10477.6Female3022.4Age group (in years)< 110.71–586.06–152014.916–252518.726–351712.736–451712.746–551611.956–65139.766+1712.7Education*High school graduate (high school is grades 9 to 12) or above96.7Secondary school graduate (secondary school is grades 6 to 8) and/or some high school2417.9Completed primary school (primary school is up to grade 5) and/or some secondary school3425.4Primary school not completed2014.9No schooling3727.6N/A (children under 6 years old)107.5OccupationFarmer3526.1Other self employed1511.2Student2417.9Housewife2014.9Unemployed1410.4Labourer139.7Retired107.5Other (governmental employee and conscript)32.2N/A (children under 10 years old)107.5*In decreasing order according to the Iranian system.Fatal and non-fatal unintentional injury characteristicsFigures 1 and Figure 2 show the characteristics of the fatal and non-fatal unintentional injury circumstances (place of injury and injury mechanism). The injuries occurred in similar proportions in the home or on the road, outside the village. Traffic injury was by far the most common injury mechanism (44.8%), followed by falls (26.1%) and thereafter burns (fire/flame/heat; 11.2%). More details of these three injury mechanisms are given in the text below.Figure 1Places of occurrence of fatal and non-fatal unintentional injury events (June 1, 2005 – May 31, 2006).Figure 2Mechanisms of fatal and non-fatal unintentional injuries (June 1, 2005 – May 31, 2006).Road Traffic Injuries (RTIs; 60 injuries from 56 households)Of 60 RTIs, 48 were sustained by males and the bulk of them were among people of working age, most often 16–35 years (28 cases). Only 10 cases were reported among children 15 years and less. The most common mode of transportation among the injured people was by motorcycle (43.3%), followed by car (26.7%) and on foot (20%); most injured people were drivers (60%). The number of people injuries as motor vehicle passenger was almost equal between males and females (5 and 4 respectively). Injuries occurred above all on roads outside the village (68.3%).As mentioned above, traffic crashes were the primary cause of fatal injuries (n = 10). Two pedestrians were killed and eight motor-vehicle riders: five drivers (1 car, 2 motorcycle and 2 tractor drivers) and three car passengers. Three crashes involved more than one injured person with the following consequences: (1) the first had four injured people, three recovered and one died; (2) the second had two injured people, one recovered and one was disabled; (3) the last one had two injured people and both recovered completely.Falls (35 injuries from 35 households)Of 35 fall injuries, 28 were among males and 29 among adults – one in three (12 cases) were among people aged 65 years and above. Falls on the same level and from a roof were the most common kinds of falls reported (8 cases each), followed by fall from tree (n = 7) and from stairs (n = 5). Among older people, fall on the same level and from stairs were most common (4 cases each). Falls from trees occurred among adults and during work activities. This is common in rural Twiserkan, especially during the walnut harvest. Four deaths were fall-related: three occurred close to the injurious event (a man fell from a walnut tree, an older one fell from a roof, a woman aged 78 years on the stairs), and one some days later (a man aged 80 years following a pelvis fracture sustained after a fall when walking).Burns (15 injuries from 8 households)As many as 11 out of 15 burns were among males and 6 cases were paediatric burns (0–15 years). All burns occurred at home. Contact with flame was the most common cause (n = 12), several injurious events occurred when manipulating gas equipment used for cooking or heating. Other burns resulted from contact with hot liquids, steam or other gas (n = 3). Three events had more than one injured person: the first had three (two recovered and one died), the second had two (one recovered and the other one disabled) and the third one led to as many as five casualties.Burns were the second cause of fatal injuries: six casualties resulting from two injurious events. The first occurred in the evening while the family members (5 persons) were sitting in a room and one of them was trying to unscrew a small gas capsule (picnic gas stove), to prevent a gas leak. According to the interviewee, the capsule came off suddenly, caught fire and all five persons received serious burn injuries. They were taken to the hospital but all five died after a few days. The second event occurred in the morning and involved three people. In circumstances similar to the preceding case – a gas leak from a gas capsule – the house caught fire suddenly. The three family members were also taken to the hospital but the family's one-year old child died after a few days.Fatal and non-fatal unintentional injury consequencesTable 2 shows the characteristics of consequences of the fatal and non-fatal unintentional injuries, all injuries aggregated and for three injury mechanisms: traffic, fall, and burn. Fracture was the most frequent nature of injury, encompassing about half of the all cases and 86% of fall-related injuries. The category \"organ system injury/internal injury\" refers to damage of some vital internal systems such as respiratory system, blood circulation system. In cases where more than one nature of injury was reported, focus was placed on the most serious one, according to WHO guidelines [29].Table 2Consequences of the fatal and non-fatal unintentional injuries by mechanism (June 1, 2005-May 31, 2006)ConsequencesAll (n = 134)Number (%)Traffic (n = 60)Number (%)Fall (n = 35)Number (%)Burn* (n = 15)NumberPhysical natureFracture66 (49.3)32 (53.3)30 (85.7)-Concussion17 (12.7)15 (25.0)2 (5.7)-Cut, bite or other open wound15 (11.2)4 (6.7)1 (2.9)-Burn15 (11.2)--15Poisoning7 (5.2)---Organ system injury/internal injury6 (4.5)3 (5.0)--Bruise or superficial injury4 (3.0)4 (6.7)--Sprain/strain3 (2.2)1 (1.7)2 (5.7)-Unspecified1 (0.7)1 (1.7)--Body regionHead/face19 (14.2)17 (28.3)1 (2.9)1Upper limb18 (13.4)5 (8.3)8 (22.9)-Lower limb27 (20.1)15 (25.0)7 (20.0)1Neck/shoulder/lower back/rib10 (7.4)4 (6.7)3 (8.6)2Pelvis/hip8 (6.0)1 (1.7)7 (20.0)-Internal system7 (5.2)---Multiple regions42 (31.3)16 (26.7)9 (25.7)10Unspecified3 (2.2)2 (3.3)-1Severity/recoveryComplete recovery54 (40.3)25 (41.7)16 (45.7)3Partial recovery31 (23.1)14 (23.3)10 (28.6)2Disability22 (16.4)11 (18.3)5 (14.3)4Death26 (19.4)10 (16.7)4 (11.4)6*As the number of burns amounted to 15, no percentages are presented.For all injuries aggregated, the first most common injured single body region was the lower limb (20.1%), followed by the head/face (14.2%) and the upper limb (13.4% respectively). The most commonly injured single body region for traffic injuries was head/face (28.3%) followed by lower limb (25%) and for fall-related injuries was upper limb (22.9%) followed by lower limb and pelvis/hip equally (each 20%).Whenever an injured person had more than one injured body region, it was considered and coded as \"multiple regions\"; which occurred in nearly one in three injuries (31.3%). As for the injury severity, as mentioned above, almost one in five injuries was fatal. Twice as many injured people recovered completely and 23.1% recovered partially.Interviewees' suggestions for preventionInterviewees came up with quite a lot of suggestions concerning what more could be done to contribute to injury prevention and these suggestions varied considerably in kind depending on what actor they were asked to reflect upon: the authorities, the Behvarzes or the people themselves. The suggestions proposed were organized in a number of different categories and are presented in Table 3, shown in percentages both by number of suggestions and by number of households and also considering all injuries aggregated and traffic, fall and burn injuries separately.Table 3Household-based suggestions about activities that can be undertaken by various actors in order to control and prevent injuriesAllTrafficFallBurnno. household=117no. household=56no. household=35no. household=8People's opinions/suggestions% suggestions (% households)% suggestions (% households)% suggestions (% households)NumberAuthoritiesEngineering/building (infrastructure/signalization/product)31.1 (48.7)37.4 (60.7)34.8 (45.7)1Provision of daily services (accessibility/availability)19.7 (30.8)19.8 (32.1)19.6 (25.7)3Financial support12.0 (18.8)7.7 (12.5)23.9 (31.4)-Enforcement10.4 (16.2)12.1 (19.6)6.5 (8.6)1Provision of emergency services (accessibility/availability)8.7 (13.7)5.5 (8.9)6.5 (8.6)5Education8.2 (12.8)8.8 (14.3)4.3 (5.7)1Maintenance/repair7.7 (12.0)8.8 (14.3)2.2 (2.9)1Other2.2 (3.4)-2.2 (2.9)-Total number of suggestions183914612BehvarzesInstruction/education/information62.4 (58.1)69.4 (60.7)63.3 (54.3)3Post trauma care (health house/home visit)16.5 (15.4)14.3 (12.5)13.3 (11.4)1Availability/accessibility of Behvarz all day round/drug and equipment of health house10.1 (9.4)10.2 (8.9)10.0 (8.6)1As good as it can be9.2 (8.5)4.1 (3.6)13.3 (11.4)1Other (Insist on of financial support by authorities and pass people's problem to them)1.8 (1.7)2.0 (1.8)-1Total number of suggestions10949307PeopleBehave in a safe manner34.1 (37.6)35 (37.5)30.0 (34.3)3Cooperation together18.6 (20.5)13.3 (14.3)20.0 (22.9)3Cooperation (with authorities)17.8 (19.7)20.0 (21.4)20.0 (22.9)1Compliance12.4 (13.7)18.3 (19.6)7.5 (8.6)-Engineering/building8.5 (9.4)3.3 (3.6)17.5 (20.0)1Education (pay attention to planned education for them and/or pass to their children)8.5 (9.4)10.0 (10.7)5 (5.7)1Total number of suggestions12960409AuthoritiesAlmost one-third (31.1%) of the suggestions concerned changes in the living and commuting environment e.g., engineering and/or building (including infrastructure such as road construction and asphalting; signalization and product design). Proposals regarding the infrastructure and modernization of the traffic were most common, followed by suggestions concerning