Upload 5000 files

Med-MDPI/cancers/cancers-12-06-01389.txt

@@ -0,0 +1 @@
+Pancreatic cancer has the worst prognosis and lowest survival rate among all types of cancers and thus, there exists a strong need for novel therapeutic strategies. Chimeric antigen receptor (CAR)-modified T cells present a new potential option after successful FDA-approval in hematologic malignancies, however, current CAR T cell clinical trials in pancreatic cancer failed to improve survival and were unable to demonstrate any significant response. The physical and environmental barriers created by the distinct tumor microenvironment (TME) as a result of the desmoplastic reaction in pancreatic cancer present major hurdles for CAR T cells as a viable therapeutic option in this tumor entity. Cancer cells and cancer-associated fibroblasts express extracellular matrix molecules, enzymes, and growth factors, which can attenuate CAR T cell infiltration and efficacy. Recent efforts demonstrate a niche shift where targeting the TME along CAR T cell therapy is believed or hoped to provide a substantial clinical added value to improve overall survival. This review summarizes therapeutic approaches targeting the TME and their effect on CAR T cells as well as their outcome in preclinical and clinical trials in pancreatic cancer.Pancreatic cancer, i.e., pancreatic ductal adenocarcinoma (PDAC), is a fatal disease with five-year overall survival rates of 1% to 5% and median survival duration of fewer than six months [1]. The poor prognosis has not substantially changed during the past decades, establishing pancreatic cancer as the fourth leading cause of cancer-related deaths in Western countries [2,3,4]. Therapeutic progress in other types of cancer will lead to its ascension in second place among all cancers within the next decade [5]. Surgery remains the only potentially curative treatment, but only a minority of patients show a resectable disease stage at diagnosis, due to invasion to the surrounding vasculature and due to lack of symptoms at an early stage [6]. Nonetheless, the median overall survival is still only 24 months for patients with resectable disease [7].Therapeutic failures of chemotherapy, targeted therapy, and immunotherapy of PDAC can be largely attributed to the special features of this cancer, which exhibits highly nutrient-poor, immunosuppressive, hypoxic and desmoplastic characteristics leading to rapid cancer progression [8]. The tumor is composed of only a minor number of malignant cells within a microenvironment of dense extracellular matrix (ECM), a barrier that prevents adequate drug delivery and might serve as a prognostic factor (Figure 1 and Figure 2) [8]. Responsible for the stromal reaction are mainly cancer-associated fibroblasts (CAFs) that develop from bone marrow-derived mesenchymal stem cells (MSCs), pancreatic stellate cells (PSCs), and quiescent resident fibroblasts through multiple pathways of activation [9]. The complex tumor vasculature in PDAC is characterized by a lack of blood vessels, leading to high levels of hypoxia in the tumor interior [10]. Furthermore, the capillaries and lymphatic vessels that are present tend to be collapsed due to high interstitial pressure, either from excess fluid or from solid stress [11]. Other non-neoplastic cancer-associated cells consist of immune-suppressor cells such as regulatory T cells (Treg), myeloid-derived suppressor cells (MDSC), and tumor-associated macrophages (TAM) that can inhibit CD8+ T cells, which play a key role in the antitumor immune response, and thereby establish an immunosuppressive tumor microenvironment [12]. Neural remodeling and perineural invasion (PNI), the neoplastic invasion of tumor cells into nerves, are further unfavourable histological features, and are considered as one of the main routes for cancer recurrence and metastasis after surgery [13]. Conventional therapies such as chemotherapy and radiation have focused on effective therapy of the malignant cell population. Thus, a concordant combination of various treatments targeting additional key cellular features of PDAC such as stroma, reversing suppressive immune reactions and enhancing antitumor reactivity may lead to more successful treatment strategies [14]. Thus, there is a clinically unmet need for new therapeutic options.Immunotherapy is a rapidly developing field within oncological research, especially since the development of chimeric antigen receptor (CAR) T cells, which are genetically engineered to express receptors targeting cancer cells for immunotherapy. CAR technology has made leaps of development since its conception in 1993, combining antigen recognizing regions from antibodies with intracellular T cell signaling domains (Figure 3) [16]. In this way, potential demasking of tumor cells by major histocompatibility complex (MHC) class I downregulation, can be overcome [17]. At first, double chimeric receptors were developed by engineering the VH and VL chains of immunoglobulins to the constant regions of the T cell receptor (TCR) [18]. Over time, CARs were modified into a single chain approach coupling a single chain variable fragment (scFv) derived from an antibody via a spacer and transmembrane domain to the CD3ζ signaling domain of the TCR [16]. The addition of costimulatory domains from CD28 or 4-1BB generated a stronger activating signal, circumventing the intracellular activation by TCR-domains, defining the second CAR generation [19]. Second-generation CARs targeting CD19 are the first CAR success-story wherein phase II study 81% of the B cell acute lymphoblastic leukemia patients demonstrated complete remission 28 days after infusion [20]. Their tremendous success in the treatment of leukemia and lymphoma patients led to the FDA approval of the first CAR T cell therapy as a second-line treatment in 2017 [21]. The incorporation of further costimulatory domains derived from CD27 or CD40 as well as the introduction of additional cytokine expression or induction of other signaling pathways established the third, fourth, and fifth generations of CAR T cells, increasing cytokine production, cell survival, and persistence [22]. In recent years, advanced CAR concepts, such as Tandem or Universal CAR approaches have been developed and enabled the targeting of challenging antigen expression profiles on cancer cells [23]. Other advanced CAR technologies explore mechanisms to switch on and off CAR expression on T cells to control possible toxic side effects [24]. Another upcoming class of engineered receptors is synthetic Notch (synNotch) receptors, which can induce transcriptional activation after target recognition [25]. Ultimately, all developmental generations of CARs offer various opportunities and challenges for prospective cell-based approaches as reviewed before [22,24,26]. Unfortunately, fewer exciting outcomes were achieved in initial clinical trials with CAR T cells targeting solid tumors, including PDAC. Successful CAR therapy for carcinomas needs to overcome the physical and environmental barriers in the tumor microenvironment (TME) [27]. The TME consists, next to tumor cells, of endothelial, immune, and inflammatory cells, stromal cells, the extracellular matrix and a broad spectrum of enzymes, cytokines, and growth factors [28]. This creates a strong physical barrier for CD8+ T cells, while their immune response is further diminished by the high amount of immunosuppressive immune cells present in the TME of PDAC [12,29]. These aspects must be considered and addressed in the field of cell-based immunotherapy against solid cancers. Here we review different strategies to overcome these hurdles for successful CAR T cell therapy in pancreatic cancer. The first obstacle for effective CAR T-cell therapy for carcinomas is the lack of suitable targets on carcinoma cells. CAR T therapy induces an ablation of all cells with a certain degree of antigen expression leading to potentially fatal side effects such as “on target/off tumor” toxicities [30]. Unfortunately, this is also the case for most of the PDAC targets tested in preclinical and clinical trials such as carcinoembryonic antigen (CEA), CD133, CD70, Claudin 18.2, epithelial cell adhesion molecule (EpCAM), receptor tyrosine-protein kinase erbB-2 (HER2), mesothelin, and prostate stem cell antigen (PSCA) (Table 1) [31].The most advanced targets for clinical consideration are CEA and mesothelin, with up to five clinical trials completed, active, or recruiting (CEA: NCT03818165, NCT02850536, NCT02416466, NCT04037241, NCT03682744; mesothelin: NCT03323944, NCT03497819, NCT03638193, NCT01897415). In contrast, the only published results from clinical trials of CAR T cells in PDAC originate from mesothelin and CD133. The mesothelin-specific CAR trial resulted in two patients with a progression-free survival of four to five months and another patient showed a reduction of liver lesions, but not of the primary tumor (NCT01355965) [33]. The CD133 CAR trial also demonstrated a partial remission in two PDAC patients with Grade II toxicity, potentially due to the expression pattern of CD133 in hemopoietic stem cells (NCT02541370) [32]. Both studies verified the feasibility, safety, and principal efficacy of CAR T cell therapy for pancreatic cancer. Nevertheless, several problems prevented the induction of full remission and improvement of survival by immunotherapy despite its efficacy against metastases, often the discriminating factor for successful cancer therapy [71]. Two of the problems that must be solved for effective CAR T cell treatment are (i) emerging exhaustion and (ii) missing persistence of CAR T cells [32,33]. Co-treatment with PD-1/PD-L1 interfering checkpoint inhibitors or multiple infusions of CAR T cells might overcome these problems [72]. This aims to precondition chemotherapy and CAR constructs modifications, e.g., different costimulatory domains for CD4+ and CD8+ CAR T cells as well as transgenic cytokine expression, might overcome these problems [72]. However expression levels of cytokines need to be steered carefully, e.g., with conditional induction, to limit the risk for toxic cytokine release syndrome (CRS) [73].The heterogeneity underlying PDAC makes therapeutic options based on one-size-fits-all approaches ineffective. Among others, Bailey et al. [15] defined for example four subtypes of PDAC, based on genomic analysis correlating with histopathological characteristics. These various PDAC types and their distinct stroma subtypes imply a specific stratification of the patients due to different behavior under the same treatment [74]. The complexity is further increased by another hurdle, which remains unchallenged: advanced targets in pancreatic cancer are usually heterogeneously expressed and are sometimes just present on 20% of the tumor cells, leading to progression of the diseases by the target-negative cells in the clinical trials [31,32,33]. Therefore, classifying patients in subtypes that could benefit from cell therapy would help improve outcomes and quality of life as well as avoid ineffective or even risky therapy approaches. These complex circumstances require the identification of new CAR targets as well as sophisticated Tandem, Universal CAR, and adapter-CAR approaches. In this way, unintentional “on target/off tumor” toxicities can be prevented for a safe and balanced application of CAR T cells in pancreatic cancer [75].A second major hindrance for cell therapy is the complex TME of solid tumors, representing an exceptional challenge in comparison to other tumor types. However, the histological key feature of PDAC is the occurrence of a unique desmoplastic reaction, comprising over two-thirds of the total tumor volume and destructing the architecture of normal pancreatic tissue [76]. Desmoplasia is marked by a dramatic increase in the proliferation of alpha-smooth muscle actin-positive fibroblasts and is also accompanied by the increased deposition of extracellular matrix molecules [77]. This has a strong impact on treatment outcomes since cytotoxic therapy can not only increase the amount of active CAFs but also increase their treatment resistance and tumor aggressiveness [78]. Another aspect of the dense tumor stroma is the limited availability of nutrients and oxygen [12]. The consequences of this deprivation for immune cells, including CAR T cells, in the stroma of solid tumors as well as major changes in the metabolic processes of the TME, have been extensively reviewed elsewhere [79,80,81] and will not be addressed in this review.Under normal conditions, stromal fibroblast cells communicate and interact with the surrounding ECM. They secrete and synthesize new ECM molecules as well as growth factors and enzymes, e.g., upon stimulation by tissue injury [82]. Under pathological conditions in the context of cancer however, the complexity of fibroblasts’ roles increases. In an early tumour stage, fibroblasts have been demonstrated to prevent tumour growth by remodeling the ECM and inducing an anti-tumour immune response [83]. Whereas at later stages with an established tumour, fibroblasts transform into activated CAFs, where they become tumorigenic and enhance metastasis-potential and chemoresistance [84]. ECM molecule expression and release of tumour-promoting cytokines can also be increased in activated CAFs, but stimuli and time point of phenotype switch are still under investigation [85]. CAFs can originate from various cell types, such as resident fibroblasts, chondrocytes, adipocytes, mesenchymal stem cells, pericytes, and mesenchymal transitioned endothelial and epithelial cells, including cancer cells and cancer stem cells [86]. In PDAC, CAFs can additionally be derived from PSCs, quiescent under normal conditions but transitioned into a myofibroblast-like phenotype under pathophysiological conditions in the pancreas [87]. Regardless of CAF origin, this cell type can constitute up to 90% of the tumour mass in PDAC, representing an inevitable hurdle for expedient treatment strategies [88].Accordingly, numerous efforts have tried to dispose of CAFs or reprogram them within the TME [89]. In the context of CARs, several groups have generated fibroblast activation protein (FAP)-redirected CAR T cells to erase FAP-expressing CAFs, resulting in a reduction of ECM molecules and tumour growth, also in a syngeneic murine pancreatic cancer model [34,35,36]. FAP is a serine protease capable of local ECM modification by changing fibronectin orientation [90]. All studies emphasized the value of co-targeting CAFs and tumour cells simultaneously for solid tumours. Nevertheless, a debate is on-going regarding the safety of FAP as a CAR target, after the demonstration of hematopoietic side effects due to FAP+ bone marrow stromal cells (BMSCs) in mice [37,38]. Other possible extracellular markers expressed on CAFs, e.g., platelet-derived growth factor receptor (PDGFR) α and β, exhibit inappropriate expression patterns [86,91]. Therefore, more convenient and safe targets or target combinations need to be evaluated for successful CAF-redirected CAR establishment.Next to cell-based CAF depletion, drug-based therapeutic options have also been proposed. Nab-paclitaxel has been shown to decrease CAFs numbers in PDAC in a clinical trial in combination with gemcitabine (NCT00398086) [92]. Small molecules inhibiting the sonic hedgehog (SHH) pathway have demonstrated promising preclinical results but failed to recapitulate these outcomes in clinical trials [93,94]. A phase II clinical trial (NCT01130142, NCT01064622) with a combination of vismodegib (GDC-0449) and gemcitabine revealed no survival benefit [39]. One possible explanation supported by the results of Özdemir et al. [95] is that the depletion of myofibroblasts in pancreatic cancer may also accelerate cancer growth and reduce survival. While the myofibroblast-depleted tumours did not respond to gemcitabine, anti-CTLA4 immunotherapy inverted the outcome and resulted in prolonged animal survival. Although FAP+ cell-depletion upon adenoviral vaccination demonstrated an improvement of CD8+ T cell function [40], remodeling of CAF expression pattern instead of CAF depletion might be a better-suited strategy for combinatorial approaches with immunotherapy in PDAC. The clinically most advanced substance to alter CAF expression phenotype is all-trans retinoic acid (ATRA), currently used as the standard treatment of acute promyelocytic leukaemia but also tested in PDAC [96]. It reduces ECM and cytokine secretion by inhibiting FAP, ACTA2 and transforming growth factor β receptor (TGF-βR) expression on CAFs [41]. Suitability of ATRA for stromal remodeling in pancreatic cancer is currently under clinical investigation (NCT03307148, NCT03878524) [42]. Another preclinical substance reducing CAF activation and expression in PDAC is JQ1; an inhibitor of the BET family of bromodomain chromatin-modulating proteins [43]. JQ1 has been demonstrated to control MYC silencing [97]. Since MYC-activated cells secrete factors, which can induce an MYC-dependent metabolic program in CAFs, JQ1 might be able to interfere with the tumour cell-CAF crosstalk [44]. Furthermore, the PDAC-specific CAF precursor cells, PSCs, can be remodeled to decrease the desmoplastic reaction. Calpeptin, a calpain inhibitor, was also able to decrease fibrosis in a subcutaneous xenografts mouse model using co-implantations of PSCs and pancreatic cancer cells [45]. A combination of metformin and gemcitabine resulted in significantly lower tumour size and reduced collagen amounts in an orthotopic mouse model [98]. Unfortunately, most of the approaches are not protein or nucleic acid-based and cannot be produced by CAR effector cells. Therefore, FAP-depleting or remodeling molecules could be applied as a pharmacological pre-treatment to reshape the therapy-inhibiting expression pattern of CAFs. Alternatively, FAP-redirected CAR T cells could be used to deliver CAF remodeling factors or antibodies to inhibit the crucial expression profil of CAFs and their autocrine feedback loops (Figure 4) [99]. Tandem chimeric antigen or synNotch receptor approaches could be appllied simultaneously or in a time-shifted manner.One of the key features of activated fibroblasts is their distinct ECM production, especially crucial in PDAC with its pronounced desmoplasia [86]. Whatcott et al. [100] observed a strong negative correlation between patient survival and high levels of ECM deposition, also a solid tumour specific hurdle for immunotherapy [101]. Thus, the composition of the ECM in combination with the capability of CAR T cells to degrade extracellular matrix proteins can have a major influence on T cell tumour-trafficking and infiltration. A major challenge, however, is the fact that ECM proteins are not necessarily tumour-specific, but exert important physiological functions in organ development, tissue integrity, and wound healing [102].Caruana et al. [46] demonstrated that ex vivo manipulated CAR T cells may downregulate ECM-degrading enzymes and overexpression heparanase improved CAR T cell infiltration and anti-tumour activity in vivo. However, heparan sulphate proteoglycans are not the only obstacle in the ECM of PDAC [103]. It is composed of collagens, non-collagen glycoproteins, glycosaminoglycans, growth factors, and proteoglycans as well as modulators of the cell-matrix interaction. Overexpressed ECM molecules, including thrombospondin, periostin, hyaluronic acid (HA), tenascin-C, vitronectin, collagens, and fibronectin increase pancreatic cancer cell migration and invasion [104]. Some of these molecules have already been exploited for possible effects on immunotherapy approaches. Collagen is the most frequent molecule in the ECM of PDAC and a major component of the desmoplastic reaction [105]. Furthermore, a collagen-derived proline can compensate as an alternative nutrient source in the resource-deprived TME [106]. However, collagen also regulates the activity, phenotype ratio and the amount of tumour-infiltrating T cells due to its dense network [107]. In this way, mammary tumours with a high collagen-density, correlated with a worse prognosis, contained a higher ratio of CD4+ to CD8+ T cells and an overall reduced amount of infiltrating CD8+ T cells. In PDAC, it was demonstrated that excessive collagen amounts abrogated tumour cell-directed movement of T cells by chemokines, but favoured T cell movement to the stroma cells in a contact guidance dependent manner [108]. These findings imply the relevance of the ECM composition for cell-based immunotherapy in solid tumours. Despite the severe impairment created by the collagen network, Ishihara et al. [47] managed to turn the presence of collagen into an advantage by increasing the delivery of cytokines with a short half-life, such as IL-2, and checkpoint inhibitors specifically and dosable to the TME through coupling to a collagen-binding domain. This enables a safe approach to shift the balance of pro- and anti-tumorigenic cytokines and stimulate the immune cells in the TME. Consequently, collagen-redirected IL-2 reduced common side effects such as vascular leak syndrome and increased tumour infiltrating CD8+ T cells in an orthotopic breast cancer mouse model [47].Fibronectin, another common molecule in the ECM of pancreatic cancer, but not in healthy tissues, is considered to be a significant hallmark of epithelial-to-mesenchymal transition (EMT) occurring in advanced tumours [109]. Fibronectin interacts with many ECM and surface molecules, creating an active interaction platform. This stimulates the EMT and multiple aggressiveness- and resistance-related signalling pathways, which in turn upregulate fibronectin expression, resulting in a strong feedback loop in the TME [110]. As in the case of collagen, intratumoural regions with low fibronectin amounts displayed high leukocyte infiltration [111].The important role of fibronectin led to the creation of several approaches inhibiting its functions or using its presence in the TME for imaging, drug delivery, and therapy [112,113]. BC-1 coupled to IL-12 was used for TME-targeted cytokine delivery in clinical studies and resulted in stable disease in 46% of melanoma or renal cell carcinoma patients [48]. However, the single-chain variable fragment (scFv) L19-based cytokine delivery is more clinically advanced than the BC-1 based IL-2 delivery [49]. L19-IL2 (DARLEUKIN®) is already in clinical trials against various solid tumour types (NCT01058538, NCT02086721, NCT02735850, NCT03705403). Despite the promising preclinical results, a clinical trial of L19-IL2 with gemcitabine in patients with advanced pancreatic cancer had to be terminated due to lack of recruitment (NCT01198522). Nevertheless, phase II trials in melanoma patients resulted in reduced metastasis and increased survival demonstrating the potential of fibronectin-redirected IL-2 [50,51]. Besides IL-2, L19 was also coupled to IL-12 and tumour necrosis factor (TNF) α, revealing equally promising results in solid metastatic cancers [58,114], in particular for L19-TNF in combination with L19-IL2 [115]. In this way, targeting fibronectin enabled TME-specific cytokine delivery to outbalance immunosuppressive cytokines. This can be exploited as a combinatorial therapeutic strategy together with CAR T cells or as a pre-treatment.Similar to fibronectin, tenascin-C is mostly present in the pathophysiological conditions of adults, building up a provisional matrix in the scar formation process [59]. It is upregulated in the ECM of solid tumours, including PDAC [60]. While the exact role of tenascin-C remains undefined, it is widely known for its modulation capacity on cell adhesion to fibronectin and its promotion of EMT, enhancing cancer cell growth and motility [116,117]. It has also been shown to interact with multiple ECM molecules and to facilitate the angiogenic switch by representing an important factor of the AngioMatrix (ECM and related protein involved in the angiogenic switch) inducing resistance to chemo- and anti-angiogenic therapy in PDAC [118]. Nevertheless, no correlation between high tenascin-C expression and survival has been determined. However, overexpression of tenascin-C together with other ECM-related factors has been shown to correlate with poor prognosis for patients of pancreatic cancer [119]. Tenascin-C pronounced importance in the context of solid tumours led to multiple approaches to modify tenascin-C in the ECM or to make use of its presence. Inhibition of tenascin-C expression is possible by blocking its natural activation pathways such as transforming growth factor β (TGF-β), but also by RNA interference resulting in only short survival prolongation [120]. Tenascin-C expression and signalling have been demonstrated to be prevented by angiotensin II type 1 receptor (AT-1) and angiotensin-converting enzyme (ACE) inhibitors, which has not yet been assessed in the clinic [120]. Another possibility would be to erase tenascin-C, as previously described for heparan sulphate proteoglycans, from the ECM of solid carcinomas, a process occurring after wound healing. Unfortunately, this mechanism has not yet been identified (reviewed by Spenle et al. [120]). Therefore, as in the case of fibronectin, multiple antibodies have been generated redirecting radionuclides and cytokines to the tenascin-C-rich ECM. F16-IL2 (TELEUKIN®), an IL-2 coupled antibody-cytokine fusion protein is the most advanced candidate with two clinical trials in solid tumours, such as breast and lung cancer (NCT01131364, NCT01134250). This recombinant protein demonstrated its ability to increase survival as well as the number of macrophages and NK cells in the tumour stroma in a BALB/c nude mice breast cancer model [52]. F16-IL2 clinical potency has also been analysed in a clinical setting in solid tumours including pancreatic tumours, demonstrating an anti-cancer activity in combination with doxorubicin [53]. Thrombospondin 1 (TSP-1) is a strong inhibitor of angiogenesis, promotes inflammatory (‘M1-type’) macrophage recruitment and prevents stemness of cancer cells. Via its crosslinking-interaction with the “don’t eat me”-signal CD47 it can directly induce tumour cell death [121,122]. However, it also releases the active form of TGF-β from its latent form, promotes Treg formation and inhibits T cell proliferation [123,124]. Several inhibitors for TSP-1 are available with the most advanced being ABT-510, CVX-045, and Trabectedin [62,63]. While ABT-510 showed a limited increase of cytotoxic T cell frequency, it did not demonstrate efficacy in various solid tumours as a monotherapy leading to its suspension from clinical development (NCT00586092) [62,125,126]. Trabectedin, approved for the treatment of sarcoma and ovarian cancer, indicated a tremendous effect on favourable cytokines/chemokine expression level, although there was no efficacy as a single agent in stage II clinical trial for salvage therapy in metastatic pancreatic cancer (NCT01339754) [64]. Nevertheless, based on the findings of Weng et al. [127] TSP-1-targeted therapy in combination with cell therapy may deserve a second chance as a more nuanced treatment. Here it was shown that downregulation of TSP-1 solely in dendritic cells increased the amount of tumour-infiltrating CD4+ and CD8+ T cells [127].Next to heparan sulphate proteoglycans, hyaluronic acid (HA) is another glycosaminoglycan, overexpressed in the ECM of PDAC [104]. HA is widely expressed in all tissues and plays an important role in multiple biological processes, e.g., cell proliferation, inflammation, and angiogenesis [128]. Nevertheless, it exerts its most important biological functions by regulating cell motility via CD44, the tissue hydration influencing the intestinal fluid pressure (IFP), tissue permeability, and drug delivery potential [100,129]. Consequently, high amounts of high molecular weight HA contribute to a stiff tumour matrix increasing the IFP and reducing the ability of chemo-, nanomedicine, and cell-based therapies to penetrate stroma-rich tumours [130]. Accordingly, HA accumulation in the ECM of pancreatic cancer patients correlates with poor survival [131]. Unlike tenascin-C, there is a specific way to remove excess high molecular weight HA from the ECM. HA disruption with the PEGylated human recombinant PH20 hyaluronidase (PEGPH20) indicated improved drug delivery and response in a mouse model of pancreatic cancer and increased CD8+ T cell infiltration and better checkpoint inhibitor efficacy in a syngeneic breast cancer mouse model [54,55]. PEGPH20 treatment also resulted in a remodeling of the TME by decreasing other ECM molecules, such as collagen and tenascin-C. The promising preclinical success was also transferred to the clinic (NCT03481920, NCT01453153, NCT01839487, NCT04058964, NCT03634332, NCT02241187, NCT02921022, NCT02910882, NCT01959139, NCT04134468, NCT03193190, NCT02715804) and was in stage III of clinical development for pancreatic cancer [56]. Unfortunately, the phase III study was not able to meet the endpoint criteria, halting further development [57]. Nevertheless, especially for cell-based therapy approaches, which are limited by larger diameters (hydrodynamic size) than chemotherapeutics, depletion of HA may have a potential of exerting a significant impact on therapy delivery. Altogether, these findings imply the importance of the ECM for the outcome of cancer therapy including immunotherapy. The impact of the ECM on the therapeutic outcome is further strengthened by the wide range of cytokines, which are bound and released by various ECM molecules after expression by CAFs and tumour cells, as recently reviewed by Tzanakakis et al. [132] for the group of the proteoglycans. Furthermore, options that failed before as monotherapies or in combination with chemotherapeutics deserve a second consideration for suitability in combination with immunotherapy. In the long-term, the latest CAR technologies could be utilized to secrete engineered proteins to increase tumoricidal immune response and CAR T cell infiltration, overcoming the complex barriers created by the ECM. The majority of the growth factors, expressed by cancer cells or CAFs in the TME, increase cell survival, proliferation, migration, and metastasis in an autocrine feedback loop or in a paracrine manner, via their associated receptors [99]. They can also be bound by ECM molecules and be released by enzymes, such as matrix metalloproteinases (MMPs) [86,133]. Aside from the close cancer cell and fibroblast communication network, some of these factors are also released by other immune cells in the TME, such as tumour-favouring M2 macrophages or neutrophils [134,135]. A thoroughly-investigated factor is the vascular endothelial growth factor A (VEGF-A), and its receptor (VEGFR2), which regulates the process of angiogenesis [28]. Unlike most hematologic malignancies, solid tumours heavily depend on the formation of new vessels for sufficient blood supply. Hypoxia in all tissues, including cells present in the intertumoral regions of PDAC, induces the expression of VEGF after hypoxia-inducible factor 1 alpha (HIF-1) translocation to the nucleus in a gradient manner, which in turn initiates the growth of new blood vessels into hypoxic regions [136,137]. Nevertheless, the relationship between angiogenesis and PDAC is far more complex. On the one hand, PSCs and CAFs secrete VEGF, which leads to increased, disorganized vascular growth and formation with enhanced IFP [11]. While on the other hand, the dense desmoplastic reaction around pancreatic tumours leads to vascular disruption, which further increases hypoxia and reduces drug administration [10]. This leads to insufficient therapeutic-dose delivery that might, to some extent, explain the low survival rates in patients with pancreatic cancer [61,138]. Cell therapy also relies on functioning vessels [79]. Fortunately, vessel function can be restored by using anti-angiogenic treatments, such as bevacizumab, to normalize vessel organization and IFP [139]. Co-treatment of angiogenesis inhibitor bevacizumab together with GD2-redirected CAR T cells increased tumour infiltration and antitumor activity in a preclinical neuroblastoma model [66]. Bevacizumab was already tested in pancreatic cancer patients in combination with gemcitabine. Despite the promising objective response rate of 21%, there was no difference in the overall survival time between the bevacizumab and the placebo group (NCT00088894) [67]. This undesirable outcome may be attributed to the ability of tumours to acquire resistance to VEGF inhibition, e.g., by the release of more proangiogenic factors, such as angiopoietin 1 (ANGPT1), resulting in increased amounts of vascular progenitor cells [140]. Recently, another mechanism dependent on the ECM molecule periostin, present in ECM of PDAC, has been revealed and induced revascularisation and macrophage recruitment [65]. The second effect was reversible by the addition of an anti-colony stimulating factor 1 receptor (CSFR1) antibody, blocking macrophage infiltration [65]. This highlights the importance of understanding the individual TME composition of each patient in order to match the most suitable anti-angiogenic treatment, because many of the early mentioned ECM molecules have been shown to modify angiogenesis in different ways, e.g., by VEGF interaction [129]. Modification of other TME molecules, such as thrombospondin-1, together with anti-angiogenic treatment has already been evaluated in the clinic by the co-treatment of advanced solid tumours with bevacizumab and ABT-510, resulting in partial response for one patient and stable disease for more than a year in five patients [68]. Hence, combining multiple anti-angiogenic approaches with cell therapy might be necessary for a successful cell-based immunotherapy of PDAC. These findings stress the importance of moving away from the current one-size-fits-all therapy approaches to more personalized combinatorial therapies, simulating personalized nanomedicine approaches [141].Tumour cells in hypoxic areas often express other growth factors next to VEGF. Their interactions with their defined receptors lead to receptor tyrosine kinase (RTK) induction, which can be antagonized by the blockage of downstream signalling pathways with RTK inhibitors [142]. RTKs are a group of cell surface receptors involved in multiple key pathways of cell proliferation, differentiation, survival, and migration. The inhibition of the RTK, Axl, attracted attention for its influence on immune cells and not on tumour cells. Axl has been associated with the traditional RTK pathways in cancer cells and with the regulation of innate immune response and a more aggressive and resistant phenotype [143,144]. These findings motivated the preclinical evaluation of the Axl receptor as a target for monoclonal antibody immunotherapy in pancreatic cancer [145]. Small molecule inhibition by BGB324 of Axl decreased immune suppression and increased chemotherapy potency in pancreatic cancer and synergized with CAR T cell therapy in B cell malignancies [69,70]. This in vivo demonstrated synergy was dependent on T helper cell type 1 phenotype polarization, expressing an anti-tumorigenic cytokine profile, induced by Axl inhibition. Given the great influence on vessel functionality and further, on immune cells, growth factor modification might have a significant influence on the improvement of immunotherapy in solid tumours. These findings encourage the application of already clinically approved drugs as supporting combinatorial approaches with immunotherapy. Upon favourable outcomes from clinical trials, biological inhibitors such as bevacizumab, could even be secreted by the CAR T cells, creating a living drug.Pancreatic cancer represents an exceptional challenge for successful cancer therapy. CAR T cells are no exception, instead, they face great obstacles but also have the capacity to offer valuable chances. Cell-based immunotherapy has shown pronounced clinical success in hematologic malignancies and its feasibility has been demonstrated in pancreatic cancer, but it needs to overcome certain barriers, such as infiltration, persistence, and exhaustion. However, the first major hurdle is the heterogeneity of pancreatic cancers in terms of proposed subtypes and varying target expression. This requires advanced CAR technology to ensure the successful targeting of all cancer cells. The complex and heterogenous TME is the second major hurdle specifically for CAR T cells against pancreatic cancer. All parts of the TME require individual strategies. Reprogramming of CAFs might be more favorable than CAFs depletion without directly powering up the therapy intensity. The presence of tumor specific ECM molecules, as described in this review, would enable a specific delivery of cytokines, using agents such as F19-IL-2 [53]. In this way, both approaches could be combined strategically to first loosen the dense stroma, before boosting up CAR T cells. This represents an option to increase the temperature of immunological “cold” tumors, similar to PDACs [146]. However, tremendous tumor growth in areas that are no longer suppressed needs to be vigorously prevented. The same holds true for situations, where CAR T cells are equipped with ECM-degrading enzymes, such as overexpressed heparanase, or tumors are pre-treated with IFP decreasing molecules such as PEGPH20 [55,56]. Restored baseline IFP and vessel function is of major importance for successful CAR T cell delivery to the tumor, even if they are provided with infiltration-increasing mechanisms, such as heparanase [46]. IFP and enhanced permeability and retention (EPR) effect in cancer nanomedicine are closely related. Hence, high-resolution 3D imaging techniques, used in nanotherapy, could be applied for translational approaches in terms of vessel functionality in vivo and later patient stratification for combinatorial cell-based therapies [147]. A high need for vessel functionality assessment is also present for the analysis of the interplay of all the ECM molecules and growth factors in the TME, which can influence vessel growth and development [11,68]. Tumors undergoing anti-angiogenic treatment strategies, such as bevacizumab, may develop resistance mechanisms. Those mechanisms can be dependent on the ECM composition, but might also be overcome by modifications of the present molecules. The availability of vessel-independent growth factors, secreted by the various players in the TME indicates a medical need for in-depth patient-stratifications based on the presence of key different TME molecules, especially when it comes to the application broad range RTK inhibitors. This research requires technically advanced organoid or tissue printing methods, combined with established immunological assays. Taken together, there is an overall need for the development of new in vivo and in vitro assays in combination with imaging strategies to facilitate combinatorial research and improve preclinical translation potential. Agents, which might have failed as monotherapies, might deserve a second look in the context of combinatorial approaches with immunotherapy, due to their characteristics as a “living drug”. Research on different TME subtypes needs to be intensified and these parameters, in addition to molecular markers, need to be taken into account to define clear subgroups of PDAC. The acquired knowledge should assist in identifying only the PDAC patient, who will benefit from a particular personalized medicine concepts (Figure 5).  Therefore, sub classifying patients would help to improve outcomes and quality of life, as well as avoid ineffective therapy and reduce financial and organizational burdens on the health systems, healthcare providers, and the patients. These efforts will hopefully utilize existing and developing pharmacological therapies, regardless of their stand-alone therapeutic success, in combination with CAR T cells to create highly improved multifactorial therapeutic strategies, that can overcome the current hurdles faced by the challenging TME in pancreatic cancer. This research received no external funding. The authors would like to thank Rita Pfeifer for the scientific advice and helpful discussion and Jeannine Mißbach-Güntner for providing the haematoxylin/eosin-stained PDAC sample.J.H., O.H. and W.A. are employees of Miltenyi Biotec B.V. & Co. KG. All other authors declare no competing interests.Complex tumor microenvironment (TME) of pancreatic cancer. The pancreatic ductal adenocarcinoma (PDAC) microenvironment is characterized by a dense desmoplastic stroma, with cancer-associated fibroblasts (CAFs) presenting the majority of the cell population (in grey). Tumor cells (round and brown) in aggressive PDACs can occur in tumor buds, small groups of cells, especially in the invasive front. A high abundance of extracellular matrix (ECM) molecules, enzymes, and growth factors is another important feature. Immune cells are often excluded from the TME or exhibit an immunosuppressive phenotype. The distribution of pro- and anti-inflammatory immune cells as well as the exact composition of the tumor stroma is dependent on the subtype of pancreatic cancer as discussed by Bailey et al. or by Karamitopoulou [12,15].Haematoxylin/eosin-stained human PDAC sample. Tumor cells (arrow) are surrounded by the desmoplastic reaction of stromal cells and few immune cells.Developmental stages of chimeric antigen receptors. The first double chain chimeric receptors were engineered to customize the variable T cell receptor (TCR) domain by using VH and VL chains of antibodies (orange and bright blue boxes) fused to the constant regions of the TCR α- and β-chains (green and blue boxes). They mimicked the TCR in appearance and functionality. Activation relies on association with intracellular CD3ζ (yellow boxes), CD3γ, CD3δ, and CD3ε chains (purple boxes). The first generation of CARs combined the antigen recognizing scFv directly with the CD3ζ-signalling domain in one construct overcoming expression difficulties by the tremendous construct length of double chain chimeric receptors. Cytotoxicity, proliferation, cytokine secretion, and persistence of CARs were increased in second and third generation CARs by the addition of further costimulatory domains (CS1 and CS2) such as CD27, CD28, CD134, or 4-1BB. Introduction of T cell redirected for universal cytokine-mediated killing (TRUCKs) or fourth generation CARs increased the flexibility in CAR design for specific challenges even further, enabling local expression of cytokines such as IL-12, which are toxic in high concentrations. Fifth generation CARs, as fourth generation CARs, are based on second generation CARs. The individual antigen response is complemented by activation of intracellular domains of cytokines (dark blue box) e.g., IL-2Rβ, which induced signal transduction in the STAT3/5 pathway. Another group of artificial antigen receptors, gaining increased interest in recent years, are synNotch receptors. These receptors use the cleavage process after Delta-Notch binding and enable an unlimited variety of responses (green box) after target recognition such as cell fate determination with transcription factors and expression of selected cytokines or therapeutic antibodies. In this way, they bring the potential of immune cells as “living drugs” a big step forward.Strategies for CAR T cells to overcome or use the TME for successful immunotherapy. CAR T cells face major hinderances created by the distinctive TME of pancreatic cancer. Some of the hinderances might be surmounted or turned into a specific targeting strategy. CAFs may represent up to two thirds of the pancreatic tumor mass. However, CAF-depletion or remodeling approaches using CAR T cells or pharmacological substances such as ATRA or nab-paclitaxel might be able to break their crucial influence in the TME. Another strategy, potentially breaking the crucial influence of CAF expression profile in the TME, could be the application of FAP-redirected synNotch CAR T cells to deliver specific antibodies for inhibition of excess growth factors. Collagen is a key molecule in the creation of the dense ECM of PDAC, while its presence could be used for specific delivery of cytokines, required to boost CAR T cell efficacy and persistence. Moreover, it has already been demonstrated that CARs, re-equipped with ECM-degrading enzymes, such as heparanase, had higher infiltration compared to the control CARs. Multiple TME components have a high potential of influencing vessels development and growth. These components need to be targeted and modified, e.g., by inhibitory antibodies to improve vessel functionality and ensure directed CAR T cell transport to the pancreatic tumor. Use of broad RTK inhibition needs to be balanced after careful consideration of their influence on different TME players. In this way, polarization of pro-inflammatory cells into anti-inflammatory cells can be prevented.Strategy flow chart for PDAC therapy. Pancreatic cancer patients are classified into one of three categories upon diagnosis. Therapy (Tx) is chosen on the basis of this classification. In case of later stage PDAC or recurrent tumor, personalized medicine approaches could be of use. Imaging of patients would be followed by tissue retrieval to perform in-depth phenotyping of the tumor and its stroma. This could be performed by the application of up and coming technologies such as patient-derived organoids analysis, RNA-Seq or multiplex immunofluorescence staining. All in all, such refined selection criteria enables the balanced and careful stratification of patients into further effective and safe therapy paths, including personalized therapy approaches, such as CAR T cell therapy, with or without conditioning of the tumor microenvironment.Therapeutic options for combinatorial stromal and immunotherapy.Abbreviations: CEA, carcinoembryonic antigen; EpCAM, epithelial cell adhesion molecule; HER-2, receptor tyrosine-protein kinase erbB-2; PSCA, prostate stem cell antigen; FAP, fibroblasts activation protein; CAR, chimeric antigen receptor; CAF, cancer-associated fibroblasts; ATRA, all-trans retinoic acid; N/A, not applicable; CBD, collagen binding domain; CPI, immune checkpoint inhibitors; PEGPH20, PEGylated recombinant human hyaluronidase; VEGF, vascular endothelial growth factor; RTK, receptor tyrosine kinase.

Med-MDPI/cancers/cancers-12-06-01390.txt

@@ -0,0 +1,2 @@
+Remodeling of the extracellular matrix (ECM) is an important part in the development and progression of many epithelial cancers. However, the biological significance of collagen alterations in ovarian cancer has not been well established. Here we investigated the role of collagen fiber morphology on cancer cell migration using tissue engineered scaffolds based on high-resolution Second-Harmonic Generation (SHG) images of ovarian tumors. The collagen-based scaffolds are fabricated by multiphoton excited (MPE) polymerization, which is a freeform 3D method affording submicron resolution feature sizes (~0.5 µm). This capability allows the replication of the collagen fiber architecture, where we constructed models representing normal stroma, high-risk tissue, benign tumors, and high-grade tumors. These were seeded with normal and ovarian cancer cell lines to investigate the separate roles of the cell type and matrix morphology on migration dynamics. The primary finding is that key cell–matrix interactions such as motility, cell spreading, f-actin alignment, focal adhesion, and cadherin expression are mainly determined by the collagen fiber morphology to a larger extent than the initial cell type. Moreover, we found these aspects were all enhanced for cells on the highly aligned, high-grade tumor model. Conversely, the weakest corresponding responses were observed on the more random mesh-like normal stromal matrix, with the partially aligned benign tumor and high-risk models demonstrating intermediate behavior. These results are all consistent with a contact guidance mechanism. These models cannot be synthesized by other conventional fabrication methods, and we suggest this approach will enable a variety of studies in cancer biology.According to the American Cancer Society, ovarian cancer ranks fifth in cancer deaths among women in the United States. If detected in early stages, the five-year relative survival rate is 92% [1]; however, less than 20% of the cases are detected at a localized stage [2]. This is due to both the prevalence of non-specific symptoms and also the lack of early sensitive and specific screening/imaging methods to detect small tumors before they become metastatic [3,4,5,6,7]. This is an especially critical problem for ovarian cancer, as the primary metastatic mechanism is exfoliation from the surface epithelium to the peritoneum, which can occur during early disease stage [8,9,10,11]. Importantly, the five-year survival rate for high grade metastatic disease is ~25%.A better understanding of the composition in the tumor microenvironment (TME) in this cancer could potentially lead to development of effective biomarkers and new therapies [12,13,14]. For example, while the dynamic interplay between cells and the extracellular matrix (ECM) influences differentiation, proliferation, and migration in both normal and tumor cells [12], it is not well understood how tumor growth depends on alterations in the matrix composition or morphology. Specifically, while migration is a hallmark of all cancers, it has not been well studied in ovarian cancer with respect to ECM remodeling. The lack of mouse models that represent human disease further complicates this problem, where these have been mainly limited to xenografts [15]. New mouse lines have been developed with specific mutations representing human disease, but their in vivo use lies in the form of implantation and monitoring subsequent tumor growth [16,17]. Thus, while promising, this application represents metastasis rather than primary disease in the fallopian tubes or ovary. There remains a clear need for biomimetic in vitro models that represent stromal alterations of the ovarian TME that allow hypothesis testing of the roles of the tumor cells and ECM morphology in disease progression.A necessary first step to create such models is accurately characterizing the underlying stromal changes. The majority of the normal ovarian stroma comprises collagen type I with nonspecific fiber orientation/alignment, other minor matrix proteins, and stromal cells (e.g., fibroblasts and myofibroblasts) [18]. Importantly, alterations in Col I architecture and loss of ECM components (e.g., collagen IV) have been associated with ovarian carcinogenesis, where these likely extend throughout disease progression [19,20]. Additional changes in other components including fibronectin and laminin have also be documented [19].Our lab has focused on examining changes in the collagen I stromal architecture using Second-Harmonic Generation (SHG) imaging microscopy. This 3D modality has great specificity/sensitivity for visualizing changes in fiber morphology as well as extracting underlying aspects of collagen architecture, for example average fibril size and macro/supramolecular helical attributes [21,22]. We have specifically examined such changes in a spectrum of human ovarian tumors and used these alterations as components in classification schemes [23,24,25,26]. For example, we have shown that six classes of ovarian tissues (e.g., normal and tumors) can be quantitatively differentiated using machine learning techniques based on the respective collagen morphology [24,25]. An overarching conclusion of all our studies is that the remodeling in high-grade serous ovarian cancer (HGSOC) is in the form of newly and improperly made collagen, creating a reactive stroma. We have further showed that these alterations are mostly present in the first 200 µm below the surface epithelium where the collagen is the most dense [26]. These observations form the rationale for creating tissue engineered scaffolds of the Col I architecture near the surface epithelium to study the corresponding cell biology. This is further supported by the fact that the large majority (~85%) of human ovarian cancers arise from the surface epithelium [27,28,29,30], and that the primary metastasis mechanism is exfoliation from the surface to the intra-peritoneal cavity.Collectively, these observations lead to questions of the role of the collagen alterations on migration, cytoskeletal dynamics, and proliferation. However, there has been a lack of in vitro cancer models that can incorporate the native collagen fiber morphology in ovarian and other cancers. For example, self-assembled gel models have been successfully used to demonstrate the influence of collagen fiber size and density on migration persistence in breast carcinomas [31]. We have also used analogous models to examine the role of different collagen isoforms (Col III and V) on collagen assembly [22,32]. However, these gels have limited control in terms of fiber length, alignment, and spacing, resulting in models that poorly replicate the topography of the stromal microenvironments in different classes of tissues. Soft and hard photolithographies provide more control in terms of spatial resolution; however, the use of masks limits the replication of complex 3D collagen structures. Flow chambers and microfluidic approaches have identified important ECM species involved in migration and extravasation dynamics, but these models cannot replicate the in vivo collagen morphology [33,34,35,36].To address these limitations, we developed a microscope-based system that utilizes multiphoton excited (MPE) photochemistry to synthesize in vitro biomimetic models [37,38,39,40,41]. The method is akin to 3D printing but produces submicron feature sizes and can utilize collagen and its analogs to reproduce the complex fiber morphology of the ovarian ECM. The fabrication resolution, or minimum feature size, is about 0.5 µm in diameter, which makes it a powerful tool to reproduce the native fiber widths and lengths [42]. We have previously shown that simple MPE fabricated ECM patterns (e.g., collagen IV, laminin, and fibronectin), and concentration gradients thereof, govern cell migration dynamics of different cell types, including breast and ovarian cancer, fibroblasts, and mesenchymal stem cells [43,44,45,46].More recently, we introduced an image-inspired approach to study normal ovarian epithelial IOSE cells on scaffolds representing human tissues and simplified models thereof [46]. Specifically, we examined migration dynamics by decoupling fiber shape and fiber alignment and found that, while both contributed to the overall response, the highly periodic fiber shape in HGSOC greatly promoted motility. We now extend that study to compare the response of ovarian cancer cells on image-based scaffolds, where we also analyze additional markers and begin a mechanistic study to decouple the respective cell and matrix contribution to migration dynamics. These studies may identify new diagnostic/prognostic targets based on the collagen fiber structure.The fabrication instrument was a purpose-built multiphoton microscope and has been described in detail previously [47]. A mode-locked titanium sapphire femtosecond laser (Mira; Coherent, Santa Barbara, CA, USA) was integrated with an upright microscope stand (Axioskop 2, Zeiss, Thornwood, NY, USA). Scanning was performed by two galvo mirrors (Cambridge Technologies, Bedford, MA, USA) for individual fields of view, which were then tiled together by a 3D motorized stage (Ludl Electronic Products Ltd., Hawthorne, NY, USA) to fabricate larger structures for cell analysis. A wavelength of 740 nm was used for two-photon excitation of the photo-initiator (see below), with approximately 100 mW average power at the plane of focus, where this was controlled by a 10 KHz electro-optic modulator (EOM, Conoptics, Danbury, CT, USA). We previously showed that the minimum feature sizes for crosslinked protein agreed with the theoretical resolution. For example, using 0.75 NA and 740 nm two-photon excitation, the respective lateral and axial resolutions were ~600 nm and 1.8 µm [42].A key technology challenge in creating the stromal models is developing the scan control mechanism that can match the relative collagen concentration between each pixel in the image to that in the fabricated construct. Due to the inertia of the scanning mirrors, random scanning is not an optimal method of controlling the laser exposure. Instead, we implemented an approach we termed modulated raster scanning. Here the galvo mirrors were raster scanned (as in conventional or multiphoton microscopy), and the laser was rapidly shuttered by a second higher-speed EOM (maximum 100 MHz, Conoptics). The “open” fraction in each pixel (~10 microseconds) was mapped to the grayscale level (bits) of the corresponding pixel in the original SHG image. Thus, increased laser exposure linearly corresponds to in increased crosslinked collagen concentration [47].We used the water-soluble Irgacure Sodium 4-[2-(4-morpholino)benzoyl-2 dimethylamino]butylbenzenesulfonate (MBS) as the photo-initiator. This moiety is non-toxic and has comparable efficiency to conventional vinyl photo-initiators soluble in organic solvents. MBS was synthesized in house using published protocols [48], where the properties were validated through standard spectroscopy characterization. Two-photon excitation of MBS (1 nM) at 740 nm drives a photochemical reaction that makes benzoyl and alpha amino alkyl radicals. At the focal volume, reactive radicals interact with the collagen and collagen analogs (described below), where the resulting radicals attack a second molecule inducing a chain crosslinking reaction and eventual termination [49].For all scaffolds, we utilized a combination of solubilized 80% gelatin methacrylate (GelMA) + 20% Col I (150 mg/mL GelMA + 10 mg/mL type I collagen) as the starting material. GelMA was prepared from well-established protocols without further modification [50]. While it would be ideal to use all collagen I, solubility and pH issues make this approach difficult. GelMA is often used as a collagen substitute in tissue engineered scaffolds as it is biomimetic [51,52], and its use makes the photochemistry more facile. While the single-stranded GelMA presents RGD cues, GFOGER is the relevant binding site in the collagen triple helix, and the GelMA/collagen mixture then offers both the structural integrity and proper binding cues of the former and latter, respectively.The substrates for the scaffolds were prepared first with a rubber hybridization chamber (~200 micron volume) secured to a silanized microscope slide [43]. To prevent non-specific adsorption, a monolayer of 30 mg/mL bovine serum albumin (BSA) was formed on the slide before the fabrication solution (MBS and GelMA/collagen) was applied [43]. The solution was kept strictly below 40 °C to avoid self-polymerization. Scaffolds were kept in phosphate-buffered saline (PBS) until cell seeding. Following the fabrication, un-crosslinked starting materials were dissolved away in a water bath at 37 °C.Three ovarian epithelial cell lines of varying characteristics were used in this study: HEY (highly metastatic; from Dr. Molly Brewer, UCONN Health Center), OVCA433 (moderately metastatic; co-author Dr. Manish Patankar, WI, USA), and Immortalized Ovarian Surface Epithelium (IOSE; from Dr. Molly Brewer, MD, UCONN Health Center) as the normal control [53,54]. These three cell lines were cultured at 37 °C with 5% CO2 in DMEM/F12 medium base (LifeTechnologies 11330, Carlsbad, CA, USA) supplemented with 10% FBS (LifeTechnologies 10082).Following fabrication and prior to cell seeding, the scaffolds were sterilized with 1X PBS containing 100 U/mL penicillin–streptomycin (Invitrogen 15140-122, Carlsbad, CA, USA). All cell lines were seeded at a density of 50K cell/mL and incubated overnight. Time-lapse studies were then performed by phase-contrast imaging (Nikon Ti-Eclipse inverted microscope with Pathology Devices, Inc., LiveCellTM incubator system). Phase-contrast images (10×, 0.25NA objective) of each seeded scaffold were collected at 30 min intervals over 72 h. Cells became too densely populated at longer times to isolate the cell–matrix interactions, and collective migration events were not tracked. At least three independent measurements were used for each cell/scaffold combination with 60–80 attached cells in each case.Tracking of cell migration was performed with Imaris (v7.6.5, Bitplane AG). For each fabricated scaffold, at least 20 cells were tracked for statistical significance. The resulting trajectory was directly exported to a spreadsheet and then analyzed in self-written code in MATLAB (Mathworks, Natick MA). This code outputs (i) cell position, (ii) the instantaneous speed, (iii) direction of the migration, and (iv) mean square displacement (MSD; d2(t)). Motility coefficients (µ) were then determined by applying non-linear least-squares regression modeling of the MSD measurement to [55]:(1)〈d2(t)〉=2ndμ[t−P(1−e−tp)]
+        where P is the directional persistence time, and nd is the dimensionality and equals 2 here. Cell shape characteristics (spread area, circularity) were determined with ImageJ software.The ovarian cells were grown on the scaffolds between 16 and 24 h prior to staining for actin stress fibers, focal adhesions, and N/E-cadherin. For actin staining, the cells were fixed with 4% paraformaldahyde in PBS for 15 min. Following two washes with 1× PBS, the cells were permeabilized with 0.3% Triton X-100 for 10 min and stained with Texas Red conjugated phalloidin for 30 min. Two-photon excited fluorescence images were collected using a 40× 0.8NA objective. This was done for both IOSE and OVCA433 cells, with cells analyzed for each scaffold. CurveAlign [56] was used to quantify the angular distribution of f-actin fibers for cells in a given pattern as well as the overall collagen alignment from the SHG images.To stain for focal adhesions, the cells were incubated with an anti-vinculin primary antibody (VIIF9 (7F9), mab 3574, Sigma-Aldrich, St. Louis, MO, USA) overnight at 4 °C, followed by incubation with a Texas Red secondary antibody (Mouse IgG (H+L), T862 1/EA, Invitrogen). Two-photon excited immunofluorescence images were collected using a 40× 0.8NA objective. This was done for both IOSE and OVCA433 cells with 20 cells analyzed for each scaffold. The number of focal adhesions per cell and integrated areas (following background subtraction) were determined in ImageJ.For cadherin staining, the cells were incubated with an anti-E-cadherin (mouse, ab1416, Abcam) and anti-N-cadherin (rabbit, ab18203, Abcam, Cambridge, UK) primary antibody (at 1:200 dilution) overnight at 4 °C, followed by incubation with Alexa Fluor 488 (goat anti-rabbit IgG (H&L), ab150077, Abcam) and Alexa Fluor 594 (goat anti-mouse IgG (H&L), ab150116, Abcam) secondary antibody, respectively, for 1 h at room temperature. Fluorescent images of each respective channels were collected using a 40× 0.75NA objective. This was done for both IOSE and OVCA433 cells with 30 cells analyzed for each scaffold. Corrected total cell fluorescence (CTCF) was determined using ImageJ by measuring the integrated staining density and subtracting the total background.Statistical analyses of migration data, cell shape data, focal adhesion, and cadherin staining were performed in Origin 2017 (OriginLab, Northampton, MA, USA) first using ANOVA, followed by two-sample t-test analysis. Watson’s U2 tests were performed on f-actin and collagen fiber distributions using Oriana (Kovach Computing Services, Pentraeth, UK) to calculate directional statistics of the distribution and mean direction. Pearson correlation coefficients between these distributions were also calculated to measure correlation of the stress fibers and the collagen fibers in the stromal models.To serve as blueprints for the scaffolds, we began with SHG images we previously collected and analyzed from normal ovarian tissues, high-risk tissues, benign tumors, and high-grade tumors, where these originated ~10 µm below the surface epithelium [23,24,25]. For statistical relevance, four images from each group were used in this study, where these were chosen at random from those properly classified by machine learning [25]. Figure 1A shows a representative SHG image of the collagen topography from each of the four groups. In general, the normal stroma has a mesh-like morphology with straight fibers, whereas the other tissues have varying degrees of alignment and periodicity [45].As fibers can overlap with the focal volume, we used image processing techniques (e.g., eigenvalues of the Hessian matrix, thresholding, and tubeness) to discretize fiber structures from the SHG images, where the resulting images were used as the design templates. Figure 1B shows the resulting fabricated structures, where the scaffolds were stained with rhodamine B for contrast and imaged via two-photon excited fluorescence. Fidelity of the fabricated structures relative to the discretized model was 90% or higher to their respective template, where this was obtained by co-localization of both the spatial pixel overlap and the respective grayscale intensities between the model of the image data and the fabricated structure [47]. Immunofluorescence confirmed that Col I was incorporated in the GelMA + Col I scaffolds [46]. For the studies to follow, the scaffolds comprised 3 × 3 repeats of the same 200 × 200 × 10 µm pattern, for an overall size of 600 × 600 µm to yield sufficient area to simultaneously monitor multiple cells.We investigated how the highly different collagen topographic patterns in the four classes influenced cell migration dynamics. In order to decouple the cell and matrix contributions, we used three cell lines of varying metastatic potential—IOSE, OVCA433, and HEY—where the first represents a “normal” immortalized ovarian surface epithelial cell, and the latter two are moderately and highly metastatic ovarian cancer cells, respectively. Figure 2 shows representative phase contrast images of IOSE (Figure 2A) and OVCA433 (Figure 2B) cells on a high-grade pattern. Migration on the scaffolds was measured for 72 h, and trajectories were mapped as described in the methods.Figure 3 shows representative migration trajectories of IOSE, OVCA433, and HEY cells (~20 in each case) on normal (Figure 3A) and high-grade (Figure 3B) models, respectively. For each cell type, we observed longer trajectories on the high-grade model relative to the normal stroma. For example, the majority of the IOSE trajectories on the latter were localized within 5 µm, where these ranged up to 200 µm on the cancer model. The cancer cells followed the same trend, although we note that the HEY cells were highly migratory in comparison to IOSE and OVCA433 lines, and the former had longer track lengths on the respective model. This agrees with our prior work studying migration of these cells on crosslinked gradients [57]. These results indicate that for all cell types, the highly aligned fibers promote cell migration to a larger extent than the more random structure provided by normal stroma.To quantify all trajectory data, we first determined the instantaneous velocity and motility coefficients. Figure 4A shows the average velocity values for the three cell lines on the four scaffolds. First, we note that the highly metastatic HEY cells were the fastest on each scaffold, which was consistent with our previous data using simple linear models [45,58]. In addition, we note that instantaneous speed did not change significantly for each of the cell lines on the different morphologies, with the only differences lying between the mesh-like normal and highly aligned high-grade models.We next examined the role of motility, i.e., the ability to migrate in one direction before changing direction, where this was determined through measuring the mean square displacement (MSD; see Equation (1), Methods Section 2.4). The averaged motility results for the three cell lines on the four structures are shown in Figure 4B. First, we can compare the cell behavior on the different morphologies. The highly polar HEY cells showed the highest motility in comparison to the other cell lines, but these also had the weakest scaffold dependence. When comparing the role of the collagen morphology, we found the lowest motilities on the normal stromal model, which had a fairly random alignment. In contrast, the more aligned fibers of the high-risk and high-grade models resulted in higher motility for each cell line. In sum, these measurements showed that both the cell phenotype and fiber architecture were important factors in the resulting migration, although the different cells followed the same overall trends on the same scaffold.We next examined cell shape on the four classes of image-based models to determine how the collagen architecture affects the resulting phenotype. We quantified the cell morphology using circularity measurements. This metric is given by 4πA2p2, where A and p are the area and perimeter, respectively, and lower values correspond to more elliptical shapes (more aligned), respectively.Figure 5 gives the circularity for the three cell lines on the four image-based structures. We found the same trends in each case, where, for example, the cells were the most aligned on the high-grade model (i.e., lowest circularity) and least on the normal stroma. Cells on the high-risk model had similar circularities to those on the high-grade model. We note that both these models shown in Figure 1 also had highly aligned collagen fibers. Similarly, cells on the normal and benign tumor models had similar circularities, consistent with the comparable random fibers from the SHG images (although the benign tumor had thicker, more fibrotic like fibers). Overall, the OVCA433 cells had lower values of circularity than the IOSE, which likely was due to the more polar initial cell shape. The absolute values for the HEY cells were higher than the other cells types. As these cells showed the least scaffold-dependent motility, lower alignment may be expected. In sum, while the absolute values of the circularity varied between the cell types, the overall trends with significant differences for the four stroma models were the same in each case, where these are related to fiber alignment in the models.We next measured focal adhesion density on image-based structures to determine the extent that aligned fibers promoted enhanced expression. Here, we specifically stained for vinculin as it is a membrane-associated protein component of integrin-mediated adhesions that connects the cytoskeleton to the ECM [59,60]. Representative immunofluorescence (anti-vinculin) images for IOSE and OVCA433 cells are shown in Figure 6. These studies could not be conducted with HEY cells as they do not form discrete focal adhesions and the vinculin staining is diffuse [44].We quantified the focal adhesion density in terms of number of adhesions per cell area, and the results are shown in Figure 7. For the IOSE cells (Figure 7A), the focal adhesion expression was significantly higher on high-grade structures in comparison to the other scaffolds. These results suggest that the aligned crimping pattern promotes focal adhesion formation. In contrast, the expression was lowest on the linear, mesh-like fibers in the normal model. The OVCA433 cells (Figure 7B) showed very similar behavior, where these results indicated the matrix morphology was the dominant factor in the cell response, rather than the initial cell type.We next analyzed the spatial distribution of cellular stress fibers relative to the collagen morphology of the different stromal models. Representative images of fluorescence images (phalloidin) for IOSE and OVCA433 cells are shown in Figure 8. These studies could not be conducted with HEY cells as the actin filaments form only shorter segments, compromising the analysis.We quantified the alignment of stress fibers and collagen fibers through Pearson correlation coefficients between the respective radial distributions (Table 1). Both cells types showed similar responses on the same respective matrix. Specifically, the highest correlations were on the high-grade (highest) and high-risk structures, and the lowest was on the benign model. Therefore, we conclude the highly aligned stromal architectures of the former promote stress fiber alignment. These results are consistent with the migration (Figure 4B) and cell shape (Figure 5) data, where both cells displayed higher motility on the high-grade models, as well as the lowest values of circularity (higher alignment). Lastly, we used the Watson’s U2 test to examine the distribution of the stress fibers for each cell/scaffold combination. For each cell type, the cells on the high-grade and normal models had the narrowest and broadest distribution, respectively. In analogy to the focal adhesion expression, these results indicate the matrix morphology drives the actin cytoskeleton characteristics of distribution width and alignment with respect to the collagen fibers.To begin to investigate the mechanism of the shape/cytoskeleton changes, we performed cell shape analysis following Rho-associated protein kinase (ROCK) inhibition (iY27632 at 10 µm). This treatment should result in further spreading and polarization as ROCK itself enhances contraction. To quantify the elongation, we measured the angle of the cells relative to the fiber axis, where lower angles correspond to increased alignment [61]. As we found the largest differences in motility, cell shape, and cytoskeleton expression/alignment for cells on the normal and high-grade models, we chose these scaffolds for this analysis. The alignment data for OVCA433 cells on normal and high-grade matrix are shown in Figure 9A,B, respectively (~n = 40 cells/scaffold). The only significant shape change in response to ROCK inhibition for cells on the normal matrix was after 24 h of adhesion, where increased alignment was observed. The data on the high-grade matrix are in strong contrast, where increased alignment (lower angle) was observed at all timepoints after 2 h. This increase may be due to an underlying contact guidance mechanism provided by the aligned, wavy fibers presented by this model. This would be consistent with both increased focal adhesion density and stress fiber alignment relative to the fibers.Modulation of both E-cadherin and N-cadherin expression has been reported in HGSOC [62,63,64,65]. Here, we quantified their relative scaffold-dependent expression to determine if/how these are influenced by the stromal morphology. Representative immunofluorescence images for IOSE and OVCA433 cells are shown in Figure 10. HEY cells exhibit negligible E-cadherin expression and were not used for the analysis. For quantification of the relative expression of individual cells on each scaffold, we measured the immunofluorescence intensity ratio relative to cells off the patterns (normalized to the same area; ~n = 30 cells/scaffold).Figure 11A shows the relative immunofluorescence intensity ratios of both E-cadherin and N-cadherin for IOSE cells on image-based patterns. The E-cadherin expression was significantly lower on the normal pattern in comparison to the other models, indicating that the more aligned collagen fiber architecture of the benign, high-risk, and high-grade scaffolds promotes its expression. Overall, the N-cadherin expression on the respective scaffold mirrored that of E-cadherin. Similar trends were found for the OVCA433 cells (Figure 11B); however, the largest differences were found between cells on the high-grade model relative to the rest, which were mostly not statistically different. This is similar to the motility results shown in Figure 4B, where the IOSE cells had a stronger scaffold-dependent motility than the OVCA443 cells.A deep understanding of the cell–stromal interactions in ovarian cancer is critical to the development of better diagnostics as well as assessments of treatment efficacy. This is a critical issue for high-grade disease as it can metastasize via exfoliation from the surface epithelium while lesions are still small [8,9,10,11]. However, studies of cell migration have been limited by the lack of biomimetic models [66]. While the role of collagen alignment in cancer (especially breast cancer) has been studied with different methods [31,67,68,69], it has not been possible to reproduce all aspects of the complex collagen architecture, specifically the fiber lengths, widths, and packing into the overall stromal morphology. The MPE fabrication approach is well matched to this task as we can exploit the freeform nature to create image-based scaffolds of different classes of tissues. Moreover, the cell–matrix interactions can be studied with cells of differing characteristics. As the ovary can be either the primary site or first metastatic site (from the fallopian tubes) in this cancer [70], we chose to model the surface of the ovary for these studies. We focused the analyses on migration and migration-related structural aspects (cell morphology, focal adhesion expression, and f-actin fiber alignment) as these processes are highly mis-regulated in ovary relative to normal tissues [45,46,57].The primary finding of our study is that increased fiber alignment and crimping morphology of the high-grade stroma models enhance motility for all cell lines. In addition, the overall trends in the other migration-based metrics were similar for the normal IOSE and cancer cell lines on the respective four scaffolds. Specifically, the motility, cell alignment, f-actin alignment, and focal adhesion expression were highest on the highly aligned high-grade model and the lowest on the random mesh-like normal stromal models. Collectively, the similarity in response for a normal ovarian line and high-grade ovarian cancer cell lines on the same respective scaffold indicate the dynamics are governed to larger extent by the matrix morphology than the initial cell type. We note that there are some differences in absolute values in the migration-related metrics between the cell lines. For example, the IOSE cells showed the largest decrease in circularity on the aligned structures. This may be due to the larger cell size, which can interact with more fibers at the same time. This is analogous to prior work by Yamada, where they showed greater elongation of fibroblasts on closely packed 1D lithographic stripes than those further apart [71]. Overall, the OVCA433 cells had lower values of circularity, which likely is due to the more polar initial cell shape. The circularity values for the HEY cells were higher, where this may be attributed to the smaller cell size interacting with fewer fibers. Moreover, these cells show a larger distribution of cell shapes, which likely is due to their fast migration without organized stress fibers and focal adhesions.One second commonality in findings is that the highly metastatic HEY cells had the fastest migration and highest motility, and these properties were only weakly substrate dependent. This may be due to these cells having already undergone an epithelial-to-mesenchymal transition (EMT) and migrate with a different mechanism than the IOSE or OVC433 cells. For example, these cells do not express discrete stress fibers or focal adhesions. We had also observed these differences using crosslinked concentration gradients [57].We can discuss the putative migration mechanism in terms of contact guidance, where this process describes cell migration in response to anisotropic physical features of the ECM [72,73]. For example, cells can elongate and migrate when sensing local changes in substrate concentration or morphology. Contact guidance has been investigated in terms of ROCK signaling, where this well-known serine-threonine kinase acts on the cytoskeleton, regulating cell shape and migration via actomyosin contractility [74]. Importantly, this specific mechanism has been reported to play an important role in cancer metastasis [75]. For example, using ROCK inhibition, Provenzano et al. showed that an aligned collagen matrix (self-assembled gel model) provided contact guidance cues that significantly enhanced cell motility [73].In our work, we drew upon such previous reports and performed a ROCK inhibition study to assess if contact guidance was implicated in the greater motility of ovarian cancer cells seeded on high-grade scaffolds (Figure 4). In Figure 9, we showed that the relative alignment of OVCA433 cells subjected to ROCK inhibition (reducing actomyosin contractility) was more influenced by the more highly aligned collagen fibers of the high-grade scaffolds in comparison to the normal model. Moreover, the crimping pattern in the former provides additional anisotropy and should further promote contact guidance, which is further consistent with both increased focal adhesion density (Figure 7) and stress fiber alignment (Table 1). We can also relate the relative N- and E-cadherin expression on these models to that observed in previous studies of ovarian cancer progression [63,76]. The cadherin extracellular domain provides cell–cell signaling, while the intracellular domain connects with the actin cytoskeleton by associating with catenins [77]. We observed that the relative expression of both E- and N-cadherin was significantly higher for IOSE and OVCA433 cells on the high-grade stromal model relative to the normal. While an EMT switch has been documented in many epithelial cancers, including ovarian, increased E-cadherin has also been associated with disease progression in HGSOC [63,76,78]. Relatedly, up-regulation of N-cadherin has been shown to promote cellular migration and increase motility [79,80,81]. These literature findings are consistent with our results on motility (Figure 4) and N-cadherin expression (Figure 11) for cells on the high-grade model. It is important to note that we are comparing the respective expression of each cadherin on the four tissue scaffolds, but not to each other, and thus we are not quantifying an EMT switch. Lastly, we performed staining for β-catenin and found a slight increase in expression for cells on the high-grade model, but the differences were not significant.While we performed this detailed study on migration in ovarian cancer, it is insightful to compare the findings to those reported in the better studied breast cancer. For the latter, it has been shown that changes in collagen architecture (specifically aligned fibers) enhanced the motility of cancer cells in vivo and ex vivo [31,82,83,84,85,86]. For example, breast cancer cells on aligned collagen in microchannel models displayed enhanced and persistent migration [87,88,89,90]. Additionally, using a breast xenograft model, Condeelis found that cancer cells on parallel collagen fibers displayed highly directed migration [86]. We point out that normal breast stroma is mostly characterized by wavy fibers that become straighter in invasive cancer [91]. However, the shape in ovarian cancer is in the opposite direction, and we found greater motility for all cell lines on wavy fibers (high-grade model) over straight fibers of normal tissue. Thus, because the form of collagen remodeling may be different between cancer types, biomimetic models are needed to decouple the respective cell and matrix morphology contributions to the migration dynamics. We further suggest the MPE fabrication approach is broadly applicable to studying this class of problems across disease states.Using multiphoton excited fabrication, we constructed image-based models of ovarian tissues. This technique is superior to other fabrication methods as the complex morphology of the collagen visualized by SHG microscopy can be recapitulated with high fidelity. Moreover, this approach affords hypothesis testing of respective cell and matrix morphology contributions to the migration dynamics. The key finding is that cell characteristics such as motility, cell shape, f-actin alignment, focal adhesion expression, and cadherin expression are mainly determined by the collagen fiber morphology to a larger extent than the initial cell type. Specifically, while the absolute values for these metrics were different for the three cell lines, the overall trends in the relative response to each scaffold were similar. Notably, we found enhanced motility and cell/cytoskeletal alignment on the highly aligned high-grade model. Conversely, the weakest corresponding responses were observed on the more random mesh-like normal stromal matrix. We suggest the scaffolds can be used for further cell biological studies and as platforms for testing of drug efficacy.Conceptualization, Manish Patankar; Data curation, S.A., R.B. and H.S.; Formal analysis, S.A.; Funding acquisition, M.P. and P.J.C.; Investigation, S.A.; Resources, D.H. and R.H.G.; Supervision, P.J.C.; Writing—original draft, S.A.; Writing—review & editing, P.J.C. All authors have read and agreed to the published version of the manuscript.This research was funded by National Cancer Institute: 1R01CA206561-01, National Cancer Institute: R01CA232517-01, and National Science Foundation: DMR-1610345.P.J.C. and M.P. gratefully acknowledge support by the Rivkin Center for Ovarian Cancer and NIH 1R01CA206561-01 and R01CA232517 -01. RHG acknowledges support under NSF, DMR-1610345. We thank Visar Ajeti for the initial design.The authors declare no conflict of interest. Ovarian stromal images and corresponding fabricated scaffolds. (A) Second-Harmonic Generation (SHG) optical sections of collagen from the four categories of ovarian tissues. (B) Two-photon excited fluorescence images of the resulting respective scaffolds. Each pattern is 200 × 200 µm in size with 10 µm in height. Scale bar = 50 µm.Phase-contrast images of IOSE (A) and OVCA433 (B) cells on a high-grade pattern. Scale bar = 150 µm.Representative trajectories of ovarian cell lines on image-based normal (A) and HGSOC (B) stromal models over 72 h. There were approximately 20 cells tracked in each case. Trajectories for these cells on the benign and high-risk models are not shown for simplicity.Migration dynamics for the three cell lines on scaffolds representing the four tissue types. (A) Instantaneous cell migration speed. (B) Motility coefficients. Migration was tracked over 72 h (* p < 0.05).Cell circularity for IOSE (A), OVCA433 (B), and HEY (C) cells on image-based patterns (* p < 0.05).Representative two-photon excited immunofluorescence images of IOSE and OVCA433 cells stained with a primary anti-vinculin antibody and secondary antibody conjugated with Texas Red. Scale bar = 30 µm.Focal adhesion density of IOSE (A) and OVCA433 (B) cells on image-based structures (* p < 0.05).Two-photon excited fluorescence of f-actin filaments of IOSE and OVCA433 cells stained with Texas Red (phalloidin) on image-based patterns. Scale bar = 30 µm.OVCA433 cell alignment on normal (A) and high-grade (B) stromal models in response to ROCK inhibition. Lower angles correspond to higher alignment, and significant differences were found at all time points after 2 h on the high-grade matrix. (* p < 0.05).Representative immunofluorescence images of IOSE and OVCA433 cells stained with primary anti-E-cadherin and anti-N-cadherin antibodies and a secondary antibody conjugated with Alexa Fluor 488 and Alexa Fluor 594, respectively. (A) Normal and (B) high-grade models. Scale bar = 50 µm.Immunofluorescence intensity ratio for cells on/off patterns of both E-cadherin and N-cadherin of IOSE (A) and OVCA433 (B) cells on image-based patterns (* p < 0.05).Pearson correlation coefficients between cellular f-actin fibers and stromal collagen fiber distributions of IOSE and OVCA433 cells for the four models.

Med-MDPI/cancers/cancers-12-06-01391.txt

@@ -0,0 +1 @@
+We evaluated the heterogeneity of the effect of known risk factors on breast cancer development based on breast density by using the Breast Imaging-Reporting and Data System (BI-RADS). In total, 4,898,880 women, aged 40–74 years, who participated in the national breast cancer screening program in 2009–2010 were followed up to December 2018. Increased age showed a heterogeneous association with breast cancer (1-year hazard ratio (HR) = 0.92, 1.00 (reference), 1.03, and 1.03 in women with BI-RADS density category 1, 2, 3, and 4, respectively; P-heterogeneity < 0.001). More advanced age at menopause increased breast cancer risk in all BI-RADS categories. This was more prominent in women with BI-RADS density category 1 but less prominent in women in other BI-RADS categories (P-heterogeneity = 0.009). In postmenopausal women, a family history of breast cancer, body mass index ≥ 25 kg/m2, and smoking showed a heterogeneous association with breast cancer across all BI-RADS categories. Other risk factors including age at menarche, menopause, hormone replacement therapy after menopause, oral contraceptive use, and alcohol consumption did not show a heterogeneous association with breast cancer across the BI-RADS categories. Several known risk factors of breast cancer had a heterogeneous effect on breast cancer development across breast density categories, especially in postmenopausal women.Breast cancer is the most common cancer in women worldwide, accounting for 24.4% of all types of cancer; moreover, breast cancer is the leading cause of cancer death in women [1]. In Asian countries, breast cancer incidence is lower; however, it has rapidly increased over the past decades. Changes in demographic factors associated with social and economic development, including lifestyle and reproductive risk factors, increase in breast cancer screening, and awareness in the region, have been attributed to this trend [1,2].Breast density is a non-modifiable risk factor for breast cancer and can be identified through breast cancer screening using a mammogram. High mammographic density increases the risk of breast cancer; women with a breast tissue density ≥ 75% have a 4–5 times higher risk compared with those with a density < 5% [3,4]. The effect of mammographic density on breast cancer risk has been investigated previously. Studies have suggested that breast density is associated with risk factors of breast cancer such as age, reproductive factors, and body mass index (BMI) [5,6,7]. Other studies investigated whether there is an interaction effect between breast density and the known risk factors on breast cancer. It has been shown that the effect of breast density can be increased or decreased by family history, reproductive factors, behavioral factors, and body mass index (BMI) [8,9,10,11,12,13].Most previous studies that investigated the interaction effect between breast density and other risk factors on breast cancer development considered other risk factors as effect modifiers and breast density as an independent variable of interest [8,9,10,11,12,13]. Considering a preventive approach in treatment, whether the effects of other breast cancer risk factors differ by breast density [14], needs to be assessed. In addition, the subjects of these studies included only women in the US, Japan, or France. Considering the differences in breast cancer epidemiology between countries, the interaction effect between breast density and other risk factors needs to be assessed in other populations.In Asian countries, known breast cancer risk factors are less prevalent; however, the prevalence of dense breasts is high [6] and screening mammography programs have only started in some countries [15]. Furthermore, the evaluation of the relationship of mammographic density and breast cancer risk, considering its interaction with other risk factors, has been limited. Therefore, in this study, we comprehensively examined the interaction effect between breast density, which is measured during mammographic screening, and various risk factors on breast cancer development by assessing the heterogeneity of the effect of these risk factors based on breast density in a large East Asian population for individualized risk assessment within the screening setting.Table 1 presents the distribution of breast cancer risk factors in participants with or without incident breast cancer according to the Breast Imaging-Reporting and Data System (BI-RADS) density category. Among 5,038,851 subjects, after excluding 141,345 women with missing BI-RADS density or with breast implants, 14.8%, 23.5%, 37.8%, and 23.9% of breast cancer incident cases were classified as BI-RADS density category 1, 2, 3, and 4, respectively. The corresponding values in controls were 27.3%, 27.4%, 30.1% and 15.2%.The associations between risk factors and the development of breast cancer stratified by breast density, and heterogeneity of the association between BI-RADS categories are presented in Table 2. For participants with BI-RADS density category 1, increased age was associated with a decreased risk of breast cancer (HR = 0.92; 95% CI = 0.90–0.93); no association was observed in participants with BI-RADS category 2 and an increased association was observed in participants with BI-RADS category 3 and 4 (HR = 1.03; 95% CI = 1.01–1.04, HR = 1.03; 95% CI = 1.01–1.06, respectively; P-heterogeneity between the four groups < 0.001). Advanced age at menopause increased breast cancer risk in all BI-RADS categories; this was significantly more prominent in participants with BI-RADS density category 1 (age > 51 years at menopause, HR = 1.46; 95% CI = 1.37–1.56) and less prominent in women with other BI-RAD density categories (HR = 1.28, 95% CI = 1.21–1.35; HR = 1.33, 95% CI = 1.25–1.41; and HR = 1.24, 95% CI = 1.12–1.38 for BI-RADS 2, 3, and 4, respectively; P-heterogeneity = 0.009). Other risk factors did not show a heterogeneous association with breast cancer across BI-RADS categories. Advanced age at menarche, post-menopausal status, and no use of hormone replacement therapy after menopause decreased breast cancer risk; a family history of breast cancer increased the risk irrespective of density category. Menopause at the age of 50–51 years compared with that at ≤ 49 years, use of hormone replacement therapy after menopause for ≥ 1 year compared with that for < 1 year, and higher BMI increased the risk of breast cancer in all density categories.When stratified by menopausal status, a heterogeneous effect of each risk factor on breast cancer based on the four density categories was not observed in premenopausal participants (Table 3). Despite the lack of heterogeneity, there was no association of BMI with breast cancer in participants with BI-RADS category 1 or 2. However, as BMI increased, the risk of breast cancer was also found to increase; this was more prominent in women with BI-RADS density category 4.In postmenopausal participants, in addition to age at menopause (> 51 years old, shown in Table 2), a family history of breast cancer, BMI ≥ 25 kg/m2, and smoking history showed a heterogeneous association with BI-RADS density categories (Table 4). Those with family history of breast cancer showed an increased risk of breast cancer with a HR of 2.08 (95% CI = 1.75–2.50), 1.72 (95% CI = 1.52–2.00), 1.59 (95% CI = 1.39–1.82), 2.22 (95% CI = 1.82–2.70) in BI-RADS category of 1, 2, 3, and 4, respectively (P-heterogeneity = 0.014). Compared with BMI < 18.5 kg/m2, those with BMI ≥ 25 kg/m2 had an HR of 2.38 (95% CI = 1.70–3.34), 2.36 (95% CI = 1.80–3.09), 1.64 (95% CI = 1.35–1.98), and 1.79 (95% CI = 1.40–2.29) for those in BI-RADS categories 1, 2, 3, and 4, respectively (P-heterogeneity = 0.008). Smoking increased the risk of breast cancer significantly in participants with BI-RADS density category 3 (HR = 1.20; 95% CI = 1.08–1.34); however, smoking did not show an association in other categories (P-heterogeneity = 0.001).To the best of our knowledge, this is the one of the largest population-based studies that has investigated the heterogeneity of the effect of risk factors on breast cancer development across BI-RADS density categories. Age and age at menopause (> 51 years vs. ≤ 49 years) showed a heterogeneous association with breast cancer across density categories. Among the participants who underwent menopause, a family history of breast cancer, BMI ≥ 25 kg/m2, and smoking showed a heterogeneous effect on breast cancer across the BI-RADS density categories. In premenopausal participants, risk factors did not show a heterogeneous association with breast cancer across the BI-RADS density categories.In the Korean population, age-specific breast cancer incidence increases up to the age of 45–49 years, and then decreases with further increases in age [16]. In the Korean population, age-specific breast cancer incidence increases up to the age of 45–49 years, and then decreases with further increases in age [16]. A higher incidence rate of breast cancer in women aged < 50 years is a distinct pattern in Asian countries where approximately 50% of breast cancers are diagnosed among women under 50 [2]. Considering that the cumulative proportion of women who experience menopause before 50 and 55 years of age is 46.0% and 90.3% [17], premenopausal breast cancer development is more common than post-menopausal breast cancer. Not only among Korean women but also among other Asian women, breast cancer incidence plateaus or decreases above the age of 50, which is in contrast to the continuous increase in age-specific breast cancer incidence rates with increased age among Western women [18]. In this study, > 80% of the participants with BI-RADS density category 1 were aged ≥ 50 years; 70% of the participants with BI-RADS density category 4 were aged 40–49. Considering the age-related decline in breast density [19], the association of age with breast density may reflect the age-specific breast cancer incidence rate in Korea. The result is in line with a previous study that showed that high breast density was related to the aggressiveness of breast tumors, especially in younger women [20].A previous study suggested that menopause is more strongly associated with breast density than age [21]. Several studies have investigated the association between mammographic density and breast cancer based on menopausal status and hormone replacement therapy. A study suggested that despite the absence of statistical significance, the strength of the association between density and breast cancer was different, with the strongest association in premenopausal women and postmenopausal women using hormone replacement therapy [8,22,23,24]. However, another study showed the opposite result; the effect of mammographic density was not modified by menopausal status or use of hormone replacement therapy [13]. Reciprocity of the interaction means that if A is an effect modifier of B, then B modifies the effect of A [25]; therefore, when the effect of breast density is significantly different based on other factors, we can expect that the associations of the factor would be different based on the breast density. Kerlikowske et al. showed that women with BI-RADS density category 3 and 4 using hormone replacement therapy had a higher risk of breast cancer than those who did not; this association was not observed in women with BI-RADS density category 1 and 2 [23], similar to our result. However, our results did not show a significant heterogeneity. Despite the lower proportion of women who had ever used hormone replacement therapy than that in a previous study in the United States [21], the heterogenous effects were comparable. More advanced age at menopause was associated with dense breasts [6] and an increased risk of breast cancer [26]; however, the association of age at menopause and breast cancer based on breast density has rarely been studied. In Korean women, the current mean age at menopause is lower than that in other countries such as the United Kingdom, Australia, or Japan but shows an increasing trend [17], which may possibly contribute to the increased breast cancer-related burden. In this study, despite a similar association, advanced age at menopause (>51 years) was associated more with breast cancer in women with low breast density than in women with dense breast tissue.BMI is a well-known risk factor for the development of breast cancer in postmenopausal women because of the aromatization of androgens into estrogens, but not premenopausal breast cancer where increased BMI may have a protective effect [27]. However, in this study, even premenopausal women with dense breasts (BI-RADS density category 3 and 4) had a slightly increased risk of breast cancer with increased BMI. The combined effect of breast density and BMI on breast cancer has been studied well. The suggested mechanism is that obesity-related insulin deregulation and the adipokine-associated inflammatory response may activate proliferation [28]. In this study, in postmenopausal women with BI-RADS density category 1, BMI had a more prominent effect on breast cancer; this effect decreased as the BI-RADS density category increased. If those with BI-RADS density category 4 had a higher baseline risk of breast cancer, then the added effect of BMI might be lower. Previous studies on the relationship of the effect of breast density and BMI showed inconsistent results; more evident interactions were observed in postmenopausal women [29] or premenopausal women due to a higher risk of estrogen receptor-negative cancer [30]. However, other studies did not find a relationship between breast density and BMI in breast cancer [31], irrespective of menopausal status [10]. In both developed and developing countries, the prevalence of obesity is increasing, especially in younger age groups in developing countries [32]. However, in Korea, prevalence of obesity, defined as BMI ≥ 25 kg/m2, has not increased in women. Especially in women aged 20–39 and 40–59 years old, the prevalence of obesity has decreased [33], which is different from the trend in other countries [32]. The unique pattern of changes in BMI in Korea might contribute to the differing results.Several studies have shown that the association between breast density and the risk of breast cancer is stronger in women with a family history of breast cancer than those without a family history [9,11,34]. The heritability of mammographic density within family members has been shown before [35,36]. A recent study suggested that genetic markers in mammographic density such as percent density and dense area show a shared genetic origin and biological pathways with breast cancer [37]. In this study, we observed the heterogeneous effect of a family history of breast cancer based on BI-RADS density category in postmenopausal women. In women with BI-RADS density category 4, those with a family history of breast cancer showed an increased breast cancer risk; however, in women with BI-RADS density category 3, the increment was less. Considering that the direction and strength of associations were comparable in all BI-RADS density categories, the significant P-heterogeneity may be due to the large sample size of the study subjects.Despite the well-known carcinogenic effect of smoking, there has been a debate on the association between smoking and breast cancer. However, recent studies demonstrated that smoking increased the risk of breast cancer [38], as well as postmenopausal breast cancer risk [39]. In this study, we did not observe an overall increased risk of smoking in all, premenopausal, or postmenopausal women. It could be attributed to the low prevalence of ever-smoking in Korean women in this study’s subjects, which was similar to that reported in previous studies [40,41]. However, in postmenopausal women, a heterogeneous effect of smoking was observed; smoking significantly increased the risk in women with BI-RADS density category 3 and marginally decreased the risk in women with BI-RADS density category 2. Smoking, due to its antiestrogenic effect on breast tissue, is associated with a decrease in breast density [42,43]. Further investigation is required to understand the effects of smoking on breast density and its association with the risk of breast cancer.The effect of the decrease in risk with increased age at menarche and a non-significant association of oral contraceptive use irrespective of the breast density category were comparable with findings in previous studies [26,44]. Lower use of oral contraceptives in the Korean population may be the cause of the non-significant association [45]. Light drinking is not associated with most types of cancer and only leads to a mild increases in the risk of breast cancer [46]. The non-significant association of alcohol consumption might be attributed to the low alcohol consumption among Korean women [47]. In Korean women, the prevalence of drinking, defined as ≥ 1 drink of alcohol per month during the last year, was around 40% [41] and the average alcohol consumption was 8 g/day [48].There are several limitations of this study. First, the study population comprised cancer screening examinees and their baseline characteristics may not be comparable with those of non-examinees. Considering that breast cancer screening participation rate in 2009–2010 was approximately 45% [49] and we included all breast cancer examinees, our study covered almost half of the target population. However, considering that women with lower levels of household income or education participate less in breast cancer screening programs [50], a possible selection bias still remains. Secondly, the BI-RADS density classification provides information to physicians regarding the likelihood of missing a lesion that may be masked by dense tissues; BI-RADS does not quantify breast cancer risk exactly [51]. The BI-RADS density categories were reported by radiologists at many different screening units. However, the inter-observer agreement on BI-RADS categories is substantial [52,53] and a mammographic education program has been conducted to standardize the performance of radiologists in Korea [54]. Furthermore, although the automated breast density and BI-RADS categories showed modest agreement, their association with breast cancer was similar [55], suggesting that the HRs in this study are robust. Thirdly, the information regarding all risk factors except BMI were obtained from self-administered questionnaires; thus, there may be a bias in the responses. However, this would lead to non-differential misclassification and the effects on the results would not be significant [56]. Regarding risk factors, due to a lack of information, some important risk factors for breast cancer, such as parity-related factors, could not be considered. Therefore, the residual confounding effect of unaccounted variables or variables with broad categories may affect the results. In addition, several risk factors, such as BMI, oral contraceptive use, or hormone replacement therapy use, can be changed during one’s lifetime; however, we considered these risk factors at a single time point at their screening during the year of 2009–2010. Fourthly, cancer incidence was identified using the ICD-10 code and catastrophic illness registry in the National Health Insurance Service (NHIS) database, which has the potential for misclassification. However, using the catastrophic illness registry is related to the reimbursement of co-payment, requiring relevant clinical information by the insurance administration. In addition, it covers more than 97% of cancer patients in the Korea Central Cancer Registry. Thus, bias regarding the definition of breast cancer incidence would be minimal. When we separated invasive breast cancer and ductal carcinoma in situ, the results were comparable with the original results, supporting the robustness of the results. The follow-up time of ≥8 years would not be enough to identify all breast cancers. Breast cancer risk assessment estimates mostly 5-year risk but long-term risk assessment for 10 years or more is also needed [57]. Thus, the follow-up period of this study represents an intermediate time point. Despite these limitations, our study has a number of strengths including the large sample size and prospective design. The follow-up period was > 8 years and enabled us to study the short- and intermediate-term effects of risk factors and breast density on breast cancer. In addition, the study population covers a large proportion of the female population in Korea and various risk factors interacting with breast density were considered together.The NHIS in Korea supports biennial health examinations for individuals aged ≥ 40 years. It also supports the screening of stomach, liver, colon, breast, and cervical cancer in the eligible population. Participants of health examinations and cancer screening are asked to complete self-reported questionnaires, which collect information on lifestyle factors, family history of chronic diseases and cancer, and reproductive factors. The information in the questionnaires and results of health examinations and cancer screening are collected through the NHIS screening database. All participants provided consent that allowed the transfer of information to the NHIS database.For breast cancer, NHIS supports biennial mammographic screening for women aged ≥ 40 years. Approximately 40% of women who are eligible participate in the screening; approximately 3 million mammographic screenings are performed each year. During screening, data on the BI-RADS density assessment and BI-RADS assessment categories were collected. Breast cancer incidence was defined via linking the database with the NHIS medical usage database to obtain the information related to either invasive breast cancer (C50) or ductal carcinoma in situ (D05) and catastrophic illness codes up to December 2018. Upon reviewing the study proposal and request to the National Health Insurance Sharing Service, the NHIS database was made available for research. The study proposal was approved by the Institutional Review Board of the Hanyang University College of Medicine (IRB no. HYI-18-175-1).We identified 5,317,312 women who participated in health examinations and breast cancer screenings between 2009 and 2010. Among these, women aged ≥ 75 and women who had a medical usage record for any type of cancer with catastrophic illnesses code before the date of breast cancer screening or within 3 months from the date of breast cancer screening (n = 278,461) were excluded. The remaining 5,038,851 were included in the analysis as subjects. Women who newly acquired medical usage record for breast cancer (C50 and D05) with a catastrophic illness code (n = 55,538) by December 2018 were defined as incident breast cancer cases. In addition, 141,345 women with missing information on breast density or with breast implants were excluded; the remaining 4,897,506 women including 54,164 breast cancer cases were included in the analysis for the analysis stratified by BI-RADS density categories.BI-RADS breast density was classified into four categories: 1, almost entirely fat; 2, scattered fibroglandular densities; 3, heterogeneously dense; and 4, extremely dense.Information on breast cancer risk factors was obtained from self-reported questionnaires. The questionnaire for health examination included smoking history in one’s lifetime, mean days of drinking per week during the last 1 year, and mean days of physical activity during the last one week. BMI was calculated based on the weight and height measured during the health examination. Reproductive factors, including age at menarche, menopausal status, age at menopause, use of hormone replacement therapy, and oral contraceptive, and family history of cancer in first-degree relatives were obtained from the questionnaire for cancer screening.The distribution of breast cancer risk factors as a whole and according to the BI-RADS density categories was presented as numbers and percentages. The effect of each risk factor on breast cancer was analyzed using the Cox proportional-hazard regression model adjusted for other risk factors. The results are presented as hazard ratio (HR) and 95% confidence intervals (CIs) for the total number of participants and stratified by BI-RADS density categories. The follow-up was considered from the date of breast cancer screening during the year of 2009–2010 up to December 31, 2018, date of death, or date of any cancer incidence (based on which date came first). Incident breast cancer was considered as an event; death, incidence of other types of cancer, and no incidence of cancer were censored. Additionally, the effect of each risk factor on breast cancer and the stratification by BI-RADS breast density categories were evaluated according to the menopausal status. Heterogeneity of the effect of each risk factor across the BI-RADS density categories was assessed using I2 statistics. Analyses were conducted using SAS 9.4 (SAS Institute, Cary, NC, USA) and R software (version 3.5.0) during March–December 2019.In summary, some known risk factors of breast cancer showed a heterogeneous effect on breast cancer across breast density categories, especially in postmenopausal women. The association of genetic risk factors or family history, known breast cancer risk factors, especially reproductive factors with breast density is under investigation. Among various hypotheses regarding the mechanism of effect of breast density on breast cancer risk, this research focused on the interaction effect. However, studies have suggested the mediation effect of breast density on the association between known risk factors and breast cancer [58,59]. The role of breast density in breast cancer risk in Asian women, including its mediation effect, needs to be considered in future studies.Conceptualization, B.P.; Data curation, J.Y.; Formal analysis, S.-E.L. and Y.S.C.; Investigation, S.-E.L., H.A. and Y.S.C.; Project administration, H.A. and J.Y.; Supervision, B.P.; Writing–original draft, B.P. and S.-E.L.; Writing–review & editing, B.P. All authors have read and agreed to the published version of the manuscript.This research was funded by National Research Foundation of Korea: NRF-2019R1H1A1079862. Korean Foundation for Cancer Research: 2019.This work was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT and Future Planning (NRF-2019R1H1A1079862) and a research grant in 2019 by the Korean Foundation for Cancer Research, Seoul, Republic of Korea.The authors declared no conflict of interest.Distribution of risk factors of breast cancer in subjects with or without incident breast cancer by mammographic density.* Women with missing Breast Imaging-Reporting and Data System (BI-RADS) density or with breast implants were included only in total.Association between known risk factors of breast cancer and breast cancer development by breast density, in consideration of their interaction on breast cancer risk.HR: hazard ratio; CI: confidence interval; * Adjusted for listed variables in addition to age.Association between known risk factors of breast cancer and breast cancer development by breast density in premenopausal women.HR: hazard ratio; CI: confidence interval; * Adjusted for listed variables in addition to age.Association between known risk factors of breast cancer and breast cancer development by breast density in postmenopausal women.HR: hazard ratio; CI: confidence interval; * Adjusted for listed variables in addition to age, age at menopause, and hormone replacement therapy use.

Med-MDPI/cancers/cancers-12-06-01392.txt

@@ -0,0 +1 @@
+The authors would like to make a correction to their published paper [1].The authors would like to change one incorrect sentence in reference [1].On page 9, in paragraph 2, the sentence “In colon cancer, CAFs release exosomes containing miR-92a-3p and promote invasion and chemotherapy resistance. miR-92a-3p directly binds to FBXW7 and MOAP1 thereby activating Wnt-induced EMT and mitochondrial apoptosis [89].” should be changed to “In colon cancer, CAFs release exosomes containing miR-92a-3p and promote invasion and chemotherapy resistance. miR-92a-3p directly binds to FBXW7 and MOAP1 thereby activating Wnt-induced EMT and inhibiting mitochondrial apoptosis [89].”The change does not affect the scientific results.The rest of the manuscript does not to be changed. The authors would like to apologize for any inconvenience caused. The manuscript will be updated, and the original will remain available on the article webpage.The authors declare no conflict of interest.

Med-MDPI/cancers/cancers-12-06-01393.txt

@@ -0,0 +1 @@
+Targeted agents have improved the efficacy of chemotherapy for cancer patients, however, there remains a lack of understanding of how these therapies affect the unsuspecting bystanders of the stromal microenvironment. Cetuximab, a monoclonal antibody therapy targeting the epidermal growth factor receptor (EGFR), is given in combination with chemotherapy as the standard of care for a subset of metastatic colorectal cancer patients. The overall response to this treatment is underwhelming and, while genetic mutations that confer resistance have been identified, it is still not known why this drug is ineffective for some patients. We discovered that cancer-associated fibroblasts (CAFs), a major cellular subset of the tumor stroma, can provide a source of cancer cell resistance. Specifically, we observed that upon treatment with cetuximab, CAFs increased their secretion of EGF, which was sufficient to render neighboring cancer cells resistant to cetuximab treatment through sustained mitogen-activated protein kinases (MAPK) signaling. Furthermore, we show the cetuximab-induced EGF secretion to be specific to CAFs and not to cancer cells or normal fibroblasts. Altogether, this work emphasizes the importance of the tumor microenvironment and considering the potential unintended consequences of therapeutically targeting cancer-driving proteins on non-tumorigenic cell types.Colorectal cancer (CRC) has been well studied over the years, leading to a relative understanding of the genetics involved in disease progression and the identification and validation of promising drug targets. For example, supplementing chemotherapy regimens with cetuximab, a monoclonal antibody that targets epidermal growth factor receptor (EGFR), is now the standard of care for the KRAS wild-type subset of metastatic CRC patients. However, such treatment offers only modest benefits. Many genetic alterations have been identified as sufficient to confer resistance to cetuximab such as mutations in KRAS, BRAF, and alterations in the phosphoinositide 3-kinase/phosphatase and tensin homolog (PIK3CA/PTEN) pathway [1,2]. However, the mechanism of an estimated 10–30% of patients with initial clinical resistance remains unknown [2,3]. While this drug has been extensively investigated in terms of its effects on cancer cells, how this targeted agent affects the surrounding tumor microenvironment is still unknown.In an era of precision medicine, biomarkers are used to optimize therapies and are thought to improve clinical endpoints. While the most studied and validated biomarkers are tumor cell intrinsic, the contributions of the surrounding microenvironment are increasingly well-recognized [4] and warrant further investigation. Cancer-associated fibroblasts (CAFs), the predominant cell type in the tumor stroma, have been implicated in various aspects of tumorigenesis including metastasis [5] and therapeutic resistance [6]. CAFs arise from multiple origins such as resident fibroblasts, epithelial cells, and distant bone marrow mesenchymal stem cells [7]. They lack mutations found within cancer cells [8] and are molecularly characterized by the expression of CAF-associated markers (such as αSMA, vimentin, fibronectin) [9]. They also express proteins fundamental to cellular processes, one of which is EGFR.CAFs share a common environmental niche with cancer cells and collectively encounter EGFR inhibition during cetuximab treatment. Previous studies have identified contexts where the secretomes of stromal cells including CAFs [10,11] and B cells [12] are changed in response to chemotherapy and radiation treatments to confer resistance to the surrounding cancer cells. The effect cetuximab has on CAFs and its potential implications on cancer cell drug response have not previously been investigated. Here, we detail our findings that cetuximab treatment causes patient-derived CAFs isolated from three human CRC tumors to increase the secretion of EGF, which subsequently leads to increased resistance of CRC cancer cells to treatment.The influence of cancer therapeutics on non-tumorigenic cells is often overlooked. Given that CAFs are known to be involved in many aspects of tumorigenesis including drug resistance, we investigated whether these cells expressed EGFR, the target of cetuximab. Immunofluorescence staining of surgical tumor resections from CRC patients showed EGFR co-localized with αSMA, a marker used to identify CAFs, suggesting that CAFs do express EGFR (Figure 1A). Furthermore, CAFs isolated from these patient tumor tissues and cultured in vitro (confirmed to be CAFs through the expression of CAF-associated markers; Figure 1B, Figure S1), also expressed EGFR (Figure 1C). In contrast to CRC cells (DiFi and LIM1215), inhibiting EGFR signaling with cetuximab treatment does not alter overall CAF viability (Figure 1D).We previously established an imaging-based methodology that allows one to study the influence of drugs on heterogeneous cell populations while distinguishing between cell types [13,14]. Briefly, we quantified live and dead cell counts over time and then fit the data to an exponential growth model to determine net birth, death, and growth rates. This allows for readouts of cellular dynamics across time as well as distinguishing between cytotoxic (increased death rate) and cytostatic (decreased birth rate) effects of the drug. When we applied this approach to co-cultures of patient-derived CAFs and cancer cells at a starting ratio of approximately 1:1, the presence of CAFs prevented cancer cell death, even at high concentrations of cetuximab (Figure 2A,B; Figure S2A). This result is comparable to in vitro cetuximab resistance due to KRAS mutations (Figure S2B) [15]. Furthermore, when the starting fraction of CAFs-to-cancer cells was increased, reminiscent of CRC clinical stromal percentages (Figure S3), we observed a stronger protective effect against cetuximab, as evidenced by an increased growth rate of the cancer cells (Figure 2C). A minimum population of approximately 30% CAFs prevented cetuximab-induced death of cancer cells. CAF-driven increased growth in the untreated conditions was not dependent on CAF proportion (Figure S4).Next, we sought to identify whether the CAF protective effect was dependent on physical cellular interactions or CAF-secreted factors. Keeping in mind that CAF secretomes may change in response to drug treatment, media were collected from untreated and treated CAFs. When added to cancer cell cultures, the conditioned media from cetuximab-treated CAFs (CMtx) provided more protection than untreated CAF conditioned media (CM) during cetuximab treatment (Figure 2D, Figure S5). This finding suggests that CAF secretomes change in response to cetuximab treatment, leading to the protection of cancer cells from the drug’s effects.In order to identify secreted factors specific to cetuximab-treated CAFs, a cytokine array was performed to compare CM versus CMtx (Figure 3A,B; Figure S6). Surprisingly, the only common differentially expressed cytokine in the treated vs. untreated CAF samples was epidermal growth factor, EGF. This cetuximab-induced increase in CAF secretion of EGF was confirmed via enzyme-linked immunosorbent assay (ELISA), with at least a two-fold increase seen for each patient-derived CAF line (Figure 3C). This secretion pattern was not affected by culture media (Figure S7) and was sustained across five days, which was the longest time point tested (Figure S8). Moreover, this CAF effect occurred regardless of the cancer cell mutations found within the tumors that the CAFs were isolated from (Table S1). We next wanted to verify whether this was a CAF-specific effect or a result seen across all cell types. CRC cell lines and patient-derived normal colon fibroblasts were found to secrete very low baseline levels of EGF and did not increase EGF secretion upon treatment with cetuximab (Figure 3C). Furthermore, EGFR inhibition by erlotinib, a small molecule inhibitor that binds to the intracellular tyrosine kinase domain, or treatment with oxaliplatin, a chemotherapy used to treat colorectal cancer, did not initiate increased secretion of EGF (Figure 3D). This suggests that increased EGF secretion by CAFs depends on cetuximab binding to the extracellular region of EGFR and not a general response to inhibition of the EGFR pathway or a general stress response.Standard culture media for 2D immortalized cancer cell lines do not contain EGF. However, when supplemented with increasing concentrations of EGF, cancer cell growth rates increased in proportion with EGF concentration at each cetuximab dose (Figure 4A, Figure S9A). We hypothesized that EGF-induced resistance to cetuximab results from sustained signaling through the MAPK pathway. In cancer cells, EGF stimulation increased levels of pERK1/2, whereas cetuximab treatment shut down this pathway, as evidenced by undetectable pERK1/2 levels. In contrast, co-treatment of cetuximab and EGF stimulation preserved MAPK signaling. Furthermore, pERK levels increased in correlation with increasing EGF concentrations (Figure 4B, Figure S9B, Figure S12A–F, Figure S12O,P). This shows that pERK1/2 is rescued with the addition of EGF, even in the presence of cetuximab.Patient-derived organoid models more accurately resemble patient tumors given their genetic and microenvironmental heterogeneity, although CAFs and other stromal cells are often not present. Culture media developed to support long term growth of 3D patient-derived organoids contain various supplements including EGF. Previous studies have warned about the potential bias these exogenous factors may impart in the context of drug response [16]. In order to translate our findings to a more physiologically relevant cancer model, we repeated our cetuximab and EGF experiments in KRAS wild-type patient-derived CRC organoids, ORG12620. When we lowered EGF concentration in the media (0.4 ng/mL from the previously defined 50 ng/mL), we restored cetuximab sensitivity in our CRC organoids (Figure 4C,D; Figure S12F) with no significant decrease in overall viability in the untreated condition after five days (Figure S10). Furthermore, the addition of EGF during cetuximab treatment preserved MAPK pathway activity with pEGFR, pHER2, and pERK levels mirroring baseline levels (Figure 4E, Figure S12G–J).To verify that EGF was the specific CMtx-factor that conferred resistance to cetuximab, we incubated CMtx with an EGF-neutralizing antibody (CMtx-EGF) (Figure S11), which led to cancer cell response to cetuximab through reduced cell viability. Specifically, cancer cells that were exposed to CMtx-EGF were re-sensitized to cetuximab at a level resembling baseline response (Figure 5A–C). The CMtx-induced resistance is likely to be due to sustained signaling through the MAPK pathway, as ERK is still active (Figure 5D, Figure S12K–N). This supports the hypothesis that EGF in the CMtx media is causing resistance, as similar results were observed in cancer cells treated with exogenous EGF and cetuximab (Figure 4).Molecular targeting agents have significantly impacted the treatment of cancer; however, a large portion of the protein targets are expressed not only in cancer cells, but also in other cell types. There is limited research being done to investigate potential phenotypic responses to targeted agents, especially effects other than viability, which may occur in cells throughout the body including the stromal component of the tumor microenvironment. We discovered CAF secretion of EGF is increased in response to cetuximab and the presence of exogenous EGF results in cancer cell resistance to cetuximab treatment. While our studies focused on cetuximab treatment, analogous results are anticipated with panitumumab, an alternative monoclonal antibody targeting EGFR, which has similar clinical efficacy and toxicity profiles to cetuximab [17]. Future work will investigate the underlying mechanism leading to CAF secretion of EGF during extracellular inhibition of EGFR. The observation that EGF can outcompete an EGFR antibody in cell models has been previously reported [16,18,19,20], however, a source of EGF secretion from the stromal microenvironment in response to cetuximab treatment has not previously been identified.The ratio of CAFs to cancer cells varies across patient tumors (Figure S3). If this ratio is low, it is likely that the concentration of EGF secreted by CAFs is not sufficient to rescue cancer cells from cetuximab treatment. However, as the tumor shrinks from cetuximab treatment, the ratio of CAFs will increase (since CAF viability is not affected—Figure 1C) and it is possible that EGF levels will be adequate for protection from cetuximab and therefore also be a cause of relapse to treatment. It has been shown that the stromal microenvironment changes over the course of cetuximab treatment. Of note, in patients with progressive disease, an increase in stromal abundance was observed when compared to baseline (i.e., prior to cetuximab treatment) [21]. This increase in stromal cells could be a culprit in treatment resistance, with the proportion of cancer cells to CAFs reaching a state where the secreted EGF levels are sufficient to sustain MAPK signaling in the presence of cetuximab. Combination treatment with MAPK inhibitors may be an attractive target to mitigate CAF-induced cetuximab resistance [22,23]. In recent years, there has also been a push to develop therapeutics that target CAFs [24]. We hypothesize that utilizing such drugs in combination with cetuximab may be another way to increase cetuximab efficacy. Furthermore, ongoing work is focusing on potential dual-targeting of receptor tyrosine kinases that may be activated in colorectal cancer cells in response to increased exogenous EGF.There have been multiple clinical studies looking at biomarkers for cetuximab response that may supplement the current genetic alterations used for treatment stratification [25]. The CMS4 subtype of CRC tumors, which are characterized by a high stromal density [3], have been found to be prognostic for poor response to anti-EGFR treatment [26,27]. Furthermore, when looking at plasma levels of EGFR ligands, an increase in EGF levels from two weeks post-cetuximab treatment compared to initial treatment levels were significantly higher in non-responders [28]. Another independent study also identified a significant increase in EGF serum levels after cetuximab treatment, which corresponded to disease progression [29]. These clinical observations support our findings that EGF can confer cetuximab-resistance, with our results homing in on the stromal microenvironment as a significant culprit.Most therapeutic agents used for treating cancer are given systemically and therefore have the potential to affect cells throughout the body. While this concept is considered extensively in the context of adverse side effects, the potential of one’s body contributing to better or worse overall response to the drug has just recently begun garnering attention. For example, microbiome composition is indicative of overall response to PD-1 based immunotherapy [30]. Our data suggest that CAF composition is important for cetuximab response, specifically highlighting EGF secretion by cetuximab treated CAFs as a previously unknown mechanism of resistance to anti-EGFR treatment in colorectal cancer.DiFi and LIM1215 cancer cell lines were obtained from Dr. Alberto Bardelli (University of Torino) and cultured in Dulbecco’s Modified Eagle Media (DMEM) (Corning Inc., Corning, NY, USA) and Roswell Park Memorial Institute (RPMI) (Corning Inc., Corning, NY, USA), respectively, supplemented with 10% fetal bovine serum (FBS) (Gemini Bio, Sacramento, CA, USA) and 1% penicillin/streptomycin (P/S) (Gemini Bio, Sacramento, CA, USA) under standard laboratory conditions (5% CO2, 37 °C).Tumor tissues were received from colorectal cancer patients under Institutional Review Board (IRB) approval at the Norris Comprehensive Cancer Center of the University of Southern California (USC). All subjects gave their informed consent for inclusion before they participated in the study. The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the IRB Ethics Committee of USC (Protocol HS-06-00678; approval date 08-02-2019). Known tumor mutations and treatment data are detailed in Table S1.Patient-derived metastatic colorectal tumor organoids were developed following previously described methods [31]. Briefly, tumor tissue was digested with 1.5 mg/mL collagenase (MilliporeSigma, Burlington, MA, USA) and 20 μg/mL hyaluronidase (MP Biomedicals, Irvine, CA, USA) for 30 min at 37 °C, then separated through a 100 μm strainer (Corning Inc., Corning, NY, USA). Isolated cells were cultured in 3D using basement membrane extract gels to allow for tumor organoid formation. CAFs were separated from the same digested tumor tissue by culturing a fraction of cells on plastic tissue culture plates and letting the fibroblasts grow out over 1–2 passages. Cells were then verified as CAFs via qPCR and immunofluorescence staining for common CAF markers: α-smooth muscle actin (αSMA) (Abcam, Cambridge, MA, USA), vimentin (VIM) (Abcam, Cambridge, MA, USA), fibronectin (FN1) (Life Technologies, Grand Island, NY, USA), and fibroblast specific protein (FSP) (MilliporeSigma, Burlington, MA, USA) (Figure 1B, Figure S1, respectively). For all experiments, primary CAFs were used between passages 2 and 8.Cells were seeded in four 384-well plates 24-h prior to treating with cetuximab (USC pharmacy, Los Angeles, CA, USA). On day 0, cells were treated with the drug at the desired concentration. Before imaging, cells were stained with 5 μg/mL Hoechst 33342 (nuclear dye) (Life Technologies, Grand Island, NY, USA) and 5 μg/mL propidium iodide (PI) (Life Technologies, Grand Island, NY, USA) to identify cells as live or dead, respectively. Individual plates were imaged on days 0, 2, 3, and 5 using the Operetta High Content Screening (HCS) system (PerkinElmer, Waltham, MA, USA)). Cells were then segmented based upon the nuclear dye using Harmony software (PerkinElmer, Waltham, MA, USA). In order to differentiate cell types in co-culture assays, morphological features were calculated and used to train a machine-learning algorithm to classify cells as either ‘CAF’ or ‘tumor,’ as described in Garvey et al. [13,14]. Propidium iodide intensity levels were calculated and cells were classified as ‘dead’ if their intensity was above the established threshold. Growth rates for each cell type were calculated as previously described [13,32] by fitting the live cell counts over time to an exponential growth model.When CAF cultures reached approximately 80% confluent, media was changed to DMEM supplemented with 1% P/S and 10% FBS. Cells were treated with 1 μg/mL cetuximab or IgG isotype control and incubated for 72 h (unless otherwise specified). Media were collected, spun to remove debris, and stored at −80 °C. Media were thawed and incubated overnight with protein A/G agarose to remove remaining drug (CMtx) or IgG isotype control (CM). After separation the media from the agarose pellets, 0.5 μg/mL EGF neutralizing antibody (CMtx-EGF) or IgG isotope control (CMtx) was added and media were incubated at 37 °C for one hour.When cells were at approximately 70% confluence, culture media were replaced with FBS and P/S free DMEM for three days. This medium was then collected, spun down to remove debris, aliquoted, and stored at −80 °C. Frozen conditioned media were thawed and subjected to cytokine arrays (R&D systems, ARY022B) or ELISAs (R&D systems, DEG00), both following the manufacturer’s instructions. Clustering, overlap analysis, and visualization: Clustering analysis and visualization were performed in the R statistical environment (v3.6.0) [33] using the cluster package (v2.0.9) [34] and BPG (BoutrosLab.plotting.general) package (v5.9.2) [35]. Modified Z-scores were generated by gene-wise scaling of the cytokine array data by median and standard deviation. These values were subsequently clustered in heatmaps. Visualization of overlaps between groups was facilitated via the VennDiagram package (v1.6.20) [36].Cells were serum-starved overnight and treated with EGF or cetuximab for the time specified. Cells that were treated with EGF and cetuximab were incubated with cetuximab for 1 h prior to the addition of EGF for the specified time. Cells were harvested on ice and needle treated using RIPA buffer supplemented with protease and phosphatase inhibitors. The protein lysates (30 μg) were then resolved on 4–12% Bis-Tris gradient pre-cast gels (Invitrogen), transferred onto polyvinylidene difluoride (PVDF) membranes via semi-dry transfer. Immunoblotting was then performed with corresponding antibodies. Quantification of protein bands with densitometry was performed using FIJI ImageJ. The analysis was added to the uncropped western blot images included in the Supplementary Materials (Figure S12A–P).Unpaired t-tests were performed using GraphPad Prism version 8.0.0 (GraphPad Software, San Diego, CA, USA) for Mac. p-value ≤ 0.001: ***; p-value ≤ 0.01: **; p-value ≤ 0.05: *.Targeted therapies have predominantly been designed to interfere with specific molecules that impact cancer cell proliferation and survival. However, often the targets expressed on the cancer cells are also found on neighboring stromal cells in the tumor microenvironment. Here, we provide the first evidence that the EGFR-targeted therapy cetuximab alters CAFs in a manner that protects CRC cells from the drug’s effects. These findings emphasize the importance of considering how targeted therapies may influence the microenvironmental milieu and ultimately alter tumor response.The following are available online at https://www.mdpi.com/2072-6694/12/6/1393/s1, Figure S1: Validation of CAF isolation from patient tumors, Figure S2: LIM1215 was also de-sensitized to cetuximab treatment in the presence of CAFs (A), or with mutated KRAS (B), Figure S3: Hematoxylin and eosin (H&E) images scored for stromal percentage, Figure S4: CAF percentage does not correlate with increased growth rate in untreated conditions, Figure S5: Dose-response curve showing conditioned media from cetuximab-treated CAFs (CMtx) is more protective than conditioned media from untreated CAFs (CM), Figure S6: Comparison of downregulated cytokines from array performed on untreated and cetuximab treated CAFs, Figure S7: Media types do not change levels or patterns of EGF secretion, Figure S8: EGF secretion by cetuximab treated CAFs was maintained over five days, Figure S9: Exogenous EGF protects LIM1215 cells from cetuximab treatment, Figure S10: ORG12620 viability is not significantly altered under various EGF concentrations tested, Figure S11: Validation of EGF neutralization by ELISA, Figure S12: Uncropped originals of the western blots depicted in the manuscript with densitometric analysis, Table S1: Clinical details of patient samples.Conceptualization, S.M.M.; Formal analysis, C.M.G., R.L., A.S., R.X.S., O.C., A.J., and B.L.; Funding acquisition, S.M.M.; Methodology, C.M.G.; Project administration, H.-J.L. and S.M.M.; Supervision, C.M.G. and S.M.M.; Validation, C.M.G. and R.X.S.; Visualization, C.M.G., E.J.F., and M.E.D.; Writing—original draft, C.M.G. and S.M.M.; Writing—review & editing, R.L., A.S., R.X.S., E.J.F., M.E.D., H.-J.L., and B.L. All authors have read and agreed to the published version of the manuscript.This research was funded by the following National Cancer Institute (NCI) Grants: Cancer Center Support Grant Development Funds P30CA014089 (S.M.) and a Physical Sciences-Oncology Trans-Network Award U54CA143907 (S.M.). It was also funded through philanthropic support as part of the Stephenson Family Personalized Medicine Center.We would like to express our deepest gratitude to our philanthropic supporters, particularly the Stephenson family, Emmet, Toni, and Tessa, for their donation of the Operetta HCS platform. We would also like to thank: R. Hill, D. Ruderman, and K. Kani for meaningful discussions; Y. DeClerck, M. Stallcup, and K. Patsch for manuscript review; O. Castellanos and team for patient data assistance; and S. Kim and E. Spiller for experimental assistance. The graphical abstract was created with BioRender.com by E.J.F.The authors declare no conflicts of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.Cancer-associated fibroblasts (CAFs) express epidermal growth factor receptor (EGFR), but are not sensitive to EGFR inhibition. (A) Immunofluorescence stained colorectal cancer tissue from a biopsy of patient 12620 displayed expression of EGFR (red) and αSMA (green). Full tissue slice is shown on top (scale bar: 5 mm), with a 20× zoomed-in section shown below (scale bar: 100 μm). (B) Quantitative polymerase chain reaction (qPCR) analysis of CAF-associated markers alpha-smooth muscle actin (αSMA), fibronectin (FN1), and vimentin (VIM) was performed on primary cultured CAFs (isolated from colorectal cancer (CRC) tissues of patients 12,905, 12,911, 13,000), normal primary fibroblasts (NCF12737) and CRC tumor cells (LIM1215). (C) Expression of EGFR in CAFs 12,905, 12,911, and 13,000 as detected by immunofluorescence. Scale bar: 100 μm. (D) Cells were treated with 1 μg/mL cetuximab (CTX) or IgG control for five days. Live and dead cell counts were obtained on days 0, 3, and 5 to calculate growth rates (cell doubling per hour), which were normalized to IgG condition for each cell type.CAFs protect cancer cells from cetuximab treatment. CAFs and DiFi cancer cells were co-cultured and treated with various concentrations of cetuximab. (A) Representative images of DiFi and DiFi + CAF13000 co-culture treated with cetuximab or IgG control were taken five days post-treatment. (B) Birth (left) and death (right) rates of DiFi cells were calculated on co-cultures with CAF starting percentages ~50% by fitting live and dead cell counts taken on days 0, 3, and 5 to an exponential growth model. (C) Starting ratios of CAF and DiFi cells were calculated before a 5-day treatment with 1 μg/mL cetuximab and DiFi cell growth rates were calculated. The dotted line represents the growth rate of DiFi monoculture treated with 1 μg/mL cetuximab. Linear fits show an increasing slope, indicating increased tumor cell growth with increased CAF percentages upon cetuximab treatment. R2 values of fit: CAF12905 = 0.414; CAF12911 = 0.716; CAF13000 = 0.543. (D) Conditioned media was collected from CAFs untreated (CM) and treated with 1 μg/mL cetuximab (CMtx) after three days. DiFi cells were then cultured with the conditioned media conditions with or without cetuximab treatment for five days. The absolute difference between treated and untreated DiFi cell growth rates was calculated for each condition. The dotted line represents the absolute difference of DiFi monoculture. p-value ≤ 0.01: **; p-value ≤ 0.05: *.Cetuximab treatment alters CAF secretion profiles. (A) Raw images of cytokine array blots performed on conditioned media collected from CAFs treated with IgG control or 1 μg/mL cetuximab for 72 h. Boxed readings indicate epidermal growth factor (EGF). (B) Cytokine and growth factor expression was evaluated via cytokine arrays. The overlap of upregulated cytokines (>0.5 fold compared to untreated) across CAF lines is shown. (C) Levels of EGF secretion were determined via enzyme-linked immunosorbent assay (ELISA) on primary CAFs (12905, 12911, 13000), cancer cells (DiFi, LIM1215), and normal primary fibroblasts (NCF12737) lines. (D) Conditioned media was collected from CAFs treated with 1 μg/mL cetuximab, 1 μM erlotinib, 5 μM oxaliplatin, or 1 μg/mL IgG control for 72 h. ELISAs were performed to evaluate levels of EGF. p-value ≤ 0.01: **; p-value ≤ 0.05: *; ns: not significant.Exogenous EGF confers cetuximab resistance in cancer cell lines and organoids. (A) DiFi cells were treated with cetuximab in media containing various spike-in levels of EGF. Images were acquired on days 0, 3, and 5. Live and dead cell counts were obtained and fitted to an exponential growth model to calculate the growth rate. (B) After being serum-starved overnight, DiFi cells were treated with 10 µg/mL cetuximab and/or increasing concentrations of EGF for 2 h. Protein expression was evaluated by western blot. (C,D) Patient-derived colon tumor organoid line ORG12620 was treated with increasing concentrations of cetuximab in low EGF (0.4 ng/mL) patient-derived organoid (PDO) defined media for five days. (C) Images were acquired and (D) CellTiter-Glo was performed on organoids to determine percent viability. (E) ORG12620 were serum-starved overnight and treated with cetuximab (10 µg/mL) and/or EGF (100 ng/mL) for 2 h. Protein expression was evaluated via western blot.EGF is the factor in CAF CMtx conferring cetuximab resistance in cancer cells. (A) DiFi cells were treated with various cetuximab concentrations while cultured in Dulbecco’s Modified Eagle Media (DMEM), 13000CMtx, or 13000CMtx-EGF (i.e., 13000CMtx treated with anti-EGF) media. Images were acquired on days 0, 3, and 5, and representative images from day five are shown. (B) Live and dead cell counts were obtained and fitted to an exponential growth model to calculate the growth rate. (C) DiFi cells were cultured with CMtx or CMtx-EGF collected from CAF12905, CAF12911, and CAF13000 with or without 1 µg/mL cetuximab. Growth rates were calculated and the absolute difference (treated-untreated) is shown. (D) Conditioned media was collected from CAF12905 treated with 1 µg/mL cetuximab (12905CMtx) in fetal bovine serum (FBS)-free media. Following overnight serum-starving, DiFi cells were cultured with 1 µg/mL cetuximab and/or 12905CMtx for 2 h. Protein expression was evaluated via western blot. p-value ≤ 0.01: **; p-value ≤ 0.05: *.