the provision of daily services (19.7 %), e.g., piped gas and even fire station for burns prevention, outdoor lighting for fall prevention and telephone for rapid contact with the health services at the time of injury event. People also mentioned the need for financial support for safety improvement in the village and even for better housing (12%). Law enforcement was also raised, in particular for traffic injury prevention (10.4%).BehvarzesFor the Behvarzes, focus was placed largely on the provision of education to the people (about 60–69%), including various forms of continuous education in the population and informing upwards in the health system. They pointed to safety education in general and also to different aspects of safety educations such as road traffic safety education for the young (especially in relation to motorcycling) and home safety (targeting childcare and supervision). The latter was also considered as a way to promote the development of safe behaviour in children. The second important category of suggestions concerning the Behvarzes was about post trauma care, including both health house and home visits (16.5 %). Greater availability and accessibility to the Behvarzes and also to drugs and equipment (10.1%) came in third place.People themselvesAbout one in three respondents considered that people themselves could behave in a safer manner to help control and prevent injuries. Thereafter, respondents raised the issue of cooperation between people (18.6%), including assisting each other in the event of injury, e.g., by transferring injured people to the health facilities. The third category of suggestions was about people's cooperation with the authorities to improve safety in the village-environment (17.8%). People also raised issues related to compliance, in-house engineering, and being good role-models for children, including teaching them safe behaviours and practices.DiscussionThe study reveals that injuries among people in rural areas affect mainly males and also people of working age, which is in line with the injury distribution in other settings [2,5,9,11,14,17,30,31]. One finding is the relative importance of injuries on the road, not only because of their severity (10 out 26 deaths), but also because of their frequency. They are indeed as numerous as injuries in the home. Similar results were obtained in earlier studies from Vietnam [5] and Nicaragua [18].Thus, traffic injuries are not only a concern among urban people in Iran but also among rural dwellers [32]. As motorized commuting is on the increase even in rural Iran, it is important to pay attention to road safety [4,32]. In particular, the safety of motorcyclists must be carefully considered as motorcycles are a popular means of transport [22] – as is the case in many Asian countries [7,34].According to the opinions of the people interviewed – coming from households that have been affected by a severe RTI during the past year -, various things can be done: roads can be better designed and maintained, individual protection legislation (e.g. compulsory safety belt and helmet wearing) could be enforced, people could comply with and adopt safer behaviours, Behvahzes could educate the population and convey information of relevance for injury control and prevention upwards in the health system. Interestingly, many of those suggestions find an echo in the recommendations found in the WHO report on road traffic injuries, in particular concerning making road safety a political priority, enacting and enforcing legislation, managing infrastructure to promote safety for all, and campaigning for greater attention to road safety [4].Injuries at home are also a concern in rural areas just as they are for instance in Pakistan, a neighbouring country, where they form the majority of injuries [9]. In the district studied herein, burns may require special attention, if not because of their frequency, because of their severity (see also earlier Iranian studies [35-38]). Burns also affect children to a greater extent, which is consistent with an earlier study from Bangladesh that showed that the incidence of burns among rural children was more than four times higher than among urban children [39].In the Twiserkan district – and perhaps even in Iran as a whole – gas equipment used for cooking or heating may warrant special attention. Some villages still do not have gas mains and people use gas capsules and/or other heating equipment that is poorly adapted to in-house use. At present, fire stations are far from most villages and people did mention this as a matter of concern.For its part, the prevention of falls is undermined by the vast number of different situations leading up to them, which is a challenge for community-based education programs. Potentially more severe falls from a height, e.g. from a roof and falls from a tree may constitute important targets. The latter occur during work activities, mainly during the walnut harvest. This work is done in a traditional, non-technical manner and every year some people fall from large trees and are injured or killed. The results of one study on safety assessment of agricultural machinery in Iran showed that in 60% of cases agricultural injuries were severe [40]. It ought to be emphasized that an important number of falls affect older people, which has been also observed in an earlier study showing that falls from standing height, falls during walking and falls on stairs were important risk factors for hip fracture for older patients [41]. Fall-related injury prevention may require not only environmental improvements in and around the house but also, in the long run, changes in health behaviours (e.g. eating, smoking, and exercising) so as to reduce individual susceptibility to fall and also recovery after fall.Generally, people frequently mentioned that Behvarzes could play an important role in safety education matters on the local level. Behvarzes already have face-to-face meetings with community members as part of their traditional duties. Also, in recent years, the Ministry of Health and Medical Education has introduced a number of home safety program [20,35,42], and provided Behvarzes with educational packages, which is consistent with international literature.To our knowledge, there is a dearth of studies conducted thus far in Iran or in other rural settings that have collected people's opinions and suggestions about injury control and prevention. This study shows that rural people have a lot of ideas which can be considered for the conception and implementation of context-relevant measures for injury prevention in their community. In particular, people from households where injuries have occurred during the past year consider that not only a change in their own behaviour but also environmental changes and the provision of information and education are needed. We hope that the suggestions highlighted, though not fully representative of the whole rural population, will be taken into account in future developments of safety measures and programs, in both the Twiserkan district and other districts.Because of the routines in place [25] and the relatively small size of the catchment areas, we have good reasons to believe that the study offers an accurate coverage of the severe injuries incurred in the population under study during the study period. It is indeed very likely that health houses do have a complete coverage of injuries leading to hospitalization and death in their community [25]. In spite of the fact that collecting injury data was a relatively new procedure when the cases were identified, we regard the likelihood of missing cases as very unlikely given that the injuries covered are relatively severe, that the Behvarzes are well anchored in their community, and that those communities are relatively small.Before concluding, it ought to be underlined that the study covers one district only and is limited to one year of observation. Because of this, it is not possible to extrapolate our results to any other time period or district. Yet, some results can be regarded as a matter for investigation in other districts as well (e.g., traffic related injuries or burns).ConclusionTraffic injury is an important cause of severe and fatal injury among people from rural areas. Its prevention requires a variety of measures under the responsibility of different actors. Behvarzes may play an important role in both injury surveillance and in identifying context-relevant means of prevention.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsFR-S has made substantial contributions to the conception and design of the study, took responsibility for and coordinated the acquisition of data, which she analyzed. She took part actively in the analysis of the data and in the writing up of the manuscript. LL, MN and MS contributed to the conception and design of the study. LL and MN were closely involved in the data collection process and took active part in the data analysis, result interpretation and manuscript writing. MS contributed to the study design, data acquisition and results interpretation. All authors read and approved the final manuscript.Pre-publication historyThe pre-publication history for this paper can be accessed here:\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2533344.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2533344",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2533344\nAUTHORS: Sarantis Gagos, George Papaioannou, Maria Chiourea, Sophie Merk-Loretti, Charles-Edward Jefford, Panagiota Mikou, Irmgard Irminger-Finger, Anna Liossi, Jean-Louis Blouin, Sophie Dahoun\n\nABSTRACT:\nMalignant melanomas are characterized by increased karyotypic complexity, extended aneuploidy and heteroploidy. We report a melanoma metastasis to the peritoneal cavity with an exceptionally stable, abnormal pseudodiploid karyotype as verified by G-Banding, subtelomeric, centromeric and quantitative Fluorescence in Situ Hybridization (FISH). Interestingly this tumor had no detectable telomerase activity as indicated by the Telomere Repeat Amplification Protocol. Telomeric Flow-FISH and quantitative telomeric FISH on mitotic preparations showed that malignant cells had relatively short telomeres. Microsatellite instability was ruled out by the allelic pattern of two major mononucleotide repeats. Our data suggest that a combination of melanoma specific genomic imbalances were sufficient and enough for this fatal tumor progression, that was not accompanied by genomic instability, telomerase activity, or the engagement of the alternative recombinatorial telomere lengthening pathway.\n\nBODY:\nIntroductionCutaneous malignant melanomas are highly aggressive tumors with unpredictable biological behavior [1]. Metastases in brain, bones and viscera with subsequent ascites development, are frequent [1]. The progression of a transformed melanocyte to malignant melanoma is accompanied by gradual acquisition of multiple genetic alterations that lead to losses of onco-suppressor genes and increased tumor hypermutability [2]. Malignant melanomas display both types of known genomic instability in neoplasia; chromosomal instability (CIN) and microsatellite instability (MIN) [2,3]. MIN has been observed in 30% of cutaneous malignant melanomas [4]. However, the great majority of malignant melanomas examined by various cytogenetic methods, exhibit increased karyotypic complexity, extended aneuploidy and heteroploidy [5-7]. Recurrent chromosomal imbalances in skin melanomas include losses of chromosomes 1p, 6q and 9p [2,8]. Tumor progression and aggressive behavior have been associated with imbalances of chromosomes 7, 10 and 17 [2,7].Most human tumors including melanomas maintain sufficient telomere length for continuous growth by expressing telomerase [9,10], the remainder are thought to utilize a variety of telomere recombination mechanisms termed alternative lengthening of telomeres (ALT) [11]. Observations on transformed and tumor cell lines that lack telomerase, linked ALT phenotype to highly increased structural chromosomal instability and extreme telomeric length deviation ranging from very long to extremely short telomeres [11]. We report a MIN, CIN, and TRAP negative (Telomere Repeat Amplification Protocol), highly aggressive melanoma metastasis to the peritoneal cavity, with unusually stable abnormal pseudodiploid karyotype, and relatively short but not dysfunctional telomeres.MethodsImmunochemistry-CytopathologyCell material from the peritoneal aspirations was subjected to routine diagnostic cytopathology protocols including Giemsa, Papanicolaou, Hematoxylin-Eosin (BDH-Chemicals) stains and immunocytochemistry using the melanocyte specific antibody S-100 (Dako). Immunocytochemical staining against S-100 was performed using Horse Radish Peroxidase (HRP) (Dako). Cell smears were re-hydrated, treated with 3% hydrogen peroxide for 15 minutes and rinsed with Tris Buffered Saline with 0.05% Tween 20 (Dako). After cooling for 20 min, sections were incubated with the primary antibody (rabbit antihuman monoclonal S-100 antibody, 1:400 dilution, Dako) for 1 hour at room temperature and then incubated for 45 min with an anti-mouse HRP labelled polymer (EnVision+System-HRP, Dako). Finally slides were treated with a diaminobenzidine (DAB) chromogenic substrate (Dako) for 10 min, counterstained with hematoxylin, dehydrated and coverslipped.Short term cultures/Cytogenetic analysisMalignant cells from two peritoneal aspirations were collected by centrifugation (10 min/1500 rpm/25°C). They were subsequently cultured in eight 25 cm2 T-flasks at 37°C and 5% CO2, in Dulbecco's Minimum Essential Medium supplemented with 10% fetal bovine serum, 0.08 mg/ml amphotericin, 25 units/ml penicillin, and 25 pg/ml of streptomycin (Invitrogen). When high mitotic index was reached, cells were exposed to colcemid (0.1 μg/ml) (Invitrogen) for 30 min, in 37°C, and harvested using trypsine (Invitrogen), after 0.075 KCL hypotonic treatment and Methanol/Acetic acid (BDH-Chemicals) fixation. For the construction of the representative karyotype, we combined G-Banding after Trypsine and Giemsa (GTG-Banding), inverted 4',6-diamidino-2-phenylindole (DAPI)-banding, subtelomeric FISH (TeloVysion-Vysis) and telomeric FISH (Dako), in a total of 400 metaphases from 8 short term monolayer cell cultures. For dual-color interphase or metaphase FISH we used satellite probes specific for chromosomes 7 and 17 (Cytocell). In brief, our general FISH protocol was based on pepsin pre-treatment, formamide or NaOH target denaturation, over-night hybridization and high stringency post hybridization washes. Telomere-specific Peptide Nucleic acid Analog (PNA) hybridizations were performed using a Cy3-(Indocarbocyanine)- conjugated (CCCTAA)3 probe (Dako), according to manufacturer's instructions. All FISH preparations were mounted and counterstained with VectaShield antifade medium (Vector), containing 0.1 μg/ml DAPI (Sigma). GTG-Banding was performed after trypsine denaturation (Invitrogen) and Giemsa (BDH-Chemicals) staining. Digital images were captured in a Perceptive Systems Imaging, a Metasystems or an Applied Imaging molecular cytogenetics workstations equipped with fluorescent Zeiss, or Nikon microscopes. Quantification of telomeric PNA fluorescence was performed in 500 chromatids on DAPI counterstained metaphase preparations in a single hybridization experiment using the Isis software (Metasystems).Microsatellite instability assayTwo mononucleotide markers, BAT-25 and BAT-26 were tested for microsatellite instability by radioactive PCR after Polyacrylamide Gel Electrophoresis (PCR-PAGE) assay, using the following primers: BAT25.1 (5'-TCGCCTCCAAGAATGTAAGT-3'), BAT25.2 (5'-TCTGCATTTTAACTATGGCTC-3'), BAT26.1 (5'-TGACTACTTTTGACTTCAGCC-3') and BAT26.2 (5'-AACCATTCAACATTTTTAACCC-3'). Experiment was monitored by controls for human microsatellite stability (normal genomic and MIN DNA from a patient with human Hereditary Non-Polyposis Colon Cancer – HNPCC-).TRAP assayTelomerase activity of cell lysates was analyzed by the telomeric repeat amplification protocol (TRAP) assay with a TRAPeze Telomerase Detection kit (Intergen) according to manufacturer's instructions. Approximately 106 cells were harvested and lysed in 400 μl of 1× CHAPS (3-[(3-Cholamidopropyl)dimethylammonio]propanesulfonic acid, 3-[(3-Cholamidopropyl)-dimethylammonio]-1-propanesulfonate) lysis buffer [Tris-HCl 10 mM, pH 7.5; 1 mM EGTA (ethylene glycol tetraacetic acid), 1 mM MgCl2, 0.5% CHAPS 10% glycerol, DEPC (Diethylpyrocarbonate) treated water on ice for 30 min. Cell debris were spun down for 20 minutes at 12,000 r.p.m at 4°C. Each reaction was carried out by using 2 μl of supernatant, 1 μl of each primer, 0.5 μl of Taq-Polymerase (TAKARA), 10 μl of solution-Q (Qiagen), 5 μl of 10× buffer, 2 μl of dNTPs, in DEPC treated water in final volume of 50 μl. The primers used for the TRAP-assay PCR, were TS-5'-AATCCGTCGAGCAGAGTT-3' and Cxa-5'-GTGTAACCCTAACCCTAACCC-3'. The PCR program consisted first of an incubation at 30°C for 30 min and then in a thermocycler, 94°C for 2 min; 94°C for 30 s, 50°C for 25 s, 72°C for 30 s (33×); 72°C for 1 min. PCR products were electrophoresed in a 10% 19:1 acrylamide gel (Sigma)/0.5× TBE (Tris/Borate/EDTA) buffer using the mini protean II gel system (Biorad). Gels were stained with 2 μl of SYBR Green (Sigma) for 15 min at room temperature in 50 ml of TBE 0.5× buffer, and then exposed to UV light and visualized by a Kodak image acquisition station.Flow FISHTo measure cellular telomere length, short term cultured cells were hybridized in situ with a fluorescent telomere-specific peptide nucleic acid probe, according to manufacturer's protocol. Briefly, cells were washed in PBS, and re-suspended to 105 cells/100 μl of a hybridization mixture (Dako) containing 70% formamide and a telomere-specific FITC (Fluorescein isothiocyanate)-conjugated PNA probe. Control samples were re-suspended in hybridization solution without probe to obtain background fluorescence values. After hybridization, cells were spun down and washed twice with 4 ml PBS (Phosphate Buffered Saline) at 40°C for 10 min and finally re-suspended in PBS containing 0.1% Bovine Serum Albumin, 10 μg/ml RNase A (Roche) and 0.1 μg/ml propidium iodide (Calbiochem-Novabiochem). Cells were analyzed on a FACScan flow cytometer (Becton Dickinson) or stored at 4°C before analysis.ResultsPatient history and ascitic fluid samplesPeritoneal fluid samples were obtained by two subsequent paracenteses (within a 12-day interval) of a 38-year-old woman, presented at the Department of Gynecology, Laikon Hospital, with ascites and solid structures at her ovaries as revealed by CT-scan. Two years ago the patient had a less than 1.5 cm large, cutaneous nevus excised from the anterior surface of her left hip. The primary tumor was characterized as a nodular melanoma, Clark's level 3, Breslow's depth 2.0 mm. One out of 14 inguinal nodes, excised in a subsequent operation, was found to be invaded. She received 6 cycles of chemotherapy (cis-platin-dacarbazine) and remained disease-free for 15 months. The cytologic examination of the ascitic aspiration confirmed the presence of malignant cells positive for the melanocyte specific antibody S-100 (Figure. 1). The patient refused to be operated, gave her written consent for further research on the specimens obtained, and expired 40 days after presentation.Figure 1The cytologic examination of the ascitic fluid showed malignant cells with high mitotic index (Giemsa × 400) (A). Immunocytochemistry against the melanocyte specific antibody S-100 confirmed the presence of malignant melanocytes (Hematoxylin and DAB × 400).Cytogenetic analysisG-Banding analysis (according to ISCN 1995) [12] from 8 short-term cell cultures of two peritoneal aspirations taken in an interval of 12 days, showed a 46,XX,del(6)(q23?qter),del(9)(p10pter),der(10)t(7;10)(q31.3qter::p13)del(10)(p14?pter),der(11)t(5;11)(q22.3qter;q23)del(11)(q24?qter),i(17q) pseudodiploid karyotype, in 94–96% of 200 mitoses examined (Figure. 