Med-MDPI/cancers/cancers-12-06-01394.txt

@@ -0,0 +1 @@
+The field of multiple myeloma (MM) imaging has evolved. The International Myeloma Working Group recently recommended performing 18F-fluorodeoxyglucose glucose (18FDG) positron emission tomography/computed tomography (PET/CT) with the aim of staging MM patients at baseline and evaluating response to therapy. Novel oncological radiotracers such as 11C-Choline and 18F-Fluorocholine, have been studied in comparison with 18FDG, mostly in MM patients presenting with refractory disease or suspected relapse. Choline-based tracers may overcome some limitations of 18FDG, which include a lack of sensitivity in depicting skull lesions and the fact that 10% of MM patients are FDG-negative. The majority of MM lesions display a higher uptake of Choline than FDG. Also, in many situations, Choline may offer better lesion visualization, with a higher tumor to background ratio; however, various patterns of Choline and FDG uptake have been observed in MM and some limitations, notably as regards liver lesions, should be recognized. Overall, Choline may provide additional detection of up to 75% more lesions. This article aims to provide a comprehensive review of the potential role of Choline in multiple myeloma, as compared to FDG, encompassing Choline physiopathology as well as data from clinical studies.Multiple myeloma (MM) imaging has rapidly evolved during the past few years. Functional imaging such as magnetic resonance imaging (MRI) and 18F-fluorodeoxyglucose glucose (18FDG) positron emission tomography/computed tomography (PET/CT) are now recommended for baseline staging by the International Myeloma Working Group (IMWG) [1]. Additionally, the IMWG recently recognized the prognostic value of 18FDG PET/CT in the evaluation of treatment response [2,3,4].18F-Fluorocholine (FCH) and 11C-Choline are PET/CT radiopharmaceuticals that were initially developed for prostate cancer imaging [5]. Indeed, Choline PET/CT can help to locate anatomical sites of disease recurrence when there is biochemical suspicion of relapse. Choline PET/CT is also recommended for the staging of well-differentiated hepatocellular carcinoma [6]. Choline-based tracers are emerging radiopharmaceuticals in the field of MM and could offer several advantages over 18FDG. This article aims to provide a comprehensive review of the potential role of Choline in MM, as compared to FDG, encompassing Choline physiopathology as well as data from clinical studies.FCH and 11C-Choline have similar biodistributions [5]. Choline is a precursor to the synthesis of phospholipids [7]. It is internalized by the cell through a choline transporter and is then transformed into phosphocholine by a Choline Kinase. It is then coupled to diacylglycerol to form phosphatidylcholine, a major component of cell membranes; therefore, Choline is thought to reflect the intensity of cell membrane synthesis. Choline can be labeled either with 11C or 18F. Even though 11C has the advantage of yielding a natural compound, the use of 11C-Choline is limited due to 11C short half-life of 20 minutes, which necessitates the presence of an on-site cyclotron. On the other hand, the longer half-life of 18F (110 min) allows FCH to be distributed to PET centers that are distant from the production site. After injection, 11C-Choline is rapidly cleared from the blood and optimal tumor-to-background contrast is reached within 5–7 min [8,9]. Choline is mainly metabolized by the liver, the main organ for lipid metabolism. Intense uptake in the liver (mean SUVmax of 11.4) and pancreas (mean SUVmax of 7.8) is observed [10]. Moderate uptake is seen in the spleen, salivary glands, and lachrymal glands. Mild uptake can be observed in the small and large intestines, breasts, and testicles as well. Bone marrow uptake is usually low (mean SUVmax of 1.7) [10], similar to 18FDG (SUVmean of 1.7 at the lumbar spine [11]). There is no relevant uptake in the adult brain, except in the choroid plexus and the pineal gland. 11C-methyl-choline can be oxidized to betaine by choline oxidase, with detectable metabolites soon after radiotracer injection [12]. Due to Choline urinary excretion, there is also an intense activity in the kidneys and the bladder. FCH has the advantage of having less urinary excretion than 11C-Choline.Malignant tumors can exhibit a high choline uptake due to their high rate of replication, implying a high turnover of the cell membrane. Besides cell replication, other metabolic pathways may be involved in choline uptake, as it seems not always related to the levels of Ki67 expression [13]. For example, Choline can be involved in the synthesis of acetylcholine and can also be metabolized to betaine [14]. Pathological Choline uptake is also not specific to cancer. Indeed, Choline uptake can be observed in cases of inflammatory conditions, as the monocyte-to-macrophage differentiation implies a high rate of cell membrane synthesis and lipogenesis. Hence, reactive lymph nodes including granulomatous disease can be a major cause of interpretation pitfalls [10].The uptake of Choline by MM cells might be related to different mechanisms. First, choline uptake can be related to the rate of replication of clonal plasma cells. Second, choline uptake may be linked to lipogenesis, as fatty acids and lipids play an important role in the pathogenesis of MM cells. Lysophospholipid levels (lysophosphatidylcholine and lysophosphatidic acid) are indeed increased in MM patients compared to healthy subjects [15]. Fatty acid synthetase expression is also upregulated in myeloma cells, contributing to proliferation and survival [16]. In relapsed and in high-risk MM, phosphatidylcholine is downregulated and is thought to be hydrolyzed to form lipid messengers, responsible for tumor dissemination [17,18]. Hence, MM cells may exhibit different patterns of uptake depending on the patient’s previous exposure to MM drugs.18FDG is a widely used PET radiotracer offering diagnostic and prognostic information in many hematological and solid cancers. It reflects the level of glycolytic activity of cancer cells, which is increased in about 90% of newly diagnosed MM [19]. 18FDG PET/CT offers a thorough evaluation of MM tumor burden at diagnosis and carries a powerful prognostic value at baseline and during therapy response assessment [20,21,22,23,24].Bellow, we discuss the findings relating to the detection of focal bone lesions and extra-medullary disease by 18FDG and Choline PET/CT.The comparison of Choline PET/CT and 18FDG PET/CT lesion detectability in MM has been performed in the past few years (Table 1). In a study conducted by Nanni and colleagues [25], 10 patients underwent both 11C-Choline and 18FDG PET/CT scans within one week. The settings were: evaluation after completion of initial therapy (n = 4), disease relapse (n = 4) and follow-up (n = 2). Six patients were positive on both examinations, with a total of 37 bone lesions detected on 11C-Choline PET/CT vs. 22 lesions on 18FDG PET/CT. Four patients had concordant negative findings. The last patient had only one lesion in the pelvis, which was 18FDG-positive but 11C-Choline negative. None of the patients had a positive 11C-Choline PET combined with a negative 18FDG PET. Choline intensity of uptake was globally superior to that of FDG, with a mean SUVmax of 5.0 vs. 3.8 respectively (p = 0.042); however, it is interesting to note that some lesions had a high 18FDG uptake but a low Choline uptake. Moreover, a mix of uptake patterns of MM lesions could be observed within the same patient. In a second study by Cassou-Mounat et al. [26], 21 MM patients underwent FCH-PET/CT and FDG PET/CT within a median time of 7 days, for progression under treatment (n = 6) or suspected relapse (n = 15). Nineteen were biochemically confirmed. One patient was detected as positive for bone involvement by FCH-PET/CT only, while no patient was positive by FDG PET/CT only. After masked reading, the mean number of foci per patient was 4.6 for FDG vs. 8.1 for FCH (p < 0.001) with a total number of 69 lesions for FDG and 121 for FCH. Among the 69 FDG-positive lesions, 65 were also FCH-positive. The 56 foci that were FCH+/FDG- were predominantly located in the skull and the torso. The median SUVmax values for the FCH-detected lesions (3.8 and 5.7) were higher than for the FDG-detected lesions (3.0 and 4.5). Additionally, the tumor/background ratio of Choline was superior to that of FDG. Overall, in relapsed/refractory MM, Choline PET/CT detects up to 75% more bone lesions than 18FDG PET/CT [25,26]. Choline also offers better visualization of MM lesions than FDG, which can contribute to a higher reading comfort but also higher inter-observer agreement. The low sensitivity of FDG for skull lesions is explained by the high glucose uptake in adjacent brain tissues (Figure 1). This limitation is overcome by Choline tracers (Figure 2). However, the high liver uptake of Choline may limit its sensitivity to depict lesions of surrounding bone areas such as the right costal grid.A difference in uptake pattern between 18FDG and Choline can be observed among patients but also within the same patient (Figure 3) [25,26]. FDGhigh/Cholinelow uptake of MM lesions are relatively rare. One needs first to exclude false-positive findings of FDG, with for example recent fractures that can bring substantial inflammation leading to increased FDG uptake [26]. MM patients are prone to bone fractures, especially of the ribs and spine, as these patients are at higher risk of having osteoporosis [28]. FDGhigh/Cholinelow uptake can be seen in aggressive or “dedifferentiated” lesions [26]. This uptake pattern has already been documented in hepatocellular carcinoma and is a signature of aggressive disease [29]. The more frequently observed pattern of Cholinehigh/FDGlow uptake of MM lesions may have different causes. It may correspond to early lesions, as only a minority of these foci were accompanied by osteolytic changes on the corresponding CT images (35%) [26] (Figure 4). It may also reflect a more indolent form of the disease [26]. It was recently pointed out that about 10% of newly diagnosed MM patients, with bone lesions on MRI, are FDG-negative, which is linked to a lower expression of hexokinase-2 [19]. Another study showed that newly diagnosed MM patients with false-negative findings on FDG PET/CT had longer PFS than FDG-positive patients [30]; however, other mechanisms also appear to be involved in the uptake as FDG-negative patients with relapsed/refractory disease were found to have normal expression of Hexokinase-2 [31]. Finally, the use of dexamethasone as a treatment option can decrease FDG-uptake by promoting MM cell apoptosis and by lowering the inflammatory environment of MM cells. Overall, these results corroborate the recent findings of multiple, spatially separated clones that coexist in the same patient, carrying different genotypic mutations [32]. These clones can be individualized even in patients with negative minimal residual disease. This highlights the complementary role of whole-body imaging, and the possible advantage of using more than one tracer in some situations that remain to be defined. Whether these different patterns are linked to patient prognosis should also be investigated in prospective studies.Although a good correlation can be observed between Choline uptake and plasma cell infiltration on BMB (R2 = 0.52) [27], the performance of Choline for the diagnosis of diffuse infiltration pattern has not yet been addressed. Studies prospectively comparing MRI and Choline PET/CT in that regard are warranted.The MYELOCHOL study is a prospective monocentric study that is aimed at comparing FCH and 18FDG PET/CT performance in newly diagnosed MM. Both of the previous studies performed have a common limitation [25,26]. Lesions were assessed without a pre-determined reference standard able to differentiate between true-positive and false-positive findings. MRI is the gold standard for bone marrow imaging [33], which makes it a good candidate for Choline PET/CT performance evaluation, as obtaining biopsy results for each lesion would not be feasible. Additionally, the performance Choline PET/CT is yet to be evaluated in a patient population that has not been exposed to MM drugs. The treatment of MM may indeed select clones that could behave differently from baseline [32].The MYELOCHOL study plans to include 30 newly diagnosed MM patients (Haut-Lévêque Hospital, University of Bordeaux, Bordeaux, France). All patients will undergo Whole-Body MRI, FCH, and 18FDG PET/CT with a maximal interval of 7 days between each imaging modality. Whole-Body MRI will be the standard of reference. Each imaging study will be read blindly by two nuclear medicine physicians with expertise in MM. A second step will include a comparative reading of each PET lesion with MRI findings, to determine if the lesion is a true positive or a false positive. In case of a Choline-positive lesion not found on MRI, a multidisciplinary team, composed of nuclear medicine, radiology, and hematology physicians, will come to a consensus as to whether the lesion can be related to MM. Study enrollment started in September 2019 and is currently ongoing.Multiple myeloma patients with extra-medullary disease (EMD) have a worse prognosis [34]. Extra-medullary lesions develop in any soft tissue and should be differentiated from soft tissue lesions expanding from adjacent bones (“breakout lesion”) [35]. In line with what was found for bone lesions, Choline uptake within EMD can either be higher or lower than FDG-uptake [25,26]. The detectability of liver EMD lesions may be hampered due to the high physiological choline uptake by the liver. Moreover, benign hepatic lesions such as nodular hyperplasia may mimic MM EMD [6]. Generally, caution must be taken when diagnosing EMD with Choline or FDG, as inflammatory or neoplastic processes are known confounders for both tracers [10,36].The value of Choline PET/CT for response assessment in MM has not yet been explored. The IMPETUS criteria have been defined for harmonizing 18FDG PET/CT readings with the use of the Deauville five-point scale [37]. This scale implies the use of liver uptake as a reference above which a lesion could be considered positive after treatment; however, Choline-specific criteria will have to be elaborated for response evaluation, mainly because of the markedly intense liver uptake of this tracer. The main features of Choline PET/CT and FDG PET/CT are summarized in Table 2Lapa and colleagues investigated the detectability rate of 11C-Choline compared to 11C-Methionine [27]. Nineteen pre-treated patients received both PET/CT scans within a median time of 10 days. No reference standard was used. Fifteen patients had positive findings, with only one patient with a methionine positive PET/CT but negative choline PET/CT. Choline and Methionine detected an equal number of lesions in 11/19 patients while Methionine detected more lesions in 8/19 patients. Seven patients had disseminated disease (>50 FL) that was identified by both tracers in four patients. In the remaining three patients with disseminated disease on Methionine PET/CT, Choline PET/CT only identified a total number of four lesions. The uptake of Methionine was superior to that of Choline with a median SUVmax of 9.3 (3.5–39) vs. 5.7 (3.5–14.6), respectively. A better tumor/background ratio was also observed with 11C-Methionine (18.7 vs. 8.8, p = 0.0001). Although 11C-Methionine appears to yield better results than 11C-Choline, its availability is currently limited, considering the difficulties related to the short half-life of carbon-11.Besides FDG, Choline, or Methionine, other radiotracers are of potential interest, with especially radiopharmaceuticals that can be used for disease imaging and treatment as well, constituting important theranostic applications [38], such as radiolabeled agents targeting the prostate-specific membrane antigen (PSMA), the chemokine receptor CXCR4 [39] or radiolabeled CD38-targeting antibodies [40,41]. A cluster of differentiation CD38 is highly expressed on myeloma cells. Daratumumab is the first fully human monoclonal antibody targeting CD38 for the treatment of MM. It improves the depth and duration of response in combination with other antimyeloma agents [42,43]. Daratumumab has been successfully labeled with 89Zr and 64Cu with promising immuno-PET imaging results in pre-clinical studies but also recently in small in-human studies [40,41]. CD38-targeted radioimmunotherapy has also been investigated preclinically [44,45]. The rate of patients with high CD38 expression will determine how many of them could beneficiate from anti-CD38 immuno-PET imaging and possibly CD38-targeted-immunoradiotherapy. In multiple myeloma, PET/CT performed with Choline-based tracers may offer several advantages over 18FDG PET/CT, showing better detectability of focal bone lesions; however, the diagnostic performance of Choline PET/CT should be further explored in larger prospective studies, which should be designed with a reference standard. These data are required before clinical routine implementation can take place. Studies focusing on the prognostic value of Choline PET/CT and its value as regards treatment response evaluation are also warranted.Original draft preparation C.M., G.M., and E.H.; review and editing, C.M., G.M., E.H., L.B., C.H., and A.L. All authors have read and agreed to the published version of the manuscript.This research received no external funding.This work has been supported in part by grants from the French National Agency for Research called “Investissements d’Avenir” IRON Labex n◦ ANR-11-LABX-0018-01, and from INCa-DGOS-Inserm_12558 (SIRIC ILIAD). We would like to sincerely thank Constantin Lapa and Malte Kircher for kindly providing additional data regarding the study they co-authored.The authors declare no conflicts of interest.A 65-year-old male with a light chain multiple myeloma (MM). 18F-fluorodeoxyglucose glucose (18FDG) positron emission tomography/computed tomography (PET/CT) (left panel) and 18F-Choline PET/CT (right panel) were performed with a 4-day interval. 18F-Choline PET/CT axial image of the skull shows an intense uptake of a skull base lesion. Because of the intense surrounding cerebral uptake on 18FDG PET/CT, the lesion is more difficult to individualize. Clinical examination at baseline identified the presence of diplopia and left ptosis, which disappeared a few days after induction chemotherapy was started.A 60-year-old female with IgG lambda smoldering multiple myeloma. An osteolytic lesion of the skull was found on a follow-up CT. 18F-Choline PET/CT was ordered to further characterize this lesion and search for additional bone lesions. 18F-Choline PET/CT axial image of the skull shows a moderate uptake of an occipital osteolytic lesion (A). Additional focal uptake of 18F-Choline was seen in the left femur, corresponding to a bone marrow lesion with no bone structural changes on CT (B).A 72-year-old male with light chain multiple myeloma. 18FDG PET/CT (left panel) and 18F-Choline PET/CT (right panel) were performed with a 5-day-interval. Different uptake patterns are seen in this patient. Lesion exhibiting high choline uptake but low FDG uptake can be seen in the left femur (A). In contrast, however, a lesion with high FDG uptake but low Choline uptake can be seen in the posterior arch of the left 6th rib (B) corresponding to an osteolytic lesion on CT (central panel).A 56-year-old female with IgG kappa MM. 18FDG PET/CT (left panel) and 18F-Choline PET/CT (right panel) were performed with a 3-day-interval. An axial PET/CT image of the pelvis shows a focal uptake of 18F-Choline within the right ischium without a corresponding uptake on 18FDG PET/CT. The corresponding axial CT image (central panel) shows no clear-cut structural changes (A). A focal lesion of the right femur exhibits a high uptake of 18F-Choline. This lesion exhibits only a faint uptake on 18FDG PET/CT (left panel) and is not visible on CT (central panel) (B).Characteristics and main results of studies exploring the use of 11C-Choline or 18F-Choline in multiple myeloma.MTH: Methionine; FL: Focal bone Lesion; EMD: Extra-Medullary-Disease; NS: Not Specified *91 FL were found in 15 patients. The remaining four patients had innumerable FL (>50) which were all matched by MTH PET/CT. ** 112 FL were found in 12 patients. The remaining 7 patients had innumerable FL (>50) with 3 of them having less than 3 lesions on Choline PET/CT.Comparative table of the main characteristics of 18F-Choline or 11C-Choline PET/CT vs. 18FDG PET/CT in multiple myeloma.MM: Multiple Myeloma; FL: Focal Lesion; EMD: Extra-Medullary Disease; IMPETUS: Italian Myeloma Criteria for PET Use.

Med-MDPI/cancers/cancers-12-06-01395.txt

@@ -0,0 +1 @@
+We evaluated tumor response at Computed Tomography (CT) according to three radiologic criteria: RECIST 1.1, CHOI and tumor volume in 34 patients with metastatic adrenocortical carcinoma (ACC) submitted to standard chemotherapy. These three criteria agreed in defining partial response, stable or progressive disease in 24 patients (70.5%). Partial response (PR) was observed in 29.4%, 29.4% and 41.2% of patients according to RECIST 1.1, CHOI and tumor volume, respectively. It was associated with a favorable prognosis, regardless of the criterion adopted. The concordance of all the 3 criteria in defining the disease response identified 8 patients (23.5%) which displayed a very good prognosis: median progression free survival (PFS) and overall survival (OS) 14.9 and 37.7 months, respectively. Seven patients (20.6%) with PR assessed by one or two criteria, however, still had a better prognosis than non-responding patients, both in terms of PFS: median 12.3 versus 9.9 months and OS: 21 versus 12.2, respectively. In conclusions, the CT assessment of disease response of ACC patients to chemotherapy with 3 different criteria is feasible and allows the identification of a patient subset with a more favorable outcome. PR with at least one criterion can be useful to early identify patients that deserve continuing the therapy.Adrenocortical carcinoma (ACC) is a rare neoplasm, with an estimated incidence in Western countries between 0.7 to 2 cases per million population per year [1]. With the increasing accuracy of diagnostic modalities and their more frequent use, nearly 15% of ACC cases are incidentally diagnosed and the risk of malignancy in incidental lesions has been related to tumor size (sensitivity and specificity values of 97% and 52% for lesions > 4 cm and 91% and 80% for tumors > 6 cm, respectively) [2]. In patients with adrenal tumors of 1–4 cm both the European Guidelines for management of adrenal incidentalomas [3] and the American College of Radiology Paper on the management of incidental adrenal masses [4] underline the role of unenhanced CT to detect microscopic fat content in solid adrenal lesions, with an attenuation threshold of 10HU, which is considered fundamental (with sensitivity and specificity values of 71% and 98%, respectively) in the differential diagnosis between benign and malignant lesions [5]. ACC, however, is frequently diagnosed in advanced stages: 18–26% of patients are in stages III and 21–46% in stage IV at presentation [6]. The respective five-year survival is 24–50% and 0–17% for stage III and IV patients, respectively [7,8]. Cortisol hypersecretion [9], complete surgical resection and proliferative activity, assessed by Ki67 expression [10] are additional independent prognostic parameters.The etoposide, doxorubicin and cisplatin combination regimen, administered in association with oral mitotane (EDP-M), is the mainstay of therapy for advanced/metastatic ACC [11,12]. According to the ENSAT guidelines [13], contrast-enhanced CT (CECT) of the chest and abdomen/pelvis is the reference imaging method, both for tumor staging and disease monitoring during therapy, while MRI of the abdomen and FDG-PET could offer additional information in selected cases. Response Evaluation Criteria in Solid Tumors (RECIST 1.1) [14] is the reference system for response evaluation of solid tumors to systemic antineoplastic treatments and is currently employed in the evaluation of the activity of chemotherapy in advanced ACC. RECIST 1.1 is based on detection of changes in tumor size, measured as the sum of the two longest axial diameters. The relationship between change in diameter and volume of tumors is based on the assumption that solid neoplastic lesions are spherical and, consequently, that proportional changes of tumor volume and product of perpendicular dimensions follow changes in tumor diameter, and vice versa. In practice, however, tumors commonly show odd shapes and not all parts equally respond to treatment. Another limitation of RECIST 1.1 is that it relies on the presumed correlation between tumor volume burden and planar dimensions. The assessment of tumor response according to volumetric criteria could be more advantageous than RECIST for assessing response to treatment in cancer patients, better showing size changes even in large, irregular lesions [15,16]. Moreover, not infrequently the tumor changes after treatment may be characterized by an increase in the necrotic component, which can be associated with stabilization or even an increase of the tumor size. The post-chemotherapy changes of tumor structure may present with the appearance of necrotic/hypo enhancing areas and these effects can be assessed as attenuation changes, measured in Hounsfield units (HUs) at contrast-enhanced CT scan. This is the basis for evaluation of response using the Choi criteria. These criteria were repeatedly found to be more helpful than RECIST to define advanced/metastatic cancer patients who benefit from target therapies such as imatinib [17] or sunitinib [18]. The introduction of multidetector-row CT (MDCT) scanners, in the clinical radiology practice, allows isotropic acquisition of extensive anatomic data, with postprocessing analysis, tumor segmentation, evaluation of tumor volume burden and its variations during therapies. These advances in imaging technology allow not only the measurement of bidimensional tumor size, but also different postprocessing analysis such as semiautomatic outlining of tumor boundaries (segmentation) and assessment of different tumor characteristics such as accurate definition of lesion volume, attenuation changes, favoring the detection of tumor response using multiple response criteria.In the present study, we prospectively evaluated disease response of advanced/metastatic ACC patients, uniformly treated with the standard EDP-M regimen, using RECIST 1.1, Choi and volume criteria. The study aimed to correlate response assessment with each criterion with progression free survival (PFS) and overall survival (OS) and to evaluate whether the combination of the three criteria could offer more helpful information on patients’ outcome.Demography and characteristics of the 34 patients included in the study are shown in Table 1. They were 24 females and 10 males, median age at diagnosis was 46.3 years (range 16–71). Sixteen patients (47.1%) did not have distant metastases and were initially addressed to surgery, while the remaining 18 patients (52.9%) had unresectable locally advanced or metastatic disease. Eighteen patients (52.9%) had secretory tumor. According to mENSAT ACC staging system [6,13,19], 4 patients (11.8%), were at stage III, 16 patients at stage IV A (47.1%), 6 at stage IV B (17.6%) and 8 at stage IV C (23.5%), before EDP-M administration, respectively. The distribution of the analyzed target lesions was as follows: 24 primary adrenal tumors (70.6%), 4 loco-regional recurrent solid lesions (11.8%), 2 liver metastases (5.9%), 2 mediastinal lymph-nodes (5.9%) and 2 peritoneal localization (5.9%).At the time of the analysis 20 patients (58.8%) were dead.In the present series no advanced ACC patients attained a complete response to the therapy with any adopted criteria. According to RECIST 1.1, a partial response (PR) was obtained in 10 patients (29.4%), 18 patients (52.9%) had stable disease (SD) and 6 patients (17.6%) showed disease progression (PD). As regard as Choi criteria, 14 patients (41.2%) were classified as PR, 13 patients (38.2%) had SD and 7 patients (20.6%) had PD. According to volumetric criteria, 10 patients (29.4%) obtained a PR, 17 patients (50%) SD and 7 patients (20.6%) PD. Overall, in 24 patients (70.5%) there was an agreement of all the three criteria in defining PR, SD or PD. Concerning the PR definition, this was recognized by all the three criteria in 8 patients (23.5%). Among discordant results, 2 patients (5.9%) considered responsive using RECIST 1.1 were not recognized as such using Choi criteria: 1 patient (2.9%) and volumetric criteria: 1 patient (2.9%). Four patients (11.7%) considered responsive applying the Choi criteria were classified as unresponsive by both RECIST 1.1 and volumetric criteria; 2 further patients (5.9%) classified as responsive according to the Choi criteria were not confirmed by RECIST 1.1 (1 patient-2.9%) and volume criteria (1 patient-2.9%).Finally, 10 patients (29.4%) considered responsive by the volumetric criteria were not confirmed by Choi in 1 patient (2.9%) and RECIST 1.1 in another patient (2.9%). An example of a large pelvic secondary implant, considered as a PR using the Choi criteria, otherwise classified as PD when applying RECIST 1.1 and volumetric criteria is shown in Figure 1.Overall, PR recognized by at least one criterion was observed in 15 patients (44.1%).We analyzed the correlation among each response criterion and patient outcome, in term of PFS (Table 2) and Os (Table 3). Responding patients, assessed with the RECIST 1.1 criteria, attained a higher OS (median 37.7 months) with respect to patients with SD (18.7 months) or PD (median 14.3 months). The corresponding median PFS was 14.9, 9.9 and 1.8 months, respectively.As regard as Choi, median OS was 25.4 months in responding patients, 18.7 months and 14.3 months in those with SD or PD, respectively. Median PFS in the 3 groups was 14.9, 9.9 and 10.8 months, respectively.With respect to Volume criteria, median OS was 37.7 months in responding patients and 18.7 and 14.3 months in those obtaining SD or PD. The corresponding median PFS was 14.9, 9.9 and 2.6 months, respectively.The Receiver operating characteristic curve (ROC) analysis revealed that all the three response criteria provided a similar moderate accuracy for predicting either PFS (area under curve [AUC] being 0.595, 0.668 and 0.654 for RECIST, CHOI and Volume criteria, respectively) or OS (AUCs: 0.717, 0.704 and 0.717, respectively).We further explored the additional information that Choi and volume response added to the RECIST response and observed that median OS was 37.7 months in patients with RECIST response, 21 months in patients obtaining Choi response, but no response according to RECIST and 12.2 months in non-responders. The corresponding median PFS values were 14.9, 12.3 and 9.9 months, respectively. Patients who obtained a response with volume criteria, but not with RECIST had a median OS of 19.2 months and PFS of 15.3 months as opposed with not responding patients in which median OS and PFS were 14.3 and 9.9 months, respectively.Moreover, we analyzed Overall Survival (OS) and Progression Free Survival (PFS) in patients attaining PR by at least one criterion versus others (SD and PD). The median OS was 25.4 months (19.4–31.4) in patients with PR and was 12.2 months (8.2–16.2) in the non-responding group. As regard as PFS, the median was 14.9 months (11.8–18) in PR patients and 9.9 months (6.8–12.9) in patients with SD or PD (Figure 2 and Figure 3).Finally, we analyzed OS and PFS stratifying our patients in three groups, as follows: (A) patients in whom the response assessed with all the three criteria were concordant; (B) responding patients according to one or two criteria, but not all the three criteria; (C) patients in whom there was no response with any of the adopted criteria.In the group A, the median OS was 37.7 months (95% CI: 19.8–55.6), versus 21 months (95% CI: 17.3–24.6) in the group B and 12.2 months (95% CI: 8.2–16.2) in the patients of the group C (SD or PD), p = 0.005 (Figure 4).Group A patients showed a better PFS: 14.9 months (95% CI: 12.5–17.2), than group B: median of 12.3 months (95% CI: 4.7–19.8) and group C: median 9.9 months (95% CI: 6.8–12.9), p = 0.107 (Figure 5).Imaging has a pivotal role in objectively defining tumor response or progression of cancer patients during systemic therapy. RECIST criteria, introduced in 2000 [20], are the most standardized, scientifically accepted and currently used system for tumor response evaluation. As mentioned in the introduction, these criteria, even after their revision [14], have several limitations linked to the heterogeneity of the forms and contours of tumor lesions and among different lesions in the same patient. These limitations can be addressed at least in part by the concomitant use of others response criteria. In patients affected by GIST, treated with the tyrosine kinase inhibitor imatinib, Choi criteria demonstrated that the degree of contrast enhancement at CT reflects vascular and interstitial volumes of the tumor, providing information about its structure and biological behavior, even in the absence of size variations [21,22]. In addition, the introduction in clinical radiology of dedicated software allows a fast and semiautomatic segmentation of the analyzed lesions with the automatic measurements of diameters as well as volume and attenuation changes. This approach aims to reduce the subjectivity of the analysis and to detect earlier subtle variations during and after systemic antineoplastic therapy, as well as to shorten the time spent by the radiologist for the manual tumor analysis, while increasing the reproducibility of the evaluation (Figure 1).In this study, we have prospectively evaluated disease response at CT imaging of advanced/metastatic ACC patients submitted to EDP-M regimen, adopting RECIST 1.1, Choi and volumetric criteria (Figure 6). Partial response assessed by each criterium significantly correlated with patient outcome, both in terms of PFS and OS. However, the concordance among these criteria in defining the disease partial response was observed in 8 patients (23.5%), while in 7 patients (20.5%) the response observed with one criterion was not confirmed by one or 2 other criteria. We therefore adopted a comprehensive approach, considering as responders all patients showing a disease response with at least one criterion. Using this definition, the proportion of responders increased from 23.5% to 44% with respect to standard RECIST 1.1, and responding patients maintained a better PFS and OS than patients with SD and PD. We subsequently evaluated the prognostic effect of the response recognized by all three criteria than discordant response between criteria and disease stabilization or progression. The results showed that patients in whom the response was agreed upon by the three criteria obtained the best prognosis in terms of PFS and OS. Patients, whose response was identified by one or two criteria, but not by all the three obtained a better survival, although not a significant advantage in terms of PFS than patients with SD and PD.These results have potential clinical implications. First, in the clinical course of patients undergoing EDP-M therapy we need to early identify which of them should stop a toxic treatment due to inefficacy with respect to those who deserve to continue the treatment. A disease response identified by at least one parameter could be useful in this respect, particularly in cases of discordant results. Noteworthy, the disease control in ACC patients submitted to EDP-M is subordinated not only to the cytotoxic effect of chemotherapy, but also to the achievement of mitotane levels in the therapeutic range, which usually occurs after 2–3 months from the beginning of therapy [23,24]. We have learned from the routine clinical practice that early progression of ACC patients to EDP-M according to RECIST 1.1 does not always mean treatment inefficacy [23,24]. The combination of the three methods of analysis could be pivotal, to identify true progressing patients from patients in which the treatment should be continued, despite initial disease progression.Perhaps more important, surgery of residual disease is often considered in many patients with ACC who achieve an objective response to cytotoxic treatment. The long-term efficacy of the surgical approach of metastatic malignant disease is known to be higher in patients with an indolent disease or in those in which the disease is made inactive by treatment. The finding that patients in whom all the three criteria confirmed the response to therapy had the best prognosis is relevant since it could identify a patient subgroup with less aggressive disease that could potentially obtain benefit from adjunctive surgery if deemed feasible.This study has several limitations, first of all it is clearly underpowered, only 34 patients were evaluated. Considering the 29% response rate obtained with the RECIST criteria alone and the 44% response obtained considering at least one of the 3 criteria, the study has a power of 0.36 with a one-tailed test (0.25 with a two tailed test) to demonstrate the observed absolute difference of 15% with an alpha error of 5%. Moreover, the few patients enrolled, the absence of a validation set and the different target lesions analyzed, with 24 primary tumors (70.6%), 4 loco-regional recurrent solid lesions (11.8%), 2 liver metastases (5.9%), 2 mediastinal lymph-nodes (5.9%) and 2 peritoneal localization (5.9%) are additional drawbacks of this study.These limitations notwithstanding, our results underline the importance to combine these three response criteria to better define the prognostic role of disease response to therapy. The progressive implementation and development of advanced software for automatic analysis is crucial to achieve an accurate, objective analysis, together with a significant reduction of the radiologist’s time spent in the segmentation process.In this prospective, observational, monocentric study we analyzed the clinical data and CT examinations of 58 locally advanced or metastatic ACC patients, treated in the Department of Oncology at ASST Spedali Civili of Brescia from November 2013 to September 2019. Among them, 24 were excluded due to incomplete/inadequate CT data (because CT examinations performed in different sites and with different imaging protocols were not always recognized from the IntelliSpace Portal Software used for the analysis), different anatomic districts (i.e., patients with exclusive lung metastases, where the semi-automatic segmentation process could be less accurate) and sizes of the lesions (very small lesions, not always recognized by the software), and/or inclusion in EDP-M chemotherapy regimen, therefore our final population consisted in 34 patients (Figure 7).EDP-M was administered according to the following scheme: etoposide 100 mg/m2 (day 2–3–4), doxorubicin 20 mg/m2 (day 1), cisplatin 40 mg/m2 (day 3–4). Oral mitotane was administered concomitantly with chemotherapy at the starting dose of 1500 mg daily, with further progressive dose increments up to the maximum tolerated dose. Serum mitotane was monitored every 4 weeks and when the patients attained the therapeutic range the mitotane dose was tapered to maintain serum concentration within 14 and 20 mg/L. All patients were studied at the baseline (before starting of chemotherapy) and every 3 months, according to the ENSAT guidelines.Multi Detector Computed Tomography (MDCT) examinations were carried out using two different scanners, Toshiba Aquilion (with 80 detectors array) and Philips Brilliance (with 64 detector arrays), before and after i.v. automatic injection of iodinated contrast agent (370 mgI/mL@3 mL/s-1.3 mL/Kg body-weight), using a bolus tracking technique. CT images of the chest and abdomen were acquired during the late arterial and venous phases after contrast injection. Image analysis was performed using a dedicated Software (Multi-Modality Tumor Tracking, IntelliSpace Philips Portal, version 10, Philips Healthcare, Best, The Netherlands) and focused to the analysis of solid lesions, as follows: primary adrenal tumors, local solid recurrence, recurrent peritoneal lesions, hepatic metastases and solid pulmonary lesions.Two experienced Radiologists (R.A., M.D.T.) performed 3D semiautomatic segmentation of target lesions on the baseline examination and on the first CT after chemotherapy, using a dedicated software (Multi-Modality Tumor Tracking, IntelliSpace Philips Portal, version 10, Philips Healthcare, Best, The Netherlands). After segmentation, the software automatically compares the examinations in terms of the long and short axis, lesion volume and CT attenuation. Furthermore, aside from the calculation of the percentages of variations, the software provides a chart where relative changes of the analyzed tumor characteristics between multiple examinations are immediately evident. In this study, we evaluated tumor response considering the sum of the longest diameter of all lesions (RECIST 1.1 criteria), the presence of attenuation changes at CT (CHOI) and volume changes after therapies (Figure 8).The assessment of response according to the RECIST 1.1, Choi and volumetric criteria are summarized in Table 4.Study population characteristics were described using classical descriptive statistics: percentage, means and standard derivation, median and extreme values. PFS was defined as the time elapsing from the beginning of the EDP-M treatment until disease progression or death. Non-progressing patients still alive were censored at the last follow-up examination. OS was defined as the time interval between the date of EDP-M treatment start and the date of death from any cause or the last known alive date. The present study aimed to obtain exploratory information on the prognostic role of disease response evaluated with 3 different criteria—and the combination of them—in patients with advanced ACC submitted to systemic chemotherapy. Due to the exploratory nature of this study, a sample size was not calculated. All survival curves were calculated by the Kaplan–Meier method and differences compared by the log-rank test. Cox’s proportional hazards regression model was employed to assess the Hazard ratios (HRs) and 95% confidence intervals (95% CIs). Missing data were dealt with by excluding patients from particular analyses if their files did not contain data for the required variables. A Receiver operating characteristic curve (ROC) analysis was conducted to assess the global accuracy of disease response by the 3 criteria for predicting PFS and OS categorized at the median value. SPSS statistical software (version 23.00, Chicago, IL, USA) was used for statistical analysis. A p value of 0.05 was considered statistically significant.The concomitant use of RECIST 1.1, volumetric and CHOI criteria in the assessment of disease response at contrast-enhanced MDCT, in advanced/metastatic ACC patients submitted to EDP-M chemotherapy, is feasible. Dedicated software and automatic procedures make it more objective and not excessively time-consuming for the radiologist. The concordance of all the three response criteria allows more accurate detection of a patient subset destined to obtain a good outcome, in which surgery of the residual disease could be justifiable, if technically feasible. The partial non-overlap between the response criteria allows instead to enlarge the number of patients deemed responsive to therapy and this can be useful in discriminating early after a few cycles, the patient for whom it is appropriate to continue therapy compared to those in whom it is more appropriate to stop treatment.Conceptualization, R.A., A.B., L.G. and S.S.; methodology, L.G., S.S. and G.A.M.T.; software, R.A., M.D.T. and M.B.; validation, L.B., F.V. and R.A.; formal analysis, S.G., S.S. and M.L.; investigation, R.A., M.C.B., L.B., F.V. and M.D.T.; resources, G.A.M.T. and A.B.; data management, M.L., D.C. and M.B.; writing—original draft preparation, R.A., L.G. and M.L.; writing—review and editing, R.A., M.L., M.C.B.; visualization, M.C.B., M.D.T. and M.B.; supervision, A.B., S.G. and L.G.; project administration, L.G., G.A.M.T. and S.S.; funding acquisition, A.B. and L.G. All authors have read and agreed to the published version of the manuscript.This research received no external funding.Acknowledgments to Marta Peroni (ICAP, Clinical Lifecycle Specialist, Philips Healthcare, Milano, Italia), for her technical support with the use of Intellispace system and applications. We are grateful to our patients and their families.The authors declare no conflict of interest.Disagreement in response assessment within the three criteria: according to Choi, the decrease in tumor attenuation was evaluated as a partial response, while according to both Volume and RECIST 1.1 its increase in planar dimensions and volume resulted in Progressive Disease.Overall Survival (OS) in patients with PR with at least one criterion [median 25.4 months (19.3–31.4)] versus no response disease [median 12.2 months (8.2–16.2)], p = 0.019.Progression Free Survival (PFS) in patients with PR with at least one criterion [median 14.9 months (11.8–18.0)] versus no response [median 9.9 months (6.8–12.9)], p = 0.093.Overall Survival (OS) in patients with PR according to all three criteria [median 37.7 months (19.8–55.6)] versus PR in one or two criterion [median 21 months (17.3–24.6)] versus SD or PD (in all 3 criteria) [median 12.2 months (8.2–16.2)], p = 0.005.Progression Free Survival (PFS) in patients with PR according to all the three criteria [median 14.9 months (12.5–17.2)] versus PR in one or two criteria [median 12.3 months (4.7–19.8)] versus SD or PD (in all 3 criteria)[median 9.9 months (6.8–12.9)], p = 0.107.Concordance of the three response criteria in the assessment of a primary metastatic ACC, with planar dimensions changes (RECIST 1.1), tumor volume and tumor size/attenuation (according to Choi).CONSORT Diagram describing the population of the study.Analysis of variations of tumor characteristics (longest diameters for RECIST 1.1, attenuation changes for Choi and Volume changes) obtained after semi-automatic tumor segmentation and displayed in a chart using the “Multimodality Tumor Tracking” Software (IntelliSpace Portal-Philips).Patients demographics and clinical characteristics.Progression free survival (PFS) univariate analysis according to RECIST 1.1, Choi and volume staging criteria.1 PR, partial response; 2 SD stable or 3 PD progressive disease.Overall, Survival (OS) univariate analysis according to RECIST, Choi and volume staging criteria.1 PR, partial response; 2 SD stable or 3 PD progressive disease.Summary of response assessed by the RECIST 1.1, Choi and Tumor Volume criteria.

Med-MDPI/cancers/cancers-12-06-01396.txt

@@ -0,0 +1 @@
+The B cell receptor (BCR) pathway has been identified as a potential therapeutic target in a number of common B cell malignancies, including chronic lymphocytic leukemia, diffuse large B cell lymphoma, Burkitt lymphoma, follicular lymphoma, mantle cell lymphoma, marginal zone B cell lymphoma, and Waldenstrom’s macroglobulinemia. This finding has resulted in the development of numerous drugs that target this pathway, including various inhibitors of the kinases BTK, PI3K, and SYK. Several of these drugs have been approved in recent years for clinical use, resulting in a profound change in the way these diseases are currently being treated. However, the response rates and durability of responses vary largely across the different disease entities, suggesting a different proportion of patients with an activated BCR pathway and different mechanisms of BCR pathway activation. Indeed, several antigen-dependent and antigen-independent mechanisms have recently been described and shown to result in the activation of distinct downstream signaling pathways. The purpose of this review is to provide an overview of the mechanisms responsible for the activation of the BCR pathway in different B cell malignancies and to correlate these mechanisms with clinical responses to treatment with BCR inhibitors.The B cell receptor (BCR) is a transmembrane signaling complex that is expressed by most normal and malignant B lymphocytes and plays a key role in regulating the growth, differentiation, and function of these cells. It is composed of a membrane immunoglobulin molecule, which functions as the antigen recognition unit, and a heterodimer of the proteins CD79A and CD79B, which functions as the signaling unit. Binding of the membrane immunoglobulin to antigen generates a signal that is transduced by a complex network of various kinases, phosphatases, adaptor proteins, and transcription factors. This signal can induce a variety of cellular responses, including proliferation, differentiation, adhesion, survival, anergy, or apoptosis. The outcome is determined by the relative activity of the various downstream signaling molecules that transduce the BCR signal. In addition to antigen binding, antigen-independent mechanisms have been shown to activate the BCR pathway in certain B cell malignancies and to contribute to the expansion and survival of the malignant B cells.In this review, we outline the mechanisms responsible for the chronic activation of the BCR pathway in various B cell malignancies and describe how these different mechanisms affect the signaling pathways and cellular responses that are regulated by the BCR signal. In addition, we summarize clinical experiences with different BCR inhibitors and correlate the clinical responses with mechanisms of BCR pathway activation. The initial events resulting in activation of the BCR have still not been fully elucidated, but most available evidence suggests that antigen binding induces a reorganization of the actin cytoskeleton leading to local convergence of monomeric or oligomeric BCR units and their assembly into signaling microclusters [1,2,3]. These clusters then recruit members of the SRC-family of kinases (SFKs), such as LYN, FYN, or BLK, that phosphorylate the tyrosine residues within the immunoreceptor tyrosine-based activation motifs (ITAMs) of CD79A and CD79B (Figure 1). The phosphorylated ITAMs then serve as binding sites for the kinase SYK, which becomes activated through a multistep process that involves phosphorylation by SRC family kinases and trans-autophosphorylation [4]. Once activated, SYK propagates the BCR signal by phosphorylating the adaptor proteins BLNK, BCAP, and SHC, which then serve as a scaffold for the recruitment of other signaling molecules that together form a large multimolecular complex defined as the BCR signalosome [5]. The BCR signal is further propagated by the lipid kinase PI3Kδ, which is recruited to the signalosome by binding to BCAP or CD19 and becomes activated through a conformational change in the regulatory p85 subunit that exposes the catalytic p110δ subunit [6]. Activated PI3Kδ then phosphorylates the membrane phospholipid phosphatidylinositol-4,5-bisphosphate (PIP2) and converts it into phosphatidylinositol-3,4,5-triphosphate (PIP3), which then recruits several important downstream signaling molecules and therapeutic targets. One of these is the kinase BTK, which is subsequently activated by phosphorylation by SRC family kinases and autophosphorylation. BTK then phosphorylates and activates PLCγ2, which hydrolyses PIP2 to generate inositol trisphosphate (IP3) and diacylglycerol (DAG). Binding of IP3 to its receptor, a calcium channel located in the endoplasmic reticulum, results in release of calcium into the cytosol and activation of the phosphatase calcineurin, which then dephosphorylates and activates the transcription factor NFAT [7]. In addition, calcium together with DAG activate protein kinase C (PKC), which phosphorylates the adaptor protein CARD11 and induces the formation of a CARD11–BCL10–MALT1 (CBM) signaling complex that activates the transcription factor NF-κB [8]. Another key signaling molecule that is recruited to the cellular membrane by PIP3 is the serine/threonine kinase AKT, which becomes activated following phosphorylation by PDK1 and the mTORC2 complex [9]. AKT has numerous substrates that play important roles in regulating cell growth and survival. Among these are the FoxO transcription factors and the kinase GSK3, which are both inactivated by AKT-mediated phosphorylation. Inactivation of FoxO transcription factors results in reduced expression of certain proapoptotic and cell-cycle inhibitory proteins, whereas inactivation of GSK3 inhibits the turnover and prolongs the half-life of several proteins that positively regulate cell survival and proliferation. In addition, AKT phosphorylates and activates the mTORC1 complex, which then increases the rate of protein translation by activating the ribosomal protein S6 kinase and the eukaryotic initiation factor 4E (eIF4E).Other signaling molecules that are activated downstream of the BCR are the mitogen-activated protein kinases ERK, JNK and p38MAPK [10]. These kinases regulate a number of transcription factors, including Elk1, c-Myc, c-Jun, ATF2, and Max, which are important for B cell proliferation and survival.The BCR signal is negatively regulated by various inhibitory receptors, such as CD22, CD72, FCγRIIB, and SIGLEC10, and phosphatases, such as SHP1, PTPN22, SHIP1, and PTEN [11,12]. SHP1 and PTPN22 terminate the BCR signal by dephosphorylating certain BCR-proximal signaling components, including CD79A, CD79B, SFKs, SYK, and BLNK, whereas PTEN and SHIP1 function by dephosphorylating PIP3. Expression of these negative regulators can vary across different B cell malignancies and can affect the capacity of the malignant cells to transduce the BCR signal. For example, SIGLEC10 is often downregulated in human B-cell lymphoma cell lines compared to normal B cells and knockdown of this receptor in a murine model has been shown to result in enhanced BCR signaling [13,14]. Similarly, downregulation of PTEN has been reported in 55% of germinal center B-cell-like (GCB) diffuse large B cell lymphomas (DLBCLs) and results in constitutive activation of the PI3K/AKT pathway [15]. Downregulation of SHP1 has also been observed across several B cell malignancies and can result in antigen-independent activation of certain downstream BCR signaling pathways (described in greater detail later). Together, these data suggest that perturbations of the BCR signaling machinery are frequent in B cell malignancies and likely to contribute to the pathogenesis of these disorders.The complexity of the BCR signaling network is further increased by the existence of parallel pathways of activation and the substantial crosstalk between the different downstream signaling molecules. These features are sometimes cell type specific and may account in part for the different activity of the various BCR inhibitors in different B cell malignancies. For example, in addition to PI3Kδ, the related kinase PI3Kα has been implicated in signaling through the BCR in Activated B cell-like (ABC) DLBCL [16,17], providing a potential explanation for the greater clinical activity of the dual PI3Kα/δ inhibitor copanlisib compared to the PI3Kδ inhibitor idelalisib in patients with DLBCL [18,19]. Similarly, a recent study from our group showed that two pathways downstream of the BCR are involved in inactivation of the kinase GSK3 in chronic lymphocytic leukemia (CLL) cells, resulting in a different capacity of SYK, BTK, and PI3Kδ inhibitors to reduce the expression of the antiapoptotic protein Mcl-1, which is negatively regulated by GSK3 [20]. Finally, although BTK is located downstream of PI3K, several studies have suggested that there is also crosstalk in the opposite direction and that BTK can enhance the activity of the PI3K/AKT pathway in BCR-stimulated cells [21,22,23,24,25,26]. The mechanism behind this effect is still not completely understood, but involvement of BTK in the phosphorylation of the adaptor BCAP and in the production of the PI3K substrate PIP2 have been proposed as potential explanations [22,24]. This crosstalk could also provide an explanation for the partial inhibition of AKT following treatment of CLL cells and lymphoma cell lines with the BTK inhibitor ibrutinib in vitro and in vivo [27,28,29,30]. Another possible explanation for the variable activity of BCR inhibitors in different B cell malignancies is the mechanism of BCR activation. Studies that were conducted over the last 15–20 years have provided compelling evidence that the BCR pathway is activated in CLL, DLBCL, Burkitt lymphoma (BL), follicular lymphoma (FL), mantle cell lymphoma (MCL), marginal zone lymphoma (MZL), and Waldenstrom’s macroglobulinemia (WM) (Figure 2). However, these studies also revealed important differences in the proportion of cases with an activated BCR pathway and the mechanism of BCR pathway activation. The following sections will describe our current understanding of these mechanisms in the different disease entities and how these relate to responses to treatment.CLL is characterized by the expansion of a subset of mature B lymphocytes that express the surface markers CD5, CD19, and CD22 and low levels of surface IgM and CD79B. Substantial evidence suggests a central role of BCR signaling in disease development and progression. Seminal studies on immunoglobulin gene structure uncovered two main subsets of CLL: unmutated CLL (U-CLL) harboring 98% or more homology to the germline immunoglobulin heavy chain variable (IGHV) gene sequences and mutated CLL (M-CLL), where less than 98% IGHV gene homology to the germline is observed [31]. The fact that U-CLL patients have significantly worse prognosis than M-CLL patients hinted that BCR structure is a key predictor of disease progression [32,33]. Apart from these characteristics related to the degree of somatic hypermutation (SHM), CLL BCRs also show a peculiar pattern of IGHV and immunoglobulin light chain variable (IGLV) gene usage which is dominated by the presence of particular IGHV and IGLV gene combinations. BCRs encoded by these IGHV/IGLV combinations and having particular HCDR3 structures have been named stereotyped BCRs and have been identified in approximately one-third of CLL cases, whereas they are rarely seen in normal B lymphocytes [34,35,36,37]. Considering that the likelihood of this happening by chance is virtually negligible, the occurrence of stereotyped BCRs has been taken as evidence that CLL cells are selected because of particular antigen-binding properties. Based on IGHV/IGLV association and HCDR3 properties, the stereotyped receptors have been divided into “stereotyped subsets”.Soluble immunoglobulins derived from CLL cells have been reported to display shared reactivity towards self-antigens, commonly more than one [38,39,40,41,42,43]. Such polyreactivity is primarily observed with U-CLL immunoglobulins and includes binding to IgG, ssDNA, dsDNA, vimentin, filamin A, cofilin-1, phosphorylcholine on oxidized LDL, cardiolipin, nonmuscle myosin heavy chain IIA, and stromal cell antigens. Importantly, many of the abovementioned self-antigens are found on apoptotic blebs [43], hinting that microenvironmental apoptosis is an important disease driver. Additionally, reactivity with foreign antigens has been reported in some cases, including Streptococcus pneumoniae polysaccharides [40], β-(1,6)-glucan on filamentous fungi [44], cytomegalovirus phosphoprotein pUL32 [45], HIV-1 envelope gp41, influenza hemagglutinin, and hepatitis C virus E2 protein [46]. Reactivity with any of these antigens could account for the chronic activation of the BCR pathway that is frequently observed by gene expression or phospho-protein profiling analysis of CLL cells. Such evidence is particularly seen in CLL cells isolated from lymph nodes, which typically display high levels of BCR and NF-κB target genes [47] and express constitutively activated BCR signaling molecules, including LYN [48], SYK [49], PI3K [50], BTK [29], PKCβ [51], ERK [52], NF-κB [53], and NFAT [52]. Importantly, enhanced activation of these molecules correlates with inhibition of spontaneous apoptosis, suggesting a pro-survival role for BCR signals [29,48,49,50]. Indeed, the BCR-induced constitutive SYK activation has been shown to upregulate the antiapoptotic protein Mcl-1 [49] by activating the PI3K/AKT pathway [54,55]. Notably, prolonged AKT activity results in increased mTORC1 and reduced GSK3 activity, with a resulting increase in Mcl-1 protein translation and inhibition of MCL1 degradation, respectively [54,56,57]. Further pointing to an important role for the BCR pathway in the pathogenesis of CLL is the fact that a number of signaling molecules that are involved in BCR signal transduction are aberrantly expressed by the leukemic cells. The ZAP-70 protein kinase, which is a SYK homologue that plays a key role in transducing signals through the T cell receptor, is aberrantly expressed mostly in U-CLL patients [58]. Importantly, ZAP-70 associates with CD79B, enhancing BCR signaling and acting as a negative prognostic factor [59]. Interestingly, although ZAP-70 is inefficiently phosphorylated following BCR stimulation, its role in recruiting downstream BCR molecules is preserved [60], hinting that it could interfere with BCR negative regulation rather than being a direct activator. Defective negative regulation is a frequent phenomenon in oncogenic signaling; accordingly, absent or substantially reduced expression of the AKT and ERK negative regulator PHLPP1 is observed in CLL cells, causing an enhanced BCR-mediated AKT, ERK, and GSK3 phosphorylation [61]. An additional mechanism accounting for aberrant AKT activation in CLL consists in the overexpression of the phosphatase PTPN22 [62]. PTPN22 quells LYN activity, thus blunting LYN-mediated activation of a negative regulatory loop involving the inhibitory receptor CD22 and the phosphatase SHIP, which by dephosphorylating PIP3 blocks AKT membrane recruitment and activation. Given that LYN is a major activator of SYK, PTPN22 overexpression also downregulates proximal BCR signaling, including PLCγ2 and MAPK cascade activation. The latter effects may seem counterintuitive given the pro-oncogenic role of the BCR. However, hyperactivation of BCR signalling above a maximum threshold can induce apoptosis in B cells, including CLL cells [63,64]. Thus, PTPN22 overexpression may serve to selectively uncouple AKT from downstream proapoptotic BCR pathways and thus protect CLL cells from tolerance mechanisms that eliminate autoreactive B cells. Another AKT regulator, TCL1, is also often overexpressed in CLL cells, especially in the U-CLL subset [65]. TCL1 is a lymphoid oncogene which associates with AKT and ZAP-70 in the proximity of the membrane. More precisely, BCR activation induces and stabilizes AKT-TCL1 complexes on the membrane, potentiating AKT-mediated signals [66]. Importantly, TCL1 is a potent negative prognostic marker in CLL. Consistently, Eµ-TCL1-transgenic mice display an emergence of clonal CD5+/IgM+ B cell expansions resembling IGVH-unmutated human CLL, thus defining TCL1 as a strong CLL oncogene [67,68]. Collectively, these studies indicate that recurrent alterations in the levels of positive and negative BCR signaling regulators intrinsically affect the nature of BCR signaling and may contribute to the pathogenesis of CLL. A major step forward in understanding how BCR signals are generated in CLL cells came from the study of Dühren-von Minden and colleagues, who identified cell-autonomous signaling consequent to BCR recognition of internal immunoglobulin epitopes as a novel mechanism of BCR pathway activation in CLL [69]. In this paradigm-shifting study, Dühren-von Minden and colleagues expressed CLL-derived immunoglobulins and an inducible BLNK adaptor in a BCR/BLNK-negative murine pre-B cell line and reported Ca2+ flux in the absence of an external BCR ligand. Interestingly, this signal was observed independently of BCR type (stereotyped vs non-stereotyped) and regardless of IGHV mutation status. Importantly, such Ca2+ flux was not observed after expression of immunoglobulins derived from other NHLs, suggesting a CLL-specific mechanism. Additional experiments revealed that the HCDR3 regions of CLL immunoglobulins recognized epitopes in the variable or constant regions of adjacent surface immunoglobulin molecules, providing an explanation for the inherent capacity of CLL BCRs to undergo autonomous interactions [69,70,71]. Consistent with cell-autonomous reactivity of CLL BCRs, a very recent study employing super-resolution microscopy provided further evidence of such interactions by visualizing oligomeric BCR nanoclusters on the surface of unstimulated CLL B cells [3]. Following the studies by Dühren-von Minden and colleagues, our group used the Eµ-TCL1-transgenic mouse model to evaluate the capacity of different types of transgenic BCRs to induce leukemia in vivo [72]. This study showed that B cells that express transgenic BCRs that become activated by low-affinity extrinsic autoantigens and/or cell-autonomous interactions enter into the leukemogenic process and become CLL cells, whereas B cells that express high-affinity transgenic BCRs do not undergo malignant transformation regardless of antigen form (soluble or membrane-tethered) or presentation (foreign or self). Subsequent studies using the Eµ-TCL1-transgenic mouse model and transgenic BCRs with other specificities reaffirmed these findings. In particular, the study of Hayakawa et al. showed that B cells expressing a transgenic low-affinity anti-Thy-1 BCR undergo malignant transformation only in the presence of the Thy-1 autoantigen, confirming that chronic stimulation with extrinsic low-affinity autoantigen can induce leukemia in vivo [73]. To further explore the role of high-affinity foreign antigens in driving leukemia development and progression, Jiménez de Oya and colleagues generated Eµ-TCL1 tg mice carrying gene-targeted immunoglobulin heavy chains from antibodies reactive with lymphocytic choriomeningitis virus or vesicular stomatitis virus [74]. Although these mice developed leukemias that express BCRs reactive with their cognate viral antigens, the presence or absence of the viral antigens had no influence on the rate of leukemia development or progression. Rather, the transgenic heavy chains preferentially paired with light chains that conferred to the leukemic BCRs the ability to cross-react with various external autoantigens or undergo cell-autonomous interactions. Taken together, the findings from these three studies provide in vivo evidence that self reactivity is a major driving force in CLL pathogenesis and suggest that only BCR signals of certain quality can promote the growth of the malignant cells.The murine studies also suggested that the capacity of the leukemic cells to respond to external antigen can influence the aggressiveness of the disease and may account for the variability in the clinical course of CLL. In particular, we observed that leukemia developed more rapidly when the malignant cells expressed BCRs that generated a weak autonomous signal but responded strongly to stimulation with external antigen, compared to cells expressing BCRs that generated a strong autonomous BCR signal but did not respond or only weakly responded to external antigen stimulation [72]. Similar findings were presented by Minici and colleagues who investigated the strength of the autonomous interactions of human CLL BCRs belonging to two distinct stereotyped subsets [71]. Importantly, in the more indolent BCR subset #4, self-recognition was found to be tight and long-lived, while in the more aggressive BCR subset #2, self-association was found to be of low-affinity and shorter duration [71]. The explanation for this correlation is that strong self-recognition pushes the cells toward anergy, while a weaker self-recognition implies higher BCR availability to bind to external self-antigens and to respond to such stimulation.The above findings also imply that the signals generated by cell-autonomous and cell-extrinsic interactions are not equivalent and may induce distinct cellular responses [75]. Cell-autonomous signals might provide a repetitive stimulation of low or intermediate strength, which may act as a continuous survival source for the leukemic cells. On the other hand, BCR stimulation with external ligand has been shown to increase the expression of the cell cycle regulators MYC, CCND2, and CDK4 and to increase the percentage of CLL cells in the G1 phase of the cell cycle, suggesting that interactions with external autoantigens may provide the initial stimulus required for leukemic cell proliferation [76,77,78]. Defining the downstream signaling pathways that are activated by cell-autonomous and cell-extrinsic interactions will be required to understand the exact nature of the cellular processes regulated by these two mechanisms of BCR pathway activation. Addressing this question is important in view of the possibility that different prognostic subgroups of CLL may be driven by different types of BCR signals and that these BCR signals may be differently targeted by available BCR signaling inhibitors. DLBCL is a heterogeneous neoplastic entity, characterized by a high mutational load: the implementation of next generation sequencing techniques revealed that the mean number of coding genome alterations exceeds 70 per single DLBCL case and identified over 150 genetic drivers [79,80]. Such genetic complexity is responsible for a very heterogeneous phenotype as well as for a significantly variable response to therapy. While the details of DLBCL pathogenesis are extensively reviewed elsewhere [81], we will mainly concentrate on the mechanisms of activation of the BCR pathway. Genomic, biochemical, and functional analyses performed in the last two decades have resulted in the identification of several different DLBCL subtypes. The first classification, made by the Staudt group [82], defined two distinct DLBCL subtypes, termed Germinal Center B cell-like (GCB) and Activated B cell-like (ABC) DLBCL, which represent lymphomas arising from different stages of lymphoid differentiation. As discussed below, the two subgroups differ profoundly when the mechanisms of BCR activation are concerned. Recently, these subgroups have been further subdivided based on combinations of recurrent genetic lesions harbored in the malignant cells, which presumably define different evolutionary trajectories leading to lymphomagenesis [83,84]. In parallel, another DLBCL classification has been proposed by the Shipp group, which distinguishes DLBCLs that depend on the BCR for their survival (BCR-dependent, containing both ABC and GCB DLBCLs) from DLBCLs that evolve through other pathogenetic mechanisms [85,86].Pioneering evidence for the involvement of BCR signaling in ABC DLBCL came from the study of Davis et al., which revealed that signals departing from the BCR were crucial for the activation of the NF-κB pathway, described to be essential for the survival of ABC DLBCL cells [87]. Indeed, knockdown of BTK, CD79A, or IgM was selectively toxic for a substantial proportion of ABC DLBCL cell lines, pointing to the importance of BCR signaling for the constitutive activation of NF-κB. Furthermore, total internal reflection fluorescence (TIRF) microscopy revealed prominent BCR clusters on these ABC DLBCL cell lines that were not present in cell lines derived from GCB DLBCL, BL, or MCL. Importantly, these clusters were also present on BCR-stimulated but not on unstimulated normal B cells, suggesting that they reflect active proximal BCR signaling. Of note, a substantial proportion of these cases were shown to harbor heterozygous mutations in the CD79B ITAM domain that prevent recruitment of LYN and disrupt LYN-mediated endocytic internalization of the BCR. Importantly, this study also showed that CD79B mutations amplify the BCR signal but are unable to induce BCR clustering and activation on their own. The mechanism responsible for BCR activation in these cases was revealed in a subsequent study by the same group demonstrating that reactivity with self-antigens expressed on the same or adjacent cells was responsible for generating and maintaining this chronic active form of signaling [88]. Although these modalities of BCR activation are exquisitely reminiscent of the mechanisms of BCR activation in CLL cells, a noteworthy difference is the absence of CD79B mutations in the latter disease. This could suggest that the quality of the BCR signal is not the same in ABC DLBCL and CLL or that the two diseases employ different strategies in the context of similar biological phenomena. As recently suggested, the role of CD79 mutations in DLBCL may be to block the LYN-mediated induction of anergy [89], providing an additional explanation of why these mutations are frequently selected in ABC DLBCL. Of note, the IGHV4-34 gene, which is used by almost one third of all ABC DLBCL tumors, was found to associate in almost 40% of the cases with CD79B mutations [83,88]. This gene typically encodes for autoantibodies that react with cell surface glycoproteins and is expressed on a subset of peripheral blood B cell that express low levels of surface IgM and have other features of anergy [90]. Thus, by blocking BCR internalization and by enhancing BCR signaling, CD79B mutations may prevent induction of anergy, which in the case of CLL may be primarily overcome by co-stimulatory signals from the tumor microenvironment [91,92]. The mechanisms responsible for activation of the BCR pathway in GCB DLBCL until very recently remained unknown, despite the long-standing evidence that the BCR-proximal kinase SYK is activated in a substantial proportion of GCB DLBCL cell lines and primary GCB DLBCL tumor samples [93,94,95,96]. The BCR dependency of these GCB DLBCL lines was not detected in the initial RNA interference screens that revealed the BCR-dependence of the ABC DLBCL cell lines [87], presumably because of the incomplete knockdown of BCR subunits with this technique, but was confirmed more recently using a CRISPR/Cas9 genome editing approach [97]. Parallel biochemical studies revealed that only the PI3K/AKT pathway is activated downstream of the BCR in GCB DLBCL, in contrast to ABC DLBCL where both PI3K/AKT and NF-κB are activated [86]. Another important difference between the two subsets was the absence of BCR clusters and the much lower frequency of CD79B mutations in GCB as opposed to ABC DLBCL [87,98]. Moreover, replacement of the endogenous BCR with a BCR specific for a foreign antigen did not affect the growth of GCB DLBCL cells but inhibited the growth of ABC DLBCL cells, further suggesting that BCR signaling is antigen-independent in GCB DLBCL and antigen-dependent in ABC DLBCL [98]. To further examine the potential mechanisms for BCR activation in GCB DLBCL, we recently investigated the expression of several negative regulators of this pathway in a series of BCR-dependent and BCR-independent DLBCL cell lines [26]. Remarkably, all of the BCR-dependent GCB DLBCL cell lines (n = 6) and 2 of the 4 BCR-dependent ABC DLBCL cell lines showed absent or markedly reduced expression of the phosphatase SHP1. As pointed above, SHP-1 is the principal negative regulator of the BCR pathway and functions by dephosphorylating several BCR proximal signaling molecules, including CD79A, CD79B, SYK, BLNK, and CD19 [99,100]. Accordingly, these SHP-1 negative lines showed increased levels of phosphorylated SYK and BLNK which were reduced by SHP-1 re-expression, providing direct evidence that SHP1 deficiency is at least in part responsible for the constitutive activation of the BCR pathway in GCB DLBCL [26].Previous immunohistochemical studies had reported that SHP1 is downregulated in approximately 40% of primary DLBCL tumors [101,102]. These findings were further corroborated in our study, which showed substantially lower levels of SHP1 mRNA in primary DLBCL tumors compared to normal GC B cells in approximately half of the cases, with significantly more frequent downregulation in GCB compared to ABC DLBCL (Figure 3) [26]. Moreover, this analysis revealed that SHP1 is also frequently downregulated in a substantial proportion of BL, FL, and primary effusion lymphoma (PEL) tumors (Figure 3), suggesting that SHP1 deficiency could represent a common mechanism of BCR pathway activation in several B cell malignancies. The mechanisms responsible for SHP1 downregulation in DLBCL have still not been fully elucidated, but two recent studies reported the presence of inactivating SHP1 mutations in approximately 5% of primary DLBCL tumors [83,84]. Other causes of SHP-1 downregulation could include inactivating genetic lesions in the histone methyltransferase KMT2D and genetic lesions resulting in overexpression of the transcriptional repressor BCL-6, each of which is affected in approximately 30% of DLBCL tumors. KMT2D has been shown to positively regulate SHP-1 expression [103], while BCL6 represses its transcription [104], suggesting that part of the oncogenic activity of KMTD2D and BCL6 alterations could be related to SHP1 downregulation. The findings described above suggest that two distinct mechanisms are responsible for activation of the BCR pathway in ABC and GCB DLBCL, respectively: an antigen-dependent mechanism resulting in a “chronic active” BCR signal that activates both NF-κB and PI3K and an antigen-independent mechanism resulting in an exaggerated “tonic” BCR signal that activates only PI3K. The different quality of the signals generated by these two mechanisms could potentially be explained by quantitative differences in the levels of SYK activation. In an earlier study, we showed that in unstimulated GCB DLBCL cells, SYK is primarily phosphorylated on the regulatory tyrosine (Y) at position 352, whereas phosphorylation of the regulatory tyrosines at 525/526 becomes mainly detectable following BCR crosslinking with an external ligand [95]. Phosphorylation of Y352 and YY525/526 represent two consecutive stages in SYK activation: phosphorylation of Y352 by SRC family kinases is the first stage, allowing SYK to adopt an open, catalytically active configuration, whereas trans-autophosphorylation of YY525/526 occurs in the second stage, resulting in stabilization of the activation loop and increased kinase activity (Figure 4). Importantly, experiments with phosphomimetics corresponding to these two stages of SYK activation revealed that phosphorylation of YY525/526 is associated with a drastic increase in PLCγ2 activation and a more modest increase in AKT activation. Considering that PLCγ2 is required for activation of NF-κB, this could provide an explanation for the higher NF-κB activity in BCR-dependent ABC compared to BCR-dependent GCB DLBCL cell lines. It is also worth noting that experiments with the IL-3 dependent murine B cell line BaF3 showed that only the fully activated SYK phosphomimetic induces growth-factor independent cell proliferation, whereas both phosphomimetics increased cell survival [95]. These findings suggest that the consequences of BCR pathway activation in ABC and GCB DLBCL are different and may provide a potential explanation for the different response rates to ibrutinib treatment [105]. BL in an aggressive germinal center B-cell malignancy divided into three subtypes: endemic, prevalent in young African children and associated with EBV infection; immunodeficiency related, mainly associated with HIV infection; and sporadic. Gene expression profiling studies have revealed that BL cells have a distinct molecular signature consistent with derivation from dark zone GC B cells [106].A hallmark feature of BL are translocations of the MYC oncogene resulting in aberrant MYC activation. In over 80% of cases, the translocation partner is the immunoglobulin heavy chain locus on chromosome 14, whereas in the remaining cases, MYC is translocated in the κ or λ light chain locus on chromosome 2 and 22, respectively. Importantly, the MYC translocation always spares the productive IGHV and IGLV genes that are used to construct the BCR, indicating that BL cells depend on signals from the BCR for growth or survival [107]. Initial evidence for BCR involvement in BL lymphomagenesis came from the study of Schmitz and colleagues, who reported recurrent mutations in the transcription factor TCF3/E2A and its negative regulator ID3 in 70% of sporadic BL tumors [108]. These mutations relieve TCF3 from the negative influence of ID3 and promote its constitutive activity, resulting in changes in expression of numerous transcriptional targets of TCF3. Importantly, these changes included upregulation of IGHV and IGLV gene expression and repression of SHP1, suggesting a similar mechanism of BCR pathway activation to the one previously described in GCB-DLBCL. Consistent with this possibility, inhibition of PI3K/AKT signaling and apoptosis induction was observed in the majority of BL cell lines following knockdown of TCF3, CD79A, or SYK. Further evidence for a role of the BCR/PI3K/AKT pathway in BL pathogenesis was provided by Varano and colleagues, who demonstrated in a MYC transgenic mouse model of BL that BCR-negative lymphoma cells can proliferate and survive in vitro but exhibit a competitive growth disadvantage compared to their BCR-positive counterparts [109]. The competitive growth advantage of the BCR-positive BL cells was driven mainly by PI3K/AKT-dependent inhibition of the kinase GSK3β, which facilitated G1 to S phase cell cycle transition and increased cell survival. These effects were more pronounced under conditions of nutrient deprivation, suggesting that BCR signals increase the metabolic fitness of the malignant cells. Similar findings were reported by another group using the human BL cell line Ramos, which showed that the signals that provide the competitive advantage to BCR-positive BL cells involve interactions between LYN, SYK, CD19, and CD79B [110]. Interestingly, the survival of BCR-dependent GCB-DLBCL cell lines also requires LYN and CD19 [97], providing another similarity between BL and GCB-DLBCL and pointing to a common mechanism of PI3K/AKT activation downstream of the BCR in these two diseases. It should be noted, however, that another recent study reported that BL cells do not have a hyperactivated PI3K-AKT pathway and are not sensitive to AKT knockdown or inhibition in contrast to GCB DLBCL cells [111]. These data argue against a growth-promoting role for the PI3K/AKT pathway in BL and underline the need for additional studies on the role of the BCR pathway in the pathogenesis of this disease. FL is an indolent B cell malignancy derived from GC B cells [112]. Its hallmark feature is the t(14;18) (q32;q21) translocation, which is present in approximately 90% of the patients. The translocation brings the BCL-2 gene into the IGHC gene locus, resulting in constitutive BCL2 expression and apoptosis evasion. As in the case of BL, the translocation always spares the productively rearranged IGHV allele, allowing for the expression of a functional BCR. An important characteristic of FL is the ongoing SHM, with the consequent acquisition of novel mutations and intraclonal variation of the IGHV and IGLV genes. The random introduction of these mutations would be expected to result in stop codons and loss of immunoglobulin expression. However, FL tumors always maintain surface immunoglobulin, indicating a selective force that favors retention of a functional BCR. Accordingly, treatment of FL patients with anti-idiotype antibodies was associated with the outgrowth of escape variants that still expressed a functional BCR but were no longer recognized by the anti-idiotype antibody because of SHM in the targeted V region sequence [113]. Moreover, in vitro studies have shown that pharmacological inhibition or RNAi-mediated knockdown of SYK in FL cell lines results in inhibition of the PI3K/AKT/mTOR pathway, with consequent growth arrest and reduced invasive capacity, further implicating the BCR pathway in the pathogenesis of FL [114,115]. Available evidence suggests that three mechanisms may be responsible for the activation of the BCR pathway in FL. Studies from the Stevenson group have revealed that 80% of FL patients carry SHM-introduced N-glycosylation sites in the functional IGHV gene, most of which are found in the complementarity-determining regions [116]. More recent next-generation IGHV sequencing studies confirmed these data, observing acquisition of N-glycosylation sites in 96% of FL subclones, with loss of negative subclones in successive events of disease progression [117]. The N-glycosylation sites represent cues for the addition of “high-mannose” glycans, which bind to the lectin DC-SIGN that is overexpressed on M2 macrophages and dendritic cells in FL lymph nodes [118]. This interaction has been shown to induce a continuous low-level BCR signal in the tumor cells, resulting in SYK, ERK, and AKT activation, Ca2+ flux, and induction of MYC expression [118,119,120]. Another potential mechanism of BCR pathway activation in FL is binding to self-antigens. Using permeabilized human HEp-2 cells as a screen for human tissue antigens, Sachen and colleagues tested 98 FL immunoglobulins and observed autoreactivity in 26% of the cases [121]. For one tumor immunoglobulin, the self-antigen was identified as myoferlin, a protein associated with cell and nuclear membranes. Interestingly, reactivity with myoferlin was also observed in another study that used a combinatorial peptide library to identify ligands for three FL immunoglobulins [122]. The myoferlin peptide sequence that was recognized by one of these antibodies was found to be identical to a sequence in a surface protein from Streptococcus mitis and Pneumocystis jirovecii, suggesting crossreactivity with foreign antigens.Consistent with these findings, another study reported reactivity of FL-derived immunoglobulins with HEp-2 cells in 11% of the 217 investigated cases [123]. A considerable proportion of these cases reacted with vimentin, which is an intracellular filament protein that is expressed on the surface of activated and apoptotic T cells, macrophages, neutrophils, and platelets. Interestingly, the percentage of cases reacting with the N-terminal region of vimentin was even higher than the percentage of cases reacting with HEp-2 cells, indicating the existence of different conformational epitopes that can be recognized by the tumor immunoglobulins. Apart from aberrant BCR engagement, the addiction of FL to BCR-related pathways could potentially be caused by mutations in molecules that transduce the BCR signal. Indeed, whole exome sequencing (WES) analysis identified recurrent mutations in the interconnected BCR and CXCR4 pathways in almost 45% of FL patients, including mutations in BTK, SYK, BLNK, CD22, PLCG2, EGR1, and EGR2 [124]. Although the functional consequences of these mutations are still unclear, an effect on BCR signal transduction would be expected. Perhaps equally relevant is the fact that KTM2D is mutated in approximately 60% of the FL tumors [112,124] and may account for the frequent downregulation of SHP1 that was previously mentioned. Accordingly, SHP-1 deficiency could cooperate with expression of mannosylated immunoglobulins to amplify the BCR signal generated by mannose-binding lectins, as already proposed [118]. Considering that N-glycosylation sites are also present in approximately 40% of DLBCL immunoglobulins [116], it is possible that these two mechanisms cooperate in activating the BCR pathway also in GCB DLBCL. A hallmark of MCL is the t(11;14)(q13;q32) translocation which brings the CCDN1 gene into the IGHC gene locus, resulting in cyclin D1 overexpression. Other frequent genetic alterations include mutation or deletion of ATM and TP53 and copy number alterations that typically affect genes related to cell cycle regulation, DNA damage response, and cell survival, including CDKN2A, RB1, MYC, CDK4, or BCL2. In addition, recurrent mutations in chromatin modifiers and genes belonging to the NOTCH and NF-κB pathways have been identified in a substantial proportion of cases (reviewed in Reference [125]).The malignant cells typically express unmutated IGHV genes (>98% homology to germline sequence), although around 30% of cases show a higher level of SHM [126,127]. In favor of an antigen-driven lymphomagenesis, expression of stereotyped BCRs has been reported in approximately 10% of MCL cases [128,129]. Additional evidence for a potential role of the BCR pathway in the pathogenesis of MCL was provided by phospho-proteomic analysis of MCL cell lines and primary tumor samples demonstrating constitutive activation of multiple BCR signaling molecules, including CD79B, LYN, SYK, BLNK, BTK, and PLCγ2 [130]. Treatment of these cell lines with SYK inhibitors induced cell cycle arrest and apoptosis, suggesting that BCR signals regulate the proliferation and survival of the malignant cells [131]. A growth inhibitory effect was also observed with ibrutinib against a subset of MCL cell lines characterized by activation of the cannonical NF-κB pathway [132]. In contrast, ibrutinib was ineffective against MCL cells lines with constitutive activation of the noncannonical NF-κB pathway caused by genetic lesions in the regulatory components TRAF2 and BIRC3.A study from Wiestner’s group also revealed the existence of two MCL subsets that differ with respect to the mechanism of NF-κB activation. In this study, transcriptome analysis of lymph node and peripheral blood MCL cells identified a subset with higher expression of BCR and NF-κB target genes in lymph node MCL cells and a subset with equal expression of BCR and NF-κB target genes in the two cell populations [133]. Cases from the first subset showed higher expression of phosphorylated SYK, PLCγ2, AKT, ERK, and the NF-κB p65 subunit in lymph node compared to peripheral blood MCL cells, consistent with a BCR-dependent mechanism of NF-κB activation.The mechanisms that activate the BCR pathway in MCL have still not been fully elucidated, although most data indicate a role for autoantigen stimulation. Similar to FL, the tumor immunoglobulins in MCL have been reported to bind HEp-2 antigens, including vimentin, in about one-third of the cases [123]. In another study, reactivity with the autoantigen low-density lipoprotein receptor-related protein-associated protein 1 (LRPAP1) was identified in 36% (10/28) of investigated tumor immunoglobulins [134]. In addition, given the similarities with CLL and the finding that MCL cells frequently present neoantigenic peptides derived from the lymphoma immunoglobulin heavy- or light-chain variable regions to idiotype-specific T cells [135], it remains possible that the BCR pathway is activated in a subset of cases by cell-autonomous BCR interactions. Such interactions were not detected by Dühren-von Minden et al. in their seminal study [69], but the small number of investigated MCL immunoglobulins in that study does not entirely exclude this possibility.MZL is a B cell malignancy comprised of three distinct entities: the extranodal MZL of mucosa-associated lymphoid tissue (MALT lymphoma), splenic MZL (SMZL), and nodal MZL (NMZL). The cells of origin are marginal zone B cells, which act as a first line of defense against infectious agents and are responsible for the mounting of a rapid, innate-like antibody response against both T cell–dependent and T cell–independent antigens. The three MZL subtypes share common genetic lesions and deregulated pathways and present subtype-specific alterations. The NF-κB pathway is activated in approximately 50% of cases from all three subsets by loss-of-function lesions in the negative regulators TNFAIP3 (A20), BIRC3, and TRAF3, or translocations resulting in deregulated expression of the positive regulators MALT1 or BCL10. Other frequent genetic lesions include mutations that activate the NOTCH pathway, which are present in approximately 40% of SMZL and NZML cases but in less than 5% of MALT lymphomas, and mutations that inactivate the transcription factor KLF2, which are detected in 20–40% of SMZL and 20% of NZML cases (reviewed in Reference [136]). MZL was one of the first B cell malignancies with evidence for a role of chronic antigen stimulation in the pathogenesis of the disease. Specific disease entities are frequently associated with certain chronic bacterial or viral infections, such as gastric MALT lymphomas with Helicobacter pylori, SMZL with Hepatitis C virus (HCV), ocular adnexal MALT lymphomas with Chlamydia psittaci, cutaneous MALT lymphomas with Borrelia burgdorferi, and immunoproliferative small intestine disease with Campylobacter jejuni infection [137,138,139,140,141]. The pathogenic role of chronic antigen stimulation in these MZL entities is further supported by the capacity of antimicrobial drugs to induce tumor regression upon eradication of the infectious agent in patients at early stages of the disease [142,143,144]. Beside infection, chronic antigen stimulation in the context of certain autoimmune disorders may also contribute to the development of MZL, given the association of Sjögren syndrome and Hashimoto’s thyroiditis with MALT lymphomas of the salivary gland and thyroid, respectively [145,146].The malignant B cells in MZL typically express somatically mutated IGHV genes with a pattern of mutations consistent with antigen selection. In addition, biased IGHV gene usage and expression of stereotyped BCRs is frequently observed, further hinting to a BCR-driven process. The most frequently overrepresented genes include IGHV1-69, which is predominantly expressed in HCV-associated MZLs and salivary gland and gastric MALT lymphomas, and IGHV1-02, which is expressed in more than one third of splenic MZLs not associated with HCV infection. In addition, the IGHV3-30, IGHV3-23, IGHV3-7, and IGHV4-34 genes are frequently overrepresented, particularly in splenic MZL and gastric and orbital adnexal MALT lymphomas [137,147,148,149,150,151,152,153].The IGHV1-69 immunoglobulins in MZL often contain the IGKV3-20 light chain and typically display rheumatoid factor (RF) activity [137,154]. RF activity has also been reported for MZL immunoglobulins encoded by other overrepresented IGHV genes, including the stereotypic combination IGHV3-7/IGKV3-15 [137,155]. Moreover, reversion of IGHV and IGKV mutations in these RFs to germline configuration reduced the affinity for IgG, suggesting that the tumor cells were selected for their capacity to bind to IgG or IgG-containing immune complexes [155]. In addition to RF activity, reactivity of MZL immunoglobulins with other self antigens has been reported, including insulin, thyroglobulin, galactosidase, galectin-3, stomach extract, and other human tissue antigens [151,156,157]. This polyreactive binding pattern resembles the reactivity of CLL immunoglobulins, although the affinity of the latter is considerably lower. Interestingly, despite the frequent association of MZL with chronic infections, most evidence suggests that the tumor imunoglobulins do not bind directly to the microbial antigens [137,149]. These findings suggest that the BCR pathway in MZL is primarily activated by autoantigens and indicate that the role of the infectious agents may be to provide costimulatory signals rather than activate the BCR pathway [158]. The signaling pathways that are activated downstream of the BCR in MZL cells have not been investigated in detail, but analysis of primary tumor samples by phospho-flow cytometry demonstrated constitutive activation of SRC family kinases, SYK, PLCγ2, and p65 NF-κB [159]. Interestingly, some of the more frequent genetic defects in MZL can potentially augment canonical NF-κB activation induced by stimulation of the BCR. These most notably include inactivating mutations in the negative regulator TNFAIP3 (A20), which are detected in ~30% of MALT lymphomas and in 10–15% of splenic and nodal MZLs, and inactivating mutations in the transcription factor KLF2, which have been reported in 20–40% of SMZL and in 17% of NZML (reviewed in References [160,161]. In contrast, other frequent genetic defects, such as the t(1;14)(p22;q32), t(14;18)(q32;q21), and t(11;18)(q21;q21) translocations, which are detected in over 25% of gastric and over 50% of lung MALT lymphomas, result in overexpression of BCL10 or MALT1 and constitutive activation of the canonical NF-κB pathway [161]. Interestingly, these translocations are not seen in cases with RF BCRs and have been associated with resistance to H. pylori eradication therapy, suggesting that they may define a subset of MALT lymphomas that do not depend on BCR-mediated NF-κB activation for their expansion [137]. WM is defined as lymphoplasmacytic lymphoma associated with monoclonal IgM. It is a rare and indolent B-cell malignancy, characterized by the infiltration of cells in the bone marrow and extramedullary sites. Studies analyzing the genomic landscape of WM have identified mutations in the Toll-like receptor (TLR)-adaptor protein MYD88 and the chemokine receptor CXCR4 (present in >90% and 30–40% of cases, respectively) as the principal hallmarks of the disease. In addition, deletions in chromosome 6q21-25 occur in about 50% of patients and affect the expression of several genes involved in BCR signaling, including the BTK inhibitor IBTK, the transcription factor FOXO3, and the NF-κB regulators TNFAIP3 and HIVEP2. Other recurrent genetic defects in WM include mutations in CD79A and CD79B, which have been reported in 7–15% of the cases (reviewed in Reference [162]). Some of the initial evidence for a pathogenic role of the BCR in WM came from its association with several autoantibody-mediated autoimmune disorders, such as cold agglutinin disease, type II mixed cryoglobulinemia, and certain polyneuropathies (reviewed in Reference [163]). The finding that type II mixed cryoglobulinemia can evolve into WM and that the tumor immunoglobulins in these cases typically bind to HCV-containing immune complexes further substantiated this possibility [164,165]. The tumor immunoglobulins in HCV-associated WM were shown to express a biased IGHV gene repertoire with predominant usage of the RF-encoding IGHV1-69/IGKV3-20 combination, consistent with antigen-driven selection [166]. Subsequent studies investigating WM without HCV infection also showed a restricted IGHV repertoire, although with overrepresentation of other genes, such as IGHV3-23, IGHV3-7, and IGHV3-74 [167,168,169]. Additional support for an involvement of the BCR pathway in the pathogenesis of WM came from observations that a number of BCR signaling molecules, including SFK, SYK, BTK, BLNK, PLCγ2, ERK, AKT, and NF-κΒ, are constitutively activated in primary WM tumor cells [170]. Moreover, treatment of WM cell lines with BTK or SYK inhibitors resulted in cell cycle arrest, apoptosis, and inhibition of AKT and ERK signaling [171,172]. Interestingly, both BTK and SYK have been shown to directly interact with mutated MYD88 in WM cells [171,173]. Moreover, knockdown of MYD88 or use of a MYD88 inhibitor reduced the levels of activated SYK, BTK, AKT, and STAT3 in these cells, suggesting a novel mechanism of BCR pathway activation that involves direct cooperation with the TLR pathway [173]. A similar mechanism has been described in ABC DLBCL, where a signaling complex composed of MYD88, TLR9, and IgM has been identified in the endolysosomal compartment of MYD88-mutated cells [97]. This complex has been shown to drive the sustained NF-κB activity in ABC DLBCL cells and presumably mediates the same effect in WM cells.The identification of the BCR as a major therapeutic target in various B cell malignancies resulted in the clinical testing and approval of multiple drugs that block signaling through this receptor. Most of these drugs target BTK or PI3K, although inhibitors of SYK, SRC family kinases, and mTOR have also demonstrated activity in clinical trials (Table 1). The primary mechanism of action of these drugs is inhibition of BCR-induced proliferation and survival signals [174,175,176,177], although inhibition of other BCR-regulated processes has also been shown to contribute to the activity of these drugs, particularly in CLL. Specifically, inhibition of BCR-mediated integrin activation has been shown to reduce the adhesion of the leukemic cells to extracellular matrix and cell surface adhesion proteins, whereas inhibition of BCR-mediated chemokine secretion has been shown to result in reduced recruitment of T cells and monocytes [178,179,180,181,182]. These effects reduce the retention of the leukemic cells in the nurturing lymph node microenvironment and prevent them from receiving growth and survival signals from accessory cells. In addition, part of the activity of these drugs has been attributed to their capacity to interfere with signaling through other receptors that regulate leukemic cell migration, proliferation, and survival, including CD40L, BAFF, IL-4, CXCR4, and TLR9 [29,183,184]. These effects would be expected to further deprive CLL cells of growth and survival signals in the microenvironment, resulting in “death by neglect” [185]. The importance of these non-BCR effects on clinical responses to BCR inhibitor treatment in B cell malignancies other than CLL are less certain at present, although similar effects on malignant cell migration and adhesion have been reported in MCL and WM [30,186].Fostamatinib was the first BCR inhibitor that was tested in B cell malignancies, demonstrating potent preclinical activity in DLBCL and CLL [49,94,174]. A subsequent phase 1/2 clinical trial in patients with relapsed or refractory (R/R) B cell malignancies showed an objective response rate (ORR) of 55% for CLL, of 22% for DLBCL, of 10% for FL, and of 11% for MCL [196]. Another phase 2 clinical trial conducted only in patients with R/R DLBCL showed an ORR of 3%, indicating low single-agent activity of fostamatinib in DLBCL [217]. Similar results were obtained with the more selective SYK inhibitor entospletinib, which showed a 61% ORR in patients with R/R CLL and a more modest effect in other B cell malignancies, including a 35.3% ORR in LPL/WM, 17.9% in MCL, 17.1% in FL, 11.8% in MZL, and no responses in DLBCL [197,204,208]. Entospletinib also demonstrated clinical activity in R/R CLL patients that had received prior treatment with a BTK or PI3Kδ inhibitor, with 32.7% of patients responding to treatment, including 2 of 8 patients with Richter Transformation [198]. Cerdulatinib, another SYK inhibitor that is currently under development, also showed promising results for CLL and FL but not in DLBCL or MCL in a phase 1 clinical trial on patients with B cell malignancies [199]. Altogether, these data suggest that SYK inhibitors are active mainly in CLL, with more modest activity in other B cell malignancies.Ibrutinib is a small molecule inhibitor that irreversibly inhibits BTK by covalently binding to cysteine 481 in the ATP-binding pocket. In vitro studies showed that ibrutinib mainly inhibits the BTK/PLCG2/PKC/NF-κB axis [218], although inhibitory effects on AKT and ERK have also been reported [27]. Ibrutinib was initially tested in a phase 1 study of patients with R/R B cell malignancies, which showed an ORR of 69% in CLL, of 78% in MCL, of 75% in WM, of 38% in FL, of 29% in DLBCL, and of 25% in MZL [189]. These remarkable results encouraged a phase 1b/2 study in 85 heavily pretreated R/R CLL patients, displaying an ORR of 71%. The ORR further increased after a 3-year follow-up, rising to 89% [219]. Ibrutinib also displayed a high ORR in phase 2 studies of MCL (68%), WM (91%), and MZ (48%%), resulting in its subsequent approval in these disease settings [209,215,220]. Lower activity was observed in phase 2 clinical trials of FL (ORR 38%) and DLBCL (ORR 25%), with the majority of the responding patients belonging to the ABC DLBCL subset (ORR of 37% in ABC DLBCL and 5% in GCB DLBCL) [105,205]. Notably, ibrutinib inhibits not only BTK but also other kinases such as ITK, TEC, EGFR, JAK3, ErbB, and SRC family kinases. Inhibition of some of these kinases has been shown to cause some of the side effects of ibrutinib, prompting the development of more selective BTK inhibitors. Acalabrutinib, a more specific molecule, in phase 2 studies of R/R CLL and R/R MCL displayed an ORR of 95% [190] and 81% [221], respectively, leading to its approval for treatment of patients with MCL that had received at least one prior line of therapy. Tirabrutinib, another more selective BTK inhibitor, also displayed considerable activity in patients with CLL and MCL, with 96% and 92% of patients responding to treatment, respectively. Tirabrutinib also displayed similar activity as ibrutinib in DLBCL, with an ORR of 35% in patients with non-GCB DLBCL [191]. Idelalisib is an orally available reversible inhibitor of PI3K with 30–400 times greater selectivity for PI3Kδ compared to PI3Kα, PI3Kβ, and PI3Kγ. Phase 1 trials in R/R CLL and R/R MCL patients showed ORRs of 72% and 40%, respectively [192,210], while a phase 2 trial in patients with relapsed indolent B cell malignancies resulted in ORRs of 61% in CLL, of 54% in FL, of 47% in MZL, and of 80% in LPL/WM [193]. A subsequent phase 3 study of idelalisib in combination with rituximab in relapsed CLL displayed an ORR of 81% [222]. These trials led to the approval of idelalisib as a single agent in FL and SLL patients with at least 2 prior therapies and in combination with rituximab for second line treatment of CLL. Several second-generation PI3K inhibitors have more recently been developed, including the PI3Kγ/δ inhibitor Duvelisib and the PI3Kα/δ inhibitor Copanlisib. In a phase 2 trial of patients with refractory indolent NHL, duvelisib treatment resulted in ORRs of 67.9% in SLL, of 42.2% in FL, and of 38.9% in MZL [207], whereas an ORR of 74% was reported in a phase 3 trial of R/R CLL [194]. Copanlisib has also shown promising results in phase 2 trials of patients with R/R lymphoma, resulting in ORRs of 75% in SLL, of 59% in FL, of 70% in MZL, and of 64% in MCL [195,211]. More recently, copanlisib was tested in a phase 2 trial of patients with relapsed/refractory DLBCL, demonstrating ORRs of 31.6% and 13.3% in ABC and GCB DLBCL patients, respectively [19]. The clinical data accumulated over the last decade demonstrate remarkable therapeutic efficacy of BCR inhibitors in patients with various B cell malignancies. Most of these data were obtained with the BTK inhibitor ibrutinib and the PI3Kδ inhibitor idelalisib, which induce clinical responses in the vast majority of patients with CLL, MCL, and WM and in approximately half of the patients with FL and MZL (Table 1). In addition, approximately one third of patients with ABC DLBCL initially respond to treatment with ibrutinib or the PI3Kα/δ inhibitor copanlisib, whereas responses are infrequent in patients with GCB DLBCL [19,105]. No clinical studies have been reported yet with BCR inhibitors in BL except for one case report [223]. Overall, the response rates in the various B cell malignancies largely correlate with the proportion of cases that have an activated NF-κB pathway downstream of the BCR (Table 2). As described previously, the majority of CLL, MCL, and WM cases demonstrate high NF-κB activity that can be reduced by treatment with a BCR inhibitor [133,171,176]. The proportion of such cases is lower in MZL and ABC DLBCL, which is consistent with the lower response rates in these diseases. Constitutive activation of the p65 subunit of NF-κB has also been detected in approximately 20% of primary FL tumors, although this has not yet been correlated with sensitivity to BCR inhibitor treatment [224]. In the majority of cases that do not respond to BCR inhibitor treatment, the NF-κB pathway is not activated or is activated by BCR-independent events. The latter mechanism appears to be particularly frequent in MZL and ABC DLBCL, where the NF-κB pathway is activated either by BTK-distal events, such as genetic lesions in BCL10, MALT1, and CARD11, or by genetic lesions in the non-canonical pathway, such as mutations in BIRC3 and TRAF3. Such mutations can be detected, although less frequently, also in MCL and CLL and have been associated with resistance to BTK-inhibitor treatment [105,132]. As previously discussed, the majority of FL, GCB DLBCL, and BL tumors have low NF-κB activity despite evidence for constitutive BCR activation. A major difference with respect to tumors with high NF-κB activity is the mechanism of BCR pathway activation. In FL, GCB DLBCL, and BL, available data suggest that the BCR pathway is activated by antigen-independent mechanisms that involve interactions with mannose-binding lectins and/or deficiency of the negative regulator SHP1. These mechanisms have been shown to induce signals of low intensity that are sufficient to activate ERK, AKT, and NFAT but apparently do not reach the threshold required for NF-κB activation [26,108,118,119,120,225]. Activation of NF-κB requires a large transient increase in intracellular Ca2+, such as occuring following acute engagement of the BCR with an external antigen [225]. Binding of the tumor immunoglobulins to autoantigens could provide such a signal, particularly in U-CLL, MZL, and WM, where the malignant cells frequently express autoreactive BCRs. The mechanism responsible for the activation of the NF-κB pathway in M-CLL is still unclear, considering that cell autonomous BCR–BCR interactions would not be expected to induce the high-amplitude Ca2+ elevations required for NF-κB activation. In this subset, it remains possible that the therapeutic activity of the kinase inhibitors is primarily mediated by BCR-independent mechanisms, such as inhibition of leukemic cell migration and adhesion. Clinical trials in tumors with an activated BCR pathway and low NF-κB activity, such as GCB DLBCL, have shown very modest efficacy of BCR inhibitors as single agents. However, such tumors typically display increased basal activity of SYK, PI3K, AKT, and GSK3, which regulate the expression of several BCL-2 family proteins, including MCL-1, BIM, and HRK [26,86,226,227]. Notably, treatment of such tumors with SYK, PI3K, and to a lesser extent BTK inhibitors results in reduced expression of the antiapoptotic protein MCL-1 and increased expression of the proapototic proteins BIM and HRK [26,227]. These changes induce relatively little apoptosis on their own but can sensitize the malignant cells to cytotoxic drugs, such as venetoclax. Together, these data suggest that combinations of BCR inhibitors with venetoclax deserve further study in tumors characterized by an activated BCR and overexpression of BCL2. Although the different BCR inhibitors have still not been compared in randomized clinical trials, available data indicate higher response rates and more durable responses with BTK inhibitors compared to PI3Kδ and particularly compared to SYK inhibitors (Table 1). This is somewhat surprising considering that SYK and PI3Kδ are positioned more proximally than BTK, and therefore, their inhibition would be expected to result in more complete inhibition of the BCR pathway, at least with respect to PI3K/AKT signaling. One possible explanation for the greater activity of BTK inhibitors is the more persistent inhibition of the BCR signal caused by the covalent interaction of these drugs with cysteine 481 in the ATP binding domain of the enzyme. In support of this possibility, mutations in cysteine 481 that alter the irreversible covalent binding of ibrutinib to a reversible interaction lead to drug resistance [228,229]. Moreover, a recent study comparing twice a day and once a day dosing of acalabrutinib in patients with CLL showed that the lower BTK occupancy in the latter cases was associated with a lower degree of BCR pathway inhibition and a lower ORR [230]. Collectively, these data suggest that the durability of BCR pathway inhibition is an important determinant of clinical response to drugs that target the BCR pathway.The substantial differences in response rates and response duration to BCR inhibitor treatment suggest variable dependencies on BCR signaling and distinct mechanisms of BCR pathway activation between different B cell malignancies. In tumors with highest response rates, the BCR is typically activated by antigen-dependent mechanisms, resulting in the activation of multiple downstream signaling pathways, including NF-κB and PI3K/AKT. In tumors where the BCR pathway is activated by antigen-independent mechanisms, the NF-κB pathway is not activated or is activated by other mechanisms and responses to single-agent BCR inhibitor treatment are considerably lower. However, targeting the BCR pathway in such tumors leads to inhibition of PI3K/AKT signaling and changes in the expression of apoptosis regulatory proteins that can sensitize the malignant cells to cytotoxic drugs. Combinations of BCR inhibitors with cytotoxic agents such as venetoclax may provide a therapeutic benefit also in these patients and warrant future clinical trials. Future work should focus on the development of clinically useful biomarkers to identify patients with an activated BCR pathway and to select the optimal BCR inhibitor for individual patient treatment. Writing—Original draft preparation, S.T.; Writing—Review and editing, L.L.; Original Concept, Writing—Review and editing, D.G.E. All authors have read and agreed to the published version of the manuscript.The study was supported by the Associazione Italiana per la Ricerca sul Cancro (AIRC), Investigator Grant IG2016 Id.19236, Milan, Italy (to D.G.E) and Progetto Ricerca Finalizzata PE-2016-02362756, Ministero della Salute, Rome, Italy (to D.G.E.).The authors declare no conflict of interest.Overview of B-cell receptor signaling pathways.B cell development and cell of origin of B cell malignancies covered in this review.SHP1 expression in different B cell malignancies: (left) Comparison of SHP1 mRNA levels in normal and malignant B cell subsets analysed in dataset GSE2350 (CB, centroblasts; CC, centrocytes; mem. B cell, memory B-cells; HCL, Hairy Cell Leukemia; PEL, Primary Effusion Lymphoma). (right) Comparison of SHP1 expression levels in germinal center B-cell-like (GCB and Activated B cell-like (ABC) diffuse large B cell lymphoma (DLBCL) tumors analysed in dataset GSE4732.Mechanisms of B cell receptor (BCR) pathway activation in ABC and GCB DLBCL.Responses to BCR inhibitor treatment in different B cell malignancies.* R/R, relapsed/refractory; TN, treatment naive; SLL, small lymphocytic lymphoma; n/a, not available.Common BCR-related features in different B cell malignancies.

Med-MDPI/cancers/cancers-12-06-01397.txt

@@ -0,0 +1 @@
+Nanotechnology-based approaches hold substantial potential to avoid chemoresistance and minimize side effects. In this work, we have used biocompatible iron oxide magnetic nanoparticles (MNPs) called MF66 and functionalized with the antineoplastic drug doxorubicin (DOX) against MDA-MB-231 cells. Electrostatically functionalized MNPs showed effective uptake and DOX linked to MNPs was more efficiently retained inside the cells than free DOX, leading to cell inactivation by mitotic catastrophe, senescence and apoptosis. Both effects, uptake and cytotoxicity, were demonstrated by different assays and videomicroscopy techniques. Likewise, covalently functionalized MNPs using three different linkers—disulfide (DOX-S-S-Pyr, called MF66-S-S-DOX), imine (DOX-I-Mal, called MF66-I-DOX) or both (DOX-I-S-S-Pyr, called MF66-S-S-I-DOX)—were also analysed. The highest cell death was detected using a linker sensitive to both pH and reducing environment (DOX-I-S-S-Pyr). The greatest success of this study was to detect also their activity against breast cancer stem-like cells (CSC) from MDA-MB-231 and primary breast cancer cells derived from a patient with a similar genetic profile (triple-negative breast cancer). In summary, these nanoformulations are promising tools as therapeutic agent vehicles, due to their ability to produce efficient internalization, drug delivery, and cancer cell inactivation, even in cancer stem-like cells (CSCs) from patients.Approximately 50% of human cancer treatments are based on chemotherapy, being one of the most common strategies, mainly in metastatic disease. However, this method causes many side effects as well as multidrug resistance (MDR), one of the most severe problems during the treatment of oncological patients [1]. In this way, chemotherapy is usually employed to reduce the tumour size or when metastasis is produced and other strategies are not possible [2,3]. Generally, MDR reflects an overexpression of ATP-binding cassette (ABC) transporters on the plasma membrane, which are capable of causing an efflux of several anticancer drugs, such as doxorubicin, epirubicin, and paclitaxel [4,5].Recently, nanotechnology-based formulations, or nanomedicine, have become a promising strategy for cancer treatment. This approach not only provides a new opportunity to overcome MDR, avoiding drug efflux, but also increases its retention time [6,7]. Magnetic nanoparticles (MNPs), which can be targeted to the tumour site with a magnetic field, gained importance in cancer therapy in recent years. This class of material includes metallic, bimetallic, and superparamagnetic iron oxide nanoparticles, which have a reactive surface that can be readily modified with biocompatible coatings and loaded with therapeutic agents [8,9]. The use of non-covalent functionalized magnetic nanoparticles with drugs to improve cancer therapy has been extensively explored to maximize the therapeutic activity of the drugs while minimizing their side effects [10,11,12]. However, there are still some drawbacks, such as the poor control of the release of the immobilized drug. Therefore, we propose the covalent functionalization of magnetic nanoparticles with drugs through stimuli-responsive linkers as a suitable way to tackle the controlled release. There is a growing interest in developing nanocarriers that can target disease sites in a specific manner and perform a gradual slow release, avoiding rapid efflux of chemotherapeutic drugs. Ideally, only cancer cells should be targeted by drug-functionalized magnetic nanoparticles, in which drugs are inactive until they are released inside the cells. In this regard, different linkers sensitive to intracellular triggering stimuli such as pH [13], reducing environment [14] and the presence of some enzymes [15], or external stimuli such as temperature [16] have been employed to attach and release drugs from magnetic nanoparticles in a controlled manner.Regarding the specific targeting of tumours, recent research suggested that cancer originates from a small fraction of tumour initiating cells with self-renewal capability, unlimited propagation and multipotent differentiation. This small population of tumour cells has been called cancer stem-like cells (CSCs) [17,18]. Moreover, CSCs in tumours evade the anticancer effects of standard chemotherapy, emerging as an underestimated biological barrier to the success of systemic chemotherapy, inducing resistance to therapy and thus the subsequent tumour recurrence [19].In this work, iron oxide magnetic nanoparticles were loaded with doxorubicin (DOX), an anticancer agent frequently used to treat different cancers [20]. Electrostatic and covalent strategies were employed to obtain DOX-loaded nanocarriers. The nanocarriers consist of a magnetic iron oxide nanoparticle (MNPs) with a coating of dimercaptosuccinic acid (DMSA) easily modifiable with different drug molecules. We applied different functionalization strategies to make a systematic comparative study to define the nanodelivery strategies that show advantageous therapeutic properties. Covalent delivery systems were designed by using three different linkers: disulfide (DOX-S-S-Pyr), imine (DOX-I-Mal), or both (DOX-I-S-S-Pyr). The purpose was to analyse which system was more effective in eliminating breast tumour cells, including breast cancer stem-like cells. Our results indicate that double-sensitive nanoparticles MF66-S-S-I-DOX have many of the properties that an ideal nanocarrier should have as stability, biocompatibility, sufficient loading capacity and ability to maintain its cytotoxic activity after their internalization by tumour cells, including against cancer stem cells.Firstly, biocompatible magnetic nanoparticles (MNPs) coated with dimercaptosuccinic acid (DMSA) called MF66 were successfully functionalized with DOX through electrostatic interactions. Secondly, DOX was covalently bound via: (1) disulfide bond (DOX-S-S-Pyr), a reducing environment sensitive linker; (2) imine bond (DOX-I-Mal), a pH-sensitive linker; (3) or both disulfide and imine bonds (DOX-I-S-S-Pyr) (see the Supporting Information for the detailed functionalization process and full physico-chemical characterization of resulting DOX-functionalized MNPs).The release of the attached molecules from the electrostatic functionalized MF66 was monitored for 120 h in milliQ® water, PBS buffer (150 mM NaCl, 10 mM phosphate pH 7.4, extracellular pH) and AcONa/AcOH buffer (150 mM NaCl, 10 mM sodium acetate pH 4.7, intracellular pH) at 37 °C (Figure 1a).These results show that physiological salt concentration in both PBS and acetate buffer caused the continuous dissociation of the molecules of DOX from the electrostatic functionalized MF66 and that the release rate was independent on the pH.The release of the DOX from the covalent functionalized MF66 was monitored during 8 h in PBS buffer (150 mM NaCl, 10 mM phosphate pH 7.4) and AcONa/AcOH buffer (150 mM NaCl, 10 mM sodium acetate pH 4.7) using intracellular reducing conditions (1 mM of 1,4-dithiothreitol (DTT)) and extracellular reducing conditions (1 μM of DTT) at 37 °C (Figure 1b1).These results show that in the case of MF66 functionalized with DOX-S-S-Pyr, a reducing environment caused the rapid dissociation of the drugs from MNPs and that the release rate was strongly dependent on the reducing environment. In the case of MF66 functionalized with DOX-I-Mal, an acidic environment caused the rapid dissociation of the drugs from MNPs, the release rate being strongly dependent on the pH (Figure 1b2).Finally, in the case of MF66 functionalized with DOX-I-S-S-Pyr, both a reducing environment and/or acidic environment caused the rapid dissociation of the drugs from MNPs. Therefore, the cargo from the described nanoparticles is only released under reducing conditions and or acidic conditions as those present in the cytoplasm and lysosomal environments and is stable under extracellular neutral and low reducing conditions (Figure 1b3).As observed in Figure 2a, a decrease in cell viability was observed in samples treated with MF66-DOX (53.0% ± 4.3%). However, the cytotoxicity induced by free DOX was clearly higher (see Figure S1 of the Supplementary Material).Staining with Prussian blue allowed us to visualize an efficient cell labelling with MF66. It also showed that incubation with MF66-DOX caused morphological alterations 72 h after treatment (24 h incubation with the different treatments). Cells incubated with bare MF66 showed no changes in cellular morphology (Figure 2b). However, in cells treated with MF66-DOX, a higher rate of mitosis appeared and most of the divisions were aberrant (multipolar, with altered mitotic spindles and misaligned chromosomes). In addition, the percentage of giant cells increased in cells incubated with DOX nanoformulation related to bare MF66 (43.3% ± 5.9% vs 1.9% ± 2.7%). Moreover, these cells had a single large nucleus or several nuclei, most of them with a smaller size or micronuclei. Finally, we also detected cells with apoptotic morphology (14.1% ± 3.9%).Flow cytometry by DNA staining with propidium iodide displayed a clear alteration of the cell cycle 72 h after the 24 h treatment with MF66-DOX (Figure 3a,b). In particular, a cell cycle arrest in the G2/M phase, as well as an increase in polyploidy rate, was observed. A peak in SubG0 phase was also detected, which could be related to the additional apoptotic cells detected with Prussian blue staining. This increase in apoptotic cell number was confirmed by caspase 3 activation assessed by immunofluorescence (Figure 3c). In addition, DNA counterstaining with Hoechst 33,258 displayed nuclear condensation and fragmentation. These are characteristic indications of apoptotic cell death.Since the proliferation of surviving cells after DOX treatment was not observed, we proceeded to perform a test of cellular senescence based on the overexpression of senescence-associated β-galactosidase enzyme in the samples. The results show positive blue staining indicating senescence in almost half of the treated cells with MF66-DOX, with higher occurrence in the largest cells (Figure 3d).All these morphological and biochemical changes were confirmed by time-lapse videomicroscopy. The full-length movies are available as Supplementary Material (Videos S1 to S3). At 48 h after treatment, cells were already larger, flatter and more frequently multinucleated (characteristic morphology of senescent cells) than control cells. In addition, movies of this sample display an apoptotic outbreak when approaching 72 h of post-incubation, and mitosis was longer than in the control or MF66 samples.The flow cytometry results indicate that electrostatic formulation bound to DOX was internalized by MDA-MB-231 cells (Figure 4a), confirming the high internalization observed by Prussian blue staining. It should be taken into account that a higher uptake is reached by “free” DOX (drug non-loaded on MF66). However, it is important to highlight that while “free” DOX has a short retention time, the intracellular DOX level was significantly retained with MF66-DOX formulation at 72 h after that 24 h incubation (Figure 4b).The AlamarBlue® assay performed 72 h after incubation for 24 h with the different formulations demonstrated that cells treated with DOX-loaded formulations show a significant reduction in cell viability in all the cases, although it was higher with MF66-S-S-I-DOX (Figure 5a).Prussian blue staining confirmed similar results to those previously described for the electrostatic formulation with DOX. At 72 h after incubation for 24 h, this staining displayed a high internalization for the three MF66-DOX covalent formulations (Figure 5b). This result was confirmed by fluorescence microscopy. In the cells incubated with any of the three covalent formulations, the red signal from DOX was visualized intracellularly in the nucleus and cytoplasm, mainly attached to MNPs (Figure S1a).A detailed study by differential interference contrast (DIC) microscopy allowed us to detect several morphological alterations in cells treated with MF66 covalently conjugated with DOX (Figure S1b). Prussian blue staining corroborated these results (Figure 5b). At this long post-incubation time, the mitotic index was tripled in MF66-I-DOX-treated cells in respect to control cells and aberrant mitosis was observed after incubation with this nanoformulation. Another relevant finding was the increase in multinucleated and giant cells. In addition, features of apoptosis (cell shrinkage, condensation, and fragmentation of nuclear chromatin) were also observed when the cells were treated with the three different samples, but mainly with MF66-S-S-I-DOX.Taking into account that MF66-S-S-I-DOX, which had immobilized half the DOX compared to electrostatic formulation, led to the highest cytotoxic effect, mainly due to apoptosis, this covalent formulation was selected for activity evaluation on breast cancer stem cells in MDA-MB-231 cell line and primary metastatic patient-derived breast cancer cells.The mammosphere colony assay was performed to quantify MDA-MB-231 cancer stem cell activity. Figure 6a displays significantly decreased mammosphere-forming efficiency 72 h after 24 h incubation with MF66-S-S-I-DOX compared to control cells. Images obtained by dark field and fluorescence microscopy revealed that mammospheres generated from treated cells with DOX formulation had a similar shape but with a lower number of cells, and most of the cells looked larger than control size ones (Figure 6b).To characterize these mammospheres, paraffin-embedded sections were stained with Prussian blue, showing MNPs as blue spots inside cells (Figure 7a). Cells incubated with bare MNPs had similar morphologies to control cells (images not shown). However, mammosphere cells treated with MF66-S-S-I-DOX were larger and demonstrated nuclear condensation and fragmentation.To confirm that several cells in mammospheres generated after treatment with DOX formulation had triggered apoptosis, immunostaining for cytochrome c and cleaved caspase-3 was performed in paraffin-embedded sections. Figure 7b shows cytochrome c as green spots in the control sections. However, cytochrome c appeared as a diffuse green signal in treated sections, indicating its translocation from mitochondria to the cytoplasm. Moreover, apoptosis in treated samples was confirmed by cleaved caspase-3 immunofluorescence, as illustrated in Figure 7b.Finally, we analysed the effect of this formulation in primary samples obtained from the pleural effusion from a patient with triple-negative breast cancer. Trypan blue staining showed a significant reduction in cell viability with MF66-S-S-I-DOX incubation for 24 h (Figure 8a). This was confirmed by direct observations of the cells by an inverted microscope, which, in addition, allowed us to detect fluorescence of the DOX (Figure 8b). In addition, the activity of cancer stem cells to form mammospheres was also significantly reduced (Figure 8c,d).MNPs have great relevance in the field of nano-oncology due to their multiple applications, including their use as mediators for different anti-tumour therapies (e.g., hyperthermia or carriers for chemotherapeutic drugs) [21]. For this purpose, two fundamental types of strategy have been used to design MNPs: non-covalent (electrostatic) and covalent interactions.Cytotoxic analysis showed reduced viability of MF66 electrostatically functionalized with doxorubicin (DOX; MF66-DOX) in relation to bare MF66 in MDA-MB-231 cells 72 h after 24 h treatments. This result indicates the significance of long-term evaluation of nanoparticles’ effectiveness when they are functionalized, to understand the kinetics of drug release. In conventional cell culture studies, nanoparticles toxicity is not usually tested much longer than 24 h. It is important to note that MDA-MB-231 is a highly aggressive, invasive and poorly differentiated triple-negative (ER, PR and HER2 negative) breast cancer cell line and one of the most commonly used cell lines in cancer research laboratories [22].It is also important to note that the cytotoxicity induced by free DOX was clearly higher compared to MF66-DOX (see Figure S1 of Supplementary Material). In relation to this fact, similar results have been described for other magnetic nanoparticles [23,24]. An adequate explanation for the differences observed in the cytotoxic effects of DOX and MF66-DOX could be their different transport mechanisms. Free DOX enters the cells very rapidly by simple passive diffusion and reaches the nucleus easily. In the case of MF66-linked DOX, an endocytic mechanism occurs, and the drug has to be released from the surface of the nanoparticle in the lysosomes. Therefore, DOX takes more time to reach the nucleus and to bind to DNA.A detailed visualization of MF66-DOX samples showed three types of morphological alterations, compared to cells incubated with bare MF66. Mostly, aberrant metaphases, increased size of the cells and multinucleation, as evidence of mitotic catastrophe and senescence, which were confirmed by the significant increase in G2/M and polyploidy phases observed in cell cycle analysis and by positive staining for the senescence-associated β-galactosidase assay. In addition, to a lesser extent, cells with apoptotic features were detected; this process was confirmed by an increase in cells in the subG0 region in the cell cycle and by the immunofluorescence of activated caspase-3, the main effector caspase in apoptosis [25].In summary, MF66-DOX treatment induced aberrant mitosis, apoptosis and polyploid giant senescent cells. In fact, the true roles of these cells have generated a certain amount of controversy. For many decades, senescence has been considered a conventional response to cancer therapy, taking into account that cellular senescence represents a state of cell cycle arrest in which cells remain viable and metabolically active but non-proliferative [26]. However, recent evidence has proposed that giant senescent cells, which remain chronically present following an anticancer therapy, may contribute to cancer regrowth and promote systemic inflammation, increasing the side effects of conventional anticancer agents (reviewed in [27]).Time-lapse videomicroscopy results demonstrate that senescent giant cancer cells can undergo apoptosis. In this sense, the use of pharmacological activators of apoptosis has been proposed as an effective approach to eliminate the potential senescent giant cells generated, and thus reduce the possibilities of cancer recurrence [27,28,29].Our results also indicate that DOX conjugated electrostatically to the MF66, once internalized, was slowly released from MF66 MNPs and was able to reach the nucleus and induce its cytotoxic effects.On the other hand, flow cytometry results show that cells internalized the DOX attached to MF66 MNPs more slowly than free DOX, and this effect might be due to the difference between the endocytosis of the nanocarrier and the passive diffusion of free DOX [30]. However, there was higher intracellular retention of DOX when it was vehiculated in nanoparticles in comparison to free DOX. This result may be explained by the rapid exit of the internalized free DOX from the cells by overexpression of drug efflux pumps, which facilitate the acquisition of chemoresistance mechanisms [31]. This effect is blocked when DOX is loaded in different nanoplatforms [32,33]. So, the administration of electrostatic DOX conjugated to MF66 could act as an intracellular reservoir of the drug that would progressively release DOX molecules to induce their cytotoxic effect, thus bypassing the action of drug-efflux pumps.In summary, MF66 MNPs possess appropriate properties to be considered as a suitable carrier for different agents (such as DOX) via electrostatic interactions. In fact, in a previous study, we analysed the effectiveness of MF66-DOX in combination with hyperthermia after intratumoural injection of the nanoparticles in vivo [34]. In the same study, we detected a greater cytotoxic effect on MDA-MB-231 cells of free DOX relative to MF66-DOX, possibly due to their different internalization mechanisms (passive diffusion vs. endocytosis). However, the mechanisms of action of MF66 with electrostatic formulations in MDA-MB-231 cells were not analysed in depth in the abovementioned research.Once the high potential of the electrostatic formulations was demonstrated, our research focused on the study of MF66 nanoparticles for DOX delivery using three different covalent linkers: disulfide (DOX-S-S-Pyr, called MF66-S-S-DOX), imine (DOX-I-Mal, called MF66-I-DOX) and both disulfide and imine (DOX-I-S-S-Pyr, called MF66-S-S-I-DOX).Our covalent strategy provides several advantages compared with the electrostatic approach, mainly due to the inherent stability of the covalent ones. For instance, these MNPs could improve the plasma circulation of the drug, preventing their non-specific release triggered by the presence of salts and proteins in the plasma, as can occur in the electrostatic approach. Furthermore, the covalent-modified systems can increase the selectivity of the release of the drug towards cancer cells, where the specific triggering stimuli are present, thus avoiding the damage in healthy tissues. In this sense, the use of disulfide moieties is particularly useful, since the release of the therapeutic molecules can be controlled by glutathione (GSH), which is present at micromolar concentrations outside the cells, but at millimolar concentrations inside them [35]. What is more, cancer cells present even higher concentrations of GSH compared to healthy cells, increasing the overall selectivity of the approach toward cancer cells [36]. Ulbrich et al. carried out a detailed review describing the most significant advantages and disadvantages related to targeted drug delivery with polymers and magnetic nanoparticles [37].Our first results show that cell toxicity depended on the type of covalent linker used to bind DOX to the MF66 nanoparticles. Thus, MF66-S-S-I-DOX MNPs with a double-sensitive linker (to pH and redox) exhibited the greatest cytotoxic effect (only 37% viable cells). This result is difficult to compare with those obtained by other authors with superparamagnetic iron oxide nanoparticles linked covalently to DOX. As can be inferred from the aforementioned review by Ulbrich et al., different parameters and experimental protocols had been used, including among others, type of cell line, size, shape, and coating of the nanoparticle and concentration and incubation time [37].On the other hand, similarly to the electrostatic formulation, Prussian blue staining allowed us to detect high retention rates of covalent formulations inside the cells, as well as, different alterations in cellular morphology depending on the type of covalent samples, mainly apoptosis and mitotic catastrophe. According to our results, several studies have reported an apoptotic response after incubation with DOX covalently bonded to different types of iron oxide magnetic nanoparticle [38,39]. However, there are very few studies related to mitotic catastrophe for DOX-loaded nanoparticles [40]. It is important to note that, at least to our knowledge, this is the first time that an iron oxide nanoparticle DOX loading covalently has been described to be capable of inducing cell death at the same time by both the mechanisms of apoptosis and mitotic catastrophe. One possible explanation is that in many studies with nanoparticles, neither the possible changes induced in the cell cycle profile nor inverted microscope observations of the cell response were usually evaluated, although they should be carried out to define the cell death mechanisms triggered by anticancer therapies [41].According to all these results, the best formulation of DOX-loaded MNPs was MF66-S-S-I-DOX because drug delivery induced the highest cell death by apoptosis and the highest decrease in cell viability and, furthermore, DOX was observable in the nucleus after 72 h post-incubation. Therefore, these MF66-S-S-I-DOX MNPs were chosen to continue in subsequent studies.In an effort to explore the in vitro properties of this covalent nanoformulation, we evaluated its effect against cancer stem cells contained in the MDA-MB-231 cell line by a mammosphere forming efficiency (MFE) assay, which is an experimental technique commonly used to analyse the efficacy of different treatments on the activity of cancer stem cells [42,43].Several articles have stressed the potential of functionalized nanomaterials to reduce this population of cancer cells, which is responsible for relapse and metastasis [44], and our formulation presented excellent properties in this regard. In particular, MF66-S-S-I-DOX induced a decrease in the efficiency of cancer stem cells to form mammospheres compared with both the control and cells treated with bare MF66, also triggering an apoptotic response in the mammospheres formed.Importantly, these double-sensitive MNPs were also able to induce a significant toxicity as well as a decrease in the formation of mammospheres in the primary BB3RC79 cells, obtained from a patient with a similar origin (pleural effusion) and genetic profile (triple-negative breast cancer) than MDA-MB-231 cells.Several studies have revealed that the mammosphere-forming cells from primary breast cancer samples exhibit resistance to multiple chemotherapeutic drugs and therefore are a useful resource to test breast cancer stem-like cells targeted therapies [45,46]. In this line, our results show a remarkable effect since the patient from whom the primary cells were obtained had been treated with epirubicin (a similar drug to DOX), inducing an ineffective response in the patient who developed metastasis. By contrast, our formulation was efficient against these cells, even reducing the cancer stem cell activity.Therefore, double-sensitive nanoparticles MF66-S-S-I-DOX have many of the properties that an ideal nanocarrier should have, such as stability, biocompatibility, sufficient loading capacity and the ability to maintain its cytotoxic activity after their internalization by tumour cells, including against cancer stem cells.These results stimulate future studies with MF66-S-S-I-DOX-nanoformulations to determine the expression of breast CSC markers after treatments with these nanoparticles, as well as with in vivo model systems, in order to confirm their cellular functions and their therapeutic potential applications.The MNPs used in this study, called “MF66” (Liquids Research Limited; Bangor, Gwynedd, UK), were produced by means of the co-precipitation technique [47]. Coating with dimercaptosuccinic acid (DMSA) was performed as described previously [48].The DOX derivatives DOX-S-S-Pyr and (5-maleimidovaleroyl) hydrazone (DOX-I-Mal) were synthesized as previously described [49,50,51]. DOX-I-S-S-Pyr was synthesized as previously described with modifications [52] (see Supplementary Material for the details on the synthesis of the different DOX derivates).All details have been included in Supplementary Material.The release of immobilized DOX onto electrostatic formulation was monitored by a spectrophotometer. Water was used to determine the stability of the formulation over time. Then, the same experiments were performed in PBS buffer (pH 7.4), AcONa/AcOH buffer (pH 4.7) and water desorption of DOX from the MNPs in the presence of salts.The release of DOX from covalent functionalized MF66 was carried out under extracellular pH conditions (pH 7.4 and 37 °C) using two concentrations of reducing agent (1 μM and 1 mM of 1,4-dithiothreitol (DTT), mimicking the extracellular and intracellular reducing power, respectively. Intracellular pH conditions (pH 4.7 and 37 °C) were carried out using two concentrations of reducing agent (1 μM and 1 mM of DTT). More details can be found in Supplementary Material.The human breast cancer cell line MDA-MB-231 was obtained from American Type Culture Collection (ATCC® HTB-26TM) and was cultured following the standard protocols for established cell lines [53]. Primary cells BB3RC79 derived from a patient with triple negative breast cancer were cultured with DMEM/F12 medium (DMEM/Ham’s F12; Thermo Fisher Scientific), supplemented with 10% FBS, 2 mM L-glutamine (Thermo Fisher Scientific), 50 U mL−1 penicillin and 50 μg mL−1 streptomycin and 10 μg mL−1 insulin (Merck), 1 μg mL−1 hydrocortisone (Merck) and 50 ng mL−1 epidermal growth factor receptor (EGFR) (Miltenyi Biotec; Bergisch Gladbach, Germany). Metastatic samples from breast cancer patients were collected at The Christie NHS Foundation Trust. Patients were informed and consented according to local National Research Ethics Service guidelines (Ethical Approval Study No.: 05/Q1402/25).The different MNP stocks at 2.4 mg Fe mL−1 were dispersed by sonication for 5 min in a 50 kHz sonicator bath (Bath Ultrasonic QS3, Scientific Laboratory Supplies; Cardiff, UK). MNPs were then resuspended in complete cell culture media at a final concentration of 0.1 mg Fe mL−1 (and drugs immobilized were at 4 µM for DOX in the electrostatic formulations and 2 µM for DOX in the covalent formulations). The mixture was sonicated for 1 min and incubated with cells for 24 h. The studies were evaluated at different post-incubation times (mainly, from 0 to 72 h).Cell viability with the different formulations was assessed by AlamarBlue® assay in MDA-MB-231 cells and trypan blue in BB3RC79 cells according to manufacturer instructions (for more details see Supplementary Material).These parameters were analysed in living cells to avoid artefacts of fixation and in cells fixed and stained with Prussian blue specific to visualize iron nanoparticles [54].MDA-MB-231 cells seeded on 25 cm2 flasks were incubated with the electrostatic MF66-DOX formulation or ‘free DOX’ (non-covalently bounded DOX) for 24 h with different post-incubation times: 0, 24 and 72 h. Then, the supernatant was removed (harvesting detached cells) and cells attached to well, were trypsinized (Thermo Fisher Scientific). These cells were added to the previously collected ones and centrifuged for 4 min at 300 g (JP Select; Abrera, Spain). The supernatant was discarded, and the pellet was resuspended in cold medium without phenol red (Thermo Fisher Scientific). Samples were analysed in a flow cytometer Cytomics FC500 (Beckman-Coulter; Brea, CA, USA) with a laser of 488 nm and a filter 620 BP.MDA-MB-231 cells, grown directly onto wells, were incubated with the different samples for 24 h, washed, collected, and analysed after 72 h. To this purpose, we added 50 μL of RNAse 4 kU mL−1 and 1 mL of propidium iodide 50 µg mL−1 of kit DNA-Prep Reagents (6607055, Beckman-Coulter). Samples were stirred and incubated for 30 min at 37 °C in darkness. They were analysed in a flow cytometer Cytomics FC500 with a laser of 488 nm and a filter 620 BP.The common protocol for indirect immunofluorescence was performed (see more details in Supplementary Material) [55].MDA-MB-231 cells were seeded on coverslips, incubated with the different formulations for 24 h and analysed 96 h after with senescence β-galactosidase staining kit (CS0030, Merck) following the instructions of the manufacturer.Cells seeded in chambered coverslips (Ibidi; Martinsried, Germany) were treated as previously explained. Frames were acquired by phase contrast microscopy every 15 min between 48 and 72 h after incubation, maintaining CO2, temperature and humidity conditions in cell culture range.Breast stem cell activity was quantified by standard sphere-forming assay (see Supplementary Material) [42]. Mammospheres bigger than 50 μm diameters were counted after 5 or 7 days for MDA-MB-231 or BB3RC79 cells, respectively.Mammospheres fixed in formalin were embedded in paraffin. Sections of 5 µm were deparaffinised and analysed by the different protocols (see Supplementary Material).Results are the mean values and standard deviation (SD) from at least five different experiments in triplicate. Statistical analysis was performed by GraphPad Prism 5 software (GraphPad Inc.; La Jolla, CA, USA) using one-way ANOVA and Tukey’s post-test. The threshold for significance was p = 0.05 and statistically significant differences were labelled as ‘*’ when p < 0.05, ‘**’ when p < 0.01 and ‘***’ when p < 0.001.In summary, this study brings essential insights into the relevance of the selection of appropriate functionalization strategies, which have significant implications on the final performance of a nanoformulation. Among others, the drug release mechanism and kinetics can be achieved, leading to different cytotoxic efficacy and cell death mechanisms. The best performing functionalized nanoparticle in this study (MF66-S-S-I-DOX) is a promising tool, which can be used to improve the efficiency of existing chemotherapeutic approaches with iron oxide nanoparticles, reducing the side effects of the chemotherapeutic drug and increasing efficiency against cancer stem cells.The following are available online at https://www.mdpi.com/2072-6694/12/6/1397/s1. Supplementary Materials: 1.1. Electrostatic functionalization of MNPs, 1.2. Covalent functionalization of MNPs, 1.3. DOX release studies, 1.4. AlamarBlue® assay, 1.5. Trypan blue assay, 1.6. Indirect immunofluorescence for cleaved caspase-3 and cytochrome c, 1.7. Mammosphere forming efficiency, 1.8. Morphology of mammospheres, 1.9. Statistical analysis, Supplementary Results: 2.1. Morphological effect of electrostatic formulations over time, Supplementary Movie S1: Videomicroscopic analysis of control MDA-MB-231 cells, Supplementary Movie S2: Videomicroscopy study of MDA-MB-231 cells incubated with MF66, Supplementary Movie S3: Videomicroscopy study of MDA-MB-231 cells incubated with MF66-DOX, 2.2. Internalization and morphological alterations of covalent formulations in living cells, Table S1: Characterization of the DOX functionalized MF66-MNP, Figure S1: Surviving fraction of MDA-MB-231 cells incubated 24 h with free unmodified DOX, Figure S2: Living cells visualized 72 h after incubation for 24 h with the different formulations linked covalently to DOX.A.L.C. (Ana Lazaro-Carrillo) performed all studies of electrostatic nanoparticles in cell cultures, analysed the data and partly wrote the manuscript; M.C. performed all studies of covalent nanoparticles in cell cultures and analysed the data; A.A. performed the synthesis and characterization of nanoparticle formulation and release kinetics of the different formulations; A.L.C. (Aitziber L. Cortajarena) designed the synthesis and characterization of nanoparticle formulation, participated in discussion of results, partly wrote the manuscript and contributed to the acquisition of funding; B.M.S. designed and supervised the experiments of mammosphere formation assay, participated in discussion of results and partly wrote the manuscript; A.L. performed the synthesis and characterization of nanoparticle formulation and release kinetics of the different formulations; Á.S. designed the synthesis and characterization of nanostructures, linkers and modified drugs, participated in discussion of results, partly wrote the manuscript and contributed to the acquisition of funding; R.B.C. participated in discussion of mammosphere-forming assays and contributed to the acquisition of funding; R.M. contributed to the acquisition of funding; A.V. designed the paper, generated figures, supervised the project, partly wrote the manuscript, reviewed the manuscript and contributed to the acquisition of funding. All authors have read and agreed to the published version of the manuscript.This research was funded by the European Seventh Framework Program (grant agreement number 262943); the European Union’s Horizon 2020 research and innovation programme (grant agreement number 685795); Ministerio de Economía y Competitividad, Spain (grants CTQ2016-78454-C2-2-R, BIO2016-77367-C2-1-R and SAF2017-87305-R); Basque Government Elkartek KK- 2017/00008; Comunidad de Madrid (IND2017/IND-7809; S2017/BMD-3867 RENIM-CM and S2018/NMT-4321 NANOMAGCOST-CM); NIHR Manchester Biomedical Research Centre (IS-BRC-1215-20007) and Breast Cancer Now (MAN-Q2); co-financed by European Structural and Investment Fund, Asociación Española Contra el Cáncer (Singulares 2014) and IMDEA Nanociencia. CIC biomaGUNE acknowledges Maria de Maeztu Units of Excellence Program from the Spanish State Research Agency (Grant MDM-2017-0720). IMDEA Nanociencia acknowledges support from the ‘Severo Ochoa’ Programme for Centres of Excellence in R&D (MINECO, Grant SEV-2016-0686).We recognize the valuable contribution of Sylvia Gutiérrez and Ana Oña (Confocal Microscopy, Centro Nacional de Biotecnología, Madrid) and Carmen Moreno-Ortiz and Sara Escudero (Flow Cytometry, Centro Nacional de Biotecnología, Madrid).The authors declare no conflict of interest.Release kinetics of the different formulations. (a) Release kinetics of the electrostatically immobilized doxorubicin (DOX) (circles) were studied by dispersing MF66-DOX in water (blue), PBS buffer (pH 7.4) (black), or sodium acetate buffer (pH 4.7) (red). (b) Release kinetics of the covalently immobilized DOX (squares) were studied by dispersing MF66-S-S-DOX (b1), MF66-I-DOX (b2), or MF66-S-S-I-DOX (b3), in PBS buffer (pH 7.4) containing 1 μM (black) or 1 mM of 1,4-dithiothreitol (DTT) (purple), or sodium acetate buffer (pH 4.7) containing 1 μM (red) or 1 mM of DTT (green).Evaluation of electrostatic nanoformulations in MDA-MB-231 cells 72 h after 24 h of incubation with the different treatments. (a) Cellular viability assessed after incubation with the different samples by AlamarBlue® assay. (b) Efficient uptake and morphological alterations analysed by Prussian blue staining. Percentages included are the aberrant mitosis over the total number of mitosis and the number of giant multinucleated cells or apoptosis over the total number of cells. Scale bar: 10 µm.Cellular inactivation mechanisms analysed after incubation of MDA-MB-231 cells with electrostatic formulation. (a) Representative cell cycle histograms 72 h after 24 h of incubation with the different formulations. (b) Bar chart displaying percentages of cells in the different phases of cell cycle 72 h after 24 h treatments. (c) Apoptotic cells visualized by indirect immunofluorescence for cleaved caspase 3 (red) and DNA staining (blue) analysed 72 h after 24 h treatment. Scale bar: 10 µm. Percentages included are the apoptotic cells over the total number of cells. (d) Senescent cells analysed by senescence-associated β-galactosidase activity (blue) 96 h after treatment incubated for 24 h visualized in differential interference contrast (DIC) microscopy. Scale bar: 50 µm. Percentages included are the senescent cells over the total number of cells.DOX efflux by MDA-MB-231 cells incubated with the drug dissolved in medium (“free” DOX) or linked to MF66. (a) Intracellular DOX level measured by flow cytometry after 24 h incubation. (b) DOX retention at 72 h after that 24 h incubation.Effect of the different covalent formulations 72 h after the incubation for 24 h in MDA-MB-231 cells. (a) Cellular viability assessed by AlamarBlue®. (b) Efficient uptake and morphological effects analysed by Prussian blue staining. Scale bar: 10 µm. Percentages included are the mitosis, number of giant multinucleated cells or apoptosis over the total number of cells.Mammospheres from MDA-MB-231 cells untreated (control) or treated with MF66 and MF66-S-S-I-DOX. (a) Quantification of mammosphere forming efficiency analysed 5 days after cell 24 h incubation with magnetic nanoparticles (MNPs). (b) Images of mammospheres in phase contrast and fluorescence microscopy 72 h after mammosphere formation assay. Scale bar: 50 μm.Morphological characterization and cell death mechanism triggered in mammospheres from MDA-MB-231 cells. (a) Prussian blue staining in control cells or treated with MF66 or MF66-S-S-I-DOX MNPs and visualized by bright field microscopy. Scale bar: 5 μm. (b) Immunofluorescence for cytochrome c (green) and DNA staining (blue) visualized by fluorescence microscopy. Scale bar: 5 μm. (c) Immunofluorescence for cleaved caspase 3 (red) and DNA staining (blue) visualized by fluorescence microscopy. Scale bar: 5 μm.Analysis of MF66-S-S-I-DOX in primary metastatic patient-derived breast cancer cells. (a) Cellular viability assessed by trypan blue performed 24 h after the 24 h incubation. (b) Images of (i) control cells or incubated with (ii) bare MF66 or (iii, iii’) MF66-S-S-I-DOX in phase contrast and fluorescence microscopy, respectively, at the same post-incubation time. Scale bar: 50 μm. (c) Quantification of mammosphere forming efficiency analysed 7 days after cell incubation for 24 h with MNPs. (d) Images of mammospheres from (i) control cells or incubated with (ii) bare MF66 or (iii, iii’) MF66-S-S-I-DOX in phase contrast and fluorescence microscopy, respectively at the same post-incubation time. Scale bar: 25 μm.