2A). Endoreduplication was observed in 4–6% of the malignant cells leading to a 92,XXXX,idemx2 karyotype. Subtelomeric FISH specific for all human telomeres except for chromosomes 16, 19, 20 and the short arms of acrocentric chromosomes, was used to assist in the description of marker chromosomes identified by G-Banding (Figure. 2A), and to verify deletions spanning up to the end of rearranged chromosomes. To examine if this remarkable karyotypic stability was not confined only to dividing mitotic cells, we performed dual color interphase FISH with probes specific for centromeres 7 and 17, in 200 interphase nuclei obtained from 2 short-term cell cultures from both aspirations. Centromeres 7 and 17 showed notable numerical stability in these populations. The rates of whole genome endoreduplication were similar to those of the karyotyped mitotic cells (Figure. 2B).Figure 2A GTG-Banding and sub-telomere specific FISH composite representative karyotype of the reported melanoma. Subtelomeric FISH verified structural integrity of most chromosomes, canonical orientation of both translocations, the deletions 6q, 10p, and 11q, as well as the isochromosome i(17q). The depicted partial dual or triple color subtelomeric FISH karyotypes derive from 23 independent pseudodiploid metaphases; each black box represents a single mitotic nucleus (Red = Spectrum Orange, Green = FITC, Purple = Spectrum Aqua ×1000) (A). Dual color interphase FISH for centromeres 7 (yellow), and 17 (green), shows remarkable numerical stability in 200 nuclei (error-bars represent the standard error of the mean) (B).Examination of factors related to chromosome stabilityIn an attempt to attribute the karyotypic stability of this metastatic melanoma to measurable parameters related to chromosome stability in the context of neoplastic continuous growth, we examined microsatellite instability (MIN) and telomerase activity. Microsatellite unstable tumors show a significantly lower rate of chromosomal instability as compared to the MIN negative [3]. To rule out underlying microsatellite genomic instability in this metastatic melanoma, we tested by PCR-PAGE the robust mononucleotide repeat markers BAT-25 and BAT-26. Both loci have been shown to be sensitive markers of MIN [13]. Compared to positive and negative controls, this metastatic melanoma displayed no micro-satellite instability (Figure. 3A). Ectopic expression of telomerase in normal fibroblasts has been connected to karyotypic stability [14]. We conducted a TRAP assay to test telomerase activity in cultured cells from 2 sub-cultures from both peritoneal aspirations. In both samples this assay was negative (Figure. 3B). To examine if these melanoma cells followed the ALT-pathway of telomere maintenance [11] we compared the relative telomeric length of our specimen by Quantitative-PNA-Flow-FISH [15] with the pseudodiploid human acute T cell leukemia JURKAT cell line, normal human fibroblasts and an ALT-positive cell line [16]. Cell material for this test was obtained from a short-term subculture that was previously karyotyped and found to be composed exclusively from chromosomally abnormal mitotic cells. This comparison revealed that the melanoma cells had relatively short telomeres (Figure. 3C). PNA-telomeric FISH on 500 chromatids from 10 randomly picked metaphase spreads showed that most of the 46 chromosomes of this metastatic melanoma were uniformly capped with telomeric repeats (Figure. 3D) and no signs of structural chromosome instability attributed to telomere dysfunction such as end-to-end fusions and dicentric chromosomes were evident.Figure 3Microsatellite instability in neoplasia (MIN) was excluded in this tumor since the microsatellite markers BAT-25 and BAT-26, showed no instability as compared to MIN positive and negative controls (A). The TRAP assay was negative for telomerase activity in cell culture material obtained from both peritoneal aspirations as compared to two well known telomerase positive human cancer cell lines (MCF-7 and HeLa) (B). Telomere length in this melanoma is relatively low as compared to the ALT U2-OS cell line, leukemic JURKAT cells and human embryonic fibroblasts (error bars represent standard error of the mean between 3 independent experiments) (C). Telomeric PNA FISH indicated uniform terminal capping with TTAGGG repeats on virtually all chromosomes in both pseudodiploid and endoreduplicated clones and low deviation of telomeric length in 500 chromatids as compared to the ALT U2-OS cell line (inverted DAPIx1000) (error bars represent standard deviation)(D).DiscussionMetastatic transition in most human tumors is accompanied by a series of complex recurrent and stochastic chromosomal anomalies. These changes reflect the evolutionary pressure held by the cancer cells to bypass natural barriers and re-establish continuous growth into unrelated histopathologic environments [17,18]. In this report, the karyotype of the primary tumor is not available, therefore the relative simplicity of genomic imbalances encountered in metastasis, permits only a hypothetical reconstruction of the chromosomal evolution of the disease. It has been proposed that melanomas develop through a mode of karyotypic evolution, common to both low and high complexity karyotypes [2]. To become malignant, an apparently normal melanocyte of this patient underwent multiple karyotypic alterations involving breakpoints in at least 7 different chromosomes as well as chromosomal losses and non-disjunctions. Although we cannot define the temporal order of the recorded rearrangements, we postulate that the hemizygous deletions 6q23qter and 9p- might be early events in the chromosomal evolution of this melanoma. Translocations and deletions involving the q-arm of human chromosome 6 have been found in more than 80% of melanomas [5]. According to Hoglund et al (2005) [19], deletions of the distal 6q should be considered early chromosomal lesions in melanomas. Moreover, the short arm of chromosome 9 is the site of several cell cycle regulators that have been linked with familial disease, or associated to melanoma progression and aggressive behavior [2]. The gains of genomic material and the additional deletions involving 10p, 11q and 17p, were by-products of unbalanced chromatid separation of balanced translocations and the isochromosome formation. These more complex alterations might represent later events in the process of the karyotypic evolution of the disease. Chromosomes 10 and 11 are frequently lost in metastatic melanomas whereas chromosome 7 is frequently gained [2,5,7,20,21].Rearrangements affecting the short arm of chromosome 17, where the p53 gene is located, have been implicated in the pathogenesis of malignant melanoma [2]. It is interesting that although p53 deficiency has been related to increased rates of numerical chromosome instability or polyploidy [22], in this melanoma hemizygosity of p53 was not associated with continuous genomic instability. MIN tumors display extremely low rates of CIN [3]. We ruled-out the possibility that this melanoma belonged to this type of tumors. We also ruled-out CIN in our specimens, since this metastatic pseudodiploid tumor was highly cytogenetically stable by all means examined. These results are compatible with those of Abdel-Rahman et al. 2001 and Fabarius et al. 2003, who observed that chromosomes of near-diploid cells are structurally much more stable than those of highly aneuploid counterparts [23,24]. Perhaps, the rare, melanoma described here, is unusually stable, because it is near-diploid, in contrast to the majority of highly aneuploid genomically unstable melanomas.The majority of human malignant melanomas and melanoma cell lines studied with the TRAP assay were found to express telomerase activity [10,25]. Furthermore, telomerase activity has been connected to aggressiveness of melanomas [26]. In continuous neoplastic growth, insufficiently protected telomeres tend to undergo end-to-end fusions and to produce numerous complex chromosome rearrangements such as dicentric chromosomes and inverted duplications [27-29]. No evidence of such lesions was found in our specimens. The transient stage of structural chromosomal instability in this case, equally involved subtelomeric, centromeric and genomic regions, and gave rise to translocations with canonical orientation. Surprisingly, this metastatic tumor was negative for telomerase activity. Moreover, no signs of recombinatorial telomere elongation were present [11] since flow FISH showed relatively short telomeres and PNA FISH displayed a uniform terminal capping of virtually all chromosomes of this melanoma with TTAGGG repeats.The remarkable stability, and telomeric integrity of the metastatic tumor presented here, can be attributed either to transient telomerase activation, or the action of an unknown but efficient telomere restoration mechanism. However, we can not exclude the possibility that adequate telomeric length for clonal expansion and metastasis was already acquired by the cancer progenitor melanocyte. This assumption might correlate with the relatively young age of the patient. A thorough examination of a series of human osteosarcomas revealed a category of tumors that do not express telomerase activity and do not display any ALT-pathway characteristics [30]. Interestingly these tumors showed low rates of CIN [30]. A similar sub-category might be also encountered in melanomas. The exceptional case reported here, suggests that metastatic progression in this melanoma, was not accompanied by genomic instability, telomerase activity, or the engagement of the classical alternative recombinatorial telomere lengthening (ALT) pathway.