Med-MDPI/cancers/cancers-12-06-01398.txt

@@ -0,0 +1 @@
+Non-epithelial ovarian tumors are heterogeneous and account for approximately 10% of ovarian malignancies. The most common subtypes of non-epithelial ovarian tumors arise from germ cells or sex cord and stromal cells of the gonads. These tumors are usually detected at an early stage, and management includes surgical staging and debulking. When indicated for advanced disease, most respond to chemotherapy; however, options for patients with refractory disease are limited, and regimens can be associated with significant toxicities, including permanent organ dysfunction, secondary malignancies, and death. Targeted therapies that potentially decrease chemotherapy-related adverse effects and improve outcomes for patients with chemotherapy-refractory disease are needed. Here, we review the molecular landscape of non-epithelial ovarian tumors for the purpose of informing rational clinical trial design. Recent genomic discoveries have uncovered recurring somatic alterations and germline mutations in subtypes of non-epithelial ovarian tumors. Though there is a paucity of efficacy data on targeted therapies, such as kinase inhibitors, antibody–drug conjugates, immunotherapy, and hormonal therapy, exceptional responses to some compounds have been reported. The rarity and complexity of non-epithelial ovarian tumors warrant collaboration and efficient clinical trial design, including high-quality molecular characterization, to guide future efforts.Non-epithelial ovarian tumors are an uncommon group of malignancies that arise from germ cells, sex cord cells, and/or stromal cells of the ovary. The term non-epithelial is used to distinguish these tumors from their epithelial counterparts, which usually arise from the external lining of the ovaries or the fallopian tube epithelium (Figure 1). This histological distinction is based on the World Health Organization’s classification of ovarian tumors [1,2] and has important genomic, epigenetic, and clinical implications [3,4].Non-epithelial ovarian tumors account for approximately 8–10% of ovarian malignancy cases [3,5,6], representing approximately 2200 new cases per year in the United States [3]. This group of tumors is heterogeneous and comprised of malignant ovarian germ cell tumors (MOGCT), malignant sex cord–stromal tumors (SCST), and other tumors. These categories are further subdivided into a multitude of histologically and clinically diverse groups. Compared with epithelial malignancies, these tumors disproportionally affect younger patients, with some types most frequently occurring in the pediatric population. The rarity and heterogeneity of non-epithelial ovarian tumors result in a paucity of high-quality data to guide clinical care of patients with these tumors.This review will address potential molecular therapeutic approaches to non-epithelial ovarian tumors, focusing on germ cell tumors and sex cord–stromal tumors of the ovary. General considerations for the management of these tumors, including epidemiology, surgical staging, and treatment, and common chemotherapy regimens will be briefly covered.Non-epithelial ovarian tumors often occur at a younger age than epithelial ovarian cancer, with some tumors occurring predominantly in children, adolescents, or young adults [3,5]. Diagnostic and therapeutic considerations for these patients include potential preservation of fertility and preventing long-term toxicity of chemotherapy, including organ dysfunction and secondary cancers. Fortunately, as a group, non-epithelial ovarian tumors are usually detected at an early stage and are associated with a favorable prognosis [7,8].Like epithelial ovarian cancer, staging of non-epithelial ovarian tumors is performed surgically and is crucial for informing prognosis, subsequent surveillance, and therapy [9,10,11]. A comprehensive review of the outcomes associated with specific staging procedures is outside the scope of this review [7,8,12]. Complete staging is recommended by most gynecologic oncology societies [9,13], but it is unclear how this practice affects survival [14]. For example, the typical surgical staging of children by pediatric surgeons is less comprehensive than that of gynecologic oncologists for adults [14]. Generally, complete staging may contribute to increased perioperative morbidity [15], while incomplete staging is associated with a higher risk of tumor recurrence [11,16,17]. For patients who have given birth or are postmenopausal, surgical staging includes bilateral salpingo-oophorectomy and total abdominal hysterectomy. However, fertility-sparing surgery with unilateral salpingo-oophorectomy is often pursued for younger patients, preserving the uterus and the contralateral ovary [15,18,19,20,21,22]. For patients with bilateral tumors, cystectomy of one ovary can be considered to preserve fertility [13].Peritoneal fluid is typically sampled. Cytoreduction of the visible tumor, random biopsies of the peritoneal surfaces and other organs, and omentectomy are also recommended, although not infrequently omitted [14,17]. Routine lymphadenectomy is controversial and depends on the tumor subtype and pre- and intra-operative findings [15,23,24]. Patients with adult granulosa cell tumors in particular have low rates of lymph node metastases and do not seem to benefit from lymphadenectomy [25,26,27,28]. In addition to obtaining diagnostic and prognostic information, surgery is a major component of therapy. Early stage tumors can be cured with the surgical procedure alone and may not require adjuvant systemic treatment [7,8,29].Patients with advanced disease typically require treatment with platinum-based chemotherapy regimens [13]. However, these regimens are associated with significant toxicity, including long-term organ dysfunction, secondary malignancy, and death.In the setting of platinum resistance, patients are encouraged to enroll in clinical trials due to limited and understudied therapeutic options [30,31]. Several second-line chemotherapeutic regimens have been proposed, including high dose chemotherapy with an autologous stem cell transplant [32,33]. But while effective for some patients, high-dose chemotherapy is highly toxic, with a 6% death rate in one of the aforementioned trials [32].Hence, incorporating effective targeted treatments for non-epithelial ovarian tumors could be beneficial for reducing long-term toxicity from chemotherapy as well as for addressing the unmet needs in the recurrent or refractory setting. Recent characterization of molecular and genetic aberrations of non-epithelial ovarian tumors has uncovered new potential therapeutic targets that differ by the tumor cell of origin and subtype.The most common group of non-epithelial ovarian tumors is malignant ovarian germ cell tumors (MOGCTs). In children, they represent the majority (75%) of malignant ovarian tumors [34]. MOGCTs are infrequent in older patients and occur predominantly in adolescents and young adults [34]. The most common types of MOGCTs (Table 1) are dysgerminoma and immature teratoma which comprise 65–70% of MOGCTs, followed by yolk sac tumors and mixed germ cell tumors [35,36]. These tumors reflect the pluripotent potential of primordial germ cells to differentiate into all somatic (endoderm, mesoderm, ectoderm) and extra-embryonic tissues [37]. MOGCTs can lead to elevations of tumor markers in the peripheral blood, including alpha-fetoprotein (AFP), beta-human chorionic gonadotropin (hCG) and lactate dehydrogenase (LDH) [7,38,39].MOGCTs represent approximately 3% of all ovarian tumors in the United States, with 4 cases per 1,000,000 women [3]. The incidence of MOGCTs is estimated to be slightly higher among Asian, Hispanic, and non-Hispanic black women than among non-Hispanic white and American Indian/Native Alaskan women [3]. The proportion of ovarian malignancies attributed to MOGCTs also varies across and within different geographic regions worldwide, with the highest proportion reported in East Asia and Central America [5,34]. The concordance of international variation with ethnicity in the United States raises the possibility of genetic susceptibility to MOGCTs, but further studies are needed to evaluate this hypothesis.MOGCTs are often diagnosed at an early stage. In the United States, approximately 69% of patients with available data are diagnosed at Stage I [3]. Five-year cause-specific survival for these patients is 99% across all races. Approximately 26.5% are diagnosed at Stage II–III, both with five-year cause-specific survival rates of over 90% with modern therapy. Less than 5% of patients with available stage data are diagnosed at Stage IV, with a 69% five-year survival [3].For early stage dysgerminoma and immature teratoma, surgery without adjuvant chemotherapy is currently recommended [13]. The reason is threefold: (i) outcomes with surgery alone are often curative; (ii) in the setting of recurrent or residual disease, responses to chemotherapy are excellent; and (iii) this approach best preserves ovarian reserve and fertility [43,44].For those who require adjuvant therapy, the most common first-line chemotherapy regimen consists of bleomycin, etoposide, and cisplatin (BEP). This regimen is curative for most patients with limited or no residual disease after surgery. For those with bulky residual disease, 50–60% achieve cure with adjuvant chemotherapy [45]. While very effective, this regimen may result in significant short- and long-term toxicities [46,47,48]. Furthermore, a substantial proportion of patients remain disease-free without adjuvant chemotherapy [46], and therapies for disease recurrence are often curative. These facts call into question the benefit of adjuvant treatment. Adverse effects of BEP occur in up to 30% of patients, including cisplatin-induced peripheral neuropathy, ototoxicity, and nephrotoxicity, potentially fatal secondary malignancies, bleomycin-associated lung injury, bone marrow suppression, and cardiomyopathy [46,49]. Most patients who receive BEP, however, retain ovarian function and can achieve pregnancy [50,51,52]. Given the toxicity of BEP, less toxic regimens have been proposed, including carboplatin and etoposide [50,53], but these remain investigational. Despite encouraging retrospective data, results from testicular germ cell tumors raise concerns that carboplatin is less effective than cisplatin for certain non-epithelial ovarian tumors [54].As is the case of all ovarian cancers, chemotherapy for recurrent diseases tends to be more effective in those with platinum-sensitive disease, with remission rates on the order of >60% with salvage platinum-based chemotherapy, but only <30% experiencing long-term survival with platinum-refractory disease. Active agents include platinum agents, vinblastine, ifosfamide, taxanes, and gemcitabine [45]. There is controversy regarding the need for adjuvant chemotherapy for early-stage, non-dysgerminoma MOGCTs, including stage IA grade 2–3 immature teratoma, stage IA or IB yolk sac tumors, and other less common histologies [14,55]. Observation alone is often proposed for these patients by pediatric oncology and European gynecologic oncology societies, whereas adjuvant chemotherapy has typically been offered by gynecologic oncologists in the United States. Protocol AGCT 1531 (NCT03067181) evaluates the risks and benefits of these approaches.The genomics of MOGCTs are understudied. Available data suggest that MOGCTs have a low mutational burden with marked aneuploidy [42]. This pattern is hypothesized to arise from abnormal segregation of chromosomes during meiosis and/or mitosis [42,56]. A whole exome sequencing study of 24 MOGCTs found a median of 2.5 (range 0–8) non-synonymous mutations per tumor; an average of 35% of the genome was affected by copy number alterations in 87 patients. The most common copy number alteration was gain of chromosome 12p, containing the oncogene KRAS [42]. This alteration was found in 82% of dysgerminomas, 58% of yolk sac tumors, and 43% of mixed germ cell tumors, but not in immature teratomas. In contrast to epithelial ovarian cancer [57], TP53 mutations were not detected in MOGCTs; the most common mutations were in the genes KIT and KRAS [42], akin to testicular germ cell tumors [58]. Similar differences in mutations between epithelial and non-epithelial ovarian tumors can be seen in the Genomics Evidence Neoplasia Information Exchange of the American Association for Cancer Research (GENIE/AACR, version 7.0) database (Figure 2) [59,60]. Importantly, any genomic analysis grouping all MOGCTs and/or SCSTs is limited by the vast heterogeneity of tumors within each group.The immune response to MOGCTs is understudied. Early evidence suggested that similar to testicular germ cell tumors, MOGCTs (primarily, dysgerminomas) are characterized by immune infiltrates that may have a prognostic value [61,62,63,64,65]. These infiltrates are comprised of several cell types, including T and B lymphocytes, that can organize as tertiary lymphoid structures with germinal center-like structures [66]. Tertiary lymphoid structures are found within MOGCTs, whereas they are commonly peritumoral in other cancers [67,68,69]. Limited data from testicular tumors suggest that tumor progression is accompanied by a decrease in T cells and natural killer cells and an increase in regulatory T cells and macrophages [70]. Programmed death ligand-1 (PD-L1) expression is also common in male germ cell tumors [71].Dysgerminoma is the most common MOGCT, accounting for 30–35% of cases. Dysgerminoma histologically resembles testicular seminoma, with correspondingly similar immunohistochemistry and chemosensitivity [30,72]. Though bilateral disease is present in approximately 10–15% of cases, fertility-sparing surgery can still be considered given its high chemosensitivity [13]. Adjuvant chemotherapy is not typically administered for stage IA disease and is controversial for patients with bilateral disease, ovarian capsule rupture, and positive peritoneal/ascitic cytology (stages IB–IC) [13]. Adjuvant BEP is typically recommended for patients with stage II–IV disease [13]. Carboplatin and etoposide is another potential regimen that can be used in these patients as discussed above [53].A minority of patients with dysgerminoma have gonadal dysgenesis. However, the karyotypic abnormalities involved in gonadal dysgenesis, including the presence of Y chromosome material, are a significant risk factor for the development of dysgerminoma and concurrent gonadoblastoma (a rare neoplasm containing both germ cell and sex cord–stromal cells) [50,73,74,75]. Therefore, bilateral oophorectomy is typically recommended for patients with Y chromosome material.KIT mutations and amplifications have been described in 30–50% of dysgerminomas [42,59,76]. KIT is a tyrosine kinase receptor that can lead to tissue-specific activation of several intracellular signaling pathways, including the RAS-RAF-MEK-ERK/JNK, PI3K-AKT-mTOR, and JAK/STAT pathways [77] (Figure 3A). KIT mutations are common in gastrointestinal stromal tumors (GIST), where they are also predictive of response to the tyrosine kinase inhibitor, imatinib [78].Chromosomal gains of the 12p arm containing KRAS, a gene commonly found in testicular germ cell tumors [37], occur in up to 80% of patients with dysgerminoma [42,79]. KRAS is an oncogene implicated in the pathogenesis of multiple cancers and is a major driver of the RAS-RAF-MEK-ERK/JNK pathway. While the direct targeting of KRAS is challenging [80], encouraging results have been reported with the MEK1/2 inhibitor trametinib in low-grade serous ovarian cancer, an uncommon epithelial ovarian cancer subtype [81].Overexpression of several genes has been described in dysgerminomas, including CASP8, CDH3, CXCL10, and IL6R and the pluripotency-related genes NANOG, POU5F1, POU5F1B, PLBD1, and PDPN [37]. Significant limitations of these analyses include the grouping of dysgerminoma with testicular seminoma and the lack of normal tissues to include as controls.Yolk sac tumors account for approximately 15% of MOGCTs [36]. Yolk sac histology is typically associated with elevated AFP and is considered an adverse prognostic factor [7,13,36,82]. Chemotherapy has traditionally been offered even for Stage IA disease, although the European guidelines suggest it is optional to forego adjuvant chemotherapy in this setting [13,83]. More advanced stages warrant adjuvant chemotherapy. BEP is the recommended regimen [13,83].Yolk sac tumors are aneuploid with characteristic copy number alterations, including chromosome 12p gain in approximately 60% of the tumors, but are not associated with specific recurrent mutations [37,42]. Alterations in the PI3K/AKT/mTOR signaling pathway (Figure 3B), which occur frequently in certain subtypes of epithelial ovarian cancers [81,84,85], seem to be enriched in tumors with a yolk sac component (72%). PIK3CA and AKT1 were amplified in 42% and 37% of tumors with a Yolk sac component, respectively [42]. Targeting this pathway in epithelial ovarian cancer has yet to significantly improve outcomes [81,84]; similar challenges may arise with yolk sac tumors. Gene expression studies suggest that TGF-β/BMP and Wnt/β-catenin signaling pathways are activated in yolk sac tumors, but not in dysgerminomas [41]. TGF-β/BMP is involved in embryonic development and possibly MOGCT development [41]. The Wnt/β-catenin is activated in many cancer types, and inhibitors of this pathway are in clinical trials [86].Immature teratoma, sometimes referred to as malignant teratoma, has comparable or higher incidence than dysgerminoma (30–35% of MOGCTs) and may also present with bilateral disease [36,87]. It is thought to be chemo-resistant and surgical management is often needed for recurrent or advanced disease [46,55]. However, adjuvant BEP is recommended for advanced stage disease and is controversial for Stage IA disease with histological grade 2–3 [46,55]. Long-term overall survival is slightly worse than in patients with dysgerminoma (84% vs. 89.1% overall survival with a median follow-up of 126 months) [36].Unlike other MOGCTs, immature teratoma is typically diploid [37,42], and chromosome 12p gain and KIT/KRAS mutations are uncommon [42]. Whole exome sequencing from 10 patients, most with advanced disease, uncovered extensive loss of heterozygosity without recurring somatic mutations. Variants without known functional significance were detected in TP53, NF1, CTNBB1, and NOTCH2 (once each). The authors show that copy neutral loss of heterozygosity results from meiotic errors at different stages that can be identified based on the copy number variation profile of individual tumors (Figure 3C) [56]. This study also suggested that while bilateral disease arises from different clonal events, spread to the peritoneum is driven by a single clone, even in bilateral disease [56]. Seventeen patients with immature teratoma are included in the GENIE/AACR database. While recurrent mutations were not common, variants of unknown significance in POLE, BRCA2 and ATM (one each) were detected.Mixed MOGCTs contain more than one histological form and account for approximately 5% of the MOGCTs. Embryonal carcinoma, choriocarcinoma, and polyembryoma cell types account for 5–10% of MOGCTs and have the worst prognosis. They rarely exist in pure form [36]. Adjuvant BEP is typically recommended for these histologies at all stages [13,83]. Their genomic landscape is thought to be determined by individual tumor components [37,42]. Embryonal carcinoma expresses CD30 in approximately 80% of cases, although expression may decrease after treatment with chemotherapy [88,89,90].Malignant sex cord–stromal tumors (SCSTs) arise from the primitive sex cord and/or stromal cells of the gonads (Figure 1), including granulosa, theca, Sertoli, or Leydig cells, as well as fibroblasts. SCSTs are rare; middle-aged women are the most commonly affected [5,6]. In the United States, they represent approximately 2% of ovarian malignancies, with 3 cases per 1,000,000 women. SCSTs appear to be most common among non-Hispanic black women and least common among Asian/Pacific Islander women in the United States [3], though the existing epidemiologic data are substantially limited [6]. Though commonly diagnosed at an early stage, five-year cause-specific survival is slightly lower than for MOGCTs, at 88% across all the stages uniformly. In the United States, Stage I, II, III, and IV disease at diagnosis was found in 69%, 12%, 14%, and 5% of the cases with available data, respectively [3]. The corresponding five-year cause-specific survival rates were 98%, 84%, 61%, and 41%.When indicated, common chemotherapy regimens for SCSTs include BEP, cisplatin and etoposide (EP), and carboplatin with paclitaxel [13]. The latter is increasingly used in clinical practice. A phase II randomized trial is underway comparing carboplatin and paclitaxel to BEP for sex cord–stromal cell tumors (NCT01042522). Response to taxanes in incompletely resected recurrent disease has been measured at 42% [91].The most common subtypes of SCSTs (Table 1) are granulosa cell tumors, accounting for over 70% of SCSTs in most series [87,92,93]. Sertoli–Leydig tumors are the next most frequent group. SCSTs often secrete hormones, including inhibin, estradiol, testosterone, and anti-Müllerian hormone, which can be measured and followed as tumor markers. These can lead to such symptoms as virilization, menstrual changes, post-menopausal bleeding, and precocious puberty [6]. SCSTs, unlike MOGCTs, are not characterized by widespread genomic instability with copy number variations [94], although recurrent chromosomal abnormalities have also been described [40].The adult and juvenile granulosa cell histological subtypes comprise the majority (>70%) of SCSTs in the adult and children/adolescent age groups, respectively [6,87,92,93]. Overall, adult granulosa cell tumors are much more common [95]. Granulosa cell tumors often secrete estradiol, which induces proliferation of the endometrium. Endometrial hyperplasia and endometrial cancer, which is associated with granulosa cell tumors, can manifest as abnormal uterine bleeding [96,97]. Surgery is the mainstay of treatment for early stage disease, commonly followed by platinum-based adjuvant chemotherapy for metastatic disease. The two most common regimens are BEP and carboplatin with paclitaxel [6,13].Genomic studies of SCSTs demonstrate that a single somatic mutation in FOXL2 (C134W) is almost ubiquitous in adult granulosa cell tumors, occurring in up to 97% of cases [94,98]. Since some adult granulosa cell tumors can be difficult to definitively diagnose based on histology and immunohistochemistry alone, FOXL2 has been suggested for the molecular diagnosis of these tumors [95,99,100]. The presence of the FOXL2 mutation in tumors with equivocal histological diagnosis may aid in the classification of the tumor as an adult granulosa cell one [100]. Conversely, the tumors classified as the adult granulosa cell ones, but lacking the characteristic FOXL2 mutation, may represent a histological misclassification [101]. FOXL2 is a transcription factor that is involved in regulation of hormone production, cell cycle, and apoptosis [102,103]. The precise mechanism by which this mutation promotes tumor formation is unclear; FOXL2 possibly serves as a tumor suppressor [40,104], but others have postulated that it acts as an oncogene [105]. The somatic mutation may lead to dysregulation of multiple cellular processes. FOXL2 normally downregulates cytochrome P450 (CYP) 17, and the altered product may lead to an increase in CYP17, with resulting increased estrogen production (Figure 4A) [106]. In addition, the mutated FOXL2 increases expression of CYP19/aromatase [107]. The FOXL2 mutation was also detected in 50% of granulosa theca cell tumors, but it is uncommon in juvenile granulosa cell tumors. FOXL2 is rarely mutated in other cancers, with mutations occurring in approximately 1% of all cancers profiled by the GENIE/AACR project and less than 5% of any individual tumor type apart from SCSTs [59,60]. Fewer than 10% of these FOXL2 mutations (33/408) are the recurrent C134W mutation, and of the 33 C134W mutations, 31 (94%) were found in SCSTs [59,60]. The functional significance of FOXL2 mutations in other cancers is outside the scope of this review.TERT mutations are also common in adult granulosa cell tumors [108,109]. A specific mutation in the TERT promoter (TERT c.-124C>T), found in up to 40% of cases, was associated with more aggressive disease and worse overall survival [108,109].Whole genome sequencing of ten granulosa cell tumors revealed no mutations in BRCA1/2 and only a few mutations (10%) in the following genes: TP53, PIK3CA, CTNNB1, and PIK3R1 [110]. These are all rare (<5%) in 86 predominantly adult granulosa cell tumors with the data available on the GENIE/AACR. Recurrent alterations in KMT2D can be found in over 20% of patients in this database [59,60]. Other recurrent alterations are found in a limited number of patients and require confirmation in larger cohorts.Most granulosa cell tumors are diploid, but recurrent chromosomal copy number alterations, including trisomy 12, 14 and monosomy 22 have been described [108,111,112]. In a small series of ten patients that requires confirmation in larger cohorts, AKT1 was the most commonly amplified gene [112], potentially leading to aberrations in the PI3K/AKT/mTOR pathway. In juvenile granulosa cell tumors, approximately 30% harbored a mutation in GNAS [113] in a cohort of thirty patients. In a small study of 16 patients, over 60% harbored a duplication of AKT1 [114] (Figure 4B). GNAS encodes a subunit of G protein-coupled receptors that are bound by follicle-stimulating hormone (FSH) on the surface of granulosa cells and stimulate adenylyl cyclase activity, increasing production of cyclic AMP. Protein kinase A is thought to be the initial protein kinase activated by cyclic AMP and one of the most important mediators of cyclic AMP signal transduction [115]. Granulosa cell tumors also frequently express vascular endothelial growth factor (VEGF) [116,117,118] and platelet-derived growth factor (PDGFR) [119]. The rare hereditary syndromes, Ollier disease and Maffucci syndrome, are associated with increased risk of juvenile granulosa cell tumors [6].Sertoli and Leydig cells are found in the normal testis. Sertoli–Leydig cell tumors are typically detected at an early stage and are often accompanied by androgen production; AFP elevations have been described [95]. If adjuvant chemotherapy is indicated, BEP or carboplatin and paclitaxel are typically recommended [13].DICER1 mutations have been described in approximately 60% of Sertoli–Leydig tumors [120] (Figure 4C). The prevalence of DICER1 mutations in Sertoli–Leydig cell tumors may be even higher when accounting for potential histological misclassification [121]. In one series, they were not found in well-differentiated Sertoli–Leydig tumors, but were found in all moderately poorly differentiated tumors [121]. DICER1 is a member of the ribonuclease III (RNAse III) family involved in transcriptional regulation via miRNA (microRNA) modulation. Some cases with somatic DICER1 mutations were also found to harbor a germline mutation in DICER1, predisposing to additional tumors including pleuropulmonary blastoma [122]. Thus, germline testing for DICER1 should be offered to patients with these tumors. DICER1 is rarely mutated in cancers that are not associated with germline DICER1 mutations [59]. FOXL2 mutations have been described in 10–20% of Sertoli–Leydig tumors [99,123]. One study found that DICER1 and FOXL2 mutations are mutually exclusive in Sertoli–Leydig tumors and that each mutation was associated with distinct clinicopathological features [99]. These data suggest that the molecular classification of Sertoli–Leydig cell tumors may be clinically relevant, but prospective trials are needed to evaluate this hypothesis.Peutz–Jeghers syndrome caused by mutations in the serine-threonine kinase 11 gene (STK11) can predispose to a specific subtype of SCSTs. These SCSTs are classified as SCSTs with annular tubules containing tubules of Sertoli cells arranged around one or more hyaline bodies [6]. When arising in Peutz-Jeghers syndrome, they are thought to be benign, but outside of this syndrome, a malignant clinical course has been described [6]. Approximately 40% of patients can be associated with Peutz-Jeghers syndrome [124]. Malignant thecomas and fibrosarcomas are very rare.The rarity of non-epithelial ovarian tumors limits the ability to develop targeted therapies and evaluate them in well-powered clinical trials. A recent search of clinicaltrials.gov (Figure 5) revealed 166 interventional trials involving MOGCTs and only 12 involving SCSTs. Of the MOGCT trials, only 27 had results posted, and only 3 of these were specific to targeted therapy of MOGCTs. Within this group, there was only one female participant. For SCSTs, there was only one trial of targeted therapies with posted results. Several agents have been evaluated and trials from male testicular cancer patients may inform therapeutic strategies for non-epithelial ovarian tumors.The alterations in signaling pathways discussed above and pan-cancer observations of kinase alterations [125] provide the rationale for using kinase inhibitors for the treatment of non-epithelial ovarian cancers. Imatinib, an inhibitor of several kinases, including c-KIT, has been evaluated in a phase 2 clinical trial including testicular and ovarian germ cells tumors (NCT00042952), but results of this trial are not currently available (Table 2). Rationale for this trial can be found in the frequent alterations in KIT in MOGCTs noted above. Imatinib has been reported to elicit a response in two anecdotal cases of granulosa cell tumors [126,127], with and without overexpression of c-KIT. Benefit, lack thereof, or harm cannot be established based on these limited data. A trial evaluating sunitinib, a multikinase inhibitor, for the treatment of germ cell tumors did not recruit any women (NCT00453310) [128]. Another trial evaluated oxaliplatin in combination with the cyclin-dependent kinase 9 (CDK9) inhibitor, alvocidib. This trial included only one female, and the results are available in the abstract form only. The trial did not meet its primary endpoint (NCT00957905). A myriad of other inhibitors have been proposed as being mechanistically relevant for the treatment of MOGCTs or SCSTs [129], including epidermal growth factor receptor (EGFR), PDGFR, insulin-like growth factor 1 (IGFR1) and VEGF. However, clinical data for their efficacy is lacking, and case reports are prone to publication bias.VEGF inhibition with bevacizumab monotherapy was evaluated in 36 patients with recurrent SCSTs [130]. The majority of them had granulosa cell tumors, and 92% had received prior chemotherapy. A 17% partial response rate accompanied by decreases in tumor markers met the pre-specified criteria for further investigation with combination regimens based on the prior retrospective data suggesting higher response rates in granulosa cell tumors treated with dual bevacizumab therapy and chemotherapy [131]. We anticipate full results of a bevacizumab/paclitaxel combination in this population (NCT01770301). Results in the abstract form suggest that the addition of bevacizumab does not improve the PFS; it was reported to lead to better response rates, but higher rates of adverse events [132].Little is known about the immune response to non-epithelial ovarian tumors. The immune response to germ cell tumors is discussed above. Putative predictors of response to immunotherapy include tumor mutational burden, microsatellite instability, programmed death-1 (PD-1) and PD-L1 expression, and the presence of a host immune response within or around the tumor core [133,134]. Expression of the PD-L1 has been reported, in the abstract form only, in 75–80% of SCSTs [135], but immunotherapy has not been reported in a clinical trial of these tumors. Non-epithelial ovarian tumors are not characterized by a high mutational burden as discussed above, and microsatellite instability has not been reported in these tumors. Tumor lymphocyte infiltration occurs in germ cell tumors, but the trial of pembrolizumab for refractory germ cell tumors achieved no responses and did not include female participants (Table 2) [136]. Another trial of the dual checkpoint blockade with durvalumab, a PD-L1 inhibitor, and tremelimumab, a cytotoxic T lymphocyte-associated protein 4 (CTLA-4) inhibitor, is ongoing (NCT03158064). Other strategies of immunotherapy may become relevant for non-epithelial ovarian cancers. For example, recurrent clonal mutations such as those found in SCSTs can be targeted using adoptive T cell therapy [137], but the feasibility of this approach remains to be demonstrated and is dependent on HLA compatibility and efficient presentation to T cells. Cellular therapies are rapidly developing, and additional research is necessary to identify additional membrane-bound targets in non-epithelial ovarian cancer [138].Endocrine therapy has been suggested primarily for granulosa cell tumors [139]. Recent discoveries about the hormonal effects of mutated FOXL2 (Figure 4), which functions as a transcription factor that plays a role in granulosa cell development and in expression of hypophyseal gonadotropin-releasing hormone (GnRH) receptor expression [94], and the physiologic presence of follicle-stimulating hormone (FSH) receptors on granulosa cells provide the rationale for this approach. Promising reports are limited to case reports and series, while small trials have proven somewhat disappointing. Prior studies suggest efficacy using hormone blockade with leuprolide, a GnRH analog [140], and aromatase inhibition [141]. In one small but promising trial of 6 granulosa cell tumor patients treated with leuprolide, two patients experienced partial responses and 3—disease stability [140].Several meta-analyses have also looked at hormonal therapy in granulosa cell tumors, including one that included 19 studies (31 patients) where patients received a variety of therapies, including aromatase inhibitors (AI) and tamoxifen, found an objective response rate of 71.0% with 25.8% complete responses. In this series, all responses (9/9) were to AI, and there were no responses to tamoxifen. Interestingly, of the 9 patients treated with AI, none had progressed at the time of publication, with the follow-up time ranging from 6–54 months after starting the treatment [142]. A more recent review in 2018 provided an updated meta-analysis including 12 different combinations of non-AI hormonal therapies (including GnRH agonists/antagonists, tamoxifen, progesterone, and diethylstilbestrol—DES), among 50 patients. The pooled analysis found clinical benefit for 33 of 50 patients and at least partial response in 17 patients [143]. The same review also compiled data for AI specifically, including letrozole, anastrozole, and exemestane. In this data set of 25 patients, 7 experienced complete response, 5—partial response, and 7—disease stability for a total of 19 patients with clinical benefit. Given the favorable side effect profile, the authors concluded that AI may be an alternative to chemotherapy [143].There has been considerable interest surrounding AI given the data above and that FOXL2 is also known to activate aromatase [143]. The phase 2 PARAGON trial, which investigated anastrozole in a variety of gynecologic cancers, included a cohort of 41 postmenopausal recurrent granulosa cell tumor patients with estrogen receptor-positive disease. Results for this cohort, though somewhat disappointing compared with the above meta-analyses, have been published in the abstract form, showing a 9.8% response rate with 59% progression-free survival at 6 months [144].A trial is evaluating the androgen receptor signaling inhibitor, enzalutamide, in these patients (NCT03464201), and a trial of the progesterone antagonist onapristone for patients with progesterone receptor-positive low-grade ovarian tumors, including granulosa cell tumors, is currently recruiting (NCT03909152).Cytochrome P17 (CYP17) converts 17-hydroxyprogesterone to androstenedione and is downregulated by FOXL2. Mutations in FOXL2 as noted in granulosa cell tumors, therefore, may result in increased androstenedione levels. Trials of cytochrome P450 (CYP) 17 inhibition with the nonsteroidal inhibitor orteronel and the anti-fungal ketoconazole for the treatment of granulosa cell tumors have been reported, but are yet to be published in peer-reviewed journals. A case report of ketoconazole for this indication suggested activity in a patient with multiple recurrences who experienced at least 10 months of disease stability following the regimen [145], leading to a clinical trial by the same group. Based on this single-arm trial of ketoconazole in six patients with adult granulosa cell tumor, only three of whom were confirmed to have the somatic FOXL2 mutation, ketoconazole achieved stable disease in five patients and was granted an orphan designation for this indication by the European Medicines Agency. This data is available in the preprint form only [146]. Ortenerol was evaluated in ten patients in a trial that was terminated early due to slow recruitment [147]. The data available in the abstract form report a clinical benefit rate of 50%, with 3 patients achieving stable disease for over 12 months.Non-epithelial ovarian tumors are not considered to harbor homologous recombination deficiency (HRD); mutations in BRCA1/2 and other HRD-associated genes are uncommon (Figure 2). Poly (adenosine diphosphateribose) polymerase (PARP) inhibitors, which can confer synthetic lethality to cancer cells with HRD, have not been reported in clinical trials for non-epithelial ovarian tumors. Therapeutic compounds that may confer synthetic lethality to tumor cells with aneuploidy, such as MOGCTs, have also not been trialed [148] in these patients. The hypomethylating agent guadecitabine is being evaluated in combination with cisplatin for refractory germ cell tumors (NCT02429466), with signs of efficacy including 2/14 patients experiencing complete response. Only one patient in this trial had a MOGCT, which has only been reported in the abstract form. The data regarding the efficacy of serine-threonine kinase inhibitors for the treatment of non-epithelial tumors is also lacking, despite the data above suggesting pathway alterations in RAS-RAF-MEK-ERK/JNK and PI3K-AKT-mTOR.The histological and molecular similarities between MOGCTs and testicular germ cell tumors suggest that the strategies that are successful in treating testicular tumors may be applied to MOGCTs; furthermore, the vast majority of testicular tumors are germ cell tumors, which are much more common than MOGCTs [37,72]. Trials of unselected testicular tumor patients with advanced or platinum-resistant disease have not shown benefit of VEGF-tyrosine kinase inhibitors (TKI), EGFR-TKIs, or c-Kit inhibitors.Imatinib, an inhibitor of several tyrosine kinases, including c-KIT, was evaluated in six patients with KIT-expressing refractory germ cell tumors [149]. A decline in AFP was seen in a single patient, who had stable disease for 3 months. A case report outlined a complete response in one heavily pretreated testicular tumor patient [150]. Pazopanib, a multikinase inhibitor with activity against VEGF receptors, c-KIT, and additional kinases, did not meet its primary endpoint of progression-free survival (PFS) in a single-arm trial including 43 patients who had failed at least two platinum-containing chemotherapy regimens [151]. Approximately 70% of patients had a short-lived decrease in tumor markers, but the overall response rate (ORR) was less than 5% [151].Sunitinib also led to tumor marker decline, but showed no clinical benefit in a trial with ten patients [128]. However, in another trial, a single patient with chemotherapy-refractory testicular germ cell tumor had a clinical and biochemical response to sunitinib that lasted 17 months, potentially related to RET amplification in his tumor [152]. The combination of bevacizumab and oxaliplatin failed to meet the primary endpoint of PFS in a trial of 29 patients with chemotherapy-refractory testicular germ cell tumors [153]. Tivantinib, a different tyrosine receptor kinase inhibitor, failed to achieve any responses in patients with relapsed or refractory germ cell tumors [154].The mammalian target of rapamycin (mTOR) inhibitor, everolimus, elicited no clinically meaningful responses as a single agent for patients with relapsed or platinum-refractory disease [155]. A trial of avelumab, a PD-L1 checkpoint inhibitor, in a similar setting also failed to meet its PFS endpoint [156], as did a trial of pembrolizumab [136]. Brentuximab vedotin, an antibody–drug conjugate targeting CD30, was evaluated in patients with chemotherapy-refractory tumors with an embryonal carcinoma component and expression of CD30 [157]. A complete and very good partial response was described.Insights into the pathogenesis, molecular features, and omics of non-epithelial ovarian tumors have been accumulating in the recent years. The leading targeted therapy candidates from these translational and bench studies have been evaluated mainly in male germ cell tumors, without encouraging results. This exemplifies the fact that the vast majority of drug–indication pairs that are tested clinically do not achieve their expected clinical benefits and that many cancer therapies exert their effect in ways that are different from their presumed mechanisms [158]. A small number of clinical trials of females reflect paucity of potential indications for rare tumors that are often cured with surgery and chemotherapy. Recruitment of patients has depended on large referral centers or collaborative efforts. An additional challenge is to achieve diagnostic accuracy for atypical cases, which may require incorporation of mutational analysis, as demonstrated above for SCSTs.Additional potential vulnerabilities of MOGCTs and SCSTs have been described, but have not been targeted in clinical trials. These include alterations that are common across multiple cancers, such as aneuploidy, TERT mutations, and activation of the RAS-RAF-MEK-ERK/JNK and PI3K/AKT/mTOR pathways [59,60]. Targeting these alterations does not rely solely on recruitment of non-epithelial ovarian tumor patients; trials of other solid tumors, including testicular tumors, may inform the treatment for MOGCTs and SCSTs. Other alterations are fairly specific to subtypes of non-epithelial ovarian tumors, including the recurrent FOXL2 and common DICER1 mutations in adult granulosa cell tumors and Sertoli–Leydig cell tumors, respectively [59,60]. Including the analysis of these and other mutations in prospective clinical trials may help delineate differences in the clinical course and response to therapy of specific molecular subtypes. Many features of non-epithelial ovarian tumors remain understudied, including hallmarks of chemotherapy-refractory disease [159], the interaction with the tumor microenvironment, the immune response, and more. The functional effects of tumor markers is also understudied, but have been shown to potentially promote tumor progression in other tumors [160].Future studies should focus on effective and collaborative clinical trial designs that minimize the number of participants needed, e.g., multi-arm trials [161]. Clinical and molecular characterization of patients may lead to identification of prognostic and predictive biomarkers and may also detect rare somatic mutations that can be targeted through precision medicine initiatives and basket trials [162,163]. If effective therapies for chemotherapy-refractory disease are identified, they could potentially be leveraged to decrease the need for chemotherapy in the upfront setting and reduce long-term toxicity of common platinum-based regimens. Conceptualization—K.M., A.M.; Writing—original draft preparation, A.M., M.A.C., S.M.; Writing—review and editing—A.M., K.M., M.A.C., S.M., M.K., L.D.R., A.K.S., D.M.G.; Supervision: K.M., A.K.S., D.M.G. All authors—A.M., K.M., M.A.C., S.M., M.K., L.D.R., A.K.S., D.M.G.—have read and agreed to the published version of the manuscript.Ensign Endowment for Gynecologic Cancer Research (K.M.); CA217685, American Cancer Society and the Frank McGraw Memorial Chair in Cancer Research.The authors would like to acknowledge the American Association for Cancer Research and its financial and material support in the development of the AACR Project GENIE registry, as well as members of the consortium for their commitment to data sharing. Interpretations are the responsibility of the study authors. A.M. would like to thank Gopal Yadavalli and David Coleman for their support.Scientific consulting, Kiyatec, Merck; shareholder, Bio-Path; research funding, M-Trap (A.K.S.); stock and other ownership interests, Celgene, Johnson & Johnson, Biogin; consulting or advisory role, Clovis Oncology; research funding, Novartis; royalties from Elsevier for book editing, royalties from UpToDate for authorship (D.M.G.); honorarium, Chugai; textbook editorial expense, Springer; investigator meeting attendance expenses, VBL therapeutics (K.M.); consultant, Quantgene (L.D.R.); research funding, MSD (S.M.); advisory board, Tesaro, GSK (K.M.).Epithelial and non-epithelial cells of the ovary. The cells from which the primary epithelial and non-epithelial ovarian tumors originate are depicted. Epithelial ovarian cancer arises from the surface epithelium of the ovary, fallopian tubes, and peritoneum. Non-epithelial ovarian tumors arise from gonadal germ cells, sex cord–stromal cells, and other non-epithelial cells.Differences in genetic alterations between epithelial and non-epithelial ovarian tumors (GENIE/AACR database). Aggregated data for 5 genes were derived from the GENIE/AACR database for high-grade serous ovarian cancer (HGSOC) (an epithelial ovarian cancer), female germ cells tumors (MOGCT) and female sex cord–stromal tumors (SCST). Alterations included mutations (excluding synonymous mutations), amplifications, homozygous deletions, and fusions. Alterations with <1% frequency across the three tumor categories were excluded. The breakdown of tumor subtypes in this database does not reflect their prevalence in the population.Common alterations in malignant ovarian germ cell tumors. Common alterations in the most prevalent malignant ovarian germ cell tumors (MOGCTs) are shown. (A) Dysgerminomas frequently demonstrate mutations in c-KIT and KRAS. (B) Yolk sac tumors have frequent amplifications of the genes PIK3CA and AKT1 in the PI3K/AKT/mTOR pathway. Both dysgerminomas and yolk sac tumors are characterized by marked aneuploidy, whereas (C) immature teratoma is characterized by near-diploid copy neutral loss of heterozygosity (LOH).Common alterations in sex cord–stromal cell tumors. (A) Adult granulosa cell tumors almost ubiquitously have a somatic mutation in FOXL2, leading to transcriptional alterations, including in cytochrome P450 (CYP) 17 and 19 expression. (B) Juvenile granulosa cell tumors have mutations in GNAS in approximately 30% of cases. AKT1 is the most commonly amplified gene. (C) Approximately 60% of Sertoli–Leydig tumors are associated with a mutation in the ribonuclease III (RNAse III) DICER1, which can be a germline mutation predisposing to several cancers. Mut—mutated.Clinical trials of targeted therapies of non-epithelial malignant ovarian tumors. Clinicaltrials.gov was searched for terms related to (A) malignant ovarian germ cells tumors (MOGCTs) and (B) sex cord–stromal tumors (SCSTs). Observational trials were not reviewed. * Trials evaluating drugs in multiple cancers, epithelial ovarian cancer and trials with chemotherapy-only interventions were classified as irrelevant. Relevant trials with results were evaluated for the number of female participants with non-epithelial ovarian tumors.Non-epithelial ovarian malignancies and their common genetic alterations.Note: The most common subtypes of malignant ovarian germ cell tumors and malignant ovarian sex cord–stromal tumors and their corresponding commonly identified alterations are noted [40,41,42]. Frequencies are estimates based on the available data, which are limited for certain alterations or tumor subtypes. * LOH—loss of heterozygosity. # NOS—not otherwise specified. Gsp—Gs-Protein, referring to the alpha subunit of G-protein (Gs).Clinical trials of targeted therapies for female MOGCTs and SCSTs.Clinical trials of targeted therapies for MOGCTs and SCSTs are summarized in the table. Trials without female participants are not listed. Abbreviations: ORR—overall response rate, CR—complete response, SD—stable disease, PFS—progression-free survival, PR—progesterone receptor. NA—not available. NCT—National Clinical Trial. 5FU—fluorouracil

Med-MDPI/cancers/cancers-12-06-01399.txt

@@ -0,0 +1 @@
+Chronic myelogenous leukemia (CML) is the most common type of leukemia in adults, and more than 90% of CML patients harbor the abnormal Philadelphia chromosome (Ph) that encodes the BCR-ABL oncoprotein. Although the ABL kinase inhibitor (imatinib) has proven to be very effective in achieving high remission rates and improving prognosis, up to 33% of CML patients still cannot achieve an optimal response. Here, we used CRISPR/Cas9 to specifically target the BCR-ABL junction region in K562 cells, resulting in the inhibition of cancer cell growth and oncogenesis. Due to the variety of BCR-ABL junctions in CML patients, we utilized gene editing of the human ABL gene for clinical applications. Using the ABL gene-edited virus in K562 cells, we detected 41.2% indels in ABL sgRNA_2-infected cells. The ABL-edited cells reveled significant suppression of BCR-ABL protein expression and downstream signals, inhibiting cell growth and increasing cell apoptosis. Next, we introduced the ABL gene-edited virus into a systemic K562 leukemia xenograft mouse model, and bioluminescence imaging of the mice showed a significant reduction in the leukemia cell population in ABL-targeted mice, compared to the scramble sgRNA virus-injected mice. In CML cells from clinical samples, infection with the ABL gene-edited virus resulted in more than 30.9% indels and significant cancer cell death. Notably, no off-target effects or bone marrow cell suppression was found using the ABL gene-edited virus, ensuring both user safety and treatment efficacy. This study demonstrated the critical role of the ABL gene in maintaining CML cell survival and tumorigenicity in vitro and in vivo. ABL gene editing-based therapy might provide a potential strategy for imatinib-insensitive or resistant CML patients. Leukemia is classified as acute or chronic and as myelogenous or lymphocytic; the subtypes include acute myelogenous leukemia (AML), chronic myelogenous leukemia (CML), acute lymphocytic leukemia (ALL) and chronic lymphocytic leukemia (CLL), and chronic leukemia grows slowly, and progressively worsens over time. CML is the most common type of leukemia in adults, comprising 15%–25% of all adult leukemia cases worldwide [1]. A chromosome translocation, between the long arms of chromosomes 9 and 22, t(9;22)(q34;q11), is found in over 90% of CML patients, in a lower proportion of ALL or biphenotypic acute leukemia cases and in rare cases of de novo AML [2,3]. This well-known Philadelphia (Ph) chromosome produces the BCR-ABL oncogenic fusion protein that activates multiple signaling pathways involved in the cell cycle, adhesion and apoptosis [4,5]. In addition, expression of this BCR-ABL oncoprotein transforms hematopoietic progenitor cells; in an animal model, this transformation event activated downstream signaling proteins that increase cell survival and proliferation, indicating the essential oncogenic role of BCR-ABL in CML cells [6].There are several treatments for CML patients, including tyrosine kinase inhibitors (TKIs), hematopoietic stem cell transplantation and chemotherapy. The predominant treatment for CML is a TKI, and the TKI imatinib (Gleevec) is the first-line targeted therapy for CML. Imatinib has been shown in recent years to be highly effective at increasing the life expectancy of CML patients [7]. Imatinib inhibits BCR-ABL, subsequently inhibiting the proliferation of abnormal leukemia cells [8]. However, this TKI has side effects, such as nausea, headache, diarrhea, fatigue, rash, hypertension and diabetes. Additionally, strict treatment compliance is essential, since any missed dose might contribute to the development of drug resistance and gene mutations [9], leading to leukemia recurrence. Accordingly [10], approximately 33% of patients with CML treated with imatinib do not achieve a complete cytogenetic response (CCyR). Therefore, resistance to TKIs is still the primary problem that needs to be solved in CML treatment.The mechanisms of imatinib resistance could be both BCR-ABL dependent (gene amplification or point mutations) and BCR-ABL independent [11]. Therefore, a more effective and precise CML therapeutic strategy is urgently needed. Recently, clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 technology has initiated a new era of genome editing. CRISPR/Cas9 technology overcomes the limitations of traditional genome editing techniques, which are considered ineffective, time-consuming and laborious; thus, it can be applied as a general-purpose gene editing system [12]. CRISPR/Cas9 generates double-strand breaks (DSBs) at target sites by recognizing 20-nt sequences that match an engineered gRNA and a 3-nt protospacer adjacent motif (PAM), located downstream of the target sequence. The subsequent cellular DNA repair process, nonhomologous end joining (NHEJ), is an error-prone DSB repair mechanism that introduces the desired genetic insertions, deletions or substitutions at the target site [13]. In other words, CRISPR/Cas9 allows researchers to rapidly generate a pool of gene knockout cells without a massive gene engineering design effort. This gene editing system has been widely used to correct mutated genes related to disease and cancer [14]. The CRISPR/Cas9 system can potentially modify disease-related genes in vitro and in vivo; however, few studies have explored the potential of this system in general cancer therapeutic applications.The BCR-ABL fusion gene is an ideal target for CRISPR/Cas9 gene therapy in CML [15]. However, the junction regions of the BCR-ABL gene are different in every CML patient [16]. Therefore, we utilized the CRISPR/Cas9 gene editing strategy to cleave the ABL gene and eliminated its oncogenic activity in vitro. To ensure gene editing efficiency, we used several assays, such as Sanger DNA sequencing, tracking of indels by decomposition (TIDE) analysis, restriction fragment length polymorphism (RFLP) of the ABL gene region and protein analysis of K562 cells. In addition, the safety of CRISPR/Cas9-mediated gene editing in human cells was addressed by an analysis of potential off-target genes and bone marrow cells. Notably, our effective anticancer results in a systemic leukemia animal model treated with virus-mediated gene editing therapy suggested an alternative treatment for clinical CML patients who are insensitive or resistant to imatinib treatment.The human leukemia K562 cell line (CML) was kindly provided by Dr. Kai-Wen Hsu, Research Center for Tumor Medical Science, China Medical University, Taichung, Taiwan. The bone marrow derived epithelial cells were kindly proved by Dr. Chia-Ling Hsieh, The Ph.D. Program for Translational Medicine, College of Medical Science and Technology, Taipei Medical University, Taipei, Taiwan. The cells were maintained in Dulbecco’s Modified Eagle Medium: Nutrient Mixture F-12 (DMEM/F-12) (Gibco, Grand Island, NY, USA). The peripheral blood of CML participants and healthy controls was obtained at Shuang Ho Hospital, Taipei Medical University, New Taipei City, Taiwan, according to a protocol approved by the Institutional Review Board (N201711069). Clinical parameters, such as RBC count, WBC count, karyotype and fluorescent in-situ hybridization (FISH) analysis, were determined.Cell viability was determined using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium (MTT), which is based on reduction of the yellow MTT to purple formazan by living cells [17]. In 96-well plates, 8 × 104 cells were seeded in 100 μL of DMEM/F12 per well and were exposed to different concentrations of Imatinib according to the experimental protocol. After 48 h of treatment, the medium was changed to fresh medium containing 1 mg/mL of MTT. Two hours later, 100 μL of DMSO was added in each well and the absorbance at 570 and 630 nM was determined. The percentage of cell viability was calculated using a formula [percentage viability = (average OD of sample/average OD of control) × 100].K562 cell proliferation was determined using the colorimetric bromodeoxyuridine (BrdU), which measures the incorporation of BrdU, a thymidine analogue, into the DNA of proliferating cells. The BrdU assay used in this study was an ELISA-based assay that was performed as recommended by the manufacturer (Merck-Millipore, USA). Imatimib treated K562 cells or ABL sgRNA virus infected K562 cells were incubated for 36 h at 37 °C, the media were supplemented with 10 μM BrdU and incubated for an additional 12 h. The cells were then stained with a peroxidase-labeled antibody against BrdU, followed by TMB Peroxidase Substrate addition for 30 min and acid stop solution exposure. The absorbance of the samples at 450 nm with a reference wavelength of 540 nm was measured using a microplate reader. K562 cells were transfected with pcDNA3 plasmids expressing the firefly luciferase gene (the gene sequences were originally from luc4.1; Chris Contag, Stanford University, Stanford, CA, USA), as described previously [18]. Briefly, 1 × 106 K562 cells were washed twice with phosphate-buffered saline (PBS) and mixed with 10 μg of plasmid. Two 1.2-kV pulses were applied for 20 milliseconds using a pipette-type Microporator MP-100 (Digital Bio, Seoul, Korea). Stable cells were selected 48 h later with G418 (1 mg/mL). Bioluminescent derivatives of K562 cells were used for further in vitro and in vivo studies.Four-week-old female severe combined immunodeficient (SCID) mice were purchased from the National Science Council Animal Center (Taipei, Taiwan) and housed in micro-isolator cages at the Laboratory Animal Service Center in the China Medical University (Taichun, Taiwan). This study was carried out in strict accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals from the National Institutes of Health. The protocol was approved by the Institutional Animal Care and Use Committee (IACUC) at China Medical University (Permit Number: 2018-030-1). The systemic leukemia animal model uses immune-competent SCID mice to mimic cancer development in humans. Six mice were anesthetized with 2% isoflurane, and 5 × 106 bioluminescent K562 cells were injected via the tail vein into each SCID mouse. The mice were then grouped (three mice/group) after three weeks, and injected via the tail vein with 5 × 108 copy number of scramble (SC)- or ABL-targeting virus in a 20 μL normal saline. Throughout the study, all mice were kept in an environmentally controlled room maintained at 21–24 °C and 43–65% relative humidity. During the experiment, all animals underwent bioluminescence imaging every two weeks to observe the CML cells. All surgeries were performed under isoflurane anesthesia, and all efforts were made to minimize suffering. During the experiment, no stress or abnormal behavior due to the cancer was observed in the mice. The health status of the animals was monitored once daily by a qualified veterinarian.Bioluminescence imaging was performed with a highly sensitive, cooled CCD camera mounted in a light-tight specimen box (In Vivo Imaging System-IVIS; Alameda, CA, USA). Fifteen minutes before imaging, the mice were injected i.p. with D-luciferin (200 mg/kg). The animals were placed on a warmed stage inside the camera box, and were continuously exposed to 2.5% isoflurane to sustain sedation during imaging. Every group of mice was imaged for 1, 5, 10, and 30 s. The light emitted from the mice was detected by the IVIS camera system, integrated, digitized and displayed. Regions of interest on the displayed images were identified, and the total photon counts were quantified using Living Image® software 4.0 (Caliper, Alameda, CA, USA). K562 cells (5 × 105 cells/dish) were plated in 6-cm dishes for infection. An equal number of virus particles and K562 cells was defined as 1-fold. K562 cells were exposed to different folds of concentrated virus harboring the pLJM1-EGFP plasmid. Lentivirus-transduced cells were harvested three days after infection, and the GFP-positive cell population was analyzed by flow cytometry (FACSCalibur, BD Biosciences, San Diego, CA, USA).Primers targeting the luciferase (forward 5′-CCGTCGTATTCGTGAGC-3′ and reverse 5′-GGTGGCAAATGGGAAGT-3′) and mouse β-glucuronidase (GUS, forward 5′-TGAACTCTTGAAAGCCTGC-3′ and reverse 5′-GAAATGGAGGACCAGCTCATA-3′) genes were used to quantify human CML DNA in the peripheral blood of the mice. Primers for the WPRE region (forward 5′-TCATGCTATTGCTTCCCGTA-3′ and reverse 5′-CCAAGGAAAGGACGATGAT-3′) were used for lentivirus quantification. All oligo primers were synthesized by Genomics BioSci and Tech (Taipei, Taiwan). A LightCycler thermocycler (Roche Molecular Biochemicals, Mannheim, Germany) was used for Q-PCR analysis. One microliter of sample and master mix was first denatured for 10 min at 95 °C and then subjected to 40 cycles (denaturation at 95 °C for 5 s; annealing at 60 °C for 5 s; and elongation at 72 °C for 10 s) with detection of fluorescence intensity. All the PCR samples underwent a melting curve analysis to detect non-specific PCR products. Luciferase gene expression from the Q-PCR analysis was normalized to mouse GUS expression as an indicator of DNA input using the built-in Roche LightCycler Software, version 4.To generate an absolute quantitative standard curve for Q-PCR analysis, we cloned the PCR product of the human GUS gene into the TA cloning vector (pTA® Easy Cloning Kit, Genomics BioSci and Tech, Taipei, Taiwan). After gene sequencing, E. coli amplification, plasmid purification and molecular weight determination, the copies of the GUS gene were calculated and diluted from 108 to 102 /μL. Each copied gene was measured for accuracy and a linear correlation.For western blot analysis, K562 cells were washed once with ice-cold PBS and lysed with radioimmunoprecipitation assay (RIPA) lysis buffer containing protease inhibitors. Fifty micrograms of protein from each sample was resolved by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to a nitrocellulose membrane. The anti-GAPDH (sc-32233), anti-p-ERK (sc-7383), P21 (sc-817) and c-Abl (sc-23) antibodies were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA), and the anti-PARP (#9541) antibodies were purchased from Cell Signaling Technology (Danvers, MA, USA). The secondary anti-mouse and anti-rabbit antibodies were purchased from Santa Cruz Biotechnology. Most of the primary antibodies were used at a 1:1000 dilution with overnight hybridization, whereas only the c-Abl antibody was used at a 1:250 dilution, followed by a one-hour incubation with a 1:4000 dilution of the secondary antibodies. All the western blotting was measured and quantified by Image J software. Original whole blot can be found at File S2. The comparison was done by fold of control cells. Lentiviral particles were produced by transient transfection of Phoenix-ECO cells (CRL-3214) using TransIT®-LT1 Reagent (Mirus Bio LLC, Madison, WI, USA). Guide oligonucleotides were phosphorylated, annealed, and cloned into the BsmBI site of the lentiCRISPR v2 vector (Addgene, 52961, kindly provided by Feng Zhang), according to the Zhang laboratory protocol [19] (F. Zhang lab, MIT, Cambridge, MA, USA). All the plasmid constructs were verified by sequencing. The lentiCRISPR construct or the pLJM1-EGFP plasmid (Addgene plasmid #19319, a gift from David Sabatini) was co-transfected with pMD2.G (Addgene plasmid #12259) and psPAX2 (Addgene plasmid #12260, both kindly provided by Didier Trono, EPFL, Lausanne, Switzerland). Lentiviral particles were collected at 36 and 72 h and then concentrated with a Lenti-X Concentrator® (Clontech, Mountain View, CA, USA). The lentivirus concentration for each gene was quantified by Q-PCR. Biohazards and restricted materials were used in this study in accordance with the “Safety Guidelines for Biosafety Level 1 to Level 3 Laboratory”. The protocol was approved by the Institutional Biosafety Committee (G-106-097) at Taipei Medical University, Taipei, Taiwan.Custom sgRNAs for ABL and mABL gene were designed using the MIT CRISPR Design website (https://www.benchling.com/crispr/) with the sequence of ABL (NM_005157). This website provides both on-target sequences and off-target possibilities. We selected the highest scoring off-target sequences in the ABL protein-coding region, sgRNA_1 and sgRNA_2.Genomic DNA was extracted, and the ABL exon 2 region was PCR-amplified using the following primers: forward GAGAGGCTGGTGACACGTAA and reverse TTTGTAGAAAGCTTCCTTTTCCCG. Off-target sequences were PCR-amplified using the following primers: ADAMTSL1, forward TTTCTTCCTTTACTCTGCCAAATTA and reverse TACAATTCCAAGCTTCCGAT; TBRG4, forward TAGGGAGTAGATGCTCGTT and reverse GGACCTGGGAATCTGAATTAT; and C17orf75, forward CATGTCCCATCACTGCTC and reverse TTCTCCGTTTCATTCTGTGT. The mABL exon 2 region was PCR-amplified using the following primers: forward GGGAACCAAGTGAGACTATAC and reverse CAGGCATTTCTGCTCTCAA. The PCR products were purified using a PCR Clean-up Purification Kit and sequenced by Sanger sequencing using the forward PCR primers. The editing efficiency of the sgRNAs and the potential induced mutations were assessed using TIDE software (https://www.deskgen.com/landing/tide.html#/tide; Netherlands Cancer Institute, Amsterdam, Netherlands), which required only two Sanger sequencing runs from wild-type cells and mutated cells. PCR products (approximately 100 ng per assay) of the ABL exon 2 region were incubated for 30 min at 37 °C with Cas9 protein (30 nM) and sgRNAs (30 nM) in 10 μl of NEB buffer 3. After cleavage, RNase A (2 μg) was added, and the reaction mixture was incubated for 15 min at 37 °C to remove RNA. Next, proteinase K (2 μg) was added, and the reaction mixture was incubated for 15 min at 58 °C to remove the Cas9 protein. The products were resolved on 2% agarose gels and visualized by ethidium bromide (EtBr) staining.Routine peripheral blood chromosomal analysis was performed using phytohemagglutinin (PHA) stimulation and standard techniques. Metaphase chromosomes were stained by giemsa-trypsin banding. Twenty metaphase cells were examined for each patient. The resolution of our protocol is around 500 bands on average. Fluorescence in situ hybridization was performed on interphase nuclei using Vysis LSI BCR/ABL dual color translocation probe set (Vysis, Downers Grove, IL, USA). Hybridization was carried out in a humidified chamber at 37 °C for at least 16 h. The slides were washed with 0.4× SSC/0.3% NP40 three times for 2 min each and then air-dried in the dark. Hybridization areas were counterstained with 20 μL DAPI (Vysis Inc, Downers Grove, IL, USA). Cells were observed under a Zeiss fluorescence microscope (Carl Zeiss Microimaging GmbH, Gottingen, Germany) and images were captured and analyzed using GenASIs Scan & Analysis platform of Applied Spectral Imaging (Carlsbad, CA, USA).BALB/c mice were anesthetized with 2% isoflurane and tail-vein injected with 5 × 108 copy number of scramble (SC)- or mABL-targeting virus in a 20 μL normal saline. After four weeks, the blood sampling was performed by cardiac puncture and stored in EDTA contained blood collection tube. The blood samples were then automatic hematology analyzed by IDEXX Procyte Dx (IDEXX Laboratories, Westbrook, ME, USA). The leukocytes from cardiac puncture collected blood, kidney, liver, lung and spleen from each mouse was isolated and Sanger sequenced for mouse ABL gene disruption analysis. Cardiac puncture collected blood were lysed of RBC and washed in FACS buffer (PBS containing 2% heat-inactivated FBS). For surface staining, the leukocytes were first incubated 5 min with anti-CD16/CD32 mouse Fc block (biolegend company, San Diego, CA, USA), followed by double staining with anti-CD4 (FITC) and CD8 (PE) antibodies (biolegend company, San Diego, CA, USA) for 30 min. Cells were washed twice, collected on a FACSCalibur flow cytometer, and analyzed using CellQuest software (BD Biosciences, San Diego, CA, USA). All data are expressed as the mean±standard error, and the differences were analyzed by Student’s t-test for pairwise samples. All statistical comparisons were performed using SigmaPlot graphing software (San Jose, CA, USA) and Statistical Package for the Social Sciences v.13 (SPSS, Chicago, IL, USA). A p-value < 0.05 was considered statistically significant, and all statistical tests were two-sided.This animal study was carried out in strict accordance with the recommendations of the Guide for the Care and Use of Laboratory Animals from the National Institutes of Health. The protocol was approved by the Institutional Animal Care and Use Committee (IACUC) at China Medical University (Permit Number: 2018-030-1).The clinical patient study was approved by the Institutional Review Board (N201711069) in Taipei Medical University, Taipei, Taiwan, and the patients provided written consent.Several cell types, including primary human fibroblasts, human umbilical vein endothelial cells, Jurkat cells, and leukemia cells, have long been considered difficult to transfect [20]. To optimize leukemia cell transduction, we transduced CML (K562) cells with different concentrations of the EGFP expression virus (from 0.3 to 200:1; virus number:cell number). After transduction, the number of GFP-positive K562 cells was determined by flow cytometry (Figure S1A); this number was significantly increased after infection at a ratio of 30:1 (purple), and reached a maximum at a ratio of 200:1 (yellow, Figure S1B). We next examined the association between virus input and the GFP-positive K562 cell number. The results showed that a 100:1 virus ratio resulted in transduction of 40% of the K562 cell population, whereas additional virus did not significantly increase the transduction efficiency (defined as MOI = 1). In subsequent experiments, we used a MOI = 1 virus concentration for gene delivery (Figure S1C).For gene editing using CRISPR/Cas9 technology on BCR-ABL junctions, we amplified a DNA sequence of K562 CML cells using primers for the e12(b1) transcript on BCR and the a3 transcript on ABL [21]. The sequence showed BCR-ABL reciprocal translocation in K562 cells: sequence tags mapped to chr9:133,607,145-133,607,558 in the ABL gene were linked to chr22:23,632,242-23,632,742 in the BCR gene (Figure 1A), producing a P210 BCR-ABL Major form fusion protein [22]. Next, we designed two sgRNAs for the CRISPR/Cas editing system; protospacer 1 (BCR-ABL_1) targets the plus strand, whereas protospacer 2 (BCR-ABL_2) targets the negative strand (Figure 1B), specifically targeting the BCR-ABL junction sequence in K562 cells. Transduction of the K562 cells with the target scrambled (SC) virus produced a wild-type ABL sequence, as shown by Sanger sequencing (Figure 1C,D), with no evidence of gene editing. However, transduction with the BCR-ABL sgRNA_1 virus or the BCR-ABL sgRNA_2 virus led to multiple gene disruptions at the predicted cleavage sites (red arrowhead, Figure 1E,F) of both protospacers. In addition, both BCR-ABL sgRNA_1 and BCR-sgRNA_2 virus infections showed considerable gene editing efficiency, with 53% and 97.5% editing of the cell pools shown by TIDE analysis, respectively (Figure 1G,H, Figure S2A,B). The most frequent mutations in the BCR-ABL sgRNA_1 cell pool were other edits (37.6%) and 2-bp deletions (10.1%), whereas those in the BCR-ABL sgRNA_2 cell pool were other mutations (77.3%) and 1-bp insertions (11.6%). The algorithm predicted the same patterns of genome repair for BCR-ABL sgRNA_1 and BCR-ABL sgRNA_2, which included mutations mainly at the cleavage sites (Figure S2C,D). Next, we determined whether cancer cell viability is influenced by BCR-ABL gene disruption in K562 cells. We then confirmed the gene editing efficiency by assessing ABL protein expression. We observed that both the 145-kDa ABL protein and the 210-kDa BCR-ABL oncoprotein present in K562 cells (Figure 1I). Upon infection with the BCR-ABL sgRNA_1 and BCR-ABL sgRNA_2 virus, the BCR-ABL oncoprotein levels were significantly decreased, compared to control and scramble virus-infected cells. Protein analysis also demonstrated that PARP cleavage and the tumor suppressor P21 were both significantly induced in BCR-ABL sgRNA virus-infected K562 cells. We also determined whether cancer cell viability is influenced by BCR-ABL gene disruption in K562 cells. The results showe that the viability of K562 cells was significantly inhibited by infection with the BCR-ABL sgRNA_2 virus (p < 0.001) compared to infection with the BCR-ABL sgRNA_1 virus (p = 0.011) and scramble virus (Figure 2J). These findings demonstrated that targeting the BCR-ABL gene caused a dramatic disruption in DNA sequence and protein suppression and eventually suppressed cell viability in CML cells, implying that BCR-ABL gene editing would be a sufficient and effective anticancer strategy. However, in clinical patients, the genomic breakpoints in both BCR and ABL of CML cells are dispersed over intervals of 3.0 kb and ~150 kb, respectively. Each patient’s fusion sequence is therefore virtually unique, indicating that a patient-specific BCR-ABL CRISPR/Cas9 gene editing strategy for anticancer therapy will be a challenge. To be more specific, due to the limited option on BCR-ABL junction sequence in CML patients, it will be difficult to design specific and effective sgRNAs with appropriate PAM (NGG) sequence for CRISPR/Cas9 based personalized medicine. Accordingly, such sites only occur once in every 128 bp of random DNA sequence [23].To develop a novel CML therapy for clinical use, we utilized CRISPR/Cas9 genomic editing by targeting two custom-designed protospacers on the human ABL locus. As shown in the ABL genomic map on chromosome 9 (Figure 2A), protospacer 1 (ABL sgRNA_1) targets the plus strand, whereas protospacer 2 (ABL sgRNA_2) targets the negative strand. Transduction of K562 cells with the target scrambled (SC) virus produced a wild-type ABL sequence, as shown by Sanger sequencing (Figure 2B,C), with no evidence of gene editing. However, transduction with the ABL sgRNA_1 virus or the ABL sgRNA_2 virus led to multiple gene disruptions at the predicted cleavage sites (Figure 2D,E). In addition, ABL sgRNA_2 virus infection was shown by TIDE analysis to have strong gene editing efficiency: 41.2% of the cell pool was edited (Figure 2G,I) compared to only 8.8% with the ABL sgRNA_1 virus (Figure 2F,H). The most frequent mutations in the ABL sgRNA_1 cell pool were 1-bp insertions (3.4%) and 1-bp deletions (3.1%), whereas those in the ABL sgRNA_2 cell pool were 1-bp insertions (38.4%) and other mutations (2.7%). Again, the algorithm predicted the same patterns of genome repair for both ABL sgRNAs, which included mutations mainly at the cleavage sites (Figure S2A,B).To confirm the gene editing efficiency of both ABL sgRNAs, we assessed the PCR amplification of the gene editing site by RNA-guided engineered nuclease-restriction fragment length polymorphism (RGEN-RFLP) analysis. To perform this assay, we first purified the Cas9 protein by affinity chromatography pull-down of the histidine tag in E. coli lysate containing NLS-Cas9 (Figure S3). The molecular weight of the purified Cas9 protein was approximately 140 kDa. RGEN-RFLP analysis of the gene editing efficiency of the ABL DNA region showed that the SC sgRNA without Cas9 did not cleave the DNA (Figure 3A); however, the SC sgRNA with Cas9 fully cleaved the DNA into fragments, indicating a 100% wild-type DNA sequence without gene disruption. In contrast, the proportion of uncut DNA from K562 cells was higher after exposure to ABL sgRNA_2 than ABL sgRNA_1, indicating that ABL sgRNA_2 had a higher gene editing efficiency than ABL sgRNA_1, with 38% versus 6.8% indels in the ABL gene region (the fragment from the DNA cleavage is indicated with an asterisk). To confirm the gene editing efficiency, we assessed ABL protein expression by western blot analysis. We observed that both the 145-kDa ABL protein and the 210-kDa BCR-ABL oncoprotein are present in K562 cells (Figure 3B). Upon infection with the ABL sgRNA_2 virus, the BCR-ABL oncoprotein levels were significantly decreased, compared to those in the ABL sgRNA_2 virus- and scramble virus-infected cells. Protein analysis also demonstrated that the tumor suppressor P21 was significantly induced in ABL sgRNA_2 virus-infected K562 cells. This observation directly explains the strong induction of apoptosis caused by PARP cleavage in the ABL-edited leukemia cells. Next, we determined whether cancer cell viability is influenced by ABL gene disruption in K562 cells. The results showed that the viability of K562 cells was significantly inhibited by infection with the ABL sgRNA_2 virus (p = 0.001) compared to infection with the ABL sgRNA_1 virus and scramble virus (Figure 3C). In addition, the cell proliferation of K562 cells shown by BrdU incorporation was significantly inhibited by infection with the ABL sgRNA_2 virus (p < 0.05), compared to that of the SC sgRNA-infected cells (Figure 3D). These findings demonstrated that the targeting ABL gene have similar anti-cancer effect of targeting BCR-ABL gene, causing a dramatic suppression of both cell viability and cell proliferation in CML cells. Notably, in difficult-to-transfect cells, a high level of gene editing is possible with a virus with an optimized sgRNA, indicating the future potential for gene therapy in all CML patients.To evaluate the therapeutic effects of ABL disruption in vivo, we established a systemic leukemia animal model to evaluate the anticancer effect of an ABL-targeted CRISPR/Cas9. Luciferase-labeled human CML K562 cells were purified and expanded under normal culture conditions. SCID mice were injected with luciferase-labeled K562 cells through the tail vein. After three weeks, the mice were grouped and then injected via the tail vein with the SC- or ABL sgRNA_2-targeted CRISPR/Cas9 virus. The bioluminescence in each mouse was detected ventrally (Figure 4A) and dorsally (Figure 4B) every two weeks, to evaluate changes in leukemia cell number. The bioluminescence images clearly showed strong K562 cell growth in animals injected with the SC-targeted CRISPR/Cas9 virus from week 3 to week 7, especially in the ventral images of the heart and brain (blood-enriched organs). However, in mice treated with the ABL-targeted CRISPR/Cas9 virus, the K562 cell number only slightly increased in the ventral and dorsal bioluminescence analyses. Radiance photon measurements were used to calculate the cell number in each mouse, and the K562 cell growth curve showed significantly greater growth in the SC group than in the ABL-targeted group in both the ventral (p = 0.02, Figure 4C) and dorsal (p = 0.03, Figure 4D) analyses. Notably, this leukemia animal model not only mimics a gene therapy application in CML patients but also indicates that one course of ABL-targeted CRISPR/Cas9 virus treatment is sufficiently effective to suppress leukemia cell growth. To confirm the bioluminescence observations, we stained peripheral blood smears from the SC and ABL groups (Figure 4E). Myoblasts and neutrophils are shown at 400× and 1000× magnification; the smears showed a significant difference in the immature differential white blood cell count in the SC and ABL groups, with average values of 35 and 9, respectively (p < 0.05, Figure 4F). In addition, enhanced platelet aggregation and an increased basophil number were found in the SC group, mirroring the clinical diagnostic features of CML patients. Finally, we purified DNA from the peripheral blood of mice in the SC and ABL groups. Absolute quantification of the DNA was performed with luciferase-specific primers, and the results were normalized to the mouse GUS gene copy number in the qPCR analysis (Figure 4G). Again, the data showed significantly higher luciferase gene (carried by the CML cells) expression in the SC-targeted CRISPR/Cas9 virus-treated mice than in the ABL-targeted CRISPR/Cas9 virus-treated mice, indicating a strong therapeutic effect of ABL-targeted gene therapy in a systemic animal model of human leukemia.To evaluate whether ABL-targeted gene editing can be used as an antileukemia therapy in the clinic, we infected the peripheral blood mononuclear cells (PBMCs) of CML patients with the ABL sgRNA_2-targeted CRISPR/Cas9 virus. The patient was a 49-year-old male who had lost 10 kg of body weight within one year, and complained of a low-grade fever, early satiety and loss of energy. The laboratory examination of this patient showed normal values for GOT, GPT, creatinine, AFP, CEA and CA199. However, the hematology analysis showed that RBC, lymphocyte and monocyte counts were far below average, with poor RBC quality, as shown in the HGB and HCT measurements. The WBC count was 320.8 × 103/μL (normal range is 4.8–10.8 × 103/μL), and the WBC differential count identified 12% band cells, 8% metamyelocytes, 4% promyelocytes and 5% blasts of all immature leukocyte stages in peripheral blood. Conventional cytogenetic analysis of products of conception revealed, at a 400 G-band level of resolution, an abnormal karyotype 46, XY, t(9;22)(q34;q11.2) (Figure 5A). Clear karyotypes of chromosome 9 and chromosome 22 are also shown (Figure 5B). To further confirm the chromosomal translocation, we performed a FISH analysis of the patient through the hybridization of combined probes targeting chromosome 9 and centromere 22 (Figure 5C). Our data revealed that the “fusion” signal resided on the derived chromosome 9 and not on the derived chromosome 22, as would typically be expected in CML with one isolated fusion signal. Next, we investigated the efficacy of CRISPR/Cas9 genome editing in clinical CML cells by SC, ABL sgRNA_1 and ABL sgRNA_2 virus infection with the optimized virus concentration. The CML cells transduced with or without the SC virus produced a wild-type ABL sequence, as shown by Sanger sequencing (Figure 5D,E), with no evidence of gene editing. However, transduction with the ABL sgRNA_1 or ABL sgRNA_2 virus led to multiple gene disruptions at the predicted cleavage sites (red arrowhead) of both protospacers, with more pronounced effects in ABL sgRNA_2-transduced CML cells than ABL sgRNA_1-transduced CML cells (Figure 5F,G). In addition, ABL sgRNA_2 virus infection was shown to have strong gene editing efficiency by TIDE analysis: 30.9% of the cell pool was edited, rather than 9.4% with a 1-bp deletion in ABL sgRNA_1 virus infection (Figure 5H–J). The most frequent mutations in the ABL sgRNA_2 CML cell pool were 1-bp insertions (27.7%) and other mutations (1.9%) (Figure 5K). Next, we used a LIVE/DEAD assay to visualize apoptosis induced by CRISPR/Cas9 genome editing of ABL in clinical CML cells (Figure 5L). The images showed a significant increase in the number of cells undergoing apoptosis after transduction with ABL sgRNA_2 (21.3%±4.65, apoptosis cell percentage), compared to that of SC (2.1% ± 1.2, apoptosis cell percentage)-transduced CML cells (Figure 5M, p < 0.05), indicating that the ABL-targeted CRISPR/Cas9 virus can be used an effective gene therapy strategy for clinical CML patients.As highly efficient CRISPR/Cas9 gene editing technology has been used in vitro, in vivo and ex vivo, it is critical to determine whether this technology causes unexpected cleavage events at similar DNA sequences or under other circumstances. If so, further analysis of the specificity of CRISPR/Cas9 technology should be undertaken. From all human genes, we chose those with highly similar sequences to the ABL sgRNA_1 and ABL sgRNA_2 sequences as candidates for off-target cleavage and evaluated these genes by Sanger sequencing. The genes with a highly similar DNA sequence to the ABL sgRNA_1 sequence, which might thus be subject to off-target effects, are ADAMTSL1 and TBRG4 (Figure 6A), and the C17orf75 gene shares sequence similarity with ABL sgRNA_2. We designed specific primers for these three genes, amplified the regions by PCR, and analyzed the DNA sequences by Sanger sequencing. No genomic editing occurred in these genes (Figure 6B). This result validates the high specificity of the CRISPR/Cas9 system in targeting the ABL gene, and the superior leukemia therapeutic effects observed both in vitro and in an animal model are anticipated to translate into the clinic in the future.Imatinib (Gleevec) is a chemotherapeutic used to treat Ph-positive CML and ALL and certain types of gastrointestinal stromal tumors, systemic mastocytosis and myelodysplastic syndrome [24]. To investigate the anticancer activity of imatinib and the essential role of ABL in K562 cells in vitro, we determined the IC50 value of imatinib in K562 cells by treating them with different concentrations of imatinib for 48 h and then performing MTT assays (Figure 7A). K562 cells were highly sensitive to imatinib (IC50 = 1.8 μM). In addition, imatinib inhibited K562 cell viability, perhaps by suppressing ERK activation and upregulating the tumor suppressors P21 and P27, eventually resulting in PARP cleavage and apoptosis (Figure 7B). Next, we used a LIVE/DEAD assay to visualize imatinib-induced apoptosis (Figure 7C). The images showed an increasing number of K562 cells undergoing apoptosis with increasing concentrations of imatinib, with significant cell death at 10 and 25 μM imatinib compared to the control (Figure 7D, p < 0.05). In addition, at these concentrations, imatinib significantly inhibited K562 cell viability and cell proliferation by approximately 50% compared to the DMSO control, whereas only sgRNA_2 virus infection had a similar inhibitory effect on K562 cell viability (Figure 7E, p < 0.05). Our recent study showed that imatinib-resistant K562 cells (K562-IR) have a much higher IC50 value for imatinib treatment than wild-type K562 cells, with a 30-fold increase in drug sensitivity [25]. With the same cells, we did not find any mutations in the entire ABL gene through Sanger sequencing, indicating that BCR-ABL amino acid substitutions inside the kinase domain may not be the main cause of disruption in the interaction of imatinib and the tyrosine kinase domain, resulting in a loss of sensitivity to the drug in our long-term imatinib-treated K562-IR cells. With imatinib treatment and ABL-edited virus infection of K562-IR cells, we found that K562-IR cells with the ABL sgRNA_2 virus showed significant cancer cell growth inhibition (Figure 7F, p < 0.05), which was even higher than that with the high concentration (25 μM) of imatinib. Finally, we confirmed that both ABL sgRNA_1 and ABL sgRNA_2 virus introduction into the normal bone marrow cell line HS27A resulted in no difference in cell growth compared to that of SC virus-infected cells (Figure 7G), ensuring the safety of the ABL target virus in vivo. The above evidence indicates that ABL plays an important role in promoting cell growth and preventing apoptosis. CRISPR/Cas9-based gene editing of ABL provides an effective anticancer strategy for imatinib-resistant CML cells.Previous studies showed that ABL gene has its own physiological roles such as regulating T-cell survival and development [26,27], which is import for anti-cancer immune response. To investigate this potential concern, we targeted ABL gene on mouse chromosome 2 to investigate whether ABL disruption through CRISPR/Cas9 affects T-cell survival and development in vivo (Figure 8A). We utilized CRISPR/Cas9 genomic editing by targeting two mouse protospacers as protospacer 1 (mABL sgRNA_1) targets the plus strand, whereas protospacer 2 (mABL sgRNA_2) targets the negative strand. Transduction of NIH-3T3 mouse fibroblast cells with the target scrambled (SC) virus produced a wild-type ABL sequence, as shown by Sanger sequencing (Figure 8B,C), with no evidence of gene editing. However, transduction with the mABL sgRNA_1 virus and the mABL sgRNA_2 virus transfection cells led to major single nucleotide deletions around the predicted cleavage sites (Figure 8D,E). Through TIDE analysis, it is showed that both mABL sgRNA_1 and mABL sgRNA_2 obtained great gene editing efficiencies with 97.4% (Figure 8H) and 98.9% (Figure 8I) of the cell pool, respectively. The most frequent mutations in the mABL sgRNA_1 cell and mABL sgRNA_2 cell pool were 1-bp delection with 86.3% (Figure 8F) and 90.1% (Figure 8G), respectively. In addition, the algorithm predicted the same patterns of genome repair for both mABL sgRNAs, which included mutations mainly at the cleavage sites (Figure S5A,B). Next, we tail-vein injected SC-, mABL sgRNA_1- and mABL sgRNA_2 virus on BALB/c mice. After four weeks, the mice were scarified and the blood samples were collected and hematology analyzed, whereas isolated leukocytes and internal organs were Sanger sequenced for mouse ABL gene disruption. The representing result showed that the mABL sgRNA_1 and mABL sgRNA_2 virus transfected mice obtained great ABL gene editing on leukocytes (Figure S6, File S1), with 55 ± 5.29% and 64.66 ± 7.13% of all leukocyte pool, respectively (Figure 8J). Interestingly, all the ABL gene sequences from internal organs, such as the kidney, liver, lung and spleen remained unedited, either in SC- or mABL sgRNAs introduced mice (Figure S6). This observation suggests that lentivirus as a gene carrier has a short-term but great transfection effect on the circulation cells through tail-vein delivery, this could prevent unwanted host targeting. In Table 1, the hematology analysis of RBC, WBC and platelet cell counts showed no significant differences between SC-, mABL sgRNA_1- and mABL sgRNA_2 targeted CRISPR/Cas9 virus injected mice. Furthermore, the percentage of neutrophil, monocyte and lymphocyte from DIFF scattergram showed no differences among these three groups (Figure S7), indicating mouse ABL gene editing through CRISPR/Cas9 has no significant effect on either blood cell survival or differentiation. To further reveal whether mouse ABL gene editing influences T-cell development, we used flow cytometry to determine CD4 and CD8 expressions of blood purified leukocytes from three groups of mice. Using unstained samples to determine CD4 and CD8 positive regions (CD4+ and CD8+), it is clear shown that the percentages of CD4+CD8+DP (double positive), CD4-CD8-DN (double negative) and CD4+ or CD8+ SP (single positive) T-cells were not significantly different between SC and ABL gene edited mice (Figure 8K). These results suggested that T-cell linage of CD4+ or CD8+ lymphocytes are not effected in the absence of mouse ABL gene.In this study, we aimed to investigate the anticancer efficacy of virus-mediated CRISPR/Cas9 gene therapy targeting the human ABL gene in CML cells in vitro and in vivo (Figure 8). First, the ABL-targeting virus was carefully produced, quantified and optimized for CML cell delivery (Figure 9A). The concentrated ABL-targeting virus was used both in vitro (Figure 9B) and in an in vivo bioluminescence imaging-based systemic leukemia animal model (Figure 9C); mouse systemic leukemia can be considered a preclinical gene therapy model (Figure 9D). Once CML cells are infected with the ABL-targeting virus (Figure 9E), the sgRNA, with the PAM sequence, targets the ABL gene locus (Figure 9F), and the sgRNA loop then carries Cas9 to cleave the nearby DNA. The competing NHEJ pathway for DNA repair is often favored and frequently leads to indels or chromosomal rearrangements, particularly in mammalian cells (Figure 9G). These DNA indels result in frameshifts in the protein-coding sequence, thus producing nonsense proteins or causing early termination of ABL translation, eventually resulting in CML cell death (Figure 9H).In the past few years, CRISPR/Cas9 gene editing technology has become an important strategy for discovering therapeutic targets in human disease. In basic laboratories, most studies continue to use plasmid transfection approaches, such as liposomes or electroporation, to deliver CRISPR/Cas9 and sgRNA for gene editing. One major limitation of those methodologies is that they may alter the physical or biological condition of the target cells, or even cause incidental cell death unrelated to gene targeting. Here, we used lentiviruses as carriers to deliver the CRISPR/Cas9 genomic editing system to achieve improved transfection efficiency and increasingly flexible treatment applications for both in vitro and in vivo experimental designs. Lentivirus use has grown exponentially both in research and in gene therapy protocols, accounting for 12% of the viral vector-based clinical trials in 2011, and the percentage is still increasing [28]. However, the safety of lentivirus use should be the highest priority in the clinic, and many questions remain about the potential harm that lentiviruses may cause. Thus, it will be difficult to balance the inherent potential risks with the potential benefits of abrogating oncogenes with this powerful tool. Therefore, the appropriate lentiviral input should be determined in an animal model of systemic leukemia to avoid unexpected side effects.Ideally, genomic editing will be optimized to enhance on-target efficiency and to reduce off-target efficiency. However, off-target cleavage and other undesirable effects on protein translocation and recombination have considerable potential to hinder CRISPR/Cas9 research and related biotechnology [29]. Off-target effects stem from nonspecific recognition of non-target sequences, which might lead to alterations in protein expression. The major concern is that gene editing-mediated chromosomal rearrangements might disable tumor suppressor genes or activate oncogenes, both of which could contribute to cellular toxicity [30]. Several techniques to measure off-target effects, such as targeted sequencing, exome sequencing, whole genome sequencing and GUIDE-sequencing, are often used. Each method has its own advantages and disadvantages. Targeted sequencing is relatively easy, fast and widely available, although it presents some potential drawbacks. For example, the results of targeted sequencing might be biased if no unexpected mutation sites are detected. Off-target measurements might be expensive and time-consuming if many candidate sites are screened. In this study, we used Sanger sequencing (targeted sequencing) to demonstrate that our developed ABL-targeted therapy had no off-target effects on similar gene sequences in the pool of gene-edited cells, indicating that the ABL-based CRISPR/Cas9 gene editing system is highly specific and would be safe for CML therapy. However, more sensitive off-target detection methods are required, especially for applications such as gene therapy that require absolute fidelity. Further research should use unbiased GUIDE-sequencing (next-generation sequencing) to ensure off-target detection with high sensitivity and high DNA coverage before applying ABL-based CRISPR/Cas9 gene editing in a clinical setting.Due to the success of CRISPR/Cas9 genomic editing, many studies have used this technology to target BCR-ABL fusion genes in different models for different purposes. For example, Lekomtsev and colleagues employed Cas9 with two guide RNAs targeting the two breakpoints in the haploid human cell line eHAP, to determine whether BCR-ABL translocations could be reverted back to the wild-type status. However, after genotyping analysis showed that the overall targeting efficiency of the BCR-ABL translocations was 0.8% of 384 clones, additional spectral karyotyping and G-band karyotyping were used to confirm and visualize the engineered chromosomal translocations [31]. Additionally, Hara and his colleagues investigated the efficiency of both on-target and off-target genome editing by introducing FokI-dCas9 (fCas9), Cas9 D10A (D10A), Cas9 WT (WT) and gRNAs into zygotes to generate mutant mice [32]. The results of their study revealed that the specificity of fCas9 is more strictly regulated than that of other Cas9 forms, enabling the generation of knockout mice with reduced unwanted off-target effects by CRISPR/Cas9 technology.In another study using CRISPR/Cas9 genomic editing of the BCR-ABL fusion gene, García-Tuñón and his group were the first to report the use of CRISPR/Cas9 genome editing to abrogate the human BCR-ABL oncoprotein in leukemia cells as a therapeutic intervention [33]. The authors designed specific sgRNAs to direct Cas9 to the BCR/ABL fusion sequence (junction sequence) in the Boff-p210 cell line, a pro-B-derived Baf/3 hematopoietic cell line that artificially expresses BCR/ABL. The CRISPR/Cas9-mediated reduction in BCR-ABL oncoprotein (p210) expression in Boff-p210 cells resulted in the loss of tumorigenicity in a CML xenograft animal model. However, several forms of BCR-ABL fusion proteins have been identified in the clinic based on three breakpoint regions—major breakpoint cluster region (M-BCR, P210, 210 kDa), minor breakpoint cluster region (m-BCR, P190, 190 kDa) and micro breakpoint cluster region (mu-BCR, P230, 230 kDa)—or, in rare cases, other nearby sites [3,34]. Thus, targeting the junction sequences of BCR/ABL may not be applicable to all clinical CML patients. Therefore, in this study, we utilized a methodology similar to that in previous work. Our study adopted approaches like those used for clinical gene therapy, such as using lentiviruses to deliver CRISPR/Cas9 genes, targeting leukemia cells with the native Ph chromosome, optimizing the ABL genomic editing sequences and evaluating cancer efficacy in an animal model of systemic leukemia. Notably, our study produced a more reliable and flexible ABL-based gene therapy than approaches targeting the BCR-ABL oncogene junction in CML patients. However, although any adverse effect was not observed with CRISPR/Cas9-directed disruption of the ABL gene in mice in this short-term study, it should be noted that the long-term effect of ABL gene disruption is not known. Furthermore, the extent to which ABL gene disruption affects the hematopoietic stem cell population is not yet known. In the future, the genomic editing of ABL described in this study could be improved by addressing the following concerns: (a) primary cultures of leukemia cells from CML patients are required to assess anticancer efficacy; (b) lentivirus transfection may vary between primary leukemia cells and the K562 cell line; (c) a primary culture immortalization system for leukemia cells is essential for future preclinical anticancer evaluation; (d) a systemic leukemia model using primary culture cells would be the best platform to mimic a gene therapy approach targeting ABL or other oncogenes. Nonetheless, to the best of our knowledge, this is the most relevant preclinical study of BCR-ABL-targeted therapy, the efficacy of which stems from virus-mediated genome editing. Such findings are useful for future studies and the optimization of gene therapy in clinical trials.In conclusion, this study established a virus-mediated ABL-targeting gene therapy that significantly reduced Ph expression and abolished leukemia cell survival and tumorigenic abilities in vivo, in vitro and ex vivo. These findings suggest that this CRISPR-Cas9-based gene therapy has strong potential and may be further applied for the treatment of CML patients who are insensitive or resistant to imatinib treatment.The following are available online at https://www.mdpi.com/2072-6694/12/6/1399/s1, Figure S1: Optimization of viral transduction conditions for human K562 cells, Figure S2: The TIDE algorithm analysis of BCR-ABL gene-targeted virus infection, Figure S3: The original TIDE algorithm analysis of ABL gene-targeted virus infection, Figure S4: Cas9 protein production and purification, Figure S5: The TIDE algorithm analysis of mABL gene-targeted virus infection, Figure S6: The representing mouse ABL disruptions of leukocyte and internal organs from SC- and mABL sgRNAs virus introduced mice, Figure S7: Hematology analysis of SC- and mABL sgRNAs virus injected mice. File S1: Complete sequencing results in .ab1 format. File S2: Original whole blot.S.-H.C., Y.-Y.H., H.-E.T. and C.-H.L. designed the research. Y.-S.C. and C.-Y.L. performed the experiments and acquired data. C.Y.-L., K.-W.H. and Y.-S.C. helped design the CRISPR protospacers and advised on the virus production and infection procedures. Y.-S.C. and C.-H.L. helped interpret the results and analyzed gene editing efficiency. Y.-Y.H. and H.-E.T. provided clinical Chronic Myeloid Leukemia patient sample. S.-H.C., S.-M.L., W.-S.H., M.-J.S. and C.-H.L. participated in clinical discussion and WBC differential counting. S.-H.C. and C.-H.L. conceived the study and supervised the project. All authors read and approved the final manuscript.This study was supported by the Ministry of Science and Technology, grant 106-2813-C-038032-B for Miss Chiang and Dr. Lee, grant 108-2635-B-038-0002 for Dr. Chen, grant 108-2628-B-039-003 for Dr. Hsu, Shuang Ho Hospital by the grant 109TMU-SHH-14 for Dr. Chen and Dr. Lee. Meanwhile, this study was financially supported by the “Drug Development Center, China Medical University” from The Featured Areas Research Center Program with in the frame work of the Higher Education Sprout Project by the Ministry of Education (MOE) in Taiwan. None of the funding bodies were involved in the design of the study, the collection, analysis, and interpretation of the data, or the writing of the manuscript.We would like to acknowledge the following for their kind services: Facility Center of Taipei Medical University; Instrument Center and the Laboratory Animal Center at the National Defense Medical Center. This work was financially supported from the Young Scholar Fellowship Program by the Ministry of Science and Technology (MOST) in Taiwan, under Grant MOST 108-2636-B-009-005, and partially supported by the Center for Intelligent Drug Systems and Smart Bio-devices (IDS2B) of the Higher Education Sprout Project by the Ministry of Education (MOE) in Taiwan.Declare conflicts of interest or state “The authors declare no conflict of interest.”Efficient and specific clustered regularly interspaced short palindromic repeats (CRISPR/Cas9) gene editing of the BCR-ABL junction regions in K562 cells. (A) DNA sequence map of the BCR-ABL junctions in K562 cells. The red column indicates the BCR gene, the red column indicates the ABL gene and the black column indicates the shared sequences for both the BCR and ABL genes. (B) Schematic representation of the human BCR-ABL junctions and two protospacer sequences (blue underline) for editing that were designed from the plus (protospacer 1) and negative (protospacer 2) DNA strands. The arrowhead indicates the expected Cas9 cleavage site. The protospacer adjacent motif (PAM, red underline) is the motif required for Cas9 nuclease activity. Scrambled (SC) sgRNA and BCR-ABL sgRNA were delivered to K562 cells by lentivirus. After transduction, DNA from the virus-infected cells was purified and subjected to Sanger sequencing of (C) plus and (D) negative DNA strands of the BCR-ABL junction in the SC K562 cells. (E) BCR-ABL sgRNA_1 and (F) BCR-ABL sgRNA_2 produced a mixture of sequences around the expected Cas9 cleavage site in a pool of gene-edited cells after lentivirus transduction. Tracking of indels by decomposition (TIDE) algorithm analysis of the ABL gene-edited sequences (indels, insertions and deletions) showed a high editing efficiency in K562 cells. The pie charts show the percentages of indels in the ABL gene edited by (G) BCR-ABL sgRNA_1 and (H) BCR-ABL sgRNA_2. The gene editing efficiency of the two sgRNAs is presented in red, while the most common other indels is presented in brown color. (I) Western blot analysis of BCR-ABL and ABL protein expression. Bands at 210 kDa and 145 kDa, corresponding to BCR-ABL and ABL, respectively, were observed in the parental and scrambled (SC) sgRNA control K562 cells, whereas these bands were significantly reduced in the BCR-ABL sgRNA_1- and BCR-ABL sgRNA_2-transduced K562 cells. The protein expression of downstream ABL targets, such as P21 and cleaved PARP (c-PARP), was also activated by BCR-ABL gene redundancy. All the western blotting was measured and quantified by Image J software. (J) Cell viability curve of the BCR-ABL sgRNA_1- and BCR-ABL sgRNA_2-transduced K562 cells determined by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium (MTT) assays. Data are presented as the mean and standard deviation. Data were analyzed with Student’s t-test.ABL gene targeting in K562 cells using the CRISPR/CAS9 system. (A) Schematic representation of the human ABL DNA locus and two protospacer sequences (blue underline) for editing. The arrowhead indicates the expected Cas9 cleavage site. The protospacer adjacent motif (PAM, red underline) is the motif required for Cas9 nuclease activity. Scrambled (SC) sgRNA and ABL sgRNA were delivered to K562 cells by lentivirus. (B,C) After transduction, DNA from the virus-infected cells was purified and subjected to Sanger sequencing of the wild-type ABL DNA locus in K562 cells. (D) ABL sgRNA_1 and (E) ABL sgRNA_2 produced a mixture of sequences around the expected Cas9 cleavage site in a pool of gene-edited cells after lentivirus transduction. TIDE algorithm analysis of the ABL gene-edited sequence (indels, insertions and deletions) showed a high editing efficiency in K562 cells. The pie charts show the percentages of indels in the ABL gene edited by (F) ABL sgRNA_1 and (G) ABL sgRNA_2. The gene editing efficiency of the two sgRNAs is presented in red, while the two most common -1 and +1 indels are presented in green and orange, respectively. The TIDE analysis of indel distribution is shown for (H) the ABL sgRNA_1- and (I) ABL sgRNA_2 virus-transfected K562 cells compared to the SC K562 cells.ABL gene disruption inhibits cancer cell growth and induces apoptosis in K562 cells. (A) The ABL gene in K562 cells was analyzed with RNA-guided engineered nuclease-restriction fragment length polymorphism (RGEN-RFLP) assays to measure the gene editing efficiency. The gel image of ABL gene cleavage induced by addition of specific sgRNA, and Cas9 shows the indel percentage in the gene editing pool. Cleaved DNA fragments are highlighted with an asterisk. (B) Western blot analysis of BCR-ABL and ABL protein expression. Bands at 210 kDa and 145 kDa, corresponding to BCR-ABL and ABL, were observed in the parental and SC sgRNA control K562 cells, whereas these bands were significantly reduced in the ABL sgRNA_1- and ABL sgRNA_2-transduced K562 cells. The protein expression of downstream ABL targets, such as P21 and cleaved PARP (c-PARP), was also activated by ABL gene redundancy. All the western blotting was measured and quantified by Image J software. (C) Cell viability curve of the ABL sgRNA_1- and ABL sgRNA_2-transduced K562 cells determined by MTT assays. (D) Cell proliferation of the ABL sgRNA-transduced K562 cells was determined by bromodeoxyuridine (BrdU) incorporation assays.ABL-targeted CRISPR/Cas9 lentivirus therapy effectively inhibits leukemia cell growth in an animal model. For lentivirus therapy, 5×106 luciferase-labeled K562 cells were injected via the tail vein into severe combined immunodeficient (SCID) mice. The mice were grouped (three mice/group) after three weeks and injected via the tail vein with SC- or ABL-targeting virus at a 100:1 ratio of virus (defined as MOI = 1) to the expected number of K562 cells (5 × 108). Bioluminescence images were taken every two weeks with the mice in the (A) ventral and (B) dorsal positions. Bioluminescence images were analyzed by photon influx, which represents the number of human leukemia cells in the mice. The photon influx of each group of mice is presented and compared in both the (C) ventral and (D) dorsal positions. (E) Peripheral blood was collected from each mouse, and Liu’s stain was used to perform a WBC differential count. Myeloblasts, basophils and neutrophils are shown at 400× and 1000× magnification. (F) The immature WBC cell counts from the mice injected with SC- or ABL-targeting virus were analyzed. (G) DNA was purified from both groups of mice, and the luciferase gene expression was measured. All data were normalized to mouse GUS expression, which was used as a DNA input control. Data are presented as the mean and standard deviation. Data were analyzed with Student’s t-test; all p-values were two-sided. p values less than 0.05 are indicated with an asterisk.Ex vivo ABL-targeted CRISPR/Cas9 lentivirus therapy of CML patients. (A) Karyogram from the products of conception showing the karyotype of 46, XY, t(9;22)(q34;q11.2). (B) Highlight of chromosomes 9 and 22 at the same level of resolution. (C) Interphase fluorescence in situ hybridization (FISH) analysis using probes for the BCR (green) and ABL genes (red) shows an abnormal pattern t(9;22) of the fusion protein (yellow, lower panel) in the CML patient cells compared to the normal cells (separate colors). SC sgRNA and ABL sgRNAs were delivered to the clinical CML cells by lentivirus. After transduction, DNA from the virus-infected cells was purified and subjected to Sanger sequencing for the target sites of (D) ABL sgRNA_1 and (E) ABL sgRNA_2 in the SC sgRNA-transfected CML cells. (F) ABL sgRNA_1 and (G) ABL sgRNA_2 produced a mixture of sequences around the expected Cas9 cleavage site in a pool of gene-edited cells after ABL sgRNA lentivirus transduction. TIDE algorithm analysis of the ABL gene-edited sequence showed a high editing efficiency in clinical CML cells. The pie charts show the percentages of indels in the ABL gene edited by (H) ABL sgRNA_1 and (I) ABL sgRNA_2. The gene editing efficiency of the two sgRNAs is presented in red, while the two most common -1 and +1 indels are presented in green and orange, respectively. The original TIDE algorithm analysis is shown for (J) ABL sgRNA_1- and (K) ABL sgRNA_2 virus-transfected CML cells compared to SC-transfected cells. (L) The LIVE/DEAD cell viability assay was performed after SC sgRNA and ABL sgRNA_2 were delivered to the clinical CML cells by lentivirus. Cells were subjected to viability assays to identify live (green) and dead (red) cells. (M) Cell death after SC sgRNA and ABL sgRNA_2 lentivirus delivery was analyzed for significance. Data were analyzed with Student’s t-test; all p-values were two-sided.Off-target investigation of the ABL-targeted CRISPR/Cas9 system. (A) The CRISPR design website was used to predict off-target candidate genes for both the ABL sgRNA_1 and ABL sgRNA_2 viruses. Similarities are presented as dots, and mismatch sites are indicated by nucleotide substitution. (B) Sanger sequencing of the K562 cells infected with the ABL sgRNA_1 or ABL sgRNA_2 virus was used to examine potential indels in off-target candidate genes.Imatinib inhibits K562 cell survival and induces apoptosis. (A) The IC50 values of the control or imatinib in K562 cells were determined using MTT assays after treatment for 48 h. (B) Imatinib significantly inhibited ERK activation and induced P21, P27 and cleaved PARP (c-PARP) protein expression in a dose-dependent manner, as evidenced by western blot analysis. All the western blotting was measured and quantified by Image J software. (C) The LIVE/DEAD cell viability assay was performed after imatinib treatment of K562 cells for 24 h. Cells were subjected to viability assays to identify live (green) and dead (red) cells at 100× total magnification (D) Cell death after imatinib treatment was analyzed for significance. Comparison of cell viability following imatinib treatment and ABL sgRNA virus infection of (E) K562 cells or (F) K562-IR cells. K562 or K562-IR cells were treated with 1, 10 or 25 μM imatinib or infected with ABL sgRNA_1 and ABL sgRNA_2 virus for 48 h. The cells were analyzed by MTT assays. (G) Cell viability curve of the ABL sgRNA_1- and ABL sgRNA_2-transduced HS27A cells determined by MTT assays. Data are presented as the mean and standard deviation. Data were analyzed with Student’s t-test; all P-values were two-sided. P values less than 0.05 are indicated with an asterisk.Mouse ABL gene targeting through CRISPR/Cas9 to investigate T-cell survival and development in vivo. (A) Schematic representation of the mouse ABL DNA locus and two protospacer sequences (blue underline) for editing. The arrowhead indicates the expected Cas9 cleavage site. Scrambled (SC) and mouse ABL (mABL) sgRNAs were delivered to NIH-3T3 cells by lentivirus. (B,C) After transduction, DNA from the virus-infected cells was purified and subjected to Sanger sequencing of the wild-type mABL DNA locus in NIH-3T3 cells. (D) mABL sgRNA_1 and (E) mABL sgRNA_2 produced a single nucleotide deletion around the expected Cas9 cleavage site in a pool of gene-edited cells after lentivirus transduction. TIDE algorithm analysis of the mABL gene-edited sequence (indels, insertions and deletions) showed a high editing efficiency in NIH-3T3 cells. The bar figures show the indel distribution of the (F) mABL sgRNA_1 and (G) mABL sgRNA_2 virus-transfected NIH-3T3 cells, compared to the SC transfected cells. The pie charts demonstrate the percentages of indels in the mABL gene edited by (H) mABL sgRNA_1 and (I) mABL sgRNA_2. The gene editing efficiency of the two sgRNAs is presented in red, while the most common -1 indel is presented in green color. (J) Gene edit efficiency (indel%) of the leukocytes was compared by SC-(n = 4), mABL sgRNA_1 (n = 3) and mABL sgRNA_2 (n = 3) introduced mice. (K) The representative T-cell subpopulations from three groups of mice were measured for leukocyte CD4 (FITC) and CD8 (PE) expressions by FACS counting. Data are presented as the mean and standard error.Schematic representation of the research strategy in this study. (A) Phoenix cells were used as the host to generate the lentiCRISPR plasmid-based ABL gene-edited virus. The collected lentivirus was purified, concentrated and quantitated by qPCR analysis. The high-quality lentivirus was used (B) in vitro and (C) in in vivo bioluminescence imaging-based animal models. (D) We injected lentivirus targeting bioluminescent human CML cells via the tail vein into the systemic leukemia xenograft mouse model. The leukemia cells were detected by IVIS. (E) Once the ABL gene knockout virus attacks the Ph chromosome in K562 cells, (F) the virus will generate sgRNA targeting the human ABL gene locus, and Cas9 will cleave the DNA. (G) The competing NHEJ pathway for DNA repair frequently creates indels. (H) CML cells eventually die due to the production of nonsense ABL sequences or early termination of the protein due to a frameshift.Cardiac puncture collected blood samples were taken from SC, mABL sgRNA_1 and mABL sgRNA_2 virus tail-vein injected mice and blood count parameters were quantified. No significant was found in these parameters.

Med-MDPI/cancers/cancers-12-06-01400.txt

@@ -0,0 +1 @@
+Extracellular vesicles (EVs) are small membrane vesicles released by all cells and involved in intercellular communication. Importantly, EVs cargo includes nucleic acids, lipids, and proteins constantly transferred between different cell types, contributing to autocrine and paracrine signaling. In recent years, they have been shown to play vital roles, not only in normal biological functions, but also in pathological conditions, such as cancer. In the multistep process of cancer progression, EVs act at different levels, from stimulation of neoplastic transformation, proliferation, promotion of angiogenesis, migration, invasion, and formation of metastatic niches in distant organs, to immune escape and therapy resistance. Moreover, as products of their parental cells, reflecting their genetic signatures and phenotypes, EVs hold great promise as diagnostic and prognostic biomarkers. Importantly, their potential to overcome the current limitations or the present diagnostic procedures has created interest in bladder cancer (BCa). Indeed, cystoscopy is an invasive and costly technique, whereas cytology has poor sensitivity for early staged and low-grade disease. Several urine-based biomarkers for BCa were found to overcome these limitations. Here, we review their potential advantages and downfalls. In addition, recent literature on the potential of EVs to improve BCa management was reviewed and discussed.Urological tumors represent approximately 25% of all human cancers [1]. Bladder cancer (BCa) is the 10th most common, and the 9th cause of death by malignancy worldwide [2]. Aging, ethnicity, and male gender are considered non-modifiable risk factors [2,3,4], but most tumors are derived from acquired environmental exposure to carcinogenic substances. Cigarette smoking is considered the main risk factor [5,6], with an estimated causal association for half of BCa in both genders [7,8,9], whereas occupational exposure accounts for 10–20% of all cases [10]. The worldwide incidence of BCa seems to reflect areas with higher exposure to risk factors [10], which explains why developed countries have a larger number of diagnosed cases [2,10]. As an example, Schistosoma haematobium infection, another known risk factor, is more prevalent in northern and sub-Saharan African countries, where there is a relatively higher incidence of BCa [10]. In addition, differences in healthcare systems also account for disparity of incidence rates, being better resources associated to an easier and faster diagnosis. Urothelial cancer originates in the epithelial cells of the urothelium, extending from the renal pelvis to the urethra [11,12,13]. The majority of these tumors are located in the bladder, accounting for 90–95% of cases, while 5–10% are located in the upper urinary tract (UUT) [14,15,16,17,18]. Tumor extension is classified according to the TNM (Tumor-Node-Metastasis) staging system. At diagnosis, approximately 75–80% of bladder tumors are non-muscle invasive (NMIBC), which includes mucosa (for stages Ta and Cis) and lamina propria (T1 stage) confined disease, while 20–25% are muscle-invasive (MIBC), when invading the muscle layer and beyond (T2–T4 stages) [1,4,14].Although clinical presentation may be suggestive, the gold standard diagnostic procedures are cystoscopy and urinary cytology [19,20,21,22,23,24]. Nevertheless, cystoscopy is an invasive, operator-dependent procedure, with low sensitivity for small papillary or Cis tumors, which, if underdiagnosed and untreated, progress to muscle-invasive disease in approximately half of the patients [19,20,21,22,23,24]. The sensitivity and specificity of white light cystoscopy is 71% (95% CI: 0.49–0.93%) and 72% (95% CI: 47–96%), respectively [24]. However, due to its invasiveness, it is frequently associated with side effects such as dysuria (50%), hematuria (19%), or urinary tract infection (3%) [25,26].As for urinary cytology, it has high diagnostic accuracy for high grade lesions and Cis, with a sensitivity of 80–90% and specificity between 98% and 100% [27]. However, it exhibits low sensitivity for low grade lesions, between 4% and 31% [28,29,30,31,32,33], and high rate of false positives, due to benign or inflammatory conditions produced by chemo or radiation therapy [34,35]. To overcome these limitations, several urinary biomarkers were developed and are currently commercially available. Compared to cytology, they have higher sensitivity but lower specificity and are, unfortunately, less useful in low risk BCa [36,37,38]. Therefore, consensus among the different international societies on these biomarkers still do not recommend them as replacements of cytology in the current clinical practice [36,37,38].The standard therapy for NMIBC is trans-urethral resection of the bladder (TURB), with both diagnostic and therapeutic purposes, complemented or not by intravesical adjuvant treatment [39,40]. However, even after complete endoscopic resection, there is a high recurrence rate, around 50–70%, and 10–30% will progress to MIBC [39,40]. This feature of BCa natural history elicits the need for a regular follow-up with cystoscopy and cytology at every 3 months interval, generally accompanied by repeated treatments due to recurrence, and which frequently result in high morbidity and economic burden [1,41,42].Driven by the invasiveness and morbidity of cystoscopy, the lack of acceptable sensitivity of urinary cytology and of specificity of the commercially available urinary diagnostic biomarkers, urge the need for extensive research on the identification of novel and more effective biomarkers, to implement better tools for diagnosis, follow-up, and screening of at risk populations [1,29,34,42,43,44].Extracellular vesicles (EVs) are small membrane vesicles which have emerged as a source of biomarkers in bladder cancer [45]. Their detection in liquid biopsies is feasible, due to their presence and stability in most human fluids, and may serve as biomarkers in bladder cancer early detection as they present similar cargo to their donor cancer cells [46]. Additionally, they have some advantages as a source of biomarkers since they are more abundant in liquid biopsies compared to circulating tumor cells (CTCs), protect their cargo against degradation and may carry molecular signatures associated with specific phenotypes [47,48,49]. The present review focus on the status of urinary biomarkers in diagnosis and follow-up of bladder cancer, pinpointing the emerging potential role of urinary EVs on bladder cancer diagnosis and management.The ideal biomarker would be cost-effective, objective, fast to process, and easy to interpret, with high sensitivity and specificity [43,50,51,52,53,54]. For urothelial cancer biomarkers, four goals have been proposed to be accomplished: (i) reduce the need for frequent invasive procedures; (ii) exclude recurrence; (iii) detect progression towards invasive disease; (iv) predict effective treatment response [43,44]. The close contact with urothelium makes urine an attractive approach to detect the presence of tumor cells, in a minimally invasive way. Importantly, this liquid biopsy approach would allow multiple longitudinal sampling of the tumor to identify presence of malignancy, grade, and genomic landscape for improved clinical follow-up. [53,54].Following this line of thought, previous research for BCa biomarkers has been conducted using mostly proteins, nucleic acids, inflammatory and metabolite markers, within the concept of liquid biopsies [55,56]. Taking into consideration that such biopsies concern the detection of any kind of molecular or cellular biomarkers in patient bodily fluids (including urine, blood, saliva, pleural, peritoneal, or cerebrospinal fluids), a novel biomarkers array emerged. These include circulating tumor cells (CTCs), proteins, metabolites, circulating nucleic acids, namely cell-free tumor DNA (ctDNA), messenger RNA (mRNA), micro RNA (miRNA), or long non-coding RNA (lncRNA). Most of these biomarkers may be found free or within extracellular vesicles (EVs) shed by tumor cells or by other elements of the tumor microenvironment [56,57] (Figure 1). There is a growing interest on the liquid biopsy concept, since (i) the biomarkers found have extensive potential for diagnosis and monitoring of disease stage and recurrence; (ii) prediction of therapeutic response/resistance and disease prognosis, with minimally invasive procedures, and (iii) helping therapeutic clinical reasoning based on identified molecular changes [56,57,58].Several interesting and promising biomarkers have been under clinical scrutiny during the past years, although only those approved by in vitro diagnostics (IVD) regulatory entities (e.g., FDA) became commercially available biomarkers, to be used as adjuncts to cystoscopy in primary diagnosis and follow-up of BCa. Taking into consideration the various reports on the subject, novel urinary biomarkers contributed to higher sensitivity but lower specificity than cytology, leaving them out of international guidelines recommendations [1,36,53,54,59,60,61,62]. Table 1 provides an overview of the biomarkers approved for clinical use and their reported diagnostic accuracy.The bladder tumor antigen (BTA) is a complement factor H related protein secreted by malignant cells, which confers them survival advantage, as it interferes in the complement cascade [79]. There are two approved versions of this test for BCa follow-up in concurrent use with cystoscopy, the BTA TRAK and the BTA Stat (Polymedco Inc., Cortlandt Manor, New York, USA) [63]. In different reviews and meta-analysis, the BTA Stat has a sensitivity and specificity of 64% and 77%, respectively, whereas the BTA Trak has 65% and 74%, respectively [64,65,66]. The sensitivity was higher in the diagnosis of symptomatic patients rather than in follow-up, but with similar specificity. Both tests demonstrated higher sensitivity than urinary cytology, despite the decreased specificity in conditions where the complement factor H related protein is present, such as in other genitourinary malignancies and benign conditions with hematuria, including lithiasis, inflammation, instrumentation, and intra-vesical therapies [31,64,65,66,80].The nuclear matrix has an important role on DNA replication and RNA transcription and splicing [81], with nuclear matrix proteins (NMP) being essential components of mitosis, with a role in tumoral proliferation. Numerous NMPs have been described in solid tumors, although NMP22 was shown to be specific for BCa [81,82]. It is released from apoptotic cells towards urine, with significantly higher release rate in cancer than in normal cells [81,83,84]. The NMP22 BC test kit (Matritech Inc.; Newton, MA, USA) is a quantitative test used for patient follow-up, whereas the NMP22 BladderChek test® (Matritech Inc.; Newton, MA, USA) is qualitative and used for both follow-up and initial diagnosis, in symptomatic patients [85,86,87]. Concerning sensitivity and specificity, the quantitative test has 69% and 77%, while the qualitative has 58% and 88%, respectively [64,67,68,69,88,89]. When compared to urinary cytology, the sensitivity of NMP22 was higher (70% versus 40%), albeit specificity was lower (81% versus 97%) [28]. Taken together, both NMP22 and cytology, resulted in sensitivity of 91% [28,89]. Notably, Grossman et al. studied approximately 2000 patients, to compare NMP22 Bladder Chek test® with cystoscopy, and observed decreased NMP22 sensitivity (50–56%) in comparison to cystoscopy (89–91%), although diagnostic accuracy was 94–99% if both tests were considered together [90]. Although NMP22 has higher sensitivity than urinary cytology, specificity is too low to replace it. The fact that it is released from apoptotic cells, which might also be seen in benign conditions, is responsible for the relatively high false positive rate [91]. However, if combined with cistoscopy, this significantly increases its diagnostic value.The ImmunoCyt™/uCyt+™ test (Diagnocure Inc, Quebec, Canada) combines cytology with monoclonal antibody immunofluorescence labelling to detect three BCa antigens, M344, LDQ10, and 19A11, specifically found in malignant exfoliated urothelial cells [92]. To be positive, it requires many exfoliated cells (>500 per field). This test is expensive, with inter-observer variation and time-consuming analysis, but less prone to be influenced by benign inflammatory conditions, comparatively to other tests [93,94]. It is recommended in BCa patients only for follow-up as adjunct test to urinary cytology [95]. Sensitivity varies between 83% and 85% and specificity between 75% and 87%. These are higher in primary diagnosis than follow-up [28,64,65,70]. Mowatt et al. [28] compared uCyt+™ with urinary cytology and showed that this test presented higher sensitivity (82% versus 44%) and lower specificity (85% versus 94%), respectively. Interestingly, the simultaneous use of both tests improved sensitivity without impacting on specificity (87% and 68%, respectively). Sensitivity and specificity, in the study of Schmitz-Dräger et al. [96], was 85% and 88% for immunocytology and 84% and 98% for cystoscopy. When combined, sensitivity increased to 100%, whereas specificity decreased to 87%. Although less prone to interference, immunoCyt™ has lower specificity than urinary cytology. Likewise, despite combination with cystoscopy increases sensitivity, the false positive rate remains elevated [96]. Pfister et al. [97] assert that due to its good sensitivity, the combined use of uCyt+™ with cytology might delay the time intervals between cystoscopies, particularly in lower risk patients. Currently, this test was approved only for patient follow-up [95].UroVysion (Abbott Laboratories, Abbott Park, Illinois, USA) is a fluorescence in situ hybridization (FISH) probe set to detect bladder cancer cells [95,98]. It uses fluorescent labelled DNA probes to assess genetic changes in exfoliated cells, namely chromosomal aberrations suggestive of BCa, aneuploidy of chromosomes 3, 7, and 17, and loss of the 9p21 locus. It has been approved for primary diagnosis and follow-up of BCa patients [95,98]. The reported sensitivity is 63–72% and specificity 85–87% [28,64]. Their diagnostic accuracy was superior in primary diagnosis than in follow-up, showing low sensitivity, similarly to urinary cytology, particularly for low grade tumors [71,72]. Compared to cytology, UroVysion had better sensitivity (72% vs. 42%) and lower specificity (83% vs. 96%) [71]. When used simultaneously, there was a significant improvement in sensitivity but still a low specificity of 50% [72,73]. UroVysion™ is more expensive than cytology and requires specialized laboratory techniques. However, it could be useful in situations of atypical cytology and equivocal cystoscopy, identifying patients that may need further investigation [62,97]. Two prospective studies found that UroVysion had high positive predictive value, supporting that patients with a positive test and negative cystoscopy are more likely to have disease recurrence within one year [99,100,101]. Thus, a FISH test that is positive may be used to anticipate BCa recurrence during follow-up, especially in low risk patients [99,100], and reduce the number of unnecessary bladder biopsies [102]. Therefore, these studies suggest that chromosomic aberrations precede the detection of malignant lesions by cystoscopy and other standard techniques [101].Analyses comparing the above-mentioned biomarkers have been reported. No differences were found in terms of sensitivity and specificity between the NMP22 test kit (cut-off > 10 U/mL) and the BTA Stat, in different stages and tumor grades [91,103,104,105,106,107,108]. The ImmunoCyt™ has higher sensitivity for low stage (Ta, T1) and low-grade tumors, although lower specificity than the UroVysion™ test [107,109,110]. However, although these tests were FDA approved for diagnosis and follow-up of BCa, together with standard techniques, most of these studies are case–control ones in populations with high prevalence of the disease, giving them an unrealistically high positive predictive value. On the other hand, the question remains how to interpret positive findings of these tests, when no significant findings are found on cystoscopy during follow-up. In fact, most positive results have not been submitted to confirmatory biopsy. Moreover, there are few external validation studies to support their use in daily practice. In summary, multicentric prospective studies are required to assess consequences from positive and negative tests in the long term, to increase the likelihood to be supported by international urology organizations. To overcome the limitations of approved diagnostics biomarkers, extensive research is ongoing to find more effective biomarkers for BCa diagnosis and follow-up. There are several commercially available tests, despite not being approved by regulatory institutions. The CxBladder (Pacific Edge Diagnostics, Dunedin, New Zeland) is a RTqPCR test in voided urine, that quantifies different mRNAs expressed in BCa, as IGFBP5, HOXA13, MDK, CDK1, and CXCR2, associated with non-malignant conditions, to reduce the number of false-positives results due to inflammation [75,111]. The Triage™, Detect™, and Monitor™ tests have specific population targets. The first was developed for screening of high-risk populations as a pre-test guiding the need for cystoscopy, the Cxbladder Detect™ was intended for aiding in diagnosis of symptomatic patients and the Monitor™ for BCa follow-up [75]. Studies using Cxbladder Detect™ found higher sensitivity but lower specificity than cytology (73.6% sensitivity and 81.7% specificity) in one study [74], while another described 82% sensitivity and similar specificity [112]. There are reports for Cxbladder Monitor™ stating a sensitivity of 93% that increases to 95% in high risk patients [75]. A large study comparing biomarkers performance for BCa detection in urine, found that the Cxbladder Monitor™ sensitivity (91%) overcomed cytology by 22%, NMP22 BC test kit® by 26% and NMP22 BladderChek® by 11%, with an estimated reduction in the number of cystoscopies needed in follow-up by 81.7% [38]. Although prospective confirmatory trials are needed, some authors suggest its use as an auxiliary test to postpone the need of repeated cystoscopies in low risk patients [38,54]. The Assure MDx™ (MDx Health, Irvine, CA, USA) is a test performed in urine to identify DNA mutations in three genes (FGFR3, TERT, and HRAS) and methylations in another three genes (OTX1, ONECUT2, and TWIST1) [113]. A multicentric study demonstrated a sensitivity of 93% and specificity of 86% for BCa diagnosis [76]. It might be useful for screening low risk patients with symptomatic hematuria, reducing by an estimated 77% the number of unnecessary diagnostic cystoscopies. The XPert® Bladder Cancer (BC) Monitor (Cepheid, Sunnyvale, CA, USA) is a RT-PCR test that measures the number of urinary transcripts in five genes, UPK1B, IGF2, CRH, ANXA10, and ABL1, and was designed for BCa patient follow-up [77]. This test was superior to cytology on NMIBC during follow-up, in terms of sensitivity (84% versus 33%), while presenting similar specificity (91% versus 94%) [77], despite controversial findings from another study, that indicated 46.7% sensitivity and 77% specificity [114]. The heterogeneity between studies and the lack of external validation makes its present use unreliable. The UBC® (Urinary bladder cancer) is a test that detects the expression of cytokeratins 8 and 18 in urine, with presentation of quantitative, UBC®-ELISA, and qualitative UBC®-rapid procedures [78]. The reported sensitivity for UBC®-rapid was 86.9% for detecting Cis, 30.4% for low grade NMIBC, 71.4% for high grade NMBC, and 60% for MIBC [78]. Other studies reported sensitivities between 30% and 87% for Cis and specificity of 63–91% [115,116,117,118]. The UBC®-rapid is a test that provides results within 10 min, but in comparison with other tests has the lowest specificity [115]. Besides these commercially available diagnostic tools for BCa detection in the urine, extensive research is underway to find more effective biomarkers [60,86,119]. The insufficient number of patients in most studies, the lack of external validation in large scale prospective studies, and absence of comparative trials between biomarkers, foster the need for both methodological improvement of existing biomarkers and uncover novel robust biomarkers. Moreover, the existing biomarkers, in general, perform poorly in low risk BCa or have low specificity, and are more accurate in the initial diagnosis of BCa than in follow-up [66]. Taken together, these limitations preclude actual recommendations by most international clinical societies, and current literature suggests that single biomarkers are insufficient to overcome this problem. Therefore, the current trends of research are focusing on the combination of different biomarker signatures, to develop more accurate diagnostic and surveillance tools in BCa, as well as to predict its behavior in order to provide prognostic information [120].Recently, tumor-derived extracellular vesicles (EVs) have received considerable interest by the biomarker research community for BCa diagnosis and follow-up. EVs are non-replicable small lipid bilayer membrane vesicles continuously released by all prokaryotic and eukaryotic cells to the extracellular surroundings. Importantly, this mechanism allows cells to exchange information (encoded in nucleic acids or proteins) between donor and target cells [121].Depending on their biogenesis mechanism and secretion, EVs are broadly divided in exosomes, microvesicles, and apoptotic bodies (Figure 2). Exosomes are the generally smallest vesicles (30–150 nm) and originate by inward budding of intracellular endosomes, later converted into multivesicular bodies (MVBs) that fuse with the cellular membrane and release their cargo into the extracellular space. Instead, microvesicles are typically larger (100–1000 nm) and shed directly from the bleebing of the outward cellular membrane [46,121,122,123]. Apoptotic bodies (1000–5000 nm) are produced by cells undergoing programmed cell death. The processes of synthesis and release of EVs are regulated by endosomal sorting complexes required for transport (ESCRT), p53/TSAP6 pathway, syndecan-syntenin-ALIX, Rab proteins, phospholipase D, sphingomyelinase, and ceramide [124]. To date, it is not possible to experimentally support the attribution of certain activities or markers to specific EV subtypes, which prompted the International Society for Extracellular Vesicles (ISEV: www.isev.org/) to publish guidelines on EVs nomenclature and characterization. The current recommendation is to report all EV subtypes generically as “extracellular vesicles”, while describing in detail the mechanisms used for their separation and characterization, their physical characteristics, biochemical composition, or descriptions of the cell of origin, unless their biogenesis pathway is confirmed [124,125,126].EVs behave ubiquitously, and have been identified in most body fluids, including blood, saliva, breast milk, urine, and amniotic fluid [125,126,127,128,129,130,131,132]. It is known that they carry a cargo from their donor cell, primarily composed of proteins, mRNAs, miRNAs, lncRNAs, small DNA fragments, and lipids. Indeed, it was postulated that EVs may reflect the biological functions of the originating cells [46,123,133], even though they were thought to be biologically insignificant or a simple vehicle for cellular waste disposal. In recent years, EVs were recognized as having physiological and pathological relevance [134]. Extracellular vesicles are key intervenients in several processes involved in cellular homeostasis [134]. Besides their physiological role on cell survival and anti-stress protection, they have a main function in intercellular communication, transporting key molecular messengers to recipient cells, thereby influencing the recipient cells function. Packaging this information into vesicles provides additional protection to the messengers (i.e., cargo) and allows simultaneous delivery to remote locations, which might be achieved through distinct mechanisms (Figure 3): (i) transfer of nucleic acids that induce phenotypic changes and affect multiple functions in recipient cells; (ii) transfer of lipids and proteins (such as cytokines, chemokines, and growth factors) to neighboring or distant cells, thus modulating the targeted recipient cells; (iii) trigger cell signaling pathways in recipient cells by exposure of ligands, proteins, and lipids, that bind to and stimulate matched receptors at the cell surface; (iv) transfer of functional receptors to recipient cells, thus allowing cell signaling in recipient cells that originally lacked that receptor or enhancing their number [47,48,49].Extracellular vesicles have been involved in thrombosis and hemostasis, as they contribute towards increased availability of negatively charged phosphatidylserine and tissue factor, thus providing the triggers for activation of the extrinsic pathway and subsequent thrombin generation, promoting pro-thrombotic effects and platelet aggregation [48]. Interestingly, bladder cancer patients are at increased risk for venous thromboembolism [135], an event associated with the worst prognosis. Similarly, the crosstalk between blood and endothelial cells mediated by EVs expose cell surface receptors and adhesion molecules, required for angiogenesis and neovascularization, of relevance upon tissue injury, post-ischemic revascularization, and regeneration [47,48,49]. A pleiotrophic effect of EVs on immunoinflammatory processes has also been described. Dendritic cell-derived EVs enhance natural killer cell cytotoxic activity and stimulate epithelial cells to release proinflammatory cytokines [49]. Others, derived from circulating leukocytes, participate in endothelium activation and upregulate the release of adhesion molecules, leading to leukocyte recruitment; EVs may also have antigen-presenting properties, exposing major histocompatibility complexes (MHCs) and co-stimulatory molecules on cell surface, initiating a pro-inflammatory response in epithelial cells and T-cell activation. Dendritic cell-derived microvesicles containing tumor necrosis factor-α (TNF-α) might initiate an innate immune response in epithelial cells and affect adaptive immunity, whereas platelet-derived microvesicles can increase B-cell production of immunoglobulins and activation of the complement system [47,48,49].Interestingly, the recognized function of EVs as critical effectors in the maintenance of physiological cell-to-cell interactions seems to be severely disturbed throughout cancer progression. Indeed, the transfer of pro-tumoral EVs cargo between cancer cells and the surrounding tumor microenvironment influences the multiple stages of tumorigenesis, namely neoplastic transformation, proliferation, migration, invasion, and metastasis to distant organs, angiogenesis, immune response, and emergence of drug resistant traits in cancer cells [47,121,136] (Figure 3). Their potential for tumor initiation was shown in prostate cancer, where EVs isolated from tumor cells were able to upregulate the pro-survival protein STAT3, converting normal into malignant epithelial cells [137]. Similarly, in other cancer models, BRCA1-KO fibroblasts treated with sera (containing EVs) from cancer patients yielded higher proliferation and malignant transformation than wild type control fibroblasts [138]. This was demonstrated in BCa as well, as healthy recipient fibroblasts gained malignant phenotypes and transformed into cancer associated fibroblasts (SMA, FAP, galectin), after exposure to cancer cell-derived EVs [139]. EVs were also able to promote tumor cell proliferation in several cancers, including BCa. When derived from cells under hypoxic conditions, EVs carry high levels of lncRNA-urothelial cancer associated 1 (lncRNA-UCA1), which promotes proliferation and invasion in human recipient cells [140]. Their ability to induce migration, invasion, and angiogenesis has also been demonstrated [141]. In fact, tumor-derived EVs can promote angiogenesis by supporting communication between cancer and endothelial cells. Indeed, extracellular vesicles isolated from high grade bladder cancer cells, and from the urine of patients with high grade bladder cancer, contained EDIL-3 and increased angiogenesis and migration of bladder cancer and endothelial cells. Indeed, when EVs originate from high grade bladder cancer cells isolated from the urine of patients submitted to radical cystectomy, EVs contain substantially higher levels of EDIL-3, a pro-angiogenesis and migration protein, than healthy controls. Following EDIL-3 knock-down in bladder cancer cells, the collected EVs had lower EDIL-3 levels, and were unable to promote angiogenesis and migration [141]. Likewise, the EVs isolated from MIBC cells had increased levels of periostin, which was capable of activating ERK oncogenic signals. This resulted in increased aggressiveness of low-grade tumor cells and was associated with worse prognosis, in both MIBC cell lines and clinical tissue samples [142]. On the other hand, Franzen et al. [143] demonstrated the ability of MIBC derived EVs to induce epithelial to mesenchymal transition, a well-known mechanism for initiation of metastasis and cancer progression. In another study, Ostenfeld et al. reported that altered secretion of EVs containing tumor suppressive miRNAs, like the miR-23b, regulated by members of the RAB family, namely the RAB27A and RAB27B, contributed to invasion, anoikis, angiogenesis, and pulmonary metastasis in BCa patients [144]. The increased amount of mucin-1 (MUC1) and epidermal growth factor (EGF) receptor HER3 in EVs were shown to be associated with a more favorable prognosis [145]. The EVs have a prominent role in damping the hosts immune cell response to the emerging cancer cells [146]. Indeed, their role in immunosuppression has been well established in several cancer types [146], once they may cooperate with cancer cells to overcome immune checkpoints [146]. Nevertheless, the specific contribution of EVs for immunosuppression of host response in BCa remains poorly understood. Currently, there is limited consensus on the standard pre-analytical procedures for the clinical validation of EV-based diagnostics in bladder cancer [147,148]. Some of these pre-analytical factors include the standardization of procedures for urine selection, collection, storage, and shipping/transportation conditions. Noteworthy, many of these factors have a direct impact on the co-elution and polymerization of Tamm–Horsfall protein (THP), one of the major contaminants of urinary EV separation. Indeed, the (i) timing of urine collection (e.g., first morning urine vs spontaneous urine vs intravesical urine); (ii) the need for inter-day urine collection due to sample variation; (iii) the necessity for stabilization of urine pH; (iv) the type and impact of THP inhibitor cocktails added to the urine (e.g., reducing dithiothreitol (DTT), detergent CHAPS or urea); (v) its storage temperature (e.g., 4 °C vs 20 °C); and (vi) maximum storage time prior EV separation are some of the unresolved issues required to improve EV-based diagnostic accuracy. So far, the consensus seems to prefer first morning urine for collection, with some reports stating that a citrate-based buffer is beneficial for controlling urine pH while reducing THP protein precipitates after thawing [148]. Nevertheless, such pre-analytical interventions may affect the size and even the native composition of isolated EVs [131,149,150]. Additionally, the high variation of EV recovery due to inter-day urine collection seems to be hampered by the addition of protease inhibitors [148]. Unfortunately, the optimal inhibitor cocktail is still an open debate with several authors showing that different protease inhibitor cocktails had different impacts on the ratio of THP polymerization and consequently on the yield of urinary EVs in the UC pellets [148,151,152,153]. Regarding urine storage/transportation, 4 h seems to be the maximal time interval upon urine collection to minimize EV sample degradation. Importantly, this time interval seems to be heavily reliant on the storage temperature and type of analyte (e.g., nucleic acids, proteins, lipds, etc.) [154]. Taken together, several studies show that several pre-separation methodological issues have a remarkable influence on the yield, purity, and cargo profile of EVs isolated from urine samples [148,154,155]. Importantly, this impact is transversal to all studies regardless of the selected EV separation method. Indeed, the standardization of pre-analytical variables is required to ensure a reliable evaluation on the reported quantity and EV profile for a given pathological state. Moreover, the development of shared Standard Operating Procedures (SOPs) would enable the comparison (and even merging the data) of urine EVs derived from bladder cancer patients from different laboratories. This would facilitate the establishment of multicentric clinical trials to validate the clinical feasibility of EV-based biomarkers. Considering this, the International Society of Extracellular Vesicles (ISEV) has been supporting several initiatives to favor the development of such SOPs [156]. Some of these initiatives includes the Minimal Information for reporting EV-related research [124], the EV TRACK: EV Transparent Reporting and Centralizing Knowledge [157], and the Clinical Wrap-Up session at ISEV2018 [158]. Regarding EV separation from urine samples, several technologies can be used for this purpose. Some of the conventional methodologies include differential ultracentrifugation (dUC) [159], density gradient ultracentrifugation (gUC) [160], chemical precipitation [161], affinity capture [162], hydrostatic filtration dialysis [163,164], ultrafiltration (UF) [155,165,166], and size exclusion chromatography (SEC) [167]. With no perfect solution in sight, the selection of the optimal EV separation method typically lies on the compromise between urine EV recovery yields, EV integrity, and purity [124,168]. Within this reality, dUC remains the most used methodology despite having a low recovery yield of EVs (1–5%) from urine samples. Moreover, dUC has a reported co-precipitation of contaminants (e.g., urine proteins, cell membrane debris, etc.) with the pellet EVs. This issue may compromise a reliable proteomic approach for biomarker discovery [160]. Even more, gUC have a reported EV recovery yield of nearly 30% from crude urine samples. However, both ultracentrifugation-based approaches have a timely and laborious nature which may compromise its generalization for some clinical applications.Facing this bottleneck issue, several authors have pursued other approaches for isolating highly pure EVs from urine samples. In this regard, several two-step combination methods (e.g., UF combined with SEC or asymmetrical-flow field-flow fractionation) have been pursued to minimize the impact of urine contaminants on the purity of isolated EV sample [165,169,170,171]. Unfortunately, despite having higher EV recovery yields (up to 60%) these techniques fail to provide the same reliable EV proteomic analysis obtained via gUC gold standard [165,171]. Direct comparisons between the different pre-analytical factors and EV separation methods on the observed results remain difficult to perform. This is mostly due to inter-laboratory variability, lack of widely accepted EV-marker normalization methodologies, backed up by the limited biological sample availability to test all these conditions in a single experiment. Recently, the use of “spike-in” fluorescent EVs standards were proposed to monitor the efficacy of pre-analytical methods and EV separation procedures [172]. Their use could enable data normalization across laboratories and a deeper comprehension on the causes of variability in the urinary EV cargo. This would increase the detection rate of artefacts originated by technical variation (e.g., sample preparation and instrumentation). Indeed, the use of such internal EV standards would facilitate the establishment of consensual and evidence-based SOPs for optimal collection, storage, and handling of urine EVs [148]. As far as new urine EV separation methods are concerned, microfluidic [173,174,175,176] and/or nanofiltration [177] miniaturized systems have recently emerged as promising approaches. These technologies separate efficiently EVs from other urine components based either on acoustic trapping [174], lateral fluidics displacement [175], immunocapture [178] and/or physical entrapment by nanowires technology [176] or double filtration meshes [173,177]. These new technologies have the advantage of enabling easy and rapid sample processing (e.g., less than 30 min) with minute amounts of urine sample (up to 1 mL), compatible with an -omics approach. These are some of the critical features of future point-of-care devices intended for clinical use. The EV-based liquid biopsy concept for bladder cancer diagnosis relies on the detection of rare EV-subsets (shed by bladder cancer cells) in the pool of isolated urine EVs (derived from virtually all cells of the body). In most cases, only small amounts of clinical samples are available, rare molecular targets have to be detected in complex biological fluids with high specificity and sensitivity in a timely fashion. Indeed, most of traditional methodologies (NTA, TEM, WB, etc.) for EV analysis become obsolete and fail to provide such diagnostic detail under these strict clinical requirements [156].To fulfill such requirements, innovative optical [179,180,181], nano-flow cytometry [182,183,184,185], Raman [164,186,187] and other plasmonic sensors methods [188,189,190,191] have recently emerged for highly sensitive single-EV detection. Nevertheless, their application has not been applied to urine samples and the analysis of single-EVs and other submicron particles has presented many challenges and has produced a few controversial results in other types of samples. Thus, consortium-based efforts are being currently implemented. This will allow a combined effort for technique optimization for EV detection, definition of data reporting criteria, and finally to forge consensual international guidelines for each technique prior clinical application [192].The role of EVs in the maintenance of tissue homeostasis, and the demonstration of their disruptive role during cancer progression and metastization, render them as an attractive source for diagnostic biomarker research, particularly in bladder cancer. Indeed, EVs have several advantages as source of cancer biomarkers. Firstly, some studies suggest that EVs secretion by tumor cells may be higher than by non-tumor cells, even though this still needs to be proved since other studies failed to demonstrate such association [193]. Secondly, the EVs presence and stability in large quantities in most human body fluids, being more abundant in liquid biopsies than CTCs, makes them easily attainable for non-invasive collection and their detection is technically feasible [45,194,195]. Thirdly, their cargo reflects the biological behavior and composition of their donor cells and, thereby, they may carry molecular signatures associated with specific phenotypes or therapeutic resistance patterns [45,46,194,195]. Finally, the lipid bilayer membrane protects their cargo against degradation [45,130,195]. EVs recently emerged as a source of biomarkers in cancer diagnosis and management [45,195,196,197,198,199,200]. EVs isolated from urine and blood have shown specific miRNA, mRNA, and protein content in different types of solid tumors [201,202,203,204,205,206]. Notably, research and knowledge on EVs is now expanding to other diseases such as hepatitis C, chronic kidney disease, and central nervous system and cardiac diseases [207,208,209,210].Using EVs as diagnostic tool in bladder cancer remains elusive, however, a large body of evidence is now accumulating and demonstrating their potential as a source of biomarkers for non-invasive diagnosis of BCa. Research has been conducted on protein and genetic content of EVs from patients with bladder cancer, providing a library for future biomarker identification [196,197,198,199,200] (Table 2). Although plasma or serum can be used, urine is the preferred body fluid for EV collection, due to its availability, low invasive procedure, and its physical contact with the bladder tumor cells. Recently, a high number of EVs were detected in the urine of BCa patients when compared to healthy controls, using a newly developed double-filtration microfluidic system as a point-of-care diagnostic device, which displayed a sensitivity of 81% and a specificity of 90% [174]. The characterization of genomics in urinary EVs, contributed towards the growing interest in its RNA content. Indeed, miRNA and lncRNA are small non-coding RNAs that regulate the expression of protein-coding genes involved in several cellular processes, including tumor development and progression [211]. Their presence in urine and other body fluids, either in their cell free form or as part of EVs, makes them interesting sources of tumor marker research with diagnostic, treatment, and prognostic objectives in bladder BCa [212]. For this purpose, Perez et al. [213] compared the urinary EV transcriptome from BCa patients and healthy controls, by performing PCR analyses of 15 genes with differential expression between both groups. The authors described four genes differently expressed in urinary EVs, where GALNT1 and LASS2 were specific of cancer patients, and the ARHGEF39 and FOXO3 transcripts were detected only in healthy controls. Other studies analyzed lncRNAs isolated from EVs, finding different genetic patterns in patients with MIBC in comparison to normal controls, particularly the HOX transcript antisense RNA (HOTAIR), implicated in tumor initiation and progression [214,215]. The interest in this panel of genetic biomarkers was later confirmed in another study that showed association to disease recurrence and poor prognosis [216]. Others studied the miRNA content of urinary EVs from BCa patients [217,218,219,220,221] and found miRNA signatures characteristic of high-grade BCa [216,220], which can be suggested as biomarkers of advanced disease. Interestingly, a study detailed one miRNA (miR-21-5p) overexpressed in urinary EVs of BCa patients with 75.0% sensitivity and 95.8% specificity for detecting disease, which was still present despite negative urinary cytology, suggesting it might detect BCa at an earlier stage of the disease, with no cytological changes [220]. Although most of the genomic research includes RNA, EVs may also be a source of tumor DNA. Lee et al. compared the genomic profiling of ctDNA and EVs DNA with tumor samples of nine patients submitted to radical cystectomy, and found the amplification of MDM2, ERBB2, CCND1, and CCNE1, and deletion of CDKN2A, PTEN, and RB1 genes, thus suggesting EVs DNA could also be another source for liquid biopsy [222].Proteomic analysis of urinary EVs content has also started to contribute with the identification of proteins, regarding its diagnostic properties. One of the first studies documented the protein content of urinary EVs from healthy donors, using liquid chromatography-tandem mass spectrometry [131]. Chen et al. [223] analyzed the EV protein content in the urine of patients with BCa, compared with inguinal hernia patients used as controls. Liquid chromatography-tandem mass spectrometry identified 107 differently expressed proteins, including tumor associated calcium-signal transducer 2 (TACSTD2), a cell-surface protein absent in blood and with minimal expression in normal cells. Another study analyzed proteomic data from 129 BCa patients and 62 healthy controls and revealed urinary BCa biomarkers for diagnosis (alpha-1 antitrypsin, SERPINA1) and prognosis (Histone H2B type 1-K, H2B1K) [224]. Smalley et al. [225] using mass spectrometry found higher urinary levels for eight proteins in EVs from BCa patients. Besides the alpha subunit of GsGTP binding protein, resistin, and retinoic acid-induced protein 3, these authors identified five proteins associated with the epidermal growth factor receptor (EGFR) pathway, namely mucin 4, the epidermal growth factor receptor kinase substrate 8-like protein 1 (EPS8L1), the Eps15 Homology (EH)-domain-containing protein 4, the epidermal growth factor receptor kinase substrate 8-like protein 2 (EPS8L2) and the Guanosine-50-triphosphate hydrolyzing enzyme NRas (GTPase NRas). Another study, by Welton et al. [196] analyzed the urinary protein content on EVs of HT1376 bladder cancer cells, as well as in patients diagnosed with bladder cancer and healthy controls, and found several proteins elevated in BCa patients, namely basigin, integrin β1, integrin α6, MUC1, CD10, CD36, CD44, CD73, and 5T4. As mentioned before, Beckam et al. [141] found BCa derived EVs had higher levels of EDIL-3, which besides promoting angiogenesis and migration in a neoplastic environment, could also serve as a prognostic biomarker. The same principle could apply to periostin, which promotes tumor aggressiveness and progression, and is also present in higher levels in urothelial cancer patients. Its presence in cancer patients was associated with a poorer clinical outcome [142].Further research on EVs is warranted, while creating the bases for accumulating evidence from past and present studies. Results should be consistently deposited in public databases, to facilitate the progress of research in this area. There are different compendiums available, namely webdomains such as the ExoCarta (http://www.exocarta.org), EVpedia (http://evpedia.info), and Vesiclepedia (http://www.microvesicles.org) which are updated databases on EVs characterization, content of proteins, mRNA, and lipids [226,227].Diagnosis and follow-up of bladder cancer currently relies on cystoscopy and cytology, despite known limitations. Cystoscopy is costly, invasive, and has reduced sensitivity for Cis or non-papillary lesions, whereas cytology lacks sensitivity for low grade tumors. Many urine-based tests have been developed to improve efficacy beyond current diagnostic tests. FDA-approved systems for diagnosis and monitoring of BCa have demonstrated higher sensitivity but lower specificity than cytology, particularly in cases of low grade and early stage or recurrent BCa; other tests are costly, limiting their use in the daily health practice (e.g., UroVysion test). Besides FDA-approved tests, the other commercially available tests remain mostly at the research level. Most tests have been assessed in inadequate conventional case-control studies, emphasizing the need for prospective cohort studies, with serial samples at different time points from a person at-risk, as well as large randomized trials, validating the biomarker clinical benefit compared to actual gold standard methods. This is the reason why their use as adjunct or surrogate to conventional cystoscopy and cytology is still not recommended by international societies’ guidelines. Moreover, positivity or equivocal results in these tests, when associated with negative cystoscopic findings, may increase patient’s anxiety and trigger further invasive medical examinations, namely biopsies or ureteroscopic procedures. Consequently, besides the need of adequate studies to validate biomarkers for early detection, the current trends of research should focus on the combination of biomarkers into signatures.Extracellular vesicles released by cancer cells carry potential cancer specific biomarkers, as they shed directly from tumor cells and contain protein and nucleic acid material that reflect their cells of origin. EVs play indeed a fundamental role in intercellular communication, being key effectors in normal and pathophysiological cancer progression. Notably, unlike biopsies of solid tumors that provide a small picture of tumor heterogeneity, EVs might provide a wider perspective of tumor heterogeneity, since they are shed by tumor cells and from the cells of the tumor microenvironment. Moreover, their stability within the biological milieu, seemingly increased concentration in some tumors, and unique molecular signatures in oncological patients, makes EVs attractive biomarkers for cancer diagnosis and follow-up. Although out of the scope of this review, EVs can also be explored as therapeutic adjuvants, as a conveyance means for drug delivery and chemosensitization in BCa.Here, we sought to describe research that has been done using EVs as biomarkers in BCa. Indeed, EVs hold promise as biomarkers, not only in BCa, but overall, in oncology. Nevertheless, further research from bench-to-bedside is still needed, as discussed next, particularly in demonstrating its clinical effectiveness. The first limitations to overcome are technical problems related to EVs separation and characterization. There is a need for consensus in EVs nomenclature, to eventually stratify exosomes, microvesicles, and apoptotic bodies [228], and the need for fast, reproducible, and effective separation methods, improving standardization and comparison between studies. Currently, the most frequently used methods for separation of EVs rely on ultracentrifugation procedures that separate them based on size and density [150]. However, this is a time-consuming, laborious, and expensive method, that needs large amounts of sample material and requires expensive equipment, halting its applicability in the clinical laboratory. Other procedures, including double-filtration microfluidic chip-based devices to separate EVs concentration at the point-of-care [174,229] have been proposed, which combine immuno-affinity, sieving, and trapping to concentrate EVs. Indeed, this approach has the advantage of needing lower sample volumes but the disadvantages of EVs structural damage and lower recovery rates, thereby hampering its clinical applicability. Further separation techniques include immune-affinity capture, using antibodies directed against EVs surface markers and bypassing ultracentrifugation [230]. Another overlooked technical issue regards the specificity of these different EVs separation methods. Some of the isolated material identified as of EVs origin by these methods may not in fact be EVs related but derived from other soluble urinary components. On the other hand, urine is enriched in contaminants, such as albumin, Tamm–Horsfall protein and different lipoproteins, and other substances that need to be fully identified and isolated [160]. Therefore, there is the need of a specific, reliable, standardized, and reproducible method, to reduce the confounding effect these contaminants on the EV separation process. A recent research identified a list of 684 of these potential contaminants and developed a bottom-up density gradient centrifugation method to separate EVs from different kinds of protein material in urine, with high specificity and methodological repeatability [160].Before clinical application, standardization of pre-analytical conditions for handling urine specimens is also required. Variables such as urine collection, use of protease inhibitors, storage, and shipping conditions should be accounted for, albeit often disregarded [155]. Bridging clinical usefulness with EVs research requires reporting guidelines to support readability, interpretation, and replication of experiments. We strongly encourage researchers to follow the International Society for Extracellular Vesicles (ISEV) guidelines [124] and the recently created EV-TRACK database (http://evtrack.org) stimulates researchers to report their methodologies for developing standardized protocols, place experimental guidelines into practice and increase research reproducibility [157,231].From a clinical analytical perspective, the use of EVs as a source of biomarkers needs to be properly validated in negative controls, since urinary EVs are produced by cells from all urinary tract. Indeed, one of the main impairments of available urinary biomarkers is the low specificity. Positive results with these tests might also occur in benign conditions such as benign prostatic hyperplasia, urinary lithiasis, endourologic stents, or urinary tract infections. Therefore, to overcome these limitations it is necessary to distinguish EVs derived from BCa cells, from other sources of EVs such as the kidney or prostate, imposing the need for including in clinical studies negative controls, such as EVS from patients with prostate and kidney cancer and hematuria, in order to fully determine the specificity for BCa. Another source of criticism that hampers EVs application as a source of biomarkers in BCa relies on the fact that most studies typically have a small and often heterogeneous cohort of BCa patients, limiting validity and comparisons. However, this is also one of the main limitations of other commercially available urine tests, that preclude their recommendation by most international scientific and clinical organizations. After the identification of robust candidate biomarkers in EVs, additional external validation in large independent multi-institutional studies are required to establish their value as useful biomarkers in bladder cancer.Cystoscopy and cytology remain the clinically approved diagnostic and follow-up procedures for bladder cancer management. This review provides a critical insight into the available urine-based biomarkers, revealing their low improvement on the precision of diagnosis due to low specificity and limiting clinical utility, and fostering the need for more reliable, sensible, and specific urinary biomarkers for BCa.Extracellular vesicles secreted from their cells of origin are vital players in the physiological and pathological intercellular communication processes and are known to promote cancer progression. In recent years, there has been growing interest in EVs as a source of biomarkers in liquid biopsies for cancer diagnosis, management, prognosis, and even as vehicles for cancer treatment. Published reports yielded encouraging findings, even though the path from bench-to-bedside still needs to be optimized, namely regarding standardizing separation protocols and including powered studies with external validation. These crucial steps are fundamental for clinical implementation of EVs as a source of diagnostic and predictive biomarkers in liquid biopsies from BCa patients.Conceptualization, R.R.; Writing—Original Draft Preparation, M.C.d.O.; Writing—Review and Editing, M.C.d.O., H.R.C., M.J.O., A.F., M.H.V., R.R.; Supervision, A.F., M.H.V. and R.R.; All authors have read and agree to the published version of the manuscript. This research received no external funding.The work of our laboratory is funded by FEDER - Fundo Europeu de Desenvolvimento Regional funds through the COMPETE 2020—Operational Programme for Competitiveness and Internationalization (POCI), Portugal 2020, and by Portuguese funds through FCT—Fundação para a Ciência e a Tecnologia/ Ministério da Ciência, Tecnologia e Ensino Superior in the framework of the project "Institute for Research and Innovation in Health Sciences" (POCI-01-0145-FEDER-007274). The authors acknowledge Christophe de Sousa for figure sketching.MHV and HRC are members of the research team of a project financed by Celgene. MHV is member of the team of a grant co-financed by AMGEN. These companies had no role in the decision to publish nor were they involved in the writing of this manuscript.Urine biomarkers for bladder cancer (BCa) diagnosis and follow up. Illustration of the distinct available approaches for the detection of urothelial cancer cells in patients’ urine. The close interaction between the bladder tumor and the urine makes this body fluid a reliable source of cancer biomarkers. A plethora of non-invasive assays exploring distinct analytes (exfoliated tumor cells; proteins; genes; metabolites and extracellular vesicles) in patients’ urine allows the longitudinal analysis of tumor progression. Some of the commercially-available tests includes FDA-approved (UroVysion™: aneuploidy of chromosomes 3; 7; 17 and the loss of 9p21 by FISH; ImmunoCyt™/uCyt+™ test: detection of carcinoembrionary antigen (CEA) and mucins by immunohistochemistry (IHC); bladder tumor antigen (BTA) TRAK/BTA Stat and NMP22 BC test kit); non-FDA approved (CxBladder™: IGFBP5, HOXA13, MDK, CDK1 and CXCR2 by RT-qPCR; Assure MDx™: FGFR3, TERT and HRAS (mutations), OTX1, ONECUT2 and TWIST1 (methylation); XPert Bladder Cancer Monitor™: UPK1B, IGF2, CRH, ANXA10 and ABL1 by RT-qPCR; and UBC™: cytokeratins 8 and 18 by ELISA) and the emerging extracellular vesicles (EV)-based biomarkers (not commercialized yet).Schematic representation of extracellular vesicles biogenesis. Extracellular and plasma membrane molecules are engulfed by plasma membrane endocytosis, creating the early endosomes. These are converted into late endosomes called multivesicular bodies (MVB) containing intraluminal vesicles (ILV). The MVBs may either fuse with the plasma membrane and empty their ILVs by exocytosis, termed exosomes, or may be converted into lysosomes and degrade their components. The process of microvesicle formation is calcium dependent and comes from direct shedding from outward cellular membrane budding; thereby carrying membrane markers of the parent cell. Apoptotic bodies are produced by secreting cells undergoing programmed cell death. Extracellular vesicle uptake by recipient cells may occur via fusion of the vesicle membrane with the cell membrane or by endocytosis. The vesicle may also transduce an intracellular signal by ligand binding to a receptor on the recipient cell. Abbreviations: MHC—major hystocompatibility complex; ER—endoplasmic reticulum; MVB—multivesicular bodies; ILV—intraluminal vesicles.Routes of EV delivery to target cells and their potential role in bladder cancer progression. (a) Transfer of EV-enclosed nucleic acids derived urothelial carcinoma cells to nearby naïve fibroblasts can induce their transformation into Cancer-Associated Fibroblasts (CAFs) with altered pro-tumoral secretome. (b) Likewise, transfer of EV-associated lipids and/or oncoproteins to neighbor or distant cells modulate targeted immune cells into an immunosuppressive phenotype and/or facilitate the transformation of healthy epithelial cells. (c) Many of the signaling molecules that integrate the EVs membrane can act directly on the surface receptors of target cells and trigger their own cell signaling pathways without the need of EV internalization. (d) The uptake of EVs derived from drug resistant tumor cells by drug sensitive tumor cells (or even supporting stroma) can mediate the transfer of functional receptors/molecules to recipient cells, thus allowing a similar phenotypic behavior in cells that originally lacked those receptors/molecules (e.g., transfer of functional drug efflux pumps).Urine-based tests to aid bladder cancer clinical reasoning.Abbreviations: BTA; bladder tumor antigen; CIA; colorimetric immunoassay; IF; immunofluorescence; NMP; nuclear matrix protein; UBC; urinary bladder cancer antigen; FISH; fluorescence in situ hybridization; RT-qPCR; reverse transcription-quantitative polymerase chain reaction; SIA; sandwich immunoassay.Extracellular vesicles-derived biomarkers for bladder cancer.Abbreviations: LC-MS/MS: liquid chromatography-mass spectrometry; RT-PCR: reverse transcription polymerase chain reaction; LC-MRM/MS: liquid chromatography-multiple reaction monitoring mass spectrometry; MALDI-TOF MS: matrix assisted laser desorption ionization-time of flight mass spectrometry; mRNA: messenger RNA; miRNA: micro RNA; lncRNA: long non-coding RNA.

Med-MDPI/cancers/cancers-12-06-01401.txt

@@ -0,0 +1 @@
+Extensive desmoplasia is a hallmark of pancreatic ductal adenocarcinoma (PDAC), which frequently associates with treatment resistance. Recent findings indicate that a combination of photodynamic therapy and the multi-kinase inhibitor cabozantinib achieved local tumor control and a significant decrease in tumor metastases in preclinical PDAC models, but the underlying therapeutic mechanisms remain unclear. This study elucidates the molecular basis of this multi-agent regimen, focusing on the role of MET signaling. Since MET activation stems from its interaction with hepatocyte growth factor (HGF), which is typically secreted by fibroblasts, we developed heterotypic PDAC microtumor models that recapitulate these interactions. In these models, MET signaling can be constitutively activated through paracrine and autocrine mechanisms. Photodynamic therapy caused significant elevations in HGF secretion by fibroblasts, suggesting it plays a complex role in the modulation of the paracrine HGF–MET signaling cascade in desmoplastic tumors. Blocking MET phosphorylation with adjuvant cabozantinib caused a significant improvement in photodynamic therapy efficacy, most notably by elevating spheroid necrosis at low radiant exposures. These findings highlight that adjuvant photodynamic therapy can augment chemotherapy efficacies, and potentially achieve improved management of desmoplastic PDAC in a more tolerable manner.Despite recent therapeutic advances, patients with pancreatic ductal adenocarcinoma (PDAC) are confronted with a dismal prognosis and a 5-year survival rate of approximately 5% [1]. The resistant nature of PDAC requires high-dose chemotherapeutic regimens, such as FOLFIRINOX, that are poorly tolerated and can only be administered to patients with sufficient performance status [2,3,4]. Alternatively, gemcitabine combined with albumin-bound paclitaxel (GEM-NAB) is a viable option for patients with metastatic PDAC [5]. A recent retrospective study compared FOLFIRINOX to GEM-NAB for the treatment of metastatic PDAC, which demonstrated that FOLFIRINOX was more effective in extending overall survival, albeit with a substantially higher toxicity profile [6]. The high toxicity noted for FOLFIRINOX is also observed for cabozantinib (XL-184), a multi-receptor tyrosine kinase inhibitor (RTKi) of MET, AXL, RET, and vascular endothelial growth factor receptor 2 (VEGFR2) [7]. Despite promising preclinical results [8,9], a recent Phase I clinical trial observed dose-limiting toxicity at low cabozantinib doses, and the trial was prematurely stopped as a consequence [10]. Adjuvant strategies that augment the efficacy of these regimens are, therefore, imperative to achieve better management of PDAC at lower chemotherapy doses.In this context, photodynamic therapy (PDT) has achieved promising clinical results for the treatment of PDAC [11]. PDT involves the selective activation of a photosensitizing agent in tumor tissues by light irradiation, resulting in the local generation of reactive molecular species that cause severe oxidative stress, vascular damage followed by tumor anoxia, and the onset of anti-tumor immune responses [12,13,14,15]. Through these distinctive mechanisms and with non-overlapping toxicities, (neo)adjuvant PDT was shown to augment the efficacy of gemcitabine, oxaliplatin, irinotecan, metformin, and EGFR-inhibitors in preclinical cancer models [16,17,18,19,20,21,22,23]. Our group recently developed a photoactivatable drug delivery system containing the photosensitizer benzoporphyrin derivative (BPD) and cabozantinib. This nanoliposomal formulation enabled a PDT–RTKi combination that significantly reduced local and metastatic tumor burdens [24]. Importantly, the combination therapy yielded efficacious PDAC eradication with less than a thousandth of the clinical oral cabozantinib dose needed, underscoring that PDT can alleviate dose-limiting cabozantinib toxicities [24].The improved treatment outcomes observed with the combination of BPD–PDT and cabozantinib were attributed to the concomitant inhibition of MET and VEGFR2 signaling by cabozantinib. However, the MET pathway is typically activated through paracrine cancer–stroma signaling, yet this interaction has not been investigated further to date. In fact, the importance of the MET signaling pathway in relation to cancer cell survival following PDT remains largely unexplored. Therefore, the current study aimed to obtain more detailed insights into the activation mechanism of this survival pathway by leveraging heterotypic 3D culture models to recapitulate the intracellular hepatocyte growth factor (HGF)–MET signaling axis. Our findings indicate that PDT can promote tumor–stroma crosstalk by promoting HGF release from fibroblasts. A combination of PDT with cabozantinib prevented MET phosphorylation and improved overall treatment outcomes.Heterotypic 3D culture models comprising PDAC cell lines and HGF-secreting fibroblasts were established to recapitulate the activation of the HGF–MET signaling axis as a result of tumor–stroma interactions. In these models, the MRC5 fibroblast cell line was used since it is known to secrete high levels of HGF [25]. We, therefore, explored the effects of MRC5 on the PDAC cell lines AsPC-1 (metastatic origin, high MET expression) and MIA PaCa-2 (primary cancer origin, low MET expression), which were co-cultured as suspended spheroids in ultra-low attachment plates.Analysis of conditioned media obtained from both 2D and 3D (spheroid) cultures confirmed the secretion of HGF by MRC5 fibroblasts, with minimal secretion by AsPC-1 and MIA PaCa-2 cells (Figure 1A,D). Growth analyses revealed a substantial impact of MRC5 cells on spheroid morphology. In both AsPC-1 and MIA PaCa-2 spheroids, the addition of MRC5 fibroblasts initially induced a notable reduction in spheroid area (Figure 1B,E), followed by accelerated spheroid growth in co-cultures (Figure 1C,F). Confocal microscopy analysis of spheroids in which the individual cell populations were separately labeled with cell tracker dyes revealed that the MRC5 fibroblasts were homogeneously intermingled with the AsPC-1 cells, while clustering of the fibroblasts within the spheroids was observed in the MIA PaCa-2+MRC5 spheroids (Figure 1E and Figure S1).We next assessed whether the HGF–MET signaling axis was activated in heterotypic PDAC spheroids. Immunoblotting revealed that the presence of MRC5 caused a substantial increase in MET phosphorylation in 2D and 3D (co-)cultures of AsPC-1 cells, but did not influence overall MET expression levels (Figure 2). Interestingly, AsPC-1 cells exhibited autophosphorylation of the MET receptor even in the absence of MRC5 fibroblasts (Figure 2A). As observed in Figure 1A, AsPC-1 cells secreted low but detectable levels of HGF in both 2D and 3D culture, thus potentially activating the HGF– MET axis through autocrine signaling. Our findings additionally report that, in comparison to 2D cultures, there were substantial reductions in MET expression levels in the 3D spheroid cultures. In contrast, neither MET nor phosphorylated MET was detected in cultures of MIA PaCa-2 cells, both in the absence and presence of MRC5 cells (Figure 2B). Taken together, spheroids composed of AsPC-1 and MRC5 cells demonstrated high activity of the HGF–MET signaling axis, whereas spheroids composed of MIA PaCa-2 and MRC5 cells did not.We next evaluated whether the presence of MRC5 in AsPC-1 spheroids influenced the efficacy of PDT. In radiant exposure dose-escalation experiments (timeline in Figure 3A), we combined in situ live/dead staining with a recently developed automated image analysis tool for multiparametric assessment of treatment effects [18,26,27]. In this study, we evaluated treatment responses by quantifying viability, necrosis, spheroid area, fractional live area, and fractional dead area.In AsPC-1+MRC5 spheroids, there was a significant reduction in PDT efficacy compared to spheroids without MRC5 fibroblasts, as reflected in the higher viabilities (Figure 3C,G), and lower necrosis (Figure 3D,H) post-PDT. Analysis of spheroid size did not indicate any dose-dependent effects, indicating that assessing spheroid size alone is not a sufficient means to measure PDT effects (Figure 3B). Fractional live and dead areas also depicted highly significant reductions in PDT efficacy in AsPC-1+MRC5 spheroids (Figure 3E,F,I,J). The fitted IC50/EC50 values for radiant exposure (Figure 3G–J) differed for the various parameters but uniformly reported a >2-fold reduction in PDT sensitivity for the AsPC-1+MRC5 spheroids compared to AsPC-1 alone.In MIA PaCa-2+MRC5 spheroids, analysis of PDT efficacies demonstrated a significant reduction in PDT efficacy compared to spheroids composed of MIA PaCa-2 alone (Figure S2). Significant differences between the two culture types were neither observed upon analysis of spheroid necrosis nor upon analysis of the fractional dead area (Figure S2D,F,I,K). In contrast to the AsPC-1 cultures, the MIA PaCa-2-based cultures displayed a clear dose-dependent reduction in spheroid size following PDT (Figure S2B). However, in the presence of MRC5 fibroblasts, the spheroids were less sensitive to this effect and remained more intact following PDT: A substantial, >10-fold reduction in the efficacy of PDT to disrupt spheroid integrity was observed for the MIA PaCa-2+MRC5 cultures compared to MIA PaCa-2 alone (Figure S2G). Analysis of viability and spheroid live area both revealed a significant, 2-fold decrease in PDT efficacy (Figure S2C,H,E,J). Together, these results depict that the integrity of MIA PaCa-2 spheroids is affected by PDT, causing fragmentation into smaller multicellular clusters, consistent with previous observations [17,26,28]. In the presence of MRC5 fibroblasts, these effects appeared to be mitigated. However, these fragmented cell clusters appeared to have comparable viabilities and extents of necrosis that appeared minimally influenced by the presence of MRC5 fibroblasts.Given the pro-survival function of the HGF–MET axis, we investigated whether this signaling pathway was influenced by PDT. We established a radiant exposure dose-response curve for BPD–PDT in MRC5 fibroblasts cultured in 2D and noted an IC50 value of 1.2 ± 0.2 J/cm2 (Figure 4A). The curve demonstrates that a radiant exposure of 10 J/cm2 reduced MRC5 culture viability to 13.9%, as measured 6 h post-PDT. At the same timepoint and PDT dose, a significant elevation in HGF was detected in the cell culture medium (Figure 4B). Results from two technical repeats revealed that HGF was secreted in a radiant-exposure dose-dependent manner, in which HGF secretion levels plateaued at 10 J/cm2 (Figure S3A). There was a linear correlation between PDT-induced cell death and HGF secretion levels from MRC5 fibroblasts (Figure S3B). We next evaluated whether MRC5 fibroblasts cultured in 3D exhibited similar behavior when treated with a dose of 10 J/cm2. As expected, there was a time-dependent accumulation of secreted HGF by non-treated MRC5 spheroids; however, HGF secretion was significantly elevated by PDT (Figure 4C). Statistical analysis revealed a significantly elevated rate constant for HGF secretion by MRC5 fibroblasts treated with BPD–PDT compared to untreated fibroblasts (K = 0.23 ± 0.09 h−1 vs. K = 0.10 ± 0.03 h−1, p = 0.003).In AsPC-1 spheroids, PDT induced an immediate release of HGF that peaked 4 h post-PDT, which was reduced at later time points (Figure S3C). Heterocellular AsPC-1+MRC5 spheroids demonstrated a delayed release of HGF following PDT, peaking 24 h post-PDT (Figure S3C). However, HGF secretion in AsPC-1 and AsPC-1+MRC5 spheroids was only significantly different 24 h post-PDT. AsPC-1 cells may sequester extracellular HGF through the high expression of MET. This sequestration could influence the extent to which HGF can be accurately detected in these cultures. We, therefore, further explored the impact of PDT on MET signaling. Results from two technical repeats indicated that MET expression was reduced in AsPC-1 and AsPC-1+MRC5 spheroids following PDT (Figure S3D,E). This effect is likely a non-specific event, as a similar decrease in EGFR levels was also detected. We hypothesized that the remaining MET in the AsPC-1+MRC5 spheroids would be activated by the PDT-induced release of HGF from MRC5, but not in the spheroids containing AsPC-1 cells alone. An investigation into phospho-MET levels provided some evidence to support this hypothesis (Figure S3D–F), although inconsistencies in the repeated analyses suggest that it remains to be fully validated. We, thus, set out to evaluate the impact of inhibiting MET signaling on PDT outcomes as an alternative approach to determine the importance of the HFG-MET cascade.To investigate whether the abrogation of the HGF–MET signaling axis could improve PDT outcomes, a combination treatment of PDT and cabozantinib (XL-184) was investigated. To prevent potential oxidative photodestruction of cabozantinib during PDT, the treatment sequence consisted of PDT followed by the addition of cabozantinib. An overview of the experimental timeline is given in Figure 5A. We found that complete inhibition of MET phosphorylation in AsPC-1 spheroid cultures could be achieved with a concentration of 10 µM cabozantinib (Figure S4A). This concentration exerted minor toxicity to both AsPC-1 and AsPC-1+MRC5 spheroids following a 72 h exposure, but had overall poor efficacy as a single therapy (Figure S4B,C). Similar results were obtained in MIA PaCa-2 and MIA PaCa-2+MRC5 spheroids (Figure S4D,E). In all culture types, we noted a minor yet significant decrease in spheroid viability when simultaneously exposed to cabozantinib and BPD in the absence of light exposure. This reduction in viability necessitated the normalization of the data in all subsequent PDT-dose response curves to either the 0 J/cm2 controls or the 0 J/cm2 + 10 µM cabozantinib controls to accurately determine the effect of cabozantinib on PDT efficacy.In AsPC-1 spheroids devoid of MRC5 fibroblasts, cabozantinib exerted only a minor yet significant effect on the overall viability of the spheroids (Figure 5D,F), and no significant dose-dependent effects on spheroid size (Figure 5B). A strong increase in spheroid necrosis was observed, translating to a highly reduced EC50 value (3-fold reduction, p < 0.001, Figure 5G,I). There was no significant effect on the fractional live area of the spheroids (Figure 5J,L), yet a strong increase in the fractional dead area of the spheroids was observed (Figure 5M). Similar to the total spheroid necrosis, this translated to a strongly reduced EC50 value (Figure 5O).Similar trends of enhanced PDT efficacy by neoadjuvant cabozantinib were also observed in AsPC-1+MRC5 spheroids. Although there was no significant improvement detected on overall spheroid size and viability (Figure 5B,E,F), the combination therapy achieved significant increases in spheroid necrosis (Figure 5H,I) and fractional dead spheroid areas (Figure 5N,O). Curve fits report a 2- and 4-fold increase in PDT efficacy, respectively. Interestingly, a significant decrease in the IC50 based on the fractional live area was also observed for the combination therapy (Figure 5L), despite that the dose-response curves appear similar (Figure 5K). Statistical comparisons of the individual PDT or PDT+XL-184 treatment groups revealed that the effects of low-dose PDT were particularly amplified by neoadjuvant cabozantinib.In MIA PaCa-2 spheroids, cabozantinib significantly improved the capacity of PDT to reduce overall spheroid area (Figure 6A,C), spheroid viability (Figure 6D,F), and the spheroid fractional dead areas (Figure 6M,O). Despite statically significant elevations in spheroid necrosis and reduction in spheroid live areas at low PDT doses (Figure 6G,J), cabozantinib did not significantly influence the overall dose-response curves based on these parameters (Figure 6I,L).In MIA PaCa-2+MRC5 spheroids, cabozantinib treatment resulted in strong and statistically significant increases in PDT efficacy, as observed by spheroid area (Figure 6B,C), necrosis (Figure 6H,I), and fractional dead area (Figure 6N,O). A minor improvement was observed based on the overall reductions in spheroid live area (Figure 6K,L), yet no effect was observed based on spheroid viability (Figure 6E,F). Similar to the AsPC-1-based cultures, it appeared that cabozantinib was most effective at elevating the efficacy of low-dose PDT.Overall, the findings convincingly demonstrate a substantial improvement in BPD–PDT efficacy when combined with cabozantinib. The treatment effects occurred in different manners in the varying spheroid types. AsPC-1-based spheroids mostly displayed elevated levels of spheroid necrosis, whereas MIA PaCa-2-based spheroids additionally exhibited strongly reduced spheroid sizes. Cabozantinib was most effective in combination with low-dose (<10 J/cm2) PDT regimens. This is likely related to the eradication of the MRC5 fibroblasts at higher doses (Figure 4A), where paracrine HGF signaling may no longer be stimulated by these cells.The current study aimed to obtain detailed mechanistic insights into the role of the HGF–MET axis in mediating PDT resistance and its rational inhibition by cabozantinib to mitigate resistance. To do so, heterotypic pancreatic cancer spheroids comprising both PDAC cell lines and HGF-secreting fibroblasts were established. However, spheroids composed of AsPC-1 cells alone showed that MET signaling can be auto-activated even in the absence of HGF-secreting fibroblasts. PDT was shown to induce the release of HGF from dying fibroblasts and may play a significant role in facilitating sustained MET survival signaling. Cabozantinib was shown to be capable of abolishing MET phosphorylation and enhancing the efficacy of PDT in 3D PDAC microtumors regardless of MET expression levels. Cabozantinib significantly increased efficacy of PDT, most notably in the low radiant exposure (0.5–10 J/cm2) range. These results underscore the potential of this combination therapy, which can yield beneficial outcomes regardless of MET expression levels in PDAC tumors.A variety of molecular signaling pathways have been identified in the adaptive responses of cancer cells that enable cells to cope with the effects of PDT [13]. These are mostly intrinsic pathways that implicate the transcription factors nuclear factor (erythroid-derived 2)-like 2 (NRF2), hypoxia-inducible factor 1 (HIF-1), nuclear factor-κB, Activator protein 1 (AP-1), and heat shock factors (HSFs) in mediating reduced sensitivity to PDT [13,29,30,31,32,33,34,35,36]. The results of this study clearly identified MET signaling as an additional auto- and paracrine survival pathway in cancer cells that can be induced and sustained by HGF secretion from perishing stromal cells. HGF is a mitogenic glycoprotein that is secreted primarily by fibroblasts [37], and binds MET that is expressed almost exclusively by epithelial cells [38]. Subsequent prolonged MET signaling following PDT may promote cell motility, proliferation, stem cell maintenance, and therapy resistance in cancer cells of epithelial origin [39], potentially resulting in an overall decrease in therapeutic efficacy. Its clinical relevance to pancreatic cancer has also been demonstrated, with various studies reporting on the overexpression of MET in pancreatic cancers [40] and a correlation between elevated serum levels of HGF and aggressive PDAC phenotypes [41]. These reports underscore that the HGF–MET signaling axis is a promising therapeutic target to overcome treatment resistance in PDAC.A study by Cuneo et al. identified that notable MET expression was detected in 86% of PDAC patients, with abnormally high expression levels noted in 64% of the total patient population [42]. Since high MET expression is associated with greatly reduced progression-free survival [42], cabozantinib appears to be a viable therapeutic candidate for PDAC. Despite promising preclinical data, the high dose-limiting toxicity for cabozantinib restricts its further clinical use for PDAC patients [10]. In contrast, toxicity profiles have been deemed acceptable for the treatment of patients with hepatocellular carcinoma, and clinical trials have yielded promising results regarding the use of cabozantinib for this disease [43]. When comparing these studies, the patients with hepatocellular carcinoma received a daily dose of 60 mg, whereas the PDAC patients received 20–40 mg per day. All patients with hepatocellular carcinoma had received sorafenib chemotherapy prior to the trial but received no additional treatment during the trial. In contrast, the advanced PDAC patients continued to receive gemcitabine during the trial. The diverging toxicity profiles may be related to the combination with gemcitabine, but further studies are necessary to provide conclusive evidence. In our study, we noted that BPD and cabozantinib had negligible (dark) toxicity when used as monotherapies at 0.25 and 10 µM, respectively, but had notable (dark) toxicity when these agents were combined. The exact nature of this toxicity warrants further investigation. Our findings, both presented here and previously [24], provide strong indications that PDT can aid in making cabozantinib more effective, and thus reduce the dose-limiting toxicities of this multi-kinase inhibitor. Other multi-kinase inhibitors have also been shown to be effective in augmenting the anti-tumor PDT effect in vivo. Inhibiting EGFR and HER-2 simultaneously using a liposomal formulation of lapatinib resulted in enhanced amino-levulinic acid-PDT efficacies and prolonged survival in an orthotopic rat model of glioma [44].Heterotypic 3D models of PDAC that include multiple cell types are valuable platforms for preclinical treatment screening as they can recapitulate stromal cues that influence treatment efficacy. This study, and others, have demonstrated that fibroblasts, including cancer-associated fibroblasts and pancreatic stellate cells, induce notable changes in spheroid/organoid morphology and growth, PDAC metabolism, and treatment susceptibility [23,27,45,46,47,48,49,50]. This study demonstrates the additional capacity of these models to recapitulate specific paracrine signaling events, such as the HGF–MET axis, although the cell lines were carefully selected to represent this specific signaling axis. In agreement with previous studies [24], cabozantinib proved effective in prohibiting MET phosphorylation and exerted an additive effect on PDT efficacy. The significant increases in PDT efficacy in 3D heterocellular cultures observed here does not fully recapitulate the promising effects observed in vivo, in which a combination of PDT and cabozantinib greatly enhanced tumor regression and reduced metastatic burdens [24]. Interestingly, these effects were observed in both AsPC-1 and MIA PaCa-2 tumors that were orthotopically implanted, despite the absence of MET expression by the MIA PaCa-2 cell lines. Therefore, the current findings suggest that the mesoscopic treatment effects of cabozantinib, such as impaired angiogenesis by VEGFR2 inhibition, are major contributors to the overall treatment efficacy that can currently not be fully recapitulated using heterotypic 3D cultures in vitro. In addition, the in vivo impact of HGF–MET signaling on angiogenesis and treatment resistance may provide further insights into the potentiating effect that cabozantinib has on PDT in vivo [51]. Future work using heterotypic desmoplastic organoids that recapitulate the vascularization of tumors may offer further understanding of the mechanisms of synergistic anti-tumor combinations, such as PDT with cabozantinib. Auxiliary inhibition of treatment escape by cabozantinib may also stem from impaired AXL and RET activity [52], although further research is needed to identify the therapeutic contributions of these effects.All cell lines were obtained from the American Type Culture Collection (Manassas, VA, USA). AsPC-1 cells were maintained in RPMI medium (Corning, Corning, NY, USA), MIA PaCa-2 were maintained in DMEM medium (Corning), and MRC5 cells were maintained in MEM medium (Corning). All types of culture media were supplemented with 1% (v/v) penicillin/streptomycin (Corning) and 10% (v/v) fetal bovine serum (FBS, Gibco, ThermoFisher, Waltham MA, USA). Culture media were changed every 3–4 days, and cells were passaged weekly. Cells were discarded after passage 25 and tested negative for mycoplasma.AsPC-1, MIA PaCa-2, and MRC5 cells were seeded at a density of 5000 cells per well per cell line in ultra-low attachment round-bottom plates (Corning) in a final volume of 100 µL in a 1:1 growth media mixture. Cells were either seeded as a single cell type (5000 cells/well) or a mixture of either AsPC-1 or MIA PaCa-2 cells, combined with MRC5 fibroblasts (5000 + 5000 cells/well). When two cell lines were mixed, they were added within 30 min of each other to ensure heterogeneous distribution.Spheroid growth was determined by daily brightfield imaging (Operetta CLS, Perkin–Elmer, Billerica, MA, USA). Image acquisition was performed using a 5× air objective (0.16 N.A.), at a resolution of 1080 × 1080 px2. From the brightfield images, spheroid areas were extracted using a custom-written script in MATLAB 2016b (Mathworks, Natick, MA, USA).AsPC-1, MIA PaCa-2, and MRC5 were seeded separately in 6-well plates at a density of 1 × 106 cells per well and were allowed to attach for 24 h. Culture media were removed, and cells were washed twice with phosphate buffer saline (PBS). A PBS solution containing 5 µM of CellTracker Red CMTPX (#C34552, Molecular probes, Thermo Fisher) or 5 µM CellTracker Deep Red (#C34565, Molecular probes) was added to AsPC-1/MIA PaCa-2 or MRC5, respectively, and cells were incubated for 45 min at 37 °C. The solution was aspirated, and labeled cells were harvested by trypsinization. Labeled cells were then used for spheroid culture initiation, as described above. Images were taken on culture day 3 using an Olympus FV1000 confocal laser scanning microscope (10× air objective, 0.4 N.A.). Image acquisition parameters were λexc = 559 nm, λem = 580 ± 50 nm (CellTracker Red) and λexc = 635 nm, λem = 750 ± 50 nm (CellTracker Deep Red). The image resolution was set up at 1024 × 1024 px2.Total protein lysates were collected from 2D monocultures or spheroids using RIPA buffer (Thermo Scientific) containing 1% protease inhibitor cocktail (Calbiochem, Merck Millipore, Danvers MA, USA) and 1% phosphatase inhibitors 2 and 3 (Sigma–Aldrich, St. Louis MO, USA). Lysates were centrifuged at 16,000 rpm at 4 °C to remove cellular debris and then transferred into a sterile 1.5 mL tube. Samples of 1–5 µg of protein were separated on 4–20% Tris-HCL gels (Bio-Rad, Hercules CA, USA). Primary antibodies rabbit-anti-MET (clone D1C2), rabbit-anti-phospho-MET (Y1234-Y1235, clone D26), and rabbit-anti-β-actin (clone 13E5) were purchased from Cell Signaling Technologies (Danvers, MA, USA) and were used at 1:1000 dilution ratios. All appropriate horseradish peroxidase-conjugated secondary antibodies were also purchased from Cell Signaling Technologies and used at 1:2000 dilution ratios.AsPC-1, MIA PaCa-2, and MRC5 were cultured as monolayers in a 6-well plate at a density of 1.5 × 105 cells per well, or as spheroid cultures, as described above. Conditioned culture media were collected after 48 h of culturing. Pro-HGF levels were measured using an ELISA (R&D Systems, Minneapolis, MN, USA) using 50 µL of culture medium per sample. We ensured that signal collection never exceeded the upper detection limit of the assay. In experiments, when HGF secretion was measured following PDT in monolayer cultures, MRC5 cells were seeded in black-walled, flat-bottom 96-well plates at a density of 1000 cells per well. The culture medium was harvested after various time intervals, as indicated.PDT was initiated on culture day 3. A 3× stock solution containing 0.75 µM benzoporphyrin derivative (BPD, US Pharmacopeia) was prepared in full growth media and added directly to each well to reach a final concentration of 0.25 µM. After 1 h incubation, 100 µL of culture media containing BPD was carefully aspirated and gently replaced with 100 µL of full growth media. A custom-made black insert that prevented light scattering was used to ensure correct PDT dosimetry in each well. Spheroids or cells were irradiated with 150 mW/cm2, 690 nm laser light (Intense, North Brunswick, NJ, USA) at cumulative radiant exposures ranging from 0–80 J/cm2 as indicated. After PDT, all cultures were maintained at standard culture conditions.Cabozantinib (XL-184, Selleckchem, Houston TX, USA) was dissolved in cell culture-grade DMSO (Sigma-Aldrich). A 3× stock solution of 30 µM XL-184 was prepared in full growth media and then filter-sterilized. For 3D cultures, each well containing 100 µL of media received 50 µL of stock solution to yield a final concentration of 10 µM XL-184. XL-184 was not present when cells were irradiated for PDT to prevent its oxidation but was added immediately after PDT from culture days 3–5, as indicated in the treatment timelines.Treatment efficacy was determined by in situ Live/Dead staining using calcein AM (ThermoFisher) and propidium iodide (Sigma–Aldrich), followed by fluorescence imaging (Olympus FV1000 confocal laser scanning microscope), as optimized for 3D cultures [26]. Derivation of outcome parameters was done using quantitative image analysis according to the CALYPSO methodology, as described previously [27]. When indicated, culture viability was assessed based on NAD(P)H oxidase activity using a CellTiter 96 aqueous non-radioactive proliferation assay (MTS, Promega, Madison, WI, USA), which was performed in accordance with the manufacturer’s recommendation.All statistical analyses were performed in Prism 5.0 (Graphpad, La Jolla, CA, USA). Growth curves were fitted using a Gompertz growth equation. Data were tested for normality using the Shapiro–Wilk method, after which the data were statistically compared using a One-way ANOVA and Sidak’s post-hoc test (normally distributed data), or using a Kruskal–Wallis and Dunn’s post-hoc test (non-normal data or data with N < 4). Dose response curves were obtained using standard inhibitor/agonist versus normalized response fits, and IC50/EC50 values were compared using extra sum-of-squares f-tests. Statistically significant differences between treatment groups are indicated with either a single asterisk (p ≤ 0.05), double asterisk (p ≤ 0.01), triple asterisk (p ≤ 0.005), or a quadruple asterisk (p ≤ 0.001).Modulating the cancer stroma has been considered a potentially effective approach for pancreatic cancer treatment [23,53]. We demonstrated that in a stroma-rich spheroid model that recapitulates the HGF–MET signaling pathway, a combination treatment that incorporates cabozantinib provides a significant additive effect on the therapeutic efficacy of PDT. The findings underscore the necessity of understanding the stromal cues that promote treatment resistance in PDAC and highlight the potential to therapeutically target these intercellular signaling pathways that are specific to desmoplastic cancers. The potentially high toxicity of such therapeutic regimens may be mitigated by developing PDT as an adjuvant therapy, as exemplified here for cabozantinib.The following are available online at https://www.mdpi.com/2072-6694/12/6/1401/s1, Figure S1: Cell tracker imaging informs on the dispersion of the MRC5 fibroblasts in spheroid co-cultures, Figure S2: PDT efficacy is reduced by the presence of MRC5 fibroblasts in MIA PaCa-2 spheroids, Figure S3: Impact of PDT on HGF secretion by MRC5 fibroblasts and MET expression and phosphorylation in AsPC-1 spheroids, Figure S4: Toxicity evaluation of 10 µM cabozantinib on heterotypic PDAC spheroids.Conceptualization, A.A., M.B., I.R., and T.H.; Data curation, A.A., M.B., S.B., and A.-L.B.; Formal analysis, A.A., M.B, and A.-L.B.; Funding acquisition, T.H.; Investigation, M.B., A.A., S.B., I.R., and T.H.; Methodology, A.A., A.-L.B., G.O., and I.R.; Project administration, T.H.; Resources, T.H.; Software, A.-L.B.; Supervision, G.O., I.R., and T.H.; Validation, M.B.; Visualization, M.B.; Writing – original draft, M.B. and A.A.; Writing – review & editing, M.B., A.-L.B., S.B., G.O., I.R., and T.H. All authors have read and agreed to the published version of the manuscript.The work was supported by grants from the National Cancer Institute/National Institutes of Health: P01 CA084203, R01 CA158415, and R01 CA160998 (T.H.), K99CA175292 and R00CA175292 (I.R.), K99CA215301 and R00CA215301 (G.O.), as well as the Bullock–Wellman Fund (A.-L.B.). M.B. and A.-L.B. acknowledge the European Society for Photobiology for support in the form of a travel award.The authors declare no conflict of interest.Establishment and characterization of heterotypic pancreatic ductal adenocarcinoma (PDAC) spheroids. (A) Quantification of hepatocyte growth factor (HGF) secretion in the conditioned media (48 h) of AsPC-1, MRC5, and AsPC-1+MRC5 2D monolayer (co-) cultures (grey bars) or 3D spheroid (co-)cultures (black bars). Depicted are the mean ± SEM of N = 6–10 from ≥3 technical repeats. (B) Brightfield images of AsPC-1 and AsPC-1+MRC5 spheroid growth. Scale bar = 250 µm. (C) Growth kinetics of AsPC-1 and AsPC-1+MRC5 spheroids based on quantitative analysis of the brightfield images. Depicted are the mean ± 95% CI of N = 16–32 from 3 technical repeats. (D) Quantification of HGF secretion in the conditioned media (48 h) of MIA PaCa-2, MRC5, and MIA PaCa-2+MRC5 2D monolayer (co-)cultures (grey bars) or 3D spheroid (co-)cultures (black bars). Depicted are the mean ± SEM of N = 4–6 from ≥2 technical repeats. (E) Brightfield images of MIA PaCa-2 and MIA PaCa-2+MRC5 spheroid growth. Scale bar = 250 µm. (F) Growth kinetics of MIA PaCa-2 and MIA PaCa-2+MRC5 spheroids based on quantitative analysis of the brightfield images. Depicted are the mean ± 95% CI of N = 16–48 from ≥3 technical repeats. Statistical significance is indicated as * p < 0.05, ** p < 0.01, *** p < 0.005.High HGF–MET signaling activity was detected in AsPC-1-based spheroids, but not in MIA PaCa-2-based spheroids. (A) Immunoblots depicting the expression levels of MET, phospho-MET (Y1234/1235), and β-actin (loading control) in 2D and 3D cultures composed of AsPC-1, AsPC-1+MRC5, MIA PaCa-2, and MIA PaCa-2+MRC5 cells. (B,C) Comparative protein band quantification of MET expression (B) and MET phosphorylation (C) in 2D and 3D cultures composed of AsPC-1 and AsPC-1+MRC5 cells. Data represent the mean + SD from four technical repeats and were normalized to their respective loading controls. Statistical significance is indicated as * p < 0.05.PDT efficacy was reduced by the presence of MRC5 fibroblasts in AsPC-1 spheroids. (A) Experimental timeline. (B–F) PDT dose-response in AsPC-1 (blue) and AsPC-1+MRC5 spheroids (red) based on spheroid viability (B), necrosis (C), spheroid size (D), fractional live area (E), and fractional dead area (F). Data depict the mean ± SEM of N = 12 obtained from three technical repeats. (G–J) IC50 and EC50 values were extracted from the dose-response curve fits. Statistical significance is indicated as * p < 0.05, ** p < 0.01, or **** p < 0.001.PDT elevates HGF secretion by MRC5 cells in 2D and 3D. (A) Efficacy of benzoporphyrin derivative (BPD)–PDT on MRC5 cells cultured in 2D. Depicted are mean ± SEM from N ≥ 12 obtained from four technical repeats. (B) Quantification of 10 J/cm2 PDT-induced HGF secretion in culture medium by MRC5 cells in 2D (mean ± SD from N = 4–8 from three technical repeats). (C) Quantification of HGF secretion in culture media by MRC5 spheroids. Depicted are mean ± SEM from N = 4–23 obtained from four technical repeats. Data were fitted using a one-phase decay equation. Statistical significance is indicated as ** p < 0.01, *** p < 0.005, or **** p < 0.001.Cabozantinib augments the capacity of PDT to elevate necrosis in AsPC-1 and AsPC-1+MRC5 spheroids. (A) Schematic overview of the experimental timeline. (B,C) Effect of cabozantinib + PDT on overall AsPC-1 (B) and AsPC-1+MRC5 (C) spheroid size. (D–F) The effect of cabozantinib on the PDT efficacy based on spheroid viability in AsPC-1 (D), AsPC-1+MRC5 spheroids (E), and a statistical comparison of the fitted IC50 values (F). (G–I) The effect of cabozantinib on the PDT efficacy based on spheroid necrosis in AsPC-1 (G), AsPC-1+MRC5 spheroids (H), and a statistical comparison of the fitted EC50 values (I). (J–L) The effect of cabozantinib on the PDT efficacy based on the fractional live area of AsPC-1 (J), and AsPC-1+MRC5 spheroids (K), and a statistical comparison of the fitted IC50 values (L). (M–O) The effect of cabozantinib on the PDT efficacy based on the fractional dead area of AsPC-1 (M), and AsPC-1+MRC5 spheroids (N), and a statistical comparison of the fitted EC50 values (O). All data are the mean ± SEM from N = 12 from three technical repeats. The error in IC50 values represents the SD. Statistical significance is indicated as * p < 0.05, ** p < 0.01, *** p < 0.005, or **** p < 0.001.Cabozantinib augments the capacity of PDT to reduce spheroid sizes and elevate necrosis in MIA PaCa-2 and MIA PaCa-2+MRC5 spheroids. (A–C) The effect of cabozantinib on the PDT efficacy based on the size of MIA PaCa-2 spheroids (A), MIA PaCa-2+MRC5 spheroids (B), and a statistical comparison of the fitted IC50 values (C). (D–F) The effect of cabozantinib on the PDT efficacy based on the viability of MIA PaCa-2 spheroids (D), MIA PaCa-2+MRC5 spheroids (E), and a statistical comparison of the fitted IC50 values (F). (G–I) The effect of cabozantinib on the PDT efficacy based on overall necrosis in MIA PaCa-2 spheroids (G), MIA PaCa-2+MRC5 spheroid (H), and a statistical comparison of the fitted EC50 values (I). (J–L) The effect of cabozantinib on the PDT efficacy based on the fractional live area of MIA PaCa-2 spheroids (J), and MIA PaCa-2+MRC5 spheroids (K), and a statistical comparison of the fitted IC50 values (L). (M–O) The effect of cabozantinib on the PDT efficacy based on the fractional dead area of MIA PaCa-2 (GM), and MIA PaCa-2+MRC5 spheroids (N), and a statistical comparison of the fitted EC50 values (O). All data are the mean ± SEM from N = 8–36 from ≥3 technical repeats. The error in IC50 values represents the SD. Statistical significance is indicated as * p < 0.05, ** p < 0.01, *** p < 0.005, or **** p < 0.001.

Med-MDPI/cancers/cancers-12-06-01402.txt

@@ -0,0 +1 @@
+Gastric cancer (GC) is a common cancer worldwide. Its incidence and mortality vary depending on geographic area, with the highest rates in Asian countries, particularly in China, Japan, and South Korea. Accurate imaging staging has become crucial for the application of various treatment strategies, especially for curative treatments in early stages. Unfortunately, most GCs are still diagnosed at an advanced stage, with the peritoneum (61–80%), distant lymph nodes (44–50%), and liver (26–38%) as the most common metastatic locations. Metastatic disease is limited to the peritoneum in 58% of cases; in nonperitoneal distant metastases, the most involved GC metastasization site is the liver (82%). The eighth edition of the tumor-node-metastasis staging system is the most commonly used system for determining GC prognosis. Endoscopic ultrasonography, computed tomography, and 18-fluorideoxyglucose positron emission tomography are historically the most accurate imaging techniques for GC staging. However, studies have recently shown renewed interest in magnetic resonance imaging (MRI) as a useful tool in GC staging, especially for distant metastasis assessment. The technical improvement of diffusion-weighted imaging and the increasing use of hepatobiliary contrast agents have been shown to increase the diagnostic performance of MRI, particularly for detecting peritoneal and liver metastasis. However, no principal oncological guidelines have included the use of MRI as a first-line technique for distant metastasis evaluation during the GC staging process, such as the National Comprehensive Cancer Network Guidelines. This review analyzed the role of the principal imaging techniques in GC diagnosis and staging, focusing on the potential role of MRI, especially for assessing peritoneal and liver metastases.Gastric cancer (GC) is one of the most common malignancies of the digestive system [1] and is the fifth most common tumor among overall cancers, with an incidence of one million new patients diagnosed every year [2,3]. Fortunately, in the 21st century, improvements in worldwide sanitary conditions, diet, and medical advances have effected a significant reduction in the incidence of GC, especially in those patients with advanced disease, usually called advanced GC [2]. However, this malignant tumor still remains a significant cause of morbidity and mortality worldwide (fifth most common malignancy in the world) due to its global diffusion, associated with common risk factors such as diet, Helicobacter pylori, tobacco, obesity, adenomatous polyps, chronic atrophic gastritis, and the male sex [4].The incidence of GC varies greatly depending on geographic area, which is similar to what has been observed in other tumors: it remains the first cause of mortality from malignant cancers in Eastern Asia, and nearly two-thirds of global GCs occur in developing areas. The highest incidence occurs in Japan and South Korea, and the most involved population is the Chinese, who represent almost 43% of all cases. This figure appears to be correlated with particular diets and hygienic conditions [5].Unfortunately, GC is characterized by a high mortality rate: the five-year overall mortality is 70–80%. At present, the best survival rate is in the Japanese (52% five-year survival), probably due to the effective screening programs for GC that began in the early 1990s [6,7,8].The term “gastric cancer” is commonly used to refer to adenocarcinoma, which represents between 90% and 95% of all gastric malignancies. However, benign lesions commonly comprise approximately 85–90% of all lesions found within the stomach [8]. There are multiple subtypes of GC, including papillary, tubular, and signed ring cell forms. This high variability in histotypes corresponds to the varying appearance of adenocarcinoma on imaging. In fact, it could appear as a bulky mass, sometimes with ulceration, such as a gastric wall thickening or a diffuse parietal infiltration (without a visible lesion), or as a particular presentation called “linitis plastica” [9]. Moreover, one distinctive but uncommon appearance of GC is that of mucinous adenocarcinoma, which can partially calcify [10]. Accurate tumor-node-metastasis (TNM) staging of GC is the cornerstone (central pin) of prognostication in GC and for performing the most accurate decision-making for treatment, reducing unnecessary surgery, and maximizing the likelihood of benefiting from selected treatments [1,11].The imaging techniques typically performed for the diagnosis of GC and for performing its staging are endoscopic ultrasonography (EUS), computed tomography (CT), and positron emission tomography (PET)-CT [3,12,13]. Historically, magnetic resonance imaging (MRI) had a limited role in GC evaluation due to its technical limitations, such as blurring and lower spatial resolution. In recent decades, however, much progress has been made in MRI technology, improving the diagnostic performance of this imaging technique in many areas of medicine, including oncology [14]. Furthermore, MRI presents many advantages over CT (the radiological method recommended for the staging of GC), including the ability to generate significantly greater soft-tissue contrast resolution and the ability to remove the risk of iodinated contrast-induced nephropathy or ionizing radiation [15]. However, in the guidelines for the management of GC, MRI does not appear among the possible imaging techniques useful for GC staging [12,13].This study aimed to analyze the role of the imaging techniques used in the diagnosis and staging of GC, focusing on the potential role of MRI, especially in the assessment of liver and peritoneal metastases from GC.The pivotal role of correct GC imaging lies in the possibility of performing different treatments in various stages of disease, especially curative treatments in the early stages. For example, selected patients with early GC could be treated with endoscopic resection, especially if the lesion is easily approachable during the endoscopic examination [16,17]. In patients with a locally advanced disease but without distant metastasis, perioperative chemotherapy, (sub)total gastrectomy, and regional lymph node dissection are the current acceptable standard of care [16,17]. The poorest clinical scenario is metastatic disease. Patients with a solitary liver metastasis can be treated with surgical resection. However, it is generally accepted that patients with GC at stage M1 are incurable and that they should be managed with noncurative intent [16,17,18,19]. The TNM guidelines recommend the use of EUS to accurately evaluate the extension of the primary tumor (T parameter) [3,12,13]. This technique is sufficient to diagnose the GC stage until T3 (tumor penetrates subserosal connective tissue without invasion of visceral peritoneum or adjacent structures) [3,12,13]. In fact, EUS has demonstrated high diagnostic performance in distinguishing the different layers that compose the gastric wall, visualize the perigastric lymph nodes and detect intraperitoneal fluid utilizing of a modern miniaturized US probe [20]. Conversely, CT has insufficient spatial resolution and soft-tissue contrast resolution to distinguish the layers of the gastric wall, and therefore it is not accurate for evaluating T1, T2, or T3 parameters of GC. However, CT is useful for identifying extragastric invasion (T4) [12,13]. Therefore, for stage T4, the use of CT and then PET-CT is still recommended to assess tumor nodularity outside the stomach or invasion of adjacent organs [12,13]. Finally, MRI is not recommended for the imaging evaluation of the T parameter in GC (Table 1) [13].According to the guidelines for the management of GC, the N parameter (lymph node involvement) is assessed by using CT or PET-CT [12,13]. In particular, potential lymph node metastasis is suspected (1) when the short axis measures greater than 1 cm on CT of the abdomen or on PET-CT and (2) when fluorodeoxyglucose-avid lymph nodes are detected on PET-CT. The criterion for differentiation between stages N1 through N3 is represented exclusively by the number of involved lymph nodes: metastases in one to two regional lymph nodes corresponds to N1, three to six correspond to N2, and seven or more correspond to N3 (N3a, metastasis in 7–15 regional lymph nodes; N3b, metastasis in 16 or more regional lymph nodes). Lymph node count should be assessed using CT or PET-CT. Moreover, pathologically involved lymph nodes inferior to the level of the renal veins are treated as systemic metastases for the purposes of staging [12,13]. Finally, MRI is not recommended for the imaging evaluation of the N parameter in GC (Table 1) [13].Until 2017, it was recommended that the M parameter (metastatic disease) be evaluated by CT or PET-CT and by the N parameter. Therefore, MRI is not recommended for the imaging evaluation of the M parameter in GC (Table 1) [13].Is there robust scientific evidence to support the current absence of MRI in the guidelines for the management of GC?A meta-analysis, published by Seevaratnam et al. [21] in 2012, compared the diagnostic performance of the most common imaging techniques used for the staging of GC—EUS, CT, MRI, and PET. This study, which included a total of 40 articles from 1 January 1998 to 1 December 2009, for a total of 3758 patients, clearly conveyed the mistaken belief that CT is more reliable than MRI for TNM staging of GC. The authors demonstrated that the overall accuracy in the assessment of T stages of MRI (82.9% ± 3.7%) was statistically superior to that of CT (71.5% ± 2.7%) (p ≤ 0.014). MRI also appeared to be better than CT in terms of sensitivity in assessing the N parameter of GC, with values of 85.3% and 77.2%, respectively. However, this study presented an important limitation: the authors did not consider the role of MRI for the diagnosis of metastasis (M parameter) because they considered MRI as “limited” due to the insufficient amount of body evaluable within a single examination and as thus unsuitable for staging the M parameter [21]. In their conclusion, however, Seevaratnamm et al. [21] declared the superiority of MRI compared with CT, despite the number of MRI studies analyzed being fewer than the CT studies.Subsequently, in 2015, a systematic review and meta-analysis conducted by Huang et al. [22] evaluated the utility of MRI in preoperative T and N staging of GC. Eleven studies were included in the analysis, and the final results revealed surprising data concerning the diagnostic performance of MRI in this setting. In fact, the reported pooled sensitivity of MRI in diagnosing the T1, T2, T3, and T4 stages of GC was 66%, 85%, 86%, and 88%, respectively, and the pooled specificity for the same stages was 97%, 90%, 89%, and 97%, respectively. Considering T3 and T4 together, the pooled sensitivity and specificity were 93% and 91%, respectively. Moreover, this study confirmed the remarkable pooled sensitivity (86%) of MRI for the correct assessment of the N parameter, despite a loss in specificity, confirming the data reported in other studies [23,24]. In a subgroup analysis, the authors [22] tested the diagnostic utility of diffusion-weighted imaging (DWI) plus MRI compared with MRI alone. Although DWI appears to be superior for the detection of lesions from GC, no statistical added value with respect to MRI alone emerged from their analysis. These early results have been partially disavowed by subsequent studies that assessed the usefulness of DWI in this setting.In a study by Soydan et al. [25] performed on a total of 46 patients who underwent abdominal DWI-MRI before surgery for GC, the sensitivity, specificity, and accuracy of DWI-MRI in T-staging produced satisfying results—in particular, that DWI-MRI could distinguish T4 from lower stages of GC with great reliability. In fact, the authors concluded that DWI-MRI could consistently help to determine an accurate staging of nonmetastatic GC. This thesis is also supported by other studies in which DWI also demonstrated better performance due to greater soft-tissue resolution than CT, reducing artifacts and with the advantage of revealing more details of GC (depicting characteristics such as necrotic, cystic, or bleeding aspects) [23,24].Joo et al. evaluated the diagnostic performance of MRI performed alone and in combination with DWI for the preoperative TNM staging of GC, especially for assessing metastatic lymph nodes. The MRI performed using DWI had demonstrated a higher sensitivity than MRI without DWI for N staging (86.7% vs. 50%) but a lower specificity (58.8% vs. 90%) [26]. Moreover, the alteration of cellular density, evaluable in metastasis using DWI, has further increased its potentiality, especially in cases of subcentimetric infiltrated lymph nodes [27].Unfortunately, none of these authors considered studies in which the possibility of using MRI as a “one stop” imaging technique to diagnose any TNM parameters during the preoperative phase was evaluated. It is difficult to determine the reasons for the failure to develop MRI for GC staging.A substantial portion of patients with newly diagnosed GC have distant metastases (M1 disease). Stage IV disease, consisting of any T or N and M1, the most advanced form of GC, is diagnosed in 35–55% of GC cases in low-incidence countries, such as the United States and Canada [28,29]. The median disease-specific survival in metastatic disease has been estimated to be approximately 10 months, and overall 5-year survival is estimated to be 3–5% [11,29,30]. Finally, it is important to perform a correct diagnosis of distant metastases in GC, because it dramatically changes prognosis and treatment plans [31].The most common sites of distant metastases from GC are the peritoneum in 61–80% of cases, the distant lymph nodes in 44–50%, and the liver in 26–38%. Following these sites, there are few minor locations for secondary malignancies from GC, such as the lung (10%), bone (6%), abdominal wall (2%), ovary (2%), brain, and prostate (approximately 0.4% for both) [32,33]. Metastatic disease was limited to the peritoneum in 58% of M1 patients, whereas 20% of M1 disease showed exclusively nonperitoneal distant metastases. In this latter case, the most involved GC metastasis site was the liver, with a weighty value of 82%, much more than the two most affected sites after the liver, lung, and bone, both equally involved in approximately 9% of cases [34].Our study aimed to highlight the emerging role of MRI in the detection and characterization of the most common metastases from GC, such as peritoneal and liver metastases. Furthermore, this study aimed to analyze whether there are robust scientific data to explain why MRI was not introduced in the guidelines as a useful tool in the staging of patients affected by GC, given scientific data on its diagnostic accuracy in the assessment of peritoneal and liver metastases from GC are robust.A systematic review by Wang et al. [35], comparing CT and MRI in the assessment of hepatic and peritoneal metastases from GC, demonstrated that CT showed high sensitivity and specificity in evaluating peritoneal and liver metastases from GC. However, MRI results also appeared to be accurate, with a diagnostic performance comparable with that of CT, potentially even better. MRI was described as a growing imaging method for the detection and characterization of most diffused repetitions from the oncologic pathologies of the abdomen, including GC. In fact, MRI demonstrated high sensitivity and specificity in detecting liver metastasis, with a sensitivity of 100% (95% Confidence Interval (CI) 0.40–1.00) and specificity of 100% (95% CI 0.89–1.00). Furthermore, DWI-MRI was more sensitive than CT in detecting liver and peritoneal metastases, and MRI functional parameters such as apparent diffusion coefficient were found to help evaluate the pathological response to neoadjuvant chemotherapy [36].Therefore, over 10 years ago, there was already the strong suspicion that MRI could be a promising method for also evaluating peritoneal metastasis. Unfortunately, in 2011, the lack of data in this diagnostic field (MRI in the detection of peritoneal metastasis) did not facilitate the demonstration of MRI reliability in this oncological setting compared with other available imaging techniques. In fact, among 33 overall studies included in the review [35], only two MRI-based papers were included along with 22 CT studies. However, as early as 2011, these first results made it possible to define MRI as a “promising technology” that could represent the future of the imaging of metastases.Similarly, a systematic review published by Laghi et al. [37] in 2017 assessed the diagnostic accuracy of CT and MRI in detecting peritoneal metastases. The authors analyzed a total of 22 studies, of which only three were MRI-based, whereas 19 were CT-based. The overall final population consisted of 630 patients, of which 275 were reported metastases from gastrointestinal malignancies. In this study, both MRI specificity and sensitivity were greater than those of CT, although without statistical significance. In particular, MRI sensitivity and specificity were 86% (95% CI 0.78–0.93) and 88% (95% CI 0.83–0.92), respectively, versus CT sensitivity and specificity of 83% (95% CI 0.79–0.86) and 86% (95% CI 0.82–0.89), respectively. However, the main limitation of this study, as in that of Wang [35], was the small sample size of the MRI population, which did not allow us to draw definitive conclusions on this imaging modality. In fact, the authors concluded that the MRI technique represents a future resource more than a current reality.In recent years, an increasing number of scientific papers have stressed the need for a correct preoperative diagnosis of distant metastases from GC, particularly peritoneal metastases, to reduce need for a second imaging technique to achieve the final diagnosis and, most of all, to prevent unnecessary surgical treatments. In this context, Borggreve et al. [1] reported that, although MRI is rarely considered for the diagnosis of peritoneal metastasis, this technique demonstrated the best sensitivity, specificity, and diagnostic accuracy for this type of diagnosis compared with other imaging modalities. In particular, the accuracy, sensitivity, and specificity for detection of peritoneal metastases of PET-CT and DWI-MRI were 80%, 84%, and 73% and 83%, 84%, and 82%, respectively.Another recent interesting study, comparing MRI and CT in terms of accuracy for detecting peritoneal ovarian metastases, demonstrated the clear superiority of MRI (93.33%) over CT (79.39%) and even further highlighted the lower “omission rate” of MRI. In fact, the authors observed that MRI missed the detection of only 12 metastatic lesions (6.67%) compared with CT, which omitted 34 lesions (20.61%), in 165 overall secondary malignancies [38]. This poor performance of CT was also confirmed by another study, in which the omission rate for diagnosis of peritoneal metastasis from GC was approximately 16%. In fact, the authors proposed a steep learning curve to improve CT performance; however, it will be some time before it is affordable and available in clinical practice. Current guidelines recommend staging laparoscopy (SL) for advanced GC [13]. Despite many developments in preoperative imaging, the suboptimal identification of advanced carcinomatosis disease could result in high rates of unnecessary laparotomies in patients with GC [39]. The addition of peritoneal lavage can identify patients with free intraperitoneal cancer cells, which suggests a poor prognosis. Although considered as a very useful asset in the preoperative cancer staging, the wider application of SL is been long debated [39]. Awaiting further robust data concerning non-invasive imaging techniques such as MRI in the management of GC, the goal of MRI could be to achieve high sensitivity in detecting peritoneal metastases to reduce the false negative cases who undergo unnecessary surgical procedures. This as a primary result could allow reducing any invasive and mini-invasive procedures, improving patient quality of life. Therefore, according to the confirmed results emerging from the literature, MRI is actually ready to become the “first choice” imaging in the evaluation of peritoneal metastasis, also from GC [40].The liver is by far the most common site of distant metastases in patients with nonperitoneal secondary lesions from GC. A recent review and meta-analysis had compared the diagnostic accuracy of MRI performed with hepatospecific contrast media, such as gadolinium ethoxybenzyl diethylenetriamine pentaacetic acid (Gd-EOB-DTPA) with that of contrast-enhanced CT (CE-CT) in patients with liver metastases [41]. It was demonstrated that the sensitivity of Gd-EOB-DTPA MRI was significantly higher than that of CE-CT. More precisely, this study investigated the diagnostic potential and precision of these imaging techniques on patients with only liver metastases from gastrointestinal and colorectal primitive cancers. The study performed by Australian Safety and Efficacy Register of New Interventional Procedures Surgical included nine diagnostic accuracy studies from 1991 to 2016, reassembling a total sample size population of 537 patients with a total of 1216 lesions. The median per-lesion sensitivity of Gd-EOB-DTPA MRI was 94.9% (range, 86.9–100.0%), and that of CE-CT was 74.2% (range, 51.8–84.6%), with statistical differences in favor of MRI (p < 0.001). Interestingly, this statistically significant advantage of Gd-EOB-DTPA MRI over CT in terms of sensitivity in detection of liver metastases does not make MRI poorer in terms of specificity (median 86.6% (range, 77.2–98%) for MRI and 94.1% (range, 80.2–98%) for CT; p = 0.44). A further per-lesion subgroup analysis for metastases other than those from colorectal cancers showed that Gd-EOB-DTPA MRI was 1.36 times more sensitive than CE-CT (95% CI 1.13–1.65; p = 0.001).Moreover, the authors [41] analyzed the imaging sensitivity according to the lesion’s diameter, dividing the patient population into two main groups: lesions <10 mm and lesions ≥10 mm. For lesions smaller than 10 mm, the median per-lesion sensitivity of Gd-EOB-DTPA MRI was 85.7% (range, 80.7–92.3%), and that of CE-CT was 50.0% (range, 26.0–64.5%). Therefore, the sensitivity of Gd-EOB-DTPA MRI appeared statistically superior to that of CE-CT (p < 0.001). As expected, these results were not confirmed for lesions larger than 10 mm, for which the median sensitivity of Gd-EOB-DTPA MRI (96.9%) remained higher than that of CE-CT (92.9%), but without statistical significance. These data confirm that Gd-EOB-DTPA MRI is more sensitive than CE-CT for detecting all liver metastases, but particularly those smaller than 10 mm [41].A milestone study in the diagnosis of liver metastases has been shown in a meta-analysis published in 2016 by Vilgrain et al. [42]. This study assessed the global diagnostic performance of DWI alone, Gd-EOB-DTPA MRI alone, and the combination of both techniques in the detection of liver metastases. The study included 39 articles, with a total final population of 1989 patients and 3854 overall lesions. The per-lesion sensitivities of DWI alone, Gd-EOB-DTPA MRI alone, and the combined techniques were 95.5%, 90.6%, and 87.1%, respectively, with a p-value < 0.001, comparing the combined technique with the other techniques alone, and Gd-EOB-DTPA MRI alone with DWI alone. These data were subsequently confirmed by other studies that analyzed these three MRI techniques [14,43]. The resultant high sensitivity of MRI justifies the use of this technique in all patients with liver metastases to ensure the most accurate preoperative planning [14,43].Moreover, another reason for the implementation of MRI and for its use in the staging of patients with cancers, including GC, also lies in the epidemiology of benign liver lesions. Benign hepatic tumors and tumor-like conditions have been demonstrated to occur in 52% of men with these lesions, and this rate is considerable because it does not include liver cysts [44]. Moreover, the number of benign lesions increases with the mean patient age [44]. This finding, discovered many years ago, is accurate; various studies about treating fortuitously discovered liver lesions have confirmed this high rate of benign lesions in the general population [45]. Therefore, it is mandatory to determine the correct stage for an oncological patient, to use an imaging technique with high diagnostic performance in detecting metastasis, and also to be able to correctly characterize hepatic lesions in both noncirrhotic and cirrhotic livers [46,47,48]. In the near future, if the data are confirmed, MRI could become the imaging of choice in the staging of oncological patients, including patients with GC (Figure 1 and Figure 2).MRI has been demonstrated to be an accurate technique for the staging of GC. In particular, the available results concerning the diagnostic performance of MRI in detecting peritoneal metastases from GC are promising, although the data are still limited. Further comparative studies between MRI and other imaging modalities for GC staging are needed in order to assess the possible added value of MRI in the management of patients affected by GC. From the point of view of liver metastases from GC, many studies have shown robust evidence that MRI with hepatospecific contrast media and with DWI currently represent the most accurate imaging technique. According to the confirmed results emerging from the literature, MRI is ready to become the “first choice” imaging in the staging of GC.Conceptualization, M.R. (Matteo Renzulli) and A.C.; Data curation, M.R. (Matteo Renzulli), A.C., G.M., and A.M.I.; Formal analysis, M.R. (Matteo Renzulli), A.C., D.S., S.B., I.P., M.R. (Matteo Ravaioli), and A.R.; Investigation, M.R. (Matteo Renzulli), A.C., and D.F.; Supervision, S.C., M.C., G.C., and R.G..; Validation, M.R. (Matteo Renzulli), S.C., M.C., G.C., and R.G.; Writing—original draft, M.R. (Matteo Renzulli), A.C., A.M.I., D.S., S.B., S.C., G.C., and R.G. All authors have read and agreed to the published version of the manuscript.No financial support nor any kind of sponsorship was received for conducting this study.The authors declare no conflicts of interest.A 62-year-old man complaining of dysphagia and epigastric pain with advanced stage gastric cancer. Axial (A) and coronal reformatted (B) contrast-enhanced computed tomography (CT) images showed a circumferential soft-tissue thickening involving the gastroesophageal junction and the cardia of the stomach (white arrows) invading the serosa layer (T-stage: T4a). Multiple (3) enlarged lymph nodes were appreciable in the aortocaval (C) (white arrow) and celiac (D) (arrowheads) region (N-stage: N2). Axial contrast-enhanced CT scan (E) did not demonstrate liver metastasis, whereas the axial 18FDG positron emission tomography computed tomography (PET-CT) image (F) revealed two metabolically active lesions with the same uptake value (SUVmax 4.7) located in the VIII hepatic segment (dotted arrows), consistent with M1 disease. The subsequent axial gadoxetic acid-enhanced magnetic resonance imaging, performed using diffusion-weighted imaging (G) and during the hepatobiliary phase (H), confirmed only one metastatic lesion (white arrow).Axial contrast-enhanced computed tomography (CT) image (A), in the same patient, did not show metastatic lesions at this liver level. The axial gadoxetic acid-enhanced magnetic resonance imaging, performed using diffusion-weighted imaging (B) and during the hepatobiliary phase (C), demonstrated a small liver metastasis in the liver segment VIII (white arrows). After six months, the axial contrast-enhanced CT image (D) confirmed the metastatic liver lesion (white arrow), enlarged.Eighth edition American Joint Committee on Cancer (AJCC) Tumor Nodes Metastasis (TNM) staging for gastric neoplasms with suggested imaging for staging (adapted from Amin et al. [12] and from the National Comprehensive Cancer Network (NCCN) Guidelines [13]).CT, computed tomography; MRI, magnetic resonance imaging; PET/CT, 18F-fluorodeoxyglucose positron emission tomography; EUS, endoscopic ultrasound.