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsSG conceived and coordinated the study, carried out the analysis of the results and wrote the manuscript. GP collected the samples, acquired informed consent and took part in the analysis of results and manuscript preparation. MC carried out conventional and FISH cytogenetics. SM-L carried out subtelomeric FISH. C-EJ performed the TRAP and Flow-FISH assays. PM and AL carried out and analyzed the cytopathology assays. II-F participated in the design of the study. J-LB performed and analyzed the microsatellite instability assays. SD participated in coordination of the study and analysis of results. All authors read and approved the final manuscript.\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2533349.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2533349",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2533349\nAUTHORS: Raphael Genolet, Tanguy Araud, Laetitia Maillard, Pascale Jaquier-Gubler, Joseph Curran\n\nABSTRACT:\nBackgroundRecent work, using both cell culture model systems and tumour derived cell lines, suggests that the differential recruitment into polysomes of mRNA populations may be sufficient to initiate and maintain tumour formation. Consequently, a major effort is underway to use high density microarray profiles to establish molecular fingerprints for cells exposed to defined drug regimes. The aim of these pharmacogenomic approaches is to provide new information on how drugs can impact on the translational read-out within a defined cellular background.MethodsWe describe an approach that permits the analysis of de-novo mRNA-ribosome association in-vivo during short drug exposures. It combines hypertonic shock, polysome fractionation and high-throughput analysis to provide a molecular phenotype of translationally responsive transcripts. Compared to previous translational profiling studies, the procedure offers increased specificity due to the elimination of the drugs secondary effects (e.g. on the transcriptional read-out). For this pilot \"proof-of-principle\" assay we selected the drug rapamycin because of its extensively studied impact on translation initiation.ResultsHigh throughput analysis on both the light and heavy polysomal fractions has identified mRNAs whose re-recruitment onto free ribosomes responded to short exposure to the drug rapamycin. The results of the microarray have been confirmed using real-time RT-PCR. The selective down-regulation of TOP transcripts is also consistent with previous translational profiling studies using this drug.ConclusionThe technical advance outlined in this manuscript offers the possibility of new insights into mRNA features that impact on translation initiation and provides a molecular fingerprint for transcript-ribosome association in any cell type and in the presence of a range of drugs of interest. Such molecular phenotypes defined pre-clinically may ultimately impact on the evaluation of a particular drug in a living cell.\n\nBODY:\nBackgroundDissecting the specific effect of a drug on a defined biological process is often complicated by the plethora of secondary effects that arise during extended exposure. This is highlighted by the anti-cancer drug rapamycin whose principle, although not exclusive, target is the rate limiting initiation step of protein translation [1]. Techniques designed to analyse the effect of the drug on mRNA-ribosome association are hampered by the long exposure times required to observe changes in the polysomal mRNA populations. Translation initiation is frequently regulated via the eukaryotic initiation factor 4E (eIF4E). It can be sequestered into an inactive complex by a family of 4E-binding proteins (4E-BP1/2/3). The affinity of these proteins for eIF4E is modulated by phosphorylation via the mTORC1 (mammalian target of rapamycin complex 1) kinase [2]. Rapamycin arrests many cells in G1 [3], and is a potent immunosuppressant [4]. It exerts its action by binding and inactivating the mTORC1 [5]. Previous studies have shown that extended exposure to rapamycin also alters transcription. Among the mRNAs regulated, a number impact directly on translation (e.g. eIF2α) [6]. These changes were minor after short drug exposure times (< 60 mins), but increased markedly after 2 hrs. Translational profiling studies (which examine the mRNAs associated with ribosomes) have also been reported [7-10]. One study, performed on the Jurkat T cell clone E6-1, revealed that after an extended exposure to rapamycin (minimum 4 hrs) almost all transcripts analysed were inhibited, and 136 of those (representing ~5% of the total) were strongly inhibited (at least 10 fold). This latter group included a number of mRNAs whose products impact directly on translation (e.g. eIF5A, eIF4A1, eEF1, eEFTu) [8]. This would predict that lengthy exposure to the drug will influence translation by modifying the levels of initiation and elongation factors. We therefore sought to develop a technique that would permit a specific analysis on how the drug alters transcript recruitment onto free ribosomes under conditions that eliminated the secondary effects associated with extended exposure. The approach has been coupled to a high-throughput microarray screen to examine how rapamycin exposure impacts on the re-seeding of the polysomal transcript populations. Results from the array have also been validated by quantitative RT-PCR.MethodsCell cultureMRC-5 cells (Coriell Cell Repository) were cultured in Minimal Essential Medium (Gibco) supplemented with 1 mM sodium pyruvate (Sigma), 0.1 mM non-essential amino acids, 10% foetal calf serum (Brunschwig), 1% penicillin/streptomycin, in a humidified atmosphere containing 5% CO2. For polysome analysis, cells in the growing phase (60% confluence) were hypertonically shocked by shifting to medium containing 300 mM NaCl for 50 min. They were then placed in normal isotonic medium for 30 min. When rapamycin was used, 100 nM rapamycin (LC laboratories) or 0.01% DMSO (the negative control) was added during the hypertonic shock, 20 min before the transfer back to isotonic conditions. Rapamycin and DMSO were kept on the cells throughout the 30 mins recovery period (total time of exposure to rapamycin was 50 mins). These conditions were based upon previously published work [11], although we have independently confirmed that they can be used on a range of cell lines including 293T [12], HeLa S3 and SK-NA5 (data not shown).Polysome gradient/RNA extractionAfter treatment, cells were scraped into the culture medium and pelleted for 4 min at 100 g. The pellets, consisting of 5 × 106 cells, were lysed for 15 min on ice in 400 μL of 100 mM KCl, 50 mM Tris-Cl pH 7.4, 1.5 mM MgCl2, 1 mM DTT, 1 mg/mL heparin, 1.5 % NP40, 100 μM cycloheximide, 1% aprotinin, 1 mM AEBSF and 100 U/mL of RNasin. Nuclei were pelleted by centrifugation in a microfuge, 10 min at 12000 rpm. The supernatant was loaded onto a 20–60% sucrose gradient (in 100 mM KCL, 5 mM MgCl2, 20 mM HEPES pH 7.4 and 2 mM DTT). Extracts were fractionated for 3 h 30 min at 35,000 rpm at 4°C in a Beckman SW41 rotor, and the gradients were recovered in 3 fractions [monosome, light polysome (2 to 5 ribosomes) and heavy polysome (> 5 ribosomes)] using a Brandel gradient fractionator equipped with an ISCO UA-6 flow cell set to 254 nm. RNA was isolated from the light and heavy polysome fractions by adding an equal volume of TriZol (Invitrogen). Samples were mixed and incubate for 15 min. on ice, then 0.3 volumes of chloroform was added. After centrifugation, the upper phase was collected and the RNA precipitated with 0.7 volumes of isopropanol. The pellet of RNA was re-suspended in water. Prior to microarray analysis the pooled RNA fractions were further purified using the Qiagen RNeasy kit. The total yield of RNA in each pooled fraction was ~2 μg. RNA quality was checked on an Agilent 2100 bioanalyser.MicroarrayTotal RNA (100 ng from each fraction) was first amplified using the two step amplification protocol of Affymetrix. cRNA (17.5 μg) was then used to probe the Gene Chip U133 Plus 2.0 with 54,675 probe sets, covering more than 47,000 transcripts. Three biological replicates were hybridised for each condition (light and heavy polysomes +/- rapamycin). Data were analysed using the GCOS normalisation of Affymetrix. After normalisation, we filtered out probe sets that were assayed \"absent\" in all the 12 arrays. For this, the expression levels in each control experiment (DMSO) were arbitrarily fixed to one, and the fold change of the corresponding probe in the treated samples was normalised to this. The variation was then tested using the Mann and Witney U statistical test. Using this approach 24,105 of the probe sets (44%) were flagged as absent. The data was further analysed using two approaches. Firstly, probe set intensity values below 100 were removed prior to GCOS normalisation, and fold changes ≥ 1.5 relative to the DMSO control were scored (the smallest score in the three independent experiments had to show at least a 20% change relative to the DMSO control i.e. a 1.2 fold increase or a 0.8 fold decrease). In a second approach, commencing with the entire data set, points were only scored if the mean of the fold difference was ≥ 2.5 (once again the smallest score in the three independent experiments had to show at least a 20% change relative to the DMSO control). The data from both screens were plotted onto biological networks using the GO onthology (Affymetrix) and Ingenuity Pathway Analysis Software packages .The microarray data are available at ArrayExpress (Accession N° E-TABM-205).Real-Time PCROne μg of total RNA from a fourth independent experiment was reverse transcribed using random hexamers (Gibco). A 1/10 dilution of the cDNA was used to perform the PCR with the SYBR Green Reagent (Roche).Primers used were:QKI AGCATCACAGTCAGAGGTCAGC, GCAGTGGCATATTAAACCAAAGC;RBM7 GTTGGAAATTCAAGCCCTACCT, AATCCTGATTGATCCAGAGGTG;ORMDL1 GTCTGGCAGAAACAACGTCTC, CAATGTGGTTGCTGTTCTGG;FAS GATGGCGAATGAGGTTCAG, CAATCCCATATCTCCCATTAAC;RBM17 GTCATCTCCGGTGATCCTTAAA, CAACCAGAGAGGCACACAGAT;PAPPA GCATCAGTTTCTCTAGCTGCAA, TATCAAACAAGCACTCCCTGTC;Actin