Med-MDPI/cancers/cancers-12-06-01403.txt

@@ -0,0 +1 @@
+These authors contributed equally to this paper.Oncolytic adenoviruses (Ads) are promising tools for cancer therapeutics. However, most Ad-based therapies utilize Ad type 5 (Ad5), which displays unsatisfying efficiency in clinical trials, partly due to the low expression levels of its primary coxsackievirus and adenovirus receptor (CAR) on tumor cells. Since the efficacy of virotherapy strongly relies on efficient transduction of targeted tumor cells, initial screening of a broad range of viral agents to identify the most effective vehicles is essential. Using a novel Ad library consisting of numerous human Ads representing known Ad species, we evaluated the transduction efficiencies in four breast cancer (BC) cell lines. For each cell line over 20 Ad types were screened in a high-throughput manner based on reporter assays. Ad types featuring high transduction efficiencies were further investigated with respect to the percentage of transgene-positive cells and efficiencies of cellular entry in individual cell lines. Additionally, oncolytic assay was performed to test tumor cell lysis efficacy of selected Ad types. We found that all analyzed BC cell lines show low expression levels of CAR, while alternative receptors such as CD46, DSG-2, and integrins were also detected. We identified Ad3, Ad35, Ad37, and Ad52 as potential candidates for BC virotherapy.Cancer is the second leading cause of death in industrialized countries and constitutes an enormous burden on the health-care system. Among all cancer types, breast cancer (BC) is the most commonly diagnosed cancer (24.2%) and the leading cause of cancer deaths (11.6%) in women worldwide [1]. Treatment options of BC patients also depend on the grading and therefore, specific features of the BC type. Hormone receptor-positive (estrogen receptors, ER+; progesterone receptors, PR+) BC can be treated with hormone-blocking agents comprising selective estrogen receptor modulators or aromatase inhibitors. Estrogen receptor-negative (ER−) BC are primarily treated with radiochemotherapy targeting fast-replicating cancer cells [2]. However, chemotherapy also causes damage to physiologic, fast-growing cells. Since 25–30% of BCs overexpress the human epidermal growth factor receptor 2 (HER2), monoclonal antibodies such as trastuzumab and pertuzumab against HER2 have been developed to target the HER2 protein. These monoclonal antibodies prevent growth factors from binding and stimulating this receptor, thus, effectively blocking the growth of the cancer cells [3,4]. Triple-negative breast cancers (TNBCs) are characterized by a negative profile for all three markers above and account for 10–15% of all BCs. Furthermore, TNBCs usually display a more aggressive growth pattern and are associated with a poorer prognosis than other BC types. TNBC is typically treated with a combination of primary surgery, radiation therapy, and adjuvant chemotherapy. Despite this treatment regimen, TNBCs are associated with a high mortality [5,6], highlighting the need for further research in this patient group.As a novel therapeutic concept, oncolytic virotherapy lately has attracted considerable attention [7,8,9]. Virotherapeutics are based on two basic concepts: delivery of therapeutic genes to target tumor cells and tumor-selective replication of oncolytic viruses, which results in tumor cell lysis [10,11,12]. Virotherapeutics originated from clinical reports more than 100 years ago already demonstrated that cancer regression was coincidental with simultaneously occurring viral infections. Based on this observation, body fluids that contained human or animal viruses were used to transmit infections to patients with cancer in early clinical trials [13,14]. To improve the therapeutic effect, the characteristics of oncolytic viruses need to be optimized to achieve high efficacy and tumor specificity. During the past 30 years, tropism determinants of many different virus families have been identified and characterized. Most importantly, modern genetic engineering systems have been developed for almost every virus family, allowing the generation of viruses with improved oncolytic properties [15,16,17]. Finally, our understanding of cancer has also improved with the availability of diagnostic markers and detailed human genome information [2,18,19,20]. Although genetically modified viruses can also be explored to treat noncancerous diseases, cancer is the most often targeted disease as evidenced from the worldwide clinical trials database. Different viruses such as adenovirus (Ad), vaccinia virus, and herpes simplex virus have been evaluated in clinical setting to treat tumor patients.Ads are medium-sized (70–100 nm in diameter), nonenveloped icosahedral viruses composed of a nucleocapsid and a double-stranded linear DNA genome with an average length between 27 and 36 kb. In humans more than 100 Ad types have been identified which can be divided into 7 genetically diverse species (A–G). Although human Ads cause significant numbers of respiratory, ocular, and gastrointestinal diseases, incidences of severe diseases caused by Ads only occur in immunocompromised individuals. Among the general population Ad infections are usually resolved quickly and result in life long immunity to the virus [21].Since its first conversion from wild-type virus to a gene delivery vector in the beginning of 1990s [22], Ads have been recognized as the most efficient vehicles to deliver genes in vivo and as highly efficient tumor-killing (oncolytic) agents. Predominantly, oncolytic vectors derived from Ad type 5 were used in clinical trials. Although Ads can effectively transfer genes in vitro and in vivo, various limitations of the commonly used Ad5 exist. For instance, efficacy is frequently hampered by the high rates of neutralizing immunity, estimated as high as 90% in some populations, that promote vector clearance and limit bioavailability for tumor targeting following systemic delivery [23,24]. Active tumor targeting is also hampered by the ubiquitous nature of the Ad5 receptor, which is the human coxsackievirus and Ad receptor (CAR). As shown in previous studies, CAR expression levels were comparably low or were not detectable on many cancer cell lines, including the commonly used BC cell line MCF-7 [25,26,27].In this study, we compared over 20 Ad types representing species B1, B2, C, D, E, and G regarding their ability to transduce human BC cell lines and breast epithelia cells. Among these Ad types, Ad5 had the highest transduction efficiency in breast epithelia cells. However, this was not true for BC cell lines. Based on this finding, we further compared the oncolytic efficacy of the selected Ad types to Ad5. Here, we identified several novel Ad candidates showing considerably higher transduction rates than Ad5 in BC cells. Analysis of major Ad receptor expression levels on BC cell lines correlated with current knowledge of Ad receptor utilization by respective Ads and therefore confirmed our findings. To our knowledge, this is the first study evaluating a broad spectrum of human Ad types in a panel of human BC cell lines and breast epithelia cells for BC targeting. The results of this study have important implications for optimized Ad-based BC virotherapy.Twenty-two Ad types representing six Ad species were utilized in this study (Figure 1A). As previously described [28], these human Ads contain an expression cassette including turbo green fluorescent protein (tGFP), nanoluciferase (Nluc), and neomycin resistance (neo) located in the early gene region E3 in the reversed direction to the adenoviral genome (Figure 1B). For each experiment, the commonly used vector type Ad5 was applied as control. In the present study, we wanted to address the question which Ad types can efficiently transduce human BC cell lines, while only modestly transducing the breast epithelial cell line M13SV1. Furthermore, we investigated whether transduction efficiencies of Ads correlate with specific features of the BC type such as grading and receptor expression and which Ad types induce efficient tumor cell lysis. To answer these questions, we measured the transduction efficiency via a high-throughput screening based on luciferase assays, quantified GFP-positive cells using flow cytometry, investigated vector genome cell entry, and analyzed expression levels of Ad receptors and integrin coreceptor on each BC cell line. Finally, in a translational approach, we performed oncolytic assay to investigate the therapeutic potential of selected Ad types within the scope of breast cancer virotherapy.In a preceding study [28], Hs 578T BC cells were screened by applying the Ad library at different viral particle numbers per cell (vp/c) and measuring the transgene expression levels via luciferase assays. Ad3-, Ad16-, Ad50-, Ad35-, and Ad37-infected cells exhibited significantly higher luciferase expression levels if directly compared to controls infected with Ad5. In the current study, three additional breast BC cell lines (SK-BR-3, MCF-7, and MDA-MB231) and a breast epithelial cell line (M13SV1) were included to investigate the oncolytic potential of Ad vectors. BC cell lines and breast epithelial cells were infected with a broad range of virus dosages (20, 200, 2000, and 20,000 vp/c). After 24 h of incubation, nanoluciferase expression levels were measured via luciferase assays (Figure 2 and Figure S1). In all BC cell lines, there were always at least one or more Ad types exhibiting higher luciferase expression levels compared to Ad5.Interestingly, analyses of luciferase expression levels in another TNBC cell line (MDA-MB-231) revealed a similar trend as observed in Hs 578T cells, which was not the case in the other two analyzed BC cell lines, MCF7 and SK-BR-3. Ad3-infected MCF7 cells demonstrated an eightfold increased luciferase level compared to Ad5. All species B Ads and a few species D Ads (Ad17, Ad37, and Ad69) showed comparable or slightly higher luciferase expression levels than Ad5. However, in SK-BR-3 cells, only Ad3- and Ad35-infected cells revealed comparable or modestly higher luciferase expression levels than Ad5. In contrast to the results obtained in BC cell lines, Ad5 demonstrated the highest transduction efficiency among all tested Ad types in the breast epithelial cells M13SV1.High-throughput screening of Ads highlighted several Ad types potentially suitable for enhanced BC targeting. To further explore these selected Ads, BC cell lines were infected with respective Ads and the percentage of transgene-positive cells was quantified. Selected Ad types were applied to the four BC cell lines and one breast epithelial cell line (M13SV1) using 1000 vp/c. GFP expression was measured via flow cytometry 24 h postinfection and representative pictures of infected cells were collected (Figure 3 and Figures S2–S4). In both TNBCs, Hs 578T and MDA-MB-231, species G virus Ad52 revealed a significantly higher percentage of GFP-positive cells than Ad5. In MCF7 cells, infected with Ad3, Ad35, and Ad52, revealed a higher percentage of GFP-positive cells than those transduced with Ad5. However, in SK-BR-3 cells, 70% of Ad5-infected cells were positive for GFP expression. Other Ad types exhibited either a similar (Ad52) or slightly lower GFP expression (Ad3, Ad21, Ad35, and Ad37) than Ad5. In concordance with the results obtained in luciferase expression measurements, Ad5 again resulted in the highest level of GFP-positive cells among all analyzed Ad types in M13SV1 cells.In the next step, the cellular entry of selected Ad types was evaluated. Cells were infected with 1000 vp/c. Briefly, 3 h postinfection, cells were washed and collected to isolate total DNA for quantification of virus genome copy numbers using quantitative PCR (Figure 4). TNBC cell lines, Hs 578T and MDA-MB-231, showed a similar trend concerning the amount of internalized virus genome copy numbers. In both cell lines, Ad3 and Ad37 demonstrated significantly higher infection rates compared to Ad 5 at 3 h postinfection. In MCF7 cells, Ad3 displayed the highest infection rates, followed by Ad37, Ad35, and Ad20. SK-BR-3 cells infected with Ad37 revealed the highest efficiency with respect to genome uptake. Other species B and D Ads also demonstrated a greater amount of intracellular adenoviral genome copies compared to Ad5. When analyzing M13SV1 control cells, the tested Ad types showed comparable (Ad14 and Ad35) or slightly higher (Ad3 and Ad37) genome entry efficiencies than Ad5.Several major receptors used by different Ad types during the process of infection were identified in the past (Figure 1A). To understand the mechanisms behind cellular infection and transduction of Ads utilized in this study, the expression levels of major Ad receptors and integrin coreceptors were also examined for each cell line. Hela cells served as positive control in this experiment (Figure 5 and Figures S5 and S6). All examined BC cell lines exhibited relatively low or even no CAR expression, whereas the control cell line (M13SV1) displayed high levels of CAR expression. Concerning the expression of the CD46 receptor, tested cell lines demonstrated a high quantity of CD46-positive cells. Roughly, 50% of Hs 578T and SK-BR-3 cell were CD46 positive. However, analyses of MDA-MB-231, MCF7, and control cells (M13SV1) revealed that almost 100% of these cells were CD46 positive. The proportion of DSG-2 receptor-positive cells was 100% for the M13SV1 cell line, 50% in the SK-BR-3 cell line, and approximately, 20% in the cell lines Hs 578T, MDA-MB-231, and MCF7. With respect to coreceptors, we examined the most often studied integrins, αvβ3 and αvβ5. SK-BR-3 demonstrated the highest expression of both receptors among all tested BC-related cell lines. However, the αvβ3 integrin expression in all cell lines was relatively low compared to the other receptor types analyzed in this study (below 10%). Both TNBCs displayed lower number of positive cells compared to the other analyzed cell lines. It is of note that the integrin expression levels in the control cell line (M13SV1) were not higher than in the MCF7 and SK-BR-3 cell lines.Efficient cell entry and transduction can represent predetermining factors enhancing the therapeutic effect of virotherapy. However, whether the viruses that enter the cells and express their genes may, meanwhile, induce tumor cell lysis is still unclear. We infected each BC cell line and the control cell line M13SV1 with selected Ad types, which demonstrated efficient transduction and infection rates in experiments described in previous sections. Ads were applied in a 10-fold dilution manner with MOIs ranging from 10,000 to 1. During a time frame of 1–2 weeks, cell lysis was observed, and the remaining cells were stained to visualize the lytic effect of the viruses (Figure 6 and Figure S7). As shown, Ad52 exhibited the most distinctive oncolytic effect in Hs 578T cells, while the MDA-MB-231 cells were lysed by Ad5, Ad35, and Ad52 to a comparable degree. In MCF7 cells, Ad3, Ad35, Ad52, and Ad69 demonstrated a 100-fold higher potency to lyse BC cells compared to Ad5. Ad3, Ad5, and Ad52 lysed SK-BR-3 cells with a similar high efficiency. Ad3, Ad20, and Ad52 showed a comparable or slightly lower lytic effect compared to that of Ad5 in the M13SV1 cells which served as control.The primary goal of this study was to identify alternative Ad types for enhanced BC virotherapy. To achieve efficient Ad-based therapy for cancer, the very first aspect to consider is the cellular entry of Ad, which is dependent on the expression levels of the primary receptors on the targeted tumor cells. Furthermore, virus-derived transgene expression levels in targeted tumor cells also represent an essential factor when considering the insertion of tumor suppressing or immunomodulatory genes into the vector’s genome. However, the most important parameter is the vector’s potency to cause cell lysis in tumor tissue, which partly depends on transgene expression and the virus replication efficiency in the targeted tumor cells.In this study, the screening of a broad range of over 20 human Ad types was performed with four BC cell lines (Hs 578T, MCF-7, SK-BR-3, and MDA-MB231) and one breast epithelial cell line (M13SV1). A luciferase and GFP reporter gene expression cassette, which was previously cloned into the deleted E3 region of the applied Ad genomes enabled the rapid screening of Ads in a high-throughput manner [28]. In concordance to our previous observation regarding one TNBC cell line (Hs 578T) [28], in the current study, one or more Ad types showed higher luciferase expression levels than Ad5 in each BC cell line. In contrast to this observation, M13SV1 cells infected with Ad5 indicated highest luciferase expression levels correlating with efficient uptake and robust vector-derived transgene expression levels in this cell line. Based on this observation, we further filtered the broad range of utilized Ad types by selecting the Ads with the highest transgene expression levels. The identified 10 Ad types were additionally analyzed regarding their GFP expression utilizing flow cytometry in BC and breast epithelial cell lines. Notably, the results of the GFP measurements exhibited a similar trend compared to the previous results based on luciferase experiments, but this study also identified other BC-cell-line-specific Ad types, which may be converted into cancer-specific oncolytic Ads. Note that some discrepancies regarding luciferase expression levels and the percentage of GFP-positive and, therefore, transduced cells for single viruses were observed. However, one parameter quantifies total transgene expression levels in a population of cells and the other parameter measures the amount of transduced cells independent of transgene expression. GFP expression in a population of cells was also analyzed and we observed a similar trend with respect to transgene expression levels compared to luciferase expression levels. It has to be emphasized that not all Ad types were evaluated using all types of assays.Besides vector-derived transgene expression levels, receptor-dependent infection efficiency is another important aspect determining the potency of an oncolytic Ad. CAR was the first Ad receptor identified and serves as a primary functional receptor for Ad5. However, it has been previously demonstrated that the therapeutic effect of Ad5 is restricted to the low CAR receptor expression levels on the targeted tumor cells [25,26,27]. To improve the therapeutic efficacy of adenoviral vectors, one promising approach is to introduce fiber modifications by replacing the fiber knob from Ad5 with a knob from an alternative Ad type [25]. Here, we explored the broad spectrum of alternative Ad types as an alternative strategy. Ads vary greatly in their tissue tropism and pathologies due to their utilization of different receptors during infection. Besides CAR, other major Ad receptors, such as the cluster of differentiation 46 (CD46), the desmoglein-2 receptor (DSG-2), and sialic acid (SA), have been identified. Most species B and some species D human Ads were confirmed to use the ubiquitously expressed membrane protein CD46 as a primary cellular entry receptor [29,30]. DSG-2 is utilized by Ad3, Ad7, Ad11, and Ad14 as major entry receptor, whereas SA is the primary receptor for some species D virus and species G virus Ad52 [31,32,33,34,35]. Here, we identified Ad3, Ad35, Ad37, and Ad52 as potential candidates for virotherapy, and the receptor usage of these vectors provides important information for tumor targeting.In the internalization assay, we measured the amount of Ad genomes in transduced cells 3 h postinfection. However, other researchers have studied even shorter incubation times such as 1 or 2 h [36,37]. We speculate that this experimental setup has no influence on the outcome of the experiments, because genome replication probably, if at all, only occurs at low levels at this time point. In future studies, different incubation times postinfection should be examined for individual Ad types to monitor genome replication. For instance, in the current study, we observed that Ad52 exhibited low infection rates 3 h postinfection, but Ad52 also resulted in the highest percentage of transgene-positive cells. Ad37 demonstrated the highest viral load 3 h postinfection, but did not exhibit the highest transgene expression levels in respective experiments. To understand the difference in infection efficiency of respective Ads, the expression of major Ad receptors by BC cell lines was analyzed (Table 1). All BC cell lines utilized in this study demonstrated low CAR (around 5% for MDA-MB-231 and SK-BR-3) or almost no CAR expression (Hs-578T and MCF7). In contrast, more than 50% of the M13SV1 cells, which served as control, were CAR positive. CD46 and DSG-2 were detected on the surface of all BC cell lines, however, some BC cell lines demonstrated only minor expression of DSG-2 (20% for DSG-2 on Hs 578T, MDA-MB-231, and MCF7).Notably, other studies showed low levels of CAR expression in primary bladder, renal, and prostate cancers cells [38,39,40,41]. However, Martin et al. demonstrated that CAR expression is elevated in primary BC [42], and that CAR expression is positively correlated with a more undifferentiated tumor histology. Additionally, BC tissue, immunostained with CAR antibody, exhibited a more robust signal compared to other tissue types in the histopathologic sample. Moreover, Martin et al. could detect significantly elevated levels of CAR transcripts in breast tumors. Another study performed by Vindrieux et al. analyzed the link between CAR expression and estrogen signaling in BC [43]. This study revealed that CAR regulation by estrogens occurs at the transcriptional level and that the ectopic expression of CAR could increase the proliferation of BC cells. Moreover, Sakurai et al. demonstrated in a recent study that the BC cell lines MCF7 and MDA-MB-231 do express intermediate levels of CAR [44]. However, Sakurai et al. did not indicate the percentage of cells expressing CAR in their sample. Interestingly, a study conducted by Shashkova et al. observed minor CAR and CD46 receptor expression levels on BC cells compared with prostate cancer cells [16].Cellular entry of Ad has been previously used as decisive parameter for infectivity. Hoffmann and colleagues evaluated 20 Ads types in 2 primary tumor models. In the first soft tissue sarcoma model, several Ad types, such as 35, 3, 7, 11, 9, and 22, demonstrated higher internalization efficiency than Ad5 [36]. In a second malignant melanoma model, the expression of CAR and CD46 receptors in primary melanoma was analyzed. All the immunohistochemical staining of primary cutaneous melanoma lesions from five patients indicated positive CD46 expression, whereas CAR expression could not be detected. Notably, the in situ immunohistochemistry data could be confirmed by flow cytometry analysis of the short-term cultures prepared from these melanoma lesions. Similar to the first study, some Ad types (35, 38, 3, 49, 21, 34, and 7) demonstrated higher internalization efficiency than Ad5 in melanoma cells [37]. Interestingly, the identified Ad types (e.g., Ad35) featuring highest internalization rates can also result in enhanced cancer therapy, in both in vitro and in vivo models.To evaluate the ability of Ad to lysate BC cells, an oncolytic assay was performed. BC cell lines were infected with a dilution series of Ad types, which had demonstrated high infection and transgene expression rates in respective experiments. In concordance with results obtained in the other assays, Ad3, Ad35, and Ad52 demonstrated the most robust cell lysis efficiencies in all BC cell lines. Ad3 has a well-defined cell entry mechanism using DSG-2 and is already used in human clinical trials as oncolytic virus [45]. Ad35, which utilizes the CD46 receptor for infection, is well studied and broadly applied in gene therapeutic approaches and vaccine studies. Shashkova and colleagues studied the anticancer efficacy of Ad35, and they observed that although Ad35 had the highest cytotoxic effect in cell culture compared to other serotypes (Ad5, Ad6, and Ad11), its in vivo anticancer activity was fairly low [16]. Ad52 is yet the only discovered member of species G Ad. It is characterized by a close phylogenetic relationship with simian Ads and exhibits the lowest frequency of detection among humans [46]. With two different types of fibers, Ad52 can utilize both sialic acid-containing glycoproteins and the CAR receptor for binding to target cells [34,47]. Ad52 should be further studied to clarify the mechanisms underlying its oncolytic potency, as it may be a candidate for enhanced Ad-based BC therapy. It is of note, the Ad types used in this work are wild-type virus related Ads, in which only the early gene 3 (E3) was replaced with reporter genes (Figure 1). To apply the identified Ad types to cancer therapy, they need to be reconstructed to convert them into real oncolytic Ads. Several strategies that have been lately successfully applied for Ad5-based oncolytic Ads may be utilized to transform the wild-type-similar Ads used in this study. A very distinctive feature of oncolytic Ads is the ability to selectively replicate in tumor cells. This can be achieved by deleting viral genes that are essential for replication in normal cells, but the function can be complemented in cancer cells. For instance, E1B55K, which binds to cellular p53 and promotes G1/S transition, is an absolutely essential protein for Ad replication in normal cells while it is not needed in most tumors due to the dysfunction of the p53 pathway [48,49]. An alternative deletion can also be included in the conserved region of E1A (E1ACR2 domain). With a small 24 base pair (bp) deletion, the E1A loses its binding ability to the retinoblastoma (pRb) protein, thereby the disability of S-phase entry. Such 24-bp deleted Ads cannot replicate in normal cells, but instead only in cancer cells with deregulated cell cycle control [50,51]. Another strategy is the utilization of tumor-selective promotors, like the human telomerase reverse transcriptase (hTERT) promoter [52]. Such tumor-selective promoters can be used to control the virus replication by placing them upstream of the E1A gene. To take advantage of immunotherapy, it may also be advantageous to incorporate transgenes-expressing immune checkpoint proteins, such as PD-1 ligands, or immune-modulatory cytokine, like granulocyte-macrophage colony-stimulating factor (GM-CSF), interferon (IFN)-α, cluster of differentiation 40 ligand (CD40L), and interleukins (IL)-12 and -18 [53,54,55]. Conditional replication and immune stimulation will provide improved features to the Ads for tumor-specific oncolytic therapy.Regarding the correlation of obtained results from different assays, we believe that multiple factors contribute to the final oncolytic effect. However, factors directly or indirectly involved in the therapeutic effect especially in the clinic remain to be discovered. If we would develop a prediction formula for the oncolytic effect, we would like to suggest the following equation: oncolysis potency = (1) viral infection + (2) transgene expression + (3) viral replication + (4) lysis of target cells. Viral infection (1) depends first on virus receptor expression, e.g., the low CAR expression on most tumors demands the exploration of alternative Ad types with CAR-independent cellular entry mechanisms. It may also be influenced by the virus cellular entry time (e.g., Ad52). Transgene expression (2) is influenced by the promoter, the expression cassette, and importantly, the type-dependent transcription and expression levels of the transgene which may be influenced by the surrounding adenoviral sequences. Note that transgene expression is of special importance if an oncolytic virus is to be armed with immune-modulatory cytokines (e.g., IFN, GM-CSF) [56,57]. Viral replication (3) can be controlled by deletion of certain viral genes or tumor-selective promotors in conditionally replicating oncolytic Ads. In general, viral replication occurs in an Ad type-dependent manner. The lysis of target cells (4) can be achieved by expressing proteins inducing necrosis and apoptosis, but can also be initiated from the lytic properties of the used virus itself [58].In summary, the broad spectrum of naturally occurring Ads provides a great resource to identify suitable candidates for BC treatment. Of the over 20 analyzed Ad types, Ad3, Ad35, Ad37, and Ad52 demonstrated the greatest potential as innovative vectors in BC virotherapy (Table 2), which should be further modified using advanced genome engineering techniques to render these vectors tumor-cell-specific and be armed with features related to immunotherapy. These oncolytic viruses can be tested regarding their potency to kill cancer cells derived from patients in vitro or in the future, even in vivo in BC animal models.Recombinant adenoviral vectors expressing turbo Green Fluorescent Protein, nanoluciferase, and neomycin resistance (GLN) (Figure 1) were generated as previously described [3,59]. In detail, adenoviral genome isolated from purified adenoviruses or adenoviruses-infected propagation cells was cloned into p15A-based plasmid backbone via linear–linear homologous recombination [60]. The reporter cassette GLN was incorporated into the early transcription region 3 (E3) by linear–circular homologous recombination (LCHR) [61]. To rescue these vectors, the p15A backbone was removed with preinserted restriction enzymes; the linearized adenovirus genome containing aimed modification was transfected into HEK293 cells via calcium–phosphate transfection agent. After serial passaging to amplify the vectors to large scale, the vectors were purified by two round of cesium chloride (CsCl) gradient [62]. The physical titer of final vector particles was determined by spectrophotometry using 260 nm.BC cell lines MDA-MB-231, Hs 578T, and MCF7 were cultured in high-glucose Dulbecco’s Modified Eagle’s Medium (DMEM, PAN-Biotech, Aidenbach, Germany); Hs 578T and MCF7 cells were supplied with 0.01 mg/mL bovine insulin. BC cell line SK-BR-3 were cultured in McCoy’s 5A Medium. M13SV1 cells (kindly provided by James Trosko, Michigan State University, East Lansing, MI, USA [63]) were cultured in McCoy’s 5A Medium and further supplemented with 10 µg/mL human recombinant EGF, 5 µg/mL human recombinant insulin, 0.5 µg/mL hydrocortisone, 4 µg/mL human transferrin, and 10 nM β-estrogen (all chemicals were purchased form Sigma-Aldrich, Taufkirchen, Germany). All the above mediums were supplemented with 10% FBS (GE Healthcare, Solingen, Germany), 100 units per mL (U/mL) penicillin, and 100 µg/mL streptomycin (PAN-Biotech, Aidenbach, Germany). All cells were maintained in a humidified atmosphere at 37 °C and 5% CO2.The transduction efficiencies of Ads in different tumor cell lines were measured by determining reporter gene (luciferase) expression levels. Individual tumor cells were grown to confluency in 96-well tissue culture plates and infected with different viral partial numbers (vp) per cell. Precisely, 24 h after infection, luciferase activity was measured with the Nano-Glo Assay System (Promega, Mannheim, Germany), and luminescence was detected with a plate reader (Tecan, Crailsheim, Germany).To detect transduction efficiencies in different BC cell lines, cells were analyzed by flow cytometry. Precisely, 1 × 105 cells were seeded in 24-well plates. After full confluency was reached, cells were infected at 1000 vp/c and incubated overnight. Briefly, 24 h later, cells were washed once with PBS and detached with trypsin-EDTA. Cells were resuspended in Dulbecco’s minimal essential medium containing 10% FBS, centrifuged (1500× g, 3 min), and washed in DPBS before they were fixed with 2% formaldehyde. Fluorescence profiles were obtained by analyzing 10,000 viable cells on Beckman Coulter Gallios Flow Cytometer. Background signal was obtained by analyzing the negative control, which was uninfected cells. The percentage of GFP-expressing cells was determined by selecting a region of fluorescence above the background of auto-fluorescence from uninfected cells.To detect CAR (coxsackievirus and Ad receptor) expression on the cell surface of different BC cell lines using flow cytometry, 1 × 105 cells were washed with PBS supplemented with 1% BSA, centrifuged (1500× g, 3 min), and resuspended in 100 μL PBS/BSA and 5 μL PE-conjugated rabbit anti-hCAR antibody (Antibody Online, ABIN2649016). Following an incubation step at room temperature for 1 h, cells were washed again with PBS/BSA, to remove unbound antibodies, and resuspended in 100 μL PBS for flow cytometry using FACS (Beckman Coulter Gallios Flow Cytometer, Krefeld, Germany). As controls, each cell line without antibody was used. The PE-conjugated mouse antihuman CD46 antibody (12-0469-42, Thermo Fisher, Schwerte, Germany) was used to detect surface expression of CD46 on different cell lines. For DSG-2 detection, PE-conjugated mouse antihuman Desmoglein 2 antibody was used (CSTEM28, Thermo Fisher, Schwerte, Germany). For integrin detection, BV480 mouse anti-integrin αvβ5 (Clone ALULA, BD, Heidelberg, Germany) and BV650 mouse antihuman CD51/CD61 (Clone 23C6, BD) antibodies were used.To quantify the cell entry efficiency, a defined number of Ad particles (vp) was used to infect preseeded tumor cells and incubated for 3 h. Cell monolayers were digested and flushed off with trypsin, followed by extensive washing with PBS. Genomic DNA was extracted by incubation in TE buffer (10 mM Tris-HCl and 10 mM EDTA; pH 8.0) with 0.5% SDS and 0.5 mg/mL proteinase K. Subsequently, a phenol–chloroform extraction and ethanol precipitation were performed. To monitor virus genome uptake efficiency, quantitative real-time PCR (qPCR) detecting the transgene (GLN gene cassette) was performed.For quantification of the turbo Green Fluorescent Protein, nanoluciferase, and neomycin resistance (GLN) (Figure 1) reporter-tagged Ads, Primer pairs GLN-qPCR-fwd (ACC AAG CGA AAC ATC GCA TCG AG) and GLN-qPCR–rev (GCG ATA CCG TAA AGC ACG AGG AAG) binding to the transgene (GLN gene cassette) were used with a my-Budget 5× EvaGreen® QPCR-Mix II reagent (Krefeld, Germany) according to the manufacturer’s protocol. PCR cycle was run and detected in the CFX Connect Real-Time PCR Detection System from Bio-Rad (Duesseldorf, Germany).Oncolytic assays were performed in 48 well plates. A 10-fold dilution series of individual Ads was prepared freshly to infect preseeded cancer cells. Cytopathic effect (CPE) was checked daily until at least one of the viruses on one plate at the lowest dosage showed CPE. At the latest, after 14 days, the experiment was stopped. The cells were first fixed with 3.7% formaldehyde and then, maintained adherent cells were detected by staining of attached cells with 0.5% crystal violet dye. After several washing steps, the stained plates were photographed. To quantify the survived cells, 400 µL of methanol was added to each well and incubated for 20 min at room temperature on a bench rocker with a frequency of 20 oscillations per minute. The optical density of each well was measured at 570 nm (OD570) using a Tecan plate reader [64].Statistical analyzes were conducted with Microsoft Excel. Experimental differences were evaluated by Student’s t test.Overall, our study provides basic information on the therapeutic potential of a panel of human Ad types regarding BC virotherapy. Our finding suggests that alternative Ad types Ad3, Ad35, Ad37, and Ad52 transduce BC in a CAR-independent manner; these Ad types should be further engineered to oncolytic virus enabling tumor-specific replication and improved immunogenic response to the tumor environment.The following are available online at https://www.mdpi.com/2072-6694/12/6/1403/s1, Figure S1: high-throughput screening of the reporter-tagged human adenovirus (Ad) library in human breast cancer (BC)-related cells with different MOIs, Figure S2: mean fluorescence intensity 1 day postinfection, Figure S3: histograms of GFP-positive cells after virus infection, Figure S4: GFP images of individual adenovirus-infected breast cancer cells, Figure S5: mean fluorescence intensity of major adenovirus receptor expression levels on BC cell lines, Figure S6: histogram of major adenovirus receptor expression levels on BC cell lines, and Figure S7: oncolytic assay using cell viability analysis.Conceptualization, W.Z., T.D., and A.E.; investigation, N.M., J.G., L.S., and W.Z.; methodology, J.G. and E.E.-S.; formal analysis, J.G.; resources, S.J., T.D., and W.Z.; writing—original draft preparation, W.Z.; writing—review and editing, N.M., J.G., L.S., T.D., and A.E.; and funding acquisition, W.Z. and A.E. All authors have read and agreed to the published version of the manuscript.This research was funded by Internal Research Funds of the Witten/Herdecke University, IFF2017-17 (W.Z.). This work was also supported, in part, by DFG grant EH 192/5-1 (A.E.).We want to thank Claudia Hagedorn for providing support for the FACs analysis. We also thank Annika Bremke for technical assistance.The authors declare no conflicts of interest.Reporter-labelled human adenoviruses (Ads). (A) Overall, 22 adenovirus types (Ad type number) from 6 species were used in transduction of breast cancer (BC)-related cells. The previously reported receptor usage is indicated (√). (B) Construction of human Ad expressing turbo green fluorescent protein (tGFP), nanoluciferase (Nluc), and neomycin resistance (neo) under the control of the hybrid construct consisting of the cytomegalovirus (CMV) enhancer fused to the chicken beta-actin promoter (CAG) This cassette was inserted in the E3 genomic region. *, it is of note that Ad52 used here is only labelled with turbo GFP (tGFP)-neo. Therefore, Ad52 is not presented in luciferase-based assays.High-throughput screening of the reporter-tagged human adenovirus library in human BC-related cells. Transgene expression levels of different adenovirus types (Ad type number) were evaluated in BC-related cell lines. Cells were infected with each virus at various virus particle numbers per cell (vp/c). Shown are (A) MDA-MB-231 cells at 2000 vp/c, (B) MCF7 cells at 200 vp/c, (C) SK-BR-3 cells at 200 vp/c, and (D) breast epithelia cells M13SV1 infected with at 20 vp/c is used as control. Results from other vp/c are shown in the Figure S1. Luciferase expression levels were measured 24 h postinfection by addition of furimazine substrate and expressed as relative light units (RLU). Levels were compared to the commonly used adenovirus type 5 (Ad5) and indicated as fold change. Error bars, ±SEM of three independent experiments. * p < 0.05; *** p < 0.001; compared to Ad5 control.Number of GFP-positive cells after virus infection. Cells were infected with 10 Ads at 1000 viral particle per cell (vp/c), and GFP expression levels were analyzed 24 h postinfection by flow cytometry analyses. Uninfected cells (negative controls) were used to set the background gate below 1%. Percentage provided indicates percent of GFP-positive cells. A total of 10,000 viable cells were counted. (A–D) BC-originated tumor cell lines. (E) Breast epithelia cells M13SV1 are used as control. Error bars represent mean ± SD (n = 2).Virus internalization efficiency in BC cell lines. Cells were infected with individual viruses at 1000 viral particles per cell (vp/c) for 3 h to quantify internalized viral genome copy numbers (VCN), which were quantified by quantitative real-time PCR (qPCR) and expressed as VCN per cell. (A–D) BC-originated tumor cell lines. (E) Breast epithelia cells M13SV1 are used as control. Error bars represent mean ± SD (n = 2).Major adenovirus receptor expression levels on BC cell lines. Cells were stained with antibodies against coxsackievirus and adenovirus receptor (CAR) (A), cluster of differentiation 46 (CD46) (B), desmoglein-2 receptor (DSG-2) (C), and integrins αvβ3 and αvβ5 (D,E). Receptor expression was measured via flow cytometry, expressed as percentage of CAR, CD46, DSG-2, and integrin-receptor-positive cells. Hela cells were used as positive control. Unlabeled cells (negative controls) were used to set the background gate below 1%. A total of 10,000 viable cells were counted.Cytotoxicity in response to infection with the different Ad types at various multiplicity of infection (MOI) (0, 1, 10, 100, 1000, and 10,000 virus particles per cell (vp/c)). Cell cytotoxicity was assessed by crystal violet staining (Figure S7) and then furthermore, quantified by measuring the absorbance of solubilized crystal violet dye at 570 nm. The data is displayed as percentage (%) of viable cells (i.e., stained cells) exposed to infection in relation to viable cells in the control (ctrl, uninfected) sample.Breast cancer (BC)-related cell lines used in this study.Summary of adenovirus infection in breast cancer-related cell lines.

Med-MDPI/cancers/cancers-12-06-01404.txt

@@ -0,0 +1 @@
+Triple negative breast cancer (TNBC) is an aggressive breast cancer with historically poor outcomes, primarily due to the lack of effective targeted therapies. The tumor molecular heterogeneity of TNBC has been well recognized, yet molecular subtype driven therapy remains lacking. While neoadjuvant anthracycline and taxane-based chemotherapy remains the standard of care for early stage TNBC, the optimal chemotherapy regimen is debatable. The addition of carboplatin to anthracycline, cyclophosphamide, and taxane (ACT) regimen is associated with improved complete pathologic response (pCR). Immune checkpoint inhibitor (ICI) combinations significantly increase pCR in TNBC. Increased tumor infiltrating lymphocyte (TILs) or the presence of DNA repair deficiency (DRD) mutation is associated with increased pCR. Other targets, such as poly-ADP-ribosyl polymerase inhibitors (PARPi) and Phosphatidylinositol-3-kinase/Protein Kinase B/mammalian target of rapamycin (PI3K-AKT-mTOR) pathway inhibitors, are being evaluated in the neoadjuvant setting. This review examines recent progress in neoadjuvant therapy of TNBC, including platinum, ICI, PARPi, phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA) pathway targeted therapies, and novel tumor microenvironment (TME) targeted therapy, in addition to biomarkers for the prediction of pCR.Triple negative breast cancer (TNBC) accounts for 15% of all breast cancer and it is characterized by the lack of expression of estrogen receptor (ER)/progesterone receptor (PR)/human epidermal growth factor receptor-2 (HER-2), earlier recurrence, tendency of visceral metastasis, and worse overall survival [1,2,3]. The mainstay of treatment for early stage TNBC is neoadjuvant chemotherapy, followed by definitive surgery. Response to initial chemotherapy predicts clinical outcomes in breast cancer [4,5,6]. Neoadjuvant therapy has become increasingly used for the treatment of tumor ≥2 cm in standard-of-care clinical practice, and pathological response is routinely assessed for the evaluation of overall prognosis. Pathological complete response (pCR) was associated with better prognosis in neoadjuvant TNBC trials and has become a surrogate marker of survival [7,8]. The prognosis of TNBC is poor, particularly when pCR was not achieved [9].Conventional neoadjuvant chemotherapy regimen composed of adriamycin, cyclophosphamide, and paclitaxel (ACT) results in a pCR rate of 35–45% [8,10,11]. Tumor heterogeneity of TNBC is well recognized [12,13,14], yet molecular subtype driven therapy has not become a routine clinical practice, largely due to the lack of effective targeted therapies. Recent clinical trials incorporating immune checkpoint inhibitors (ICI) or targeted therapy, such as poly-ADP-ribosyl polymerase inhibitors (PARPi), may have the potential to personalize neoadjuvant therapy in early stage TNBC [15,16,17]. Other agents, including the phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA) pathway inhibitors and androgen receptor (AR) targeted therapies, may be applicable for certain molecular subtypes of TNBCs. Novel approaches in conducting neoadjuvant clinical trials, such as I-SPY 2 and ARTEMIS, may accelerate the progress to bring effective targeted therapies to the neoadjuvant setting [18,19,20,21]. In this review, we will summarize recent neoadjuvant trials focusing on the following perspectives: (a) chemotherapy optimization with the addition of carboplatin; (b) the addition of ICI to chemotherapy backbones; (c) clinical trial design by the evaluation of novel targeted agents such as I-SPY 2; (d) biomarker driven identification of clinically relevant patient subgroups to enable a more precise treatment approach (ARTEMIS). Many promising targeted therapies and approaches that are discussed in this review may lead to a paradigm shift of neoadjuvant therapies for early stage TNBC.A subset of TNBC is chemosensitive and 35–45% of patients achieve pCR despite the poor prognosis and aggressive nature. This might be explained by the tumor molecular heterogeneity of TNBC. Several molecular classifiers that are based on mRNA profiling of TNBC have been identified. Lehmann et al. reported TNBC-7 subtypes [12]: basal-like 1 (BL1), basal-like 2 (BL2), immunomodulatory (IM), mesenchymal (M), mesenchymal stem-like (MSL), luminal androgen receptor (LAR), and unstable (UNS). The BL1 subtype expresses genes that are related to cell cycle and proliferation, which is consistent with the elevated DNA damage response pathway observed in this subtype. This molecular property explains the robust response to neoadjuvant chemotherapy of the BL1 subtype [14], specifically sensitivity to platinum agents [22]. The BL2 subtype expresses genes that are associated with growth factor signaling. The IM subtype highly expresses immune cell signaling genes. The M and MSL subtypes express genes that are involved in cell motility, and the MSL subtype is also associated with cell differentiation. The LAR subtype demonstrates distinctive gene expression that is enriched in hormone-regulated pathways, such as AR signaling, and it is the least proliferative subtype resulting in enhanced chemotherapy resistance. The clinical relevancy of the TNBC 7-subtype was further investigated by determining pCR rates after neoadjuvant chemotherapy. Of 146 patients with TNBC, molecular subtype and pCR status were significantly associated (p = 0.04379) and TNBC subtype was an independent predictor of pCR status (p = 0.022) by a likelihood ratio test [23]. The BL1 subtype had the highest pCR rate (52%); BL2 and LAR had the lowest (0% and 10%, respectively). Similarly, in a study conducted by Santonja et al., 125 TNBC patients treated with neoadjuvant anthracyclines and/or taxanes +/− carboplatin showed BL1 tumors had the highest pCR to carboplatin containing regimens (80% vs. 23%, p = 0.027) and LAR tumors had the lowest pCR to all treatments (14.3% vs. 42.7%, p = 0.045 when excluding MSL samples) [22]. Later, these seven subtypes were refined into four types (TNBC type-4): BL1, BL2, M, and LAR with evidence of IM and MSL subtypes representing tumors with substantial infiltrating lymphocytes and mesenchymal cells, respectively. The BL1 subtype demonstrated the highest pCR rate of 40–50% [14]. Burstein et al. subdivided TNBCs into TNBC-4 subtypes: LAR, mesenchymal (MES), basal-like immunosuppressed (BLIS), and basal-like immune-activated (BLIA) [13]. The LAR subtype demonstrated molecular evidence of ER activation suggesting response to anti-estrogen or anti-androgen therapies, as described in Lehmann’s subtypes [12]. MES subtype was characterized by pathways of cell cycle, mismatch repair, and DNA damage repair. The BLIS subtype exhibited a downregulation of immune and cytokine pathways that are associated with the worst clinical outcomes. Contrary to BLIS, the BLIA subtype showed the best clinical outcomes with upregulated immunoregulation pathways. Despite the fact that TNBC subtyping provides an in-depth understanding of the tumor heterogeneity of TNBC [24,25,26], its clinical application has been limited due to the complexity of gene signatures. Table 1 summarizes molecular subtypes of TNBC and potential targets for therapies.Platinum salts are DNA damaging agents that show an increased efficacy in the tumors with a defected DNA repair system. The platinum salts react with DNA inside cells and distort the double helix of DNA inducing single-strand breaks (SSB) and double-strand breaks (DSB). When these damages cannot be efficiently repaired, it results in cell death [28]. These agents have shown activity in cancers with germline BRCA mutation, as BRCA 1/2 proteins have an essential role in repairing DNA damage [29,30,31].A high proportion of TNBC exhibits BRCA-like status (BRCAness), which indicates that these tumors are highly sensitive to platinum salts [32,33]. Platinum-based chemotherapy has been investigated in the neoadjuvant setting, with the goal to increase pCR and improve clinical outcome. Carboplatin-containing regimens demonstrated superior pCR rates when compared with standard regimen in two large randomized trials. The CALGB 40603/Alliance trial studied the clinical benefit of adding carboplatin +/− bevacizumab to neoadjuvant chemotherapy in stage II/III TNBC [34]. A total of 443 patients with stage II/ III TNBC received paclitaxel 80 mg/m2 weekly × 12, followed by dose dense doxorubicin and cyclophosphamide (ddAC) × 4, and were randomly assigned to concurrent carboplatin (AUC 6) once every three weeks × 4 ± bevacizumab 10 mg/kg every two weeks × 9. With carboplatin, the percentage of patients who achieved pCR increased significantly from 41% to 54% (odds ratio (OR), 1.71; p = 0.0029). The trial was not powered to detect long term overall survival (OS) and the addition of carboplatin to standard chemotherapy did not improve long term OS [35]. In the GeparSixto trial, 595 patients with stage II and III TNBC were randomized to receive either carboplatin or no carboplatin on a backbone regimen with paclitaxel, liposomal doxorubicin, and bevacizumab [36]. The pCR rates were significantly improved in the carboplatin group: 53.2% vs. 36.9 (p = 0.005). In both the CALGB 40603 and GeparSixto trials, hematological toxicities, including neutropenia and thrombocytopenia, were increased in the carboplatin group. The result from a meta-analysis of nine randomized controlled trials (RCTs) (N =  2109) showed that platinum-based neoadjuvant chemotherapy significantly increased pCR rate from 37.0% to 52.1% (OR 1.96, 95% confidential interval (CI) 1.46–2.62, p < 0.001) [37]. In addition, an increased pCR rate persisted after restricting the analysis to the three RCTs (N = 611) that used the same standard regimen in both groups of weekly paclitaxel (with or without carboplatin), followed by doxorubicin and cyclophosphamide (AC) (OR 2.53, 95% CI 1.37–4.66, p = 0.003). In two of the RCTs (N = 748) with survival data reported, no significant difference in event free survival (EFS) (hazard ratio (HR) 0.72, 95% CI 0.49–1.06, p = 0.094) and OS (HR 0.86, 95% CI 0.46–1.63, p = 0.651) were observed. Significantly increased grade 3/4 hematological adverse events (AEs) were observed with platinum-based neoadjuvant chemotherapy. Our single center phase II trial of carboplatin plus nab-paclitaxel (carboplatin AUC6 every four weeks × 4 and weekly nab-paclitaxel at 100 mg/m2 × 16 week) in stage II-III TNBC (N = 67) demonstrated a pCR rate of 48% with reasonable tolerability [38]. Sharma et al. reported a pCR rate of 55% with the combination of carboplatin and docetaxel (carboplatin AUC6, docetaxel 75 mg/m2 every three weeks × 6, N = 190) [39]. Table 2 shows the pCR rates from clinical trials that explored the efficacy of carboplatin.In addition to studies with carboplatin, cisplatin 75 mg/m2, every three weeks × 4 was evaluated in a randomized phase II study of neoadjuvant cisplatin vs. AC in germline BRCA carriers with HER2-negative (TBCRC 031) [46]. The pCR rate was 18% with cisplatin and 26% with AC, yielding a risk ratio (RR) of 0.70 (90% CI, 0.39 to 1.20).Other chemotherapy agents were evaluated to improve clinical outcomes in TNBC without compelling results. Nab-paclitaxel, an albumin-bound particle form of paclitaxel, has shown preferential tumor uptake and favorable safety profiles when compared to paclitaxel [47,48,49]. The clinical benefit from nab-paclitaxel in TNBC is still controversial [50,51]. When nab-paclitaxel was used instead of paclitaxel in phase III trial, the improvement of pCR was not statistically significant, showing pCR 41.3% with nab-paclitaxel vs. 37.7% with paclitaxel in the TNBC group (OR 0.85; 90% CI, 0.49–1.45) [52]. The GeparSepto trial that included about 20% of TNBC demonstrated significantly higher pCR in nab-paclitaxel subgroup than in the paclitaxel subgroup (48% vs. 26%, p = 0.00027) [53]. Additionally, nab-paclitaxel showed superior pCR when given with carboplatin as compared with gemcitabine [44].Introducing immunotherapy in the oncology field has changed the landscape of cancer treatment. Programmed death-1 (PD-1) is a T-cell inhibitory receptor that regulates the immune system by downregulating T-cell response upon binding with its ligand, programmed death ligand-1 (PD-L1) expressed on cancer cells. While the activation of this pathway prevents cancer cells from immune mediated cell death, the inhibition of PD-1 or PD-L1 can restore anti-tumor effects of T-cells. Among all of the breast cancer subtypes, TNBC is especially immunogenic, exhibiting increased expression of PD-L1 [54,55]. This immunogenicity observed in TNBC attracted ICI as a treatment option. PD-L1 inhibitor atezolizumab showed progression free survival (PFS) benefit in metastatic TNBC (mTNBC) and a complete response (CR) of 10%, which lead to the first ICI approval in TNBC [56]. ICIs have been investigated in several neoadjuvant trials in TNBC. The primary objective of these trials is to test pCR from adding ICI to chemotherapy, mainly taxane and anthracycline. Notably, carboplatin has been utilized when considering its efficacy in TNBC based on previous trials. KEYNOTE-173 is a phase Ib study that showed improved pCR rate from programmed cell-death 1 (PD-1) inhibitor pembrolizumab combined with neoadjuvant chemotherapy [16]. KEYNOTE-522 is a phase III study of neoadjuvant chemotherapy (paclitaxel and carboplatin, then doxorubicin or epirubicin and cyclophosphamide) combined with pembrolizumab or placebo, followed by adjuvant pembrolizumab or placebo in patients with TNBC [15]. An interim analysis reported significantly higher pCR rate in the pembrolizumab combined group (64.8% vs. 51.2%) than in the chemotherapy alone group, regardless of PD-L1 status. Event-free survival (EFS) was significantly higher in the pembrolizumab group during median follow up of 15.5 months. Grade 3 or higher AEs were 76.8% and 72.2%, respectively, with neutropenia as the most common serious AEs in both groups. This is the first phase III trial supporting the role of ICI in neoadjuvant and adjuvant treatment, and a long-term survival result is expected. In I-SPY 2, the overall pCR rate reached 60% when pembrolizumab was given with paclitaxel followed by AC [57].Durvalumab has also been evaluated for neoadjuvant treatment. In a phase I/II trial, durvalumab with nab-paclitaxel and sequential ddAC carried 55% of pCR in the PD-L1 positive group [58]. GeparNuevo trial randomized patients with TNBC to nab-paclitaxel with or without durvalumab [59]. All of the patients then received epirubicin and cyclophosphamide (EC) as neoadjuvant chemotherapy. Among 174 patients, the pCR rate with durvalumab was 53.4% vs. placebo 44.2%. Interestingly, this increased pCR rate was seen exclusively in patients that were treated with durvalumab alone before the initiation of chemotherapy (pCR 61%). In the NeoTRIP study (NCT002620280), 280 patients with TNBC were randomized to receive neoadjuvant carboplatin AUC 2 and nab-paclitaxel at 125 mg/m2 intravenously (IV) on days 1 and 8 with or without atezolizumab at 1200 mg intravenously on day 1 [60]. The pCR rate was 43.5% (95% CI, 35.1%–52.2%) with atezolizumab and 40.8% (95% CI, 32.7%–49.4%) without atezolizumab in the intent-to-treat population, which led to an odds ratio of 1.11 (95% CI, 0.69–1.79; p = 0.066). In this study, 49% of patients had cT2 disease, 59% had cN1 nodal status, and 56% were PD-L1 positive. The role of atezolizumab in the neoadjuvant setting is currently being investigated in the GeparDouze/NSABP B-59 (NCT03281954) trial. TNBC patients will receive neoadjuvant atezolizumab combined with chemotherapy (carboplatin plus paclitaxel and AC or EC), followed by adjuvant atezolizumab [61]. Clinical trials evaluating ICIs in early stage TNBC are described in Table 3.Based on these encouraging pCR rates, ICIs may play an important role in neoadjuvant therapy for TNBC and eventually become standard-of-care for a subset of TNBCs. However, well defined biomarkers for the better identification of appropriate patients remain lacking. Long-term survival benefits from adding ICIs need to be evaluated in order to adopt ICIs as neoadjuvant treatment in clinical practice. The tumor microenvironment (TME) is associated with immune suppression, escape from immune detection and development of drug resistance, and is being increasingly recognized as a potential target for treatment of TNBC [63]. Tumor associated macrophages (TAMs) promote the progression and metastasis of TNBC by releasing inhibitory cytokines, reducing functions of tumor infiltrating lymphocytes (TILs), promoting TREG (regulatory T-cells), and modulating PD-1/PD-L1 expression in TME [64]. TME targeted therapies are undergoing active clinical trial investigation. The combination of cabiralizumab, an antibody that inhibits the colony stimulating factor-1 receptor (CSF1R) and it blocks the activation and survival of macrophages, and ICI with neoadjuvant chemotherapy might improve efficacy by decreasing TAMs and increasing TILs in early stage TNBC [65,66]. Currently, cabiralizumab is being used in combination with nivolumab and neoadjuvant chemotherapy in patients with localized TNBC (NCT04331067).DNA repair deficiency and PARP inhibitors: BRCA 1/2 mutation is one of the greatest genetic risk factors of developing breast cancer. BRCA 1 and BRCA2 are tumor suppressor genes that play a major role in the DNA repair system, specifically in homologous recombination, which repairs double-stranded breaks (DSBs). When homologous recombination does not function (homologous recombination deficiency, HRD), commonly seen in cases of BRCA 1/2 mutations, DSBs result in genomic instability. Poly ADP-ribose polymerase (PARP) 1 is a protein that binds to single stranded breaks (SSBs) during the DNA repair process. PARPi traps PARP1 and induces cell death by preventing SSB repair, followed by DSBs without functional homologous recombination in patients with BRCA mutations. PARPi showed efficacy in patients with BRCA mutations. The OlympiAD trial is a randomized phase III trial that compared olaparib with standard chemotherapy in patients with metastatic breast cancer and germline BRCA mutation [67]. The significantly longer PFS shown in the olaparib group was more prominent in the TNBC (HR 0.43 in TNBC vs. 0.82 in non-TNBC) and made up of 50% of this study. Talazoparib is aPARPi that is approved for advanced breast cancer with BRCA mutation through the EMBRCA trial [68]. In this trial, patients with germline BRCA mutation were randomized in a 2:1 ratio to receive talazoparib or single-agent therapy of the physician’s choice. Among a total of 287 patients who received talazoparib, 45% patients were TNBC. The median PFS was significantly longer in the talazoparib group. The efficacy of PARPi was further investigated in the neoadjuvant setting. I-SPY 2 trial studied the neoadjuvant PARPi veliparib and carboplatin followed by AC as compared with standard neoadjuvant chemotherapy (paclitaxel followed by AC) [69]. The estimated pCR rate in TNBC was 51% in the veliparib-carboplatin group vs. 26% in the control group. This study demonstrates that patients with TNBC can benefit from PARPi and carboplatin as neoadjuvant treatment. However, the grade 3 or 4 hematologic toxicity was much higher in the veliparib-carboplatin group than in control group. BrighTNess trial is a phase III randomized trial to confirm clinical benefit of adding veliparib and carboplatin in TNBC [45]. A total of 634 patients with stage II-III TNBC were randomized (2:1:1) to receive veliparib/carboplatin/paclitaxel, carboplatin/paclitaxel, or paclitaxel, followed by AC after the randomized portion. The pCR rate in veliparib/carboplatin/paclitaxel group (53%) was significantly higher than the paclitaxel group (31%), but adding veliparib to carboplatin/paclitaxel did not improve pCR (58%). Grade 3–4 hematology toxicity was significantly increased from adding carboplatin, regardless of using veliparib. Several studies have been conducted to evaluate the role of PARPi as neoadjuvant treatment in early stage BRCA mutated or HRD breast cancer. Most patients enrolled in these studies are TNBC. MD Anderson reported a pilot study of neoadjuvant talazoparib in patients with germline BRCA mutations [70]. Fifteen of 20 enrolled patients were TNBC, and 50% achieved pCR after six months of single agent talazoparib. Hematologic toxicity was the most common AE with 40% Grade 3 anemia and 15% Grade 3 neutropenia. This enhanced pCR from single agent talazoparib offers a different approach for patients with early stage BRCA mutated TNBC. A confirmatory trial is currently ongoing in order to verify the benefit of single agent talazoparib (NCT02401347). GeparOLA trial (NCT02789332) is a phase II randomized trial to evaluate neoadjuvant paclitaxel and olaparib in patients with HRD [71]. The study randomized 107 patients, including 77 TNBC, to either weekly paclitaxel and daily olaparib or weekly paclitaxel and carboplatin for 12 weeks, and then followed with EC. Interestingly, an improved pCR rate in the olaparib group was achieved in the hormone receptor-positive group (29 patients). In TNBC, the olaparib group showed a pCR rate of 56% and the carboplatin group showed 59.3%. Carboplatin is a DNA damaging agent, and works in a similar way to PARPi by inducing DNA damage through DSBs in HDR resulting in a comparable pCR. Table 4 summarizes clinical trials utilizing PAPRi. The benefit of adding PARPi to neoadjuvant chemotherapy is still controversial for all TNBCs, despite its correlation with genomic instability. The improved pCR from adding carboplatin (also a DNA breaking agent) makes the use of PARPi more controversial. Large randomized trials are needed to determine whether adding carboplatin, PARPi, or both can achieve better pCR. Importantly, the toxicities from adding carboplatin or PARPi should be considered as increased hematologic AEs were observed in previous trials.Phosphatidylinositol-3-kinase (PI3K)/AKT/ mammalian target of rapamycin (mTOR) signaling is the most commonly activated cancer driver pathway, leading to cell proliferation and survival (Figure 1). The mutation of PIK3CA, the gene encoding the subunit p110α of PI3K, or deactivation of phosphatase and tensin homolog (PTEN), negative regulator of PI3K, can contribute to the progression of cancer [72,73,74]. PIK3CA mutation is found in 20–40% of breast cancer and it is associated with increased resistance to chemotherapy [75,76,77]. The inhibition of this pathway has been actively investigated [78], and the mTOR inhibitor everolimus and the PI3K-α inhibitor alpelisib were FDA approved for hormone receptor-positive metastatic breast cancer [79,80]. The alterations of the PI3K/PTEN/AKT pathway (including PIK3CA mutations, PTEN inactivating mutations, and AKT1 activating mutations) occur in 25% of primary TNBC and possibly at a modestly higher frequency in mTNBC [81,82,83]. In the phase II LOTUS trial, the patients were randomly assigned (1:1) to receive intravenous paclitaxel 80 mg/m2 (days 1, 8, 15) with either ipatasertib, a pan-AKT inhibitor at 400 mg or placebo once per day (days 1–21) every 28 days. Median PFS in the intention-to-treat (ITT) population was 6.2 months (95% CI 3.8–9.0) with ipatasertib versus 4.9 months (3.6–5.4) with placebo (HR 0.60, 95% CI 0.37–0.98; p = 0.037). In the 48 patients with PTEN-low tumors, the median PFS was 6.2 months (95% CI 3.6–9.1) with ipatasertib vs. 3.7 months (range 1.9–7.3 months) with placebo (HR 0.59, 95% CI 0.26–1.32, p = 0.18) [84]. The Phase III IPATUNITY130 trial (NCT03337724), where patient receive either paclitaxel and ipatasertib or paclitaxel and placebo, will confirm survival benefit [85]. The AKT inhibitor Capivasertib showed significantly longer PFS and OS in mTNBC when added to paclitaxel as first line treatment of mTNBC [86]. Ipatasertib has been studied in neoadjuvant TNBC in a phase II neoadjuvant FAIRLANE study of weekly paclitaxel plus ipatasertib or placebo with the following endpoints: pCR rate, PTEN-low population assessed via IHC and PIK3CA/AKT1/PTEN-altered tumors using next generation sequencing (NGS) [87]. The addition of ipatasertib showed a small increase in pCR rate of 17% vs. 13% in ITT. The clinical response rate by breast MRI of ipatasertib was numerically improved, but not statistically significant in the biomarker-selected patients: PTEN-low tumors (32% vs. 6%) and PIK3CA/AKT1/PTEN-altered tumors (39% vs. 9%). In the adaptive neoadjuvant phase II I-SPY 2 trial, the AKT inhibitor MK-2206 plus standard neoadjuvant chemotherapy of weekly paclitaxel followed by AC achieved an estimated pCR rate of 40% when compared with 22% from chemotherapy alone in the TNBC subgroup [88]. These results support further evaluation of AKT inhibition + paclitaxel and AC neoadjuvant chemotherapy in patients with PIK3CA/AKT1/PTEN-altered tumors.AR is a potential therapeutic target considering 10–40% of TNBC express AR of 1 to 10% of stained tumor cells [89,90,91]. The efficacy of AR inhibitors has been studied in AR-positive mTNBC. The AR inhibitor enzalutamide has demonstrated clinical benefit rate at 16 weeks of 33% in mTNBC with AR ≥ 10% [92]. Abiraterone, an inhibitor of 17α-Hydroxylase/C17,20-lyase (CYP17) required enzyme for androgen biosynthesis, had modest objective responsive rate (ORR) of 6.7%, PFS of 2.8 months and six month clinical benefit rate (CBR) of 20% in AR-positive (≥10% IHC) mTNBC [93]. Single agent AR targeted therapy appears to be modest, and combination therapy with other targeted agents are currently under investigation. Enzalutamide plus paclitaxel neoadjuvant therapy is currently ongoing (NCT02689427). Enzalutamide in combination with taselisib (NCT02457910) or alpelisib (03207529) trials in mTNBC are actively accruing patients. While immunotherapy is successful across a variety of tumor types, biomarkers precisely predicting response to therapy remain to be identified. Understanding the tumor immune microenvironment holds promise for optimal cancer therapy. TILs and PD-L1 and are the most commonly used biomarkers to evaluate the response to ICI. The presence of stromal tumor infiltrating lymphocytes (sTILs) is widely recognized as a good prognostic factor in both adjuvant and neoadjuvant chemotherapy [94,95,96]. Loi et al. reported that higher levels of TILs were associated with decreased distant recurrent in TNBC, and improved disease free survival (DFS) and OS [97]. Two pooled analyses with a large number of patients demonstrated that increased TILs predict pCR and improved survival in TNBC. The German Breast Cancer Group analyzed pretreatment core biopsies from 3771 patients for sTILs following the guidelines of the International TIL working group [98]. TILs were predefined in three groups: low (0–10%), intermediate (11–59%), and high TILs (≥60%). Increased TIL percentile predicted response to neoadjuvant chemotherapy in TNBC: pCR was achieved in 80/260 (31%) of patients with low TILs, 117/373 (31%) of patients with intermediate TILs, and 136/273 (50%) of patients with high TILs (p < 0.0001). A 10% increase in TILs was associated with longer DFS in TNBC (HR 0.93 (95% CI 0·87–0·98), p = 0.011) and longer overall survival in TNBC (HR 0·92 (95% CI 0·86–0·99), p = 0.032). These findings were reproduced in a different pooled analysis with 2148 patients from nine studies for adjuvant chemotherapy [99]. Mean sTILs was 23% and increased sTILs were significantly associated with improved survival: HR for a 10% increase in sTILs was 0.83 (95% CI, 0.78–0.87) for distant DFS and 0.83 (95% CI, 0.79–0.88) for OS. sTILs significantly decreased in metastatic TNBCs as compared with matched primary [100,101]. Higher TIL PD1 expression was associated with better prognosis in early stage TNBCs [102]. These results further support the approach of introducing ICIs early in the neoadjuvant or adjuvant setting, since primary tumors are more immunogenic. PD-L1 expression on tumor cells or immune cells has been evaluated as a biomarker of treatment response to anti-PD-1 or anti-PD-L1 therapies [103,104]. Measuring PD-L1 expression remains controversial due to different methods and antibodies. The expression of PL-L1 in TNBC was estimated to be 40–65% on immune cells [105,106]. Mittendorf et al. reported 19% (20 /105 TNBC) of tumor cells were PD-L1 positive, defined by >5% of membranous staining by IHC [54]. In IMpassion 130 trial, intratumoral CD8 correlated with PD-L1 immune cell expression, and was therefore predictive of prolonged PFS (HR, 0.74; 95% CI, 0.61–0.91) and OS (HR, 0.66; 95% CI, 0.50–0.88) with atezolizumab and nab-paclitaxel vs. placebo and nab-paclitaxel [107]. sTILs were not well correlated with PD-L1 immune cell expression, and only predicted prolonged PFS with atezolizumab when compared with placebo (HR, 0.66; 95% CI, 0.50–0.86). There is a lack of quantitative association between PD-L1 expression and response. Indeed, the response to ICI is not linearly associated with increasing levels of expression, and the methods and antibodies used for PD-L1 assessment remain controversial [108]. It has been observed that PD-L1 negative patients may still derive benefit from ICIs. The knowledge gap in PD-L1 testing across different trials needs to be mitigated in order to best characterize patients who might benefit from ICIs.In addition to TILs and PD-L1, multi-gene signatures have been studied as a more comprehensive tool capturing the immunogenicity of TNBC. The GeparSixto trial was analyzed for mRNA markers from pretherapeutic formalin-fixed paraffine embedded core biopsied samples [109]. A GeparSixto immune signature (GSIS) composed of seven immune-activating genes (CXCL9, CCL5, CD8A, CD80, CXCL13, IGKC, CD21) and five immunosuppressive (IDO1, PD-1, PD-L1, CTLA4, FOXP3) genes was validated as a marker for immune reaction. GSIS revealed that the increased mRNA expression level of these genes, including immunosuppressive genes, was associated with pCR. In our neoadjuvant carboplatin and nab-paclitaxel trial, GSIS was significantly associated with pCR and residual cancer burden (RCB) in a multivariate model (submitted) [38].Using laser capture microdissection gene expression profiles, the tumor immune microenvironment (TIME) was captured and subclassified from therapy-naïve TNBC tumors. An “immune hot” TIME exhibited tumor infiltration of granzyme B+CD8+ T cells (GzmB+CD8+ T cells), a type 1 IFN signature, and elevated expression of multiple immune inhibitory molecules, including indoleamine 2,3-dioxygenase (IDO) and PD-L1, was associated with good outcomes. An “immune-cold” TIME with an absence of tumoral CD8+ T cells was defined by elevated expression of the immunosuppressive marker B7-H4, signatures of fibrotic stroma, and was associated with poor outcomes [110]. This laboratory approach appears to be labor-intensive and might not be easily adapted in the clinic. Recent advanced technology can capture functional HRD beyond BRCA 1/2 mutations (BRCAness) that shares molecular features of BRCA alteration with a scoring system. myChoice HRD® by Myriad Genetics (Salt Lake City, UT, USA) is a commercially available test for assessing HRD. This is a NGS-based in vitro test that determines genomic instability that is based on an algorithmic scoring system of loss of heterozygosity (LOH), telomeric allelic imbalance (TAI), and large-scale state transitions (LST) [111]. TNBC is strongly related with BRCA mutation and HRD. Among all of the patients with TNBC, 10–15% of patients have germline BRCA 1/2 mutations [112,113] and 40–60% of patients are positive for HRD [114,115,116]. Other genes that are involved in the HR process, such as PALB2 and RAD51, have been discovered to play an important role in TNBC [117]. TNBC with BRCA mutation or HRD is more sensitive to chemotherapy or PARPi [111,118,119,120]. However, evaluating HRD has not been standardized in clinical practice. Currently there are few commercially available methods to evaluate HRD, and this needs to be further studied before being used in clinical practice. Several ongoing trials have been evaluating the addition of novel targeted therapy agents to standard chemotherapy in the neoadjuvant setting, including I-SPY 2 and ARTEMIS trial. The I-SPY 2 trial utilizes an adaptive design for evaluating the addition of novel agents to paclitaxel, followed by AC (P-AC) in high-risk early stage breast cancer [121]. The addition of veliparib to P-AC had an estimated pCR of 51% [69] and adding pembrolizumab to P-AC had an estimated pCR rate of 60% [57]. Although improved pCR is encouraging, the addition of veliparib or pembrolizumab has not been conclusively shown to improve long-term outcome. This might be attributed to the small sample size.Precision medicine based on the genomic tests has been adopted in clinical trials. In the metastatic setting, the utility of genomic mutation driven therapies has been tested in basket trials, such as NCI-MATCH (Molecular Analysis for Therapy Choice), which contains a multi-arm design with each arm testing a single drug on a histology-agnostic fashion [122,123]. Despite the appealing concept of precision medicine for management of metastatic breast cancer, the implementation of such approaches in the neoadjuvant setting remains challenging. While TNBC patients with pCR/RCB-0 or RCB-1 have better survival, those with extensive residual disease (RCB-II or RCB-III) after neoadjuvant chemotherapy (NACT) have poor prognoses [124,125,126]. The ARTEMIS (NCT 02276443) is a randomized phase II trial to determine whether precision neoadjuvant therapy (P-NAT) impacts the rates of pathologic response (RCB 0–I) while using a CLIA-certified chemosensitivity mRNA gene signature (GES) and subtyping of TNBC by IHC to select targeted therapy trials for chemotherapy-insensitive tumors [19,20,21]. The initial study plan was to randomize 360 patients with TNBC as 2:1 ratio to “know” vs. “not know” P-NAT. Chemotherapy-sensitive tumors received chemotherapy, and chemotherapy-insensitive tumors were enrolled in a clinical trial. The first interim analysis (N = 133 patients with RCB status) revealed a RCB 0–1 rate of 56% (“know” P-NAT) vs. 62% (“not know” P-NAT); p = 1.0; thus, randomization was discontinued for futility [19]. A total of 232 patients were enrolled, including 168 evaluable for RCB. In the ultrasound-resistant cohort (N = 43), RCB 0–I rates were higher in patients treated with targeted therapy (N = 30) vs. AC-T (N = 13); (30% vs. 8%; odds ratio = 5.1 with 95% CI, 0.6–45.7; p = 0.11). GES failed to improve the rates of RCB 0–I in TNBC; however, in patients with resistant disease identified by ultrasound after AC, RCB 0–I rates were higher in patients that were treated with targeted therapy as compared to chemotherapy alone. This trial again demonstrated a persistent gap between tumor biology and the clinical application of precision medicine in the neoadjuvant setting.The optimization of a neoadjuvant chemotherapy regimen in early stage TNBC continues to evolve. The key question remains to be the appropriate selection of a neoadjuvant regimen based on patient and disease characteristics. Recently, the promising pCR rate that was reported from KEYNOTE 522 using pembrolizumab plus carboplatin/paclitaxel followed by AC potentially shifts the standard-of-care regimen for early stage TNBC neoadjuvant therapy toward more intensive chemotherapy backbone, such as weekly carbo/taxol followed by AC, although significant Grade 3–4 AEs raised the question of whether every patient requires such an intensive regimen. There is existing evidence showing the carboplatin/taxane regimen remains highly active and it could serve as a chemotherapy backbone for immunotherapy or targeted therapy combinations (de-escalated chemotherapy). Recent trials demonstrated a promising pCR rate of “anthracycline-free regimen”. Our single center phase II trial of carboplatin plus nab-paclitaxel (carboplatin AUC6 every four weeks × 4 and weekly nab-paclitaxel at 100 mg/m2 × 16 week) in stage II-III TNBC (N = 67) demonstrated a pCR rate of 48% with reasonable tolerability [38]. Sharma et al. reported a pCR rate of 55% with the combination of carboplatin and docetaxel (carboplatin AUC6, docetaxel 75 mg/m2 every three weeks × 6, N = 190) [39]. WSG-ADAPT-TN trial reported by Gluz et al. also reflected a de-escalation concept with a 12 week neoadjuvant regimen [44]. When patients were treated with carboplatin AUC2 with nab-paclitaxel 125 mg/m2 on day 1 and 8 for four three-week cycles, the pCR rate was 44.9 (N = 154). The expression of immunological genes (CD8, PD-L1), basal-like mRNA expression profile, and high Ki-67 were associated with pCR in a multi-variate model (p < 0.05) [127]. All three trials are consistent with a favorable toxicity profile and high efficacy using carboplatin and taxane based anthracyclin-free regimen. These data support further research while using de-escalated chemotherapy backbone for combination therapy with ICI or targeted therapy. Currently, the ongoing NeoPACT trial combining carboplatin AUC6, docetaxel 75 mg, and pembrolizumab 200 mg every three weeks is actively enrolling patients, and the results of the trial result are eagerly awaited (NCT03639948). Confirmatory analysis of biomarkers predicting patients who can achieve pCR without the use of anthracycline and/or ICIs is critical for patient selection.Recent progress has been made in neoadjuvant therapy for early stage TNBC. ICI and PARPi may become standard-of-care for appropriate subtypes of TNBC. Carboplatin remains an important treatment in BRCA-associated tumors or HRD tumors. Novel clinical trial design, such as I-SPY 2 or ARTEMIS, might vastly facilitate testing novel targeted therapy in the neoadjuvant setting. The many promising targeted therapies and approaches that are discussed in this review may lead to a paradigm shift of neoadjuvant therapies for early stage TNBC. Conceptualization, J.S.L. and Y.Y.; writing—original draft preparation, J.S.L.; writing—review and editing, S.E.Y. and Y.Y.; visualization, S.E.Y.; supervision, Y.Y. All authors have read and agreed to the published version of the manuscript.This research received no external funding.Y.Y. has contracted clinical trials and research projects sponsored by Novartis, Eisai, Pfizer, Merck, Genentech, and Puma. Y.Y. is an advisory board of Immunomedics, Pfizer, Genentech, and Novartis and a speaker bureau of Eisai, Novartis, AstraZeneca, Genentech, and Daiichi. The other authors declare no conflict of interest. Mechanisms of PI3K/AKT/mTOR pathway activation and targeted therapies. Activating mutations in the α catalytic domain of PI3K and/or PTEN mutation lead to pathway activation. PI3K signaling pathway linking RTK signaling leads to downstream activation of PI3K/AKT/mTOR, promoting cell proliferation and survival. RTK receptor tyrosine kinase, PI3K phosphatidylinositol-3-kinase, PTEN phosphatase and tensin homolog, AKT Protein Kinase B, mTORC mechanistic target of rapamycin complex.Triple negative breast cancer (TNBC) molecular subtype and potential targets for therapy.Modified from Lehmann et al. [12,14] and Collignon et al. [27];AXL, tyrosine-protein kinase receptor UFO; CSC, cancer stem cells; EGFR, epidermal growth factor receptor; EMT, epithelial-mesenchymal-transition; FGFR, fibroblast growth factor receptors; IGF-1R, insulin-like growth factor receptor; IL, interleukin; MET, hepatocyte growth factor; mTOR,, mammalian target of rapamycin; NGF, nerve growth factor; PARP, poly ADP ribose polymerase; PDGFR, platelet-derived growth factor receptors; PD1, programmed cell death 1; PD-L1, programmed death-ligand 1; PI3K, phosphatidylinositol 3-kinase; TGFβ, transforming growth factor beta. Pathological complete response (pCR) rate in neoadjuvant trials with carboplatin in early stage TNBC.E: epirubicin; C: cyclophosphamide; T: docetaxel; P: paclitaxel; NPLD: non-pegylated liposomal doxorubicin; Bev: bevacizumab; A: doxorubicin; dd: dose-dense; G: gemcitabine; N/A: not applicable *pCR in the both breast and axilla (ypT0/is ypN0).Immune checkpoint inhibitor trials in early stage TNBC.P: paclitaxel; Cb: carboplatin; A: doxorubicin; C: cyclophosphamide; E: epirubicin; T: docetaxel; dd: dose-dense; N/A: not applicable. *pCR in the both breast and axilla (ypT0/is ypN0).Poly-ADP-ribosyl polymerase (PARP) inhibitors in neoadjuvant TNBC trials.P: paclitaxel; Cb: carboplatin; AC: doxorubicin and cyclophosphamide; EC: epirubicin and cyclophosphamide; N/A: not applicable. *pCR in the both breast and axilla (ypT0/is ypN0).

Med-MDPI/cancers/cancers-12-06-01405.txt

@@ -0,0 +1 @@
+These authors contributed equally.Survival rates for Ewing sarcoma (ES) patients with metastatic disease have not improved in over 20 years. Tumor growth and metastasis are dependent on tumor vasculature expansion; therefore, identifying the regulators that control this process in ES may provide new therapeutic opportunities. ES expresses high levels of repressor element 1 silencing transcription factor (REST), which is regulated by the EWS-FLI-1 fusion gene. However, the role of REST in ES growth and the regulation of the tumor vasculature have not been elucidated. To study this role, we established REST-knockout human TC71 ES cell lines through CRISPR/Cas9 recombination. While knockout of REST did not alter tumor cell proliferation in vitro, REST knockout reduced tumor growth and metastasis to the lung in vivo and altered tumor vascular morphology and function. Tumor vessels in the REST-knockout tumors had a punctate appearance with significantly decreased tumor vascular pericytes, decreased perfusion, and increased permeability. REST-knockout tumors also showed increased apoptosis and hypoxia. These results indicate that REST plays a critical role in ES vascular function, which in turn impacts the ability of ES tumors to grow and metastasize. These findings therefore provide a basis for the targeting of REST as a novel therapeutic approach in ES.Ewing sarcoma (ES) is the second most common malignant bone tumor in children and young adults, and the lung is the most common site of metastasis. Increasing the dose and frequency of chemotherapy administration has improved the survival rate to 75% for patients with a tumor in an extremity and no detectable metastases [1,2,3,4,5]. However, there have been no major advancements in treatment for more than 10 years. The survival for patients presenting with metastases is much worse (<25%) and has not improved in over 30 years [6,7]. Salvage chemotherapy protocols for relapsed disease are ineffective, as patients typically die within 1 year of relapse. Tumor vessels play an important role in providing nutrients and oxygen to the tumor, which are required for sustained growth. Understanding the mechanisms that contribute to the successful formation of functional tumor vessels in ES and understanding how ES tumor cells contribute to this process, may allow the identification of new therapeutic targets that inhibit vascular expansion.The origin of ES is still controversial. Several studies suggest that its origin is in the neural crest [8,9,10]. This origin would indicate that genes that play a role in neuronal tumors may also contribute to the tumorigenesis of ES. Repressor element 1 silencing transcription factor (REST) is a neuronal repressor gene that regulates neuronal stem cell differentiation [11,12]. REST also plays a multifunctional role in the regulation of non-neurogenic cells [13,14]. In tumor growth, REST has a dual function depending on the cellular context. We have previously shown that REST is upregulated in both ES cell lines and patient samples and that EWS-FLI-1, the hallmark fusion gene protein in ES tumors, regulates REST expression [15]. We also showed that inhibition of REST expression by shRNA reduced tumor growth and increased tumor hypoxia and apoptosis without modifying the expression of EWS-FLI-1 or decreasing cell proliferation in vitro. Inhibiting REST also altered the morphology of the tumor vessels. While neither CD31 nor VEGF expression in the REST shRNA ES tumors was decreased, tumor vascular pericyte coverage decreased significantly. This decrease was associated with increased hypoxia and tumor cell apoptosis. Pericytes are important for vascular stabilization and efficient blood flow, suggesting that REST in ES is critical for maintaining vascular perfusion and that the decreased tumor growth in the REST shRNA tumors was due to an effect on the tumor vasculature rather than a direct effect of the tumor cells.The REST shRNA expression vector that we used previously decreased REST expression by 50–70% [15]. However, we were unable to completely downregulate the gene or maintain complete inhibition in vivo. To confirm and extend these investigations focusing on the role of REST in tumor vascular expansion and function, we used CRISPR/Cas9 technology to knock out REST expression in ES cells. In addition to analyzing the effect of REST knockout (KO) on tumor growth and vascular morphology, we evaluated the effect of REST KO on lung metastasis, vascular perfusion, and vascular permeability. These data confirmed that REST plays a critical role in maintaining and expanding ES tumor vessels that are required for tumor growth. These findings together with our previous studies show that interfering with vascular formation and expansion severely retards not only tumor growth but also metastasis to the lung [16,17,18,19,20,21]. REST is therefore a potential target for ES therapy.We previously showed that inhibiting REST expression in TC71 and A4573 ES cells using shRNA slowed tumor growth and altered vascular morphology [15]. To further investigate the effects of REST on ES tumor vasculature, TC71–REST-KO clones were generated using CRISPR/Cas9 recombination. Five single-guide RNA (sgRNA) oligonucleotides targeting five specific human REST genomic regions were synthesized and cloned into CRISPR/Cas9 vector pX458-U6-chimeric_BB(+85)-CBh-NLS-hSpCsn1-2A-GFP. The sgRNA expression clones were verified by restriction enzyme digestion and sequencing. TC71 cells were either transfected with a single sgRNA expression vector for single-nicking recombination or cotransfected with two sgRNA expression vectors for double-nicking recombination. Double-nicking CRISPR/Cas9 by cotransfection of two sgRNA expression vectors had much higher knockout efficiency than single-nicking recombination (Supplementary Materials Figure S1A,B). Western blot analysis confirmed that REST protein expression was dramatically reduced in several different CRISPR/Cas9–REST-KO clones compared with DAOY positive control cells and the TC71 parental cells; densitometry analysis demonstrated that REST protein expression was decreased by at least 80% in the R1106 and R1606 REST-KO cells compared with the parental cells (Figure 1A).Immunofluorescence staining confirmed that REST expression was significantly inhibited in the R11106 and R1606 cells compared with the TC71 parental and TC71 CRISPR/Cas9 recombination clone with normal expression of REST (RC-control) cells (Figure 1B).The inhibition of REST had no effect on cell proliferation. The in vitro doubling time did not significantly differ between TC71 parental cells, RC-control cells, and REST-KO clones R1106 and R1606: 22.0 h (standard deviation, ±5.0), 26.0 ± 5.0 h, 21.7 ± 6.9 h, and 26.7 ± 6.1 h, respectively.To determine the effect of REST KO in vivo, TC71 parental cells, RC-control cells, and R1106 or R1606 REST-KO cells were injected into the tibias of mice. Twenty-four days after injection, R1106 and R1606 REST-KO tumors were significantly smaller on average than RC-control or TC71 parental tumors (Figure 2A). Immunofluorescence staining of the tumor samples for REST expression confirmed that REST was significantly reduced in the R1106 and R1606 tumors compared with the RC-control tumors (Figure 2B). Expression of the proliferation marker Ki67 was also quantified in the tumor samples. Echoing our in vitro findings, the Ki67 expression levels did not significantly differ between the REST-KO tumors and the RC-control tumors (Figure 2C). These results suggest that the decreased growth in R1106 and R1606 REST-KO tumors was not the result of decreased tumor cell proliferation. R1106 and R1606 REST-KO tumors also showed decreased metastatic potential to the lung (Figure 2D).Tumor growth requires a robust functional vasculature. Since the growth of ES tumors in vivo was not explained by an effect on tumor cell proliferation, we next evaluated the tumor vasculature morphology in the R1106 and R1606 REST-KO and RC-control tumors, first by staining with CD31. Vessels in the RC-control tumors showed open lumens, whereas the vessels in the R1106 and R1606 REST-KO tumors had a punctate morphology (Figure 3A, white arrows).Since pericytes are critical for vascular stabilization and maintaining an open lumen, we next evaluated tumor vessel pericyte coverage. Using CD31 to identify tumor vessels and the pericyte marker NG2, we evaluated both the total number of pericytes in tumors and the ratio of pericytes to tumor vessel cells (Figure 3B, yellow color highlighted by white arrows identifies the double positive staining). The total number of pericytes was significantly reduced in the REST-KO tumors compared with the RC-control tumors (Figure 3B,C). In addition, the ratio of NG2-positive to CD31-positive cells was significantly decreased in the REST-KO tumors (Figure 3B,D). These findings suggest that REST regulated the tumor vascular morphology by decreasing pericyte coverage.Decreased pericyte coverage on tumor vessels can affect vascular perfusion and permeability. Since REST KO was associated with a decreased NG2:CD31 ratio (Figure 3D), we next evaluated vascular function by assessing the effect of REST KO on perfusion and vascular permeability in the R1606 tumors and RC-control tumors. A DyLight 594–labeled lectin perfusion assay was used to evaluate the effect of REST down-regulation on tumor vascular perfusion. Consistent with our data in Figure 3A, vascular morphology was altered in the R1606 REST-KO tumors; the tumor vessels had a punctate appearance with few open lumens (Figure 4A). In addition, the mean perfused area in the R1606 REST-KO tumors was significantly decreased compared with that in RC-control tumors (Figure 4B, p < 0.01). To determine the percentages of functional perfused and non-functional vessels in the different tumor samples, CD31 staining (fluorescein isothiocyanate (FITC)) was used to identify all vessels, and co-localization of lectin and CD31 was used to identify functional vessels (Figure 4C, yellow color indicated by white arrows shows the double positive staining). The percentage of functional perfused vessels among all vessels in R1606 REST-KO tumors was significantly decreased compared with that in RC-control tumors (Figure 4D, p < 0.01).These data demonstrate that down-regulation of REST decreased tumor vascular perfusion.In addition to decreasing vascular perfusion, decreased pericyte coverage increases vascular permeability, which is an indication of decreased vascular function. To determine whether REST KO affected vascular permeability, four weeks after tumor cell injection mice bearing TC71-RC-control, R1106, and R1606 tumors were injected intravenously with green FITC-labeled dextran (which leaks from permeable vessels into the tumor tissues) five minutes before the mice were sacrificed. Normal vessels are not permeable to this high molecular weight dextran. Decreased vascular pericytes will result in increased vascular permeability, which in turn will increase the leakage of the FITC labeled dextran into the tumor tissues. CD31 staining of the tumor samples also was performed as before. CD31-positive vessels in R1106 and R1606 REST-KO tumors again had a punctate and irregular morphology compared with the RC-control tumors (Figure 5A). FITC-labeled dextran was increased in R1106 and R1606 tumors compared with RC-control tumors. Quantitative analysis confirmed that down-regulation of REST increased tumor vessel permeability, as FITC-labeled dextran was significantly increased in R1106 and R1606 tumors compared with RC-control tumors (Figure 5B, p < 0.01).Decreased vascular perfusion and increased vessel permeability have been shown to increase tumor tissue hypoxia and apoptosis. We therefore quantified tumor hypoxia and apoptosis in the REST-KO and RC-control tumors. Hypoxia-inducible factor-1α (HIF-1α) was used to detect hypoxic areas. HIF-1α expression was significantly increased in both R1106 and R1606 REST-KO tumor tissues compared with RC-control tumors (Figure 6A,B).A terminal deoxynucleotidyl transferase dUTP nick end labeling (TUNEL) assay showed that apoptosis was also significantly increased in the R1106 and R1606 tumors compared with control tumors (Figure 6C,D).Taken together, these data indicate that inhibition of REST decreased the number of functional tumor vessels and increased vascular permeability, thus increasing tumor hypoxia and apoptosis.There is an urgent need for new therapeutic approaches for patients with ES, particularly those with metastatic or relapsed disease. Both the primary tumor and the metastases require a functional vascular system. Inhibiting vascular expansion compromises blood delivery, which hinders tumor growth and may therefore be an alternative therapeutic approach to treat tumors that do not respond or relapse after chemotherapy. Developing such therapies requires that we understand and define the molecular mechanisms that regulate tumor blood vessel expansion and functionality. This understanding is critical for identifying the potential targets that interfere with effective tumor vascular formation, blood flow, and oxygenation. We previously demonstrated that the REST gene is overexpressed in ES tumors and is regulated by EWS-FLI-1 [15,22]. In this study, we show that down-regulating REST using CRISPR/Cas9 impacts tumor growth and metastasis by targeting tumor blood vessel structure and function. While knockout of REST did not alter tumor cell proliferation in vitro, REST knockout reduced tumor growth and metastasis to the lung in vivo and altered tumor vascular morphology and function. Tumor vessels in the REST-knockout tumors had a punctate appearance with significantly decreased tumor vascular pericytes, decreased perfusion, and increased permeability. REST-knockout tumors also showed increased apoptosis and hypoxia. These results indicate that REST plays a critical role in ES vascular function, which in turn impacts the ability of ES tumors to grow and metastasize.Our previous studies demonstrated that a portion of ES tumor vascular pericytes are derived from bone marrow progenitor cells [22], that VEGF165 was involved in the chemoattractant recruitment of these bone marrow cells to the tumor for formation of new blood vessels, and that the Notch-DLL4 signaling pathway regulated their differentiation into vascular pericytes [19,20,21]. Together with our previous report [15], the data presented here implicate REST in the regulation of the ES tumor vasculature.We used CRISPR/Cas9 to knock out REST, creating two REST-KO clones with ≥80% reduction in REST expression, confirmed in vitro and in tumor samples (Figure 1A,B). Down-regulation of REST had no effect on cell proliferation in vitro but significantly inhibited tumor growth and metastasis in vivo and altered both the morphology and the function of the tumor vessels.We also demonstrated a significant reduction in vascular pericytes and vascular perfusion in the REST-KO tumors. The ratio of pericytes to endothelial cells was significantly decreased. Pericytes play a critical role in blood vessel functionality and maturation, protection of endothelial cells, and regulation of endothelial cell viability and proliferation. Without pericytes, vessels are leaky and poorly perfused [23]. Indeed, loss of pericyte coverage in the REST-KO tumor vessels decreased vessel perfusion and increased vascular leakage. Pericyte depletion in tumor vessels also has previously been shown to induce tumor hypoxia [24]. Indeed, the decrease in tumor vascular pericyte coverage in the REST-KO tumors was associated with increased tumor hypoxia and apoptosis. Taken together, these data suggest that the vascular effects induced by REST KO, rather than an effect on cell proliferation, resulted in the reduced tumor growth and metastatic potential. These in vivo findings of decreased pericyte coverage of the tumor vessels in the REST-KO tumors are supported by our previous in vitro studies showing that REST regulated the expression of the pericyte markers desmin and NG2 in ES cells [15].REST was originally described as a neuronal repressor gene that regulates neuronal stem cell differentiation [11,12]. Later data demonstrated that REST targets multiple genes in non-neuronal systems [13]. Recent studies indicated that REST modulates the vasculature in diffuse intrinsic pontine glioma (DIPG). REST has been shown to up-regulate the pro-angiogenic molecule gremlin-1 [25]. Inhibition of REST in DIPG caused a substantial decline in tumor vasculature, as measured by decreased CD31 and VEGFR2 staining, and reduced tube formation. Our results are in line with these findings but also show that the down-regulation of REST reduced both endothelial cells and the tumor vascular pericytes.We have previously shown that tumors from ES patients express high levels of REST [15] and that REST controls the transdifferentiation of ES stem cells into pericytes in response to hypoxia [22]. Therefore, REST inhibition may compromise the ability of ES to contribute to the needed pool of vascular pericytes for tumor vascular expansion in an effort to recover from damage and cell death caused by chemotherapy or radiation therapy. A better understanding of the molecular mechanisms that are involved in tumor vascular formation and expansion, which impact both tumor growth and metastasis, may reveal new specific targets for anti-cancer therapy. Taken together with our previous work, our new results confirm that REST is a critical regulator of ES tumor vasculature integrity and function, pointing to REST as a new therapeutic target for ES.TC71 human ES cells were cultured in Dulbecco modified Eagle medium with 10% fetal bovine serum and authenticated by short terminal repeat fingerprinting at the Cytogenetics and Cell Authentication Core facility, The University of Texas MD Anderson Cancer Center. Human medulloblastoma cell line DAOY was purchased from the American Type Culture Collection. The cells were cultured in ATCC-formulated Eagle’s Minimum Essential Medium with 10% FBS according to the manufacturer’s instructions. All the cells were mycoplasma free as determined by the MycoAlert Mycoplasma Detection Kit (Lonza, Rockland, ME, USA).Five pairs of DNA oligos were synthesized by Sigma-Aldrich (St. Louis, MO, USA) for REST-KO sgRNA expression, as follows:REST-RY1F: caccgGTTATGGCCACCCAGGTAAT and REST-RY1R: aaacattacctgggtggccataacc; REST-RG2F: caccgAGACATATGCGTACTCATTC and REST-RG2R: aaacgaatgagtacgcatatgtctc; REST-RG4F: caccgCGCACCTCAGCTTATTATGC and REST-RG4R: aaacgcataataagctgaggtgcgc; REST-RG5F: caccgCAACAGTGAGCGAGTATCAC and REST-RG5R: aaacgtgatactcgctcactgttgc; REST-RG6F: caccgGTCTTCTGAGAACTTGAGTA and REST-RG6R: aaactactcaagttctcagaagacc.Annealed double-stranded DNA fragments were cloned into a pX458-U6-chimeric_BB(+85)-CBh-NLS-hSpCsn1-2A-GFP vector (kindly provided by Dr. L. Copper from the Department of Pediatrics Research, MD Anderson Cancer Center, Houston, TX), and sgRNA expression clones were verified by restriction enzyme digestion and sequencing. TC71 cells were either transfected by a single sgRNA expression vector for single-nicking recombination or cotransfected by two sgRNA expression vectors for double-nicking recombination. After 48 h of transfection, single GFP expression cells were sorted into 96-well cell culture plates (one cell per well) with complete culture medium for continuing culture. Each single-cell culture was expanded, collected, and analyzed for REST expression by Western blotting analysis. REST-KO single-cell clones were selected for further in vitro and in vivo experiments. A random chosen clone with similar expression of REST as the parental TC71 cells was named TC71-RC-control and used as recombination control.Four- to five-week-old athymic (T-cell deficient) nude mice were purchased from the National Cancer Institute and maintained in a specific pathogen-free animal facility approved by the Association for Assessment and Accreditation of Laboratory Animal Care International. The animal experiment protocol was approved by the Institutional Animal Care and Use Committee of MD Anderson Cancer Center (IACUC No. 00001400-RN01). TC71 parental cells, TC71-RC-control cells (with normal REST expression), R1106 REST-KO cells, and R1606 REST-KO cells in the mid-log growth phase were harvested by trypsinization. Cell suspensions (0.5 × 106 cells in 0.01 mL of Hank’s balanced salt solution) were intra-tibially injected into the right leg of the nude mice. At day 24 after injection, the maximum diameters and 90° diameters of the tumors on the right leg were measured with a caliper and recorded. Tumor volume was calculated by the formula (a/2)2 × b minus the volume of the left leg, measured and calculated the same way, where a and b are the two largest diameters. The right leg was amputated at 4 weeks, and mice were maintained until 15 weeks after injection, when the mice were sacrificed, lungs were harvested, and visible lung tumor nodules were counted and recorded.Lectin binds to glycoproteins located in the glycocalyx and in the basal membrane of endothelial cells. Intravenous injection of DyLight 594–labeled Lycopersicon esculentum (tomato) lectin is perfused through blood flow and bound inside the vessels. Functional vessels display red fluorescence color from bound DyLight 594–labeled tomato lectin, which can be visualized and quantified [26,27]. The endothelial marker CD31 shows total vessels, including both the perfused and non-perfused vessels. Thus, 24 days after the subcutaneous injection of RC-control or R1606 REST-KO cells (5 mice per group), each mouse was injected in the tail vein with 100 uL of DyLight 594–labeled tomato lectin (Vector Laboratories). The mice were euthanized 5 min later, and tumor tissues were separated and frozen in optimal cutting temperature medium (OTC) on dry ice. The tumor samples were sectioned and mounted in Vectashield HardSet with 4′,6-diamidino-2-phenylindole (DAPI) (VECTOR LABORATORIES, Burlingame, CA, USA). Perfused vessels were visualized as red tumor vessels under Leica fluorescence microscopy (Leica Microsystems, Buffalo Grove, IL, USA). The tumor tissues were then stained with FITC-labeled CD31 (green). Functional vessels were indicated by yellow fluorescence, which is the co-localization of FITC-labeled CD31 green and lectin-perfused red vessels. Co-localization of lectin and CD31 double-positive vessels and total CD31-positive vessels were quantified in 10 randomized fields from different tumor samples using SimplePCI software (Hamamatsu Photonics, Bridgewater, NJ, USA). The percentage of perfused vessels among total vessels was calculated.Dextran labeled with FITC (Sigma-Aldrich) was used to assess vessel permeability. Normal blood vessels are not permeable to substances with high molecular weight, including dextran (2 × 106 kDa); when vessel permeability increases, dextran leaks to the tissue outside of the vessels [28]. Therefore, 100 μL of FITC-labeled dextran (10 mg/mL) was intravenously injected into the mice bearing RC-control, R1106 REST-KO, or R1606 REST-KO tumors 5 min before the mice were sacrificed (three groups tumor-bearing mice, 5 mice per group). The harvested tumor tissues were then frozen and sectioned, and CD31 immunofluorescence staining was performed. Texas Red-conjugated anti-rat IgG was used as the secondary antibody. The FITC-labeled dextran area (green) outside of vessels (red, CD31-positive) indicated leaky blood vessels. The FITC-positive area was quantified in 10 different microscopic fields from different tumor samples using SimplePCI software (Hamamatsu Photonics, Bridgewater, NJ, USA), and average positive area was calculated.Different CRISPR/Cas9–REST-KO clones, DAOY human medulloblastoma cells, and CRISPR/Cas9 RC-control cells (as positive control) were cultured in 100-mm diameter dishes. Cell lysate was collected. The protein (40 μg each lane) was loaded onto 8% sodium dodecyl sulfate–polyacrylamide gel. Specific protein bands were detected with anti-human REST (Millipore, Burlington, MA, USA) and β-actin (Sigma-Aldrich, St. Louis, MO, USA) antibodies. Densitometry analysis was performed, and the values were normalized with β-actin as loading control.Frozen tumor sections were fixed with acetone and chloroform. The sections were incubated with REST antibody (Millipore); NG2, HIF-1α, or Ki67 antibody (Abcam, Cambridge, UK); or rat anti-mouse CD31 antibody (BD Biosciences, San Jose, CA, USA). Anti-rabbit cyanine 5, anti-rat Texas Red, or FITC was used as the secondary antibody. All stained slides were analyzed by fluorescence microscopy (Leica Microsystems). Relative expression was quantified in at least five different microscopy fields from different tumor samples using SimplePCI software (Hamamatsu), and the average expression was quantified.All values are reported as means ± SD (standard deviation). A two-tailed Student t-test was used to statistically evaluate the in vitro experimental results. p < 0.05 was considered statistically significant. One-way ANOVA analysis of variance was used for statistical analysis of animal experiment results by Graphpad Prism 8 software (San Diego, CA, USA).In conclusion, the inhibition of REST reduced tumor vessel pericytes and altered vasculature morphology and functionality. Tumor vessels from REST-KO tumors were punctate and poorly perfused and showed increased permeability and leakiness. The vascular changes were associated with increased tumor hypoxia and apoptosis, as well as decreased tumor growth and metastasis. These results indicate that REST may be a key regulator of ES vascular development and expansion and that inhibiting REST could compromise the ability of the tumor to grow and metastasize by compromising the tumor’s vascular function. Inhibiting vascular function by blocking REST may also compromise the tumor’s ability to recover and regrow following chemotherapy or radiation therapy, as the oxygen and nutrients required for recovery may not be delivered. Therefore, the critical vascular functions that are controlled by REST provide a rationale for considering REST as a potential target for the treatment of ES.The following are available online at https://www.mdpi.com/2072-6694/12/6/1405/s1, Figure S1: Representative WB analysis of CRISPER Clones.Conceptualization, Z.Z., Y.Y. and E.S.K.; Data curation, Z.Z., Y.Y., F.W. and E.S.K.; Formal analysis, Z.Z., Y.Y. and E.S.K.; Funding acquisition, E.S.K.; Investigation, Z.Z., Y.Y. and F.W.; Methodology, Z.Z., Y.Y. and F.W.; Project administration, E.S.K.; Resources, Y.Y. and E.S.K.; Software, Z.Z. and Y.Y.; Supervision, E.S.K.; Validation, Z.Z. and Y.Y.; Visualization, Z.Z., Y.Y., F.W. and E.S.K.; Writing—original draft, Z.Z., Y.Y. and E.S.K.; Writing—review and editing, Y.Y. and E.S.K. All authors have read and agreed to the published version of the manuscript.This work was supported in part by the Kayton Ewing’s Sarcoma Research Fund, the Mary V. and John A. Reilly Distinguished Chair (Eugenie S. Kleinerman, MD), and the National Cancer Institute P30CA016672 institutional core grant.The authors declare no conflicts of interest.Assessment of REST in the TC71 CRISPR/Cas9–REST-KO clones. (A) Western blot analysis of REST protein expression in several CRISPR/Cas9–REST-KO clones. Densitometry indicated REST protein expression was reduced by at least 80% in the R1106 and R1606 clones compared with the TC71 parental cells. DAOY human medulloblastoma cells were used as the positive control. (B) Immunofluorescence staining of REST expression in RC-control cells and the R1106 and R1606 clones. REST expression (red) was significantly down-regulated in R1106 and R1606 cells compared with the RC-control cells. DAPI was used for nuclear staining. Scale bar: 50 µm.Down-regulation of REST inhibited tumor growth. (A) Totals of 5 × 105 TC71 parental, RC-control, R1106, or R1606 cells were injected into the tibias of nude mice (10 mice per group). Tumor size was measured 4 weeks later, and the average tumor volume was calculated. One-way analysis of variance (ANOVA) showed a significant difference in tumor volume among groups (p = 0.001) or between groups (* p < 0.05). (B) Immunofluorescence staining for REST expression in the different tumor tissues confirmed that REST was down-regulated in the R1106 and R1606 tumor samples. Scale bar: 50 µm. (C) The Ki67 cell proliferation marker was used to assess cell proliferation. Ki67 expression was quantified in the different tumor samples by PCI software. Ki67 expression was not significantly different in the R1106 and R1606 tumor tissues compared with the RC-control tumors. (D) Leg amputations were performed at 4 weeks after tumor cell intra-tibial injection, and mice were maintained until 15 weeks after injection, when mice were killed, lungs were harvested, and visible lung tumor nodules were recorded. One-way ANOVA analysis of variance showed a significant difference in lung metastasis among groups (p = 0.0197) or between groups (* p < 0.05).Inhibition of REST decreased the number of vascular endothelial cells and pericyte coverage in tumor vessels. (A) The endothelial marker CD31 (red) was used to identify tumor vascular structure by immunofluorescence staining. White arrows indicate the punctate lumens. Scale bar: 50 µm. (B) Double staining for the pericyte marker NG2 (green) and endothelial marker CD31 (red) was performed in the different tumor tissues to assess vascular pericytes coverage. NG2 expression and NG2 + CD31 dual-positive vessels were decreased in the R1106 and R1606 tumors compared with the RC-control tumors. White arrows indicate double positive staining (yellow color). Scale bar: 50 µm. (C) Mean NG2 expression in each of the tumor groups was quantified. Bars represent standard deviation. * p < 0.01. (D) The ratio of NG2 to CD31 was calculated in each tumor. Bars represent standard deviation. * p < 0.01.Down-regulation of REST reduced tumor vascular perfusion. (A) The mice (two groups tumor-bearing mice, 5 mice per group) were injected with DyLight 594–labeled Lycopersicon esculentum (tomato) lectin, then euthanized after 5 min. Tumor tissue sections were analyzed for tumor vessel perfusion (red). Scale bar: 100 µm. (B) The average number of perfused tumor vessels quantified for each group. Bars represent standard deviation. * p < 0.01. (C) CD31 immunofluorescence staining was performed in the lectin-perfused samples. Co-localization of CD31 (green) and lectin (red) vessels indicated the perfused vessels (yellow) as indicated by white arrows. Scale bar: 100 µm. (D) Functional vessels (yellow) and total perfused vessels (red) were quantified. The percentage of functional vessels was calculated in each group. Bars represent standard deviation. * p < 0.01.Down-regulation of REST increased vessel permeability. (A) Dextran-FITC vessel permeability assay and CD31 staining were performed in the different tumor samples from tumor-bearing mice (three groups of mice, 5 mice per group). CD31 expression indicates tumor vessels, and FITC-labeled dextran (green) indicates vascular leakage due to increased permeability. Scale bar: 50 µm. (B) The averages of FITC-labeled dextran in the different tumor samples were quantified. Bars represent standard deviation. * p < 0.01.REST-KO increased tumor hypoxia and apoptosis. (A) Hypoxia marker HIF-1α expression was detected by immunofluorescence staining (scale bar: 100 µm). (B) The average HIF-1α expression was quantified by assessment of five random fields in each group. Bars represent standard deviation. * p < 0.05. (C) TUNEL assay was performed to detect apoptosis (scale bar: 50 µm), and (D) the average apoptosis area was quantified by assessment of five random fields per group. Apoptotic cells were significantly increased in R1106 and R1606 REST-KO tumors compared with RC-control tumors. Bars represent standard deviation. * p < 0.05.

Med-MDPI/cancers/cancers-12-06-01406.txt

@@ -0,0 +1 @@
+These authors contributed equally.Colorectal cancer (CRC) is the third most common cancer worldwide and the leading cause of cancer-related deaths. Recently, several studies have demonstrated that gut microbiota can alter CRC susceptibility and progression by modulating mechanisms such as inflammation and DNA damage, and by producing metabolites involved in tumor progression or suppression. Dysbiosis of gut microbiota has been observed in patients with CRC, with a decrease in commensal bacterial species (butyrate-producing bacteria) and an enrichment of detrimental bacterial populations (pro-inflammatory opportunistic pathogens). CRC is characterized by altered production of bacterial metabolites directly involved in cancer metabolism including short-chain fatty acids and polyamines. Emerging evidence suggests that diet has an important impact on the risk of CRC development. The intake of high-fiber diets and the supplementation of diet with polyunsaturated fatty acids, polyphenols and probiotics, which are known to regulate gut microbiota, could be not only a potential mechanism for the reduction of CRC risk in a primary prevention setting, but may also be important to enhance the response to cancer therapy when used as adjuvant to conventional treatment for CRC. Therefore, a personalized modulation of the pattern of gut microbiome by diet may be a promising approach to prevent the development and progression of CRC and to improve the efficacy of antitumoral therapy.Microbiota is composed of different bacterial populations with a mutualistic relationship that reside in the epithelial barriers of different organs in the host. Microbiota is a metabolically active ecosystem that interacts with epithelial and stromal cells, with a critical role in human health. Microbiota carries out different functions such as the production of diverse important metabolites, the prevention of infestation by pathogens, and the control of the overgrowth of some bacterial groups to prevent the modulation of the local environment by toxic bacteria [1]. In addition, microbiota is essential for the activation of the host immune system [2].The quantity and diversity of microbial species in the gut increase longitudinally from the stomach to the colon, being the colonic microbiota the most dense and metabolically active community [3]. Although the composition of microbiota is influenced by genetics [4] and may be considered relatively stable within healthy adults over time [5], there is a large variation in the microbiota composition among individuals; this variation is conditioned by different external environmental factors such as diet, chemical exposure, and antibiotic/medication consumption. Dietary changes have been shown to have significant effects in gut microbiota, and the switching from a high-fat/low-fiber diet to a low-fat/high-fiber diet may cause important changes in the gut microbiota within 24 hours [6].In the last decade numerous works have established a clear relationship between alterations in the gut microbiota composition and diverse human pathologies. In particular, obesity and associated metabolic disorders (e.g., type 2 diabetes and non-alcoholic fatty liver), autoimmune diseases (e.g., type 1 diabetes and inflammatory bowel disease), and several types of cancer are characterized by changes in the microbiome and gut dysbiosis [7]. The gut microbiota produces a diverse metabolite repertoire that may harm or benefit the host. Alterations in the intestinal bacteria balance could lead to changes in the levels of gut microbial metabolites such as short-chain fatty acids (SCFAs), polyphenols, vitamins, tryptophan catabolites and polyamines [8], which could be related to the pathogenesis of the human diseases described above. In particular, abnormal levels of SCFAs and molecules related to the metabolism of amino acid like polyamines have been involved in cancer progression and metastasis in different types of tumor [9]. In this review we discuss the potential role of gut microbiota in the carcinogenesis of colorectal cancer (CRC), the possible role of bacterial metabolites in CRC development and progression, and the influence that certain dietary mediators exert over the intestinal microbiota and CRC risk. CRC is the third most common cancer worldwide; nevertheless its exact aetiology is still unknown [10]. Most of the CRC cases are sporadic (nearly 90%), and some genetic and environmental factors have been identified as potential risk factors. Lifestyle factors that increase the risk of CRC in developing countries include physical inactivity, smoking, unhealthy dietary habits (e.g., diets rich in processed and red meat, high fat diets, low intake of fibre), alcohol consumption, and obesity [11,12,13]. Importantly, all these enviromental factors are able to produce changes in the gut microbiota composition [14].Emerging evidence suggests that in animal models gut microbiota may contribute to CRC development through the production of microbial metabolites that interact with the host-immune system and induce the release of genotoxic virulence factors [1,2,15,16]. Recent works have reported that patients with CRC display a lower bacterial diversity and richness in fecal samples and intestinal mucosa compared to healthy individuals [17,18]. In addition, CRC patients show significant alterations in specific bacterial groups with a potential impact on mucosal immune response with respect to healthy controls [18]. In particular, CRC patients exhibit a significant increase in Bacteroides fragilis, Fusobacterium nucleatum, Enterococcaceae or Campylobacter, Peptostreptococus, Enterococus faecalis, Escherichia coli, Shigella and Streptococcus gallolyticus, and a decrease in Faecalibacterium, Blautia, Clostridium, Bifidobacterium and Roseburia [19]. These changes might produce enrichment in pro-inflammatory opportunistic pathogens and a decrease in butyrate-producing bacteria, which may lead to an imbalance in intestinal homeostasis (dysbiosis) that could ultimately lead to tumor formation [11,20,21]. Ahn et al. described a decrease in bacterial diversity in fecal samples of CRC patients, with an increase in Fusobacterium nucleatum and Porphyromonas and a decrease in Gram-positive fiber-fermenting Clostridia [22]. Moreover, it has been shown that patients with colorectal tumours at an early stage (advanced adenoma) have a different microbiota composition compared with those with advanced stage tumours (definitive CRC) [19,23], suggesting that gut microbiota could participate in tumor progression.Tjalsma et al. proposed a bacterial driver-passenger model for microbial involvement in the development of CRC, in which colonic mucosa contains bacterial species that differ in their temporal associations with developing tumours [24]. In this regard, early signs of dysbiosis in adenoma and an increased abundance of F. nucleatum were associated to a higher expression of pro-inflammatory cytokines in colonic tissue from CRC patients [25,26,27].In mouse models of genetically predisposed CRC, it has been demonstrated that microbiota can elicit protumorigenic responses. For instance, Li et al. described the role of gut microbiota in the acceleration of tumor growth in APC (Min/+) mice by triggering the c-Jun/JNK and STAT3 signaling pathways in combination with anemia [28]. On the other hand, it has been shown that in IL-10 deficient mice, an increased microbiota-specific Th1 response exacerbated colitis, resulting in adenocarcinoma formation [29]. In germ-free mice the transfer of stool from patients with CRC enhanced intestinal cell proliferation, suggesting a promotive effect of microbiota on tumour formation [30,31,32]. Nevertheless, the gut microbiome is not limited only to bacteria but also includes viruses and fungal species. Many studies have reported a higher viral DNA load in tumors in comparison to normal noncancerous tissue. A number of studies have aimed to assess the potential contribution of viral infections, such as infections with human papillomaviruses, human polyomaviruses and human herpesviruses, to the risk of CRC [33,34]. Community-based viral shotgun NGS techniques have revealed alterations in the colon virome diversity in CRC patients. In particular CRC cohorts displayed a higher viral diversity in CRC cohorts, with enrichment in members of the genera Orthobunyavirus, Inovirus and Tunalikevirus. Remarkably, the last two virus genera are known to infect Gram-negative bacterial hosts, including bft-positive enterotoxigenic Bacteroides fragilis, Fusobacterium nucleatum, and pks-positive genotoxic Escherichia coli, which are implicated in CRC development. The fecal virome profile has been shown to be able to predict CRC status and segregate individuals at early and late stages of CRC [35]. By contrast, another study performed by Hannigan et al. did not find virome community differences in alpha diversity (richness and Shannon diversity) and beta diversity (Bray-Curtis dissimilarity) between healthy and cancerous states, although they detected strong associations between the colon virus community composition and CRC. The identified viruses were lysogenic bacteriophages belonging to the Siphoviridae and Myoviridae taxa, which can alter the composition of gut bacterial communities [36]. These authors propose a theory on how bacteriophage-bacterium dynamics may promote a novel colonization niche for cancer-associated bacteria. Thus bacteriophages could alter bacterial populations in the colon by promoting bacterial lysis, which would allow the production of biofilms by the opportunistic species anchored to the epithelium. This would favor the penetration of oncogenic bacteria in the intestinal lumen, triggering the inflammatory immune response and promoting the transformation of tumor cells [36].Furthermore, phage therapies that exploit the co-existence of specific bacteria within cancerous tumors to induce a specific anti-tumor immune response could be used in the treatment of CRC. In fact, Zheng et al. developed a phage-guided biotic–abiotic hybrid nanosystem that could increase the chemotherapeutic potency of irinotecan against CRC cells, selectively killing the F. nucleatum population and allowing the butyrate-producing bacteria to increase their abundance at the same time [37]. On the other hand, apart from the virome, metagenome of the fungal microbiota has also been studied in CRC. The fungal genera Phoma and Candida have been detected in higher quantities in colorectal adenoma biopsies, implicating altered host-associated fungal populations in the development of CRC [38]. In another study, a fungal dysbiosis in CRC patients have been described, with enrichment in the Basidiomycota/Ascomycota ratio and the class Malasseziomycetes in CRC patients when compared with healthy controls. On the contrary, in cancer patients a decrease in the relative abundance of Saccharomyces cerevisiae, yeast known for its anti-inflammatory and regulatory properties of the immune system, was observed, which could make it a potential therapeutic route. Ecological analysis also revealed a higher number of co-occurring fungal intra-kingdom correlations, and more co-exclusive correlations between fungi and bacteria in CRC compared with healthy controls [39]. Similarly, Gao et al. observed a fungal dysbiosis in colon polyps and CRC, with an increase in the Ascomycota/Basidiomycota ratio and in the opportunistic fungi Trichosporon and Malassezia, which might favor the progression of CRC. Subsequent analysis showed a lower diversity and significant mycobiota alteration in early-stage tumors [40]. All these studies revealed that CRC is not only characterized by a dysbalance in the composition of gut bacteria but also by a disruption of the gut virome and mycobiome homeostasis.Chronic inflammation has been proposed to be involved in the promotion of cancer. Thus, it is estimated that up to 20% of all tumours are preceded by chronic inflammation [41]. During carcinogenesis, inflammatory cytokines and chemokines produced by cancer cells attract immature myeloid cells or pro-inflammatory helper T cells. This pro-tumorigenic microenvironment is characterized by the synthesis of growth and angiogenic factors and tissue remodelling enzymes, and the suppression of antitumor T-cell responses [42], favouring tumour progression. Gut microbiota dysbiosis and increased intestinal permeability are highly associated to colon inflammation, which could be a key factor for the initiation and/or progression of CRC [43]. Thus, when intestinal permeability is increased, the lipopolysaccharides of the outer membranes of some types of bacteria penetrate the host organism, which induces the immune system to secrete cytokines and start a cascade of reactions that ultimately leads to inflammation. Local inflammation contributes to tumor progression through protumorigenic cytokines and chemokines that act as growth factors and promote angiogenesis [42]. In mouse models it has been recently demonstrated that the development of polyps was associated with defects in the colon barrier integrity, bacterial invasion, and an increased expression of several inflammatory factors such as IL-17, Cxcl2, Tnf-α, and IL-1. Moreover, alterations in the intestinal barrier allowed microbes to induce local inflammation, promoting polyp formation and cancer development in mice [44]. Hu et al. demonstrated that aberrant inflammasome-induced microbiota plays a critical role in CRC development, where mice deficient in the NOD-like receptor family pyrin domain containing 6 (NLRP6) inflammasome exhibited enhanced inflammation-induced CRC formation [45].In addition to a shift in the microbiota composition, pathogenic bacterial species may also have a role in the development of CRC. There are different pathogenic microbes associated to the promotion of CRC, including several Bacteroides species (B. vulgatus and B. stercoris), Bifidobacterium species (B. longun and B. angulatum), Eubacterium species (E. rectale 1 and 2, E. elignes 1 and 2, and E. cylindroides), Ruminococus species (R. torques, R. albus, and R. gnavus), Streptococo hansenii, Fusobacterium prausnitzi, and Peptoestreptococo productus 1 [46]. All these microbes may drive CRC tumorigenesis by inducing proliferation of the epithelial cells, producing damage in the epithelial barrier, and causing inflammation. In addition, different toxins may damage DNA inducing a protumorigenic effect. For instance, Bacteroides fragilis toxin is known to activate Wnt and NF-kB signaling pathways and enhance epithelial release of pro-inflammatory molecules [47,48], E. coli toxin (colibactin toxin) causes DNA crosslinks and double strand DNA breaks [49], and Salmonella protein AvrA has recently been shown to induce β-catenin signaling and enhance colonic tumorigenesis by activating STAT3 pathway in a colon cancer mouse model [48].  Similarly, F. nucleatum has emerged as a potential candidate for CRC predisposition, because of its ability to bind to E-cadherin on the surface of colon cells through FadA adhesion, leading to the activation of Wnt/B-catenin signaling and the production of an inflammatory and oncogenic response [50]. Fap2, another adhesin from F. nucleatum, is able to bind to the inhibitory immune receptor TIGIT (T cell immune receptor with Ig and ITIM domains) and alter the function of natural killer cells and tumor infiltrating lymphocytes [51]. F. nucleatum has also been associated with resistance to the CRC chemotherapy agent oxaliplatin by inducing authophagy via Toll-like receptor 4 [51]. The gut microbiota produces different metabolites after anaerobic fermentation of exogenous undigested dietary components. These metabolites interact with the epithelial cells of the mucosal interface, influencing immune responses and the potential development of different diseases. The gut microbiota derived metabolites with pro-carcinogenic effects include products of protein fermentation such as polyamines [52]. Remarkably, a recent metagenomic analysis reported that the CRC-associated microbiome showed an association with alterations in polyamine metabolism [53], indicating that these metabolites could be particularly important in CRC development and progression. On the other hand, CRC has been associated to alterations in the metabolism of SCFAs [8,9], which have been shown to exhibit potential anti-carcinogenic effects in cellular and animal models of colon cancer. Polyamines are aliphatic amines essential for normal cell growth. It is widely accepted that polyamine metabolism is frequently dysregulated in cancer, including CRC [54,55]. In colon cancer, ornithine decarboxylase (ODC), the key enzyme of the polyamine biosynthetic pathway, is expressed at higher levels in tumor tissue than in adjacent normal mucosa [56,57], suggesting that increased polyamine production could be involved in the tumorigenesis of CRC. The restriction of polyamine availability by alpha-difluoromethylornithine (DFMO) treatment, a chemical inhibitor of ODC, in combination with non-steroidal anti-inflammatory drugs has been shown to exhibit promising effects as therapeutic option for colorectal adenoma incidence [58,59]. A limitation of the monotherapy with DFMO, however, is that tumor cells can replace endogenously synthetized polyamines by taking extracellular polyamines from the colon lumen. This is particularly important in the case of CRC, which is surrounded by intestinal bacteria that are able to produce high levels of polyamines [60,61]. Remarkably, a recent metagenomic analysis has established that the CRC-associated microbiome showed an association with the conversion of amino acids to polyamines (e.g., L-arginine and L-ornithine degradation to putrescine) [53]. In mice, the administration of antibiotics enhanced the cytostatic effect of DFMO on tumor cells [62,63], suggesting that reduction of bacterial polyamine biosynthesis together with the inhibition of the polyamine biosynthesis route could be considered as an anti-tumoral strategy. Another promising strategy to limit the availability of polyamines in the tumor could be the combination of DFMO and the polyamine transport inhibitor AMXT 1501, which has shown to be effective in mouse models of neuroblastoma [64].On the other hand, a metabolomics screen comparing paired colon cancer and normal tissue samples from patients with CRC revealed that bacteria biofilm formation, even in the normal colon tissue, was associated with increased colonic epithelial cell proliferation and host-enhanced polyamine metabolism [65]. In addition, bacteria-generated polyamines in biofilms may contribute to the inflammation and proliferation of colon cancer [58]. Following antibiotic treatment, resected colorectal cancer tissues harbored disrupted bacterial biofilms and lowered N1, N12-diacetylspermine tissue concentrations compare to biofilm-negative colon cancer tissues, suggesting that gut microbes can induce an increase of host generated N1, N12-diacetylspermine [66] (Figure 1).SCFAs, especially butyrate, propionate and acetate, are products of the fermentation of dietary fiber by anaerobic gut microbiota with an essential role in the health of colonic mucosa through the modulation of the local immune response and the protection of the intestinal barrier. Recent studies have reported lower levels of butyrate-producing bacteria in CRC patients [18,67]. In addition, metabolomic analyses have described significant perturbations of SCFA metabolism in CRC compared to adjacent mucosa [68]. Butyrate has been shown to be able to induce IL-18 production in intestinal epithelial cells by activating GPR109a receptor, which stimulates the mucosal tissue repair through the regulation of the production and availability of IL-22 [69]. Remarkably, in mice the absence of IL-18 has been associated with gut microbiota dysbiosis, alterations of the inflammatory response, and a dysregulation of the homeostatic and mucosal repair [70,71], resulting in increased susceptibility to carcinogenesis. In fact, some experiments with mice that are unable to respond to IL-18 have shown a high incidence of intestinal dysbiosis and elevated susceptibility of chemically induced CRC carcinogenesis [69,72,73]. In addition, butyrate can induce the expansion of T reg lymphocytes to regulate the local immune response and suppressing colonic inflammation and carcinogenesis [74] (Figure 1).The use of antibiotics generally has broad effects on the gut microbiota and indirectly affects CRC progression. The suppression of microbiota by antibiotics has been related to a decrease in crypt height and heme-induced colorectal carcinogenesis in rats [75]. Dick et al. described in a nested case–control study that the use of antibiotics with both anti-anaerobic or anti-aerobic activity (such as penicillins and quinolones) was associated with a dose-dependent increased risk of CRC development [76]. Interestingly, another nested case-control study in UK has shown that bacterial or fungal outgrowth after multiple penicillin treatments slightly increases the risk of CRC development [77]. ZacKular et al. demonstrated that manipulation of the gut microbiota with different antibiotic cocktails during the onset of inflammation can significantly decrease tumorigenesis in mice [78]. Bullman et al. showed that the treatment with metronidazole of mice xenografted with CRC decreased both the load of F. nucleatum and the growth of the tumor [79]. In an azoxymethane (AOM)/dextran sodium sulfate (DSS)-induced CRC murine model the alteration of gut microbiota using antibiotics attenuated colon tumorigenesis, but only when gut microbial changes were maintained throughout the entire period of inflammation [80]. Recently, Ma et al. described that the alterations of gut microbiota after antibiotic use could contribute to the long-term dysregulation of host immune homeostasis and affect CRC pathogenesis [81]. In another meta-analysis, Wang et al. suggested that a higher number of antibiotic prescriptions were associated with a higher risk of CRC. By contrast, they described that the risk of rectal cancer was inversely associated with antibiotic exposure, possibly due to the differences in the composition of gut microbiota between colon and rectum [82].On the other hand, the use of antibiotics in early childhood has been associated with increased colonic adenoma formation (a precursor lesion to CRC) in later life, suggesting that a dysbiotic microbiota is acquired and hold over a longer period of time [83,84,85]. In a recent study based on the Clinical Practice Research Datalink (CPRD), the use of oral antibiotics with anti-anaerobic activity has been associated with increased CRC risk in a dose-dependent fashion in the UK population, although the effects differed depending on the anatomical location, being greatest in the proximal colon [86]. In this regard, a systematic review and meta-analysis of observational studies was performed to assess whether the use of antibiotics was associated with the development of pre-cancerous or cancerous lesions in adults [87]. In this review the authors only found a weak association between cumulative antibiotic use and risk of CRC. These results could be explained by confounding factors within the studies, such as heterogeneity in how antibiotic exposure was registered, the variability in the route and setting of antibiotic exposure among studies, the relatively short time between antibiotic exposure and the development of CRC in the majority of studies, and that none of the included studies tried to differentiate microbiome-associated events between initiation of CRC as polyp prevalence and progression through more advanced stages [87]. Another recent work, Armstrong et al. showed that patients prescribed antibiotics in up to 15 years preceding diagnosis were associated with a higher risk of CRC [88]. On the other hand, the use of antibiotics often leads to dysbiosis, facilitating the acquisition of drug resistance. In this context, Yuan et al. found that antibiotic treatment-induced gut microbiota dysbiosis decreased the therapeutic efficacy of 5-fluorouracil (5-FU) for tumor treatment [89]. Remarkably, antibiotic use before (but not following) the start of 5FU-based chemotherapy has been associated with worse progression-free and overall survival among patients with metastatic colorectal cancer [90]. In addition, the disruption of microbiota in MC38 colon carcinoma-grafted mice with a broad-spectrum antibiotics impaired tumor response to anti-CTLA4 immunotherapies [91]. Nevertheless, the efficacy of anti-CTLA4 treatment in antibiotic-treated MC38-grafted mice could be rescued by colonizing mice with B. fragilis, immunizing with low dose of a recombinant BFT-2 enterotoxin (a major virulence factor of B. fragilis), or performing adoptive transfer with B. fragilis-specific T cells [92]. Dietary fiber has been shown to beneficially affect metabolic activities in the gastrointestinal tract [93,94]. In the observational EPIC study, Bingham et al. found that dietary fiber was inversely associated with the incidence of large bowel cancer, although no significant differences were observed between various food sources of dietary fiber intake on the protection against CRC [95]. In another large prospective cohort study, the protective effect of whole-grain consumption was associated with a slight reduction in the risk of developing CRC [96]. In a prospective case-control study nested within seven UK cohort studies using food intake questionnaires, both the intake of absolute fiber as well as the fiber intake density were inversely associated with the risk of colorectal and colon cancers in both age-adjusted models and multivariable models adjusted for age, anthropomorphic and socioeconomic factors, dietary intake of folate, alcohol consumption, and energy intake [97]. The protective effect of dietary fiber remained evident in the 11-year follow-up of the EPIC study, being the total dietary fiber intake still inversely associated with colorectal cancer [98]. Similarly, a prospective study in the Scandinavian HELGA cohort showed that the intake of dietary fiber (especially from cereals) was associated with a reduction in the incidence of CRC [99]. Moreover, a higher intake of dietary fiber and whole grains after CRC diagnosis has been associated with better survival rates [100] Moen et al. compared the effects of several dietary interventions (inulin, cellulose or brewers spent grain) in AOM -treated A/J Min/+ mice, finding that the mice fed with inulin displayed lower incidence of colonic tumorigenesis and a distinct cecal microbiota profile associated with low colonic tumor load [101]. On the other hand, Mehta et al. found that diets rich in whole grains and dietary fiber were associated with a lower risk of developing F. nucleatum-positive colorectal cancer but not F. nucleatum-negative CRC, supporting a potential role for intestinal microbiota in mediating the association between diet and the development of colorectal neoplasms [102]. More recently, the fermentation of soluble fibers such as lignan and β-glucan to SCFAs by gut microbiota also plays a critical role in cancer prevention. The ingestion of dietary fiber has been associated with the presence of fecal butyrate-producing bacteria [103,104]. Remarkably, lower fecal SCFA levels as a consequence of a lower dietary fiber intake and lower prevalence of Clostridium, Roseburia, and Eubacterium spp. were found in CRC risk subjects compared to healthy individuals [104]. In agreement, a dietary intervention consisting in a higher intake of dietary fibers in African Americans increased saccharolytic fermentation and butyrogenesis, and suppressed secondary bile acid synthesis, resulting in the reduction of biomarkers of colon cancer risk [105]. In line with this work, another study using a gnotobiotic mouse models demonstrated that dietary fiber protected against colorectal tumorigenesis in a microbiota- and butyrate-dependent manner via inhibition of histone deacetylase activity [106]. In addition, high fiber diets given to mouse models of polyposis also produced a significant increase of SCFA-producing bacteria and ameliorated polyposis [107]. The possible mechanism that could explain the role of dietary fiber in CRC prevention could be that fiber reduces concentrations of intestinal carcinogens due to the reduction of intestinal transit time and increased faecal bulk, which would lessen the potential for faecal mutagens to interact with the colon mucosa [83]. In addition, the increase bacterial fermentation of resistant starch to SCFAs (especially butyrate) has been shown to lower fecal pH in the colon, and this pH reduction can inhibit pathogenic organism proliferation and DNA damage induction, and enhance apoptosis and prevent proliferation of cancer cells [108,109,110]. On the other hand, it has been decribed that long-term fiber dominant diet may increase the density of Firmicutes, which may have immune modulatory and anti-inflammatory effects in the host [111,112] (Figure 1). Altogether, a higher fiber intake could not only prevent the disturbances in the community structure and function of the gut microbiota, but could also stimulate the production of bacterial metabolites with anti-CRC activity such as butyrate (see below). Moreover, the intake of fiber after CRC diagnosis has been associated with better survival rates. Nevertheless, more clinical and preclinical studies are necessary to establish the most appropriate conditions (dose and duration) of these dietary interventions involving high-fiber intake to prevent CRC (Table 1).Diverse studies have recently defined an impact of dietary omega-3 polyunsaturated fatty acids (PUFAs) on the gut microbiota [113]. In particular, the supplementation of PUFAs has been associated with a decrease in Faecalibacterium and an increase of Bacteroidetes and butyrate-producing bacteria [114]. In addition, PUFAs are able to reduce intestinal microbial dysbiosis by increasing the proportions of beneficial bacteria and decreasing the proportions of pathogenic bacteria in the gastrointestinal tract [115]. PUFAs have been extensively studied due to their role in their protective effect from CRC carcinogenesis, mainly through mechanisms that regulate differentiation and apoptosis of the colonocytes [116,117,118,119]. For instance, in C57BL/6J mice bearing azoxymethane-dextran sulfate sodium–induced CRC, the relative abundance in the gut of beneficial bacteria such as Lactobacillus increased after eicosapentaenoic acid treatment, in parallel with a reduction in the size of colorectal tumors, a decrease in the number of proliferating cells, and an increase of apoptotic cells within the tumors [120]. These PUFAs could also alter the cell cycle components, act on the immune system and modulate CRC-related genes expression [121]. The protective role of PUFAs in colorectal carcinogenesis prevention may also relate to the decreased risk of microsatellite instability (MSI) and the enhancement of DNA repair systems mismatch pathways [122]. On the other hand, several randomized control trials have reported that PUFAs are often subjected to peroxidation, process by which free radicals are frequently generated [123]. In addition, Yang et al. found that the PUFA composition is different between normal and cancerous tissues in the same CRC patient, suggesting that the metabolism of PUFAs might play a significant role in the evolution of inflammation driven tumorigenesis in the CRC [124]. In rats exposed to azoxymethane, a potent carcinogen used to induce colon cancer in animal models, the consumption of dietary fish oil (which is rich in omega-3 PUFAs) led to a lower rate of CRC adenocarcinoma incidence [125]. Song et al. reported that high marine ω-3 PUFA intake after CRC diagnosis is associated with a lower risk of CRC-specific mortality, indicating that an elevated consumption of marine ω-3 PUFAs after diagnosis may confer additional benefits to patients with CRC [126]. Furthermore, other investigations have demonstrated that in animals with carcinogen-induced CRC tumors fed with a diet of fish oil plus pectin had increased colonocyte apoptosis compared with those fed with corn oil cellulose as control diet [127]. Moreover, it has been recently shown that docosahexaenoic acid in combination with butyrate enhances mitochondrial lipid oxidation and reduces mitochondrial membrane potential, contributing to the induction of apoptosis in colonocytes [109,128]. A recent clinical trial has shown a CRC incidence reduction of 22% among pesco-vegetarian subjects compared with non-vegetarian individuals [129]. Finally, a very recent study performed by Aglago et al. analyzed the association between fish consumption and dietary and circulating levels of PUFAs with CRC incidence using data from the EPIC cohort. They found that the regular intake of fish at recommended levels was associated with a lower risk of CRC, possibly through the exposure to high PUFA content [130]. Several in vivo studies have described that PUFAs can reduce 5-FU-related toxicity and potentiate 5-FU anti-cancer activity though the reduction of tumor burden and DNA damage and the increase of apoptosis [131,132,133]. A recently study in rat models showed that combining PUFAs with 5-FU and irinotecan could help restore lipid stocks, thus potentially limiting 5-FU-associated side effects [134]. Cai et al. showed that PUFAs have the potential to radio-sensitise HT29 colon cancer cells, possible due to an increase in lipid peroxidation products within the cells [135]. Granci et al. reported an increase in apoptosis in colon cancer cells when combining 5-FU, oxaliplatin, and irinotecan with a fish oil emulsion with PUFAs [136] (Figure 2). A recent double-blind, randomized, placebo-controlled trial investigated the effect of the combination treatment of PUFA and a probiotic supplement on the tolerability of capecitabine/oxaliplatin chemotherapy and on inflammatory markers in CRC patients, finding an improved overall quality of life and reduced chemotherapy-induced symptoms such as diarrhea and fatigue in these study subjects [137]. Then, PUFAs have a protective effect from CRC carcinogenesis and in combination with chemotherapeutic agents could be an effective approach to the treatment of CRC patients (Table 1). Most fruits and vegetables contain phytochemicals with anti-microbial and anti-inflammatory properties [138]. Phytochemicals are able to maintain the balance of the gut microbiota and exhibit anti-tumoral properties (e.g., decrease cell proliferation and stimulate apoptosis of cancer cells, inhibit angiogenesis and delay metastasis) [139]. Polyphenols are a structural class of phytochemicals with multiple phenolic units that are found at high concentrations in coffee, tea, wine, fruits, vegetables and whole grains [140,141]. Because polyphenols are poorly absorbed in the small intestine they usually tend to accumulate in the colon, where they can be hydrolyzed by the enzymatic activities of the gut microbial community into lower molecular-weight bioactive compounds before absorption [142,143,144]. Moreover, polyphenols present in the colon have been found to significantly alter the gut microbiota, particularly by suppressing the growth of Clostridium and Bacteroides species [145,146,147].On the other hand, red wine polyphenols have been linked to CRC prevention by their capacity of inhibiting the growth of pathogenic bacterial species such as F. nucleatum and P. gingivalis [148], as well as the adhesion to oral cells [149]. Several studies have described the effects of certain polyphenolic compounds in CRC prevention and treatment both in vivo and in vitro. Quercetin, a flavonol present at high concentrations in certain vegetables and fruits such as onions or apples, has been shown to exert some anticancer effects in colon cancer cells, mainly by inhibiting cell proliferation and inducing apoptosis [150]. Anthocyanin-rich tropical fruits such as Cocoplum (Chrysobalanus icaco L.) have also demonstrated anti-inflammatory activity (through the reduction of TNF-α, IL-1β, IL-6, and NF-κB expression levels) and pro-oxidant effect in the human colorectal adenocarcinoma cell line HT29 [136,151]. In HCT116 colon cancer cells, the activity of apigenin was correlated with a blockage in cell cycle progression, induction of apoptosis and inhibition of autophagy [152]. The antitumor effects of several polyphenols present in high amounts in blueberries, red grapes and cocoa (such as anthocyanins and tannins) have been related to their capability of inducing adaptive immune cells to target tumor cells in preclinical models [153,154,155]. In addition, the combination of curcumin and resveratrol has been shown to be highly effective in inhibiting the growth of colon cancer cell both in vitro and in vivo [156]. In some clinical studies, the intake of flavonols and flavan-3-ol monomers has been associated with a decreased risk in colorectal cancer [157]. However, the association between the regular intake of either total flavonoids or any flavonoid subclass and CRC risk and tumor subsites could not be corroborated in other human cohort studies [158,159]. Emerging evidence suggests that the combination of conventional chemotherapy treatment for CRC with some natural dietary polyphenols can significantly enhance the chemotherapeutic effect. For instance, quercetin has been tested in vitro in combination with 5-FU in CO115 human colon carcinoma cells and HCT15 colorectal adenocarcinoma cells increasing apoptosis levels in CO115 cell line, in a synergistic manner, but as an additive effect in HCT15 cells [160]. Also, the combination of 5-FU against a colon adenocarcinoma cell line treatment with phenolic acid rich-extracts such as Gelam honey and ginger (Zingiber officinale) enhanced the anticancer activity of 5-FU [140,161]. On the other hand, Montrose et al. demonstrated in DSS-mice that the chemopreventive effect of black raspberries was mediated by the downregulation of the expression of pro-inflammatory cytokines (TNF-α and IL-1β) and the decrease of COX-2 and plasma prostaglandin E2 levels [162]. On the other hand, the chemopreventive effect of curcumin through the reduction of colonic tumor burden due to the maintenance of a high microbial diversity was proven in IL-10-deficient mouse [163]. Moreover, curcumin is able to enhance chemosensitization to 5-FU-based chemotherapy by targeting cancer stem cell subpopulations that could be responsible for tumor relapse and resistance to conventional therapies [164]. The effect of 5-FU also increased in combination with resveratrol due to its chemosensitizing properties [165]. Moreover, resveratrol could be used to overcome drug resistance in combination with other chemotherapeutic drugs, due to its ability to downregulate multidrug resistant protein 1 by preventing the activation of NF-κB signalling and suppressing cAMP-responsive element transcriptional activity [166] (Figure 2).These findings suggest that polyphenols and their derived microbial metabolites could be used as a complementary therapy not only for CRC prevention, but also in potentiating the efficacy of chemotherapy against CRC (Table 1). Probiotics are live microorganisms that contribute to the health benefit of the patients and are able to inhibit CRC through different mechanisms (Figure 2). Several studies have suggested that regular consumption of probiotics may improve the diversity and richness profile of the intestinal microbiota, downregulate chronic inflammation, and reduce the production of carcinogenic compounds during intestinal dysbiosis [167,168].Hatakka et al. demonstrated that the consumption of certain strains of probiotic bacteria can reduce the activity of intestinal enzymes that can convert aromatic hydrocarbons and amines in active carcinogens and prevent colon cancer [169]. The peptidoglycan, polysaccharide and secreted glycoproteins on the surface of probiotic bacteria, combined with carcinogenic mutagens could be responsible for biotransformation aiming to detoxification [170].On the other hand, probiotics can also regulate the immune system response through the activation of phagocytes to eliminate cancer cells in their early stages of development, contributing to the maintenance of the immune-vigilance state [171,172]. For instance, the probiotic strain Bifidobacterium animalis subsp. Lactis can produce mycosporin-like amino acids, which are able to modulate host immunity by regulating the proliferation and differentiation of intestinal epithelial cells, macrophages and lymphocytes and the production of cytokines [173]. Nevertheless, not all probiotics are able to regulate the immune system and to prevent the development of CRC. To induce immunostimulation on the host, both a dosage of around 109 CFU/day and an intestinal transit time between 48 and 72 h are necessary [168]. Probiotics are also known to stimulate the production of a variety of compounds that improve the intestinal barrier function. The perioperative administration of CRC patients with a probiotic consisting in a mixture of Lactobacillus plantarum, Lactobacillus acidophilus and Bifidobacterium Longum increased the expression of mucosal tight junction proteins, improved the integrity of gut mucosal barrier and reduced enteropathogenic bacteria, resulting in decreased infectious complications after colorectomy [174]. In one randomized, double-blind, placebo-controlled trial, patients with colon cancer and polypectomized patients, Rafter et al. demonstrated that oral treatment with a probiotic mixture of Lactobacillus rhamnosus and Bifidobacterium Breve was able to induce changes in gut microbiota, reduce several cancer biomarkers such as colorectal proliferation, and improve intestinal epithelial barrier permeability [175]. Interestingly, when the bacterial probiotic strains L. acidophilus NCFM and B. animalis subsp. lactis Bl-04 were used in a prospective intervention study with CRC patients, it was observed that the patients with colon cancer that received probiotics had a unique profile of bacterial populations in their gut microbiota, which was mainly characterized by an increased abundance of butyrate-producing species in tumor, mucosa and fecal samples [176]. Other probiotic strains derived from Bifidobacterium have been shown to restore the equilibrium of the gut dysbiosis in patients with CRC [177]. The consumption of probiotics has been also associated with the induction of a proapoptotic activity in cancer cells of human CRC patients. Wan et al. found that the consumption of Lactobacillus Delbrueckii increased the expression of caspase-3, leading to apoptosis of human colon cancer cells [178]. Furthermore, Konishi et al. described that L. casei strain ATCC 334 produced ferrichrome, a molecule that inhibits the progression of colon cancer by inducing the apoptosis of cancer cells via c-Jun N-terminal kinase pathway [179]. In addition, several studies have described a relationship between gut microbiota and the efficacy and/or toxicity of both chemotherapies and immunotherapies [180,181]. Probiotics has been also shown to affect the response to immunotherapy by systemic priming and regulation of different myeloid-derived cell functions in the tumor microenvironment. Tumor-infiltrating myeloid cells appear to be primed by bacterial LPS through the TLR4 receptor for responsiveness to the TLR9 ligand CpG-ODN [181]. In this regard, Chang et al. described that L. casei variety rhamnosus (LCR35) attenuated 5-FU/oxaliplatin-induced intestinal mucositis in CRC-bearing mice [182]. Moreover, recent studies have described that the combination of PD-L1 inhibitors and oral therapy with bifidobacteria had a synergistic inhibitory effect on tumor growth compared with the effect of either intervention alone [183,184,185]. In CRC patients undergoing chemotherapy the supplementation with L. rhamnosus decreased the frequency of diarrhea and abdominal distress and avoided the dose reduction caused by intestinal toxicity compared to patients who received placebo [186]. Finally, probiotics have been studied in the setting of radiation therapy used in the treatment of CRC patients. Previous studies have demonstrated that intestinal bacteria can repair injuries and reduce the incidence and severity of diarrea and bowel movements induced by the radiation therapy [187]. Taken together, the probiotic interventions could be therapeutically used to target the gut dysbiosis frequently observed in CRC patients, due to their beneficial effects on the immune system and the intestinal barrier function, as well as, their antitumoral properties by releasing metabolites that are able to get rid of potential carcinogens. Moreover, the modification of the composition of CRC microbiome with probiotics might enhance the effectiveness to cancer chemotherapy and immunotherapies and the reduction of toxicity associated to radiation therapy. Noteworthy, not all probiotic strains showed anti-CRC effects and their beneficial impact depends on the bacterial strain, the dosage, the duration of intervention and the intestinal transit time. Further investigations are therefore necessary to clarify the action mechanism and the potential of probiotics in CRC prevention (Table 1).Different animal and human studies have revealed that the microbial composition has been altered in precancerous colorectal lesions and in CRC. Moreover, a dysbiosis in gut microbiota has been found in CRC patients compared with healthy controls, with enrichment in pro-inflammatory opportunistic pathogens and a decrease in butyrate-producing bacteria. The proposed mechanisms by which the gut microbiota dysbiosis could participate in colorectal carcinogenesis are the impairment of the intestinal epithelial barrier function, the triggering of proinflammatory responses, the biosynthesis of genotoxins that can interfere with cell cycle regulation, and the production of toxic metabolites by pathogenic bacteria. Moreover, some lysogenic bacteriophages from gut virome could alter bacterial populations (by promoting bacterial lysis) in the colon, which could indirectly result in tumor progression. In addition, ecological analyses revealed synergistic intrafungal and antagonistic bacterial–fungal interactions in colorectal carcinogenesis, suggesting that gut mycobiota may also contribute to colorectal tumorigenesis.In patients with CRC bacteria biofilm formation was associated with a host-enhanced polyamine metabolism, which may significantly contribute to the inflammation and cellular proliferation of colon cancer cells. Antibiotics were associated with CRC risk, but the effect depends on anatomical location and the type of antibiotics. On the other hand, previous epidemiological and clinical research studies have demonstrated that diet plays an important role in the promotion or inhibition of CRC, and gut microbiota is one of the most important links between them. High-fiber diets can significantly reduce the risk of CRC development. Soluble fiber is fermented into SCFAs by bacteria in the large intestine, and SCFAs (especially butyrate) has been shown to exhibit potential anti-carcinogenic effects in in vivo colon cancer models by modulating the local immune response and the protection of the intestinal barrier. Moreover, a high-fiber intake can increase the number of butyrate-producing bacteria in the gut. In addition, supplementation with PUFAs, polyphenols and probiotics could be used as therapeutical approaches for the reduction of CRC risk in a primary prevention setting, and it may also be used as adjuvants to conventional treatment for CRC, given the fact that the intestinal microbiota may modulate and enhance response to cancer therapy and reduce toxicity.  Thus, taking all this evidence together, gut microbiota should be considered as a key factor that can contribute to both the initiation and development of CRC. In addition, the dietary modulation of cancer-associated microbiome, through the intake of dietary components able to avoid dysbiosis and intestinal inflammation or to help modulate response to cancer therapy, could be an efficient strategy to prevent the development and progression of CRC and improve the efficacy of therapy.A Conceptualization, M.I.Q.-O. and J.G.-M.; methodology, L.S.-A., B.R.-M and A.L.-I.; validation, A.O., R.O. and J.A.M.; writing—original draft preparation, M.I.Q.-O., J.G.-M., L.S.-A. and B.R.-M.; writing—review and editing, all authors; funding acquisition, M.I.Q.-O.; supervision, M.I.Q.-O. and J.G.-M. All authors have read and agreed to the published version of the manuscript.This work was supported by PI15/00256 from the Institute of Health “Carlos III” (ISCIII), co-funded by the Fondo Europeo de Desarrollo Regional-FEDER. Maria Isabel Queipo-Ortuño was supported by the “Miguel Servet Type II” program (CPI13/00003, ISCIII, Spain; co-funded by the Fondo Europeo de Desarrollo Regional-FEDER), and by the “Nicolas Monardes” research program of the Consejería de Salud (C-0030-2018, Junta de Andalucía, Spain. Bruno Ramos Molina was supported by a “Miguel Servet Type I” program (CP19/00098, ISCIII, Spain; co-funded by the Fondo Europeo de Desarrollo Regional-FEDER). Lidia Sanchez-Alcoholado was recipient of a predoctoral grant (PE-0106-2019) from the Consejería de Salud y Familia (co-funded by the Fondo Europeo de Desarrollo Regional-FEDER, Andalucia, Spain). Aurora Laborda-Illanes was recipient of a predoctoral grant PFIS-ISCIII (FI19-00112) co-funded by the Fondo Europeo de Desarrollo Regional-FEDER, Madrid, Spain.We thank Richard Carlsson for help with the English language.The authors declare no conflict of interest.Mechanisms of action of polyamines and SCFAs (microbiota-derived metabolites) in the inflammation and cellular proliferation of colon cancer cells. SCFAs: short-chain fatty acids; ODC: ornitina descarboxilasa; DFMO: alpha-difluoromethylornithine; AMXT 1501: polyamine transport inhibitor; GPR109a: G-protein–coupled receptors; IL: interleukin.Beneficial effects of dietary supplementation with PUFAs, polyphenols and probiotics on the intestinal microbiota and colon cells for the reduction of CRC risk or to enhance the response to cancer therapy when are used as adjuvant to conventional treatment. PUFAs: omega-3 polyunsaturated fatty acids; 5-FU: 5-fluorouracil; SCFAs: short-chain fatty acids.Interactions between dietary mediators, gut microbiota and CRC.

Med-MDPI/cancers/cancers-12-06-01407.txt

@@ -0,0 +1 @@
+Non-alcoholic fatty liver disease (NAFLD) leads to steatohepatitis (NASH), fibrosis, and hepatocellular carcinoma. For sedentary patients, lifestyle interventions combining exercise and dietary changes are a cornerstone of treatment. However, the benefit of exercise alone when dietary changes have failed is uncertain. We query whether exercise alone arrests the progression of NASH and tumorigenesis in a choline-deficient, high-fat diet (CD-HFD) murine model. Male C57Bl/6N mice received a control diet or CD-HFD for 12 weeks. CD-HFD mice were randomized further for 8 weeks of sedentariness (SED) or treadmill exercise (EXE). CD-HFD for 12 weeks produced NAFL. After 20 weeks, SED mice developed NASH and hepatic adenomas. Exercise attenuated the progression to NASH. EXE livers showed lower triglycerides and tumor necrosis factor-α expression, less fibrosis, less ballooning, and a lower NAFLD activity score than did SED livers. Plasma transaminases and triglycerides were lower. Exercise activated AMP-activated protein kinase (AMPK) with inhibition of mTORC1 and decreased S6 phosphorylation, reducing hepatocellular adenoma. Exercise activated autophagy with increased LC3-II/LC3-I and mitochondrial recruitment of phosphorylated PTEN-induced kinase. Therefore, exercise attenuates the transition from NAFL to NASH, improves biochemical and histological parameters of NAFLD, and impedes the progression of fibrosis and tumorigenesis associated with enhanced activation of AMPK signaling and favors liver autophagy. Our work supports the benefits of exercise independently of dietary changes. Non-alcoholic fatty liver disease (NAFLD) affects nearly 25% of the world population [1], and its economic burden on society is projected to increase along with its co-morbidities [2]. Clinically, NAFLD covers a spectrum of conditions ranging from non-alcoholic fatty liver (NAFL) to decompensated liver cirrhosis, which can progress to hepatocellular carcinoma (HCC). NAFL is defined by steatosis with or without inflammation but without ballooning injury. In contrast, non-alcoholic steatohepatitis (NASH) is defined by the presence of ballooning injury with inflammation in addition to steatosis [3]. The transition from fatty liver to NASH is an important milestone in the evolution of the disease. In a prospective study from the NASH clinical research network, changes in the NASH activity score, which includes ballooning injury as a marker, were associated with concordant changes in fibrosis [4]. In turn, the degree of hepatic fibrosis was related to both liver-related mortality of NAFLD and to the overall mortality [5]. Currently, no pharmacological treatments can arrest the progression of NAFL to NASH or reverse NASH once it is established, although several experimental drugs are being tested [6]. NAFLD in general and NASH in particular are associated with high caloric, high-fat diet, and sedentariness. Hence, NAFLD is linked to a metabolic overload of the liver. Consequently, clinical trials proposing weight loss and other lifestyle interventions as therapeutic measures have enrolled patients with NASH and fibrosis and sought to demonstrate the resolution of NASH and the reversal of fibrosis. In general, weight loss is positively correlated with an improvement in histologic features of NASH, but for a resolution of NASH and regression of fibrosis, weight loss must be superior to 10% [7,8]. One specific lifestyle intervention documented to improve the hepatic metabolic situation is regular physical activity [9]. In fact, physical activity reduces steatosis in patients with NAFLD, even when it is not associated with weight loss [10,11]. Moreover, resistance exercise in sedentary adults improves steatosis without an impact on body weight [12]. Nonetheless, the interpretation of these clinical studies is confounded when dietary and exercise interventions cannot be dissociated. Whether selective intervention of an exercise regime in patients who maintain their high-fat diet is beneficial and whether exercise alone can impede the transition from NAFL to NASH and thus prevent the evolution of fibrosis are uncertain. In addition, it is also unclear which signaling pathways are operative in transducing the beneficial effects of exercise when the NASH-inducing diet remains in place.We previously reported that in a genetic PTEN knockout mouse model of NASH, regular physical activity decreases the incidence of HCC but without a change in the NAFLD activity score (NAS) score [13]. We also reported that in a rat model of orthotopic syngeneic tumor implantation, regular physical activity downregulated the expression of hepatic genes associated with the development of HCC [14]. However, it remains unknown whether an exercise regimen can intervene at all points, or only at specific points, along the continuum from NAFL to NASH to fibrosis and HCC and change the course of the disease, especially within the context of continued nutritional overload. To explore the selective benefits of physical activity in NAFLD triggered by a high-fat diet and to elucidate its mechanism, we queried whether daily exercise alone impedes the transition from NAFL to NASH and impairs the progression of fibrosis. We chose a choline-deficient, high-fat diet mouse model that displays a progressive phenotype of NAFL to NASH to hepatic fibrosis and, eventually, HCC and compared the disease outcome in sedentary and exercised mice [15,16].We initially assessed the effect of CD-HFD and exercise (Figure 1) on liver lesions by grading the degree of steatosis. Compared to control mice, all CD-HFD-fed mice developed extensive steatosis (>90% steatosis, grade 3; Figure 2A). This macrovesicular steatosis was evident in the early 12-week treated CD-HFD group and in both the sedentary and exercise 20 week CD-HFD groups. Hepatocellular ballooning, a requisite histological feature for diagnosing NASH, was absent in 12-week CD-HFD mice. However, ballooning was present after 20 weeks of CD-HFD but was significantly attenuated by exercise. Ballooning was observed in all eight mice of the sedentary group (grade 2 in 90% of mice) predominantly in zone 3, whereas it was recorded in only 28% of the exercise group (Figure 2B). Neither Mallory–Denk bodies nor apoptotic bodies were detected in either group. Consistent with the development of NAFL and NASH, all CD-HFD fed mice showed hepatic lobular inflammation of grade 3. The NAS score increased significantly in the sedentary mice when compared to the exercised mice (Figure 2C). Oil Red O staining assessed the levels of triglycerides and other neutral lipids in the livers (Figure 2D). In both the 12-week and 20-week CD-HFD groups, the lipid staining was abundant. The signal was decreased in the exercised livers.To investigate the influence of exercise on liver fibrogenesis, we assessed total collagen in the liver and quantified circulating levels of three biomarkers, Pro-C3, Pro-C4, and C6M, in the plasma. After 12 weeks, NAFL mice displayed diffuse pericellular fibrosis but no portal fibrosis (Figure 3A). After 20 weeks, sedentary mice displayed even more fibrosis, but this change was halted by exercise (Figure 3B). In fact, the extent of fibrosis in the exercised mice was comparable to that observed in 12-week NAFL livers. Pro-C3, the proteolytic propeptide released during the formation of type III collagen, was significantly elevated after 12 weeks of CD-HFD, increased further after 20 weeks of CD-HFD in sedentary mice but this rise was halted in exercised mice (Figure 3C). Another biomarker, Pro-C4, which reflects the synthesis of the basement membrane collagen, type IV, was less affected by early CD-HFD. However, it increased after 20 weeks of CD-HFD in sedentary mice, a rise that was prevented by exercise. C6M, a degradation marker for type VI collagen, was slightly increased after 12 weeks of CD-HFD, rose further after 20 weeks of CD-HFD in sedentary mice, but this rise was prevented by exercise (Figure 3C). Thus, the changes in biomarkers were in keeping with the histological patterns observed.Biochemical markers indicative of liver disease were measured in the plasma. Concentrations of both transaminases, ALT and AST, were increased at 12 weeks and further increased after 20 weeks of CD-HFD, but only in sedentary mice (Figure 4). Total bile acids increased after 12 weeks of CD-HFD, but the differences between sedentary and exercised mice at 20 weeks were not statistically significant. Exercise tended to lower plasma triglycerides. Cholesterol increased modestly after 12 weeks of CD-HFD and tended to be lower at 20 weeks in the exercised group. Free fatty acids (FFA) concentrations were increased after 12 weeks of CD-HFD, but the 20-week sedentary and exercise groups were not different. Fasting blood glucose was not different between the groups.The histological evidence of decreased lipid content after exercise was confirmed with biochemical measurements (Figure 5A). After 12 weeks of CD-HFD, hepatic levels of triglycerides rose 13-fold. A further increase at 20 weeks was evident only in sedentary mice, where levels remained significantly higher than those in exercised mice. The hepatic levels of free fatty acids showed a different pattern. After 12 weeks, levels increased but thereafter remained stable and did not change with exercise (Figure 5A). In all mice fed the CD-HFD, the mRNA level of the scavenger receptor and lipid transport facilitator, CD-36, was elevated at least six-fold and was unaffected by exercise (Figure 5B). However, the expression of the long-chain fatty acid carriers fatty acid transport protein (FATP), FATP2, and FATP5 showed the reverse trend (Figure 5C). The mRNA of FATP2 and FATP5 were both decreased by the CD-HFD. Exercise only partially reversed this trend for FATP2.The hepatic expression of lipogenic enzymes was decreased by the CD-HFD (Figure 5). After 12 weeks, both fatty acid synthase (FAS) and ATP citrate lyase (ATPCL) decreased significantly. Moreover, the phosphorylation of ATPCL was decreased by CD-HFD. Exercise also decreased the ratio of P-ATPCL/ATPCL relative to sedentary livers. Similarly, the hepatic lipid catabolic enzymes were downregulated. CD-HFD severely depressed the levels of hormone-sensitive lipase (HSL), but the balance of phosphorylation at sites 563, relative to 565, was changed by exercise. The expression of adipose triglyceride lipase (ATGL) was generally depressed in all CD-HFD fed livers (Figure 5). The expression of perilipin 2 was decreased and remained depressed by the CD-HFD. The nuclear receptor peroxisome proliferator-activated receptor α (PPARα) and fatty acid catabolism enzymes, CPT1α and MCAD, were generally decreased by the CD-HFD. However, exercise was associated with a modest increase in MCAD relative to sedentary mice. The triglyceride synthesis enzyme DGAT2 was decreased in all CD-HFD livers. The mRNA expression of several markers of inflammation was measured in the liver. The expression of the proinflammatory cytokine tumor necrosis factor α (TNF α) increased after 12 and 20 weeks of CD-HFD in comparison to controls, but exercise significantly blunted this rise (Figure 6A). The expression of transforming growth factor β1 (TGFβ1) was similarly elevated in all CD-HFD treated livers (Figure 6B). However, exercise did not influence its expression. Insulin-like growth factor 2 (IGF-2) was undetectable in control samples and was significantly elevated in the sedentary NASH group relative to all other groups (Figure 6C). In contrast, IGF-1 was uniformly decreased in all CD-HFD groups and not affected by exercise. Interleukin-6 was increased in the NAFL group relative to controls but not affected by exercise. Similarly, monocyte chemoattractant protein 1 (MCP-1) increased all CD-HFD groups and was not affected by exercise.Selected markers for endoplasmic reticulum (ER) stress were examined by immunoblotting in homogenates of liver. Binding immunoglobulin protein (BiP) tended to decrease in the exercised group (Figure 7). After 12 weeks of CD-HFD, the expression of X-box binding protein 1 (XBP-1s), which results from the inositol-requiring enzyme (Ire)1 mediated splicing of XBP-1 mRNA, was severely downregulated. Levels of XBP-1s remained depressed at 20 weeks in both exercised and sedentary groups. Conversely, after 12 weeks of CD-HFD, the expression of CCAAT-enhancer-binding protein homologous protein (CHOP) was induced. This upregulation was even more pronounced after 20 weeks of CD-HFD and was not affected by exercise. Since CHOP can trigger the intrinsic apoptotic pathway, we quantified the expression of the pro-apoptotic protein, BAX, and the anti-apoptotic protein, BCL2. The CD-HFD was associated with an upregulation of BAX, such that the ratio of BAX to Bcl2 increased significantly (Figure 7). Exercise tended to decrease this ratio.To evaluate whether the autophagy activation in the liver contributes to the improvement in hepatic steatosis after exercise, we evaluated the level of autophagy-specific microtubule-associated protein light chain 3 (LC3) by immunoblotting of liver homogenate in exercised and sedentary CD-HFD mice. The analysis of LC3BII and LC3BI demonstrated a significantly higher ratio of LC3II/LC3I in the exercised group, which is indicative of an accumulation of autophagosomes (Figure 8). Since mTOR acts upstream to inhibit autophagy, we evaluated the expression of mTOR and its phosphorylation at S2448, which correlates with the activity of mTORC1. Both the expression of mTOR and P-S2448-mTOR were lower in livers from exercised mice than in those of sedentary mice. To evaluate selective autophagy in the mitochondria, we tested for the recruitment of active, phosphorylated PTEN-induced kinase (PINK) to the mitochondrial compartment. P-PINK was higher in the exercise group (Figure 8).Extended exposure to a CD-HFD can lead to hepatic tumors. Indeed, all mice in the 20 weeks CD-HFD sedentary group displayed liver nodules, histologically compatible with hepatocellular adenoma, in accordance with the histological tumor classification of the mouse hepatobiliary system [17] (Figure 9A). In particular, the lesions were sharply demarcated from surrounding liver parenchyma, with loss of the normal lobular architecture and irregular, solid, growth patterns. Macrovesicular steatosis was present in most nodules. Despite the presence of hepatocytes of varying sizes, areas of frank cellular atypia, as well as necrosis or blood vessel invasion, were not observed. The incidence of tumor nodules was significantly reduced in the exercised group, wherein only 70% of mice developed liver nodules (SED vs. EXE, p < 0.01). Furthermore, exercise negatively affected tumor burden; the mean number of nodules per liver was reduced, 2.9 ± 2.5 vs. 7.7 ± 4.4 for sedentary and exercise, respectively (Figure 9B). To elucidate the mechanism mediating these tumor-suppressive effects of exercise, we examined the AMP-activated protein kinase (AMPK)—mTOR signaling pathway in liver homogenate. First, we quantified the phosphorylation of AMPKα on T172, as a measure of AMPK activation, then the activating phosphorylation of S6, responsible for ribosomal biogenesis and translation (Figure 9C). After 20 weeks of CD-HFD, phosphorylation of AMPK (T172) increased, whereas phosphorylation of S6 decreased in exercised mice (Figure 9C).Since dysfunctional mitochondria contribute to the progression of NAFLD, we queried whether exercise halted the progression of NAFL to NASH by improving mitochondrial respiration. In respirometry studies, mitochondria from NAFL livers showed no difference in combined complex I- and II-driven respiration but displayed higher leak respiration, an indicator of dissipated membrane potential, and significantly lower maximal respiration and complex IV respiration than did control mitochondria (Figure 10A). These differences were not detected when NASH sedentary and NASH exercise groups were compared. However, the cytochrome c control factor was significantly increased relative to control in all NAFL and NASH groups (Figure 10B). In addition, citrate synthase (CS) enzyme activity and expression were monitored as an indicator of the tricarboxylic acid cycle (Figure 10C,D). Whereas the protein expression of CS remained constant in all groups, the activity increased in NAFL relative to control (Figure 10C) but not in exercise relative to sedentary. To determine whether a decrease in cytochrome c content could account for the increase in cytochrome c control factor, we quantified the protein expression in mitochondria. The expression of cytochrome c decreased in NAFL relative to control but remained constant in NASH sedentary relative to exercise (Figure 10D). To test whether the decrease in maximal respiration and complex IV respiration could be attributed to the downregulation of respiratory proteins, we quantified the expression of cytochrome c oxidase (COX) subunits 1 and 4. The expression of both COX subunits was lower in NAFL than in controls but similar in a comparison between NASH sedentary and exercise groups (Figure 10D). To ascertain whether mitochondrial biogenesis was modified in the liver, we measured levels of PGC1α and PGC1β that were decreased slightly compared to controls (Figure 10D).After 12 weeks of CD-HFD, the fenestration of the liver endothelium was reduced 10-fold, as determined by the decrease in porosity of the sinusoidal endothelial cells (Figure 11). This morphological change featured throughout the 20 weeks of CD-HFD and was not affected by exercise. The expression of the gene product of NOS3, endothelial nitric oxide synthase (eNOS), did not change, but its phosphorylation tended to decrease in all CD-HFD treated livers, although this did not reach statistical significance.Exercise is well known to induce mitochondrial biogenesis in skeletal muscle. To test whether mice treated with the choline-deficient HFD respond to exercise as expected, we compared the expression of transcription coactivator proteins, PGC1α and PGC1β in skeletal muscle in control, and CD-HFD treated mice (Figure 12A). Both PGC1α and PGC1β were increased by exercise relative to the sedentary group, suggesting an increase in mitochondrial content. The exercise was also associated with the activation of the AKT signaling pathway in skeletal muscle (Figure 12A). Since skeletal muscle secretes peptides affecting liver function, we measured the mRNA expression of fibroblast growth factor 21 (FGF21), which can be released by both skeletal muscle (Figure 12B) and liver (Figure 12C). In skeletal muscle (Figure 12B), the CD-HFD increased FGF21 with exercise, further augmenting expression. Conversely, in the liver, the increase relative to controls was less pronounced, and the expression of FGF21 tended to decrease with exercise.In the present study, we explored the selective benefits of physical activity as a means of modifying the outcome of NAFLD triggered by a high-fat diet. We report that daily exercise alone attenuates the transition from NAFL to NASH, reduces the hepatic accumulation of triglycerides, impedes the progression of fibrosis, and decreases the incidence of tumor formation. Previous studies evaluating the effect of lifestyle interventions have relied on post-intervention liver biopsy to gauge the improvement in histologic features of NASH [7,18,19,20]. However, because dietary changes designed for weight loss and exercise programs were inextricably linked, improvement in liver function attributable to exercise could not be discerned. Conversely, our study design has circumvented this ambiguity, and we report that exercise without dietary intervention lessens ballooning, a hallmark of NASH, and arrests the fibrotic progression of disease leading to HCC. Since these beneficial effects of an imposed exercise regimen occurred without concomitant weight loss (Figure S1), we claim that exercise is a positive therapeutic measure despite the continued burden of overnutrition. The experimental model we selected was a non-obesogenic, nutrient-deficient, high-fat diet model (Table S1), which offers the advantages of a rapid onset of disease and mirrors the complexity of clinical NASH with fibrosis and development of tumors [15,16,21] but obviates the influence of weight gain [15] and peripheral insulin resistance [22,23]. Hence, it was a robust choice in which to test the short-term effects of regular exercise. In addition, choline-deficiency presents the added liability of a disturbance in the phosphatidylcholine content of hepatic membranes, including mitochondrial membranes, which is a source of mitochondrial dysfunction [24] and contributes to the pathogenesis of NAFLD in this experimental model.Exercise lowered hepatic levels of triglycerides in our CD-HFD mice (Figure 2D and Figure 5A), which is in line with non-invasive data collected from clinical studies [25]. The genesis of steatosis in NAFLD is multi-factorial and could arise from increased fatty acid uptake, increased lipogenesis de novo [26], decreased β-oxidation, impaired triglyceride lipolysis or defective assembly and secretion of VLDL particles. In our CD-HFD mice, an increase in fatty acid uptake by CD36 at the expense of FATP2 and FATP5 was presumed, given the upregulation of CD36 gene expression and the downregulation of FATP2 and FATP5. The upregulation of CD36 is consistent with the increase in plasma and hepatic FFA levels (Figure 4 and Figure 5) and is a feature observed in other experimental models [25,26]. This increased CD36 expression promotes the accumulation of unsaturated FAs, which are a driving force for steatotic triglyceride formation [27].Unlike CD36, the transcriptional downregulation of both FATP2 and FATP5 was not expected. In a study of mice fed a choline-sufficient HFD for 12 weeks, FATP2 and 5 were unchanged [28]. However, when mice were fed a lipogenic methionine choline-deficient diet for 4 weeks, FATP2 and 5 were marginally decreased, although increases in FATP 1 and FATP4 were noted [29]. Apart from any dietary stimulus, the genetic deletion of FATP2 in the liver provokes commensurate increases in CD36 and FATP1 expression, indicating compensatory mechanisms to coordinate and regulate the uptake of lipids [27]. Although we did not investigate the complete profile of fatty acid uptake systems in our CD-HFD model, we deduce that the negative correlation between exercise and hepatic lipid accumulation was independent of the changes in FATP2, FATP5, and CD36. Increased lipogenesis de novo was not a feature of our CD-HFD experimental model [26], since lipogenic enzymes were downregulated (Figure 5). In fact, exercise extended this downregulation further by suppressing the phosphorylation of ATPCL. Impaired lipolysis of hepatocyte triglyceride stores likely contributed to steatosis in our model, since two lipases involved in the sequential hydrolysis of triglycerides, ATGL and HSL, were downregulated in all groups of CD-HFD (Figure 5). Exercise did not change this trend. As a counter-measure to cytoplasmic lipid storage within droplets, expression of perilipin 2 tended to decrease in all groups of CD-HFD (Figure 5), thus rendering triglycerides more available to catabolism via autophagy [30]. Hence, since autophagy is, in turn, stimulated by exercise [31] (Figure 8), we attribute the lower steatosis in the CD-HFD exercised group in part to an increase in autophagic processes. In addition, the modest reduction in PPARα combined with those of CPT1α and MCAD protein levels in the NAFL group suggest that a decrease in fatty acid beta-oxidation could contribute to steatosis in this CD-HFD model. Although slight, exercise was associated with an upregulation of MCAD protein (Figure 5). The downregulation of DGAT2 was more pronounced (Figure 5). While the upregulation of DGAT2 has been associated with increased triglyceride synthesis [32], the consequences of its downregulation in the liver, as observed here, is unclear [33]. DGAT2 has been ascribed a role in VLDL-triglyceride secretion. Therefore, we cannot exclude that exercise has partly reversed the CD-HFD induced DGAT2 suppression and improved VLDL secretion from the liver.In NAFLD, proinflammatory TNFα released by Kupffer cells and steatotic hepatocytes mediates liver injury in part by activating NFkB signaling pathways in stellate cells [27,30]. Exercise reduced the TNFα release elicited by CD-HFD (Figure 6A). Reduced levels of TNFα have been linked to an amelioration of NAFLD, since anti-TNFα antibodies were reported to improve liver histology, to reduce circulating levels of AST and ALT, and to diminish hepatic fat content [34]. Moreover, a correlation has been shown between plasma levels of TNFα and the presence of ballooned hepatocytes [35]. Thus, we find that notwithstanding the continued CD-HFD diet, exercise prompted a decline in hepatocyte steatosis combined with an attendant decrease in TNFα expression leading to a drop in the incidence of ballooning (Figure 2B) and, consequently, to a lower NAFLD activity score (Figure 2C) along with decreases in ALT and AST concentrations (Figure 4). A novel biomarker, IGF-2, is purported to be negatively correlated with the extent of NAFLD and, in particular, to the degree of ballooning [35]. However, in our model, IGF-2 was undetectable in control livers, was highest in the sedentary 20-week CD-HFD group linked to the highest NAFLD score, and was significantly reduced in the exercised group (Figure 6C). Consequently, IGF-2 was positively rather than negatively correlated with disease. In fact, IGF-2 was reported to increase in tumor-bearing livers of mice subjected to a choline-deficient diet and CCl4 treatment [36]. For this reason, the reduced levels of IGF-2 produced in the exercise group likely reflect the fewer numbers of nodules formed. Unlike IGF-2, the levels of IGF-1 were negatively correlated with NASH (Figure 6D). This observation is consistent with clinical reports of NAFLD [37] and with experimental findings in choline-deficient diets [38]. Similarly, neither liver-derived interleukin-6 nor MCP-1 could distinguish exercise from sedentary NASH livers.The ability of exercise to prevent the histological progression of fibrosis (Figure 3A) was confirmed by the changes in circulating markers of hepatic fibrosis, namely, Pro-C3, Pro-C4, and C6M (Figure 3C). Certainly, the histological assessment of fibrosis correlated well with these circulating markers [39]. Pro-C3, a defined epitope of the NH2-terminal propeptide of type III procollagen, is a marker of active fibrogenesis and is released by the protease ADAMTS-2 during collagen maturation, which is a prerequisite for efficient incorporation of collagen type III into collagen fibrils [40]. Pro-C4, a marker of collagen type IV formation, reflects pericellular fibrosis and not bridging reticular fibrotic bands, as does Pro-C3 [41]. C6M detects an internal epitope in the collagen type VI that is exposed by multiple matrix metalloproteinases when the collagen structure is degraded [42]. It is severely upregulated in the fibrotic space of Disse and portal tract stroma and engages in signaling related to the metabolic syndrome and fibrogenesis [42]. These fibrosis biomarkers were all significantly decreased in exercised mice compared to sedentary mice (Figure 3C). In fact, the plasma concentrations of all three biomarkers in the 20-week CD-HFD exercised group were comparable to those of the 12-week NAFL group.Given that fibrosis plays a role in the pathogenesis of HCC and its presence portends a poor prognosis, the significantly lower extent of fibrosis in the exercised CD-HFD groups (Figure 4) predicts that the development of NASH-related tumors would be attenuated in this group. As expected, both the number of animals bearing tumors and the number of hepatocellular adenomas per liver were significantly reduced in exercised mice (Figure 9). In an earlier long term study of a similar CD-HFD diet administered to C57Bl/6J mice, Yoshida et al. queried whether features of the disease spectrum of NAFL-NASH-HCC could be reversed if a standard diet was imposed for 12 weeks after a 36-week regimen of a CD-HFD. Although steatosis and lobular inflammation regressed with a change in diet and fibrosis was partially reversible, the incidences of hepatocellular adenoma and carcinoma progressed [16]. This reveals a difference between the strategies of enforcing an exercise regimen while a deleterious diet continues to expose the liver to the acute consequences of high fat and of merely restricting the offending diet once the inexorable alterations in the molecular machinery, leading to HCC, have taken root. We have previously shown exercise to carry benefits in other models of HCC. For instance, regular physical activity significantly decreases the occurrence of tumors, from 100% down to 70%, in a model of NASH induced by the ablation of hepatocellular PTEN [13]. We also confirmed that the exercise-related mechanism was the activation of AMPK and inhibition of the mTOR/S6K pathway [13,14,37]. This mechanism holds for the current model of dietary-induced NASH. Exercise increased the activating phosphorylation of AMPK, resulting downstream in less phosphorylation of S6 (Figure 9D). However, additional mechanistic changes in other pathways cannot be excluded. The autophagy machinery is relevant when considering tumor development and progression. Autophagy can be modified in both genetic and dietary models of obesity. It is suppressed in the liver, at least in part due to a reduction in the expression level of key autophagy molecules, such as Atg7 [38,39]. In fact, mice with systemic mosaic deletion of Atg5 and liver-specific deletion of Atg7 develop multiple liver tumors [31]. Autophagy deficiency is accompanied by defective insulin signaling and elevated ER stress. In our dietary model, the ER stress was chronic, as reflected by the absence of XBP1 expression and the induction of the pro-apoptotic transcription factor CHOP that transactivates pro-apoptotic proteins (Figure 7). As expected, evidence of an activated pro-apoptotic pathway was detected since Bax was upregulated, and the ratio of Bax to Bcl2 was significantly elevated. However, exercise relieved the autophagic block that can accompany ER stress. Autophagy appeared to be activated by exercise, as shown by the tendency for mTOR downregulation, an increase in LC3BII/LC3BI, and an increase in mitochondrial recruitment of phosphorylated PINK indicative of mitophagy (Figure 8). The benefit of an activated mitophagy process derives from the need to clear hepatocytes of dysfunctional mitochondria that accumulate in NASH [41,42]. Indeed, mitochondria in our CD-HFD model were compromised (Figure 10). The cytochrome c control factor was significantly increased in all respirometry runs, likely because of activation of the pro-apoptotic pathway, loss of cytochrome c, and mitochondrial uncoupling (Figure 10). In addition, the decrease in maximal respiration and complex IV activity is perhaps explained by modest changes in the expression of at least two components of complex IV. Overall, our findings support the general view that exercise stimulates selective autophagic processes in the liver to alleviate hepatocytes of its deleterious burden of lipid overload and dysfunctional mitochondria [43].LSECs undergo morphological and functional changes in NAFLD, leading to sinusoidal capillarization [44]. These phenotypic changes stemming from endothelial defenestration appear early, are features of most experimental models of NASH [45], and have been linked to exposure to dietary fat and circulating free fatty acids [46]. With the progression of NAFLD, LSECs acquire a proinflammatory phenotype and thereby become effectors of liver inflammation in NASH and promoters of fibrosis [47]. LSECs have also been attributed a pro-oncogenic role in HCC through their release of the adipokine fatty acid binding protein-4, which in turn induces hepatocyte proliferation [48]. In our design, sinusoidal capillarization was firmly established at the 12-week NAFL stage before exercise was implemented, and exercise did not reverse the morphological changes in LSECs (Figure 11). Therefore, the anti-tumorigenic actions of exercise are unlikely to implicate pathways central to LSECs.Exercise directly affects skeletal muscle, as shown by the upregulation of PGC1α and PGC1β and activation of Akt and avenues exist for crosstalk between muscle and liver. We have probed FGF21 as a candidate secreted by both tissues and which is exercise responsive. FGF21 is beneficial in NASH [49], and muscle-derived FGF21 could wield hepatic effects, given that in our CD-HFD model the expression in the muscle, but not the liver, increased in response to exercise (Figure 12). Regardless of the mechanism, our findings show that exercise can change the outcome of NAFLD. Exercise offers both protective and therapeutic effects as it intervenes to decrease triglyceride accumulation in hepatocytes even though nutritional overload persists. Consequently, the decrease in proinflammatory cytokines, hepatocyte ballooning, and fibrosis curb the incidence of tumor formation. In parallel, exercise mitigates tumor progression through a decrease in phosphorylated ribosomal protein S6.Mice received humane care and experiments were approved by and conducted according to the regulations of the Bern Animal Welfare Committee, Canton of Bern, Switzerland (BE132/17: 29699 dated 19 February 2018).Male C57Bl/6N mice (Charles River, Freiburg, Germany) were chosen to avoid the Nnt (nucleotide nicotinamide transhydrogenase) mutation carried by C57BL/6J mice. Loss of NNT enzymatic activity has been linked to reduced mitochondrial NADPH/NADP+ ratio, mitochondrial redox abnormalities [50], as well as impaired mitochondrial peroxide metabolism [50] and glucose homeostasis [51]. Mice aged 8 weeks were housed under controlled temperature (22 ± 2 °C) and lighting (12-h light–dark cycles), acclimatized to the facility for one week, then randomly assigned to one of the following four groups and subjected to a diet and activity protocol (Figure 1): (1) mice (n = 11) fed a standard diet for 12 weeks (control group); (2) mice (n = 11) fed a choline-deficient high-fat diet (CD-HFD) for 12 weeks (baseline NAFL group); (3) mice (n = 11) fed CD-HFD for 20 weeks but without exercise (CD-HFD sedentary group); mice (n = 11) fed CD-HFD for 20 weeks but with treadmill exercise from weeks 13 to 20 (CD-HFD exercise group) (Figure 1). Food intake and body weight were monitored weekly.All mice were fed ad libitum. The CD-HFD contained 9% protein, 60% fat, including 2% cholesterol and 31% carbohydrate (HF-CDAA diet, E15673-94, Ssniff Spezialdiäten GmbH, Germany) (Table S1). The standard chow diet contained 12% protein, 16% fat, and 72% carbohydrate (Control diet, E15668-04, Ssniff Spezialdiäten GmbH).After 12 weeks, the CD-HFD exercise mice were placed on a treadmill (running speed of 12.5 m/min) (Förderband GFB, Elmotec, Kleindöttingen, Switzerland) for 60 min from 08.00 h to 09.00 h, corresponding to their waking time. The exercise was imposed 5 days/week for 8 weeks. Sedentary mice remained in their cages.The control and baseline NAFL groups were killed after 12 weeks. The CD-HFD sedentary and exercise groups were killed after 20 weeks, 2 days after the last exercise session. Mice were weighed (Figure S1), then anesthetized deeply with pentobarbital (100 mg/kg i.p.), and blood was collected from the inferior vena cava into heparinized tubes then centrifuged (3000× g, 15 min, 4 °C). Plasma was stored at −80 °C for less than 1 month. Before anesthesia, tail blood lactate and glucose levels were measured with a Lactate Scout Analyzer (Senslab, Leipzig, Germany) and an automated glycemia reader (Ascensia Contour, Bayer Health Care, Zürich, Switzerland). After euthanasia, liver tissue was weighed and divided and either immediately snap-frozen in liquid nitrogen or placed in RNAlater (Sigma-Aldrich R0901, St. Louis, MO, USA) and stored at −80°C, or fixed in 4% phosphate-buffered formaldehyde. Hepatic tumors were counted, sized, and fixed. Activity of alanine transaminase (ALT) and aspartate transaminase (AST), and concentrations of triglycerides, total cholesterol, and bile acids were measured (Cobas analyzer 8000, Roche Diagnostics GmbH, Mannheim, Germany). PRO-C3, the N-protease mediated cleavage of the N-terminal propeptide of type III collagen, PRO-C4, an internal epitope in the 7S domain of type IV collagen, and C6M, a neo-epitope of the proteolytic degradation of type VI collagen, were measured by means of a competitive enzyme-linked immunosorbent assay (Nordic Bioscience, Herlev, Denmark), as described [52].Formaldehyde-fixed, paraffin-embedded liver tissues were stained with hematoxylin and eosin (H&E) and examined for steatosis, NASH lesions, and tumors by a pathologist blinded to treatment conditions (LMT). The NAS score was determined as previously defined by Kleiner et al. [53]. The degree of fibrosis was assessed on sections stained with Sirius Red and visualized with a panorama scanner and case viewer (3D Histech) and 10× objective. Eight digital images were collected from different areas of the left, median, and right lobes, and the signals were quantified with MetaMorph® image analysis software (Molecular Devices, Sunnyvale, CA, USA). Tumor types were assessed as previously described [17]. Oil Red O staining was performed on frozen sections as previously described [54].Total triglyceride content was measured with the PicoProbeTM triglyceride fluorometric assay (BioVision, Milpitas, CA, USA). Total free fatty acid content was quantified by means of the fluorometric FFA kit (BioVision).Livers were homogenized in RIPA buffer (150 mM NaCl, 1% NP-40, 0.5% Na-deoxycholate, 0.1% SDS, and 50 mM Tris-HCl pH 7.4) containing protease and phosphatase inhibitors (Roche, Rotkreuz, Switzerland). Protein concentration was measured with the PierceTM BCA assay (Thermo Fisher Scientific, Rockford, IL, USA). Equal amounts of proteins were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to nitrocellulose membranes, blocked for 1 h with 5% nonfat milk or BSA, then incubated overnight at 4 °C with primary antibodies (Table S2). After incubation with peroxidase-conjugated secondary antibody (Thermo Fisher Scientific, Rockford, IL, USA), signals were revealed with enhanced chemiluminescence (Amersham ECL Prime, GE Healthcare, Glattburg, Switzerland) and a Fusion CCD camera coupled to a computer equipped with Fusion Capt Fx Software (Vilber-Lourmat, Marne-la-Vallée, France). Signals were quantified with the Bio-1D Advanced software (Vilber-Lourmat). Total RNA was extracted with the RNeasy Mini Kit (Qiagen, Hombrechtikon, Switzerland) and stored at −80 °C. RNA was reverse-transcribed (SuperScript III Reverse Transcriptase, Invitrogen, Basel, Switzerland). The gene primer, FAM-labelled probe and the TaqMan Universal PCR Master Mix were obtained from Applied Biosystems (Beverly, MA, USA) (Table S3). Amplification was performed with a CFX Connect Real-Time System (Bio-Rad, Hercules, CA, USA). The ΔCt values were calculated relative to β2-microglobulin as the housekeeping gene. Values are triplicates and are reported as fold increase or decrease relative to the controls and calculated as 2-ΔΔCt.Livers were perfused through the portal vein with fixation solution (2.5% glutaraldehyde, 2% formaldehyde, 2 mM CaCl2, 2% sucrose and 0.1 M sodium cacodylate (pH 7.4)) for 5 min. Fixed tissue was cut in blocks (1 × 1 × 5 mm) and stored in 2% formaldehyde at 4 ℃ until processed. For SEM, fixed livers were treated with 1% osmium tetroxide, dehydrated in a graded series of ethanol, and dried. The sections were coated with platinum/palladium and visualized under an S-4700 electron microscope (Hitachi, Tokyo, Japan). For evaluation of capillarization, the percent of open space area in the liver sinusoidal endothelial cells (LSECs; porosity) was measured in 15 randomly selected fields at ×10,000 magnification on at least three animals per group, using ImageJ software.Oxygen flux was measured in freshly isolated mitochondria by respirometry (Oxygraph-2k; Oroboros Instruments, Innsbruck, Austria). Mitochondria (200 µg) were added to 2 mL of respiration buffer (110 mM sucrose, 60 mM K+-lactobionate, 0.5 mM EGTA, 3 mM MgCl2, 20 mM taurine, 10 mM KH2PO4, 20 mM HEPES (pH 7.1), at 37 °C). Oxidative phosphorylation was estimated with complex I (pyruvate 5 mM, malate 2 mM, glutamate 5 mM) and complex II (succinate 10 mM) substrates in the presence of ADP (2.5 mM). Leak respiration was recorded after the addition of oligomycin (2.5 μM). For maximum uncoupled respiration, the protonophore carbonyl cyanide m-chlorophenyl hydrazine (CCCP) was titrated in 0.5 μM increments until maximal stimulation of respiration. The protocol was terminated by assessing non-mitochondrial respiration with the complex I and III inhibitor, rotenone (0.5 mM), and antimycin A (2.5 mM), respectively. Finally, the activity at complex IV was recorded with the artificial substrate N,N,N9,N9-tetramethyl-p-phenylenediamine dihydrochloride (TMPD; 0.5 mM) and ascorbic acid (2 mM), and inhibited with azide (100 mM). The cytochrome c control factor was measured after simulation of respiration with exogenous cytochrome c (10 µM). Respiration states were corrected for non-mitochondrial respiration, and complex IV activity was corrected for azide inhibition. Values were normalized for protein, as described previously [54].Data are presented as the mean values ± standard deviations (SD). Statistical comparisons were made between control and CD-HFD groups terminated at 12 weeks, and between SED and EXE groups terminated at 20 weeks, except for assessment of cytochrome c control factor and citrate synthase activity, where all four groups were compared. The normality of data was assessed by the Kolmogorov–Smirnov test. The nonparametric Mann–Whitney U-test was applied in the case of non-normal distributions. Fischer’s exact test was applied to frequency tables. A p-value ≤0.05 was considered statistically significant.Our work complements the small number of studies that have evaluated the positive effect of lifestyle interventions on the histological features of NASH [18,19,20]. In addition, our study confirms in yet another model that exercise not only arrests the development of liver tumors [13] but attenuates progression [14]. Furthermore, we have reinforced the notion that the benefit of exercise in suppressing tumors is sustained in NAFLD well after sinusoidal capillarization, ER stress, apoptotic processes, and evidence of mitochondrial dysfunction are established. Finally, we provide direct evidence that exercise alone can be a therapeutic measure and not only a preventive measure in NAFLD, and this should offer hope to patients who fail at sustained, consequential dietary changes, a situation routinely confronting the clinician [55].The following are available online at https://www.mdpi.com/2072-6694/12/6/1407/s1, Figure S1: Comparison of body weight and of liver to body weight ratios; Figure S2: Unabridged western blot images; Table S1: Composition of standard and choline-deficient high-fat diets; Table S2: Description and source of antibodies used in the study; Table S3: Description and source of semi-quantitative PCR reagents. J.-F.D., M.V.S.-P., M.G., B.H., and M.F. designed the experiments. M.G., P.K., A.F., S.G.-M., and L.M.T. performed the experiments and analyzed the data. M.V.S.-P., M.G., and J.-F.D. wrote the manuscript. M.G. and P.K. prepared the figures. J.-F.D., L.M.T., A.F., and J.-M.N. provided a critical revision of the manuscript. All authors read and approved the final manuscript. This research was funded by the University Federico II of Naples and Compagnia di San Paolo, as part of Program STAR (M. Guarino), by the Stiftung für Leberkranheiten (S. Guixé-Muntet and M. St-Pierre) and by the Swiss National Foundation CRSII3_160717 (J-F. Dufour, B. Humar, and M. Foti) and 310030_185219 (J-F. Dufour).We thank Philipp Kellmann, David Bélet, and Marco Amsler for their laboratory support. We also thank Dr. Nico Ruprecht, Radiology-Department of Biomedical Research, the University of Bern for instruction and use of the CFX Connect Real-Time System for semi-quantitative PCR. We also thank Prof. Jordi Gracia-Sancho and the Advanced Microscopy Unit, Faculty of Medicine, University of Barcelona for scanning electron microscopy. We are grateful to Nordic Bioscience, Herlev, Denmark, for measuring the Pro-C3, Pro-C4, and C6M fibrosis markers in the plasma.The authors declare no conflict of interest.Schematic outline of the study design. Male C57Bl/6N mice were randomized to one of four groups: (1) the control group (n = 11) was fed a standard diet and tissues were collected after 12 weeks; (2) the non-alcoholic fatty liver (NAFL) group (n = 11) was fed a choline-deficient high-fat (CD-HFD) diet for 12 weeks before tissue collection; (3) the non-alcoholic steatohepatitis (NASH) group (n = 11) received a CD-HFD for 20 weeks and remained sedentary before tissue collection; (4) the NASH + exercise (EXE) group (n = 11) received a CD-HFD for 20 weeks but with treadmill running at 12.5 m/min imposed from weeks 12 to 20.Effect of exercise on liver histology in mice fed a choline-deficient high-fat diet (CD-HFD). (A) Microscopy of hematoxylin and eosin (H&E)-stained liver sections showing diffuse macrovesicular steatosis in the NAFL, NASH and NASH + exercise (EXE) groups and the presence of ballooned hepatocytes only in the NASH sedentary group ). (B) Frequency table and dot plot comparing the ballooning score in the NASH sedentary and NASH+EXE groups. Ballooning was significantly lower in the NASH + EXE group (Fisher’s exact test, p = 0.005). (C) Frequency table and dot plot showing the NAFLD activity score in the NAFL, NASH sedentary, and NASH + EXE groups. The score was significantly lower in NASH + EXE than in the NASH sedentary group (Fisher’s exact test with Freeman–Halton extension, NASH vs. NASH + EXE, p < 0.0001). (D) Oil Red O staining comparing neutral lipid content of control, NAFL, NASH sedentary, and NASH + EXE livers. The quantification of lipid staining (right panel) was done with MetaMorph® analysis software. Lipid content was lower in the NASH + EXE livers (unpaired t-test; * p < 0.05).Effect of exercise on liver fibrosis. (A) Microscopy of Sirius Red-stained liver sections from C57BL/6N mice fed a control diet or a choline-deficient high-fat diet (CD-HFD). Fibrosis was absent in controls. Diffuse lobular pericellular fibrosis was present in the NAFL and NASH + exercise (EXE) groups but was highest in the NASH sedentary group. (B) Quantification of fibrosis shown in panel A. Images were quantified with the MetaMorph® analysis software. Fibrosis was higher in NAFL than in control groups and higher in NASH sedentary than in NASH + EXE group (unpaired t-test; ****p < 0.0001). (C) Fibrosis biomarkers in plasma. PRO-C3, PRO-C4, and C6M concentrations were measured in the plasma of mice from the control, NAFL, NASH, and NASH + EXE groups. PRO-C3 and C6M were significantly higher in NAFL than in controls and in NASH sedentary than in NASH+EXE (unpaired t-test; *p < 0.05; **p < 0.005).Effect of exercise on biochemical values in plasma. Plasma concentrations of alanine transaminase (ALT), aspartate transaminase (AST), total bile acids, triglycerides, cholesterol and free fatty acids (FFA), and fasting blood glucose were compared in mice fed a control diet or a choline-deficient high-fat diet (CD-HFD) for 12 weeks (NAFL), and in mice fed a CD-HFD for 20 weeks with (NASH + EXE) or without exercise (NASH). ALT and AST were higher in NAFL than in controls, and higher in NASH sedentary than in NASH + EXE (unpaired t-test; *p < 0.05; ***p < 0.001; ****p < 0.0001). Bile acids were elevated in NAFL vs. controls (p < 0.001). Triglycerides were lower in NASH + EXE than in NASH sedentary (p < 0.05). Cholesterol was higher in NAFL than in controls. FFA was higher in NAFL than in controls (unpaired t-test; p < 0.05).Effect of exercise on hepatic lipid metabolism. (A) Liver content of triglycerides and free fatty acids (FFA) were compared in mice fed a control diet or a choline-deficient high-fat diet (CD-HFD) for 12 weeks (NAFL) and in mice fed a CD-HFD for 20 weeks with (NASH + EXE) or without exercise (NASH) (unpaired t-test; ** p < 0.005; *** p < 0.001; **** p < 0.0001). (B) Semi-quantitative PCR measurement of CD36 mRNA and fatty acid transport protein 2 (FATP2) and fatty acid transport protein 5 (FATP5) levels in liver extracts relative to β2-microglobulin. CD36 was higher in NAFL than in controls (p < 0.001). FATP5 and FATP2 were lower in NAFL than in controls. (C) Immunoblots of liver homogenates comparing control and NAFL mice, and NASH and NASH + EXE groups. The signals from immunoblots were quantified and normalized with vinculin and are reported as mean ± SD (n = 3 per group). (ATGL, adipose triglyceride lipase; ATPCL, ATP citrate lyase; FAS, fatty acid synthase; HSL, hormone-sensitive lipase; CPT1A, carnitine palmitoyltransferase alpha; MCAD, medium-chain acyl-CoA dehydrogenase (unpaired t-test; * p < 0.05; ** p < 0.005; *** p < 0.001). Details of Western blots are given in Figure S2.Effect of exercise on expression of liver biomarkers. Semi-quantitative PCR analysis of mRNA expression relative to β2 microglobulin in livers from control, NAFL, NASH, and NASH + EXE groups. (A) Tumor necrosis factor α (TNFα) was increased in NAFL vs. control (p < 0.001) and decreased in NASH + EXE vs. NASH sedentary (p = 0.05). (B) Transforming growth factor β (TGFβ1) was increased in the NAFL vs. control group (p < 0.001) but not changed in NASH vs. NASH + EXE. (C) Insulin-like growth factor 2 (IGF-2) was increased in the NASH sedentary relative to NASH + EXE. (D) IGF-1 was decreased by CD-HFD. (E) Interleukin-6 (IL6) increased in NAFL. (F) Monocyte chemoattractant protein 1 (MCP-1) increased in NAFL. Groups are described in Figure 1 (unpaired t-test; *p ≤ 0.05; ***p < 0.001; ****p < 0.0001).Effect of diet and exercise on endoplasmic reticulum (ER) stress and apoptosis. Immunoblots (upper panel) and quantification (lower panel) of proteins expressed in livers from control, NAFL, NASH, and NASH + EXE groups. Comparisons were made between the control and NAFL groups at 12 weeks, and between the NASH and NASH + EXE groups treated for 20 weeks. Binding immunoglobulin protein (BiP) was not different between groups. X-box binding protein (XBP-1) was lowered in NAFL vs. control and remained very low in both NASH groups. C/EBP homologous protein (CHOP) expression was higher in NAFL than in controls and increased further in the NASH groups. The ratio of Bcl-2 associated X protein (Bax) to Bcl-2 was elevated in NAFL vs. control but was not significantly changed by exercise. Vinculin served as the loading control. Groups are as described in Figure 1 (unpaired t-test; *p < 0.05; ***p < 0.001). For more details of Western blots, please view Figure S2.Effect of exercise on autophagy. Immunoblots of protein expression (upper panels) and quantification (lower panels) in liver homogenates (left panel) and mitochondria (right panel) for markers of autophagy in the NASH sedentary and NASH + EXE groups. mTOR and its phosphorylation tended to decrease in the NASH + EXE group. Autophagy related gene 5 (ATG5) tended to decrease in the NASH + EXE group. The ratio of light chain 3BII (LCBII) to LCBI was increased in the NASH + EXE group (p < 0.05). Vinculin served as the loading control. The phosphorylation of PTEN-induced kinase (P-PINK) increased in mitochondria of the NASH + EXE group (p < 0.05). Citrate synthase served as the loading control. Groups are as described in Figure 1 (unpaired t-test; *p < 0.05; **p < 0.005). For more details of Western blots, please view Figure S2.Effect of exercise on liver tumors. (A) Representative histological images of hematoxylin and eosin (H&E)-stained liver sections with adenomas in NASH sedentary (left panel) and exercise (right panel) mice (magnification x40). (B). Comparison of the number of liver adenomas in NASH sedentary and NASH + EXE mice. NASH + EXE mice developed fewer nodules than did NASH sedentary mice (Mann–Whitney U-test; *p < 0.05). (C). Immunoblots showing protein expression and phosphorylation status of AMP-activated protein kinase (AMPK), regulated associated protein of mTOR (Raptor) and ribosomal protein S6. Signals were quantified and normalized for vinculin. Phosphorylation of AMPK was increased and phosphorylation of S6 was decreased in the NASH-EXE group. Groups are as described in Figure 1 (unpaired t-test; p < 0.05). For more details of Western blots, please view Figure S2. (D) Schematic representation of the AMPK–Raptor–S6 cascade.Effect of exercise on mitochondrial bioenergetics. (A) High-resolution respirometry of oxygen consumption (O2 flux) in mitochondria isolated from livers of control vs. NAFL group (12-week treatment) (left panel) and NASH sedentary vs. NASH + EXE groups (right panel). O2 flux was measured with an O2k Oroboros instrument. Mitochondria were exposed to sequential additions of pyruvate/malate, glutamate, succinate, ADP, cytochrome c, oligomycin, rotenone, antimycin A, TMPD, ascorbate, and azide. Coupled complex I- and complex II–driven, leak respiration, maximal respiration, and complex IV OCRs were recorded and normalized for protein (n = 3 per group). (B) Comparison of cytochrome c control factors. The ratios were measured as the fractional change of O2 flux after the addition of excess cytochrome c and calculated as (Flux CI + IIcytc -Flux CI + II)/Flux CI + IIcytc. The ratio was significantly elevated in NAFL and NASH groups relative to control. (C) Comparison of mitochondrial citrate synthase activity in control vs. NAFL group and NASH sedentary vs. NASH + EXE groups. Activity was normalized for mitochondrial protein and was higher in NAFL vs. control (one way ANOVA; *p < 0.05). (D) Immunoblot showing expression of cytochrome c and of cytochrome c oxidase (COX) subunits 1 and 4 in mitochondria. Citrate synthase served as the loading control. Cytochrome c and COX1 and COX4 were lower in NAFL vs. control groups. Expression of PGC1α and PGC1β was evaluated with vinculin as the loading control (unpaired t-test; *p < 0.05; **p < 0.005; ***p < 0.001). For more details of Western blots, please view Figure S2.Effect of exercise on liver endothelium. (A) Scanning electron microscopy images of sinusoids showing endothelial cells from livers of control, 12-week CD-HFD treated (NAFL), 20-week CD-HFD sedentary (NASH), and 20-week CD-HFD exercised mice (magnification ×15,000) (upper panel). Porosity, defined as the percentage of endothelial cell membrane perforated by fenestrations, was quantified (bottom panel) (unpaired t-test; ***p < 0.001). (B) Immunoblots of the gene product of nitric oxide synthase 3 (NOS3) and its phosphorylated form expressed in liver homogenates of control and NAFL groups (12 weeks treatment), and NASH and NASH + EXE groups (20-week treatment). Immunoblots were quantified and normalized with vinculin. The NAFL, NASH, and NASH + EXE groups are described in Figure 1. For more details of Western blots, please view Figure S2.Effect of exercise on skeletal muscle. (A) Immunoblots of peroxisome proliferator-activated receptor gamma coactivator 1α and 1β- (PGC1α, PGC1 β) and phosphorylated Akt in skeletal muscle. (B) FGF21 mRNA expression in skeletal muscle. (C) FGF21 mRNA expression in liver. Vinculin served as the loading control (unpaired t-test; *p < 0.05; ****p < 0.0001). For more details of Western blots, please view Figure S2.

Med-MDPI/cancers/cancers-12-06-01408.txt

@@ -0,0 +1 @@
+These authors contributed equally to this work.The incidence of obesity and colorectal cancer (CRC) has risen rapidly in recent decades. More than 650 million obese and 2 billion overweight individuals are currently living in the world. CRC is the third most common cancer. Obesity is regarded as one of the key environmental risk factors for the pathogenesis of CRC. In the present review, we mainly focus on the epidemiology of obesity and CRC in the world, the United States, and China. We also summarize the molecular mechanisms linking obesity to CRC in different aspects, including nutriology, adipokines and hormones, inflammation, gut microbiota, and bile acids. The unmet medical needs for obesity-related CRC are still remarkable. Understanding the molecular basis of these associations will help develop novel therapeutic targets and approaches for the treatment of obesity-related CRC.Obesity is associated with various metabolic disorders [1], such as diabetes, non-alcoholic fatty liver diseases, cardiovascular diseases, hypertension, and obstructive sleep apnea syndrome, as well as with some cancers [2,3,4], including esophageal adenocarcinoma, multiple myeloma, cardia cancer, colorectal cancer (CRC), cholangiocarcinoma, pancreatic cancer, breast cancer, endometrial cancer, ovarian cancer, and renal cancer. Obesity is closely related to increased incidence and progression of these cancers, and it is estimated to cause about 20% cancer-associated deaths [5,6]. In this review, we mainly focus on the epidemiology of obesity and CRC in the world, the United States, and China, and the molecular mechanisms of obesity contributing to CRC.Obesity has become a worldwide health burden. Body mass index (BMI) is a typical value derived from the weight and height to define overweight (25 ≥ BMI < 30) and obesity (BMI ≥ 30) in adult men and women. According to the World Health Organization (WHO) reports, the rate of obesity has nearly tripled globally since 1975. In 2016, about 2 billion adults were overweight, and more than 650 million of them were obese. The worldwide prevalence of overweight was 22.7% in women, and 20.7% in men in 1975; it was markedly increasing to 39.0% and 38.3% in 2016 and it will arrive at 49.6% and 51.7% in women and men respectively in 2035 (Figure 1A). The global prevalence of obesity was 6.3% for women, and 2.9% for men in 1975; this proportion rose to 15.1% (women) and 11.1% (men) in 2016 and will reach 21.6% (women) and 18.1% (men) in 2035 (Figure 1B). The regions with the highest prevalence of obesity are American and European [7]. With an estimated 89.6 million obese, China has the largest population of obese in the world [8]. Since 1975, the prevalence of overweight and obesity in men and women every two decades in China and the United States is shown in Figure 1C,D. If the current trends continue, as predicted, the prevalence of overweight and obesity in the USA will reach 76.9% (women) and 87.1% (men), and 48.1% (women) and 46.7% (men) in 2035, respectively. The prevalence of overweight and obesity in China will reach 43.3% (women) and 58.3% (men), and 12.8% both in women and men in 2035, respectively. Obesity has been a serious threat to human health and a heavy financial burden of health insurance, which affects the normal physiological function of humans.CRC is the third most prevalent cancer and is also the third leading cause of cancer-associated death globally in both men and women from the 1980s [9,10]. In 2018, there were 1.8 million new CRC cases, causing 0.86 million deaths worldwide, according to global cancer statistics [11]. Currently, there are more than 1 million CRC survivors in America. Based on American Cancer Society statistics 2020, the estimated numbers of new CRC cases and deaths in the United States are approximately 150,000 and 54,000, respectively [9]. Global Burden of Disease Study 2017 (GBD 2017) reported the numbers of incident cases and deaths of CRC globally, in the USA, and China from 1990 to 2017, as shown in Figure 2. We observed that in 1990, incident cases and deaths of CRC are about 107,000 and 76,000 in China, and about 432,000 and 200,000 in 2017, respectively [12]. Over the past 27 years, the incidence cases of CRC have doubled worldwide, and been increased three times in China. The unmet medical needs of CRC have been a growing public health issue.Growing epidemiological data indicated a strong positive correlation between obesity and colorectal carcinogenesis [13,14,15]. General obesity causes a higher risk of colon cancer in males compared to females, and it has a stronger association with colon cancer than rectal cancer in both genders [16,17]. Dose-response meta-analysis reported that body weight gain of 10 kg was accompanied by approximately 8% increased risk of CRC [18,19]. Early-life obese individuals are at greater risk of developing CRC in adulthood [13,18,20]. As expected, body weight loss by bariatric surgery reduces about 27% risk of CRC [21,22]. Understanding the association between body weight and the risk of CRC is essential to guide body weight management for CRC patients.Although increasing evidence suggests the positive correlation between obesity and CRC, the underlying molecular mechanisms are still not fully understood. Obesity-induced abnormal lipid metabolism, adipokines and hormones, chronic inflammation, gut microbiota dysbiosis, and disrupted bile acid homeostasis may play important roles in the complex metabolic regulation of CRC tumorigenesis.Obesity is excess body adiposity, especially ectopic deposition of white adipose tissues. Mature adipocytes (white adipocytes) act as an energy bank to store and release energy [23]. Systemic and local energy metabolic homeostasis is primarily controlled by adipocytes [24,25]. Tumor cell growth requires a lot of energy. Understanding whether and how tumor cells get energy directly from the adipocytes helps develop new therapeutic strategies.Nieman et al. reported [26] that intra-abdominal tumors are more likely home to and proliferate in the omentum majus, which is an organ mainly composed of white adipocytes. Adipocyte-tumor cell coculture induces lipolysis in adipocytes and β-oxidation in tumor cells, resulting in the rapid proliferation of tumor cells. An emerging concept in cancer metabolism is that the adipocytes surrounding tumors provide energy or nutrients for the anabolic growth of cancer cells [27,28,29]. We validated this concept by observing more adipocytes surrounding colorectal tumor tissues than normal tissues in clinical pathological sections [30]. In in vitro experiments, we found adipocyte-conditioned medium promotes proliferation and migration of colon cancer cells (SW480 and C26) through retinoic acid-related orphan α (RORα), which is a lipid metabolism-associated nuclear receptor [30]. Sadahiro et al. reported that primary adipocytes, preadipocytes, and adipose tissues enhanced the growth of colon cancer cells (CACO-2, T84, and HT29) in the cocultured system [31]. Adipocytes are part of tumor microenvironment. It is domesticated to produce and transfer energy-rich metabolites to tumor cells, including free fatty acids, glutamine, ketones, and L-lactate, and promote the growth and migration of tumors [29]. The summarized crosstalk between CRC cells and adipocytes in nutriology is shown in Figure 3. CRC cells domesticate adipocytes which supply energy or nutrients to cancer cells for further rapid growth.Cancer cachexia (CC), also known as wasting syndrome, is characterized by weight loss in cancer patients. It is caused by tumor factors and regulated by catabolic metabolism [32]. This complex multifactorial metabolic syndrome often accompanies increased lipolysis in adipose tissues. A total of 54% of colon cancer patients suffer from CC that causes about 20% of cancer-associated deaths [33,34]. It might be a piece of evidence that adipose tissues provide nutrients for tumor growth in systemic nutriology.Understanding the role of adipocytes in tumor microenvironment is critical to the discovery of new strategies. Targeted blocking energy transfers might be novel therapies for the treatment of CRC.Adipose tissues have long been thought to be energy storage tissues as the body accumulates excess nutrients and to resist cold temperature [35]. It is currently regarded as a highly active endocrine or metabolic organ [36]. It liberates more than twenty kinds of hormones and adipokines, such as estrogens, insulin, insulin-like growth factors (IGFs), leptin, adiponectin, apelin, visfatin, resistin, chemerin, omentin, nesfatin, vaspin, inflammatory cytokines (e.g., tumor necrosis factor-alpha (TNF-α), chemokine (C-C motif) ligand 2 (CCL2), plasminogen activator inhibitor-1(PAI-1), and the interleukin families (e.g., IL-1β, IL-6, IL-8, IL-10, IL-27, and IL-31). The related adipokines and hormones and their functions in the development and progression of CRC are introduced below.The insulin/IGFs system is a major driver in the pathogenesis of CRC. This system consists of insulin, insulin receptor (IR), IGF-1 and -2, IGF-1 receptor (IGF-1R), IGF-binding protein (IGFBP)-1 and -2, and IR substrates (IRS) 1 and 2 [37]. Overweight generally increases the levels of insulin and IGF-I and decreases the levels of IGFBP-1 and IGFBP-2 in serum [38]. Insulin and IGFs have been reported to promote the proliferation of HCT116 and HT29 colon cancer cell lines through activation of the phosphoinositide 3-kinase (PI3K)/Akt signaling pathway [39,40,41]. PI3K/Akt signal pathway is an important therapeutic target for treating colon cancer [42,43]. Tyrosine-protein kinase Src is a non-receptor tyrosine kinase encoded by the SRC gene in humans [44]. It regulates PI3K/Akt pathway through phosphorylation of PI3K. Src also plays a critical role in the transformation and growth of CRC cells. Knockdown or inhibition of Src inhibited cell metastasis and proliferation in human cancer cells SW480 and HT29 [45,46]. Phosphorylated IR (pIR) was highly expressed in low-grade colorectal adenocarcinoma, which indicated activation of IR is an early event in CRC tumorigenesis [47]. The expression levels of IGF1 and IGF-1R were increased in colorectal carcinomas, compared with normal colonic mucosa. Overexpression and activation of IGF1-R can activate Src, leading to elevated proliferation and migration of colon cancer in vitro [48]. Renehan et al. reported that IGF-2 SD scores (SDS) were slightly increased in CRC patients compared to healthy controls, and it showed a more dramatic increase in advanced colonic carcinomas compared with earlier stages, but the scores dropped down immediately after curative resection [37]. Taken together, the insulin and IGFs system plays an important role in the pathogenesis and prognosis of CRC through independent or joint signaling networks.Leptin, a peptide hormone encoded by Ob gene, is mainly secreted by adipose tissues, which informs the brain that the energy runs out in the liver through binding to leptin receptors [49,50,51]. Obese individuals have high levels of circulating leptin, because of leptin resistance [52]. Leptin is a risk factor for CRC [53,54]. The expression of leptin is increased in human colorectal tumors and is associated with tumor progression and clinic pathological parameters [55]. Soluble leptin receptor (sOB-R) is a potential marker of leptin resistance. European Prospective Investigation into Cancer and Nutrition (EPIC) cohort also showed circulating sOB-R inversely correlated with the risk of CRC [56,57]. In azoxymethane (AOM) induced murine colon cancer model, Leptin-deficient (ob/ob) and leptin receptor-deficient (db/db) mice showed inhibited tumor growth through Wnt signaling pathway [54]. Leptin increases cell proliferation and prevents apoptosis in HT29 cells through phosphorylation of c-Jun NH2-terminal kinase (JNK). JNK phosphorylation stimulates a cascade of downstream protein phosphorylation, including Janus kinase 2 (JAK2) and PI3K/Akt, then activates signal transducer and activator of transcription (STAT3) and activator protein 1 (AP-1) [58]. Leptin promotes cell migration and lamellipodial extension in human CRC cell lines LS174T and HM7 through activation of Rho family of GTPases, including ras homolog family member A (RhoA), cell division control protein 42 (Cdc42), and ras-related C3 botulinum toxin substrate 1 (Rac1) [59]. Adipose tissues secreted leptin inhibits mitochondrial respiration rate in HCT116 cells [60,61]. Leptin provides a link between obesity and the risk of CRC, it is a sensitive marker of obesity-induced hormonal aberrations and may be directly involved in CRC tumorigenesis.Adiponectin is a protein hormone encoded by ADIPOQ gene in humans [62]. It is one of the most abundant hormones released from adipose tissues and performs an essential function in obesity-associated cancers. The expression and circulating levels of adiponectin are reduced in most obese individuals and animal models of obesity [63,64,65]. Epidemiology studies showed that decreased plasma adiponectin levels are inversely correlated with the risk of colon cancer [66,67]. Adiponectin knockout (APNKO) mice exhibited more tumor numbers and areas in dextran sodium sulfate (DSS)  and  1,2-dimethylhydrazine (DMH) induced colon cancer model through increasing the differentiation from epithelial cells to goblet cells and inhibiting goblet cell apoptosis. It indicated that adiponectin protected against chronic inflammation-induced colon cancer [68]. High-fat diet treated mice had more and larger colorectal tumors than chow-diet mice. Adiponectin administration decreased tumor growth through inhibiting angiogenesis [69,70]. In vitro experiments, adiponectin inhibits colon cancer cell growth in adiponectin receptor (AdipoR1- and -R2) positive HCT116, HT29, and LoVo cells through the AMP-activated protein kinase (AMPK)/mammalian target of rapamycin (mTOR) signaling pathway [71,72]. Moon et al. demonstrated that adiponectin directly regulated cell proliferation, migration, adhesion, and colon formation through regulation of metabolism, inflammation, and cell cycle in MCA38, HT29, HCT116, and LoVo cells [69]. These results indicate the potential inhibitory effect of adiponectin on the development of CRC. Together, leptin and adiponectin generally show opposite molecular effects on obesity and cellular behaviors. They are relevant but reverse players in obesity-related CRC.It is well established that estrogen contributes to obesity-associated hormone-responsive cancers, especially breast cancer [73,74]. The role of estrogen in obesity-associated CRC is complicated. First, estrogens have been found to reduce the risk of CRC [75]. Hormone replacement therapy confers protection against CRC, especially for lean women, as indicated by epidemiological data [76]. Estrogen replacement therapy in postmenopausal women reduces CRC-related mortality [77]. These cohort studies indicated estrogens may play a protective role in the pathogenesis of CRC. Interestingly, adipose tissues are also partial source of estrogen in addition to ovaries. Plasma estrogen levels are increased in obese men and postmenopausal women, because adipose tissue aromatase transforms androgenic precursors to estrogens [78]. However, several studies have shown that high BMI increased the risk of CRC in men and premenopausal women, but not postmenopausal women [79,80]. Adiposity also positively correlates with blood insulin, leading to increased IGF-1. The inducible effect of insulin/IGF-1 axis on CRC appears to be compromised by estrogen released from adiposity in postmenopausal women. In premenopausal women, the primary source of estrogen is ovary compared to adiposity. Thus, more hormone supplement cannot provide more benefits [79,81]. This concept has been suggested by several cohort studies showing a positive correlation between BMI and CRC risk in younger women (<55-year-old) but not in older women [79,80,82]. This association was further confirmed by the study subjected between BMI and CRC risk in premenopausal and postmenopausal women. The risk of CRC in postmenopausal women is independent of BMI [83]. Although the relationship among hormones, obesity, and CRC is not fully understood, these observations and reasonable speculation emphasize the same importance of weight control in both genders.The effect of estrogen is mediated by its receptors, estrogen receptor (ER)-α and ER-β. The expression of ER-α is very low in normal colorectal tissues. However, the ER-α expression is increased with the development of colon cancer, and it positively correlates with CRC stages and worse survival [75]. ER-β is enriched in colon tissues [84]. The expression of ER-β is lower in colon tumor tissues compared with normal tissues and inversely correlates with the progression of CRC [85,86]. ER-β overexpression induced cell-cycle arrest and inhibited cell proliferation and tumor growth in SW480 cells and mouse xenografts model [87]. In the ApcMin/+ mouse model, estrogen treatment protected against CRC and increased the ratio of ER-β to ER-α [88]. Ablation of ER-β in ApcMin/+ mice significantly increased tumor formation, and treatment with estrogen could not prevent this phenotype [89]. These results indicate that ER-β is responsible for the protective effect of estrogens on colon tumorigenesis.We summarize the signaling pathways of obesity-secreted adipokines and hormones in the pathogenesis of CRC (Figure 4).Obesity, as a characteristic of metabolic syndrome, is related to chronic low-grade inflammation in obese subjects, because of various pro- and anti-inflammatory cytokines produced by adipose tissues, including IL-6, TNF-α, CCL2, PAI-1, and others [90,91]. Chronic inflammation is a major link between obesity and tumor microenvironment in CRC.Obesity is associated with circulating levels of IL-6. It has been reported that about 30% of circulating IL-6 was secreted from adipose tissues [91,92]. Circulating IL-6 is an important inflammatory factor in the acute inflammatory reaction which stimulates C-reactive protein (CRP) synthesis and secretion in the liver [91]. IL-6 is found in the tumor microenvironment of both murine and human colon cancer [93,94]. Prediagnostic plasma CRP, a general marker for inflammation, is also a reliable biomarker for CRC clinically [95]. Elevated levels of circulating CRP or IL-6 in CRC patients were associated with cancer progression, relapse, and worse survival [96,97]. IL-6 may act as a CRC-promoting cytokine due to its inflammatory property.TNF-α is also secreted from adipose tissues. TNF-α expression in adipose tissues is positively correlated with the degree of obesity and associated type 2 diabetes mellitus(T2DM) [98,99]. The production of TNF-α is elevated in IBD patients and it is involved in the pathogenesis of IBD and associated CRC [100,101,102]. TNF-α can stimulate NF-κB activation, and the activation of IKK/NF-κB pathway is indispensable for colitis and colorectal carcinogenesis [103,104]. TNF-α promoted the proliferation and migration of CD44+CD133+ HT29 cells by activation of Wnt/β-catenin signaling pathway [101]. Treatment with low concentrations of TNF-α (20 µg/L) enhanced cell migration and invasion in HCT116 cells through upregulating tumor-associated calcium signal transduction protein 2 (TROP-2) by phosphorylation of extracellular signal-regulated kinase (ERK)1/2 signaling pathway [105].CCL2, also known as monocyte chemoattractant protein 1 (MCP-1), is secreted by adipocytes and plays a crucial role in inflammatory reaction [106]. Circulating levels of pro-inflammatory CCL2 is also increased in obese subjects [107]. Tumor-associated macrophage induced inflammation is related to poor prognosis of CRC. CCL2 is an imperative monocyte-attracting chemokine stimulating the recruitment of macrophages into the sites of tumors [108]. Knockout CCL2 in ApcMin/+ mice (ApcMin/+/CCL2−/−) inhibited tumor growth and immune infiltration in colon cancer [109]. CCL2 facilitated the accumulation of polymorphonuclear myeloid-derived suppressor cells (PMN-MDSCs) into the tumor microenvironment and increased MDSC-mediated inhibition of T cells in a STAT3-dependent manner, and blocking CCL2 using antibodies reduced tumor growth and MDSC infiltration in a murine model of colitis-associated CRC [109]. Targeting inhibition of CCL2 may provide therapeutic benefits for the prevention and interception of CRC.PAI-1 is encoded by SERPINE1 gene in humans. It is secreted mainly by hepatocytes and endothelial cells, and partly by adipose tissues [110]. Clinically, PAI-1 expression in various tumors is higher than that in normal tissues [111]. The plasma PAI-1 level was increased in CRC patients, but it was not correlated with the risk of colonic carcinogenesis [112]. Recently, Gerard reported [113] that PAI-1 aggravated mucosal damage through PAI1–tPA axis and activation of transforming growth factor β (TGF-β) in human and murine colitis. Knockout PAI-1 or treatment with PAI-1 inhibitor reduced inflammation and mucosal damage in DSS- and Citrobacter-induced colitis [114]. PAI-1 is an important inducer of the inflammatory reaction in colonic epithelial cells.Low-grade chronic inflammation is a main feature of obesity, it mediates most of the obesity-related complications [115]. Inflammation also plays an important role in the tumor microenvironment of CRC to activate the signaling of proliferation, migration, and metastasis [116,117]. Therefore, obesity triggered chronic subclinical inflammation is a bridge linking obesity to colorectal carcinogenesis. We summarize the mechanisms by which obesity-induced chronic inflammation leads to the carcinogenesis of CRC (Figure 5).Gut microbiota has become increasingly important for health with the launch of the National Microbiome Initiative and Human Microbiome Project in America in recent years [118,119]. In humans, about 1.5 kg microbes reside in the gut and make up half of the fecal matter biomass [120]. Increasing evidence indicated that gut microbiota is considered as a potential factor in the pathogenesis of obesity and associated metabolic disorders, even cancers [121,122,123]. Understanding the role of gut microbiome in obese and CRC individuals will provide potential molecular insights and therapeutic targets to prevent or treat both diseases.Germ-free animals are critical for studying the effect of microbes on host physiological and pathological processes. In different high fat and carbohydrate diets induced obesity models, germ-free animals have more food intake but gain less body weight compared with the conventional controls [122]. In carcinogen-induced and spontaneous colon cancer models, germ-free animals also show inhibited tumorigenesis in most cases [124]. Vannucci et al. reported that germ-free rats exhibited reduced tumor formation and enhanced anti-cancer immune response in AOM-induced CRC compared with conventional conditions [125]. T-cell receptor β chain and p53 double-knockout (TCRβ−/−, p53−/−) mice can spontaneously form colorectal tumors. The rate of tumor formation is about 70% in conventional mice. Whereas, there is almost no tumor in the germ-free mice [126]. Tomkovich et al. found that germ-free ApcMin/+ and IL10−/− mice had less colorectal tumors compared to specific-pathogen-free and gnotobiotic controls, and polyketide synthase (pks)+ Escherichia coli promoted carcinogenesis mediated by colibactin [127].Lipopolysaccharide (LPS) is an endotoxin produced by gram-negative bacteria in the gut and is associated with low-grade chronic inflammation [128]. Circulating LPS was elevated in high-fat diet (HFD) induced obesity due to a gut microbiome remodeling [129,130]. We recently found HFD increased the abundance of LPS-producing pathogens Desulfovibrio in mice [129]. Bacteria and endotoxins are prevented by intestinal mucosal barrier [131]. Increased intestinal permeability and systemic endotoxemia aggravated colitis and associated CRC. Bacteria secreted LPS directly exacerbates extracellular matrix adhesion and invasion in SW480, SW620, and CACO2 cells through activation of the urokinase plasminogen activator (u-PA) system in a TLR-4/NF-κB dependent manner [132]. Wenting et al. found that LPS increased the migration and invasion of colorectal cancer cells in vivo and in vitro by promoting epithelial-mesenchymal transition (EMT) and activation of SDF-1α/CXCR4/ NF-κB axis [133]. LPS participates in the enhancement of CRC malignant behaviors, and it may serve as a biomarker for CRC metastasis.Gut microbiota can produce some beneficial metabolites, such as short-chained fatty acids (SCFAs). SCFAs are key mediators linking diet and gut microbiota to prevent obesity and related metabolic disorders [134,135]. SCFAs are the major source of energy for colonocytes, and important for gastrointestinal health to maintain intestinal barrier function [136]. SCFAs also play a beneficial role in CRC clinically [137]. Mechanically, SCFAs inhibited cell growth and differentiation, promoted cell-cycle arrest and apoptosis, and regulated histone acetylation to protect against CRC [138]. Given the potential benefits of SCFAs, they are also considered as useful probiotics to prevent CRC.Akkermansia muciniphila, a genus of the phylum Verrucomicrobia, is a probiotic for preventing both obesity and CRC [139,140,141,142]. The Patrice group found Akkermansia protected against HFD-induced obesity through increasing intestinal endocannabinoids that reduced inflammation and enhanced gut barrier function [143]. Further, they found Amuc_1100, a specific membrane protein purified from Akkermansia, improved metabolic syndrome in obese and diabetic mice through TLR2 signaling [144]. In obese human volunteers, Akkermansia administration improved insulin sensitivity and inflammation, mildly reduced body weight, compared to placebo [145]. Akkermansia is also a crucial player in gastrointestinal disorders. Treatment with Akkermansia inhibited DSS-induced colitis in mice [146]. Amuc_1100 and pasteurized Akkermansia blunted colitis and associated CRC tumorigenesis through regulation of macrophages and CD8+ cytotoxic T lymphocytes in mouse colon [147]. Lactobacillus casei is a genus of Lactobacillus. Oral administration of Lactobacillus casei enhanced CD8+ T cell infiltration and inhibited colon carcinoma growth in tumor-bearing mice [148]. These data indicate that potential probiotic bacteria and beneficial metabolites are promising therapeutic agents for treating obesity and CRC.Bile acids (BAs), amphipathic molecules, mainly mediate intestinal dietary fat absorption. The primary BAs are synthesized from cholesterol in the liver and secreted into the intestine where pancreatic lipase is activated to form micelles and promotes nutrient absorption [149]. BAs also serve as signaling molecules to regulate farnesoid X receptor (FXR) and G protein-coupled receptor (GPCR) signaling, thereby maintaining energy and metabolic homeostasis [150]. BAs play key roles in lipid metabolism. The synthesis of BAs is associated with circulating triglyceride levels in patients with hyperlipoidemia [151]. Cholestyramine is a BA sequestrant commonly used for reducing high serum cholesterol levels in patients [152]. In the obesity models, total BAs were slightly increased, while conjugated BAs and deoxycholic acid (DCA) were dramatically elevated in plasma and liver [153,154]. The level of total BAs is correlated with BMI in obese patients [149]. BAs, especially secondary BAs, are potent carcinogens or promoters for CRC. Numerous studies reported that BAs are strong inducers for CRC tumorigenesis by damaging colonic epithelium, stimulating inflammatory reactions [155], inducing reactive oxygen species (ROS) production [156], promoting genomic instability, and resisting apoptosis [157]. Targeting BAs might be an effective strategy for the prevention and treatment of CRC.FXR is a bile acid receptor (BAR), encoded by the NR1H4 gene and highly expressed in the liver and intestine tissues [158]. FXR is a double-edged sword in obesity. Evans et al. reported activation of intestinal FXR by fexaramine inhibited obesity and increased adipose tissue browning through fibroblast growth factor 15 (FGF15) signaling [159,160]. On the other hand, Frank et al. found Glycine-β-muricholic acid (Gly-MCA), an intestine-specific FXR inhibitor, reduced obesity and associated metabolic dysfunction through inhibition of ceramide metabolism [161]. Additionally, FXR is a therapeutic target to protect against colorectal tumorigenesis. FXR inhibited colonic tumor growth in vivo. Knockout FXR in the ApcMin/+ mice promoted tumor progression and accelerated mortality through activation of Wnt/β-catenin signaling pathway [158,162]. T-β-MCA, a known FXR antagonist, was reported to promote CRC progression in HFD-induced APCmin/+ mice through damaging DNA and increasing proliferation in leucine-rich repeat-containing G protein-coupled receptor 5 positive (LGR5+) cancer stem cells [163]. Given the anti-tumor activity, intestinal FXR has promising therapeutic value in treating CRC.Both obesity and CRC are global health burdens currently. Epidemiologic data indicate a positive correlation between obesity and CRC. Obesity plays a direct and independent role in colorectal carcinogenesis. In the present review, we described the epidemiology of obesity and CRC respectively, and then summarized the potential underlying mechanisms linking obesity to CRC in different aspects, including nutriology, adipokines and hormones, inflammation, gut microbiota, and bile acids as shown in Figure 6. In nurtiology, adipocytes in tumor microenvironment are an energy source for CRC growth. In adipose tissue secreted adipokines and inflammation, elevated levels of insulin, IGFs, leptin, and inflammatory cytokines (e.g., IL-6, TNF-α, CCL2, and PAI-1), and decreased levels of adiponectin in obese, which alone or together contribute to the formation and development of CRC. Interestingly, circulating estrogen level is increased in obese individuals. It is known that estrogen contributes to obesity related breast cancer, but the role of estrogen in obesity-associated CRC is controversial. Cohort studies showed that BMI affects males stronger than females in the carcinogenesis of CRC, indicating estrogen may have a protective effect in CRC. In gut microbiota, obesity induced gut microbiota dysbiosis increases harmful microbiota and metabolites (LPS) and decreases beneficial microbiota (Akkermansia) and metabolites (SCFAs), which might lead to CRC tumorigenesis. In bile acids, bile acids promote CRC progression, especially DCA and T-β-MCA which are increased in obesity. The carcinogenesis of CRC is promoted by the bile acid-dependent inhibition of FXR, which is a target for anti-CRC. Therefore, obesity induces complex biological activities to promote CRC tumorigenesis.Besides obesity, epidemiologic evidence showed dietary and lifestyle factors include red/processed meat diet, low-fiber and high-fat diet, alcohol drinking, smoking, sedentary, and low physical activity are important environmental factors for CRC risk [164]. Genetic risk factors include familial adenomatous polyposis (FAP), and certain genetic mutations [165] (e.g., mutL homolog 1 (MLH1), adenomatous polyposis coli (APC), K-Ras (KRAS), and tumor protein p53 (TP53) genes). Environmental and genetic factors commonly contribute to CRC development. Reducing weight, improving diet, decreasing alcohol intake and smoking, and in addition to reducing sedentary time and increasing physical activity are likely to improve CRC incidence and mortality.In summary, we mainly focus on the role of obesity in CRC. The potential underlying biological mechanisms linking obesity to CRC are warranted, although great strides have been made to understand the biological mechanisms in obesity and the pathogenesis of CRC, respectively. Obesity induces insulin, IGFs, leptin, IL-6, TNF-α, CCL2, and PAI-1, reduces adiponectin, and disturbs gut microbiota and bile acid homeostasis. These altered factors promote CRC carcinogenesis mediated by downstream signaling pathways. Our increased understanding of the link between obesity risk factors and CRC carcinogenic processes will help to uncover more promising therapeutic targets and approaches for obesity-related CRC treatment in the future.Conceptualization and supervision: P.X.; writing—original draft preparation: P.Y. and Y.X.; figures: P.X., Y.X.; writing—review and editing: P.Y., Y.X., Z.H., and P.X. All authors have read and agreed to the published version of the manuscript.This research received no external funding.The authors declare no conflict of interest.The prevalence of overweight and obesity in women and men. The global prevalence of overweight (A) and obesity (B) in women and men from 1975 to 2016 (left), and the value in 1975 and 2016, and the prediction in 2035 (right). The prevalence of overweight (C) and obesity (D) for women and men in 1975, 1995, 2015, and the prediction in 2035 in China and the United States. The predicted values were boxed with the dashed line. Data are from the WHO website.The incident cases and deaths of colorectal cancer (CRC) from 1990 to 2017 in the world, the USA, and China. Data are from Global Burden of Disease Study 2017 (GBD 2017).The crosstalk between CRC cells and adipocytes in nutriology.Obesity secreted adipokines and hormones contributing to pathogenesis of CRC.Schematic mechanisms of carcinogenesis of CRC induced by obesity-elicited chronic inflammation.A schematic model of mechanistic insights linking obesity with CRC carcinogenesis. Red arrow indicates promotion, green arrow indicates protection.

Med-MDPI/cancers/cancers-12-06-01409.txt

@@ -0,0 +1 @@
+In the past two decades, an extensive rollout of colorectal cancer (CRC) screening programmes has been initiated in European countries with a large heterogeneity of screening offers. Using data from a population-based cross-sectional survey conducted between 2013 and 2016 in all European Union countries, we analysed the utilisation of faecal tests and colonoscopy among people aged 50–74 years and the factors associated with uptake by type of screening offer. We observed the highest utilisation of either test for countries with fully rolled out organised programmes with faecal tests (ranging from 29.7% in Croatia to 66.7% in the UK) and countries offering both faecal tests and colonoscopy (from 22.7% in Greece to 70.9% in Germany). Utilisation was very low for countries with no programme (from 6.3% in Romania to 30.5% in Norway). Younger age (50–54 years), longer time since last consultation with a doctor and a lifestyle score associated with increased CRC risk were significantly associated with lower test use, a pattern observed across all types of screening offers. Our results suggest that more countries should implement organised programmes with faecal immunochemical tests, in combination with alternative endoscopy offers where resources allow. Furthermore, there is a large potential for increasing screening use in Europe by better reaching the younger eligible individuals, those who have not been to the doctor recently and those at increased risk for CRC.Colorectal cancer (CRC) was estimated to be the third most common cancer and the second leading cause of cancer death worldwide in 2018 [1]. With age-standardised incidence and mortality rates of 45.9 and 18.5 (per 100,000), respectively, CRC poses a particularly high burden for the European Union (EU) countries [2]. However, due to the slow development of the disease and the existence of effective screening strategies, a large proportion of CRC cases and related deaths could be either prevented or detected early and successfully treated [3,4,5]. Specifically, faecal tests [6,7,8] and lower gastrointestinal endoscopy (colonoscopy and flexible sigmoidoscopy) [4] have been shown to considerably reduce CRC incidence and mortality. Faecal tests are non-invasive tests that detect occult blood in the stool by targeting either haem (guaiac-based faecal occult blood test, gFOBT) or human haemoglobin (faecal immunochemical test, FIT); lower gastrointestinal endoscopy consists in the examination of the entire large bowel (colonoscopy) or its distal part (flexible sigmoidoscopy) by an endoscope that is also able to remove precancerous lesions. Other tests, namely stool DNA tests, have been recommended as alternative methods by several US expert groups and societies [9,10,11]; however, no European country has considered them in their screening programmes to date [12,13].Already in 2003, recognising the effectiveness of CRC screening, the EU Council called upon its Member States to implement organised CRC screening with faecal tests targeting the population aged 50–74 years [14]. Some EU countries have meanwhile followed the recommendations and implemented either regional or nationwide organised programmes, while others have been offering CRC screening mainly in an opportunistic manner (on an ad hoc basis, essentially faecal tests and/or colonoscopy as primary screening modalities) or no CRC screening [12]. These differences in screening offer are expected to be reflected in different utilisation rates and patterns across different demographic, socioeconomic and at-risk groups in the EU countries.Between 2013 and 2016, socioeconomic and health-related factors, including the use of faecal tests and colonoscopy, were ascertained in all EU Member States, Iceland, Norway and the UK through the second wave of the European Health Interview Survey (EHIS) [15]. We first reviewed relevant characteristics of CRC screening offers in the EHIS-participating countries and subsequently, using the EHIS, analysed the use of faecal tests and colonoscopy, as well as factors associated with uptake by type of screening offer.A summary of relevant characteristics of CRC screening programmes in the EHIS-participating countries is presented in Table 1. The timing of enrolment, the type of programmes, the screening tests used and the target age groups vary considerably across countries. By the time the EHIS was conducted, four countries had fully rolled out their organised programmes with faecal tests (Croatia, France, Slovenia and the UK) and eleven were in the process of rolling out their organised programmes or only offered screening in certain regions (Belgium, the Czech Republic, Denmark, Finland, Ireland, Italy, Malta, Lithuania, the Netherlands, Spain and Sweden). Seven countries also offered faecal tests but predominantly in an opportunistic manner (Austria, Germany, Latvia, Luxembourg, Lithuania, Portugal and Slovakia). Furthermore, colonoscopy was offered as a primary screening modality in eight countries (Austria, the Czech Republic, Germany, Greece, Iceland, Luxembourg, Portugal and Slovakia). The remaining countries did not have any screening programme in place or small-scale pilot programmes only.Among the four countries where organised programmes had been rolled out nationally, three were found to have faecal test utilisation rates higher than 50% (UK: 59%, Slovenia: 56%, France: 51%); the exception was Croatia with only 22% (Figure 1). When looking at countries where organised programmes were being rolled out at the time of the survey, large variations in faecal test use were observed, ranging from 10% in the Netherlands to 42% in the Czech Republic. As for countries where faecal tests were offered mainly through opportunistic approaches, the lowest utilisation rate was observed for Greece (11%) and the highest rates for Germany (51%) and Austria (49%), where utilisation levels were close to those reported for the UK, Slovenia and France. Faecal test use was very low (below 15%) for all countries/age groups where no programme was in place. A proportion of older adults, no longer targeted by screening programmes, ranging from 8% in Spain to 30% in Italy, reported to still be up-to-date with faecal tests.As far as colonoscopy use is concerned, utilisation levels below 32% were observed for all countries where colonoscopy was not offered as an alternative primary screening modality and below 20% for the majority of countries where no screening programme was available (Figure 2). Among countries offering colonoscopy as an alternative primary screening test, utilisation levels were highest in Austria (52%), Germany (51%) and Luxembourg (49%) and lowest in Greece (15%) and Slovakia (15%). When looking at the utilisation of either faecal tests or colonoscopy, we observed the highest utilisation rates for countries where faecal tests were offered within an organised programme fully rolled out nationally and for countries with offers of both faecal tests and colonoscopy (Figure 3). Among the former, the UK was found to have the highest utilisation rates (67%) and Croatia the lowest (30%); among the latter, the highest utilisation rates were observed for Germany (71%) and the lowest for Greece (23%). Utilisation was very low (< 31%) for all countries with no programme.From all the potential predictors studied, age, health care use and the lifestyle score were found to be the strongest predictors of test use (Tables S1–S3). Specifically, when compared to those aged 60–64 years, individuals aged 50–54 years were much less likely to have undergone either faecal tests or colonoscopy (ORs ranging from 0.57 (95% CI: 0.50 to 0.64) among countries offering colonoscopy as an alternative primary screening modality to 0.77 (95% CI: 0.64 to 0.92) among countries with no programme in place). Furthermore, those who reported not having had a consultation with a physician within the previous 12 months were overall 40%–60% less likely to have undergone any test (ORs ranging from 0.44 (95% CI: 0.37 to 0.53) to 0.62 (95% CI: 0.53 to 0.73)), a pattern observed across all types of screening offers. A similar pattern was observed for people with lower lifestyle scores (especially those with scores 0 or 1) who had 22%–36% lower likelihood of having undergone either test (ORs ranging from 0.64 (95% CI: 0.48 to 0.86) to 0.78 (95% CI: 0.67 to 0.92)).Besides age, health care use and the lifestyle score, self-perceived health was found to strongly predict colonoscopy use. In particular, when compared to people who perceived their health as good or very good, those who reported a less than good health status were found to be more likely to have undergone colonoscopy within the previous 10 years. This association was especially pronounced in countries with national coverage of organised programmes (OR: 1.58, 95% CI: 1.34 to 1.86) and in countries with no programme in place (OR: 1.70, 95% CI: 1.55 to 1.87). This study provides an overview on the use of faecal tests and colonoscopy among the general average-risk population aged 50–74 years in the EU countries, Iceland, Norway and the UK by the type of CRC screening offer. Large differences in utilisation were observed between the different countries, with only five (Germany, Austria, the UK, Slovenia and France) reaching proportions of the eligible population up-to-date with either test higher than 60%. Overall, countries with nationwide coverage of organised programmes with faecal tests and countries which provided both faecal tests and colonoscopy as alternative screening options were found to have comparable, and the highest, proportions of the population up-to-date with either test. Utilisation was much lower or almost nonexistent in countries where no screening programme was in place. Of the various potential predictors investigated, being in the youngest age range eligible for screening, the absence of a recent consultation with a physician and being at higher risk for CRC based on lifestyle characteristics were strongly associated with the decreased use of faecal tests or colonoscopy, irrespective of the type of CRC screening offer.The EHIS constitutes the reference source of evidence to support health-related policy making in the EU region. A major strength is that all countries had to follow detailed rules and recommendations for data collection in order to ensure high comparability levels and to use nationally representative probability samples [60]. Furthermore, to our knowledge, no international comparison on the use of faecal tests and colonoscopy has been carried out across all EU countries by the type of CRC screening and using nationwide population-based data [61]. The utilisation rates reported in this study will be of great importance when it comes to analysing and comparing the potential effects of screening on CRC incidence and mortality in the EU countries. These data may also be used in future studies as model input to project CRC incidence and mortality and to potentially improve screening offers.The study also has some limitations that should be considered when interpreting its findings. First, the data were mostly collected from self-reports which may have led to reporting and recall biases and therefore the over- or underreporting of test use and the explanatory factors included in the analysis [62]. Second, given that the countries which relied on self-administered data collection had overall higher non-response rates, selection bias may have been introduced. Third, the data did not allow for discrimination between screening and diagnostic colonoscopies. Fourth, the cut-offs used to dichotomise the individual items of the lifestyle score, especially smoking and alcohol consumption, were slightly different from the ones outlined in the recommendations, and no information on diet was available [63]. This may have led to the attribution of a less appropriate score to a small proportion of respondents. Lastly, some countries offer other screening modalities that we were not able to analyse. Specifically, in England and Italy (Piedmont and Veneto regions), flexible sigmoidoscopy is also offered as an alternative screening test [38,64]. Therefore, especially for the UK, the proportion of individuals up-to-date with CRC screening is likely to be higher than the one based on faecal tests and colonoscopy only. Comparisons across European countries were first performed in the context of the Survey of Health, Aging and Retirement in Europe (2004/2005) when most countries had not yet implemented CRC screening programmes [65]. Apart from Austria and Germany, where faecal tests and colonoscopy had already been offered in an opportunistic manner, utilisation rates were very low for all countries (below 25%). Similar utilisation rates were now found for the group of countries which still do not have CRC screening programmes.More recently, the status [13], coverage [16] and performance [41] of CRC screening have been disclosed for organised programmes implemented in some EU countries in the context of the second European screening report (data collection from 2011 to 2014) [20]. For countries with fully implemented organised programmes nationally, whose estimates of faecal test use can be compared to the ones from this report, similar results were observed for Slovenia and the UK (utilisation rates about 50%–60%). For France, we found considerably higher utilisation rates (51% compared to < 30% reported by Senore et al.) [41] which may be in part explained by the discrepancy between the more recent data collection for EHIS (2013–2016) and the screening report (2012). The much lower utilisation rates observed for Croatia are in line with those described in national reports and have been partly attributed to a lack of CRC screening awareness in the population [21,66]. Other issues that have been raised relate to the absence of colonoscopy facilities on the Croatian islands, which make it difficult to follow up on a positive faecal test [21].As for countries where organised programmes were still being rolled out, or where not all regions were covered, our national estimates reflect such gradual implementations. For instance, in Spain and Sweden CRC screening had only been implemented in some regions, and in the Netherlands the EHIS data collection took place in the year the programme was launched. Therefore, low utilisation rates would be expected. Nonetheless, the high participation rates that have been reported for these countries among those invited to screening [41] suggest that similar utilisation rates to the ones observed for France, Slovenia and the UK are likely to be achieved once the programmes are fully implemented nationally. On the other hand, the comparatively high utilisation of faecal tests observed for the Czech Republic may be explained to a large extent by the existence of a parallel opportunistic programme already adopted in 2000. For other countries with higher utilisation rates (e.g., Denmark, Ireland and Italy), the EHIS was implemented when the nationwide screening programmes had already been rolled out for at least one year.Large variations in test use were also observed across countries providing CRC screening mainly in an opportunistic manner and seem to be highly related to differences in health care systems and available resources. Specifically, a lack of prioritisation of cancer prevention and low involvement of physicians in early detection may help explain the low proportions of the population up-to-date with either test observed for countries such as Greece and Slovakia [67,68,69,70]. In countries with much higher proportions, namely Austria and Germany, CRC screening has been in place for decades and awareness of CRC screening is high [12,17,71]. The high proportions observed for the latter additionally suggest that, for countries where the necessary resources are available, offering colonoscopy as an alternative option may contribute towards achieving high CRC screening utilisation rates. This is further supported by a recent study from Germany which has found an increase in CRC screening uptake rates after the introduction of screening colonoscopy as an alternative to faecal tests [32].Looking at the utilisation among countries where no programme or only a small-scale programme was in place, the very low levels observed for all countries may mainly reflect test use for other purposes. In particular, a large proportion of the population up-to-date with colonoscopy is likely to have undergone colonoscopy for the clearance of symptoms rather than for screening purposes. This may also explain, to a large extent, the considerably higher odds of having undergone colonoscopy among people who reported a less than good health status in comparison to those who perceived their health as good or very good. Furthermore, the lower likelihood of being up-to-date with faecal tests or colonoscopy for all types of screening offers among the youngest age groups eligible for screening, and those at the highest risk for CRC, suggests a large potential for reducing the high CRC burden in Europe by efforts to better reach these people by screening offers. General practitioners could play a major role in this context, as supported by the strong association that was consistently seen between having seen a doctor in the last year and the use of CRC screening offers even within organised programmes, in which the eligible population is informed with an invitation letter and the faecal test is sent along. Given the very recent implementation of screening in most countries, its effects on CRC mortality are expected to be fully disclosed only in the next decades. However, the reported variations in the timing of implementation, screening offers and uptake across countries suggest that large differences in the progress towards CRC mortality reduction will be observed across EU countries.This study used data from the second wave of the EHIS. EHIS is a cross-sectional survey aimed at providing harmonised and highly comparable data across EU countries to support health policies and address health inequalities and social exclusion [72]. The survey was conducted in 30 countries (all EU countries, Iceland, Norway and the UK) under the Commission Regulation (EU) No 141/2013 between 2013 and 2016 and targeted non-institutionalised individuals aged 15 years and older residing in these countries at the time of data collection. In order to ensure national representativeness and comparability of the data, each country used probability sampling techniques and followed methodological recommendations set out by Eurostat [60]. A median response rate of about 60% was achieved across countries, ranging from less than 50% in Austria, Denmark, Finland, Germany and Luxembourg to over 90% in Cyprus and Portugal. Potential non-response bias was addressed by calculating and attributing to each respondent a weighting factor that ensures each country is considered in proportion to its demographic distribution. Further details can be found in the Eurostat quality report [72]. For this analysis, only individuals aged 50–74 years, who are commonly considered eligible for CRC screening, were included [12,13]. Moreover, only the participants whose survey responses were provided by themselves were included; proxy interviews were excluded to rule out potential, and very likely, inaccurate responses. In total, 128,496 individuals aged 50–74 years participated in the study. Among them, 3121 were drawn from proxy interviews and another 3479, 2419 and 4210 did not provide data on the utilisation of faecal tests, colonoscopy and either test, respectively. Hence, 121,896 respondents were included for analyses of faecal test use, 122,956 for colonoscopy use and 121,165 for either faecal test or colonoscopy use (Figure 4).The data collection period varied from 3 to 21 months across the different countries. Various data collection methods were adopted, consisting of face-to-face interviews, telephone interviews, post and a combination of different methods. The survey was translated from English into the local languages of each individual country and pre- or pilot-tested in most countries [72]. The outcome measures of faecal test use (either gFOBT or FIT) within the previous 2 years, colonoscopy within the previous 10 years and the use of either test in the respective time frames were ascertained by enquiring the respondents about the last time these tests were undertaken. These time frames are the most widely used screening intervals for the average-risk population in the EHIS-participating countries [12,13]. The explanatory variables tested as potential determinants of test use encompassed demographic and socioeconomic factors (sex, age, location of residence, marital status, education and household income), health care use (last time of a consultation with a general practitioner and a medical or surgical specialist) and health-related factors (self-perceived health and a healthy lifestyle score). A healthy lifestyle score was created based on the established evidence on CRC risk and protective factors, notably smoking, alcohol consumption, physical activity and body mass index (BMI) [63] (Table S4). It was adapted from a score created by Carr et al. that was associated with lower risk for all stages of CRC [73,74]. Specifically, respondents were attributed one point for the following low-CRC-risk behaviours: non-smoking or occasional smoking, consumption of less than two alcoholic drinks per day, being physically active (at least 150 min per week spent on doing sports, fitness or recreational physical activities, in line with World Health Organization Global Recommendations on Physical Activity for health [75]) and having a BMI below 25 kg/m2, i.e., not being overweight or obese [63]. Data on physical activity were not available for Belgium and on alcohol consumption for France and Italy. Therefore, a score ranging from 0 to 3, instead of 0 to 4, was attributed to each respondent from these three countries. For the Netherlands, data on both alcohol consumption and physical activity were not available, thus it was not considered in the multivariate analyses for which the lifestyle score was used as a covariate.A review of the relevant characteristics of CRC screening programmes in the 30 EHIS-participating countries was carried out. Information was retrieved from articles on the second European screening report [20], websites from national governments and cancer registries, as well as from a literature search in PubMed using the keywords: colorectal cancer screening, colonoscopy, guaiac faecal occult blood test and faecal immunochemical test, alongside the term “Europe” and each country name [12,13,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59]. The characteristics that we reviewed and summarised served as a basis for the analyses of EHIS data.For the analyses of faecal test use, the following five categories were created: (A) Nationwide fully rolled out organised programmes with faecal tests; (B) Organised programmes with faecal tests in partial rollout or with regional coverage only at the time of EHIS data collection; (C) Opportunistic programmes with faecal tests; (D) No programme with faecal tests or a small-scale pilot programme only; (E) Other, that is, age groups not targeted by screening programmes but to whom faecal tests may have been offered within the preceding two years when these individuals were eligible for screening.To analyse colonoscopy use or the use of either faecal tests or colonoscopy, the abovementioned groups (A) and (B) were kept and two additional groups were created, namely: (C) Colonoscopy offered as an alternative primary screening modality; (D) No programme, small-scale organised programme, or opportunistic programme with faecal tests as the first-line method only. Unlike for faecal test use, an additional category (E) with older age groups not targeted by the screening programmes was not created, as these older birth cohorts had within the past 10 years been among the age groups eligible for colonoscopy (either as a primary screening modality or as a follow-up test). These were therefore included together with the younger age groups targeted by screening programmes in their respective category. Table S5 provides an overview of the countries and age groups included in the different categories created to analyse the utilisation of faecal tests, colonoscopy and either test. These categories reflect the status of CRC screening implementation at the time the EHIS was carried out.Prevalence estimates of faecal test use within the previous 2 years, colonoscopy use within the previous 10 years and the use of either of the two tests were determined for each country according to the abovementioned categories. In order to explore the potential determinants of test use, multivariate logistic regression models were employed and odds ratios (ORs) and confidence intervals (CIs) were obtained. The log (OR) and their standard errors were subsequently computed for each country and subgroup meta-analyses were performed to estimate ORs and CIs by category/type of screening offer using the metagen function in RStudio Version 1.2.1335 (RStudio, Inc., Boston, MA, USA). The estimates obtained from the random-effects subgroup meta-analyses were extracted and summarised.For both the prevalence estimates and odds ratios, individual weights were applied, accounting for the units’ probability of selection, non-response, over- or under-representation of certain population groups and calibrated with country-specific distributions of the population with regards to demographic characteristics [60]. Variances were calculated using the Taylor series linearisation method that takes into account the complexity of the survey design.Except for the meta-analyses, all other analyses were conducted using SAS 9.4 (SAS Institute Inc., Cary, NC, USA).The EHIS was carried out according to the Regulation (EC) No 1338/2008 of the European Parliament and of the Council of 16th December 2008. The data collected by each country have been specified in the Commission Regulation (EU) No 141/2013. Ethics approval was obtained on a national level by the institutions responsible for the survey implementation. Further information can be found in the quality report released by Eurostat: https://ec.europa.eu/eurostat/documents/7870049/8920155/KS-FT-18-003-EN-N.pdf/eb85522d-bd6d-460d-b830-4b2b49ac9b03. The proportion of the population up-to-date with faecal tests or colonoscopy had remained low for the majority of the EU countries in 2013–2016. Nevertheless, in line with previous evidence, our study suggests that, when completely rolled out, organised programmes with faecal tests have the potential to reach a high proportion of the target population. In programmes with faecal tests, FITs rather than gFOBT, should be offered, given their superior diagnostic performance [76,77]. Moreover, offering the choice to undergo either faecal tests or colonoscopy was shown to attain similar proportions of the population up-to-date with CRC screening even if the offer was provided in an opportunistic manner only. Therefore, in the EU context, and based on the resources available in each country, it seems plausible that organised programmes offering faecal tests and colonoscopy as alternative test options might lead to even higher CRC screening uptake levels than organised programmes solely offering faecal tests. It also seems plausible that even higher uptake rates could be achieved by complementary offers of faecal tests and flexible sigmoidoscopy as an almost equally effective, but less invasive, endoscopy alternative [4,78]. Our results furthermore underline the important role physicians could have in motivating people to use available CRC screening offers, in particular the younger age groups eligible for screening and those at increased risk for CRC.The following are available online at https://www.mdpi.com/2072-6694/12/6/1409/s1, Table S1: Odds ratio estimates and 95% CIs from random-effects subgroup meta-analyses of the association between demographic, socioeconomic, health care use and health-related factors, and faecal test use within the previous 2 years by type of CRC screening offer, Table S2: Odds ratio estimates and 95% CIs from random-effects subgroup meta-analyses of the association between demographic, socioeconomic, health care use and health-related factors, and colonoscopy use within the previous 10 years by type of CRC screening offer, Table S3: Odds ratio estimates and 95% CIs from random-effects subgroup meta-analyses of the association between demographic, socioeconomic, health care use and health-related factors, and having undergone either faecal tests within the previous 2 years or colonoscopy within the previous 10 years by the type of CRC screening offer, Table S4: Healthy lifestyle score, Table S5: Categorisation of countries/age groups by the type of colorectal cancer screening offer.Conceptualization: R.C. and H.B.; Methodology: R.C. and H.B.; Software: R.C.; Formal Analysis: R.C.; Investigation: R.C.; Resources: H.B.; Data Curation: R.C.; Writing—Original Draft Preparation: R.C.; Writing—Review & Editing: R.C., F.G., T.H., M.H. and H.B.; Visualization: R.C., F.G., T.H., M.H. and H.B.; Supervision: H.B.; Project Administration: R.C. and H.B.; Funding Acquisition: H.B. All authors have read and agreed to the published version of the manuscript.This study was supported in part by a grant from the German Cancer Aid (Deutsche Krebshilfe, grant no. 70112095). We used data from the second wave of the EHIS, which we gratefully acknowledge. We are also thankful to Thomas Hielscher from the Division of Biostatistics at the German Cancer Research Center for the valuable statistical input provided at an early stage of this analysis.The authors declare no conflict of interest. The sponsors had no role in the design, execution, interpretation, or writing of the study.The results and conclusions in this manuscript are authors’ responsibility and do not represent the views of Eurostat, the European Commission or any national statistical authority.Data were granted by Eurostat. Information on how to access the data can be found at https://ec.europa.eu/eurostat/web/microdata/european-health-interview-survey.Prevalence estimates of faecal test use within the previous 2 years among the general population aged 50–74 years in all EU countries, Iceland, Norway and the UK, by type of screening offer (data: EHIS, 2013-2016).Prevalence estimates of colonoscopy use within the previous 10 years among the general population aged 50–74 years in all EU countries, Iceland, Norway and the UK, by type of screening offer (data: EHIS, 2013–2016). For Sweden, colonoscopy use was determined within the past 5 years.Prevalence estimates of faecal test use within the previous 2 years or colonoscopy use within the previous 10 years among the general population aged 50–74 years in all EU countries, Iceland, Norway and the UK, by type of screening offer (data: EHIS, 2013–2016). For Sweden, colonoscopy use was determined within the past 5 years.Flowchart of respondents included in, and excluded from, the analyses due to a lack of data on the outcome measures or proxy interviews.Characteristics of colorectal cancer screening programmes in the EU countries, Iceland, Norway and the UK a.Abbreviations: FIT, faecal immunochemical test; gFOBT, guaiac-based faecal occult blood test; NA, not applicable. a Information was ordered by (i) country; (ii) type of programme; (iii) year of programme initiation. b Proportion of invited individuals from the annual target population based on EUROSTAT data from 2013. For Belgium, Malta, the Netherlands and Portugal, the index year was 2014; for Slovenia, the index years were 2011–2012. Data were retrieved from Vale et al. [16]. c In Austria, gFOBT is undertaken within the freely available annual health check-up for persons aged 40+. d In Croatia, expenses are reimbursed by health insurance. e In the Czech Republic, invitations are sent to individuals up to 70 years of age only. f In Finland, CRC screening was introduced as a randomised, organised health services programme in 2004. By the end of 2012, 22% of the Finnish target population had been invited to a screening. A new programme was initiated in 2019 in nine municipalities (Jyväskylä, Muurame, Orivesi, Oulu, Paimio, Sauvo, Säkylä, Tampere, Ylitornio) and will be gradually expanded to include all 60–74-year olds. g In Iceland, an organised programme with FIT is foreseen. h Screening started in 1982 in Florence and in 2000–2004 in other regions. FIT replaced gFOBT in 1996. i Once only flexible sigmoidoscopy is offered in the Piedmont region and in a small area of the Veneto region. The non-responders are invited to undertake FIT. j All eligible individuals on the NHS lists are invited for sigmoidoscopy. k In Norway (counties of Østfold, Akershus and Buskerud), in 2012, about 140,000 individuals aged 50-74 years were invited to participate in a randomised controlled trial to undergo either sigmoidoscopy or FIT. l In Poland, colonoscopy screening was implemented in the context of randomised health services research and roll-out is ongoing. m Population aged 55–64 years divided by the number of years of the target age range. The figure was retrieved from Vale et al. [16]. n In Portugal, despite FIT being the primary screening test recommended by the Ministry of Health, colonoscopy has been found to be the most recommended method by physicians. In 2017, the Portuguese government legislated the implementation of a national organised programme; national extension of the current pilot programme is expected within the next few years. p In Spain, CRC screening is implemented on a regional level. The programme started in 2000 in Catalonia and has been extended to another eight regions. FIT replaced gFOBT in 2009–2010. q In August 2018, ministers agreed to further include the younger age groups (50–59 years) in the bowel screening programme, which is now being planned by the National Health Service and Public Health England.