CTGACGGCCAGGTCATCACCATTG, GCCGGACTCGTCATACTCCTGCTTG;L27 GTGACAGCTGCCATGGGCAAG, TCAAACTTGACCTTGGCCTCCCG,Cyclin D1 AAGCAGGACTTTGAGGCA AG, CCTCTGAGGTCCCTACTTTCAA.Primer sets were designed to amplify regions within the 3' UTR of each transcript since this generally corresponded to the site of the probe sets used on the Affymetrix chip. The specificity of each primer set was confirmed by standard RT-PCR on total cell RNA, followed by analysis of the DNA products by agarose gel electrophoresis (data not shown).Western blot analysisCells were lysed in CSH buffer (50 mM Tris-Cl pH 7.5, 250 mM NaCl, 1 mM EDTA, 0.1% Triton X-100) and the nuclei were pelleted by centrifugation at 12,000 rpm for 5 mins. Twenty μgs of protein was resolved on a 15% polyacrylamide-SDS gel and electrotransfered to a PVDF membrane. Antibodies used in this study were the anti-4EBP1 (Cell Signalling), the anti-phosho4EBPI (Thr37/46) (Cell Signalling), the anti-p70 S6 kinase (Cell Signalling), the anti-phospho-p70 S6 kinase (Thr389) (Cell Signalling) and mouse anti-actin (Chemicon). Blots were developed using the Super Signal Substrate (Thermo Scientific).ResultsRapamycin delays mRNA recruitment onto polysomesTo directly examine the effect of rapamycin on mRNA recruitment we decided to exploit a novel approach, an approach that analyses the ability of cellular mRNAs to compete for free ribosomes in-vivo. Hypertonic shock provokes a rapid inhibition of protein synthesis, disaggregation of polysomes (Figure 1A), dephosphorylation of eIF4E, 4E-BP1, S6 and an increased association of eIF4E and 4E-BP1 [11,13]. Upon restoration of isotonic conditions the polysomal fraction is rapidly reconstituted (Figure 1B and [11]). Using this methodology it was possible to examine what effect rapamycin had on the recruitment of mRNA populations onto free ribosomes following very short drug exposure times. It was in substance an in-vivo competition assay performed under two defined physiological conditions. Drug treatment appeared to delay recruitment as evidenced by the reduction in the heavy polysome peak (≥ 6 ribosomes: compare Figure 1B and 1C), and this effect was correlated with a modification of the downstream signalling targets of mTOR, including 4E-BP1 and S6 kinase. (Figure 1D, compare the second and third lanes). This confirmed that despite the relatively short time of exposure to rapamycin, the recruitment assay monitored transcript:ribosome re-association under two conditions in which eIF4E availability was altered.Figure 1Ribosomal re-recruitment in the presence of rapamycin. (A). High salt provokes a rapid disaggregation of polysomes. (B). Upon restoration of isotonic conditions the polysomal fraction is reconstituted (C). Pre-treatment with rapamycin delays the re-recruitment of ribosomes. The position of the ribonucleoprotein (RNP), the monosomal (Mono) and polysomal (Poly) fractions are indicated. (D) Western blot analysis of phospho-4EBP-1, 4EBP-1, phospho-S6K, S6K, and actin was performed on extracts isolated under the different conditions depicted in panels (A); (B) and (C).A profiling screen identifies changes in the light and heavy polysomal mRNA populationsEqual amounts of RNA isolated from the light (2 to 5 ribosomes) and heavy (> 5 ribosomes) polysomal fractions were used to probe the Affymetrix Gene Chip U133 Plus 2. Triplicate independent gradients under each experimental condition were examined. After data analysis, two subpopulations of transcripts were clearly discriminated: those dominant in the light polysomal fraction and those dominant in the heavy polysomal fraction, an important criterion since it validated the initial experimental approach (Figure 2A). Despite the fact that the polysomal peaks were smaller in the presence of rapamycin, consistent with a global repression, the two sub-populations were essentially conserved (i.e. no major movement of mRNA populations between the two fractions as a consequence of the treatment was evident). To analyse the data we used two approaches. Firstly, we filtered out all transcripts giving low probe set intensity values on the chip using the default settings of the Agilent analysis software package (values < 100). We then scored for mRNAs whose polysomal occupancy was altered by greater than ×1.5 fold (listed in Additional File 1). This produced 437 transcripts within which was found the majority of the repressed TOP mRNAs (see below). Curiously, within this group of transcripts almost equal proportions were up and down-regulated (46% and 54%, respectively) (Figure 2B).Figure 2Microarray analysis. (A). Hierarchical clustering of relative expression. Each column represents the different conditions. On the right the vertical bar indicates probe set intensity values (indicated as Expression) in arbitrary units. The zero value indicates absence, with blue indicating a low level and red a high level of expression (the maximum value being fixed as 5). The vertical brackets on the right indicate that transcripts have been grouped into those over-represented in the light polysomes (lower) or in the heavy polysomes (upper). (B). After removal of probe set intensity values < 100, and the application of a ×1.5 fold cut-off, the regulated genes were classified into four groups. The values indicate the number of transcripts in each group. (C). In a second approach, a ×2.5 fold cut-off was applied. (D) and (E). Functional classification using Gene Ontology of the genes either up-regulated or down-regulated (as depicted in panel b and in Additional File 1, and panel c and in Additional File 2, respectively). The unknown fraction represents genes not annotated in the Gene Ontology database.As a second approach to analyse the data we applied a ×2.5 fold change cut-off point to the entire data set. The rationale for this alternative analysis is based upon the fact that many of the genes that are regulated at the level of translation are frequently transcribed at low levels. This includes proto-oncogenes and other factors that regulate cell growth [14]. The majority of these transcripts are found within the lower intensity range and we therefore tested if meaningful information could be extracted from this region by applying a more stringent selection. We observed that 1160 mRNAs (3.8%) showed increased or decreased polysomal distribution in the presence of rapamycin, suggesting that in this small fraction of transcripts the affinity for the cap binding complex was changed (Figure 2C and Additional File 2). Over 2/3 of the mRNAs responding to the drug were down-regulated, whereas 1/3 showed increased polysomal occupancy relevant to the non-treated control. Only a few transcripts were regulated in both the heavy and the light polysomes (55 mRNAs) (Figure 2C). These results are not unlike those reported in a translational profiling study performed on the two tumoural cell lines, LAPC-4 (prostrate cancer) and U87 (glioblastoma). Applying the 2.5 fold cut-off point, ca. 6% of the 3,000 transcripts screened showed altered polysomal occupancy, and amongst these, 60% were down-regulated and 40% up-regulated [7].Those mRNAs showing significant redistributions (both increased and decreased: as listed in Additional Files 1 and 2) were then plotted onto cellular networks using the GO onthology (Affymetrix) software package. Results revealed that transcripts up- or down-regulated affected more or less the same biological processes (Figure 2D and 2E). However, using the Ingenuity pathway analysis a number of features were immediately evident in the rapamycin treated cells: (a) A group of mRNAs involved in the inflammatory response were regulated, consistent with the immunosuppressive activity of rapamycin. Interestingly, we also observed down-regulation of several transcripts linked to phagocytosis. It has recently been reported that rapamycin down-regulates phagocytosis in a murine macrophage cell line (Table 1) [15]. (b) We observed a number of mRNAs involved in cell growth and proliferation. Among them, a number of anti-apoptotic mRNAs were up-regulated (e.g. relA and mdm2), whereas pro-apoptotic ones were down-regulated (e.g. fas, faslg and faf1) (Table 1). This is consistent with the observation that rapamycin did not induce apoptosis in MRC-5 cells even after extended exposure (unpublished observations), and corroborated recent studies showing the anti-apoptotic properties of rapamycin [16,17].Table 1List of transcripts regulated by rapamycin (Ingenuity classification)InflammationUpregulatedDownregulatedapobec3flst1adra1aepha4il1rnptpn22tlr7blnkmdm2adra2cf9irf2ptprczeb1cald1oas1alppfasitgb3ptprz1cmklr1parvgbcl11afaslgklf2rbl2ctla4pdgfcbcl2l1fcgr2alama3rbm15cxcr4rag2castfolr1lilra2rbpjcyp3a4relacd28fybmllrelfynsat1cd36galmposatb1hla-gtcf12clec1bgap43pcgf2siglec8ifih1tfap4cul4agnrh1pla2g6smpd1ifna2thbs1cxcl11gnrhrplcg1spnil16tlr7cyp3a4hal-dqb1prdm16sykil28bunc119ddl1hckpscdbpthraitga4vtcn1dok2ifne1ptgs2thrbCell growth and proliferationUpregulatedDownregulatedadam12mdm2adra1adok2gnrh1olig2rfflblnkpappabcl2l1f12hckp53aip1sec14l2cdc2l5rag2ccl27fashmga2pcgf2sykctla4relacd28faslgirf2pdgfatfr2cxcr4s100bcdca7fbxo2itgb3piwil1tgif1erbb3ss18clca2fcer2klf2pla2g6thrafyntcf12cltcfgf18lzts1ptgs2thrbhla-gthbs1csh2folr1mdm4ptk2tnfsf15igf1runc119cyp2c9foxo1mllptpn22zfn10il28bvtcn1dccgalnab2ptpraitga4ddx17gap43nos1ptprclst1dll1glmnnovrbl2Cell deathUpregulatedDownregulatedblnkitga2abcd2dll1hmga2p53aip1rbm17cdk6mdm2acvr1berafifne1pigtrelctla4rag2adora2afaf1il1rnpiwil1satb1cul3relaatrxfasirf2pla2g6serpinb4cxcr4rnase1atxn3faslgitgb3pou4f1siglec8cyp2e1sgpp1bcl2l1fcer2klf2ppp1r9bsmpd1cyp3a4tfap4bircabpfoxo1klra1ppp2r1bspnerbb3thbs1castgallrp5prdm2stk4ergtraf4ccl27gimap5mdm4ptgs2sykfynzmym2cd28gng2mllptk2thrahla-gcdk6gria2mpoptpn22tnfsf15ifih1cyp2e1grik2nol3ptprctraf5ifna2cyp3a4grm1nos1ptprz1trps1igf1rdcchint1nrtnrbl2znf10PhagocytosisUpregulatedDownregulatedcd36klf2fasmpofcgr2apla2g6hcksykitgb3The 1160 mRNAs identified in the microarray screen (as listed in Additional File 2) were plotted onto biological networks using the Ingenuity Pathway Analysis Software package. Those whose protein products are involved in inflammation, phagocytosis, cell growth/proliferation and cell death are listed.Rapamycin is known to have a marked effect on the expression of a subset of transcripts referred to as terminal oligopyrimidine (TOP) mRNAs [18]. This includes ribosomal protein mRNAs (estimated to be ~15% of the total cellular mRNA) and translation elongation factors. In previous translational profiling studies, a number of TOP mRNAs were clearly repressed. We also observed repression of ribosomal transcripts, although the effects were less extensive than in the earlier reports. This may reflect both the shorter drug exposure times and the experimental approach that was employed (i.e. mRNA re-recruitment onto free ribosomes). With the cut-off threshold at ×2.5 fold we observed down regulation of rpl14 and rpl21 in the light polysomes and rplp0 in the heavy polysomes (ribosomal protein large). However, if the threshold was reduced to ×1.5 fold the number of hits significantly increased. Most significantly, ribosomal transcripts were only ever observed down-regulated (Table 2 and Additional File 1).Table 2List of TOP mRNAs detected in the array.LIGHT POLYSOME UPLIGHT POLYSOME DOWNHEAVY POLYSOME DOWN(NONE)rpl5 (-1.5)rplP0 (-2.7)rpl14 (-2.5)rpl36 (-1.5)rpl21 (-5.1)eef2 (-1.5)HEAVY POLYSOME UPrpl38 (-1.5)(translation elongation factor)(NONE)rps11 (-2.1)rps19 (-1.5)rps21 (-2.1)rps28 (-1.6)eftuD1 (-5.5)(translation elongation factor)Independent confirmation of the array results by RT-PCRWith the aim of validating the microarray we performed real-time RT-PCR on selected transcripts across the spectrum of probe set intensities (Figure 3A). We were particularly interested in those transcripts that gave low probe set values. For the RT-PCR, RNA was extracted from a fourth independent experiment. Nine mRNAs were initially selected (Transcripts up-regulated = qki, rbm7, ormdl1. Transcripts down-regulated = fas, rbm17, pappa, Transcripts not regulated = β-actin, L27, cyclin D1: see Additional Files 1 and 2). Transcripts from the lower values of the data set (see Additional File 2 and Figure 3A) were selected because of the large fold difference between the rapamycin and DMSO control (> 2.5 fold). The results, with the exception of those obtained with the pappa transcript, largely confirmed the micro-array data (Figure 3B). This was particularly encouraging for transcripts such as those coding for QKI and FAS, indicating that application of the ×2.5 fold cut-off approach permitted the extraction of useful hits even in the lower end of the probe intensity set (Figure 3A). Note that the absence of a light RT-PCR value for RBM17 simply reflected its low levels in this fraction (Ct > 35 cycles). The microarray study also indicated that this transcript was regulated only within the heavy polysomal fraction, a result confirmed by the RT-PCR. In addition, these studies also demonstrated that no significant changes in total mRNA levels had occurred within the selected transcripts as a consequence of drug exposure (Figure 3B).Figure 3Real-Time RT-PCR analysis. (A). A schematic representation of the intensity values of transcripts selected for the RT-PCR validation. In the upper panel, the horizontal bars represented the value range for each mRNA on the array. The lower panel plots the distribution of regulated genes (×2.5 fold cut-off selection) relative to the probe set values. (B). RT-PCR values were normalised to those obtained from two housekeeping genes and the fold change indicates the difference in the DMSO and rapamycin values after normalisation. The DMSO value was arbitrarily set at 1. The results are compared with those obtained from the microarray. The variation in the total mRNA extracted is also represented. The 2.5 fold difference used for the screening is represented by the dotted lines (2.5 and 0.4). Each bar is representative of 2 independent RT-PCR assays performed in triplicate. Bars indicate the SEM.DiscussionIn this technical report, we have outlined a novel approach to translational profiling that follows the impact of a drug on the de-novo re-association of mRNAs and ribosomes in living cells in culture. An intracellular pool of free ribosomes was generated by a short hypertonic shock. Although this undoubtedly induced a stress response, previous work has demonstrated that the translation initiation machinery recovers very rapidly after the cells are transferred back to isotonic conditions as monitored by the reconstitution of the polysomal fraction [11,13]. The rapidity at which the polysomes are reformed has permitted us to examine the effect of a drug on this process using a short time window of exposure. This has the effect of limiting undesirable secondary effects that arise upon extended exposure, an effect highlighted by the drug selected for this study, namely rapamycin (see Introduction). The short exposure time warranted a drug concentration higher than that generally used in animal studies [19], however, some studies in cell culture systems have employed rapamycin concentrations as high as 15–20 μM before observing an impact on cell growth [20]. The selection was also dictated too by the fact that rapamycins' effect on translation initiation has been extensively studied. We have demonstrated that this approach can be coupled to a high-throughput analysis of the polysomal transcript populations.Treatment with rapamycin limits the availability of eIF4E via its sequestration into an inactive complex with the hypophosphorylated 4E-BPs. Such a scenario represses global translation rates but with an effect more marked on those mRNAs containing structured 5' UTRs and those containing TOP elements. Indeed, when TOP containing transcripts were identified in our array they were always down-regulated (Table 2). Additionally, although the polysome profiles demonstrated an overall translational repression, the position of the vast majority of mRNAs on the gradient (i.e. the transcript populations in both the heavy and light polysomes) was largely unperturbed by the drug (i.e. there was little movement of transcripts from heavy to light polysomes, a somewhat unanticipated response), indicating that the affinity of re-recruitment of ribosomes onto these mRNAs (as reflected by the number of ribosomes per transcript) was largely unchanged.A number of translational profiling studies examining the effect of rapamycin in mammalian cells have already been reported. In one of these, the effect of rapamycin on the polysomal distribution of mRNAs was demonstrated to be coupled to the activity of AKT [7], a result that demonstrates the extent to which a drugs effect can be modulated by the physiological status of the cell. This interpretation has become even more convoluted following the observation that prolonged rapamycin treatment may inhibit AKT signalling by interfering with the assembly of the second mTOR complex, mTORC2 [21]. However, with regards to the two transcripts characterised in this work as strongly up-regulated, namely cyclin D1 and c-myc, the former was down-regulated in our screen (×1.5 fold), and the latter gave values that were not considered statistically significant. Other studies have also failed to observe changes in the polysomal occupancy of the cyclin D1 mRNA in the presence of rapamycin [9,10,22]. These differences may reflect the cell lines and/or the experimental procedures employed. Both cyclin D1 and c-myc were proposed to carry IRESes within the 5' UTR, and IRES activity has been reported to show cell type specificity linked to the availability of ITAFs (IRES Trans-Acting Factors) [23]. Furthermore, the responsiveness of these IRESes to rapamycin was shown to be tightly coupled to the cellular activity of the AKT and RAF/MEK/ERK signalling cascades, features that may also show cell-type variation [24]. However, the slight reduction that we observed in the polysomal levels of the cyclin D1 mRNA would be consistent with other reports indicating that its expression was sensitive to the levels of eIF4E [25,26]. Finally, the earlier profiling study followed changes in the steady-state polysomal populations after extended exposure to the drug, whilst in the current work we have followed a competitive re-association. These processes may have altered initiation factor requirements, which could impact on the mRNA populations that respond. Indeed, it has been proposed that eIF4GII but not eIF4GI is required for re-initiation subsequent to a hypertonic shock [27]. Nonetheless, a listing of transcripts detected in both studies demonstrated that the majority behaved similarly (see Additional File 3).ConclusionIn summary, a major effort is underway to use high density microarray profiles to study how different drug regimes impact on the polysomal mRNA populations. These studies provide insights into how cellular gene expression is regulated at the level of translation initiation, the rate limiting step in protein expression. Changes in this read-out are a very rapid cellular response to physiological perturbations. The method that we have outlined permits a specific analysis of how a drug impacts on transcript-ribosome association. This early response almost certainly conditions subsequent cell behaviour during extended exposure. The choice of rapamycin for this \"proof-of-principle\" work was not arbitrary since the impact of this drug on translation initiation has been extensively studied. The technique offers the possibility of establishing molecular fingerprints for different tumour derived cell types and drug regimes [28]. In addition, it provides a very powerful technique to analyse the early events in translational control at the level of mRNA:ribosome association.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsRG and PJ-G prepared the polysomal RNA for the microarray analysis. RG and TA were involved in the analysis of the microarray data. RG and LM performed the RT-PCR control. JC prepared the manuscript. All authors have read and approved the manuscript.Pre-publication historyThe pre-publication history for this paper can be accessed here:Supplementary MaterialAdditional File 1List of transcripts from filtered data set. Transcripts with probe set values < 100 were removed from the data set. After GCOS normalisation a ×1.5 fold selection cut-off was applied to those that remained.Click here for fileAdditional File 2Complete list of polysomal transcripts regulated by rapamycin. After GCOS normalisation a ×2.5 fold selection cut-off was applied to all the regulated transcripts independent of probe set intensity values.Click here for fileAdditional File 3Comparison with the array of Gera and co-workers (7). Transcripts detected in both screens are listed with the fold change. Values are indicated in red when they differ between the two studies.Click here for file\n\nREFERENCES:\nNo References"