Med-MDPI/cancers/cancers-12-06-01410.txt

@@ -0,0 +1 @@
+Cholesterol plays an important role in cellular homeostasis by maintaining the rigidity of cell membranes, providing a medium for signaling transduction, and being converted into other vital macromolecules, such as sterol hormones and bile acids. Epidemiological studies have shown the correlation between cholesterol content and cancer incidence worldwide. Accumulating evidence has shown the emerging roles of the dysregulation of cholesterol metabolism in cancer development. More specifically, recent reports have shown the distinct role of cholesterol in the suppression of immune cells, regulation of cell survival, and modulation of cancer stem cells in cancer. Here, we provide a comprehensive review of the epidemiological analysis, functional roles, and mechanistic action of cholesterol homeostasis in regard to its contribution to cancer development. Based on the existing data, cholesterol homeostasis is identified to be a new key player in cancer pathogenesis. Lastly, we also discuss the therapeutic implications of natural compounds and cholesterol-lowering drugs in cancer prevention and treatment. In conclusion, intervention in cholesterol metabolism may offer a new therapeutic avenue for cancer treatment.Cholesterol is gaining increasing attention in cancer research due to its targetable therapeutic implications in both the prevention and treatment of cancer. However, the role of cholesterol in tumorigenicity remains controversial [1]. Researchers have reported a distinctively contradictory role of cholesterol in cancer development, showing that the correlation of cholesterol in carcinogenicity can be cancer-type specific [2]. High cholesterol or hypercholesteremia has positive correlations in breast and prostate cancers [3,4], while some prospective cohort studies show an inverse association [5,6]. Therefore, this review aims to discuss the current understanding of cholesterol homeostasis, to summarize the key findings of recent pre-clinical and clinical studies investigating cholesterol metabolism in cancer, and to provide up-to-date therapeutic implications of natural compounds and cholesterol-lowering drugs in cancer treatment.As a subtype of lipid, cholesterol exists in every type of mammalian cell, ranging from fundamental components of cell membranes by maintaining integrity and stability to precursors of different forms of vital sterol compounds, like vitamins and hormones. Akin to other important molecules, cellular cholesterol levels are tightly regulated through metabolic processes, namely de novo biosynthesis, intake, export, and esterification of excess free cholesterol [7].The de novo cholesterol biosynthesis, or the mevalonate pathway in some contexts (Figure 1a), consists of more than 20 enzymatic steps. It all starts from combining three acetyl-CoA molecules to form one 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA). Under the first rate-limiting catalytic enzyme, HMG-CoA reductase (HMGCR) converts HMG-CoA into mevalonate, which is then transferred to farnesyl pyrophosphate (FPP), squalene, and finally cholesterol through a series of enzymatic reactions. Noteworthy, FPP, apart from transferring into downstream sterols and all other nonsterol isoprenoids, is capable of converting into geranylgeranyl pyrophosphate (GGPP), which are both important effectors in protein prenylation [8,9].Meanwhile, the mevalonate pathway is supervised by a master transcriptional regulatory protein called sterol-regulatory element binding protein 2 (SREBP2) in a negative feedback loop [10]. SREBP2 is synthesized in a premature format in the endoplasmic reticulum (ER) membrane [11]. The maturation of SREBP2 requires a two-step proteolytic cascade that occurs in the Golgi apparatus, where site 1 protease (S1P) and S2P act consecutively to release the N-terminal fragment of SREBP2 [11,12]. This N-terminal fragment, or nuclear SREBP2 (nSREBP2), enters the nucleus and binds to the genes containing sterol regulatory elements at the promoter region, subsequently enhancing their transcriptional levels [13]. Those genes are therefore involved in cholesterol biosynthesis, such as HMGCR, low density lipoprotein receptor (LDLR), and squalene synthase [13]. The activation of SREBP2 occurs only when the intracellular cholesterol level is low so that the SREBP cleavage-activating protein (SCAP), another ER-anchored protein, is freed from cholesterol and insulin-induced gene protein 1 (INSIG1) [14,15]. The detachment of INSIG1 induces a closed conformational change of SCAP and allows its binding to COPII-coated vesicles [16]. The SCAP-SREBP2 complex is transported to the Golgi apparatus together with the COPII vesicles for SREBP2 activation. However, when the intracellular cholesterol level is high enough to bind to SCAP, which sequentially recruits INSIG1, the INSIG1 attachment relaxes the conformation of SCAP, thus prohibiting the complex from binding to COPII vesicles [16]. The transcriptional activity of nSREBP2 can also be increased by the master regulator of anabolic reactions, mammalian target of rapamycin complex 1 (mTORC1), via inhibition of nuclear entry of lipin1, which downregulates nSREBP2 [17]. Apart from that, the mevalonate pathway, which is an energetically expensive metabolic process, can also be regulated through rate-limiting enzymes. HMGCR can be phosphorylated by 5′ adenosine monophosphate-activated protein kinase (AMPK) to abolish its activity when intracellular ATP levels are low [18]. Recently, squalene epoxidase (SQLE) has been considered as another rate-limiting enzyme in this pathway, in which it converts squalene into squalene epoxide [19]. The E3 ubiquitin ligase MARCH6 is recruited to degrade squalene epoxidase when excess cholesterol is present [19].When de novo biosynthesis remains the main source of intracellular cholesterol, most cells acquire cholesterol from low density lipoprotein (LDL) in the circulatory system via LDLR-mediated endocytosis [20]. Free cholesterol is then dissociated from LDL when lysosome is digested. Yet, proprotein-convertase-subtilisin-kexin type-9 (PCSK9) induces lysosomal degradation to LDLR [21]. The very-low-density lipoproteins, the precursors of LDL, are composed in liver, where the dietary cholesterol is transported for compartmentation. In contrast to LDLR-mediated endocytosis, enterocytes in the intestinal lumen absorb dietary cholesterol via Niemann–Pick type C1-like 1 protein (NPC1L1) through a clathrin-dependent pathway [22]. The upregulation of NPC1L1 contributes to cardiovascular diseases and symptomatic gallstone diseases [23].When cholesterol has served its intracellular purposes, the excess cholesterol is exported via ATP-binding cassette (ABC) subfamily A member 1 (ABCA1) or ABC subfamily G member 1 (ABCG1) to lipid-poor apolipoprotein A-I (ApoA-I) and generates high-density lipoproteins (HDLs) that are transported back to the liver [24,25,26,27]. The transcriptional level of ABCA1 is upregulated by nuclear liver X receptor (LXR) when the intracellular cholesterol level is high [27]. Surplus cholesterol can also be esterified by acyl-CoA:cholesteryl acyltransferase 1 (ACAT1) into cholesteryl esters (CEs), which are a less toxic form and can be stored as lipid droplets or for further processing into lipoproteins [28].The discovery of microRNAs (miRNAs), a class of non-coding RNAs, has added complexity into cholesterol homeostasis through regulation of different key components in the system [29]. miR-33a, an embedded intronic microRNAs, is located within SREBP2 gene [30]. In a low sterol condition, akin to SREBP2, miR-33a is transcribed up to 2- to 3-fold higher to regulate cholesterol export and HDL metabolism gene by targeting ABCA1 for post-transcriptional repression [30]. On the other hand, miR-223 controls cholesterol level by inhibiting the synthesis and enhancing the cholesterol efflux by elevating the expression of ABCA1 [31]. miRNA-122 is highly expressed in hepatocytes, accounting for 70% of all liver miRNA [32]. Inhibition of miRNA-122 substantially suppresses total plasma cholesterol [32]. However, the direct target of miRNA-122 is yet to be elucidated [32]. Apart from these miRNAs, miR-27a has been shown to specifically interact with HMGCR 3′ untranslated region to inversely regulate HMGCR expression by posttranslational inhibition followed by mRNA degradation [33]. Strikingly, the application of genome-wide association studies has allowed the discovery of more miRNAs in abnormal levels of cholesterol-lipoprotein circulation, such as LDLR and ABCA1 [34]. These mRNAs include miR-128-1, miR-148a, miR-130b, and miR-301b [34]. Taken all together, these findings have suggested the potential involvement of miRNAs in regulation of cholesterol metabolism, and they may contribute to an abnormal cholesterol level if left unregulated.As an essential macromolecule in metabolism, cholesterol has been suspected to play an important role in inducing cancer. Since the 1980s, such a relationship has been extensively monitored and examined in different clinical cohort studies. However, the results have indicated that the relationship between cholesterol and cancer is type- and stage-specific, both to the tumor originating site and the form of lipoproteins being examined. Cholesterol circulates in the body mainly in two different forms: LDL or HDL. Researchers have examined them separately or inclusively as total cholesterol (TC) to determine their tumorigenicity effect (Table 1).LDL cholesterol (LDL-C) level has been suggested to be a prognostic factor of breast cancer progression at diagnosis [49]. A prospective study on 244 women with operable breast cancer in Portugal showed that patients with LDL-C levels as high as 117 mg/dL had poor prognosis due to the higher proliferative rate, histological stage, and more advanced clinical stage [49]. Meanwhile, patients with LDL-C levels above 144 mg/dL can suffer from lympho-vascular invasion as well as lymph node metastasis [49]. Yet, in two meta-analyses involving over 1 million patients each, LDL-C showed no association with breast cancer risk irrelevant to menopause in women [50,51]. On the other side of the world, in a large population-based case-control study conducted in Shanghai, China, researchers showed the relationship of LDL-C in biliary tract cancers [43]. Patients contracting bile duct cancer had significantly higher LDL-C levels than control patients, while those patients who suffered from gallbladder cancer had lower LDL-C. However, there was no significant difference in LDL-C level in patients contracting carcinoma of the ampulla of Vater, a small region connecting the duodenum, bile duct, and pancreatic duct, compared with the control group [43]. Moreover, LDL-C was positively associated with liver metastases in colorectal cancer patients [52].In contrast, HDL cholesterol (HDL-C) may have a clearer effect on reducing the cancer risk. A prospective follow-up study of participants who were enrolled in the ATBC Cancer Prevention Study showed a strong inverse association between HDL-C and non-Hodgkin lymphoma (NHL). The researcher claimed that the risk of NHL was reduced by 15% for each 5 mg/dL increase in HDL-C level [39]. Participants with HDL-C levels above 55 mg/dL had over 60% lower risk of developing NHL [39]. A similar situation was observed in biliary tract cancers in the study mentioned above. Patients with a high HDL-C level (>40 mg/dL) had 11.6- and 16.8-fold lower risks of gallbladder and bile duct cancers than the patients with low HDL-C levels (<30 mg/dL) [43]. Additionally, a modest association of HDL-C with <50 mg/dL could increase the breast cancer risk among premenopausal women in a prospective study examining over 7000 patients [53].Taking both circulating forms into account and measuring them as the TSC level provides a complementary, yet more elusive, picture to examine the relationship of cholesterol and cancer risk in clinical studies. In an all-inclusive prospective study of nearly 1.2 million Korean participants, the investigators examined several common types of cancers, including stomach, liver, pancreas, lung, prostate, or colon cancers [41]. After adjusting for body mass index, alcohol consumption, fasting glucose, hypertension, smoking, and physical activity, inverse associations were observed between all-cancer incidence with total cholesterol in men (Hazard Ratio (HR), 0.84; 95% Confidence Interval (CI), 0.81 to 0.86) and in women (HR, 0.91; 95% CI, 0.87 to 0.95) [41]. A similar result was observed in another large prospective study consisting of seven cohorts from Norway, Austria, and Sweden, including nearly 600,000 participants [37]. Inverse associations were demonstrated between total cholesterol level and all cancer risk in men (HR, 0.94; 95% CI, 0.88 to 1.00) and in women (HR, 0.86; 95% CI, 0.79 to 0.93) [37]. In some individual cancer types, positive correlations could be established in both sexes, such as prostate, colon, pancreatic, and breast cancer [37]. Particularly in prostate cancer, men with higher cholesterol levels were at greater risk of developing a higher clinical stage of prostate cancer [36].Dysregulation of key molecules in cholesterol homeostasis or cholesterol itself has not only been associated with several well-known oncogenic pathways, but also related to inflammasome- and miRNAs-mediated cancer development (Figure 1b). By understanding the interplay between these parties, more effective drug interventions can be developed.In the mevalonate pathway, production of FPP and GGPP could induce onco-protein prenylation, which is involved in the activation of several oncoproteins, such as Ras GTPases [8,9,54]. Meanwhile, proprotein-convertase-subtilisin-kexin type-9 (PCSK9) induces lysosomal degradation to LDLR [21]. However, the overexpression of PCSK9 contributes to hypercholesterolemia and sequentially correlates with hepatocellular carcinoma development [55]. Hyperactivity of LXR, another key player in cholesterol homeostasis induced by its agonists, has been shown to exert an anti-proliferative effect in gastric cancer cells [56]. Yet, though CEs serve as a cholesterol reservoir, the accumulation of CEs or overexpression of ACAT1 have supported a pro-tumor role. In hepatocellular carcinoma, ACAT1 elevation is identified by proteomic and phospho-proteomic analyses [57]. In xenograft models of glioblastoma, ACAT1 ablation has reduced the tumor progression [58]. CP-113818, the ACAT1 inhibitor, suppresses the migration capacity of breast cancer cells [59]. Furthermore, inhibition of ACAT1 has also been shown to decrease prostate cancer progression [60].One of the most intensively studied oncogenes, TP53 gene mutation, arising from deletion or truncation, aggressively promotes tumor survival, invasion, migration, metastasis, and chemoresistance in many cancers [61]. With functional p53 protein, SREBP2 activity is suppressed due to upregulation of ABCA1, hence reducing the transcriptional levels of enzymes in the mevalonate pathway [62]. However, with respect to breast cancer, p53 disrupts the acinar morphogenesis, or tissue architecture, of breast cells, aided by the upregulated expression of the cholesterol biosynthesis pathway. By harvesting tumor-derived mutants of p53 in an organoid culture system, through Ingenuity Pathway Analysis and Gene Ontology Analysis, cholesterol biosynthesis was shown to be the most overrepresented regarding p53 downregulation [63]. A rescue experiment supplementing the essential intermediate metabolites in the mevalonate pathway could significantly inhibit the disordered phenotype caused by silenced p53 in breast cancer cells in 3D culture [63]. Moreover, TP53-mediated SREBP2 cholesterol synthesis can also enhance the prenylation of Rho GTPases, which, in turn, enhance the proliferation and self-renewal of breast cancer cells [54,63]. Conversely, simvastatin strikingly decreased cancer cell growth, increased cell death, reduced invasiveness, and mimicked the mutant p53 depletion in terms of morphological changes [63].Another example is the phosphoinositide 3-kinase (PI3K)/protein kinase B (Akt)/mTORC1/SREBP signaling axis, which induces overall cell growth. While the PI3K and Akt signaling pathway is responsible for the accumulation of mass, the inhibition of mTORC1 can attenuate Akt-dependent lipogenesis and eventually cause reductions in cell size in vitro and in vivo [64,65]. This finding was further elaborated. The abnormal activation of the PI3K/Akt signaling pathway maintains a high intracellular level of cholesterol via mTORC1 in inhibiting the ABCA1 efflux activity and via SREBP overexpression, resulting in LDLR-dependent cholesterol import [66]. Moreover, a high intracellular cholesterol level could further drive mTORC1 recruitment and activation in lysosomes via lysosomal transmembrane protein SLC38A9 [67]. Meanwhile, in prostate cancer, the loss-of-function phosphatase and tensin homolog (PTEN) can activate the PI3K/Akt pathway and lead to the accumulation of CEs as a result of excess intracellular cholesterol levels after upregulation of LDLR-mediated cholesterol influx [68]. A similar result was also seen in nonalcoholic fatty acid induced hepatocellular carcinoma. The overexpression of SQLE can suppress PTEN activity and subsequently induce the accumulation of CEs through Akt signaling [69]. Collectively, such an alteration is related to cell proliferation, tumor formation, and cancer aggressiveness in terms of invasion and metastasis in cancers.Cholesterol and its oxygenated derivatives have shown strong affinity to G protein-coupled receptor (GPCR), i.e., Smoothened receptor (SMO), which activates the sonic hedgehog (SHH) pathway [70]. The SHH pathway is considered an oncogenic signaling cascade, as it is capable of promoting cell cycle progression and stem cell proliferation through increased activity of GLI1 and subsequent activation of hedgehog targeted genes, therefore enhancing tumor formation [71]. Inhibition of cholesterol synthesis by statins can successfully arrest SHH signaling in medulloblastoma cells and fibroblasts, thus attenuating the proliferation of tumors [72].Inflammation is an immune response to endogenous danger signals which helps to combat different stresses [73]. The causal relationship between chronic inflammation and cancer is widely established nowadays [74]. Inflammasomes, the large intracellular multi-protein signaling complexes, are formed under inflammation which help to activate inflammatory protease caspase-1, pro-inflammatory cytokines interleukin (IL)-1β and IL-18 [74]. Nod-like receptor protein 3 (NLRP3) is one of the most well studied inflammasomes among the families and its dysregulation is associated to cancer development. The uncontrolled formation of NLRP3, arising from different cellular challenges such as presentation of lipopolysaccharides, viruses, or abnormal ion fluxes, induced IL-1β and IL-18 productions, resulting in development of various cancer types, including head-and-neck squamous cell carcinoma [75], oral squamous cell carcinoma [76], and breast cancer [77]. In colorectal cancer, cholesterol promoted colon carcinogenesis through activating the NLRP3 inflammasome and suppression of AMPKα in macrophages, resulting in significant increase of mitochondrial reactive oxygen species, which in turn enhanced the NLRP3 inflammasome activity [78]. A similar positive feedback loop was observed in hepatocellular carcinoma [79]. The enhanced production and accumulation of cholesterol in liver cancer cells activated NF-κB signaling, which could promote the overall cholesterol production via activating SREBP2, HMGCR, and LDLR [79]. Yet, in other cells like endothelial cells, SREBP2 is an important mediator for NLRP3 inflammasome activation and amplification via SREBP2-TIFA and SREBP2-NOX2 cascade [80,81], further strengthening the relationship between cholesterol homeostasis and inflammation.Accumulative evidence has demonstrated the crucial role of miRNAs in cancer development [82,83]. MiR-122 was found to regulate cholesterol homeostasis, and its overexpression is required for hepatitis C virus propagation and accumulation through binding to the 5′ UTR of the hepatitis C virus genome [84,85]. Meanwhile, miR-183 promoted proliferation and anti-apoptotic properties in colon cancer cells, through retaining a high level of intracellular cholesterol via direct degradation of ABCA1 mRNA [86]. Similarly, miR-27 also exerts anti-apoptotic function in cancer cells by blocking cholesterol efflux or targeting ABCA1 [87]. Furthermore, MYC exerts its oncogenic effects in part by altering mevalonate metabolism in glioma cells via induction of miR-33b [88]. However, on the other hand, miRNAs could act as onco-suppressors. Inhibition of miR-612 induced HADHA overexpression which in turn modified cholesterol biosynthesis via SREBP2/HMGCR cascade, eventually leading to invadopodia formation and metastasis of HCC [89]. Lastly, miR-33a has shown to be suppressed in tumors derived from lung cancer [90], breast cancer [91], and colorectal cancer [92]. Particularly in colorectal cancer, cholesterol can regulate cancer development, cell cycle progression, and anti-apoptosis via miR-33a-PIM3 signaling pathway [92].Cancer stem cells (CSCs), or tumor-initiating cells (T-ICs), have been proposed to play important roles in tumor initiation, recurrence, and chemoresistance, in which dysregulated cholesterol metabolism is shown to be involved [93]. Though as a small subset inside cancer cells, a growing body of evidence of the utility of T-ICs or cells showing stem-like characteristics has tried to explain the failure of current conventional chemotherapy in which patients or cancer cells can acquire resistance to chemotherapeutic drugs after certain periods of drug administration, eventually leading to tumor relapse [94]. Therefore, isolation of T-ICs and identification of their critical signaling pathways would bear clinical significance in light of registering new targeted therapies against virtually all types of cancer.By transforming immortalized human fibroblasts into cells bearing CSC phenotypes, cells overexpressing some of the stemness regulator genes, such as sex determining region Y-box 2, octamer-binding transcription factor 4, and homeobox protein, can form tumorspheres in an anchorage-independent manner and develop tumors in immunodeficient mice [95]. Of note, a global genome expression microarray has recognized atypical metabolic pathways when comparing the sphere-forming cells against their differentiated counterparts. The sterol biosynthetic process, or cholesterol biosynthetic process, has ranked in the top five out of 15 biological processes, showing the abnormal exploitation of cholesterol in tumor formation, particularly in supporting the growth of CSCs [95].Likewise, in cancer cell lines, cholesterol biosynthesis has intimate linkage to CSC population proliferation. In colorectal cancer cells, LDL, which is the main carrier form of cholesterol in blood vessels, was shown to regulate stemness in vitro by promoting stem-like characteristics, including spheroid formation capacity, stemness-regulating genes and migration capacity [52]. Interestingly, LDL enhanced colorectal cancer progression via the MAPK pathway, which was associated with cell proliferation and differentiation. Apart from the elevation of cholesterol intake receptor, key enzymes involved in cholesterol de novo biosynthesis are also shown to be altered in every aspect. In glioblastoma, by running and comparing the RNA sequencing of patient-derived glioblastoma sphere cells and their differentiated counterparts, the super-pathway of cholesterol biosynthesis was shown to be predominantly upregulated [96]. Among those gene lists, enzymes involved in the mevalonate pathway, which mainly synthesize sterols as end-products, including farnesyl-diphosphate farnesyltransferase 1, farnesyl diphosphate synthase (FDPS), and 3-hydroxy-3-methyglutaryl-CoA synthase 1, were highly upregulated when compared to differentiated glioblastoma cells [96]. Addition of inhibitors (alendronate and zoledronate against FDPS) has rescued the tumor progression effects [96]. Similarly, breast cancer stem cells were also shown to be tightly regulated by this mevalonate metabolism pathway [97]. Intriguingly, metformin, an anti-diabetes drug, has prohibited cancer cell growth by lowering cellular cholesterol content and the stemness properties in breast cancer cells [98] and also by reducing the numbers of tumor-initiating epithelia cell adhesion molecule (EpCAM)+ hepatocellular carcinoma cells [99]. Meanwhile, excess cholesterol can inactive lysophosphatidylcholine acyltransferase 3 (Lpcat3), which is responsible for polyunsaturated phospholipid synthesis and drives stem cell proliferation in intestinal cancer in vivo and ex vivo [100]. Alternatively, inhibition of Lpcat3 or overexpression of master regulator of mevalonate pathway, SREBP2, markedly promotes intestinal tumor formation in tumor suppressor gene adenomatous polyposis coli (Apc) multiple intestinal neoplasia (Min), or Apc min-induced tumor mice [100].As an important element in the mammalian phospholipid bilayer membrane, cholesterol helps to maintain the membrane’s rigidity and to provide a medium for proper cellular signal transmission, particularly in lipid rafts [7,101]. To achieve high levels of proliferation and activation, the propagation of the cell membrane is critical not only for cancer cells but also for the maturation of immune cells in response to adverse stress. Furthermore, the receptor relocation or co-localization in the immune cells is also critical for proper activation, in which cholesterol or its derivatives participate [102]. Regarding cancer scenarios, the relationship of cholesterol to immune cells is again cell-type specific, as we will discuss in this section.T cell proliferation and activation require massive amounts of energy to support various forms of biosynthesis, and it is reported that fatty acid biosynthesis as well as cholesterol biosynthesis are highly upregulated in T cells [103]. When T cells receive an activation signal, SREBP2 maturation is granted, while LXR is inactivated [104]. Following LXR inactivation, the cholesterol efflux transporter ABCG1 is also suppressed [104,105]. Another protein inhibiting the cholesterol efflux transporter, mTOR, has also been shown to regulate CD8+ T cell differentiation through regulating cholesterol metabolism [106]. The overall intracellular cholesterol retention is beneficial for T cell activation, and SREBP is demonstrated to show an important role in CD8+ T cell activation and proliferation [107]. Alteration of cholesterol concentration can also result in insufficient composition of lipid rafts, which allow the interaction of membrane-associated proteins. Under the administration of the lipid raft disruption mediator Miltefosine, T cell proliferation is retarded by over half when compared to the control [108]. Cholesterol itself can also be regarded as a signaling molecule in the T cell community. The administration of squalene, a precursor of cholesterol, increases the population of CD4+ T cells and predisposes T cells in response to inflammatory action [109]. In contrast, upon removal of cholesterol from either growth medium or the mouse diet, total T cell activation and proliferation are retarded [110]. Meanwhile, the bloodstream levels of LDL or HDL could compromise initial T cell development, as total cholesterol is reduced [111]. The inhibition of ACAT1, which esterifies cholesterol into CEs, activates CD8+ T cell as the total plasma cholesterol level increases [112]. Oxysterols, such as 27-hydroxycholesterol, have been found to attract γδ T cells but exhaust CD8+ T cells, eventually prompting breast cancer metastasis [113]. Therefore, the accumulation of cholesterol can facilitate nanoclustering in T cells, ultimately promoting the antigen-presenting capacity and upregulation of cholesterol synthesis and influx [114]. Akin to the controversial effects of cholesterol in cancer, cholesterol has been shown to negatively regulate T cell activities. In the tumor microenvironment, high cholesterol levels can lead to CD8+ T cell exhaustion while inducing immune checkpoints, such as programmed cell death protein 1 (PD-1), natural killer cell receptor 2B4 (CD244), T-cell immunoglobulin and mucin-domain containing-3 (TIM-3), and lymphocyte-activation gene 3 (LAG-3), via the ER stress sensor XBP1 signaling cascade [115]. By suppressing the XBP1 transcriptional capacity or reducing cholesterol content in the microenvironment, the anti-tumor activity of CD8+ T cells can be restored [115]. In addition, CD8+ T cells are capable of differentiating into different subsets with various cytokine expression profiles. Among those subsets, interleukin-9-secreting T (Tc9) cells are reported to exert stronger antitumor effects when compared to Tc1 cells [116]. However, cholesterol has been shown to attenuate IL-9 expression via activating the LXR signaling cascade and inhibiting Tc9 cell activity in antitumor responses [116].Apart from T cells, other tumor-infiltrating cells have also been shown to be regulated by cholesterol or its oxidative derivatives. Neutrophils are attracted by hypoxia-inducible factor-1α (HIF1α) under the elevation of 24-hydroxycholesterol, ultimately inducing angiogenesis [117]. 25-hydroxcholesterol has been shown to advance gastric cancer metastasis in vitro by enhancing matrix metallopeptidase expression while interacting with GPCRs to trigger macrophages [118]. Cholesterol is also found to accumulate in natural killer cells and to aid in lipid raft formation and immune signaling activation [119]. The overall maturation of natural killer cells will eventually retard mouse hepatoma cell development, thus demonstrating a strategy in combating hepatocellular carcinoma [119]. The effect of oxysterol in an immunosuppressive role can be restored in dendritic cells by expression of sulfotransferase 2B1b (SULT2B1b), which converts oxysterols into impotent sulfated oxysterol [120]. Moreover, the antigen presentation efficiency of dendritic cells can be enhanced by utilizing the natural influx of cholesterol into the cells [120]. Cholesterol-modified antimicrobial peptide (AMP) DP7 (DP7-C) can efficiently deliver various antigen peptides into dendritic cells via clathrin- and caveolin-dependent pathways, thus inducing dendritic cell maturation [121]. This novel antigen-presenting technique can be utilized as a personalized cancer immunotherapy that has demonstrated excellent antitumor effects in mouse tumor models.Accumulating evidence suggests that anticancer drugs may exert their anti-proliferative activities at least in part by reducing cholesterol content/biosynthesis. For instance, it has been demonstrated that doxorubicin induced cancer cell death by decreasing HMGCR expression and reducing cholesterol levels, which was mediated by downregulation of HMGCR via inhibition of EGFR/Src pathway [122]. Other studies indicated that tamoxifen modulates cholesterol metabolism in breast cancer cells [123]. In addition, a recent article highlighted that BRD4 inhibitor JQ1 severely impacts the expression of proteins involved in cholesterol metabolism, thus leading to a strong decrease of cholesterol content in the human liver cancer cell line HepG2 [124]. Interestingly, the same report showed that administration of cholesterol counteracts the anti-proliferative effect induced by JQ1 in hepatocellular carcinoma cells, and the acquisition of JQ1-resistance is accompanied by a compensatory upregulation of proteins belonging to cholesterol homeostasis [124]. All these solid data showed the critical role of cholesterol homeostasis in tumor cell survival in response to various anti-cancer drug treatments.Apart from anti-cancer drugs, many natural compounds also exhibit a therapeutic role in cancer prevention and therapy. Many of them, including terpenoids, green tea, garlic extract, and curcumin, were found to target cholesterol homeostasis in cancer cells. Isoprenoids, also known as terpenoids, are a class of naturally occurring phytochemicals found in fruits, vegetables, and unrefined cereal grains. Several isoprenoids such as δ-, γ-, and α-tocotrienol [125], β-ionone [126], geranylgeraniol [127], and geraniol [128], were shown to suppress the growth of tumor cells by inhibiting the transcription and activity of HMGCR in various cancer types. In a large-scale compound screening, ursolic acid, a pentacyclic terpenoid, was also found to exert anti-cancer effects in hepatocellular carcinoma cells via suppression of cholesterol biosynthesis [129]. Green tea polyphenol (EGCG) was also widely reported to exert anti-cancer role in various cancers. EGCG modulates cholesterol metabolism by increasing the efflux of cholesterol and directly inhibiting HMGCR [130,131]. The cholesterol-lowering effect of EGCG was further confirmed in human clinical studies [132]. Garlic extract was reported to decrease cholesterol biosynthesis by inhibiting sterol 4alpha-methyl oxidase [133]. In addition, several garlic-derived organosulfur compounds, including S-allylcysteine and ajoene, have been found to inhibit HMGCR activity [134]. Curcumin has a long history of use as an anti-inflammatory agent. The active component of curcumin was found to induce cell death of tumor cells. Recently, curcumin was found to suppress cholesterol biosynthesis superpathway via targeting squalene monooxygenase, which was found to complement the effect of statin in cancer therapy [135]. In addition, curcumin was able to suppress cholesterol uptake in colon cancer cells by downregulation of NPC1L1 expression [136]. Collectively, natural compounds can be used to regulate cholesterol homeostasis not as a primary cancer therapy but also as an adjuvant to complement current molecular therapies.Considering the elevated cholesterol levels in different types of cancer, as discussed in previous chapters, no matter if they are the result of the overexpression of cholesterol biosynthetic genes, enhancement of the cholesterol import mechanism or suppression of cholesterol export activity, abnormal cholesterol content can eventually activate oncogenic pathways, or their derivatives can exert immunosuppressive roles. Therefore, it is practical to apply cholesterol-lowering drugs to prevent cancer incidence and to treat cancer.Regarding the de novo biosynthesis of cholesterol, the process involves more than 20 enzymes that are potential candidates of drug intervention to regulate the overall cholesterol biosynthesis. Among those proteins, HMGCR has been well established as a rate-limiting catalytic enzyme in the mevalonate pathway, converting HMG-CoA to mevalonic acid. Given its important characteristics, statins, first marketed in 1987, have been used to inhibit HMGCR by functioning as a HMG-CoA analogue, eventually decreasing the cholesterol content [137]. Different subclasses of statins have been synthesized, arising from distinctive lipophilicity and the capacity to cross the blood–brain barrier. Since it is beyond the scope of this review, the differential properties can be read in other papers [138,139,140]. Statins were first used for treating atherosclerosis, cardiovascular diseases, and liver diseases, which arise from the excess deposition of cholesterol [141,142]. However, with accumulating clinical evidence, statins have been considered as an anti-cancer drug in recent decades [143]. In prostate cancer, statin use could effectively reduce the mortality and reduce the risk of prostate-specific antigen (PSA) recurrence in a dosage-dependent manner after radical prostatectomy [144]. Patients with metastatic renal cell carcinoma benefited from statin use in terms of improved overall survival (25.6 versus 18.9 months) [145]. The clinical data are further supported by several epidemiologic studies. The incidence rates of cancers, for example, liver, gastric, colorectal, pancreatic, and prostate cancers, are reduced under the administration of statins (Table 1). Meanwhile, laboratory experiments are also able to show that the use of statins could decrease proliferation and viability of human cancer cell lines [146,147,148].Other enzymes involved in the mevalonate pathway can also be targeted (Table 2). Bisphosphonates have been used to inhibit FPP synthase in converting mevalonate into farnesyl diphosphate [149]. Lapaquistat is used to inhibit squalene synthase [150], while Lamisil is used to attenuate SQLE, which is considered to be an oncogene [151]. Zaragozic acids, which also inhibit the production of squalene, show inhibitory effects in both lung carcinoma and lymphoma growth [152]. Another potential inhibitor to the oxidosqualene cyclase (OSC), Ro 48-8071, effectively reduces the progression and metastasis of pancreatic and colorectal cancers via prohibiting the production of lanosterol, thus limiting cell proliferation and migration [153].Apart from inhibiting the cholesterol synthesis inside the body, the intake from dietary cholesterol can be intervened by taking Ezetimibe, which disrupts the NPC1L1 protein on enterocytes and lowers LDL-C [163,164]. This has been shown to inhibit the tumor angiogenesis in prostate tumors and, hence, progression [163]. Previously, we have shown that CEs assume a pro-tumor position. Therefore, inhibition of the production of CEs shows promising anti-tumor effect. For example, the administration of the ACAT1 inhibitor avasimibe suppresses CE production and even restores the imatinib sensitivity in a myelogenous leukemia cell line and, hence, retards the growth [165]. A similar effect of ACAT1 inhibition in tumor suppression has also been demonstrated in prostate and triple-negative breast cancer cells [166,167]. Methyl-β-cyclodextrin [MβCD], a cholesterol depletion chemical, is used to disrupt the lipid rafts, which are important segments in the cell membrane for proper signaling transduction and oncoprotein embedment [162]. MβCD has been shown to induce apoptosis in breast cancer cells via activating the pro-apoptotic caspase-3 signaling cascade [168]. Given that MβCD disrupts the membrane integrity, it can synergize the efficiency of tamoxifen, as the drug can easily pass through the membrane [169,170,171].Finally, a combination treatment using conventional anti-cancer drugs and drugs targeting cholesterol metabolism proposes a promising result in treating cancers. For example, in sterol hormone-related cancers, the administration of statins sensitizes anti-hormonal drugs in breast and prostate cancers [22,172,173]. Ro 48-8071, an inhibitor of OSC, has improved tissue perfusion and thus synergizes the 5-fluoroouracil anti-tumoral effect in human colon carcinoma [153]. Avasimibe, the inhibitor of cholesterol esterification, has been well documented with chemotherapeutic drugs such as doxorubicin in inducing apoptosis in a tumor model [174]. Avasimibe is also considered an immunotherapeutic, as it can boost adaptive anti-tumor immunity in head-and-neck cancer cells when combined with a dendritic cell vaccine, or it controls melanoma progression when combined with anti-PD-1 immunotherapy [175].This review has discussed the important role of cholesterol metabolism in cancer development. Elevated cholesterol contents are observed in different cancers and in the reprogramming of cholesterol biosynthesis. In addition, cholesterol and its metabolites have been involved in several oncogenic pathways, which echo uncontrolled cholesterol metabolism. The capacity of cholesterol for modulating cancer stem cells has stoked the discussion of cancer recurrence and drug resistance, while the immunomodulatory effect can contribute to promising immunotherapy.Though insightful advancements have been made in cancer research, there are still questions to be answered. Are there any other hidden factors that, left unknown, contribute to the inconsistent results in epidemiological and clinical studies of cholesterol in cancer, despite the promising anti-tumor effects in laboratory data? What are the factors determining the preferential utilization of cholesterol between cancer cells and immune cells? How do they complete with each other in harvesting cholesterol as their utmost nutrient? Can any other cholesterol derivatives contribute to cancer development or perform immunosuppressive roles? Moreover, if cholesterol is understood as an essential compartment in cellular integrity, any artificial alterations in cholesterol metabolism could be compensated by the cell itself. Such alterations, similar to every medication, take statins as an example, could result in tremendous side effects, such as aching muscles and disruptions in liver and stomach functions. Could cholesterol homeostasis or other metabolisms be restored naturally without any drug interventions? By answering these questions, we shall gain more knowledge of the molecular mechanisms of cholesterol homeostasis and cancer development, which may potentially shine more light on cancer eradication.Conceptualization, E.H.K.M. and T.K.W.L.; writing-original draft preparation, E.H.K.M.; writing-review and editing, E.H.K.M. and T.K.W.L.; supervision, T.K.W.L.; funding acquisition, T.K.W.L. All authors have read and agreed to the published version of the manuscript.The study was supported by the RGC General Research Fund (15104119) and the Theme-based Research Scheme project (T12-704/16-R).The authors declare no conflict of interest.5′ adenosine monophosphate-activated protein kinase, AMPK; ATP-binding cassette subfamily A member 1, ABCA1; ATP-binding cassette subfamily G member 1, ABCG1; Apolipoprotein A-I, ApoA-I; Protein kinase B, Akt; Acyl-CoA: cholesteryl acyltransferase 1, ACAT1; Cancer stem cells, CSCs; Cholesteryl ester, CE; Confident Interval, CI; Endoplasmic reticulum, ER; Farnesyl pyrophosphate, FPP; Farnesyl diphosphate synthase, FDPS; G protein-coupled receptor, GPCR; Geranylgeranyl pyrophosphate, GGPP; High-density lipoproteins, HDL; HDL cholesterol, HDL-C; 3-hydroxy-3-methylglutaryl coenzyme A, HMG-CoA; HMG-CoA reductase, HMGCR; Hazard Ratio, HR; Insulin-induced gene protein 1, INSIG1; Interleukin, IL; Low density lipoprotein receptor, LDLR; LDL cholesterol, LDL-C; Lysophosphatidylcholine acyltransferase 3, Lpcat3; Liver X receptor, LXR; Methyl-β-cyclodextrin, MβCD; Mammalian target of rapamycin complex 1, mTORC1; Niemann-Pick type C1-like 1 protein, NPC1L1; Nod-like receptor protein 3, NLRP3; Non-Hodgkin lymphoma, NHL; Nuclear SREBP2, nSREBP2; Oxidosqualene cyclase, OSC; Proprotein-convertase-subtilisin-kexin type-9, PCSK9; Phosphoinositide 3-kinase, PI3K; Phosphatase and tensin homolog, PTEN; Sterol-regulatory element binding protein 2, SREBP2; Site 1 protease, S1P; Site 2 protease; S2P; SREBP-cleavage activating protein, SCAP; Squalene epoxidase, SQLE; Smoothened receptor, SMO; Sonic hedgehog, SHH; Tumor-initiating cells, T-ICs; Interleukin-9-secreting T, Tc9; Total cholesterol, TC; Green tea polyphenol, EGCG; microRNAs, miRNAs; Nod-like receptor protein 3, NLRP3.Cholesterol metabolism and key oncogenic pathways related to cancer development. (a) Cholesterol de novo biosynthesis. Starting from three molecules of acetyl-coenzyme A (CoA), cholesterol is synthesized in more than 20 enzymatic steps, whereas 3-hydroxy-3-methylglutaryl-CoA reductase (HMGCR) and squalene epoxidase (SQLE) act as rate-limiting enzymes. (b) Systematic diagram showing the cholesterol metabolism in relation to key oncogenic molecular pathways. Sterol-regulatory element binding protein 2 (SREBP2) regulates the transcriptional activity of cholesterol biosynthesis genes, low density lipoprotein receptor (LDLR)-mediated cholesterol influx, and Nod-like receptor protein 3 (NLRP3) inflammasome-associated inflammation. Embedded in SREBP2 gene, microRNA (miRNA)-33 can positively regulate SREBP2 expression. The over-activated cholesterol biosynthesis contributes to uncontrolled cell growth. Overexpressed proprotein-convertase-subtilisin-kexin type-9 (PCSK9) facilities the lysosomal degradation of LDLR, induces hypercholesterolemia, and eventually leads to the development of hepatocellular carcinoma. Excess cholesterol is exported via ATP-binding cassette (ABC) subfamily A member 1 (ABCA1) under liver X receptor (LXR) activation. However, in cancercells, ABCA1 is prohibited by the phosphoinositide 3-kinase (PI3K)/protein kinase B (Akt)/mammalian target of rapamycin complex 1 (mTORC1) pathway. The overall retention of intracellular cholesterol facilitates acyl-CoA:cholesteryl acyltransferase 1 (ACAT1), converting cholesterol into cholesteryl esters, leading to the development of different types of cancer. ABCA1 can also be inhibited by miRNA-27 and miRNA-183. TP53-mediated SREBP2 activation increases the production of farnesyl pyrophosphate (FPP) and geranylgeranyl pyrophosphate (GGPP) in the mevalonate pathway, resulting in prenylation of small Ras family GTPases and their downstream effectors. An increased SQLE level under high nuclear SREBP2 (nSREBP2) induction inhibits phosphatase and tensin homolog (PTEN) activity and sequentially allows the PI3K/Akt/mTORC1 signaling cascade. Lastly, cholesterol or its oxidative derivatives activate Smoothened receptor (SMO) in the sonic hedgehog (SHH) pathway. The overall alterations in these pathways increase the proliferation rate and the migration and invasion capacities, allow cell survival, and induce tumor formation.Clinical and epidemiological studies linking cholesterol, statin use, and cancer risks.Therapeutic targets of cholesterol homeostasis.

Med-MDPI/cancers/cancers-12-06-01411.txt

@@ -0,0 +1 @@
+Macrophages are key innate immune cells in the tumor microenvironment (TME) that regulate primary tumor growth, vascularization, metastatic spread and tumor response to various types of therapies. The present review highlights the mechanisms of macrophage programming in tumor microenvironments that act on the transcriptional, epigenetic and metabolic levels. We summarize the latest knowledge on the types of transcriptional factors and epigenetic enzymes that control the direction of macrophage functional polarization and their pro- and anti-tumor activities. We also focus on the major types of metabolic programs of macrophages (glycolysis and fatty acid oxidation), and their interaction with cancer cells and complex TME. We have discussed how the regulation of macrophage polarization on the transcriptional, epigenetic and metabolic levels can be used for the efficient therapeutic manipulation of macrophage functions in cancer.Tumor microenvironment (TME) is the place of intimate crosstalk between all cellular components, including malignant, endothelial, stromal, and immune cells [1]. The TME shapes the intracellular program of cells, regulating their functionality. Signals that are involved in such regulations are cytokines, growth and transcription factors, oxygen levels, and nutrients [1,2]. Immune cells in the TME reprogram their phenotype to a tumor-associated one, maintaining the survival, growth, and proliferation of tumor cells [3]. In this context tumor-associated macrophages (TAMs) are one of the main cellular components involved in tumor progression, by regulating angiogenesis, initiation and growth of tumors, angiogenesis, lymphangiogenesis, local and distant metastasis [4,5,6]. Macrophages are extremely plastic cells that respond to stimuli from the local microenvironment acquiring a specific phenotype and reflecting the functionally distinct macrophage populations [7].Macrophages can be classified into two major subtypes that reflect two major vectors of functional polarization: classically activated pro-inflammatory, or M1 macrophages, and alternatively activated anti-inflammatory, or M2 macrophages [8,9]. However, this nomenclature is artificial and reflects in vitro generated subtypes, while macrophages in vivo (including TAMs) are highly diverse cells, and can combine M1 and M2 molecular characteristics and functions. A number of studies have shown that changes in metabolism, transcriptome, and epigenetic-associated mechanisms provide macrophages with unique functional plasticity that is detrimental when they respond to cancer cell-derived signals and start to support tumor progression. However, such plasticity makes TAMs highly attractive targets for therapeutic reprogramming. Complex interaction in TME often involves extracellular metabolites that act as communication signals [2,10]. By changing the metabolism and transcriptome of macrophages, it will be possible to modulate their functions making them beneficial for the treatment of patients with cancer. For example, depending on the stimuli, macrophages can switch from the oxidative phosphorylation to the glycolysis, and vice versa [10,11]. Recent studies have shown a number of transcriptional factors participating in the differential activation of macrophages [12,13]. A class of small noncoding RNAs, microRNAs, also were found to participate in macrophage polarization [14]. Moreover, programs for the differentiation of monocytes and maturate macrophages are based on the significant epigenetic modifications (DNA methylation, histone modifications, miRNA.) [15].In this review, we focus on three major mechanistic levels that define macrophages phenotype and functional polarization: transcriptional factors, epigenetic modifications, and metabolic pathways. We discuss these three mechanistic levels in the context of programming of macrophage functions in cancer, and outline the perspectives for the reprogramming of TAMs to develop complex and personalized anti-cancer therapeutic approaches.Transcription factors (TFs) respond to virtually all stimuli of the tumor microenvironment including cytokines, growth factors, extracellular matrix (ECM) components, metabolic factors, and control gene expression through the transactivation or transrepression domains [16]. TF activity is mediated by complex of functional domains, through which TFs binds to the appropriate DNA strand, interacts with other TFs, coactivators, and RNAII polymerase enzyme [17], chromatin remodeling complexes, and small noncoding RNAs [14]. More than half of known TFs in the genome are expressed in macrophages under the different states of polarization, and functional activation of macrophages is controlled by number of TFs [2,18]. Below we summarize the knowledge about major transcription factors that define development, activation and plasticity of macrophages in the context of the TME (Table 1, Figure 1).PU.1 is a prominent transcriptional regulator of myeloid cell development and phenotype plasticity [19,63,64]. It is a principal transcription factor that activates promoters of Csf1r gene encoding a key receptor for the macrophage lineage commitment and regulation of macrophage, differentiation and functional activation [65]. Both interferon (IFN) regulatory factors IRF8 and IRF4 bind PU.1 cooperatively at the IRF/PU.1 site in RAW264.7 cells [66]. PU.1 promotes macrophage differentiation toward alternatively activated macrophages and is involved in the development of many types of tumors including breast cancer [67], myeloma [68], acute myeloid leukemia [69], glioma [70] and hepatocellular carcinoma [71].PU.1 mediates monocyte/macrophage differentiation via activation of miR-22 in human leukemia cell line (HL-60), human monocytic cell line (THP1) cells and CD34+ hematopoietic stem/progenitor cells [72]. In vitro, PU.1 was found to be a critical regulator of M2 polarization via the IL-4/STAT6 signaling pathway in murine bone marrow-derived macrophages (BMDMs) [48] (Table 1). PU.1-deficient murine macrophages displayed decreased expression of IL-4-induced specific markers, chitinase 3-like 3 (Ym-1) and resistin-like molecule alpha 1 (Fizz-1) [48]. PU.1 knockdown resulted in reduced alternative activation of macrophages that was associated with decreased expression of CCL22, while lipopolysaccharide (LPS) treatment resulted in up-regulation of PU.1 expression accompanied by increased level of CCL22 in murine BMDMs [49]. There is also evidence about the involvement of PU.1 in the regulation of M1 polarization. Thus, miR-181a induces macrophage polarization to M2 phenotype through suppression of the expression of PU.1, C/EBPα and KLF6 in human macrophages [50]. PU.1 is a transcription factor required for the efficient inflammatory reactions in macrophages. Thus, in a mouse model with functional PU.1 knockout in mature macrophages, the inhibition of inflammatory gene expression (COX-2, iNOS, TLR4) and inflammatory cytokine secretion (IL-6, MCP-1, IL-1β, TNF-α), as well as significant decrease in systemic inflammation, was identified [51]. Although during the last decade significant progress in the study of PU.1-mediated plasticity of macrophages was achieved, the mechanism of PU.1-shaped phenotypes of macrophages in the tumor microenvironment remains to be incompletely understood.Signal transducers and activators of transcription (STATs) are a family of transcription factors that were originally identified as classic effectors of interferon-induced signaling. STATs affect macrophage phenotypes in response to cytokines and growth factors through the different signaling pathways underlying the role of STATs on TAM functional programming [26,51,54,55,57,58,59,60,61,73,74,75,76,77] (Figure 1). Thus, STAT1 mediates M1 macrophage polarization via the IFNγ and TLR signaling pathways [54]. In patients with locally advanced cervical cancer the increase in the amount of CD68+pSTAT1+ cells, defined as M1 macrophages, in tumor mass was associated with a longer disease-free survival (DFS) and overall survival (OS) [55]. However, in contrast to human studies, STAT1, but not STAT3 or STAT6, was responsible for immunosuppressive activity of TAMs derived from colon CT-26 tumor-bearing BALB/c mice [56].STAT3 is involved in angiogenesis and tumor progression through polarization of TAMs to the M2 phenotype [57,73]. In RAW264.7 cells (mouse macrophage cell line) and in BMDMs, STAT3 phosphorylation mediates IL-4 and TGFβ1-induced macrophage polarization toward the M2 phenotype that is exacerbated by Wnt3a [58]. In a co-culture of hepatocellular carcinoma (HCC) cells and macrophages, IL-6/STAT3 signaling pathway was suppressed in M1 macrophages but was activated in M2 macrophages [59]. Similar result was obtained in monocytes of healthy donors cultivated in the presence of PC3 (prostate cancer cell line) conditioned medium where M2 phenotype was characterized by IL-10-induced phosphorylation of STAT3 [57] (Table 1). In tumor-associated myeloid-derived suppressor cells (MDSCs), STAT3 was required for the induction of angiogenic factors, including VEGF and bFGF, and increased angiogenesis in vitro [60]. Phosphorylated STAT3 and STAT6 together cooperated to increase cathepsin expression in TAMs resulting in the enhanced tumor invasion in vivo [74]. STAT6 mediates the stimulation of M2-like polarization of macrophages in response to IL-4 and/or IL-13, mediators of Th2 immune responses [75] (Figure 1). IL-4-driven activation of STAT6 leads to the inhibition of TRIM24 activity in macrophages, supporting polarization of macrophages toward the tumor-associated phenotype in a murine model of melanoma [61]. In a murine model of colorectal cancer, activated STAT6 and KLF4 are involved in MFHAS1-induced M2 polarization of TAMs leading to tumor progression [76]. In murine mammary carcinoma, TAMs facilitate metastatic colonization by secretion of IL-35 through activation of JAK2–STAT6-GATA3 signaling [77].There are several therapeutic approaches suggested for the inhibition of tumorigenic action of STATs in macrophages. For example, liposome-encapsulated STAT3 inhibitor can activate reprogramming of CD163+TAMs toward pro-inflammatory phenotypes with increased secretion of IFNγ, IL-12, TNFα, IL-2 in vitro [78]. Another study demonstrated that herbal acidic polysaccharide IAPS-2 inhibits the phosphorylation of STAT3 and enhances STAT1 phosphorylation in TAMs from S180 tumor tissues (a syngeneic sarcoma) promoting macrophage polarization toward the M1-like phenotype [79]. Inhibition of the STAT6 pathway by using small interfering RNA or the pharmacologic inhibitor AS1517499 inhibited the differentiation of murine RAW264.7 macrophages into the M2 phenotype, as demonstrated by the reduction of ARG1 and CD206 expression [77]. Besides, AS1517499 significantly attenuated tumor growth and early liver metastasis in 4T1 mammary carcinoma mouse model [77].Thus, transcription factors from STAT family are involved in the macrophage plasticity by programming phenotypes towards M1 or M2 in response to the temporal and spatial stimuli in the tumor microenvironment (Table 1).Transcription factors of nuclear factor-κB (NF-κB) family regulate the expression of genes that control inflammation, immune responses, cell survival, cell proliferation and differentiation [80]. Inflammation has a dual role in cancer progression [81]. Inflammation in the microenvironment supports cell transformation and intratumoral mutagenesis [82]. On the other hand, induction of inflammation may have a potent anti-tumor effect. NF-κB is a key transcription factor of M1 polarization which is required for induction of a number of pro-inflammatory cytokines [46,47,83]. Thus, RAW 264.7 cells stimulated by IFN-γ are polarized to M1 macrophages via NF-κB signaling pathway [84]. In the tumor microenvironment, TAM-derived IL-10 inhibits IL-12 production associated with the lack of NF-κB activation promoting tumor survival, while blocking of IL-10 restores the IL-12 production in a mouse model of fibrosarcoma [85] Tumor-promoting activation of NF-κB in macrophages was also demonstrated [46]. TAMs polarized to immunosuppressive phenotype with high expression of IL-10, TNF-α, and ARG1, but low expression of NOS2, IL-12, and MHC II, that was mediated by the IL-1R and MyD88 via NF-κB activation, resulted in increased tumor invasiveness and tumor growth in ovarian cancer in vitro and in vivo [46]. IL-17 promotes THP-1 cell differentiation towards M2-like phenotype (characterized by increased expression of CD163 and CD206, TGF-β, VEGF and IL-10 production) through NF-κB signaling pathway [47] (Table 1). Another study showed that expression of PD-1 in RAW264.7 cells can be regulated by TLR/NF-κB signaling [83].Despite the fact that NF-κB is considered as a potential activator of pro-inflammatory M1 phenotype, it seems that the role of NF-κB signaling in TAM plasticity depends on the stimuli from the TME and from the type of cancer.c-Myc was identified in 1981 as a gene activated by avian leukosis virus that was implicated in the development of bursal lymphomas [86]. c-Myc is a member of the Myc family of transcription factors that regulate broad range of cellular processes  including cell cycle, metabolism, epithelial–mesenchymal transition (EMT), metastasis and angiogenesis, thereby playing a crucial role in genesis of tumor disease and tumor progression [87]. c-Myc was identified as M2-polarizing transcription factor in murine macrophages [20]. Transcriptomic analysis of murine BMDMs demonstrated that c-Myc is a marker of M2 macrophages activated by IL-4 [20]. c-Myc modulates M2-polarization via IL-4–dependent induction of genes involved in alternative activation of human macrophages (e.g., SCARB1, ALOX15, and CD206) [21]. c-Myc inhibition by treatment with 10058-F4 or transduction of c-Myc by c-Myc/shRNA in human macrophages stimulated by tumor-conditioned medium from PANC-1 (human pancreatic cancer cell line) suppresses expression of protumoral genes (ALOX15, CD206, TGF-β, VEGF, HIF-1α and MMP9) [21] (Table 1). c-Myc is expressed in CD68+ TAMs [21]. STAT6 is required for c-Myc modulated alternative type of macrophage polarization [88]. The recent study in mature murine BMDMs cultured in conditioned medium of Hepa1-6 (murine hepatoma cells) demonstrated that Wnt/β-catenin signaling mediates polarization of M2 macrophages through activation of c-Myc that supports the progression of hepatocellular carcinoma (HCC) [22]. Interestingly, in the co-culture model of human monocytes and HCC cells, IL-12 inhibits c-Myc and STAT3 transcription factors in monocytes, mediates M1 polarization and suppresses the HCC growth [89]. Moreover, deletion of c-Myc in macrophages resulted in the reduced expression of pro-tumor genes (e.g., VEGF, MMP9, and HIF1a) in TAMs and reduced tumor development in a mouse model of melanoma [23]. Thus, c-Myc is an essential transcription factor that defines development of pro-tumoral phenotype of TAMs.Other transcription factors involved in the polarization of macrophages include a family of interferon regulatory factors (IRFs) that have been originally identified as transcription activators and repressors of interferon (Table 1, Figure 1). The family of IRFs includes nine members: IRF1, IRF2, IRF3, IRF4/PIP/LSIRF/ICSAT, IRF5, IRF6, IRF7, IRF8/ICSBP, and IRF9/ISGF3γ, that participate in the regulation of both development and activation of the immune system cells [90]. Notably, IRF1, IRF5, and IRF8 contribute to pro-inflammatory polarization of macrophages while IRF3 and IRF4 regulate M2 polarization macrophages [24,25,31,33]. Thus, IRF1 is involved in M1 polarization in human macrophage cell line U937 in response to IFNγ and LPS by upregulation of IL-12, IL-6, IL-23 and CD86 and downregulation of M2-specific marker CD206 [24]. The knockdown of IRF1 in macrophages induces their pro-tumor activity regarding to hepatocellular carcinoma cell lines HepG2 and SMMC-7721, promoting proliferation and invasion of tumor cells [24].IRF3 promotes M-CSF-mediated differentiation of monocytes toward M2 type macrophages [25]. IRF3 activates PI3K/Akt signaling mediating the inhibition of pro-inflammatory genes (IL-1α, IL-1β, TNFα, IL-6, IL-8, and CXCL1) and stimulation of anti-inflammatory genes (IL-1RN, IL-10, and IFN-β) in human fetal microglia [26]. Under LPS treatment, TAMs isolated out of murine fibrosarcoma showed impaired MyD88-dependent NF-κB activation and activation of the MyD88-independent IRF-3 pathway [27]. This was consistent with low expression of several pro-inflammatory cytokines (e.g., IL-1β, IL-6, TNF-α) and up-regulation of immunosuppressive cytokines (IL-10, TGFβ) and IFN-inducible chemokines (CCL5, CXCL9, CXCL10, and CXCL16) [27]. In a murine model of nasopharyngeal carcinoma, cancer progression is mediated by EBV encoded RNAs (EBER)-triggered inflammation dependent on the phosphorylation of p38 and IRF3 [91]. However, TAM polarization to pro-inflammatory M1 status can also dependent on TLR3- and TLR4-IRF3 signaling [28,29]. IRF3 phosphorylation and transcriptional activity is regulated by Smad2 and Smad3 [92]. Double knockdown of Smad2\3 in BMDMs is critical for the phosphorylation of IRF3 and STAT1 transcriptional activities and IFN-β production in response to LPS [92]. Another study has demonstrated the inhibition of pro-tumorigenic genes encoding VEGF and MMP2 in IRF3- and IRF7-transduced macrophages. Additionally, IRF7 displayed cytotoxic activity of macrophages in breast cancer (SK-BR-3, MCF-7) and colorectal cancer (COLO-205) cell lines [30]. IRF7 also can serve as a factor, regulating IL-10 response [35]. In human monocyte-derived macrophages IRF-7 knockdown by siRNA increased LPS-induced IL-10 production, indicating that IRF7 induction blocks early IL-10 response [35].IRF4 and histone demethylase Jumonji domains containing-3 (Jmjd3) are important players in IL-4-induced M2 polarization of macrophages acting through the activation of M2-specific genes (ARG1, FIZZ1, Ym1, and CD206) [31]. The level of IRF4 protein together with STAT3 and P-STAT3 proteins was elevated in monocyte-derived M2 macrophages induced by IL-6 in vitro [32]. ChIP assay demonstrated that IRF4 can be recruited to the PU.1 site and trans-activated by the MR enhancer reporter (pGL3-MR) in RAW264.7 cells, while transfection of macrophages with miR-125a suppressed IRF4 expression and pGL3-MR transactivation [93]. Interestingly, the miR-23a/27a/24-2 cluster reduced the production of M2 type cytokines by directly targeting JAK1/STAT-6 pathway with miR-23a and by targeting IRF4 and PPAR-γ with miR-27a [94]. qRT-PCR analysis of tumor samples from renal cell carcinoma patients revealed positive correlation between M2-associated genes (CD163, FN1 and IRF4) and reduced survival [95].High expression of IRF5, in contrast, is associated with activation of inflammatory gene expression (IL-12p40, IL-12p35 and IL-23p19) and inhibition of anti-inflammatory genes that promotes M1 polarization in macrophages [34]. Furthermore, co-expression of IRF5 and IKKβ (a kinase that phosphorylates and activates IRF5) mediates TAM polarization towards M1 phenotype, supressing tumor development in model systems of advanced-stage ovarian cancer, metastatic melanoma, and glioblastoma [33].Another study demonstrated that Notch-RBP-J signaling regulates expression of IRF8 inducing the expression of M1-specific genes in RBP-J deficient mice [18]. Moreover, IFN-γ-induced IRF8 is involved in the activation of transcription of pro-inflammatory genes [96]. Inhibition of IRF8 in macrophages reduces expression of inflammatory mediators associated with M1 macrophage (IL-1b, IL-6, iNOS, and TNF-α) and delayed wound healing in vivo [97]. IRF8 deficiency in macrophages significantly increased metastasis and expression of metastatic-associated genes in the mouse models of mammary cancer and melanoma, and correlated with reduced survival in human breast and lung cancers and melanoma [36]. High levels of IRF8 expression is associated with prolonged DFS in renal cell carcinoma patients [37].Thus, IRFs play an essential role both in the polarization of macrophages and in the formation of tumor-associated phenotype. The direction of the pro- or anti-tumoral effects depends on the type of IRF.The regulatory role of macrophage polarization was also found for SNAIL family members, differentially expressed both in TAMs and in cancer cells [98]. This family consists out of three members: SNAIL1 (SNAIL), SNAIL2 (SLUG) and SNAIL3 (SMUC) that contain a zinc finger-binding domain. Transcriptional regulation by SNAIL has been involved in various biological processes in cells, including modulation of EMT via the inhibiting E-cadherin transcription, and regulation of cell adhesion [99]. In THP-1 cells, SNAIL participates in TGF-β induced activation of M2-like phenotype through the PI3K/AKT and Smad2/3 signaling pathways [53] (Table 1). At the same time, M1 polarized macrophages displayed reduced expression of lysine-specific demethylase 1 (LSD1) and SNAIL. The LSD1 inhibitor phenelzine increased expression of M1-like signatures both in vitro and in vivo in a murine model of triple-negative breast cancer [100]. Overexpression of SNAIL in human head and neck cancer cells regulates the transcription of microRNA-21 that promotes the production of miR-21-containing exosomes from tumor cells [98]. When CD14+ monocytes engulf tumor-derived miR-21-containing exosomes, they display increased expression of M2-like markers (CD206, CD163, IL-10) and down-regulation of M1-like markers (IL-18, IL-12B, HLA-DR). Knockdown of miR-21 in cancer cells attenuated the SNAIL-mediated M2 polarization, angiogenesis, and tumor growth [98].Maf family of transcription factors comprises MafA, MafB, Maf (also known as c-Maf), NRL11, MafF12, MafG13 and MafK. Maf family belongs to the AP-1-type superfamily bZip and participates in the proliferation and differentiation of hematopoietic cells [101]. MafB (v-maf musculo-aponeurotic fibrosarcoma oncogene homolog B) and c-Maf are well-known transcription regulators of macrophage differentiation and polarization in both human and murine models [43,102,103,104]. In BMDMs from adult wild-type mice the expression of MafB was induced by IL-10 or IL-4/IL-13 and suppressed by LPS or GM-CSF. In the same model, c-Maf expression was induced by IL-10 and suppressed by IL-4/IL-13 or GM-CSF [102] (Table 1). MafB induced by IL-10 in human primary macrophages activated STAT3 signaling pathway leading to the increased expression of MMP9 and IL-7R genes [42]. LPS-stimulated peritoneal macrophages derived from macrophage-specific dominant-negative MafB transgenic mice showed increased expression of IL-6 and TNF-a [104]. MafB+ macrophages expressed high levels of IL-10, ARG1 and TNF-α in Lewis lung carcinoma (LLC) of MafB-GFP knock-in heterozygous mice [43]. Besides, strong expression of MafB was identified by immunostaining analysis in CD204+ and CD68+ TAMs on stage 3 of human lung cancer [43]. Elevated expression of MafB in TAMs was also demonstrated in a mouse model of breast cancer [105]. A recent study found that M2 macrophages induced by IL-4 and IL-13 express high levels of c-Maf that regulates expression of M2-related genes (IL-12, IL-1b, IL-6, ARG1, IL-10, VEGF, TGFb, IRF4, and CCR2) [106]. c-Maf is expressed by TAMs in human non-small cell lung carcinoma (NSCLC), and promotes M2-mediated T cell suppression and tumor progression by controlling M2-related genes in vivo [106]. Deletion of c-Maf in macrophages resulted in reduced tumor size and enhanced antitumor T cell immunity in vivo [106]. Thus, the tumor-supporting role of Maf in TAMs was found in several cancer models in mice as well as in human tumors.The Microphthalmia family of bHLH-LZ transcription factors (MiT/TFE) is a family of four leucine zipper transcription factors: MITF, TFEB, TFE3 and TFEC [107]. The MiT family members are involved in many basic cellular processes including lysosomal biogenesis and autophagy [108]. MITF family members are expressed in macrophages, and TFEC is a macrophage-specific transcription factor [109]. TFEB regulates TAM polarization in the tumor microenvironment. Knockdown of TFEB with TFEB shRNA lentivirul vector in mouse peritoneal macrophages resulted in the suppression of expression of M1 markers (NOS and TNF-α) and stimulation of expression of M2 markers (ARG1 and YM-1) [62]. In co-culture experiment of breast cancer cell line and macrophages, TFEB-knockdown in macrophages promoted their polarization to the M2-like phenotype through the downregulation of SOCS3 production and STAT3 activation. TFEB knockdown in EO771 or LLC-derived C57BL/6 mice resulting in enhanced angiogenesis, tumor growth and reduced infiltration of CD8+ T cells [62]. Besides, the activation of TFEB by hydroxypropyl-β-cyclodextrin in macrophages suppressed their M2 polarization and inhibited breast tumor growth in mice [62].Kruppel-like factors (KLF) family is comprised of 17 zinc-finger transcription factors [110]. KLF4 and KLF6 regulate key cellular processes, such as differentiation, proliferation, and programmed cell death [38,40,41,111]. KLF4 induces M2-like polarization via STAT6 signaling and reduces M1-like activation depending on NF-κB activation in RAW264.7 cells [38]. In murine peritoneal macrophages, KLF4 and STAT6, induced by IL-4, promoted M2 polarization of macrophages via MCPIP (monocyte chemotactic protein-induced protein) activation and up-regulation of expression of ARG1 and FIZZ1 [39]. KLF4 and MCPIP suppressed LPS-induced expression of NF-κB target genes (iNOS, IL-1β, TNFα and IL-6) and inhibited M1 polarization [39]. Deletion of KLF4 in murine myeloid cells resulted in suppression of expression of M2 markers (ArRG1, CD206, IL-10, TGF-β1, and Chil3) and reduction of HCC growth [112]. KLF4 stimulates M2 polarization of TAMs via Hedgehog signaling pathway in LLC1-derived mice [112].KLF6 is required for LPS and IFN-γ-induced macrophage polarization to M1-like phenotype acting in cooperation with NF-kB signaling [40]. It inhibits anti-inflammatory gene expression by downregulating PPARγ expression in macrophages (RAW264.7 cells and BMDMs) in vitro [40]. KLF6 mediates activation of pro-inflammatory gene signature through activation of NFκB signaling, and inhibits anti-inflammatory gene expression through the downregulation of STAT3 signaling in vitro in RAW264.7 cells and in vivo in KLF6-KO mice [41].Transcription factor NFAT5 drives pro-inflammatory activation of both M1 (activating IL-12) and M2 (activating FIZZ-1 and ARG1) macrophages [45]. NFAT5-deficient macrophages had reduced pro-inflammatory status, followed by the reduced infiltration of cytotoxic CD8+ T cells into the tumor and the enhanced tumor growth of LLC and ID8 ovarian carcinoma models [45].Thus, we can conclude that the polarization of macrophages toward pro-inflammatory or anti-inflammatory phenotypes depends on the variety of transcription factors. At the same time, TFs are activated by different signals from the microenvironment resulting in functional reprogramming of macrophages. Targeting of transcription factors in macrophages is a promising strategy to use macrophage plasticity for the reprogramming TAMs by blocking their tumor supporting activity and by activating their intrinsic anti-tumor functions (recognition and killing of transformed cells). However, specific delivery of the inhibitors to TAMs avoiding other cell types in various organs is still a biotechnological challenge.The epigenetic level of regulation is critical for the differentiation and functional programming of macrophages [15,113]. There are three levels of epigenetic control of macrophages differentiation and activation: DNA methylation, histone modifications, and microRNA [15,114]. DNA methylation is essential for the macrophage differentiation [115,116]. Histone methylation is a principal epigenetic mechanism for activation of inflammatory reactions in macrophages. The regulatory role of epigenetic remodeling by microRNA has been observed in differentiation and functional activation of macrophages [117,118,119]. Epigenetic differences between M1 and M2 macrophages act as important functional determinants [15,120,121].DNA methylation is methylation of 5′-carbon on cytosine bases located frequently in CpG islands of promoters [122,123,124]. DNA methylation prevents transcriptional machinery from the assembling on the altered promoter that leads to the silencing of gene transcription [122]. There are two states of DNA methylation: hypermethylation (gain–CH3) and hypomethylation (loss–CH3). Hypermethylation is characterized by the transfer of a methyl group to the cytosine ring in DNA by DNA methyltransferases (DNMTs) to form 5-methylcytosine. DNMT1, DNMT-3A and DNMT-3B are involved in this reaction [125,126]. Hypomethylation is a removal of methyl groups by ten-eleven translocation (TET) proteins [127,128]. DNA methylation in CpG islands is an active mechanism of the repression of gene expression [129,130,131]. Moreover, CpG methylation prevents also aberrant intragenic transcriptional initiation [130,131]. In cancer, DNA methylation is critical for the suppression of the expression of tumor suppressor genes while loss of DNA methylation leads to the overexpression of oncogenes.There are evidences that DNA methyltransferases have specific effect on the formation of macrophage phenotypes (Figure 1). DNMT3b knockdown promotes macrophage polarization to alternatively activated M2 phenotype in RAW264.7 cells [132]. DNMT1 is implicated in M1 polarization by silencing the SOCS1 gene and a subsequent increase in TNF and IL-6 production [133]. Overexpression of DNMT1 promotes LPS- and IFN-γ-induced M1 activation whereas inhibition of DNMT1 attenuates it [133] (Table 2). Upregulation of DNMT1 correlates with decrease in peroxisome proliferator-activated receptor gamma (PPAR-γ) and with the increased production of pro-inflammatory cytokines in peripheral blood monocytes isolated from patients with atherosclerosis and in macrophages from adipose tissue [116,134]. In type 2 diabetic mice, decrease in the ability of macrophages to support wound healing was associated with microRNA let-7d-3p, which was up-regulated by DNMT1 resulting in the differentiation of cells toward the M1 phenotype [116]. However, the effect of LPS in BMDMs during M1 activation is also associated with a significant reduction in the expression of DNMT 1, 3a and 3b, and a significant increase in the expression of TET2 and TET3 [116] (Table 2, Figure 1). TET2 expression is increased in intratumoral myeloid cells, both in a mouse model of melanoma and in melanoma patients, that is dependent on an IL-1R-MyD88 pathway [135]. Recently, the combination of mass spectrometry and single molecular imaging demonstrated that LPS induces global changes in DNA methylation of the genome of murine macrophages [113,134].Despite the clearly established role of DNA methylation in the classical inflammatory macrophage models, its role in the formation of TAM phenotypes in various tumor types is not understood. There are only some isolated reports showing that DNA methylation is involved in the MDSC function [176]. Since DNA methylation is a critical factor for cancer cell biology, there are a number of studies trying to identify epigenetic enzymes as targets for anti-cancer therapy. Therefore, understanding of the mechanism and functional consequences in DNA methylation in TAMs is urgently needed.Histone modifications, also known as histone code, provide a highly flexible mechanism for activation and deactivation of transcription in macrophages in response to the changing context of stimuli in the TME. Histone modifications in various cell types include a number of post-translational modifications such as methylation, acethylation, ubiquitination, arginine citrullination, sumoylation. The histone code is an essential mechanism that controls the activity of cancer cells [177]. The most frequent histone modifications, also found in macrophages, include acetylation and methylation, and the most frequently modified amino acid is a lysine [178]. Histone modifying enzymes regulate macrophage phenotypes through the addition or removal of acetyl/methyl groups. Acetylation and deacetylation are initiated by histone acetyltransferases (HAT) and histone deacetylases (HDACs), respectively [178]. Histone acetylation is associated with the activation of transcription, whereas histone deacetylation is associated with transcriptional repression. Bromodomain-containing proteins (BRD) and some extraterminal-motif containing proteins (BETs) are also involved in transcriptional regulation by recognizing histone acetylation sites via bromodomain acetyl-binding pocket [175,179]. BETs inhibit or activate the assembly of the transcriptional machinery regulating inflammatory cytokine (IL-1b, IL-6, TNFa, MCP-1) production [175,180]. BRD4 and BRD9 act in the SWI/SNF chromatin remodeling complex in the context of inflammatory stimulation of macrophages [181].Methylation and demethylation of histones are catalyzed by histone methyltransferases (HMT) and histone demethylases (HDM), respectively. Histone methylation can induce both transcriptional activation and repression, depending on the number and location of the methyl groups [129]. An active transcriptional state is characterized by the presence on the gene promoters or enhancers of activating histone marks such as H3K4me1 and H3K4me3. The repressed state of transcription is associated with the increase in labeling in H3K9me2/me3 and H3K27me3 [182].Remarkably, histones code acts not only on the promoters, but also on the enhancers that are critical for the differentiation and activation of myeloid precursors and mature macrophages. Single-cell RNAseq demonstrated that various populations of myeloid cells are formed already at the level of bone marrow precursors, that are controlled by a variety of transcription factors (PU.1, Cebp-a, -b and –ε, IRF8, ATF3) [183,184]. The activity of these transcription factors is regulated by histone modifications on the enhancers [183]. Moreover, di- or tri-methylation of histone H3 in lysine-4 and -79 is associated with gene activation, while the methyl group (H3K9me2/3 and H3K27me3) relates to transcriptional repression [183]. Depletion of PU.1 in primary macrophages resulted in the decreased activation of methylation of H3K4 in many enhancers [150]. The importance of gene function regulation using H3K4me2 in enhancers and promoters of IRF8 and CSF1R genes has been established for monocyte progenitors [183]. According to the ChIP-seq data obtained in the projects of the BLUEPRINT consortium, differences between monocytes and macrophages for histones H3K4me3 (promoters), H3K4me1 (enhancers) and H3K27ac (active promoters and enhancers) were revealed [185]. Monocytes gain about 5000 enhancers and lose 3000 enhancers compared with the hematopoietic stem cell (HSC) precursors, while macrophages gain and lose 6000 enhancers when differentiated from monocytes [185]. It was shown that 2547 promoters were changed in histone acetylation status in monocytes compare with macrophages, while a differential pattern of histone acetylation was found in 4036 enhancers [186].Enzymes that control histone modification, such as HMTs [137,138,139,140,141,143,144,145,146,148,149,150,151,152], HDMs [31,154,155,156,157], HDACs [15,66,121,136,150,152,158,159,160,161,162,163,164,165,166,167,168,169,170,171,172,173], BETs [174,175] are involved in the epigenetic regulation of M1 and M2 macrophage polarization (Table 2, Figure 1). Activation of the TLR-dependent pathway in macrophages and THP1 cells is accompanied by an increase in the expression of the H3K79 inhibitor–disruptor of telomeric silencing-1-like (Dot1l) [187]. SIRT1, a specific type of HDAC, suppresses macrophage activation through TFs such as p65, LXR, and IRF8, and SIRT1 expression is downregulated in LPS-stimulated macrophages [188]. SIRT1 and SIRT2 are rapidly activated during macrophage differentiation, and their inhibition results in the upregulation of many inflammation-related genes. SIRT1 and SIRT2 interact with DNMT3B and bind to the promoters of genes that become hypermethylated during macrophage differentiation that was shown in human macrophages in vitro [161]. IL-4-activated STAT6 acts as a transcriptional repressor in an HDAC3-dependent manner in BMDMs [189].Most of the data indicates the involvement of histone modification in TAMs in the formation of immunosuppressive M2-like phenotype in tumors (Table 2). For example, activation of extracellular signal–regulated kinases-1/2 (ERK-1/2) results in the inhibition of MyD88 via interleukin 1 receptor-associated kinase 3 (IRAK M) disrupting TLR signaling in TAMs of C57BL/6 mice. Histone phosphorylation of the IL-10 promoter depends on ERK-1/2 and increases IL-10 production, but not IL-12 [190]. BET bromodomain inhibitor, JQ1, blocks the association of bromodomain-containing protein 4 (BRD4) with promoters of arginase and other IL-4-dependent macrophage genes inducing immunosuppression in the TME [191]. When combining JQ1 with a PI3K inhibitor, or using the double PI3K/BRD4 inhibitor SF2523 (previously reported as a strong inhibitor of tumor growth and metastasis in various cancer models), tumor growth was suppressed in syngenic and spontaneous mouse cancer models. This effect was accompanied by the decrease in myeloid suppressor cell infiltration, restoration of the activity of CD8+ T cells, and stimulation of the antitumor immune response [191] (Table 3).Decoy receptor 3 (DcR3) regulates the expression of HLA-DR in TAMs by affecting the expression of the main regulator of HLA-DR, CIIT-A, through the ERK- and JNK-induced histone deacetylation of CIITA promoters [192]. This is the mechanism responsible for the DcR3-mediated suppression of HLA-DR and polarization of TAMs to M2-like phenotype. The level of DcR3 expression in cancer cells was inversely correlated with HLA-DR expression levels in TAMs and with the overall survival period in patients with pancreatic cancer [192] (Table 3).The classical (M1) polarization of macrophages is accompanied by a decrease in the expression of lysine-specific histone demethylase 1A (LSD1) (demethylation of H3K4 and H3K9 essential for the myeloid cell differentiation), nuclear REST corepressor 1 (CoREST) and zinc finger protein SNAIL [193]. Treatment with phenelzine (an LSD1 inhibitor) reduced the activity of H3K4 and H3K9 nuclear demethylase that resulted in the activation of the transcription and expression of M1-like markers both in vitro and in vivo in the mouse model of triple negative breast cancer. Additionally, in vivo chemotherapy reduced tumor volume and, in combination with an LSD1 inhibitor, canceled the mesenchymal signature and stimulated an innate M1-like antitumor immune response [100,193] (Table 3).Histone deacetylases (HDACs) have an ambivalent effect on the regulation of gene expression in TAMs. Pan inhibition of HDAC by suberoylanilide hydroxamic acid (SAHA) reduced NO production in RAW264.7 cells and mouse peritoneal macrophages [194]. SAHA regulates pro-tumor TAM function and induces EMT in prostate cancer cells [194]. The use of this inhibitor together with tritepinoid as anticancer drug led to the decrease in the level of macrophage infiltration into the mammary gland in MMTV-polyoma middle T (PyMT) mice and a subsequent decrease in tumor formation [195]. A similar result was obtained for murine models of lung and pancreatic cancer where inhibition of HDAC had an antitumor effect by acting through the mechanisms of regulation of nitride oxide (NO) production in TAMs [195]. More recently, a class IIa HDAC inhibitor, TMP195, was found to reduce the tumor burden and metastasis by modulating TAM phenotypes to the antitumor, highly phagocytic cells in tumor-bearing MMTV-PyMT mice [203].Histone modifying enzymes are definitely involved in the cross-talk between cancer cells and TAMs. Thus, Jumonji domain-containing histone demethylases 1A (JMJD1A) regulated by hypoxia and nutrient starvation of cancer cells, stimulates tumor aggressiveness by enhancing the amounts of TAMs and their pro-angiogenic activity [196]. However, whether JMJD1A acts also in TAMs directly still has to be clarified. Despite the accumulated data about the critical role of histone code and histone modifying enzymes in macrophage activation, the role of the histone modifying enzymes in TAM activation in tumor-specific context has to be analyzed for the development of optimal tumor targeting strategy.Small noncoding single-stranded RNAs are evolutionarily conserved and are involved in the multistep processes of transcription, nuclear export and cytoplasmic cleavage [204]. MicroRNAs act primarily as posttranscriptional repressors via the targeting the 3′-untranslated region of mRNA, inducing its degradation or the repression of its translation. More than 60% of all protein-coding genes are directly regulated by microRNAs [114].Different miRNAs are involved in the regulation of macrophage tumor-supporting and tumor-killing activities. miR-155, miR-181 and miR-451 was found in M1 macrophages and miR-146a, miR-125a and miR-145-5p—in M2 macrophages [117,118], (Table 3, Figure 1). High expression of miR-155, miR-146a, miR-127, miR-125b in M1-polarized macrophages was confirmed in BMDMs isolated from BALB/c mice, the RAW264.7 macrophage cell line and in C57Bl/6 mice [119,205,206] (Table 3, Figure 1). miR-511-3p, miR-223 and let-7c contribute to the polarization of monocyte-derived macrophages into the M2-like phenotype [118,207]. It was demonstrated that increased levels of miR-720 resulted in the inhibition of GATA3 expression, which is important for the polarization of M2 macrophages [29]. Moreover, knockdown of miR-146a promoted polarization of macrophages into M1-like phenotype and decreased polarization to M2-like phenotype [208]. miR-99a inhibits the phenotype and function of M1 macrophages by targeting TNF-α in BMDMs of mice [17]. In P388D1 and RAW264.7 cells miR-511-3p, which was found to be highly expressed in CD206+ macrophages in N202 tumors in mice, regulates the expression of IRF4, thereby supporting expression of genes associated with the M2-like phenotype [201] (Table 3).In human monocytes stimulated by human larynx epithelioma cancer cell supernatants, and in CD14+ cells obtained from blood of patients with HCC, increased expression of miR-17 and miR-20a resulted in the stimulation of angiogenesis by IL-6-dependent production of hypoxia-induced factor 2α (HIF2a) [200]. Increased expression of miR-511-3p leads to the suppression of the transcriptomic protumoral gene signature detecting by RNAseq in human and mouse CD206+ macrophages, that is associated with the inhibition of tumor growth [202]. In addition, microRNA-19-a-3p inhibits tumor progression by downregulation of human fos-related antigen 1 (FRA-1) gene (acting as a pro-oncogene by supporting the invasion and progression of breast tumors) and the FRA/STAT3 signaling pathway in RAW264.7 cells [202].Remarkably, in number of epigenetic mechanisms were found to support M2 functions of TAMs that can be explained by the fact that M1 functions are usually activated in the acute phase of inflammation and do not require epigenetic support. The majority of the data are a still coming from the animal tumor models, and a similar role for epigenetic mechanisms in TAMs in human cancers has to be analyzed. The availability of the inhibitors of histone modifying enzymes would be an interesting approach to block M2 polarization of TAMs; however, the specific delivery of such drugs to TAMs, similarly to the delivery of drugs targeting transcription factors, remains to be developed.Numerous studies showed distinct metabolic characteristics for the two main subtypes of macrophages (M1 and M2). Movahedi and colleagues indicated that M1 macrophages are mainly normoxic, while M2 macrophages reside in hypoxic areas of tumor and have a proangiogenic activity in vivo [209]. M1 polarization displays highly glycolytic metabolism through the pentose phosphate pathway (PPP), fatty acid synthesis (FAS) which organizes the plasma membrane for inflammatory signaling, and impaired mitochondrial oxidative phosphorylation (OXPHOS) and tricarboxylic acid (TCA) cycle [210]. It is commonly considered that M1 macrophages are characterized by enhanced antimicrobial activity mediated by the upregulation of reactive oxygen species (ROS), generation of reactive nitrogen intermediates (NO), an increased production of antimicrobial peptides, and pro-inflammatory cytokines, such as IL-1β and TNFα [2,211]. M1 macrophages are able to accumulate both citrate-supported NADPH and prostaglandin E2 (PGE2), and succinate stabilized hypoxia-inducible factor 1α (HIF-1α) [2] (Figure 2). In contrast, traditionally M2 macrophages undergo a metabolic reprogramming toward oxidative metabolism for bioenergetic purposes (OXPHOS), fatty acid oxidation (FAO), decreased glycolysis, decreased metabolism via the PPP and upregulation of arginase 1 (ARG1) which is processed into ornithine to produce polyamines (Figure 2). Such metabolic features are associated with the ability of M2 macrophages to resolve inflammation and to support tissue repair [2,211,212].However, recent evidences demonstrated that FAO is also essential for inflammasome activation in M1 macrophages, while glycolysis was found to be utilized by M2 macrophages [213]. Below we describe key metabolic pathways of M1 and M2, as well as the examples of mixed metabolism that can be used by macrophages in the complex pathological conditions.It is well accepted that the key feature of inflammatory macrophages is the induction of glycolysis by the up-regulation of the glucose transporter (GLUT1) which mediates glucose uptake [214]. Overexpression of GLUT1, which is a member of GLUT family, in macrophages is associated with increased glycolysis and PPP intermediates that induce ROS production and expression of pro-inflammatory mediators such as TNFα and IL-6 [215]. Overexpression of GLUT1 in murine macrophage cell line RAW 264.1 resulted in elevated secretion of pro-inflammatory mediators, such as G-CSF, IL-6, TNF-α, IL-1ra, increase in ROS production and simultaneously in enhanced glucose metabolism [215]. Moreover, in macrophages, GLUT is controlled by HIF1α which regulates the expression of genes encoding for glycolytic enzymes as well as inflammatory mediators [10]. Thus, the upregulation of GLUT1 promotes glucose uptake that is crucial for the glycolytic activity of M1 macrophages [10,215]. ROS is a prominent factor in the activation of NFkB and p38 MAPK signaling pathways inducing pro-inflammatory gene expression in M1 macrophages [216]. Besides, ROS is involved in the activation of the nucleotide-binding oligomerization domain (NOD)-like receptor containing pyrin domain 3 (NLRP3) inflammasomes [217].LPS-activated M1 macrophages express 6-phosphofructo-2-kinase B (PFKFB3) and the pyruvate kinase M2 (PKM2) [218]. PKM2 was found to activate the LPS-induced pro-inflammatory phenotype of M1 macrophages in murine model via the production of HIF-1α, IL-1β and other HIF-1α-dependent genes as well as to promote inflammasome activation by modulating eukaryotic translation initiation factor 2 alpha kinase 2 (EIF2AK2) phosphorylation in macrophages [218,219]. Pyruvate dehydrogenase kinase 1 (PDK1) was demonstrated as a critical component of glucose metabolism, which was involved in LPS-induced macrophages activation [220]. Knockdown of PDK1 in murine BMDMs suppressed M1 by attenuating glycolytic flux, the expression of pro.inflammatory cytokines (TNF-α and IL-6) and consequently aerobic glycolysis, but enhanced M2 activation by mitochondrial respiration [220]. Moreover, combined deletion of two forms of pyruvate dehydrogenase kinase PDK2 and PDK4 in myeloid cells prevents M1 polarization and correlates with the improved mitochondrial respiration in mouse models [221]. Similarly, PDK1 was identified as a HIF-1α target gene, and HIF-1α-PDK1 axis induced active glycolysis with up-regulation of glycolytic genes, such as GLUT1, phosphoglycerate kinase 1 (PGK1) or lactate dehydrogenase A (LDHA) [222].However, there are recent evidences about the crucial need for glycolysis in M2-like macrophages both for the activation of M2-specific gene expression and for the tumor support [223,224]. Analysis of different components of Akt signaling revealed that Akt mediates enhanced glucose consumption in murine IL-4-stimulated BMDMs [225]. Depletion experiments showed that IL-4 treatment enhanced global acetylation of H3 and H4 histones at promoters of M2 genes (ARG1, Retnla, MGL2) in an Akt-mTORC1-dependent manner. Moreover, Akt controls the production of Ac-CoA, the metabolic substrate for histone acetylation. Inhibition of histone acetylase p300 as well as knockdown of Raptor, a main subunit of the mTORC1 complex, reduced induction of Akt-dependent M2 genes [223]. Increased aerobic glycolysis was also found in murine BMDMs synergistically stimulated with M-CSF and IL-4 [226]. Glycolysis and mitochondrial pyruvate import were essential for M2 activation, possibly because they were used to fuel FAS for increased FAO and OXPHOS. mTORC2-mediated phosphorylation of Akt was critical for M2 activation. Deletion of Rictor, a subunit of mTORC2 complex, diminished the expression of a number of M2-specific genes (CD301, RELMα, ARG1, Chil3 (Ym1), IL-10, LIPA, CD36, FABP4, PPARG, and PPARGC1B) and glucose uptake in IL-4-stimulated macrophages. Besides, Rictor-deficient macrophages showed inhibition of activity of transcription factor IRF4, indicating the role of mTORC2 in the expression of IRF4 in IL-4-stimulated macrophages. In an in vivo mouse model of melanoma, loss of the mTORC2 in TAMs diminished M2 activation and suppressed tumor growth [224]. Interestingly, in vitro knockdown experiments revealed that STAT6 and Akt-mTORC signaling may operate in parallel and independently in response of BMDMs to IL-4 [223,224]. Despite that it is well-known that Akt-mTORC signaling is involved in the regulation of glucose consumption and glycolysis, there is limited evidence about regulating glucose metabolism via STAT6 activation [225,226]. Further investigations of the interaction of these two significant pathways in the regulation of glucose metabolism are urgently needed.The metabolic value of pentose phosphate pathway (PPP) in M1 polarization includes conversion of glycolytic intermediates to precursors of nucleotides and amino acids. The PPP generates NADPH required for the inducible nitric oxide synthase (iNOS) to catabolize arginine into nitric oxide (NO) and l-citrulline as well as for the generation of ROS [227,228]. Suppression of PPP in macrophages attenuates oxidative stress responses and LPS-induced inflammatory cytokines that were shown in a hyper cholesterolemic mouse model [229].A truncated TCA cycle was considered as a metabolic feature of M1 macrophages leading to the accumulation of citrate and succinate [230,231,232]. Citrate can be involved in fatty-acid synthesis, which is essential for membrane biogenesis [230], and in the generation of inflammatory effector molecules such as NO and prostaglandin that negatively modulate mitochondrial activity by disrupting electron transport chain [231,233]. Pyruvate dehydrogenase (PDH) activity is needed to synthesize citrate from glucose-derived pyruvate, while citrate is used for lipogenesis and for the production of the pro-inflammatory mediators such as NO [210]. Succinate is associated with the pro-inflammatory function of M1 macrophages [210]. LPS-induced succinate in macrophages enhanced IL-1β production by stabilizing HIF-1α [232]. Succinate may indirectly stabilize HIF-1α via the induction of ROS [210].Moreover, hyperglycemia was found to induce production of pro-inflammatory cytokines and S100 proteins in human primary macrophages [234,235,236]. One of major pro-inflammatory cytokines is IL-1beta that has a complex role in tumors and promotes tumorigenesis, tumor invasiveness and immunosuppression [237,238]. S100A9 and S100A12 that are induced by high glucose in primary human macrophages have multiple cellular targets and link inflammatory processes in cancer [239]. We have recently demonstrated that hyperglycemia induces activating histone code on the promoters of these genes in primary human macrophages, that shows that there is a link between glycolytic metabolism and the epigenetic level of regulation in macrophages [236]. However, it remains to be understood how these processes interact in TAMs.A key metabolic signature of alternatively activated macrophages is the consumption of fatty acids and the increase in the mitochondrial oxidative phosphorylation (OXPHOS) [210]. Using BMDMs from CD36–/– mice it was shown that the uptake of low-density lipoproteins (LDL and VLDL) is mediated by the scavenger receptor CD36 leading to their subsequent liposomal lipolysis activating OXPHOS and FAO in M2 macrophages. Furthermore, elevated CD36 expression is substantial for the up-regulation of gene expression defining for IL-4-induced macrophages (CD206, CD301, PD-L2 and RELMαin) [240]. Surprisingly, FAO was detected as the key metabolic process involved in inflammasome activation, a key signaling event in pro-inflammatory macrophages. Inhibition of FAO by etomoxir treatment suppressed NLRP3 and consequent secretion of IL-1b and IL-18 in human and mouse macrophages [241]. FAO was shown to be required for palmitate-induced NLRP3 inflammasome activation, which involves mitochondrial ROS [242]. Additionally, in vivo delivery of CpG oligodeoxynucleotide, a Toll-like receptor 9 agonist, to tumor-bearing mice with pancreatic ductal adenocarcinoma (PDAC) cells resulted in the suppression of tumor growth in pancreatic cancer models enhancing the anti-tumor activity of F4/80+ TAMs through the induction of phagocytosis of tumor cells [243]. The anti-tumor activity of TAMs is implemented by the upregulation of FAO that is a key feature of M2 macrophage metabolism, however increased pro-inflammatory cytokines (TNF, IFNγ and CCL2) in the serum of mice were also detected. FAO inhibition by etomoxir did not alter the abundance of F4/80+ macrophages in the tumor microenvironment, however, it was associated with decreased engulfment of PDAC cells by F4/80+ macrophages [243].These numerous studies demonstrated the regulation of M2 polarization of macrophages through the impact on the key metabolic pathways. Interestingly, simultaneous stimulation with LPS and IFNγ blunted mitochondrial oxidative respiration in macrophages which cannot be restored by subsequent IL-4 stimulation that was demonstrated in mouse BMDM and human monocyte-derived macrophages [233]. The main metabolic effect was accompanied by NO which impeded M1→M2 repolarization by blunting mitochondrial respiration and preventing plasticity in M1 macrophages. Inhibition of NO improved mitochondrial function and promoted IL-4-induced repolarization of M1 into M2 [233].Thus, macrophage metabolism is not strictly limited to the glycolysis in M1 and FAO in M2 phenotypes, and examples of mixed metabolism in macrophages were also identified [210]. However, most studies are based only on in vitro data, and analysis of TAM metabolism in mouse tumor models and in patient’ material is needed to understand the complex metabolic response of macrophages to the stimuli of microenvironment in various types of cancer, and the role of TAM metabolism in their pro- and anti-tumor activities.In the tumor microenvironment cancer cells adapt their cellular metabolism to the hypoxic conditions to maintain a high proliferation rate and invasive activity. Tumor is highly limited in the energy suppliers, and cancer cells and other cells of TME compete for the oxygen and nutrients [2]. The altered metabolism of cancer cells is called the Warburg effect and is characterized by an increase in glycolysis even under aerobic conditions [227]. Cancer cells preferentially convert pyruvate into lactate. TAMs can respond to the products of altered cancer cell metabolism by changing their functional program to support tumor progression and metastasis [214].Growing evidence indicates that extracellular accumulation of lactate produced by cancer cells stimulates expression of pro-angiogenic and tumor-promoting factors in TAMs, and consequently induces TAM-mediated immunosuppression [244,245]. The importance of lactate in the activation of tumor-promoting activity of TAMs was demonstrated in co-culture system of human monocytic cell line THP-1 with MDA-MB-231 and MCF-7 human breast cancer cells [67]. Lactate programmed TAM-like phenotype of THP-1 cells (upregulation of CD206 and CD163 expression and elevated production of TGF-b1, IL-10, VEGF) and stimulated the expression and secretion of CCL5. CCL5, in turn, induced an invasive phenotype of breast cancer cells by enhancing migration, EMT and aerobic glycolysis [244]. The pro-metastatic phenotype of macrophages was also shown in a model system of TAMs differentiated from human monocytes in the presence of conditioned medium of the pancreatic ductal adenocarcinoma cells (PDAC) [245]. TAMs promoted vascular network formation and supported EMT and extravasation of cancer cells. PDAC conditioned medium stimulated glycolysis in macrophages by up-regulation of a number of glycolytic genes, including hexokinase (HK2), glucose-6-phosphate isomerase (GPI), aldolase A (ALDOA), triosephosphate isomerase 1 (TPI1) and phosphoglycerate kinase 1 (PGK1). Inhibition of glycolysis in TAMs using inhibitor of HK2, 2-deoxiglucose (2DG), significantly suppressed pro-metastatic phenotype of TAMs [245]. Analysis of TAMs from MMTV-PyMT mice and BMDMs stimulated by tumor extract from MMTV-PyMT mice revealed the significant increase in HK2, enolase 1 (ENO1), and 6-phosphofructokinase (PFKL), a key mediators of aerobic glycolysis [246] (Table 4).There are also evidences of TAM-dependent metabolic re-programming of tumor cells to aerobic glycolysis. For example, in human breast cancer tissues the positive correlation between CD68+ TAM infiltration and glycolytic enzyme expression GLUT1, GLUT3 and HK2 in cancer cells was demonstrated by immunostaining [261]. In the same study, MDA-MB-231, MDA-MB-468, MCF-7 and BT474 breast cancer cells co-cultured with TAMs polarized by conditioned medium from breast cancer cells showed enhanced aerobic glycolysis by the increase in extracellular acidification rates (ECARs), glucose consumption and lactate production [261]. Besides, breast cancer cells co-cultured with TAMs showed high expression of glycolytic enzymes, including GLUT3, HK2, PKM2 (pyruvate kinase isozyme M2) and LDHA. In this case, activated aerobic glycolysis in breast cancer cells is mediated by stabilizing HIF-1α protein [261]. TAM-enhanced aerobic glycolysis in cancer cells was also shown in lung cancer [262]. A strong correlation between CD68+ macrophages and the expression of GLUT1 and HK2 in cancer cells was found in patients with non-small-cell lung carcinoma (NSCLC) [262]. In the same study, Lewis lung cancer (LLC)-cells co-cultured with BMDMs showed active glycolysis and increased lactate production. TAM-derived TNFα facilitates glycolysis and inhibits mitochondrial biogenesis in LLC cells [262]. Moreover, TAMs can compete for oxygen with cancer cells contributing to tumor hypoxia. In LLC mouse model, TAMs isolated out of tumor expressed significantly increased levels of hypoxic factors including VEGFR, Slc2a1, PDK1, and C-X-C motif chemokine receptor-4 (CXCR4), M1-polarized marker (NOS-2) as well as M2-polarized marker (ARG1), and immunosuppressive cytokines such as TNFa and IL-10. Depletion of TAMs switched the tumor metabolism from aerobic glycolysis to OXPHOS, significantly decreased expression of glycolytic gene, reduced the amount of lactate, and decreased GLUT1 protein expression [262] (Table 4).AKT1/mTOR pathway is important for activation of glycolysis in TAMs [223]. Mechanistic target of rapamycin (mTOR) complex 1 (mTORC1) inhibitor REDD1 (regulated in development and in DNA damage response 1) was up-regulated in hypoxic TAMs of a murine model of LLC [249]. Inhibition of mTORC1 by REDD1 resulted in the shift of the macrophage phenotype towards the immunosuppressive and pro-angiogenic phenotype that was due to the inhibition of glucose uptake and glycolysis and enhancing glucose availability for endothelial cells. REDD1 deletion in TAMs from murine LLC tumor promotes tumor vessel normalization and inhibits metastasis, providing evidence about the link between TAM metabolism in hypoxia and tumor vessel morphogenesis [249].In an in vitro model of TAMs where human blood monocytes were stimulated with the conditioned medium of human melanoma cells (MV3), TAMs expressed M2 specific markers (CD206 and CD163), however they were metabolically distinct from typical M2 and had metabolic features of M1-like macrophages. TAMs polarization resulted in the increased GLUT1 and HK2 expression, increased glycolysis, and high amounts of lactate by Akt–mTOR-dependent pathway that was comparable with M1 macrophages. In parallel, TAMs were characterized by the supporting OXPHOS, presenting a high basal and maximal oxygen consumption rate (OCR), while showing low rates of FAO [247] (Table 4). This study showed that macrophages can produce lactate in response to soluble factors from condition medium of tumor cells, however the role of TAM-derived lactate in tumor progression remains to be identified.Immunohistochemical analysis (IHC) of tumors of patients with thyroid cancer (TC) also validated the increase in glycolytic enzymes and lactate receptor (GBR18, PFKFB3, PKM2) in TAMs [248]. Stimulation of human macrophages with TC-conditioned medium or co-cultivation of macrophage with TC cells induced increased glycolysis in human macrophages by elevation of ECAR in an mTOR-dependent manner. RNA-sequencing confirmed on the transcriptional level enhanced expression of genes regulating glycolysis in TAMs [248].A combination of lactate and hypoxia in TME results in the induction of ARG1 expression and increased secretion of VEGF-A by ischemic macrophages [250]. In an MMTV-PyMT mouse model of breast tumor, TAM-derived VEGF were required for the response of endothelial cells for vascular morphogenesis [250]. Interestingly, in breast cancer tissue TAMs expressing CD206 are located in well-nourished perivascular regions, whereas macrophages produced high levels of ARG1 located within hypoxic regions, far from the vasculature [250,251]. Upregulation of ARG1 in TAMs results in the production of polyamines critical for the stimulation of cancer cell proliferation [251].Thus, not only tumor cells-derived lactate stimulates tumor-promoting function of TAMs, but in turn, cancer cell-activated macrophages activate aerobic glycolysis in cancer cells leading to their survival, proliferation, and long-term maintenance. Such a metabolic feedback loop provides beneficial conditions for tumor progression.In different in vitro models hypoxia stimulated expression of HIF-inducible pro-angiogenic genes, such as VEGF, basic fibroblast growth factor (βFGF) and CXCL8, as well as glycolytic enzymes in TAMs [263,264]. As a rule, macrophages infiltrate hypoxic regions in tumors in association with increased expression of pro-migratory factors CCL2, CCL5, CSF1 [265]. It was shown that melanoma cancer cells in vitro released damage-associated molecular pattern High-Mobility Group Box 1 protein (HMGB1) in response to hypoxia [252]. HMGB1 is significantly increased in metastatic melanoma in patients, and drives the accumulation of M2-macrophages with elevated expression of YM1, FIZZ1, IL-10 in murine model of melanoma. However, the depletion of HMGB1 with shRNA in mice with B16 melanoma cells-derived tumor significantly reduced tumor growth and the amount of TAMs [252]. The significant influence of hypoxia was shown in macrophages differentiated in vitro from human peripheral blood or BMDMs isolated from mice bearing deletions in the HIF-1α or HIF-2α genes [266]. Under the hypoxia condition, primary human and murine macrophages displayed the upregulation of the cell surface receptors, CXCR4 and GLUT1, and tumor-promoting cytokines VEGFA, IL-1β and IL-8, adrenomedullin, CXCR4 and angiopoietin-2, indicating the importance of both HIFs 1 and 2 in response of macrophages to hypoxia [266].Hypoxia-inducible factors (HIFs) play a key role in the regulation of cellular responses to hypoxia. Notably, up-regulation of HIF1α promotes immunosuppressive activity of TAMs and differentiation of MDSCs to TAMs [267]. LPS was found to activate HIF-1α in murine AHA-1 macrophage cells under hypoxic conditions in vitro. LPS induced transcriptional activity, but not protein expression and DNA binding activities of HIF-1α in macrophages by a ROS-dependent pathway [268]. It was shown that hypoxia influences mitochondria electron transport chain (ETC) and drives ROS increase by acting on complexes I, II, and III of the ETC [269]. Although ROS is a key metabolic marker of M1 polarization, it was shown to play a crucial role in the differentiation of monocyte to M2 macrophages in response to M-CSF and IL-4 in vitro [270]. Inhibition of ROS generation by antioxidant butylated hydroxyanisole (BHA) specifically affects the polarization of macrophages to M2, and dramatically inhibits the expression of the M2 cytokines IL-10, CCL17, CCL18 and CCL24, but not M1 cytokines. Additionally, ROS inhibitor BHA significantly reduced the accumulation of F4/80+cells and tumor burden as well as numbers of metastatic foci in K-RAS-induced lung cancer and MMTV-PyMT-induced breast cancer in vivo [270]. ROS production is regulated by NADPH oxidases. NADPH oxidase 4 (NOX4)-overexpressed lung cancer cell lines A549 and Calu-1, induced the recruitment of murine M2-like TAMs via the ROS/PI3K signaling-dependent pathway [253]. ROS produced by cancer cells stimulates various cytokine production, including CCL7, IL-8, CSF-1 and VEGF-C, that all contribute to enhanced NSCLC cell growth. IHC analysis of clinical specimens confirmed the positive correlation of NOX4 and CD68 or CD206 [253]. ROS accumulation in BMDMs that was reached by ROS inducer, glutathione synthesis inhibitor buthionine sulphoximine (BSO), results in increased expression of programmed death ligand-1 (PD-L1) and production of IL-10, IL-17, IL-4, IL-1b, insulin-like growth factor-binding protein 3 (IGFBP-3), and chemokine (C-X-C motif) ligand 1 (CXCL1) that are associated with an immune-suppressive phenotype of macrophages [271].In another study, hypoxia promoted THP-1 cells polarization to M2 phenotype in HIF-1α-independent way, by decreasing IL-1β expression and increasing VEGF and CD206 expression [272]. In patients with glioma, IHC analysis revealed the positive correlation between HIF-1α expression, periostin (POSTN) expression, and the infiltration of TAMs (CD11b+) and M2 type TAMs (CD206+) in tumor sections. The density of TAMs increased in higher grade gliomas and in hypoxic HIF-1α-positive regions. In vitro supernatants from hypoxia-treated U87 or U251 glioma cell induced strong chemotactic effect toward THP-1 cells, upregulation of M2 marker expression (IL-10 and CCL-22) and downregulation of M1 markers (IL-6 and TNF-α), indicating the activation of M2-like phenotype under hypoxic condition. Hypoxia-inducible expression of POSTN, tumor-promoting factor and chemoattractant for macrophages, in U87 and U251 cells was increased by TGF-α via the RTK/PI3K pathway in vitro [254]. Conversely, in the model of lung adenocarcinoma, hypoxia induced the metabolic shift in TAMs from glycolysis toward TCA cycle and OXPHOS activation [255]. Thus, exosomes derived from hypoxic B16-F0, A375, A431, and A549 lung adenocarcinoma cells were highly enriched with CSF-1, CCL2, FTH, FTL, and TGFβ that induced macrophage recruitment and promoted M2 polarization. In vivo, exosome-treated BMDMs showed a shift of cell population to F4/80+CD206+ population, increased B16-F0 tumor cell proliferation and viability. ATP-linked mitochondrial OCR assay demonstrated that M2-like macrophages, polarized by hypoxic exosomes, exhibited enhanced OXPHOS activity, inhibiting AKT and mTOR and increasing expression levels of mTOR negative regulator REDD1 [255]. These numerous studies indicate that hypoxia promotes the tumor-supporting function of TAMs, which is associated with a strong induction of immunosuppressive and proangiogenic phenotype.The importance of FAO for TAMs has been recently shown in a murine colon carcinoma model and in human colon cancer where unsaturated fatty acids (oleate) induced polarization of immunosuppressive TAMs by supporting mitochondrial respiration [257]. The up-regulation of M2 specific markers (CD206, IL-6, VEGFα, MMP9, ARG1) was observed upon oleate treatment dependent on lipid droplets (LD), that play an essential role in the catabolism of free fatty acids for mitochondrial respiration. The formation of LDs in TAMs was found in tumor tissue of patients with colon cancer [257]. However, inhibition of DGAT, an enzyme responsible for the formation of lipid droplets in myeloid cells, prevented oleate-induced immunosuppressive M2 phenotype in murine BMDMs and human monocyte-derived macrophages. Besides, mTOR inhibition in myeloid cells eliminated specific lipid droplet-dependent mitochondrial respiration in M2-like macrophages [257]. In contrast, the LD formation in TAMs from a mouse mammary adenocarcinoma model was associated with significantly inhibited tumor growth. LDs were formed particularly in M1-like (MHCII+CD11c+) TAM population in E0771 breast cancer-bearing mice. This subset of macrophages demonstrated up-regulation of epithelial fatty acid binding proteins (E-FABP), a lipid chaperon. Furthermore, the expression of E-FABP in human breast tumors is reduced in macrophages of invasive tumors as compared to normal stroma, and decreased TAMs in parallel with the disease progression [258]. IFNγ induces LD accumulation in MafB/c-Maf double deficient (Maf-DKO) macrophages that depends on exogenous lipids, while de novo synthesis of fatty acids from glucose plays a minor role in this process [273] (Table 4).Other pathways are also involved in metabolic changes in TAMs. Thus, TAMs isolated from human renal cell carcinoma produce pro-inflammatory chemokine CCL2 and immunosuppressive cytokine IL-10 that is dependent on the increased metabolism of 15-lipoxygenase-2 (15-LOX2) LOX-dependent arachidonic acid [259]. TAMs isolated from tumor-bearing mice (B16 melanoma and ID8 ovarian carcinoma) induced itaconate accumulation which is catalyzed by the enzyme encoded by immunoresponsive gene 1 (IRG1) [260]. Itaconic acid stimulates OXPHOS and ROS production in TAMs. Interestingly, IRG1 protein expression was found in TAMs from tumor-bearing mice, but was not detected in B16 or ID8 tumor lysates, and Irg1 shRNA treatment significantly reduced tumor burden in both tumor models. These results indicate once again that tumors profoundly alter the metabolism of TAMs, to potentiate tumor growth [260]. Other authors reported glutamine-synthetase (GS) as mediator of the proangiogenic, immunosuppressive, and pro-metastatic M2-like macrophages. It was reported that glutamine-synthetase (GS) controlled mTOR signaling and activated IL10-stimulated M2 macrophages with pro-tumor properties [256]. Moreover, deletion of GS in macrophages promotes vascular normalization, accumulation of cytotoxic T cells, and metastasis inhibition and skews TAMs toward the M1-like phenotype in mice implanted with Lewis lung carcinoma (LLC) cells. Deletion of GS in macrophages leads to the reduced expression of M2-specific markers (ARG1, CD206, CCL17, and CCL22) and upregulation of M1 marker MHC class II [256]. GS-targeted human monocyte-derived macrophages display reduced glutamine and enhanced succinate accumulation, increasing glucose flux through glycolysis, partly through the stabilization of HIF-1α [256]. The elevated expression of GS was also revealed in TAMs isolated from glioblastoma resections and TAMs co-cultured with glioblastoma cells [274].Thus, the available data indicate that tumors can program the metabolism of intratumoral macrophages to potentiate tumor growth. Although the molecular profile of TAMs is very close to M2-prototype, in TME they obtain mixed metabolism with pronounced glycolysis, a metabolic feature of M1 macrophages. Among the number of metabolites in TME, the essential tumor-promoting role of TAMs in different cancer models was assigned to lactate released by cancer cells. Lactate increases the ability of TAMs to induce angiogenesis, tumor growth and immunosuppression. The importance of FAO, a metabolic feature of M2 macrophages, has been also demonstrated in TAMs. However, there are some contradictory results concerning the lipid droplets involved in fatty acid metabolism. The majority of studies were performed using in vivo or in vitro models, and almost no results can be found for patients. Analysis of TAM metabolism in human tumors is required in order to find therapeutic targets to stimulate the anti-tumor activity of TAMs.Each tumor is a complex organ with individual dynamics of growth, metabolism, immune status, vascularization and spread within the organism. Macrophages are key innate immune cells in the TME and at metastatic sites that have the intrinsic capacity to block cancer progression, but in the majority of tumors they are reprogrammed by cancer cells to support tumor growth and spread. Programming of macrophage functional phenotypes is controlled on the transcriptional, epigenetic, and also on the metabolic levels. Close interplay of transcriptional factors and epigenetic enzymes is responsible for the activation of pro- or anti-tumor programs, and is utilized by cancer cells to give instructions to macrophages to support tumor progression. The progress in our understanding of essential elements and mechanisms that control interaction between transitional factors and epigenetic mechanism in complex TME resulted in the identification of a promising target for therapy. For example, inhibition of some TFs, such as STAT3 or STAT6, c-Maf, c-Myc, in macrophages can significantly attenuate tumor growth and metastasis of tumors [21,77,78,79,106].Metabolism of macrophages attracted more recently strong attention of the research community mostly due to the role of macrophages in development of diabetes and its complications. However, cancer cells can control macrophage activation also by modulation of their metabolic pathways. Despite that TAMs are considered to have an M2 phenotype; in TME they can have mixed metabolism with pronounced glycolysis, a metabolic feature of M1 macrophages, and less pronounced FAO. Metabolic re-writing is an attractive idea for therapeutic inhibition of tumor-promoting activity of TAMs but needs a deep understanding of which types of metabolism (glycolytic or FAO) are beneficial for the tumor and which for the patient.There are several immunomodulatory approaches based on the targeting of macrophage metabolism. A clinical study based on the administration of oleic acid combined with Vitamin D-binding Gc-globulin-derived macrophage activating factor (GcMAF) in patients with advanced cancer (including colorectal cancer, breast cancer, melanoma, thyroid cancer, renal carcinoma) was performed [275]. Administration of the OA-GcMAF complex resulted in a significant reduction in tumor size, demonstrating greater anticancer effects and immunotherapeutic activity than GcMAF alone. One of the possible mechanisms of this effect is releasing of NO responsible for the anti-cancer properties of activated macrophages [275]. Another vitamin D binding protein-macrophage activating factor (DBP-maf) was demonstrated to inhibit the growth of hepatocellular carcinoma in tumor-bearing severe combined immunodeficiency (SCID) mice [276]. In vitro DBP-maf inhibited the proliferation of endothelial cells and activated phagocytosis by macrophages [276]. Targeting glutamine metabolism using glutamine antagonist JHU083 demonstrated the inhibition of metastasis and enhanced anti-tumor immunity in 4T1 (breast cancer) tumor-bearing mice resulting in the improvement of the efficacy of anti-PD1 and anti-CTLA4 therapy [277]. Glutamine antagonist JHU083 induced the repolarization of MDSCs to inflammatory macrophages and enhanced immunogenic tumor cell death and antigen presentation of TAMs [277].In conclusion, understanding the complexity of the mechanism of the interaction between transcriptional, epigenetic and metabolic programming of macrophages is the next challenge that will allow identifying pharmacological targets for immunomodulatory therapy in specific tumor types. However, the development of delivery systems for specific targeting for pro-tumoral TAMs in different types of cancer is the next task for biotechnology.Conceptualization, I.L. and J.K.; Writing—Original draft preparation, I.L., E.K. and M.P.; Writing—Review and editing, I.L. and J.K.; Figure preparation, E.K.; Supervision, J.K.; Funding acquisition, J.K. All authors have read and agreed to the published version of the manuscript.This research was funded by Russian Science Foundation, grant number 19-15-00151. The publication costs were covered by the Tomsk State University competitiveness improvement programme.The authors declare no conflict of interest.Transcription factors and epigenetic enzymes involved in macrophage polarization. ARG1—Arginase 1; HDAC—Histone deacetylase; HDM—Histone demethylase; HMT—Histone methyltransferase; P—phosphorylated form. Figure created in biorender (http://biorender.io).Metabolic characteristics of M1 and M2 macrophages. ARG1—Arginase 1; FAO—fatty acid oxidation; FAS—fatty acid synthesis; G6P—Glucose 6-phosphate; GLUT1—glucose transporter; NADH—Nicotinamide adenine dinucleotide; OXPHOS—oxidative phosphorylation; PDK—Pyruvate dehydrogenase kinase; PPP—pentose phosphate pathway; ROS—reactive oxygen species; TCA—tricarboxylic acid. Figure created in biorender (http://biorender.io).The role of transcription factors in macrophage polarization.Notes: BMDMs—bone marrow-derived macrophages; DFS—disease-free survival; HCC—hepatocellular carcinoma; LLC—Lewis lung carcinoma; LPS—lipopolysaccharide; MDMs—monocyte-derived macrophages; OS—overall survival; RCC—renal cell carcinoma.Epigenetic effectors involved in macrophage polarization to M1 or M2 direction.Notes: ARG1—Arginase 1; LPS—lipopolysaccharide; ROS—reactive oxygen species; TLR—Toll-like receptor.Mechanisms of epigenetic regulation in tumor-associated macrophages (TAMs).Notes: BRD4—bromodomain-containing protein 4; DcR3—decoy receptor 3; EBV—Epstein-Barr virus; EMT—epithelial-mesenchymal transition; FGF—fibroblast growth factor; HCC—hepatocellular carcinoma; HDAC—Histone deacetylase; HDM1A—Histone demethylase 1A; JMJD1A—jumonji domain-containing protein 1A; LLC—Lewis lung carcinoma; LSD1—lysine-specific histone demethylase 1; MMP—matrix metalloproteinase; SCID—severe combined immunodeficiency; TET2—Tet methylcytosine dioxygenase.TAM reprogramming by tumor-derived metabolic factors.Notes: BMDMs—bone marrow-derived macrophages; ECAR—extracellular acidification rate; EMT—epithelial-mesenchymal transition; FAO—fatty acid oxidation; GS—glutamate synthesis; LLC—Lewis lung carcinoma; NSCLC—non-small-cell lung carcinoma; OXPHOS—oxidative phosphorylation; PDAC—pancreatic ductal adenocarcinoma; TC—thyroid cancer.