+}

batch_8/PMC2535587.json

@@ -0,0 +1,4 @@
+{
+    "id": "PMC2535587",
+    "text": "This is an academic paper. This paper has corpus identifier PMC2535587\nAUTHORS: Georgios Paraskevas, Paris Agios, Marios Stavrakas, Alexandra Stoltidou, Alexandros Tzaveas\n\nABSTRACT:\nAn abnormal origin of the left common carotid artery from the initial portion of the brachiocephalic trunk was found in the superior mediastinum in a 81-year-old Caucasian male cadaver during dissection practice. We report on the exact morphology of that variant that is appeared in an incidence of 0,2% in the literature. We discuss the relative literature and pay attention on the significance of such a variation for clinicians in its recognition and protection.\n\nBODY:\nIntroductionIncreasing activity in the fields of cardiac and vascular surgery has served to revive interest in the developmental and adult anatomy of the aortic arches and the great vessels derived therefrom. In the course of study of specimens in the Laboratory of Gross Anatomy, it became abundantly and strictly evident that the \"standard\" type of branching from the aortic arch not only existed in the preponderant number of cases, but also, when placed with a rather ordinary variation thereof, gave a combined total that represented over 90 per cent of cases in a series of 1000 specimens [1].The summit of the arch is usually 2,5 cm approximately below the superior sternal border but may diverge from this. Sometimes the aorta curves over the right pulmonary hilum (as a right aortic arch) descending to the right of the vertebral column, accompanied by a transposition of thoracic and abdominal viscera. Less often, after arching over the right hilum, it passes behind the oesophagus to its usual position; this is not accompanied by visceral transposition. The aorta may divide into ascending and descending trunks, sometimes dividing near its origin and the two branches soon reuniting; the oesophagus and trachea usually pass through the interval between them [2].As far as the branches of the aortic arch are concerned, there is a plenty of variations in the origins of them. An analysis of variation in branches from 1000 aortic arches showed the following findings: In 27%, the left common carotid artery originates from the brachiocephalic trunk. In 2,5%, each of the four arteries originate independently from the arch of the aorta, while in 1,2%, right and left brachiocephalic trunks originate from the arch of the aorta. The most common pattern in 65% is formed by the separate origination of three branches springing from the vessel's convex aspect: the brachiocephalic trunk, left common carotid and left subclavian arteries [3]. In our work we present a rare type of left common carotid artery origin from the initial portion of the brachiocephalic trunk.Case presentationWe dissected a 31-year-old, Caucasian, male, formaline-fixed cadaver. His ethnicity was Greek. His weight was 83 kg and the height 1,78 m. He had no past medical history and was on no medication. He did not use to smoke or drink (according to his next of kin). The dissection was carried out as part of the practice for the medical students and was approved by the Ethical Committee of the University. During the anatomical preparation we came across a variation referring to the branches springing from the aortic arch. After resection of the anterior thoracic wall we removed carefully the fat tissue and the pericardium covering the ascending aorta and the great vessels arising from it. Having obtained a clear view of the great vessels we noticed the presence of a left common carotid artery arising from the left surface of the origin site of the brachiocephalic trunk (Figure 1).Figure 1Picture of the cadaveric preparation. The left common carotid artery (b) is shown arising from the initial portion of the brachiocephalic trunk (c) (a: aortic arch, d: ascending aorta).DiscussionThree branches, as it is known, spring from aortic arch convex aspect: the brachiocephalic trunk, left common carotid and left subclavian arteries [4]. The ascending aorta arises from the base of the left ventricle behind the left sternal margin opposite the third costal cartilage. The arch of the aorta lies behind the lower part of the sternal manubrium. It begins behind the right border of the sternum at the level of the second rib cartilage, and extends dorsally to the left to reach the spine at the left of the body of the fourth thoracic vertebra [2]. The excision of the right lobe of the thyroid gland reveals the relation of the carotid and the subclavian arteries and the intervening portion of the aortic arch to the trachea [1]. They may branch from the beginning of the arch of the upper part of the ascending aorta; the distance between these origins varies, the most frequent being approximation of the left common carotid artery to the brachiocephalic trunk [2]. Primary branches may be reduced to one, more commonly two, the left common carotid arising from the brachiocephalic trunk (7%) [2], while Anson [3] rise this incidence to 27%. However Anson [3] referred to the presence of a left common carotid artery arising from the initial portion of the brachiocephalic trunk in a frequency of 0,2%. Because of the many changes involved in transformation of the embryonic aortic arch system into the adult arterial pattern, it is understandable that variations may occur. Most anomalies result from the persistence of parts of the aortic arches that normally disappear or from disappearance of parts that normally persist. Several kinds of uncommon defect occur when arches persist instead of becoming obliterated or vice versa. A right aortic arch occurs when the left fourth arch and dorsal aorta disappear. If the left fourth arch alone (and not the dorsal aorta as well) disappears, the condition of interrupted arch arises; the first part of the arch gives off the brachiocephalic and left common carotid vessels, and beyond the gap the pulmonary trunk and a persistent patent ductus (sixth arch) are required to complete an \"arch\" with the left dorsal aorta. If the right dorsal aorta persists as well as the left, a double arch ensues with the trachea and oesophagus clasped between the two [4,5]. Of course the analysis of aortic arch variability in morphology is beyond the aim of our study.In Anson's analysis of variation in branches from 1000 aortic arches there was a 65% of the usual pattern, a 25% of the four large arteries branching separately, the remaining 5% showed a great variety of patterns, the commonest (1,2%) being symmetrical right and left brachiocephalic trunks [3].More rarely, the left common carotid and subclavian arteries arise from a left brachiocephalic or right common carotid and subclavian arteries arise separately, in which case the latter more often branches from the left end of the arch and passes behind the oesophagus [1,2,6,7]. This anomaly assumes some importance in the adult as well as in the child, as a cause of esophageal compression. The abnormal course of the \"recurrent\" laryngeal nerve, which accompanies this anomaly, is also important [1].The left vertebral artery may arise between the left common carotid and the subclavian. Very rarely, external and internal carotid arteries arise separately, the common carotid being absent on one or both sides, or both carotids and one or both vertebrals may be separate branches. In about 12% the right common carotid artery arises above the level of the sternoclavicular joint or it may be a separate branch from the aorta; again it may arise with its fellow. The left common carotid artery varies in origin more than the right.When a \"right aorta\" occurs, the arrangement of its three branches is reversed. The common carotids may have a single trunk, the subclavians separate, the right arising from the left end of the arch. Other arteries may branch from it, most commonly one or both bronchial arteries and the arteria thyroide ima [2].As it is known specific interest is shown in surgery with respect to the relation of an anomalous arch or arches to the viscera in the neck and the thorax. Additionally, a variant of origin and course of a great vessel arising from the aortic arch is of great clinical value, because the ignorance on behalf the surgeon of such a variation may cause serious surgical complications during procedures occurring in the superior mediastinum and the bare of neck.Competing interestsThe authors declare that they have no competing interests.Authors' contributionsGP did the dissection and supervised the manuscript writing. AP performed the literature review and obtained the written consent. MS and AS obtained the photos and wrote the draft of the manuscript. AT helped to the final writing of the paper. All authors read and approved the final manuscript.ConsentA written consent was obtained by the cadaver's next of kin for publication of the article.\n\nREFERENCES:\nNo References"
+}