Med-MDPI/cancers/cancers-12-06-01412.txt

@@ -0,0 +1 @@
+Kirsten-RAS (KRAS) has been the target of drugs because it is the most mutated gene in human cancers. Because of the low affinity of drugs for KRAS mutations, it was difficult to target these tumor genes directly. We found a direct interaction between KRAS G12V and tumor suppressor novel H-REV107 peptide with high binding affinity. We report the first crystal structure of an oncogenic mutant, KRAS G12V-H-REV107. This peptide was shown to interact with KRAS G12V in the guanosine diphosphate (GDP)-bound inactive state and to form a stable complex, blocking the activation function of KRAS. We showed that the peptide acted as an inhibitor of mutant KRAS targets by [α-32P] guanosine triphosphate (GTP) binding assay. The H-REV107 peptide inhibited pancreatic cancer and colon cancer cell lines in cell proliferation assay. Specially, the H-REV107 peptide can suppress pancreatic tumor growth by reduction of tumor volume and weight in xenotransplantation mouse models. Overall, the results presented herein will facilitate development of novel drugs for inhibition of KRAS mutations in cancer patients.RAS was identified by the extensive study of retroviral oncogenes isolated from the genome rat-derived Harvey and Kirsten murine sarcoma viruses [1]. RAS is a proto-oncogene that is mutated in human cancer, and the RAS protein is encoded by three expressed genes: Harvey-Ras (HRAS), Kirsten-RAS (KRAS), and neuroblastoma-Ras (NRAS) [2]. The H−, K−, and NRas proteins are small GTPases that serve as master regulators of countless signaling cascades involved in particularly diverse cellular processes, such as cell division, differentiation, cell–cell adhesion, growth, and apoptosis. The small GTPase RAS proteins operate as “molecular switches” that fluctuate between inactive and active states. These switches can be activated by the exchange of guanosine triphosphate (GTP) for guanosine diphosphate (GDP), which is promoted by guanine nucleotide exchange factors (GEFs). In contrast, they are inactivated when GTP is hydrolyzed to GDP, which is promoted by GTPase-activating proteins (GAPs) [3].Activating mutations in RAS are found in certain human cancers. Some 9–30% of human tumors have RAS-activating mutations, which are common in KRAS (86%), NRAS (11%), and HRAS (3%) [4,5]. KRAS has been a target for drug design for more than 30 years because it is the most commonly mutated oncogene in human cancers, including pancreatic (90%), colon (40%), and non-small cell lung cancers (20%) [6,7]. The National Cancer Institute (NCI) recently highlighted the importance of this drug goal to bring together researchers who are developing new ideas to stop Ras by announcing a USD 10 million effort called the RAS project. The project is designed to promote the development of new drugs or treatment that can benefit cancer patients.Point mutations in the KRAS gene are present in over 90% of pancreatic ductal adenocarcinoma (PDAC) cases and are thought to be an early event in the development of PDAC that is already occurring in PanIN 1A lesions of the pancreas [5,8]. Kulemann et al. compared KRAS mutations in pancreatic circulating tumor cells (CTC) and corresponding tumors, and evaluated their significance as prognostic markers [9]. Cancers with RAS mutations are aggressive and respond poorly to standard therapies; accordingly, they have earned a reputation as being “undruggable” because scientific researchers have failed to design a drug that successfully targets the mutant gene. These mutations render RAS proteins insensitive to GTP-induced hydrolysis of GTP to GDP and lock them in the activated state. Carcinogenic mutations cause functional activation of Ras family proteins by impairing GTP hydrolysis [6]. Thus far, potential inhibition molecules have been reported to almost indirectly target RAS-functional interactions without binding to RAS.The KRAS mutations are usually found to affect residue G12 and less commonly residues G13 and Q61 [4,5]. For example, the most common KRAS mutation types are G12C, G12D, and G12V accounting for almost 83% of all KRAS mutations [10]. Recently, Liu’s group showed that a covalent inhibitor specific for G12C mutant KRAS induces tumor regression in in vivo models [11]. KRAS mutations in ovarian serous borderline tumors (OSBTs) and ovarian low-grade serous carcinomas (LGSCs) have previously been reported [12,13]. Interestingly, ovarian carcinoma patients with the KRAS G12V mutation appeared to have shorter overall survival than those without this mutation [14]. For these reasons, selectively targeting the KRAS G12V mutant is among the highest priorities of ovarian cancer therapy. The number of mice developing lymph node metastases was also found to be higher in KRas G12V (73%) and KRas G13D (29%) than in KRas wild-type (11%) mice. KRas G12V showed higher tumor cell survival, invasion, and CXCR4 expressing intravasated tumor emboli than KRas G13D. In human CRC tumors, of 12 different point mutations found at KRas codon 12 or 13, only the KRas G12V mutation conveyed an increased risk of recurrence and death [15]. The mutation at codon 13 has been found to occur predominantly in a subset of CRCs defective in DNA repair genes, while Q61 is essential for GTP hydrolysis and its mutation blocks Ras-mediated GTP hydrolysis and leads to tumor formation [16,17,18].The H-REV107 gene is a known member of the class II tumor suppressor gene family that has been identified as a growth inhibitory RAS target capable of suppressing anchorage independent growth in vitro an in vivo [19]. Several studies have shown that H-REV107 is ubiquitously expressed in most normal tissues, however, expression is lost in human tumor cell lines and tumor samples. Furthermore, H-REV107 expression is strongly reduced in approximately 50% of human ovarian carcinomas. Loss of the human H-REV107 in ovarian carcinoma cells is a result of reversible down-regulation via the MAP/ERK pathway. In contrast, induction of H-REV107 expression resulted in growth inhibition of RAS-transformed cells in vitro and in vivo [20,21,22,23]. We previously modeled the binding of H-REV107 protein to GNP-bound KRAS mutation (Q61H) [24].Because of the very low affinity of the drug for KRAS mutations, it was difficult to target these tumor genes directly. Here, we found direct interaction and inhibition between KRAS G12V and novel H-REV107 peptide. It acts as an inhibitor of mutant KRAS targets (G12V, G12D, G12C, G13D, and Q61H) by [α-32P] GTP binding assay. Treatment with the H-REV107 peptide effectively inhibited pancreatic cancer and colon cancer cell lines in cell proliferation assay by inducing apoptosis. We also determined the first crystal structure of an oncogenic mutant—KRAS G12V protein bound with novel H-REV107 peptide. The peptide directly interacts with KRAS G12V at the GDP-bound inactive state and peptide bound KRAS G12V forms a stable open form complex that can block RAS activation function. Specifically, the peptide can suppress pancreatic tumor growth by reduction of tumor volume and weight in xenotransplantation mouse models. Overall, the results of this study will facilitate development of new effective drugs for the inhibition of KRAS mutations in patients with cancer.First, we expressed and purified KRAS G12V and H-REV107 to understand their molecular interactions (Figure 1A–C and Figure S1A,B). The oncogenic KRAS mutants (G12V, G12D, G12C, G13D, and Q61H) were then constructed and purified, after which their secondary structures were studied. The circular dichroism (CD) spectra of the KRAS mutants showed that each mutation affected the conformation of KRAS to a different extent. Size-exclusion chromatography (SEC) analysis revealed the presence of KRAS G12V and H-REV107 complex in the final purified peak (Figure 1C). Using SEC-MALS (multi-angle light scattering), we found that the KRAS G12V and H-REV107 complex had a molecular mass of 46 kDa on the SDS-PAGE gel. The band of His-KRAS G12V (amino acids 1–188) was located at 27 kDa, while that of His-H-REV107 (aa 1–125) was at 19 kDa. The peak of the KRAS G12V-H-REV107 complex was shown as the sum of the molecular mass of a 1:1 complex (Figure 1D). The complex of the KRAS G12V and H-REV107 protein was purified; however, the complex crystal was not obtained. Nevertheless, we successfully obtained the crystal of KRAS G12V and H-REV107 peptide complex (Figure S2E). The H-REV107 peptide is highly soluble in water. The binding affinity of the H-REV107 protein/peptide to KRAS G12V was determined by Biacore biosensor analysis. In our previous work, the KD values of WT or Q61H KRAS were determined by surface plasmon resonance (SPR) to be 1–9 nM [24]. These were shown high binding affinities with the H-REV107 protein in the previous study. In this paper, the H-REV107 peptide also showed high binding affinities to G12V or G12D, with a KD of 1–3 μM. By using isothermal titration calorimetry (ITC) analysis, KRAS G13D, Q61H, and G12C showed binding affinities to the peptides, with KD = 17–50 μM. KRAS G12V showed binding affinity to H-REV107 protein with KD = 30 μM. The ITC results in Figure S3A,B showed ambiguous binding affinities (Figure 1E,F and Figure S3).The reduction of GTP binding affinity to KRAS mutant was predicted to be linked to the interaction with KRAS oncogenic mutants (G12V, G12D, G12C, G13D, and Q61H) and H-REV107. We evaluated this hypothesis by [α-32P] GTP binding assay using KRAS mutant and H-REV107 protein with bovine serum albumin (BSA) as a negative control. The addition of KRAS G12V and BSA both resulted in high levels of [α-32P] GTP binding in the absence of H-REV107 protein. However, the [α-32P] GTP binding activity of KRAS G12V was decreased (9%) in the presence of H-REV107 protein (Figure 2A). We also evaluated the binding ability of H-REV107 peptide in a [α-32P] GTP binding assay using KRAS mutants (G12V, G12D, G12C, G13D, and Q61H) with unlabeled guanosine 5’-tirphosphate (GTP) as a negative control. The results showed that [α-32P] GTP binding to the KRAS mutants was greatly decreased in the presence of H-REV107 peptide (Figure 2B–G). Following the addition of H-REV107 peptide, GTP binding to the KRAS mutants (G12V, G12D, and G12C) decreased to 50%, 40%, and 10%, respectively, of the GTP binding activities to KRAS mutants. In addition, the GTP binding activities to KRAS Q61H, G13D, and wild-type were decreased to 20%, 7%, and 12%, respectively. These findings indicate that the H-REV107 peptide (65LYDVAGSDKY74) inhibits the interaction between KRAS mutant targets (G12V, G12D, G12C, G13D, and Q61H) and GTP. Among their interactions, GTP binding to the KRAS G12V mutant by the H-REV107 peptide showed the greatest decrease (Figure 2B), indicating that the H-REV107 peptide inhibits the oncogenic mutant KRAS G12V well.Five tumor cell lines of different lineages were used to test the inhibitory effect of H-REV107 peptide on their proliferations. To evaluate the inhibitory effect of this peptide on cells that were in proliferative state, a longer incubation period was also tested. When the peptides were incubated with cells for up to 72 h, the decrease in cell viability was observed on SW480 (colon cancer), AsPC-1 (pancreatic cancer), NCI-H23 (lung cancer), NCI-H460 (lung cancer), and HCT116 (colon cancer) (Figure 3). H-REV107 peptide exhibited a GI50 value of 358 μM against SW480 cell line while it showed a GI50 value of 417 μM against AsPC-1 cell line. In addition, H-REV107 peptide also displayed antitumor activities with GI50 values in millimolar ranges for other tumor cell lines (NCI-H23, NCI-H460, and HCT-116). MTT results showed that higher concentration state of H-REV107 peptide had an effect on the proliferation of wild-type KRAS A549 cell line than other tumor cell lines. On the basis of the inhibitory effects of H-REV107 peptide on tumor cell proliferation, we examined H-REV107 peptide impact on the MEK/ERK signaling pathway in SW480 and AsPC-1 cells. We observed that H-REV107 peptide could decrease the phosphorylation levels of MEK/ERK in SW480 cell line, but the tendency to decrease of the phosphorylation levels of MEK/ERK in the AsPC-1 cell line was little. Western blotting had little effect on the expression level of the total MEK/ERK protein (Figure 3F,G).To investigate the anti-tumor effect of H-REV107 peptide in vivo, we inoculated pancreatic AsPC-1 cancer cells on the skin in a xenograft nude mouse model and injected the peptide into the abdominal cavity.AsPC-1 cell lines were cultured using the RPMI-1640 medium (10% FBS, 1% P/S), and the AsPC-1 cell lines were transplanted into the right flank skin with 1 × 106/animal to produce a subcutaneous xenotransplantation animal model. The separation was conducted when the average volume of tumors grew to more than 100 mm3 after the transplantation of cancer cell lines, and consisted of three groups: Vehicle, H-REV107 peptide 50 mg/kg, and H-REV107 peptide 200 mg/kg. Five were deployed per group. The H-REV107 peptide was administered 15 times in the abdominal cavity for 3 weeks and weighed once a week during the test period and twice a week, and the tumor size was measured. During the experiment, no weight loss, infirmity, or behavioral abnormalities were observed and normal weight levels were displayed in all groups (Figure 4D,E). For tumor volume, from day 3 to day 32 after inoculation, the tumor volume was observed to decrease in all H-REV107 peptide doses compared to the Vehicle dose group, and the mean weight of the tumor measured during autopsy was 68% and 75% of the H-REV107 peptide 50 mg/kg and 200 mg/kg, respectively, compared with the Vehicle dose (Figure 4B,C). In this experiment, when combined with the above results, the H-REV107 peptide experimental group observed a lower growth rate of tumors compared to the Vehicle administration and the lowest growth rate in the group administered at a high concentration (200 mg/kg) (Figure 4A).After studying the molecular interaction of KRAS mutants and H-REV107, we attempted to make their crystals. First, we obtained the crystal of KRAS G12V (1–168)-MgGDP and determined its structure by the molecular replacement method using the dimeric structure of KRAS G12V with GDP (PDB ID: 5UQW) as a search model (Figure S2C) [25]. This structure contains the sequence of the full-length KRAS G12V-MgGDP protein, except for the last 20 residues in the hypervariable region (residues 169–188). The KRAS G12V (1–168)-MgGDP was obtained as a new monomeric crystal form that belongs to the hexagonal space group P63, with unit cell parameters a = b = 82.547, c = 40.804 Å, α = β = 90, and γ = 120°. The crystallographic parameters and data collection statistics are summarized in Table 1. The structure of KRAS G12V-MgGDP consists of five α-helices (α1–α5) and six β-strands (β1–β6). The P-loop, and switch I and II regions in the structure are located in residues 10–17, 30–38, and 60–76, respectively (Figure 5A). The folding type of KRAS G12V-MgGDP is an α/β doubly wound shape, which is a mostly parallel sheet with helices on both sides. In this structure, KRAS G12V contains one molecule of MgGDP per asymmetric unit.Next, the crystal structure of H-REV107 was determined by the molecular replacement method using the crystal structure of human HRASLS3 with a search model (PDB ID: 4DOT) [26]. We obtained the fraction crystal of the N-terminal domain (residues 1–125), but the electron density for residues 1–4 and 40–56 was poor (Figure S2D). The crystal belonged to the orthorhombic P212121, with unit cell parameters a = 42.935 Å, b = 52.997 Å, c = 62.796 Å, and α = β = γ = 90 o (Table S1). The structure of H-REV107 comprises four α-helices (α1–α4) and six β-strands (β1–β6) (Figure 5B). We designed a structural-based inhibitory peptide for targeting of KRAS mutants (Figure 5C). When the structures of our H-REV107 and HRASLS3 (PDB ID: 4DOT) proteins were superimposed over Cα atoms, the root mean square deviation (RMSD) showed a difference of 0.535 Å.Finally, we successfully obtained the crystal of KRAS G12V-H-REV107 peptide (65LYDVAGSDKY74) complex and determined it at 2.3 Å resolution by the molecular replacement method using the coordinates of our KRAS G12V structure (Figure 5D and Figure S2E). The H-REV107 peptide was synthesized in a water-soluble form and its fraction was designed from α1 to β5 (amino acids 65–74). The H-REV107 peptide was bound to one molecule per four molecules of the KRAS G12V with a solvent content of 47% per asymmetric unit (Figure S2F). This structure comprises residues 1–168 of KRAS, 4 MgGDP, and 83 water molecules with an H-REV107 of 10-mer peptide. A ribbon representation of the crystal structure of the KRAS G12V-H-REV107 peptide complex was shown and the peptide and MgGDP electron density maps were clearly indicated (Figure 5E,F). The refinement is summarized in Table 1.Notably, the H-REV107 peptide was strongly docking in close proximity to the switch I and II binding pockets and the near P-loop of KRAS G12V. Seven residues (L65, Y66, D67, G70, D72, K73, and Y74) of the H-REV107 peptide bound to the KRAS G12V protein (Figure 5G,H). The L65, D72, K73, and Y74 residues of the H-REV107 peptide interacted with KRAS G12V and G13, as well as with Q61. Additionally, KRAS G13 on the P-loop and H-REV107 peptide G70 strongly bound to β-phosphate of the GDP and make crucial contact with the phosphate of the GDP. The hydroxyl group of the H-REV107 peptide Y66 strongly bound to S17, D33, and A59 via hydrogen bonds, and interacted with the phosphate binding P-loop, and switch I and II regions. Charged H-REV107 peptide D67 also interacted with P34 and T35 in the switch I region, while charged H-REV107 peptide K73 interacted strongly with KRAS N85. Mg2+ bound to the OH groups of the S17 in the P-loop, T58 around switch II, and oxygen of the GDP β-phosphate (Figure 6A,B and Table 2). Proper metal coordination is crucial for tight nucleotide binding, with mutation of magnesium-coordinating residues leading to a preference for GDP over GTP [27,28]. Surprisingly, the H-REV107 peptide in the binding pocket surrounded and bound to KRAS mutations and the phosphate of GDP, and stabilized the KRAS mutants in an irreversible inactive GDP binding state. Specifically, it acted as an inhibitor of specific mutant KRAS targets (G12V, G12D, G12C, G13D, and Q61H). A surface representation of the KRAS–H-REV107 is shown in Figure 6C. The relative distribution of the surface charge is shown with the acidic region in red, the basic region in blue, and the neutral region in white. This peptide contains two negative charges of D67 and D72 and mainly binds to positively charged surfaces.When the structures of our KRAS G12V and KRAS G12V-H-REV107 peptide were superimposed on Cα atoms, the root mean square deviation (RMSD) showed a difference value of 0.503 Å (Figure 6G). Moreover, large root mean square deviations exceeding about 0.8 Å were observed on residues D30, E31, and D33-T35 in the switch I region; residues Q61-M67 and M72-T74 in the switch II region; and the V103-D108 surface loop region near α3 (Table 3). Large conformational changes of KRAS G12V were shown in the binding of H-REV107 peptide to KRAS G12V. Most large deviations were in the E63 and Y64 residues in the switch II region and affected peptide binding near Q61. The Q61 residue is essential for GTP hydrolysis and its mutation blocks Ras-mediated GTP hydrolysis and leads to tumor formation [16,17,18]. The structural differences between the KRAS G12V and KRAS G12V-H-REV107 peptide are mainly in the switch I and II regions. Interestingly, H-REV107 peptide binding can lead to rearrangement of the active sites (the switch I and II regions) of KRAS G12V. In this structure, the solvent accessible surface area of the KRAS G12V was found to be 18,587, while that of the H-REV107 peptide was 1975 Å2. Because the surface area of the KRAS G12V-H-REV107 peptide was 19,219 Å2, we can estimate that a large amount of the surface area (1343 Å2) was buried in the interface during the complex formation. Notably, the catalytic domain of KRAS around G12V on the P-loop region was surrounded by the H-REV107 peptide in the wide binding pocket, and D67 of the H-REV107 peptide penetrated deeply into the hole between the switch I and II pockets. In the H-REV107 peptide complex, the switch I region including P34 and T35 and the switch II region including Q61 were more open when compared to the form in the KRAS G12V MgGDP. Surface representations in the front and top of the KRAS G12V-H-REV107 peptide complex are shown in Figure 6C–F. When the H-REV107 peptide was bound, the switch I and II regions relaxed to a stereogenic form that no longer interacted with the active GTP nucleotide. The peptide wide binding pocket was located within the loop near the β2 (switch I), α2 (switch-II), and α3 helices of KRAS. Large conformational changes were observed in the switch I and II binding pocket regions of the KRAS G12V-H-REV107 complex from the structure of the KRAS G12V protein (Figure S4).The structure of KRAS G12V with the peptide in the binding pocket contained P-loop residues with the highly conserved G-domain sequence GXXXXGKS; residues V29, D30, and D33-T35 of switch I; and residues T58, A59, and Q61 near the DXXG motif (switch II), which is essential for interaction with the phosphate group of the GDP or Mg2+ ion [29]. Residues N116 and D119 of the conserved NKXD motif were also important for binding to the guanosine moiety of the GDP (Figure 6B). The peptide inhibitor bound to MgGDP and stabilized an inactive GDP-bound state. These findings indicate that H-REV10 peptide strongly and irreversibly binds to oncogenic KRAS G12V and fits well inside the binding pockets, leading to the opened rearrangement of its inactive sites (switch I and II) and blockage of the association of KRAS mutants with GTP. Finally, it can successfully inhibit the activity of oncogenic KRAS mutants; however, because it is non-toxic, this peptide leaves the normal protein intact.In our structure, inhibitor H-REV107 peptide was covalently bonded to seven residues (S17, D33, P34, T35, A59, Q61, and N85) of the KRAS. In addition, two residues (G12V and G13) interacted with the H-REV107 peptide. Five residues (G13, V29, D30, N116, and D119) of the KRAS were bonded to GDP, and H-REV107 peptide was also bonded to GDP (Table 2). The structure of the KRAS G12V-H-REV107 peptide was found to be an opened conformation with an inactive GDP-bound state, and large conformational changes were shown at the regions of switch I (T35) and switch II loops (aa 60–74).Characterizing the molecular details describing how H-REV107 binds KRAS G12V is essential to understanding the molecular mechanism through which H-REV107 inhibits oncogenic KRAS mutants. It is known that the H-REV107 protein family can inhibit the RAS signaling pathway, but the molecular mechanism responsible for this is unknown. In this study, we investigated the direct interaction between KRAS and H-REV107 protein/peptide. In surface plasmon resonance (SPR), immobilized KRAS mutants (G12V and G12D) and H-REV107 protein/peptide showed strong binding affinity. Guanine nucleotide binding was regulated by the H-REV107 peptide, and the binding of H-REV107 peptide to KRAS mutants showed decreased GTP binding affinity of KRAS mutants. Notably, both KRAS G12V and G12D were sensitive to H-REV107 peptide, while H-REV107 peptide blocked GTP binding to the KRAS mutants. Treatment with the H-REV107 peptide effectively inhibited pancreatic cancer and colon cancer cell lines in cell proliferation assay by inducing apoptosis. Compared to small inhibitor compounds, the working concentration of cellular activity of peptides is usually higher. We plan to further experiment by modifying the peptide in the future in order to improve the potency, and to include control peptides that have reduced RAS binding. Moreover, we found that the H-REV107 peptide downregulated the phosphorylation of the MEK/ERK signaling pathways. Specially, the H-REV107 peptide suppressed pancreatic tumor growth by reduction of tumor volume and weight in xenotransplantation mouse models.On the basis of the information obtained from these studies, we investigated the structural aspects of KRAS G12V and H-REV107 interaction and determined the crystal structure of H-REV107 peptide-bound KRAS G12V in the MgGDP state at a resolution of 2.3 Å. In the present study, the H-REV107 peptide was strongly docking in close proximity to the switch I and II binding pockets, and near the P-loop of KRAS G12V. Seven residues (L65-D67, G70, and D72-Y74) of the H-REV107 peptide bound to KRAS G12V and four residues (L65 and D72-Y74) of the H-REV107 peptide interacted with KRAS G12V, G13, and Q61. The H-REV107 peptide interacted with the residues on the P-loop (G12, G13, and S17), switch I (D33, P34, and T35), switch II (A59 and Q61) and N85 of KRAS G12V. In addition, hydroxyl groups of residues S17 and T58 interacted with Mg2+ ion and residues G13, V29, and D30 of the P-loop, while the switch I regions interacted with N116 and D119 of the conserved NKXD motif and were found to be important for binding to the guanosine moiety of the GDP (Figure 6B–H).As indicated by the structures of the KRAS G12V and H-REV107 peptide complex, the binding of H-REV107 peptide moves the switch I and II regions even further away and interferes with GTP binding itself. On the basis of these considerations, a function of the H-REV107 peptide is to attenuate KRAS signaling by blocking the GTP binding. The currently available studies suggest that solving the outstanding issues regarding KRAS could lead to development of effective drugs that have a significant impact on cancer treatment [30]. Our data provide detailed information regarding the molecular mechanism responsible for KRAS and H-REV107 interaction that improve our understanding of the biological activity of oncogenic KRAS mutants and may lead to development of a novel KRAS inhibitor. Future research will explore other KRAS mutations and inhibitors of cancer development.His-tagged human KRAS wild-type and mutants (G12V, G12D, G12C, G13D, and Q61H) and H-REV107 were transformed into Escherichia coli BL21 (DE3) cells. Each individual colony was inoculated into 5 L of Luria-Bertani (LB) medium enriched with 10 μg/mL kanamycin, after which the bacteria were grown for 16 h at 37 °C. These cells were then added to 2 L of LB containing antibiotics and grown at 37 °C until the OD600 reached 0.5–0.6. The expression of the proteins was induced by 0.5 mM isopropyl-thio-β-D-1-thiogalactopyranoside (IPTG) at 25 °C, after which the bacterial cells were harvested by centrifugation at 3830 × g for 25 min at 4 °C, then disrupted by sonication in lysis buffer (50 mM HEPES (pH 7.5), 100 mM NaCl, and 2 mM MgCl2). The supernatant was subsequently incubated with Ni-NTA resin (BioRad) for 10 min. After washing, the bound protein was eluted from the beads with elution buffer (50 mM HEPES (pH 7.5), 500 mM NaCl, 2 mM MgCl2, and 400 mM imidazole). The eluted protein was further purified by fast protein liquid chromatography (FPLC) using a Superdex 200 10/300 GL column equilibrated with 50 mM HEPES (pH 7.5), 100 mM NaCl, and 2 mM MgCl2. The purity and identity of the proteins were determined by 15% SDS-PAGE. All mutagenesis experiments were conducted using the QuickChange method.The cDNA fragment for the N-terminal domain of H-REV107 (residues 1–125) was amplified by PCR and cloned into pGEX-4T-1 vector (in frame with N-terminal GST tag). The respective plasmid was then transformed into E. coli BL21 (DE3) cells and grown to an optical density (OD) of 0.5 in LB medium with 50 μg/mL ampicillin. Next, 0.5 mM IPTG was added to the culture and protein was overexpressed for 16 h at 25 °C. Cells were subsequently lysed in 1× PBS buffer (4.3 mM Na2HPO4, 1.47 mM KH2PO4, 137 mM NaCl, and 2.7 mM KCl (pH 7.4)). The clear GST-H-REV107 supernatant was then loaded onto a Glutathione-Sepharose 4 Fast Flow column (GE healthcare) at a flow rate of 0.5 mL/min, after which the column was washed extensively using 30 mL of 1× PBS. The bound proteins were subsequently eluted in buffer (50 mM Tris-HCl (pH 8.0), 200 mM NaCl, and 5 mM glutathione). Finally, gel filtration was performed by FPLC using a Superdex 200 10/300 GL column.KRAS G12V was mixed in a 1:1 molar ratio with purified H-REV107 protein (1–125). The bound mixture was then applied onto a size-exclusion chromatography column TSK-gel-G3000SWXL (Tosoh) connected to a DAWN HELEOS II multi-angle light scattering detector (WYATT Technology). The column was subsequently equilibrated with buffer (50 mM HEPES (pH 7.5), 100 mM NaCl, and 2 mM MgCl2) until the baseline of the MALS detector was stable. The run was performed using 2 mg/mL of sample applied at a flow rate of 0.5 mL/min.The peptide (65LYDVAGSDKY74) of H-REV107 was designed on the basis of interaction modeling between KRAS and H-REV107 proteins. The peptide was produced using Fomc solid-phase peptide synthesis (SPPS) (Peptron Inc., Korea) and purified by reverse-phase high performance liquid chromatography (RP-HPLC) to >95% purity. Finally, the peptide was identified using liquid chromatography/mass spectrometry (LC/MS; HP 1100 Series; Agilent Technology, Santa Clara, CA, USA).Measurement of the apparent dissociation constant (KD) between H-REV107 protein/peptide and KRAS mutants (G12D and G12V) was conducted using a Biacore T100 biosensor (GE Healthcare, Sweden). To accomplish this, each mutant KRAS protein in 10 mM sodium acetate (pH 5.0) was coupled to a CM5 sensor chip (GE Healthcare) at a concentration corresponding to 2300 response units (RU) using an amine coupling method. A flow path including two cells was then used to concurrently measure the kinetic parameters from one flow cell containing the mutant KRAS-immobilized sensor chip to another flow cell containing an underivatized chip. For kinetic measurement at room temperature, H-REV107 protein/peptide mixtures at concentrations ranging from 1.5 to 200 μM were set up by dilution in HBS buffer (150 mM NaCl, 3 mM EDTA, 10 mM HEPES, and 0.005% surfactant P20; pH 7.4). Each sample was subsequently injected into the flow cell at a rate 10 μL/min, after which the immobilized ligand was regenerated by injecting 50 mM NaOH.The dissociation constant and stoichiometry between His-tagged KRAS mutants (G12V, G12D, G12C, G13D, and Q61H) and H-REV107 protein/peptide were determined from auto-isothermal titration calorimetry measurements. The proteins were dialyzed in buffer (50 mM HEPES (pH 7.5), 100 mM NaCl, and 2 mM MgCl2) at a concentration of 0.1 mM. H-REV107 protein/peptide were solubilized in the same buffer at a concentration of 1.0 mM. Titrations measurements that consisted of 20 injections with 200 s spacing were performed at 25 °C while the syringe was stirred at 1000 rpm. The determined K and ΔH values were used to calculate ΔS from the standard thermodynamic equation. Auto-ITC experiments were performed using the MicroCal AutoITC200 (GE Healthcare, Sweden) and the data were analyzed using Origin 7.0 program.Circular dichroism (CD) spectrometry was used to estimate the secondary structure of the proteins [31]. Samples were analyzed using a J-1500 Spectrometer (JASCO Inc, USA) with a 1 mm pathlength cell over the 190–260 nm range (Far UV). The concentrations of recombinant wild-type KRAS or mutant KRAS (G12V, G12D, G12C, G13D, and Q61H) proteins were 0.5 mg/mL in 50 mM HEPES (pH 7.5), 100 mM NaCl, and 2 mM MgCl2. The CD spectra of the native and irradiated proteins were acquired every 0.1 nm with a 1 s averaging time per point and a 1 nm band pass. Each spectrum was obtained as an average of three scans to reduce noise and smoothed before structure analysis was performed. Secondary structure prediction was performed using the JASCO Spectra Manager Version 2 CD Multivariate Secondary Structure Estimation (SSE) program.His-tagged KRAS mutants (G12V, G12D, G12C, G13D, and Q61H) were incubated in H-REV107 protein/peptide diluted into binding buffer (50 mM HEPES (pH 7.5), 100 mM NaCl, 2 mM MgCl2, 1 mM EDTA, and 1 mM DTT) at a 1:1 molar ratio, then applied to Ni-NTA resin and allowed to bind at 4 °C. In addition, the protein complex was incubated with [α-32P] GTP (2500 cpm/pmol) and GTP at 30 °C. To terminate the binding, ice-cold wash buffer (20 mM Tris-HCl (pH 7.4), 100 mM NaCl, and 2 mM MgCl2) was added and elution of the bound protein was achieved using 200 mM imidazole. Protein-bound radioactive nucleotide was quantified by liquid scintillation counting.Cell Counting Kit-8 (Dojindo, Japan) was used to measure the inhibition effect of H-REV107 peptide on the tumor cell proliferation. Cells (2 × 103/well, 100 μL final volume per well) for HCT116 and NCI-H460, 5 × 103 cells/well for SW480 and NCI-H23, and 1 × 104 cells/well for AsPC-1 cell lines were used in this assay (Table 4). All cells were allowed to attach for 24 h after plating, and the cells were treated with fresh media the next day. The cells were then treated with H-REV107 peptide at increasing concentrations for 72 h. After the 72 h incubation, cell counting kit-8 (CCK-8) solution (10 μL/well) was added, and cells were incubated for a further 4 h at 37 °C CO2 incubator. The optical density (OD) of each well at 450 nm was measured by using the Microplate reader (BioTek, USA). Experiments were performed at least three times with representative data presented.Cells were plated in a 100 mm cell culture dish and incubated under standard conditions (37 °C under a humidified atmosphere containing 5% CO2) in RPMI-1640 media with 10% fetal bovine serum, 100 μ/mL penicillin, and 100 μ/mL streptomycin (Welgene, Korea). After 24 h, the media was removed and replaced with fresh media (containing 0.1% DMSO) and then H-REV107 peptide was added to each dish. The treated H-REV107 peptide concentration was determined according to cell proliferation assay and GI50. After treatment of 24, 48, and 72 h, the cells were lysed using protein extraction solution PRO-PREPTM (iNtRON Biotechnology, Korea) containing a protease inhibitor cocktail for 30 min at 4 °C. After this, the cells were centrifuged at 15,115 x g for 30 min. Protein content in the supernatant was measured with Bradford protein assay using bovine serum albumin (BSA). The protein was separated by 10% SDS-PAGE and transferred onto an immobilon-P membrane (Millipore, USA). Membrane was blocked in ProNA Protein-free 5X general or phospho-block solution (TransLab, Korea) for 1 h at room temperature. The membrane was incubated with primary antibodies against phosphorylated (p)-MEK1/2 (Ser217/221), total (t)-MEK1/2, p-ERK1/2 (Thr202/Tyr204), and t-ERK1/2 (Cell Signaling Technology, USA) at 4 °C overnight, washed with 1X PBS containing 0.1% Tween (PBST) three times, and incubated with anti-IgG secondary antibodies (Santa Cruz Biotechnology, USA) in PBST for 1 h. The transferred protein band was visualized with EzWestLumi plus (ATTO corporation, Japan). Probing for ß-acitn was used as a loading control.A total of 25 male five-week-old nude mice (weight: 14–17g) were purchased from Orient Bio Inc. The symptoms were observed once a day during the 7 day period. The weight was measured at the end of the purification period and the general symptoms were observed.The cells used in the cancer animal model used human-specific pancreatic cancer cell liquor AsPC-1, which was defrosted more than 2 weeks before the animal model was created to obtain stable cell alcohol conditions, and were incubated three times. The growth medium used for cell culture was used by adding 10% fetal bovine serum and 1% penicillin/streptomycin to the RPMI-1640 medium used for AsPC-1 cell lines. On the day of the production of the cancer animal model, AsPC-1 cell line was removed from the cell culture flask by 0.25% Trypsin/EDTA and washed twice with PBS. The cell line was finally harried to HBSS + matrigel (1:1). Since the number of cells to be used in the experiment was 1 × 106/head, the finally disheveled cell concentration was 1 × 107/mL. It was stored in ice until the test was used. In the case of cell transplantation, 100 μL was transplanted. The separation was performed when the average volume of the tumor had grown to more than 100 mm3. A random separation of groups was conducted to ensure that the average tumor volume was equal in the day of separation, and the separation of groups was carried out into a total of three groups.All animals were weighed and recorded once a week after transplanting cancer cells (weight unit: g). After transplanting cancer cell lines, the tumor size was measured twice a week on the basis of the start of administration. The tumor volume calculation method measured the longitude (L, length) and short diameter (W, width) and then applied the equation (L × W2)/2. All experimental procedures followed the Guidelines for the Care and Use of Laboratory Animals of the National Institutes of Health of Korea (Law No. 4379 on 31 May, 1991, Partial Amendment on 20 January, 2015, No. 12053), and were approved by the Institutional Animal Care and Use Committee (IACUC) of the Daegu Gyeongbuk Advanced Medical Industry Promotion Foundation, Republic of Korea (approval number: DGMIF18041702-01).Crystals of the KRAS G12V were grown by hanging drop vapor diffusion at 20 °C for 2 weeks with polyethylene glycol 3350, 0.2 M potassium nitrate (pH 6.8), and acetonitrile in the well. The H-REV107 crystal was grown in a reservoir consisting of polyethylene glycol 3350 and 0.2 M potassium nitrate at pH 6.8. The crystal of the KRAS G12V-H-REV107 peptide complex was grown by soaking with polyethylene glycol 3350 and 0.2 M potassium nitrate (pH 6.8) and acetonitrile at 20 °C [32]. X-ray diffraction data of the KRAS G12V, H-REV107, or KRAS G12V-H-REV107 peptide complex were collected on the Pohang Light Source (PLS), beam-line 7A, Republic of Korea. The crystal was soaked in cryo-protectant solution containing additional glycerol and flash-frozen in liquid nitrogen for data collection under 100 K. Diffraction data were processed with the HKL-2000 software [33] and the structure was solved by molecular replacement with CCP4 [34]. The final model was produced by rounds of building in COOT [35], followed by refinement using PHENIX [36]. All structure figures were generated using the PyMOL program (http://pymol.org/2/). Final statistics of the collected data and refinement of the structures are shown in Table 1.We demonstrated that H-REV107 peptide can suppress the KRAS activation function through the blocking of GTP binding to KRAS mutants. These results suggest that the blocking of GTP-binding by H-REV107 peptide could inhibit tumor cell proliferation through the downregulation of the KRAS pathway. Consequently, the reduction of GTP binding affinity of KRAS mutants by H-REV107 peptide in cancer cells influenced cell proliferation by inhibiting the RAS signaling pathway and inducing apoptosis. Our data will facilitate the development of novel drugs for inhibition of KRAS mutations.The following are available online at https://www.mdpi.com/2072-6694/12/6/1412/s1, Figure S1: (A) The SDS-PAGE gel bands of the purified KRAS and H-REV107 proteins are shown. Lane 1, protein marker; lane 2, wild-type KRAS; lane 3, H-REV107, lane 4, KRAS G12V; lane 5, KRAS G12D; lane 6, KRAS G12C; lane 7, KRAS G13D, and lane 8, KRAS Q61H. (B) Circular dichroism (CD) of wild-type and mutant KRAS was used to investigate the stability of the secondary structure of KRAS in different types of point mutations. The CD spectra of the KRAS mutants showed that each mutation affected the conformation of KRAS to a different extent. The CD spectra were measured from 260 to 190 nm using a 0.05 pathlength cell, and CD signals were merged to CDNN. (C) The sequences and secondary structures of KRAS and H-REV107 are shown. The α-helices are shown as blue ellipses, β-sheets as orange arrows, and linker loops as gray lines. The G12V mutation of KRAS is shown as a red triangle. The binding residues of the KRAS to H-REV107 peptide are indicated with stars. Every 10 residues is indicated by a point, Figure S2. (A,B) SEC-MALS spectra of the KRAS and H-REV107 proteins are shown. The inset shows the value of the molecular weight of KRAS and H-REV107 determined from the MALS data analysis (black line: MALS, dashed line: molecular weight). (C-E) The crystals of the KRAS G12V, H-REV107, and KRAS G12V-H-REV107 peptide complex are shown. (F) The molecular packing of the KRAS G12V-H-REV107 peptide complex is shown. H-REV107 peptide was bound to one molecule per four molecules of the KRAS G12V per asymmetry unit, Figure S3. Isothermal titration calorimetry (ITC) analysis of the KRAS mutant and H-REV107 peptide/protein. The H-REV107 peptides or H-REV107 protein were titrated into the KRAS mutant solutions, and the measured KD values are shown. (A) KRAS G12V with H-REV107 protein, (B) KRAS G13D with H-REV107 peptide, (C) KRAS Q61H and H-REV107 peptide, (D) KRAS G12C with H-REV107 peptide, Figure S4. (A) The Cα traces of the KRAS G12V (orange) and KRAS G12V-H-REV107 peptide (green) are superimposed and shown in stereoview. The switch I and II regions of the KRAS G12V-H-REV107 peptide were more opened than those of the KRAS G12V with MgGDP. (B-D) Surface representations of the front and top sides of the KRAS G12V-H-REV107 peptide complex are shown. (E) Surface representation of the KRAS G12V with MgGDP is shown.Performed biological experiments, made the crystals, and collected diffraction data C.W.H., Solved the crystal structures C.W.H. and S.B.J., Helped diffraction data collection S.C.H., Designed the research and analyzed the structural data S.B.J., Wrote the manuscript and participated in analysis of biochemical data M.S.J., C.W.H., M.S.J., and S.B.J. All authors have read and agreed to the published version of the manuscript. Diffraction data were collected at the Pohang Light Source (PLS-7A), Republic of Korea. This study was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (2018R1D1A1B07043701) to S.B.J. and (2016R1D1A1B02011142) to M.S.J.The authors declare no conflict of interests.Schematic representation and biophysical properties of KRAS and H-REV107. (A) Domain structures of full-length KRAS and H-REV107 are shown. (B) Cartoon model of KRAS in closed (left: guanosine triphosphate (GTP)-bound), opened (middle: guanosine diphosphate (GDP)-bound), and further opened (right: GDP- and peptide-bound) forms is shown. (C) Size-exclusion chromatography of KRAS G12V and H-REV107 proteins. The interaction band (red) of KRAS G12V (green) with H-REV107 (blue) after application of size exclusion chromatography from the Superdex 200 column was analyzed by SDS-PAGE. (D) Size-exclusion chromatography–multi-angle light scattering (SEC-MALS) spectrum of the KRAS and H-REV107 complex. The inset shows the value of the molecular weight of KRAS and H-REV107 determined from the MALS data analysis (black line: MALS, dashed line: molecular weight). (E,F) Biacore biosensor analysis of KRAS mutant (G12V and G12D) binding to H-REV107 peptide at 25 °C. H-REV107 peptide sensorgrams for 25, 50, and 100 µM are shown.Inhibitor assay of KRAS mutants and H-REV107 protein/peptide. (A) The binding of GTP to KRAS mutant (G12V) in the presence of H-REV107 protein was determined using α32-labeled GTP. (B–G) The binding of GTP to KRAS mutants (G12V, G12D, G12C, G13D, and Q61H) or wild-type in the presence of H-REV107 peptide was determined using α32-labeled GTP. The non-bound 32α-GTP was washed out and the radioactivity was examined using a scintillation counter. Non-labeled GTP was used as a competitor. Data are presented as mean ± SD. Statistic tests were performed using a Student’s t-test. * p < 0.05, ** p < 0.01, and *** p < 0.001 compared to the control under α32-labeled GTP (second column).Cell proliferation assay and Western blot analysis. H-REV107 peptide was incubated for 72 h in five different cell lines, the (A) SW480, (B) AsPC-1, (C) NCI-H23, (D) NCI-H460, and (E) HCT116. Ponatinib was used for reference in five cell lines. Ponatinib to inhibit RAS-RAF-MAPK-ERK pathway was used to study cell proliferation in the KRAS mutant cancer cells for reference. The H-REV107 peptide was tested in triplicate and the data represent mean ± SD. (F,G) Western blot analysis of phosphorylated (p)-MEK1/2, total (t)-MEK1/2, p-ERK1/2, and t-ERK is shown in SW480 and AsPC-1 cells incubated with H-REV107 peptide for 24 and 48 h. Equal loading is shown by ß-actin.Suppression of tumor growth by H-REV107 peptide in mice inoculated with pancreatic AsPC-1 cancer cells. AsPC-1 cells were inoculated on the skin in a xenograft mouse model and injected the H-REV107 peptide into the abdominal cavity on a daily basis. (A) Photograph of tumor treated with the indicated dose of H-REV107 is shown. (B–C) Volume and weight of tumor were measured. Data are presented as means ± SD. Statistic tests were performed using Student’s t-test and Dunnett’s test. The minimum level of statistical significance was set at a p-value of 0.05 for all the analyses. (D,E) Body weight of each mouse was measured daily during injection of H-REV107 peptide.Crystal structures of the KRAS G12V, H-REV107, and KRAS G12V-H-REV107 peptide complex. (A) Ribbon representation of monomeric crystal structure of KRAS G12V with MgGDP (orange) is shown. Mg2+ ion is indicated by orange circle and the GDP is shown as yellow rod. MgGDP molecule is color-coded with C in gray, O in red, N in blue, P in purple, and Mg2+ ion in orange. (B) The crystal structure of the H-REV107 protein is shown in yellow and the fraction of the H-REV107 peptide is shown in purple. (C) The 2Fo-Fc electron density map of H-REV107 peptide was calculated at 2.3 Å resolution and contoured at 1σ (gray) at the peptide site. (D,E) Crystal structure of KRAS G12V-H-REV107 peptide complex is shown in green. Switch I and II regions in orange and P-loop in blue and peptide in purple are shown. (F) Electron density (2Fo-Fc) for bound MgGDP is contoured at 1σ (gray). (G,H) The binding regions of the KRAS and H-REV107 peptide are zoomed in on, and the hydrogen bonds in KRAS G12V and H-REV107 peptide complex are indicated by the black dashed lines.Detailed binding and structural characterization of KRAS G12V and H-REV107 peptide. (A) Binding region of the KRAS G12V and H-REV107 peptide is shown. The KRAS G12V, G13, and Q61 residues are coordinated and stabilized by L65, D72, K73, Y74, and MgGDP. The KRAS residues and H-REV107 peptide on the hydrogen-bonding network are labeled. (B) Interaction of the MgGDP and KRAS G12V-H-REV107 peptide is shown. MgGDP bound to residues of the P-loop, and switch I and II regions and NKXD motif of the KRAS G12V. (C) The relative distribution of the surface charge of the KRAS-H-REV107 is shown with the acidic region in red, the basic region in blue, and the neutral region in white. (D,E) Surface representations of the KRAS G12V-H-REV107 peptide complex are shown from the front and top sides. G12V (red), G13 and Q61 (yellow), switch I (green) and II (orange) regions, P-loop (blue), and peptide (purple) are shown. Peptide D67 deeply penetrated into the hole between the switch I and II pockets. (F) Surface representation of the KRAS G12V·MgGDP is shown. (G,H) Superposition of the Cα chain traces of the KRAS G12V (orange) and KRAS G12V-H-REV107 peptide (green) is shown. The boxed regions in the left panel are enlarged in the right panels, where the large conformational changes are drawn as arrow representations. The P-loop, and switch I and II regions of the KRAS G12V-H-REV107 peptide complex are more opened than those of KRAS G12V. Large conformational changes were observed in the switch I and II binding pocket regions of the KRAS G12V-H-REV107 complex from the structure of the KRAS G12V protein.Crystallographic statistics.Values in parentheses are for the highest resolution shell. a Rmerge = ∑|Ii Im|/∑Ii, where Ii is the intensity of the measured reflection and Im is the mean value of all symmetry-related reflections. b Rcryst = Σ||Fobs| − |Fcalc||/Σ|Fobs|, where Fobs and Fcalc denotes the observed and calculated structure factor amplitude. Rfree = ∑T||Fobs| − |Fcalc||/ΣT|Fobs|, where T is a test data set of about 5% of the total reflections randomly chosen and set aside prior to refinement.Interaction distances between KRAS G12V and H-REV107 peptide.Root mean square deviation (RMSD) values between KRAS G12V and KRAS G12V-H-REV107 peptide.Characteristics of KRAS mutant cell lines.

Med-MDPI/cancers/cancers-12-06-01413.txt

@@ -0,0 +1 @@
+Patients with papillary thyroid carcinoma (PTC) have excellent survival, but recurrence remains a major problem in the management of PTC. We aimed to determine the prognostic impact of the expression of CD10 and CD15 in patients with PTC. Immunohistochemistry for CD10 and CD15 was performed on the tissue microarrays of 515 patients with PTC. The expression of CD10 and CD15 was detected in 201 (39.0%) and 295 (57.3%) of 515 PTC cases, respectively, but not in the adjacent benign thyroid tissue. Recurrence was inversely correlated with CD15 expression (p = 0.034) but not with CD10 expression. In 467 PTC patients treated with radioiodine remnant ablation, the CD15 expression had an adjusted hazard ratio of 0.500 (p = 0.024) for recurrence-free survival and an adjusted odds ratio of 2.678 (p = 0.015) for predicting long-term excellent therapeutic response. CD10 expression was not associated with clinical outcomes. In the Cancer Genome Atlas dataset, the expression level of FUT4 (CD15) mRNA was higher in the low/intermediate-risk group for recurrence than in the high-risk group and exhibited positive correlation with SLC5A5 (NIS) mRNA expression (p = 0.003). Taken together, CD15 expression was identified as an independent prognostic marker for improved prognosis in PTC patients.Papillary thyroid carcinoma (PTC) is the most common histologic type of thyroid cancer, accounting for 86% in the USA and 93% in Korea and Japan [1,2,3]. Most of the patients with PTC are not likely to die of this disease. The 5-year survival rate has been reported as 97.9% in the USA and 100% in Korea [1,4]. However, up to 20% of patients with PTC have a locoregional recurrence or distant metastasis following thyroidectomy [5]. The 2015 American Thyroid Association (ATA) guidelines classify patients with differentiated thyroid cancer into low, intermediate, and high-risk groups for recurrence after initial complete therapy based on the specific histology, tumor size, tumor encapsulation, multifocality, the extent of extrathyroidal extension, number of vascular invasion foci, number and size of the metastatic lymph nodes, and mutational status of BRAFV600E and/or TERT promoter when available [6]. However, the ATA three-tiered system is unable to accurately predict the likelihood of recurrence in an individual patient. Increasing knowledge of tumor biomarkers has led to several efforts to more accurately predict the recurrence risk for thyroid cancer patients.The cluster of differentiation (CD) antigens has gathered increased recognition not only as diagnostic biomarkers but also as a target for molecular therapy. A wide range of CD markers was initially described in various immune cells and hematological malignancies [7]. Later on, a spectrum of tissues expressing CD antigens was expanded to various solid tumors. We recently described the aberrant expression of CD20 in thyroid cancer, particularly the aggressive types, which may offer promise for translational implications [8]. This finding prompted us to further explore CD markers in thyroid tumors.CD10, also known as membrane metalloendopeptidase (MME), neprilysin, common acute lymphoblastic leukemia antigen (CALLA), neutral endopeptidase (NEP), enkephalinase, or EC 3.4.24.11, is a membrane-bound zinc metalloproteinase [9,10]. Immunohistochemistry for CD10 has been used for diagnostic purposes in lymphomas, leukemias, and solid tumors including clear cell renal cell carcinoma, solid and pseudopapillary neoplasm of the pancreas, skin tumors, urothelial tumors, endometrial stromal tumors, and mesonephric tumors [10]. Conflicting results have been reported for the association of altered expression of CD10 with adverse prognosis in various solid tumors including lung cancer, prostate cancer, head and neck cancer, colorectal cancer, melanoma, and ovarian cancer [9,10]. Several previous studies also reported selective expression of CD10 in differentiated thyroid cancers and not in normal thyroid tissues and benign thyroid nodules [11,12,13]. It has recently been reported that CD10 is highly expressed in anaplastic thyroid carcinoma, but has low expression in PTC and is absent in follicular thyroid carcinoma and medullary thyroid carcinoma [14]. However, these studies were performed on a relatively small number of cases. Little is known about the diagnostic and prognostic values of CD10 expression in PTC.CD15, also known as fucosyltransferase 4 (FUT4), LeuM1, Lewis X, or stage-specific embryonic antigen 1 (SSEA-1), is a carbohydrate antigen with the common trisaccharide structure 3-fucosyl-N-acetyl-lactosamine [15]. CD15 is a marker for human granulocytes and Reed–Sternberg cells of Hodgkin’s lymphoma [16]. CD15 expression has also been found in various solid cancers including lung cancer [17], breast cancer [18], colorectal cancer [19,20], ovarian cancer [21], and renal cell carcinoma [22,23]. In the thyroid gland, CD15 expression was reported in various types of thyroid cancers, but not in normal thyroid tissue or benign thyroid neoplasms [24]. CD15 was suggested as a marker for thyroid cancer-initiating cells or cancer stem cells [24,25,26].This study aims to investigate the prognostic potential of the expression of CD10 and CD15 by immunohistochemistry in a large series of PTC. Furthermore, the correlation between mRNA expression of both the markers and clinicopathologic features using The Cancer Genome Atlas (TCGA) data was examined, and in silico gene expression analyses of the related genes were performed.The clinicopathologic characteristics of patients and the association between clinicopathologic parameters and CD10 and CD15 expression are summarized in Table 1. A total of 515 PTC cases were studied consisting of 385 classic PTCs, 68 classic PTCs with tall cell features, 17 follicular variant PTCs (including five invasive encapsulated follicular variants and 12 infiltrative follicular variants), 21 tall cell variant PTCs, and 24 PTCs of other variants (11 oncocytic, eight Warthin-like, two hobnail, one solid, one diffuse sclerosing, and one cribriform-morular).Immunohistochemical staining for CD10 showed membranous staining at the apical surface of tumor cells (Figure 1). The staining pattern for CD15 was both cytoplasmic and membranous in tumor cells (Figure 1). Normal thyroid tissue adjacent to tumors was negative for both CD10 and CD15 staining. Positive staining of CD10 and CD15 was found in 201 (39.0%) and 295 (57.3%) of 515 PTC cases, respectively.The relationship between clinicopathological characteristics and the expression of CD10 and CD15 is shown in Table 1. The expression rates of CD10 and CD15 were higher in female patients (p = 0.049 and p = 0.009, respectively). Tall cell variant PTC had higher positive rates of CD10 and CD15 than classic PTC and follicular variant PTC. Compared with patients without extrathyroidal extension, patients with minimal (microscopic) extrathyroidal extension had higher positive rates of CD10 and CD15 (p = 0.003 and p = 0.014, respectively). However, the expression of CD10 and CD15 was not associated with gross extrathyroidal extension, pathologic T stage, or lymph node metastasis. BRAFV600E mutation was significantly associated with the expression of CD10 and CD15 (p = 0.011 and p < 0.001, respectively). The rate of structural recurrence was lower in patients with CD15 expression than those without CD15 expression (p = 0.034) but was not associated with CD10 expression (p = 0.656). Distant metastases developed in 11 patients as synchronous (n = 3) or metachronous (n = 8) lesions. Although not statistically significantly different, CD15 expression was less frequently found in patients with distant metastasis (p = 0.062).For the analysis of recurrence-free survival, nine patients with initial distant metastasis (n = 3) or less than six months of follow-up data (n = 6) were excluded from the study. The median follow-up time was 112 months (range 6–143 months). During the follow-up, structural recurrence occurred in 44 (8.7%) of 506 patients. The median time to the first recurrence was 15 months (range, 6–23 months). The recurrence occurred in 43 (9.2%) of 467 patients who underwent surgery and radioactive iodine (RAI) remnant ablation whereas the incidence was noted in one of 48 patients who were not indicated for the RAI therapy. Therefore, the analysis of recurrence-free survival was performed in 467 patients who underwent total thyroidectomy and RAI remnant ablation.On univariate survival analysis, pT3-4 stage (p = 0.022) and lymph node metastasis (p < 0.001) were significantly associated with decreased recurrence-free survival. The expression of CD15 was associated with increased recurrence-free survival (p = 0.037) (Figure 2). In the stratified survival analyses, the expression of CD15 was also a predictor for the increased recurrence-free survival in subgroups of PTC patients with extrathyroidal extension (n = 357, p = 0.005) and lymph node metastasis (n = 306, p = 0.023) (Figure 2). However, CD10 expression was not correlated with recurrence-free survival (p = 0.920).Multivariate analysis revealed an independent correlation between lymph node metastasis and poor recurrence-free survival (p = 0.001). Patients with CD15 expression had an adjusted hazard ratio of 0.500 (95% confidence interval (CI): 0.274–0.911, p = 0.024) (Table 2).The clinical outcomes in 467 patients who underwent total thyroidectomy and RAI remnant ablation were further analyzed. At the time of the last follow-up, 438 (93.8%) patients achieved an excellent response and were considered to have no clinical evidence of disease (NED). All patients without lymph node metastasis achieved an excellent response.On univariate logistic regression analysis, NED was significantly associated with lower pT stage (p = 0.006) and the presence of CD15 expression (p = 0.029) (Table 3). On multivariate logistic regression analysis, patients with CD15 expression had an adjusted odds ratio of 2.678 (95% CI: 1.215–5.902, p = 0.015) for predicting NED compared to those without CD15 expression (Table 3).A high level of MME (CD10) mRNA expression was associated with minimal extrathyroidal extension (p < 0.001), pathologic T stage (p = 0.006), lymph node metastasis (p < 0.001), and advanced cancer stage (p = 0.007). A high level of FUT4 (CD15) mRNA expression was associated with younger age (<45) (p = 0.001) and histologic variant (p = 0.031), as shown in Table 4. PTCs with a BRAF-like molecular phenotype had higher levels of MME and FUT mRNA than did RAS-like tumors (p < 0.001 and p = 0.002, respectively). According to the ATA recurrence risk stratification, the expression level of MME mRNA expression was higher in the intermediate/high-risk group than in the low-risk group (p < 0.001) whereas the expression level of FUT4 mRNA expression was higher in the low/intermediate-risk group than in the high-risk group (p = 0.023).Since CD15 showed clinical and prognostic significance in PTC based on protein and mRNA expression in the two independent datasets, as described above, we further addressed the possible relationship between the FUT4 (CD15) mRNA expression and tumor microenvironment in PTC. The gene expression profiles of 505 primary PTCs were obtained from the TCGA project (Figure 3). Initially, the level of infiltration for immune and stromal cells was measured by ESTIMATE (Estimation of STromal and Immune cells in MAlignant Tumor tissues using Expression data) algorithm [29]. A substantial level of correlation with FUT4 mRNA expression was observed, i.e., Spearman correlation of 0.532 and 0.578 for immune and stromal cells, respectively (Figure 3A). These findings suggest that FUT4 expressing PTCs are highly infiltrated with immune and stromal cells.To further examine the immune cell composition concerning FUT4 mRNA expression, the CIBERSORT algorithm was employed [30]. The relative abundance of CD4 memory T cells, B cells, and dendritic cells was positively correlated with FUT4 mRNA expression while that of neutrophils, plasma cells, and NK (Natural Killer) cells showed inverse correlation (Figure 3B). Of note, distinct macrophage polarization with respect to FUT4 mRNA expression was observed; macrophages M1 and M2 showed a positive and inverse correlation with FUT4 mRNA expression, respectively, suggesting that high FUT4 mRNA expression favors the infiltration of antitumorigenic M1 macrophages in PTC microenvironments.The mRNA expression levels of three immune checkpoint markers (PD-1, PD-L1, and CTLA-4) were further examined with the cytolytic score as the geometric mean of GZMA and PRF1 mRNA expression (Figure 3C); all the parameters demonstrated a substantial level of correlation with FUT4 mRNA expression. This is suggestive of the high level of infiltration of cytolytic T lymphocytes (cytolytic score) and their exhaustive states with a high level of expression of immune checkpoints. Gene set enrichment analysis of 50 hallmark gene sets revealed a positive correlation between immune-related gene set scores and FUT4 mRNA expression, consistent with ESTIMATE results. In addition, a positive correlation between inflammatory response and expression of epithelial-mesenchymal transition (EMT) associated genes, explaining the high level of infiltration of stromal cells with high FUT4 mRNA expression in PTC, was observed (Figure 3D). Consistently, transcription factors known to promote EMT such as TWIST and SLUG showed a high level of correlation with FUT4 mRNA expression (Figure 3E). These findings suggest that FUT4 mRNA expression in PTC might be associated with distinct immune and stromal composition of the tumor microenvironment in PTC.From the above results, initially it was hypothesized that CD15 overexpression might be associated with poor prognosis in PTC patients because it was associated with some unfavorable clinicopathological variables (aggressive histology, minimal extrathyroidal extension, and BRAFV600E) in our study cohort. However, patients with CD15 expression were more responsive to RAI therapy and had better recurrence-free survival compared to those without CD15 expression. To address the contradictory results and explain the high therapeutic response rate in patients with CD15 expression, the correlation between expression levels of FUT4 and SLC5A5 mRNA in the TCGA cohort was analyzed. Recently, it was demonstrated that the SLC5A5 mRNA expression is more reliable than immunohistochemical expression of sodium iodide symporter (NIS) coded by SLC5A5 gene regarding tumor behavior, therapeutic response, and prognostic outcomes [31]. Herein, a positive correlation between the two variables (Spearman Rho = 0.141, p = 0.003) was observed, as shown in Figure 4.In this study, the prognostic implication of CD10 and CD15 expression was investigated in PTC patients. The patients with CD15 expression had a good recurrence-free survival at about the 10 years median follow-up and excellent therapeutic outcomes after total thyroidectomy and RAI remnant ablation for PTC. However, CD10 expression was not identified to be associated with clinical outcomes.Overexpression of the FUT4 gene has been reported in different malignancies; however, the exact role of FUT4 expression in carcinogenesis still remains unclear. Many previous studies reported that FUT4 gene expression is associated with pro-tumorigenic function, such as tumor growth and invasion, and drug resistance [32]. In contrast, FUT4 overexpression in lung cancer cells has been reported to suppress the EGFR signaling pathway and attenuate EGFR-mediated invasion of tumor cells [33].In our study, we found that high expression of FUT4 mRNA in PTC from the TCGA database was associated with a high level of tumor infiltration both for immune and stromal cells as exhibited by ESTIMATE analysis. FUT4-overexpressing PTCs showed transcriptional up-regulation of EMT-associated genes (ZEB2, SLUG, TWIST, ZEB1, and SNAIL). Therefore, FUT4-overexpressing PTCs may have activated EMT during their progression and are also highly infiltrated with stromal cells. However, recent studies suggest that transcriptional signals of mesenchymal tumor subtypes largely come from stromal cells instead of tumor cells, and EMT-like tumor signatures may be largely determined by tumor purity [27,34]. Therefore, it will require further investigation to see whether the FUT4-overexpressing PTCs activate EMT or represent those with a high level of tumor-infiltrating stromal cells and low tumor purity.In terms of immune cells, we observed that FUT4-overexpressing PTCs had a high level of M1 macrophages infiltration and depleted M2 macrophages. The polarization of M1 and M2 macrophages with immunostimulatory and immunomodulatory phenotypes, respectively, define key components of tumor microenvironments [35]. Therefore, the correlation of FUT4 expression with M1 and M2 macrophages is suggestive of the immune microenvironment exhibiting antitumor activities associated with FUT4 expression. The cytolytic scores estimated from the transcript level of two key cytolytic effectors of GZMA and PRF1 [36] also showed the correlation with FUT4 expression, suggesting that the FUT4-overexpressing PTCs are likely to be infiltrated with cytotoxic T cells. In contrast, FUT4-overexpressing PTCs were associated with up-regulation of CTLA-4, PD-1, and PD-L1 in the tumor microenvironment that is known to exert immunosuppressive and pro-tumorigenic activities [37]. CTLA-4 and PD-1 negatively regulate T-cell activation and alter the motility and migration of T-cells [37]. Therefore, FUT4-overexpressing PTCs could possess the immune signatures indicating T cell exhaustion (i.e., high expression level of immune checkpoints) as well as activated T cell effectors (i.e., high level of cytolytic scores).However, despite the potential cancer-promoting effects in PTC, the CD15 expression was found to be an independent prognostic predictor for reduced disease recurrence and excellent therapeutic response in our original institutional cohort. As most of the PTC patients at our hospital underwent total thyroidectomy and RAI remnant ablation in accordance with a standard protocol during the study period, we could not analyze the long-term prognosis and therapeutic response rate according to the CD15 expression status in a subgroup of PTC patients without RAI therapy. To address the question about excellent therapeutic response in patients with CD15 expression, the gene expression patterns of CD15 and NIS were compared using the TCGA data, and positive correlation was observed between expression levels of both the genes. Therefore, it is hypothesized that good response to RAI therapy in PTC patients with CD15 expression might be associated with increased expression of SLC5A5 mRNA.The prognostic value of positive expression of CD15 for the clinical outcome in thyroid cancer patients remains controversial. In one previous study based on tissue microarray, CD15 immunostaining was usually focal in PTC and follicular thyroid carcinoma but diffuse in anaplastic thyroid carcinoma [24]. CD15 expression was associated with worse survival in anaplastic thyroid carcinoma, but not in PTC or follicular thyroid carcinoma [24]. In another report, CD15 expression was associated with recurrence or metastasis and shorter progression-free survival [25]. However, limitations of both these studies included a relatively small number of cases, selection bias, short-term follow-up, and different treatment strategies and follow-up methods. In our study, CD15 expression was associated with minimal extrathyroidal extension and BRAFV600E mutation, but not with gross extrathyroidal extension and lymph node metastasis, which are more important parameters for prognosis in PTC [38].CD15 expression was independently associated with improved recurrence-free survival and excellent therapeutic response when we analyzed PTC patients who underwent total thyroidectomy with RAI remnant ablation and thyroid-stimulating hormone (TSH) suppression. In the TCGA dataset, the expression levels of FUT4 (CD15) mRNA were higher in the low (47.0%) and intermediate (53.2%) recurrence risk groups compared to that in the high-risk group (25%). These results therefore indicate that high expression of CD15 protein and mRNA may represent prognostic markers that predict favorable clinical outcomes of patients after the treatment of PTC. A better understanding of the implications of CD15 expression may be crucial for developing biomarkers for monitoring and treating PTC patients.Previous studies have reported different prognostic values of CD15 expression in various solid tumors besides thyroid cancers. In Chinese patients with clear cell renal cell carcinoma, CD15 expression was associated with improved overall survival after surgical treatment [23]. In Japanese patients who underwent radical nephrectomy for renal cell carcinoma, high CD15 expression was associated with recurrence and shorter metastasis-free survival [22]. In metastatic colorectal cancer patients treated with cetuximab or bevacizumab plus chemotherapy, CD15 expression in cancer cells was associated with worse progression-free survival and overall survival [19]. Therefore, it is conjectured that CD15 may play various roles in solid cancer development and progression depending on the type of cancer.Until now, only limited data were available on the role of CD10 in thyroid carcinoma. Most of the previous studies focused on the diagnostic significance of CD10, for example, they were aimed at differentiatimg thyroid cancers from benign thyroid lesions [11,12,13,14]. The positivity of CD10 has been reported to be significantly higher in PTC than in benign thyroid tumors, but lower than in anaplastic thyroid carcinoma [11,12,14]. Anaplastic thyroid carcinoma shows diffuse positivity for CD10, whereas most of the cases of PTC had focal expression of CD10 [13,14]. Anaplastic thyroid carcinoma can develop from differentiated thyroid carcinoma through the stepwise process of dedifferentiation and can involve accumulation of genetic abnormalities [39]. PTC is the most common type of differentiated thyroid carcinoma in the case of anaplastic thyroid carcinomas comprising both the tumor components [40]. In the present study, we observed the focal staining pattern for CD10 in PTC cells. The expression of CD10 was significantly associated with minimal extrathyroidal extension (but not with gross extrathyroidal extension) and BRAFV600E mutation. TCGA data showed similar results with high expression of MME (CD10) mRNA being associated with minimal extrathyroidal extension and a pathologic T stage as classified by the American Joint Committee (AJCC) 7th edition and BRAF-like molecular signature. However, analysis of long-term follow-up in our study cohort showed that the expression of CD10 was not associated with tumor recurrence and clinical outcome of PTC patients. BRAFV600E mutation is an early event in the development of PTC [28,41,42]. Apparently, these findings indicated that CD10 overexpression might be a relatively early event in the development and progression of PTC. As CD10 expression has been reported to be associated with cancer progression and chemotherapy susceptibility in other cancers [9], our findings still merit further experimental investigations.A total of 515 patients who underwent surgery for PTC with tumor size ≥1 cm at Seoul St. Mary’s Hospital from 2008 to 2010 were enrolled in this study. The institutional review board approved this study (KC16SISI0104 and KC16SISI0709) and all the patients provided informed consent. Clinical information was obtained from the medical records. All hematoxylin and eosin slides of thyroidectomy specimens were reviewed by an endocrine pathologist (Chan Kwon Jung). In cases of multiple tumor foci, the largest one was considered as the index lesion. All cases were classified according to the World Health Organization criteria [1]. TNM (Tumor-Node-Metastasis) staging was done according to the 8th edition of the AJCC cancer staging manual [38]. The same data set was used in our previous studies [8,43,44], but clinical follow-up data were updated as of March 2020. All patients with total thyroidectomy took levothyroxine for suppression of thyroid-stimulating hormone (TSH). A total of 471 patients with PTC underwent radioactive iodine (RAI) remnant ablation after total thyroidectomy.The therapeutic response was assessed by serum thyroglobulin (Tg), stimulated Tg, and anti-Tg antibody levels, imaging studies such as neck ultrasonography, and diagnostic RAI whole-body scan, or cytologic/histologic examination [6,45]. Excellent response to therapy was defined according to the 2015 ATA guidelines [6] and was based on a combination of the following characteristics: non-stimulated Tg level <0.2 ng/mL, stimulated Tg level <1 ng/mL, undetectable anti-Tg antibody, and negative imaging.Patients who demonstrated an excellent response to therapy at the time of the last follow-up were considered NED at the final follow-up. A structural recurrence was defined as structural evidence of disease on imaging or biopsy-proven disease that was detected following any period of NED. Only biochemical evidence was not considered as recurrence in this study.Tissue microarrays were constructed from formalin-fixed, paraffin-embedded (FFPE) tissue blocks. The tissue microarray design was a 5 × 10 subarray with 2-mm cores at the 1-mm spacing. A 2-mm-diameter tissue core was punched out from a representative tumor area of each patient’s block and transplanted to a recipient block using a manual microarrayer (Quick-Ray set, Unitma, Seoul, Korea). Each tissue microarray block contained 47 cases of PTC, one case of normal thyroid tissue, one case normal tonsil tissue, and one case of placental tissue.Serial 4-μm-thick sections were cut from tissue microarray blocks, deparaffinized, and rehydrated in serial graded ethanol washes. Immunohistochemistry for CD10 and CD15 was performed using an automated Dako OMNIS GI100 stainer (Dako, Agilent, Santa Clara, CA, USA) in accordance with the manufacturer’s instructions. Antigen retrieval was performed on sections for 30 min at 97 °C using the EnVision FLEX Target Retrieval Solution High pH (Dako). Tissue sections were incubated with mouse anti-human CD10 monoclonal antibody (1:50, clone 56C6, NCL-L-CD10-270, Novocastra, Newcastle Upon Tyne, UK) and mouse anti-human CD15 monoclonal antibody (1:1000, clone Carb-3, Dako) for 20 min, followed by visualization with Dako EnVision FLEX /HRP detection reagent for 20 min and substrate chromogen for 5 min. The specimens were then counterstained with hematoxylin for 3 min. For positive tissue controls, all tissue sections included tonsil and placental tissue (Figure 5). Normal thyroid tissue was used as negative tissue controls. One tissue section per each run was simultaneously incubated with the antibody diluent in place of the primary antibody.Two pathologists (Eun Ji Oh and Chan Kwon Jung) independently assessed CD10 and CD15 immunohistochemical staining. In the case of discrepancy, a consensus was reached between the two observers. Immunoreactivity of CD10 and CD15 was interpreted as the staining proportion of tumor cells. The cytoplasmic immunoreactivity for CD10 and CD15 in more than 5% of tumor cells was considered as positive [24].Genomic DNA was isolated from 10 μm-thick FFPE whole tissue sections enriched for the tumor cells by macrodissection using a RecoverAll Total Nucleic Acid Isolation Kit for FFPE (Life Technologies, ThermoFisher Scientific, Carlsbad, CA) according to the supplier’s instructions.PCR reaction for the amplification of exon 15 of the BRAF gene was performed under the following conditions: 1 cycle of 3 min at 95 °C for denaturation, 35 cycles consisting of 30 s at 94 °C, 30 s at 55 °C, and 30 s at 72 °C, followed by a final cycle of 7 min at 72 °C. The PCR primers for BRAF were as follows: forward, 5’-TCATAATGCTTGC TCTGATAGGA-3’ and reverse, 5’-GGCCAAAAATTTAATCAGTGGA-3’. Sanger sequencing of PCR amplicons was performed using the same PCR primers as described previously [8,43].We obtained RNAseq-based, gene-level normalized RSEM (RNA-Seq by Expectation Maximization) scores for PTC (n = 505) from PanCancer Atlas resource (https://gdc.cancer.gov). The clinical, pathological, and molecular data for each TCGA sample were obtained from the article by the TCGA research network [46]. The expression levels of MME (CD10) mRNA and FUT4 (CD15) mRNA were grouped into low (<median expression of the gene) and high expression (≥median expression) based on the median value of mRNA expression.In the TCGA dataset, thyroid cancer staging was initially based on the 7th edition of the AJCC cancer staging system [46,47]. The extrathyroidal extension was classified into three categories, namely minimal (T3), moderately advanced (T4a), and very advanced (T4b). The minimal extrathyroidal extension refers to the extension to sternothyroid muscle or perithyroidal soft tissues. In the AJCC 8th edition, minor extension through the thyroid capsule seen only on histologic examination is not used for staging, and gross extrathyroidal extension involving strap muscles is classified into T3b disease. The presence of minor extrathyroidal extension involving strap muscle detected only by microscopy does not constitute T3b disease [38]. Although most T3 thyroid cancers by the AJCC 7th edition are reclassified into lower stages (T1 or T2) by the 8th AJCC staging system, some of them remain T3 with sternothyroid muscle invasion. We could not apply the 8th AJCC staging system in the TCGA dataset because there was no detailed information relating to cases with minimal extrathyroidal extension classified by the 7th AJCC system. Therefore, TNM staging in the TCGA dataset was done according to the 7th edition of the AJCC cancer staging manual [47].To infer the relative abundance of tumor infiltrating immune cells in PTC as we described previously [48], a support vector regression-based CIBERSORT algorithm was employed [30]. The default set (LM22) was used to estimate the relative abundance of 22 immune cell types. Ten cell types with positive and inverse correlation with FUT4 expression were selected for demonstration. A cytolytic score representing the activity of immune cytolytic effectors was calculated as the geometric means of expression of GZMA and PRF1 as previously described [36]. We used ESTIMATE R packages to estimate the score representing the proportion of immune and stromal cells in PTC [29]. Single sample gene set enrichment analysis was performed to estimate the expression-based activity of 50 hallmark gene sets as available in MSigDB Collections (https://www.gsea-msigdb.org/gsea/msigdb/collections.jsp).Chi-square test and Fisher’s exact test were used to analyze the association between CD10 and CD15 protein expression and clinicopathologic features. Survival curves were plotted using the Kaplan–Meier method. Statistical differences between survival curves were calculated using the log-rank test. Cox proportional-hazard model was employed for multivariate analysis of recurrence-free survival rates. Univariate and multivariate logistic regression analyses of variables were performed to determine whether clinicopathological variables were significantly associated with long-term clinical outcomes. The independent variables were entered simultaneously using the enter method.All statistical analyses were performed by IBM SPSS Statistics for Windows, Version 21.0. (IBM Corp., Armonk, NY, USA) and GraphPad Prism (version 6.05, GraphPad Software, La Jolla, CA, USA). A two-sided p value < 0.05 was considered as statistically significant.The expression of CD10 and CD15 might be implicated in the early stage of PTC development. CD15 expression was an independent prognostic marker for improved recurrence-free survival and excellent clinical outcomes after treatment, but CD10 expression had no impact on prognosis in PTC patients. The good therapeutic response to RAI therapy in PTC patients with CD15 expression might be attributed to the increased NIS expression.Conceptualization: A.B. and C.K.J.; methodology: E.J.O., H.C., T.-M.K., J.S.B., D.-J.L., and C.K.J.; software: E.J.O., H.C., T.-M.K., and C.K.J.; validation: E.J.O., A.B., H.C., T.-M.K., J.S.B., D.-J.L., and C.K.J.; formal analysis: E.J.O., H.C., T.-M.K., J.S.B., D.-J.L., and C.K.J.; investigation: E.J.O., A.B., H.C., T.-M.K., and C.K.J.; resources: C.K.J.; data curation: E.J.O., A.B., H.C., T.-M.K., J.S.B., D.-J.L., and C.K.J.; writing—original draft preparation: E.J.O. and C.K.J.; writing—review and editing: E.J.O., A.B., H.C., T.-M.K., J.S.B., D.-J.L., and C.K.J.; visualization: H.C., T.-M.K., and C.K.J.; supervision: T.-M.K. and C.K.J.; project administration: C.K.J.; funding acquisition: C.K.J. All authors have read and agreed to the published version of the manuscript.This research was funded by a grant (2017R1D1A1B03029597) from the Basic Science Research Program through the National Research Foundation of Korea (Daejeon, Republic of Korea) funded by the Ministry of Science and ICT.The authors declare no conflict of interest.Immunohistochemical staining for CD10 and CD15 on the tissue microarray of papillary thyroid carcinoma. (A) The tissue core contains tumor and non-tumor (NT) areas (hematoxylin and eosin stain). (B) The higher power view of the tissue core shows papillary thyroid carcinoma and NT adjacent to tumor. (C) NT in the same case is negative for both CD10 and CD15. Tumor cells in the same case are positive for both CD10 (D–F) and CD15 (G–I) immunostaining. (F) Inset shows CD10 staining at the apical surface of tumor cells. (I) Inset shows CD15 cytoplasmic and membranous staining in tumor cells; ×40 (A,D,G), ×100 (B,E,H), ×200 (C,F,I), inset (×400).Recurrence-free survival analysis of CD10 and CD15 expression in 467 patients with papillary thyroid carcinoma treated with total thyroidectomy and radioiodine remnant ablation. The recurrence-free survival was not correlated with CD10 expression (A) but was significantly correlated with CD15 expression (B). Stratified survival analyses show a significant association between CD15 expression and recurrence-free survival in patients with minimal and gross extrathyroidal extension (C) and lymph node metastasis (D).Correlation between FUT4 (CD15) expression levels and expression patterns of tumor microenvironment-related genes in 505 papillary thyroid carcinomas (PTCs) from the TCGA cohort. (A) The expression of FUT4 (CD15) mRNA is positively correlated with the infiltration level of immune and stromal cells based on the estimation by the ESTIMATE algorithm. The relative abundance of tumor infiltrating immune cells analyzed by the CIBERSORT algorithm (five cell types with positive correlation and five cell types with inverse correlation with FUT4 expression selected) (B), the expression level of immune checkpoints and cytolytic score (C), single sample enrichment score of MsigDB hallmark gene sets (10 positive and 10 inverse correlations with the selected FUT4 expression) (D), and epithelial mesenchymal transformation-associated genes (E).Correlation between mRNA expression of FUT4 and SLC5A5 in the TCGA database.Immunohistochemical staining for CD10 and CD15 on tissue microarray. Tissue microarray consisted of 47 cores of tumor tissue and three cores of control tissue (normal tonsillar, placental, and thyroid tissues).Relationship between clinicopathologic characteristics and the expression of CD10 and CD15 in 515 patients with papillary thyroid carcinoma.AJCC, American Joint Committee on Cancer. The AJCC 8th edition cancer staging was used in this study. * Minimal extrathyroidal extension by AJCC 8th edition refers to the extension to perithyroid adipose tissue, strap muscles, nerves, or small vascular structures identified only by microscopic examination [27].Multivariate Cox regression analysis of recurrence-free survival in 467 patients with papillary thyroid carcinoma treated with total thyroidectomy and radioiodine remnant ablation.Aggressive histologic subtype includes 21 tall cell variants, two hobnail variants, and one solid variant. CI: confidence interval.Univariate and multivariate logistic regression analyses of clinicopathological variables associated with long-term excellent therapeutic response in 467 patients with papillary thyroid carcinoma treated with total thyroidectomy and radioiodine remnant ablation.All patients without lymph node metastasis achieved an excellent therapeutic response. Aggressive histologic subtype includes 21 tall cell variants, two hobnail variants, and one solid variant. OR, odds ratio; CI: confidence interval.Relationship between clinicopathologic characteristics and the expression of MME (CD10) and FUT4 (CD15) mRNA expression in 454 patients with papillary thyroid carcinoma in The Cancer Genome Atlas (TCGA) database.AJCC, American Joint Committee on Cancer. The AJCC 7th edition cancer staging was used in the TCGA dataset. * Minimal extrathyroidal extension by AJCC 7th edition refers to the extension to sternothyroid muscle or perithyroid soft tissues [28].

Med-MDPI/cancers/cancers-12-06-01414.txt

@@ -0,0 +1 @@
+These authors share first-authorships.Purpose: Checkpoint inhibitors have significantly improved treatment of metastatic melanoma. However, 40–60% of patients do not respond to therapy, emphasizing the need for better predictive biomarkers for treatment response to immune checkpoint inhibitors. Prorammed death-ligand 1(PD-L1) expression in tumor cells is currently used as a predictive biomarker; however, it lacks specificity. Therefore, it is of utmost importance to identify other novel biomarkers that can predict treatment outcome. Experimental design: We studied a small cohort of 16 patients with advanced-stage melanoma treated with first-line checkpoint inhibitors. Plasma samples were collected prior to treatment initiation and continuously during the first year of treatment. Circulating tumor DNA (ctDNA) level and the expression of ten inflammatory cytokines were analyzed. Results: We found that the ctDNA-level in a blood sample collected after 6–8 weeks of therapy is predictive for response to checkpoint inhibitors. Patients with undetectable ctDNA had significantly longer progression-free survival (PFS) compared with patients with detectable ctDNA (median 26.3 vs. 2.1 months, p = 0.006). In parallel, we identified that high levels of the cytokines monocyte chemoattractant protein 1 (MCP1) and tumor necrosis factor α(TNFα) in baseline blood samples were significantly associated with longer PFS compared to low level of these cytokines (median not reached vs. 8.2 months p = 0.0008). Conclusions: These findings suggest that the levels of ctDNA, MCP1, and TNFα in baseline and early follow-up samples can predict disease progression in metastatic melanoma patients treated with checkpoint inhibitors. Potentially, these minimally invasive biomarkers may identify responders from non-responders.Metastatic melanoma is an aggressive cancer for which the incidence rate continues to rise worldwide [1]. Treatment options have been limited for metastatic melanoma patients but with the emergence of novel immunotherapeutic approaches over the past several years, the management of this disease has been vastly transformed. Immunotherapeutic agents targeting programmed cell death protein 1 (PD-1) or cytotoxic T-lymphocyte-associated protein 4 (CTLA-4) on the T cells have demonstrated prolonged overall survival (OS) for advanced-stage patients independent of mutational status [2,3,4,5]. Nonetheless, 40–60% of patients do not respond to immunotherapy and no clearly defined biomarker is available for predicting if patients will benefit from treatment [3,4,6]. The most widely accepted, clinically used predictive biomarker is the expression of PD-1 ligand (PD-L1) on tumor cells, which has been associated with higher likelihood of response to therapy [7]. Yet, up to 41.3% of melanoma patients with PD-L1 negative tumors respond to immunotherapy [4]. Furthermore, PD-L1 expression has been found to display considerable intra- and intertumoral heterogeneity [8], suggesting that PD-L1 expression is not a reliable and specific biomarker. Thus, it remains of utmost importance to uncover better predictive biomarkers for early assessment of response to immunotherapy. This will contribute to the prevention of unnecessary exposure to adverse events related to immunotherapy.Circulating cell-free tumor DNA (ctDNA) has been extensively studied as a non-invasive biomarker in melanoma and other cancer types. Analysis of ctDNA can identify tumor-specific mutations and has been shown to correlate with tumor burden and clinical outcome [9,10]. Several studies have found that undetectable ctDNA prior to or early after initiating treatment is associated with improved progression-free survival (PFS) and OS [11,12,13,14,15,16]. Furthermore, ctDNA can also be used to uncover mechanisms of acquired resistance at the time of disease progression, allowing therapy to be adapted accordingly [12,16]. Overall, this renders ctDNA a valuable tool for the real-time monitoring of response during treatment. Nonetheless, ctDNA detection can be limited by the sensitivity of detection methods and ‘non-shedding’ tumors.Immunotherapy is dependent on an activated adaptive immune system, where the focus is to improve the functionality of T cells. However, the activation level of the adaptive immune system is highly dependent on the function of the innate immune system. In this regard, activation of innate immune pathways by damage-associated molecular patterns (DAMPs) is of particular importance for further activation of the adaptive immune system and the generation of an anti-tumor immune response [17,18]. Specifically, DAMP-induced innate immune activation results in the production of inflammatory cytokines that support activation of the adaptive immune system and increase the influx of immune cells into the tumor microenvironment [19,20]. The ability to mount an innate immune response, and thereby also the ability to activate the adaptive immune system, varies between individuals. Here, intrinsic levels of blood inflammatory cytokines may reflect immune activation status of the individual. In this regard, high baseline levels of interferon γ (IFNγ), interleukin 6 (IL-6), and IL-10 have been associated with response to nivolumab [21]. Furthermore, a high level of transforming growth factor β (TGF-β) has been associated with increased response to nivolumab but worse outcome to ipilimumab [22,23]. Despite considerable research within the field, no consensus has been reached so far.Here, we investigated whether ctDNA and immune-related cytokines can predict treatment outcome in metastatic melanoma patients treated with first-line checkpoint inhibitors. We find that undetectable ctDNA at 6-8 weeks after treatment initiation as well as high baseline levels of the cytokines IFNβ, MCP1, and TNFα are predictors of superior PFS. Collectively, these data suggest that certain cytokines can synergize with ctDNA and function as minimally invasive predictive biomarkers for the effect of checkpoint inhibitor therapy.Checkpoint inhibitors have significantly increased survival for metastatic melanoma patients. Yet, 40–60% of patients do not respond to checkpoint inhibitor therapy. To provide the best possible treatment, it is of utmost importance to identify responders from non-responders. Currently, there is a lack of well-founded clinically available biomarkers for identification of responders from non-responders. Thus, it is essential to identify new biomarkers for predicting treatment response. Here, we performed an exploratory biomarker analysis in a cohort of advanced-stage melanoma patients treated with checkpoint inhibitors. Through investigation of a broad cytokine panel and ctDNA levels, we demonstrated that plasma levels of MCP1 and TNFα as measured at baseline, and ctDNA levels measured at 6–8 weeks of therapy were strong predictors for treatment response. These findings indicate a potential for using ctDNA, MCP1, and TNFα plasma levels to identify checkpoint inhibitor responders and calls for further investigations.Patients with unresectable, previously untreated stage III or IV melanoma who received systemic treatment with immune checkpoint inhibitors were eligible for the study. Key inclusion criteria were absence of uveal melanoma, absence of another primary cancer, and no previous diagnosis with cancer. A total of 33 patients were initially enrolled in the study, out of which 17 were later excluded as they did not meet the inclusion criteria: 6 due to earlier treatment; 4 due to other cancer type; 4 receiving first-line BRAF inhibitors, 2 due to lack of baseline samples, and 1 not receiving systemic treatment.Patients were enrolled between October 2016 and August 2017. All consecutive patients referred to systemic treatment at the Department of Oncology in Aarhus (Denmark) were included. All patients gave informed written consent before inclusion, and the study was approved by the Central Denmark Region Committees on Biomedical Research Ethics (no. 1-10-72-374-15) and performed in accordance with the Declaration of Helsinki.Patients received pembrolizumab at a dose 2 mg/kg every 3 weeks or nivolumab 1 mg/kg plus ipilimumab 3 mg/kg every 3 weeks, followed by maintenance nivolumab 1 mg/kg.Patient demographics and clinicopathologic features included: performance status and metastatic sites at baseline, lactate dehydrogenase (LDH), and any adverse events requiring steroid treatment during one year of treatment. Elevated LDH level was defined as levels above 205 units/liter (U/L) for patients below the age of 70 and above 255 U/L for patients above the age of 70. Tumor biopsies were routinely screened for BRAFV600E mutation status and PD-L1 expression level (</> 1%). Treatment responses were evaluated by Position emission tomography/computed tomography (PET/CT) scans and/or CT of chest, abdomen, and pelvis, and magnetic resonance imaging (MRI) in case of known brain metastases.Peripheral blood samples (3 × 10 mL Ethylenediamine tetraacetic acid (EDTA) tubes, BD Vacutainer, Plymouth, United Kingdom) were obtained at baseline (immediately before treatment initiation) and every 3–4 weeks during treatment for up to one year after treatment initiation. Plasma was isolated from peripheral blood samples within 2–3 hours after blood collection by 1800× g for 10 min at room temperature (RT). Plasma was cryopreserved at −80 °C until analysis.cfDNA was extracted from 4 mL plasma using the QIAamp Circulating Nucleic Acid Kit (Qiagen, Hilden, Germany) according to the manufacturer’s protocol. The isolated DNA was eluted in 100 μL elution buffer and stored at −80 °C until analysis.The QX200™ AutoDG™ Droplet Digital™ PCR System (Bio-Rad, Copenhagen, Denmark) was used to perform ddPCR. Samples were analyzed for the following mutations: BRAF p.V600E, NRAS p.Q61K/p.Q61R, and TERT C228T. The BRAF and NRAS assays were wet-lab validated assays purchased at Bio-Rad. The TERT assay was designed and validated in-house. Due to technical limitations with the TERT assay, only the C228T mutation was approved for this study. The reaction volume of 22 μL consisted of 2× Supermix for probes (no UTP), 900 nM primers, 250 nM probes, and 5 μL of purified cfDNA for the BRAF and NRAS assays. For TERT, it consisted of 1x Supermix for probes (no UTP), 1200 nM primers, 250 nM probes, 0.5 M betaine, 1 mM EDTA, and 5 μL of purified cfDNA. All reagents were purchased from Bio-Rad. All samples were conducted as duplicates and each run included positive and negative control samples. DNA extracted from the cell lines SK-MEL-28 and T24 was used as positive control for the BRAF assay and TERT assay, respectively. Genestrands (Eurofins Genomics, Ebersberg, Germany) diluted in cfDNA extracted from blood samples from anonymous donors collected from the blood bank at Aarhus University Hospital were used as positive controls for the NRAS assays. The limit of detection (LoD) for each assay was determined using donor cfDNA according to (Milbury biomol detect quant 2014). LoD and assay information can be found in Table S1. Data analysis was performed using QuantaSoft v.1.7.4.0917 software (Bio-Rad, Copenhagen, Denmark) and reported as copies per ml plasma.Blood levels of cytokines were analyzed by the proximity extension assay O-link (Immuno-oncology panel, BioXpedia, Aarhus, Denmark). DNA oligo-coupled antibodies targeting 92 different proteins were used in qPCR assay where the number of PCR target fragments were proportional to the concentration of target protein in the input sample. Normalized Protein expression (NPX) units were calculated from the Ct-values obtained from the qPCR run, depictured on a log2 scale. The following cytokines were excluded from the analysis due to undetectable levels in the input samples; FGF2, IL1α, IL2, IL5, IL21, IL33, IL35, and TNFα.A customized, high-sensitive 10-cytokine U-plex panel (Meso Scale Discovery, Rockville, MMD, USA) was used to analyze plasma levels of IFNβ, IFNγ, IL10, IL1β, IL21, IL6, IL8, IP10, MCP1, and TNFα in cryopreserved plasma samples according to the manufacturer’s protocol. Data were acquired using a QuickPlex SQ 120 instrument (Meso Scale Discovery, Rockville, MD, USA). The lower limit of quantification of each cytokine can be found in Table S2.Correlation between ctDNA and LDH or number of metastatic lesions was performed using Spearman’s rank correlation coefficient. ctDNA levels were dichotomized according to the median concentration and an unpaired t-test was used to assess the association between ctDNA level (log-transformed) and progression. Fisher’s exact tests was used to evaluate the distribution of patients according to biomarker status (ctDNA, cytokine score, and LDH). Patients with both BRAF p. V600E negative tumors and undetectable baseline ctDNA were excluded from all analysis on ctDNA.Time-to-event analysis were reported using PFS by the Kaplan–Meier method. PFS was defined as time from treatment initiation to the date of first reported progression or death due to any cause. Patients without disease progression or who were still alive at last follow-up were censored at the last follow-up date (2nd of October 2019). Log rank test was performed to assess differences in survival. The assumption of proportional hazard was tested by visualization of Kaplan–Meier plots before any statistic tests were performed. Univariate cox proportional hazards regression was calculated to estimate hazard ratios (HR) for PFS if the assumption was met. Each variable in the univariate cox proportional regression models was tested for proportionality. p-values less than 0.05 were considered significant. Analyses were carried out using StataIC version 15.1 (Stata Nordic, Stockholm, Sweden) and Graphpad Prism 8 (version 8.2.0).A total of 16 patients with unresectable stage III or IV melanoma and treated with first-line checkpoint inhibitors were enrolled at the Department of Oncology at Aarhus University Hospital. One out of 16 patients (6%) was diagnosed with stage III melanoma, while 15/16 (94%) patients were diagnosed with stage IV melanoma. Eleven patients received treatment with pembrolizumab and five patients received treatment with ipilimumab and nivolumab (Table 1). At the time of study evaluation, the median follow-up was 26 months (range 6.3–35.6 months).Tumor PD-L1 expression was analyzed on tumor-cells in 11/16 (68.7%) of the patients. Among these 11 patients, five (45.5%) patients displayed PD-L1 expression ≥1% (Table 1). The BRAF p.V600E mutation was detected in the tumor biopsy for 10 out of the 16 patients (Table 1). At database lock, nine (56%) patients had experienced disease progression and six (37.5%) patients had died (Table 1), out of which one patient on first-line ipilimumab and nivolumab died of a non-cancer related or treatment-related event. Five patients displayed elevated LDH levels measured at baseline, but this was not associated with worse PFS (p = 0.36) (Figure S1).cfDNA was analyzed for BRAF p.V600E, NRAS p.Q61K/p.Q61R, and TERT C228T mutations, as these are frequently found in melanoma patients [24,25,26,27]. ctDNA was detected by ddPCR at baseline in 9/16 patients (56%). The BRAF p.V600E mutations initially identified in the primary tumor biopsy were confirmed in ctDNA for 7/10 patients (70%). It should be noted that for the three patients with BRAF p.V600E mutated tumors, but undetectable ctDNA, the only metastatic sites were the brain, lung, and subcutaneous tissue, which are sites known to shed less ctDNA [12,28,29,30]. Mutations in NRAS were detected in two patients with one harboring NRAS p.Q61K and the other NRAS p.Q61R. One patient with a BRAF p.V600E mutation had a concurrent TERT C228T mutation. For four patients, mutations were identified in neither tumor biopsy nor ctDNA, and these patients were excluded from further analysis. An overview of the identified mutations is shown in Figure 1A. For the nine patients with detectable ctDNA at baseline, the median mutant allele concentration was 133.3 copies/mL plasma (range 10.20–3388) and the median allele frequency was 3.5% (range 0.60–24.6%). Patients who experienced disease progression did not have significantly higher baseline mutant allele concentration than patients without disease progression (p = 0.22). Furthermore, higher mutant allele concentrations (defined as above the median concentration) were not associated with PFS (median not reached vs. 6.35 months, p = 0.25) and neither was the presence of ctDNA at baseline. However, the baseline mutant allele concentration was significantly correlated with the number of metastatic sites (Spearman r = 0.76, p = 0.006) (Figure 1B).To assess whether early changes in mutant allele concentration were associated with PFS, we analyzed blood samples longitudinally collected during therapy. A total of 130 plasma samples were available from the 12 patients (median 9 samples per patient, range 4–19) with detectable mutations in tumor or in the baseline blood sample, and ctDNA was identified in 36 (28%) out of the 130 samples. The mutant allele concentration decreased within 2 months upon treatment initiation for the majority of patients, and it remained undetectable in particular for patients who did not experience disease progression (Figure 2A). Next, we divided the patients into two distinct subgroups according to the ctDNA level measured at week 6–8 after treatment initiation. One subgroup consisted of patients who had undetectable ctDNA at week 6–8 regardless of their baseline ctDNA status (n = 8), and the second subgroup consisted of patients with detectable ctDNA at week 6–8 (n = 4). Patients with undetectable ctDNA at week 6–8 had a significantly longer PFS (median 26.3 months) than patients with detectable ctDNA (median 2.1 months, p = 0.006) (Figure 2B). In a univariate Cox analysis, the presence of ctDNA at week 6–8 was a predictor of shorter PFS with a hazard ratio (HR) of 7.89 (95% confidence interval (CI): 1.40–44.6, p = 0.019).Examples of longitudinal ctDNA monitoring for different clinical situations are shown in Figure 2C. We observed for all patients, who experienced disease progression within one year, that the mutant allele concentration rose either prior to or at progression with a mean lead time of 85 days (range 0–192 days, n = 5). The four patients without disease progression had no detectable ctDNA within 6 weeks following treatment initiation and ctDNA remained undetectable in all following samples, except in a single sample for one patient where a diminutive level of ctDNA was detected.To explore whether other biomarkers could be used for predicting disease progression, especially in patients not applicable for ctDNA monitoring, we evaluated a large group of blood circulating cytokines. Various cytokines are known to shape and control the immune response and have been associated with anti-tumoral responses. To construct a possible biomarker cytokine panel of potential interest for the patients enrolled on checkpoint inhibitors, we first performed a broad screening of 92 immuno-oncology related proteins using the O-link technology. The O-link panel was validated on plasma samples from a small group of patients who received first-line pembrolizumab treatment. The analysis was conducted on both baseline and follow-up samples to identify baseline levels and discrepancy of cytokines over time. The screening demonstrated large plasma level variation among a large range of different cytokines (Figure 3A). To pinpoint cytokines that could potentially be used as predictive biomarkers, we next searched for cytokines that displayed large inter- or intrapatient variation. In-depth analysis of the cytokine concentrations revealed considerable interpatient variation for a number of the cytokines, including Interferon γ-induced protein 10 (IP-10), IL-8, IL-6, and MCP1 (Figure 3B). In addition to the cytokine screening, we searched the literature for cytokines of potential interest that either reflected innate immune activation or cytokines associated with anti-tumor immune responses [21,31,32,33,34]. Based on these findings, a panel of 10 cytokines (IFNβ, IFNγ, IL-1β, IL6, IL8, IL10, IL21, MCP1, and TNFα) was established for further validation in the patient cohort.For eight out of ten cytokines (IFNβ, IFNγ, IL6, IL8, IL10, MCP1, and TNFα) in the biomarker panel, we dichotomized the patients into a high and a low cytokine expression group using the median cytokine expression value determined in the baseline sample. For IL21, 9/16 (56%) patients had undetectable levels, and therefore patients were dichotomized according to undetectable (low) and detectable (high) IL21. IL1β was undetectable in the baseline samples in 13/16 (81%) patients and was excluded from further analysis. The interpatient variation for each cytokine is shown in Figure S3. Based on the dichotomization, we then analyzed whether any of the baseline cytokine levels were associated with PFS (Figure 4A and Figure S2). We found that both IFNβ, TNFα, and MCP1 demonstrated a significant difference in PFS between the low and high cytokine group (Figure 4A). The remaining six cytokines analyzed did not demonstrate any difference between low and high cytokine groups (Figure S2). Cox regression analysis revealed that the HR for PFS for patients having high versus low expression levels was 0.081 for TNFα (95% CI: 0.0098–0.66, p = 0.019), 0.073 for MCP1 (95% CI: 0.0088–0.61, p = 0.016), and 0.22 for IFNβ (95% CI: 0.044–1.09, p = 0.063) (Figure 4A). The median PFS was: 8.2 months for low IFNβ and not reached for high IFNβ; 4.2 months for low MCP1 and not reached for high MCP1; 4.2 months for low TNFα, and not reached for high TNFα.Based on the cytokine profiles evaluated above, we next speculated whether an “immune cytokine score” comprised of multiple cytokines could be used to strengthen the association to PFS. MCP1 and TNFα baseline levels provided the best stratification of patients with regard to PFS. Thus, patients having high baseline levels of both MCP1 and TNFα were assigned a cytokine score of 2, while patients having high baseline levels of either MCP1 or TNFα were assigned a cytokine score of 1. Finally, patients having low baseline levels of both MCP1 and TNFα were assigned a cytokine score of 0. Using this immune cytokine score, patients were then dichotomized into a low and a high cytokine score group. Patients with a cytokine score of 0 or 1 were assigned to the low group (n = 9), and patients with a cytokine score of 2 were assigned to the high group (n = 7) (Figure 4B). Consistent with the results for the individual cytokines, we found that patients with a high cytokine score demonstrated a significantly longer PFS compared with patients with a low cytokine score (log rank test: p = 0.0008) (Figure 4B). Median PFS was not reached for the high cytokine score group but was 8.2 months for the low cytokine score group. In contrast to PFS data on individual cytokines, only one patient classified with a high immuno cytokine score experienced disease progression during the follow-up period. Similar to ctDNA, we also evaluated the cytokine score at 6-8 weeks and found that this cytokine score also predicted longer PFS (Figure 4C). Taken together, high levels of MCP1 and TNFα measured at baseline and 6–8 weeks post treatment initiation were predictors of superior PFS in patients treated with checkpoint inhibitors.We next investigated if there was an association between ctDNA detection and the cytokine score in terms of defining patients with a favorable biomarker status (undetectable ctDNA and high cytokine score). As both ctDNA and cytokine score were expressed as binary variables, we applied Fisher’s exact test to test for correlation between ctDNA and cytokine score. Among the 12 patients in this study, we did not observe an association between undetectable ctDNA and high cytokine score (p = 0.0808) (Table S3), suggesting that there is a discrepancy between patients identified with undetectable ctDNA and patients identified with a high cytokine score. Moreover, we tested if there were association between ctDNA (undetectable vs. detectable) and LDH (normal vs. elevated) (p = 0.55) or between cytokine score (low vs. high) and LDH (normal vs. elevated) (p = 0.99); but neither ctDNA nor cytokine score were significantly associated with LDH level (Figure S1). These results suggest that ctDNA detection and the cytokine score are independent predictors of outcome and can thus supplement each other.The present study investigated the association between PFS and the levels of ctDNA and cytokines in baseline and early follow-up samples from melanoma patients treated with first-line checkpoint inhibitors. We identified a favorable ctDNA profile, defined as undetectable ctDNA levels after 6–8 weeks of therapy, as a good predictor of prolonged PFS. Furthermore, we demonstrated that high baseline levels of the cytokines MCP1 and TNFα predicted longer PFS.There is a general focus in the field on identifying biomarkers that can predict treatment response to checkpoint inhibitor thereby improving personalized medicine. Baseline LDH level has been accepted as a strong prognostic marker in melanoma [35,36,37,38] and tumor PD-L1 expression has also been suggested as a biomarker for response to anti-PD-1 therapy [39]. However, it has previously been shown that many patients benefit from checkpoint inhibitor therapy, despite their tumors being classified as PD-L1 negative, emphasizing the continued need for better and more reliable biomarkers [4,40].Here, we investigated the potential of using ctDNA as a predictive biomarker. We observed that undetectable ctDNA at 6–8 weeks of therapy predicted longer PFS, verifying that ctDNA can be a valuable biomarker in melanoma patients receiving checkpoint inhibitors. This is in agreement with other studies showing that ctDNA is a predictor for PFS in melanoma patients treated with checkpoint inhibitors but also targeted therapy [12,13,15,16,41]. We did not find an association between ctDNA detection at baseline and PFS, a finding that is both in agreement with [13] and in contrast to findings of other studies [12,14]. We furthermore observed that monitoring ctDNA levels during therapy with checkpoint inhibitors can inform on treatment response in real-time. Importantly, the mutant allele concentration rose either prior to or at progression for all patients, who experienced disease progression during the ctDNA monitoring period. This is in agreement with several other studies, indicating that evaluation of ctDNA during therapy may be a feasible supplement to conventional radiological imaging [12,14,16].Nevertheless, one limitation using ctDNA analysis is the so-called non-shedders, which are patients where ctDNA cannot be detected. In our study, we were unable to detect ctDNA harboring mutations in BRAF, NRAS, or TERT in the baseline sample for seven (44%) of the patients. For four of these patients, this is most likely explained by the limited number of tumor-associated mutations that we assessed. However, the remaining three patients had BRAF mutations in their tumor, which we could not validate in plasma. Most of these patients had isolated metastases located in sites known to shed low levels of ctDNA, such as the brain and lungs [12,28,29,30]. Thus, alternative biomarkers would actually be needed to monitor these ctDNA negative patients. For this purpose, we chose to evaluate the association between baseline inflammatory cytokine levels and PFS. Here, we found that high baseline levels of MCP1 and TNFα were predictors of superior PFS in metastatic melanoma patients treated with checkpoint inhibitors. Combining MCP1 and TNFα into a cytokine score provided an even stronger predictor of PFS, suggesting that assessing multiple cytokines may be a very robust method of choice. Notably, based on our data, a cut-off value for high and low cytokine level could not be determined, and thus discrimination between high and low cytokine level solely applies to this specific cohort. Studies including larger cohorts are needed for further validation and determination of defined cytokine cut-off values. Of note, we found the Mesoscale TNFα assay to be more sensitive than the O-link TNFα assay. This explains why TNFα was detectable in samples analyzed by Mesoscale, but not O-link, emphasizing the importance of using sensitive methods in biomarker studies.We hypothesize that the levels of the proinflammatory cytokines can be used as predictors of treatment response to checkpoint inhibitors owing to their important role during innate and adaptive immune activation. It has previously been shown that response to checkpoint inhibitor therapy is strongly associated with a high density of tumor-infiltrating lymphocytes (TILs) and an intra-tumoral IFNγ gene signature [31,34,42,43,44], suggesting that endogenous T cell activation is important for eliciting an effective response to checkpoint inhibitors. T cell activation is dependent on activated antigen presenting cells (APCs), and APCs are activated through the innate immune system. APC activation results in cytokine production and presentation of antigens by the APC to naïve T cells, which in turn become activated. Here, it is relevant to notice that dendritic cells secret several different cytokines upon activation, including MCP1 and TNFα. Therefore, we speculate that high levels of immune-related cytokines may reflect an activated innate and adaptive immune system within the tumor microenvironment. Hence, patients with high cytokine levels may be more primed to respond to checkpoint inhibitors.ctDNA negativity and the cytokine score represent two distinct areas of biomarkers; ctDNA is tumor-specific and reflects the molecular composition of the tumor and possibly tumor burden, while cytokines are indicators of immune activation status. In the present study, we demonstrated that it was favorable for patients to have either undetectable ctDNA or a high cytokine score. However, we found no significant overlap between these two patient groups, indicating that ctDNA and cytokines can be used as independent biomarkers. One reason for this discrepancy can be the inability to detect ctDNA in patients with unknown mutations. Cytokines, on the other hand, are measurable if present in concentrations above detection limit. Thus, the cytokine score may synergize with ctDNA, yielding a strengthened predictive biomarker.We are aware of the limitations of the study. The relatively small number of patients limits the extrapolation of the findings to the general clinical setup. Further validation of our findings in larger cohorts is therefore needed to validate results. However, it is worth noting that the association between ctDNA and PFS found in this study, is in agreement with a previous study and in general seems to be a well-established association [13]. Evaluation of cytokines as predictive biomarkers for responses to checkpoint inhibitors is an emerging field. A previous study found six cytokines (TRAIL, MCP1, IL2, TNFα, IL8, and IP10) to be associated with overall survival in two discovery cohorts, but were unable to verify the results in a validation cohort of 49 patients, emphasizing the need for validation [32]. The identification of cytokine-specific cut-off values using Receiver Operating Characteristic (ROC) analysis and validation of these in larger validation cohorts is important for verifying cytokines as predictive biomarkers.In summary, in our cohort, undetectable ctDNA after 6–8 weeks of therapy and high baseline levels of MCP1 and TNFα are all individual predictors of superior PFS in metastatic melanoma patients treated with first-line checkpoint inhibitors. This current study suggests that ctDNA, MCP1, and TNFα can synergize and may in the future be utilized as a minimal invasive predictive biomarker for disease progression. Nonetheless, our results should be further validated in others melanoma cohorts, and preferably also explored in other cancer types treated with checkpoint inhibitors.The following are available online at https://www.mdpi.com/2072-6694/12/6/1414/s1, Figure S1: Survival analysis according to LDH level, Figure S2: Survival analysis of baseline cytokine levels in patients treated with checkpoint inhibitors, Figure S3: Interpatient cytokine variation, Table S1: ddPCR assay information, Table S2: Lower limit of quantification, Table S3: Contingency tables comparing ctDNA vs. cytokine score, LDH vs. ctDNA, and LDH vs. cytokine score.J.G.P., A.T.M., B.S.S., T.H.Ø. and M.R.J. conceived and designed the study. J.G.P., A.T.M., K.R.G., and N.A.-P. performed experiments and biostatistical analysis. J.G.P., A.T.M., T.H.Ø. and M.R.J. wrote first manuscript draft. All authors participated in data analyses, data interpretations as well as manuscript preparation. All authors have read and agreed to the published version of the manuscript.This study was partly funded by grants from Danish Cancer Society (R167-A10737-17-S2); Lundbeck foundation (R238-2016-2708); and Danish Health Authority.We thank Ane Kjeldsen, Department of Biomedicine, Aarhus University for assistance in the laboratory and as well as Lene Dabelstein and Birgit Mortensen; department of Clinical Biochemistry. We thank the medical laboratory technicians who collected blood samples and the patients who participated in the study.The authors declare no conflicts of interest.Detection of circulating tumor DNA (ctDNA) in baseline samples. (A) Overview of the patient cohort and the ctDNA mutations detected. The upper panel shows patient characteristics, while the lower panels show tissue and baseline ctDNA mutations detected. (B) Correlation between the baseline ctDNA level and number of metastatic sites (n = 12) analyzed with Spearman’s correlation coefficient (r).Early changes in circulating tumor DNA (ctDNA) levels. (A) Changes in ctDNA levels during the initial three months of first-line therapy. Lines discontinued before three months represent the last sample prior to or at disease progression. The ctDNA ratio reflects changes from the baseline sample. Only patients with ctDNA detected at baseline are represented (n = 6 with progression, n = 3 without progression). (B) Survival analysis for PFS according to whether ctDNA was detected at week 6–8 after therapy initiation. The difference between the groups was calculated using the log-rank test. (C) Examples of longitudinal monitoring of ctDNA levels during therapy. Time is depicted on the x-axis as days since treatment start. Abbreviations: Ipi/Nivo, ipilimumab/nivolumab; Dab/Tram, dabrafenib/trametinib; Pembro, pembrolizumab; PD, progressive disease.O-link screening for cytokines as potential biomarkers. (A) Baseline and follow-up blood samples from four patients were analyzed by O-link for the levels of 92 immuno-oncology proteins. All four patients were treated with first-line pembrolizumab. Normalized Protein eXpression (NPX) expression values (arbitrary units) are visualized on the heatmap. White color indicate undetectable cytokine level. The following cytokines were excluded as their concentrations were below detection limit: Fibroblast growth factor 2 (FGF2), Interleukin (IL) 1α, IL2, IL5, IL21, IL33, IL35, and tumor necrosis factor α (TNFα). (B) Graphs display plasma level changes for Interferon γ-induced protein 10 (IP10), monocyte chemoattractant protein 1 (MCP1), IL6, and IL8 over timer. A.U. = arbitrary units. T.I. = treatment initiation.High cytokine levels predict longer progression-free survival (PFS) in checkpoint inhibitor treated patients. Baseline and follow-up blood samples were analyzed for levels of interferon β (IFNβ), monocyte chemoattractant protein 1 (MCP1), and tumor necrosis factor α (TNFα). (A) Survival analysis for PFS according to cytokine level in baseline blood sample. For each cytokine, the patients were dichotomized by the individual median cytokine concentration into a low and a high cytokine group. (B) Survival analysis for PFS according to cytokine score calculated from baseline MCP1 and TNFα levels. (C) Survival analysis for PFS according to cytokine score calculated from MCP1 and TNFα levels measured at week 6–8 after therapy initiation. Tick marks denote censored patients. (n = 16). HR; hazard ratio.Patient and disease characteristics.

Med-MDPI/cancers/cancers-12-06-01415.txt

@@ -0,0 +1 @@
+Melanoma is one of the most aggressive types of cancer and the most deadly skin cancer. According to World Health Organization, about 132,000 melanoma skin cancers occur globally each year. Thanks to the efficacy of new therapies, life expectation has been improved over the last years. However, some malignant melanomas still remain unresponsive to these therapies. The β-adrenergic system, among its many physiological roles, has been recognized as the main mediator of stress-related tumorigenic events. In particular, catecholamine activation of β-adrenergic receptors (β-ARs) affects several processes that sustain cancer progression. Among the β-AR subtypes, the β3-AR is emerging as an important regulator of tumorigenesis. In this review, we summarize data of different experimental studies focused on β3-AR involvement in tumor development in various types of cancer and, particularly, in melanoma. Taken together, the preclinical evidences reported in this review demonstrate the crucial role of β3-AR in regulating the complex signaling network driving melanoma progression. Therefore, a need exists to further disseminate this new concept and to investigate more deeply the role of β3-AR as a possible therapeutic target for counteracting melanoma progression at clinical level.Melanoma is still one of the most aggressive and chemotherapy-resistant human cancers, originating from melanocytes. Melanoma development, as for the majority of skin cancers, is related, at least in part, to ultraviolet radiation exposure and to years of any protracted sun/UV exposure. This relationship explains why melanoma location is related to sun exposure [1]. Further risk factors for melanoma development include a particularly fair skin, hair, and eye color phenotypes associated with increased sun sensitivity, the presence of dysplastic nevi, and genetic predisposition [2].The incidence of invasive melanoma has been rising worldwide over the past two decades, despite numerous efforts to enhance primary prevention and early detection [3]. The overall incidence has been increasing particularly in the United States, with an increase by 270% over the past 30 years [4]. Melanoma predominantly affects the Caucasian population, whose risk is 10-fold increased compared to people with dark skin pigmentation [4], and represents the fifth most frequent cancer in men and the sixth most common cancer in women in the United States [5]. The region with the highest incidence of melanoma is Australia with the age-standardized incidence rate increased by 181% between 1982 and 2016, from 27 cases per 100,000 in 1982 to an estimated 50 cases per 100,000 in 2016 [6]. It has been assessed so far that adrenergic system, via catecholamine release following stress events, is able to sustain cancer progression and influence the outcome of patients affected by several malignancies. In this contest, β-adrenergic receptors (β-ARs) have been identified as the main responsible actors of stress-enhanced tumor-related pathways [7]. β-ARs are G protein-coupled receptors responsible for mediating many physiological and pathological responses in humans and animals. Although initially two β-AR subtypes have been identified, β1-AR and β2-AR [8], the existence of the third β-AR subtype, firstly described as “atypical β-adrenoreceptor” [9], and named β3-AR, has been acknowledged much later. We know that all three β-ARs can be activated by noradrenaline and adrenaline, although with different affinity. The two catecholamines are, indeed, equipotent at the β1-AR, while adrenalin is 100-fold more selective for β2-AR. On the contrary, noradrenaline is more potent than adrenaline as a β3-AR agonist [10]. All three β-ARs are expressed in numerous cell types of the tumor microenvironment; therefore, dissecting the effects related to each specific β-AR subtype involved in tumor progression is a complex but required objective.Since in melanoma, especially in resistant subtypes, new therapies are needed; in light of a continuous search for alternative therapeutic strategies, β-AR modulators cannot be forgotten. In particular, recent preclinical studies evidence the emerging role of β3-AR in melanoma.For cutaneous melanomas localized and not outspread beyond the site of origin, surgical excision remains the primary modality of treatment, with a high five-year relative survival rate but yet dependent on the tumor mass thickness and ulceration. However, survival percentage drops dramatically for patients diagnosed with advanced or metastatic melanoma (65% for regional to 25% for distant melanoma lesions), because treatment options are limited [11]. Chemotherapy, such as Dacarbazine (DTIC) [12] or Temozolomide [13], has been shown to offer only modest benefits and severe adverse effects, and has not been able to significantly change the prognosis. However, in the past decade, the approval of many new treatments for metastatic disease has contributed significantly to the recent increase in survival rate, as recently reported by the American Cancer Society [14].In 2011, the US Food and Drug Administration (FDA) approved Ipilimumab for metastatic disease therapy, the first immune checkpoint inhibitor (an anti-CTLA-4 drug), for the treatment of advanced melanoma, and Vemurafenib, a BRAF inhibitor, for the treatment of unresectable or metastatic melanoma with BRAF V600E mutation. Then, two new monoclonal antibodies, Nivolumab and Pembrolizumab targeting the anti-programmed cell death protein 1 (PD-1) were approved and marketed in 2014 [15]. At the same time, another drug able to inhibit BRAF, Dabrafenib, was approved by the FDA in 2013 [16], but also MEK inhibitors, drugs able to inhibit the mitogen-activated protein kinase enzymes MEK1 and/or MEK2, such as Trametinib or Cobimetinib, were made available and approved, respectively, in 2013 and 2015. All these new drugs have been used alone or in combination [17]. These new pharmaceutical opportunities, and the ability to combine these drugs to target different immune checkpoints, have contributed to the sharp drop in mortality [18]. The overall survival rate at 5 years was 52% in patients treated with Nivolumab and Ipiliumubab versus 44% in patients treated only with Nivolumab and 26% in those treated only with Ipilimumab [19]. Alongside the higher cost of biological therapies compared to traditional ones for the treatment of melanoma, the significant increase in survival has been associated with an impressive increase of expenses, raising the question of financial sustainability, mainly for public health systems [20]. For this reason, the availability of new options, hopefully less expensive, represents a matter of priority. Despite the significant improvements in life expectations for patients with historically poor outcomes attributable to new immunotherapies, a significant number of patients still remain far from significant long-lasting benefits. Therefore, alternative therapies are definitely required.The longstanding hypothesis of the existence of a relationship between stress (for example related to surgery, but also linked to psychosocial factors, such as depression) and tumor onset, has been confirmed in the last years through numerous experimental and clinical investigations [21,22]. The explanation of this relationship must be sought in the activation of the sympathetic nervous system and the release of stress-related mediators such as catecholamines, whose concentrations increase in the tumor microenvironment [23]. In particular, sympathetic nervous system and catecholamine exert pro-tumorigenic actions through the modulation of β-ARs, which in turn affect several biological processes related to cancer progression or metastasis, such as tumor cell proliferation, invasiveness, migration and vascularization [7,22,24,25]. Psychological aspects affecting melanoma patients, such as fear of cancer recurrence, anxiety, depression and risk of suicide, have been documented by several studies [26,27,28]. Therefore, in light of the existing link between stress and cancer progression in melanoma patients, interventions aimed to provide efficient psychological support are of uttermost importance in management of melanoma patients [29]. In preclinical murine models, β-ARs antagonists have shown the ability to block stress-induced enhancement of tumor progression in several malignancies, including melanoma [30,31], breast [32], prostate [33], and leukemia [34] cancers. On the contrary, β-ARs agonists have been found to enhance in vivo tumor progression and metastasis [32,33,35], confirming the critical involvement of β-ARs in cancer biology. Among the direct effects exerted on tumor cells, some reports demonstrated that β-ARs are also able to regulate pathways in non-tumoral cells of tumor microenvironment [36]. Despite the first data highlighted a prominent role of the β2-AR subtype in regulating tumor-relating pathways [37], recent studies have directed attention to β3-AR subtype as an important player in cancer biology. The involvement of stress in melanoma development and progression is definitely clear and well established by biological and clinical evidences [38]. A large amount of in vitro and in vivo experimental data together with epidemiological studies have shown that β2-AR is the β-AR subtype mostly involved in mediating the effects of catecholamines in cancer [39].In melanoma, the role of β-adrenergic system has been evaluated by in vitro and in vivo studies before testing β-adrenergic blockers in humans. In vitro studies confirmed the presence of β-ARs in animal and human melanoma cells. These studies also confirmed the role of β-ARs in the modulation of angiogenesis and demonstrated that non-selective β-AR antagonists, but not β1-AR selective antagonists, promote apoptosis of cancer cells [40,41,42]. Propranolol exerts potent anti-tumoral effects, attenuating migration, reducing vascular endothelial growth factor (VEGF) secretion and inducing apoptosis in both cutaneous and uveal melanoma in a dose-dependent manner [43]. Anti-angiogenic effect of propranolol has also been well described in infantile hemangiomas (IHs) and retinopathy of prematurity [39]. In fact, a significant reduction of serum VEGF levels has been demonstrated in infants affected by IHs after one or two months of propranolol treatment [44,45]. A similar effect has been demonstrated in hypoxic retina [46]. Accordingly, VEGF expression is down-regulated by propranolol in a dose-dependent manner in hemangioma-derived endothelial cells [47] and in hemangioma-derived stem cells [48]. All these data suggest that the anti-tumor activity of propranolol relies in part on its ability to block tumor angiogenesis. Besides its anti-angiogenic activity, propranolol affects activities of other cells in tumor microenvironment. In ovarian carcinoma, propranolol decreases macrophage recruitment by tumor mass [49], and in a model of breast cancer it prevents macrophage M2 polarization [50]. In melanoma, propranolol decreases the infiltration of immunosuppressive myeloid cells, such as neutrophils, in the tumor microenvironment and favors the cytotoxic tumor-infiltrating lymphocytes’ activity [51]. Moreover, the pathogenic role of stress in promoting melanoma growth in human cells and in mice was described by the efficacy of propranolol used alone [30,42,52] or associated with other drugs [53]. Of particular interest is the recent demonstration of the existing relationship between stress and immune depression, which is indicated as a decisive factor for tumor progression and metastasis development [54,55,56]. The β-adrenergic system has been identified as one of the major players in the regulation of the immune system. In particular, it was demonstrated that increased catecholamine levels induced both suppression of natural killer (NK) cell cytotoxicity [57,58,59], leading to tumor metastasis [34,60], and reduction of cytolytic killing ability of antigen-specific CD8+ T cells [61]. Moreover, β-adrenergic system promoted an increase of T regulatory (Treg) cell suppressive activity, through β2-AR signaling [62] and an accumulation of myeloid-derived suppressor cells (MDSC) in mice [63] and in human patients [64]. These effects were mediated by an increased production of norepinephrine [65]. Interestingly, a preclinical study demonstrated that in mice bearing B16-F10 melanoma cells, treatment with propranolol was able to potentiate immune-based therapies [66]. These preclinical evidences suggested to better explore the possible efficacy of β-blockers as anti-cancer adjunctive treatment [51,67] and to improve the efficacy of cancer immunotherapy [68].The strongest evidence of the role played by β-ARs in regulating distinct aspects of melanoma cancer comes from clinical investigations showing the efficacy of non-selective β-ARs’ antagonists in reducing the malignancy progression. Numerous observational studies suggested that patients treated initially with β-ARs blockers for cardiovascular diseases and hypertension have also shown reduced melanoma risk [69,70,71]. De Giorgi et al. demonstrated that patients with thick malignant melanoma (thickness > 1 mm) and concomitantly treated with β-blockers for one year or more, were associated with a reduced risk of tumor progression. In fact, the risk of disease progression (assessed by the presence of sentinel lymph node metastases and lymphatic, in-transit, or visceral metastases) was significantly reduced by 36% for each year of β-blockers use. Although the number of patients included in this study was relatively small a significant reduction in mortality was observed in patients treated with β-blockers [69]. Clinical follow-up of these patients after eight years from treatment with β-blockers confirmed the reduced disease progression from 45% to 30%, and reduced mortality from 35% to 17% [70]. The study of Lemeshow et al. supports a similar conclusion: a large population-based cohort study showed that the exposure to β-blockers for more than 90 days prior to diagnosis induced a lower mortality due to melanoma compared to the no-exposure group [71]. However, two large epidemiologic studies performed in patients who received β-AR antagonists (the majority assuming β1-AR selective blockers) did not confirm these results, even though the risk improved with larger amounts of cumulative daily dose [72,73]. A clinicopathological study evaluated the prognostic significance of β2-AR expression on surgically resected cutaneous malignant melanoma. This study revealed that β2-AR expression was positively correlated with tumor thickness, ulceration, disease stage, and finally with a poor overall survival [74]. A recent study observed a significant survival benefit for patients treated with immunotherapy and taking unselective β-blockers compared to either those receiving no β-blockers or β1 selective-blockers [66]. Moreover, a more recent prospective study performed in patients with cutaneous melanoma demonstrated that propranolol treatment reduced the risk of recurrence by 80% [75].The first report, showing that β3-AR influences the risk of cancer, suggested that a polymorphism in codon 64 of the β3-AR gene, that features replacement of tryptophan by arginine (Trp64Arg), decreased the risk of breast cancer in Japanese women. In particular, individuals who simultaneously carried a glutamic acid polymorphism in β2-AR gene (Gln27Glu) together with the Trp64Arg β3-AR polymorphism had the most markedly reduced risk of breast cancer, with an odds ratio of 0.37 [76]. In a second study, the same β3-AR polymorphism (Trp64Arg) was associated with susceptibility to endometrial cancers in overweight/obese individuals [77]. From these first studies, an involvement of β3-AR in the biology of cancer was beginning to emerge. However, until few years ago, the poor knowledge regarding the β3-AR distribution and pharmacology, and the lack of selective tools suitable for the study of this β-AR subtype, has made difficult to clarify its contribution in the complex landscape of tumor biology. Indeed, the presence of β3-AR was first established in physiological tissue, primary in brown adipocytes where it mediates thermogenesis [9], and, for many years, its expression in tumor tissues was completely unknown.For the first time, in 2008, a significant up-regulation of the β3-AR mRNA was described in a tumor tissue; a study involving 41 patients affected by colorectal cancer suggested a possible involvement of the β3-AR in the pathogenesis of this malignancy [78].In 2013, in a very interesting study, Magnon et al. [79] investigated the role of adrenergic signals in prostate cancer development. To assess the contribute of the β-ARs, human prostate PC-3 cancer cell line was injected in the prostate of mice genetically deficient for β2-, β3-, or both β2- and β3-ARs. The first relevant observation was that tumor growth in the prostate was slightly delayed when mice were lacking β2- or β3-AR singularly, but it was severely compromised in ADRB2−/−ADRB3−/− mice. Notably, in the double knockout mice, prostate cancer cell dissemination into the lymph nodes and other distant organs was significantly reduced. These results were also confirmed by using the human prostate LNCaP cell line in the same animal model, suggesting that both β2- and β3-ARs, expressed in stromal cells of the tumor microenvironment, are critically involved in tumor development and metastatic dissemination of this malignancy.Recently, β3-AR mRNA and protein expression have been reported across different tumors including vascular tumors, breast cancers and human leukemia cells [80,81,82]. Notably, in these diseases, β3-AR mRNA or protein expression were strongly increased compared to the healthy counterpart tissues. Moreover, new evidence on β3-AR expression was obtained in many other tumors [83], confirming the hypothesis that this β-AR subtype could play a pivotal role in the onset and/or progression of numerous malignancies [Table 1]. Accordingly, a β3-AR gene variant has been found implicated in the predisposition to gallbladder cancer, the most common and highly aggressive biliary tract malignancy [84]. In addition to several studies on melanoma, discussed below, we recently demonstrated that β3-AR is expressed in both murine and human neuroblastoma (NB) cell lines, and in tumor biopsies from NB patients; in this study, pharmacological antagonism of β3-AR, in a murine syngeneic model of NB, was able to reduce tumor growth by affecting the neuronal differentiation of NB cancer cells [85].A prospective cohort study has demonstrated that the most used β-ARs antagonist, propranolol, an approved drug with a non-oncology primary purpose, protected patients with thick cutaneous melanoma from melanoma recurrence [75]. Propranolol is reported to have an order of affinity for β2-, β1- and β3-AR respectively of 0.8, 1.8, and 186 nM [93,94]. The protective effects observed in melanoma patients treated with propranolol suggested, therefore, a clear involvement of the β2-AR subtype in melanoma progression. However, it should be emphasized that an involvement of the other β-AR subtypes in the observed effects related to melanoma progression cannot be excluded, owing to the broad dosage spectrum with which propranolol and other β-ARs antagonists have been used in clinical practice [95], and considering that a clear-cut selectivity remains questionable for most β-blockers [96]. Unfortunately, due to the lack of clinical studies using selective β3-AR antagonists in humans, the role of β3-AR subtype in melanoma cancer has not been clarified so far at clinical level; nevertheless, its contribution to processes related to melanoma progression is becoming evident, as suggested by pre-clinical evidences.Studies on murine B16-F10 melanoma cells demonstrated for the first time that the pharmacological β3-AR blockade was able to reduce proliferation and induce apoptosis of melanoma cells in vitro; these effects were also reproduced by using a siRNA molecular approach targeting specific β-ARs [86]. In the same study, thorough a murine syngeneic experimental model (B16-F10 inoculated in C57BL mice), β3-AR antagonists SR59230A and L-748,377, were both able to significantly reduce melanoma growth in vivo, in comparison with propranolol, by reducing cell proliferation and inducing apoptosis of cancer cells. In addition, the β3-AR blockade was also able to reduce tumor vasculature through apoptosis of endothelial cells [87]. The effects of β3-AR modulation on melanoma cell proliferation and survival were found mediated by the inducible nitric oxide synthase (iNOS) demonstrating that iNOS-produced nitric oxide (NO) is a downstream effector of β3-AR signaling in melanoma [88]. Notably, melanoma cell proliferation was inhibited by β3-AR blockade either in the presence or not of noradrenaline stimulation, indicating that a partial constitutive β3-AR activity, already hypothesized [97], was present in melanoma cells and may contribute to the proliferation process [86].Both β2- and β3-AR were found strongly expressed and actively functional in different stromal cells of the TME, such as cancer-associated fibroblasts (CAF), macrophages and endothelial cells [89]. Indeed, besides the direct effects elicited by the β3-AR modulation in melanoma cells, several studies have shown how this receptor is able to regulate stromal, inflammatory, vascular and immune cells of the melanoma microenvironment [87,89,90], thus contributing to the regulation of numerous processes related to melanoma malignancy. In particular, β3-AR is able to elicit stromal reactivity, sustain secretion of pro-inflammatory cytokines and drive de novo angio/vasculogenesis [89]; the same study has confirmed that β3-AR instructs melanoma cells to respond to environmental cell signals and to sense CAFs and macrophages enhancing their tumorigenic and stem-like traits. In regard to the immune regulation, pharmacological and molecular approaches with β-blockers (propranolol and SR59230A) and specific siRNA targeting of β2- or β3-ARs injected in B16-F10 melanoma-bearing mice, suggested an involvement of β3-AR subtype in the regulation of the immune-tolerance in melanoma microenvironment [90]. Indeed, β3-AR blockade increased the number of NK cells and lymphocytes CD8+ as well as their cytotoxicity, M1/M2 macrophages ratio and N1 granulocytes, while it abrogated Treg and MDSC sub-populations in tumor mass. By reducing the immune-suppressive and increasing the immune-competent subpopulations of cells in the TME, the β3-AR blockade proved the hypothesis that β3-ARs might play a role in the promotion of immune tolerance of melanoma. Taken together, these data confirm the pivotal role played by the β3-AR in regulating several biological processes related to melanoma progression (Figure 1).Recently, it has been demonstrated that in murine B16-F10 melanoma-bearing mice, the pharmacological β3-AR blockade was able to reduce the expression of cancer stem cell (CSC) markers, and to induce a differentiated phenotype of numerous hematopoietic progenitors recruited in TME [91]. The differentiation of melanoma and various stromal cells involved in pro-tumorigenic processes, brought about by the β3-AR blockade at the expense of stemness traits, thus hitting the metastatic potential of melanoma, could represent an efficacious strategy to counteract the progression to advanced stages of this malignancy.In human A375 melanoma cells, β3-AR stimulation through the selective agonist BRL37344 was able to induce a shift from an oxidative to a glycolytic metabolism, sustaining a metabolic process typical of tumor cells and known as Warburg effect [92]. Notably, β3-AR expression was found increased in melanospheres of A375 melanoma cells compared to the parental cell line, once again confirming that the β3-AR expression may correlate with pathways related to stemness features of tumor cells. β3-AR activation, indeed, induced the expression of specific glycolytic enzymes, such as hexokinase 2 (HKII), monocarboxylate transporter-4 (MCT-4), and glucose transporter-1 (Glut-1), which reflected elevated glucose uptake and lactate overproduction, two key metabolites of the Warburg effect. Moreover, the β3-AR/UCP2 axis strongly affected the mitochondrial activity by reducing ATP synthesis and mitochondrial reactive oxygen species (mtROS) content in melanoma cells. All these effects were reverted by using the β3-AR antagonist SR59230A, highlighting the crucial role played by the β3-AR in regulating molecular signaling that sustain metabolic and energetic processes typical of cancer stem cells [92].Hypoxic induction of the β3-AR protein has been reported in murine B16-F10 and human A375 melanoma cells [86,89]. Hypoxia is a well-known condition of solid tumors, including melanoma, able to orchestrate at cellular level a complex program, which leads to pro-tumorigenic events [98,99]. The induction of β3-AR expression in a hypoxic microenvironment could suggest that tumor cells exploit the activation of β3-AR pathways to achieve aggressiveness features required in the tumorigenic process.Despite studies investigating the role of β3-AR at clinical level still not being available, the expression of this receptor has been confirmed in melanoma biopsies from different patients. An immuno-histochemical analysis for the expression of β3-AR has been assessed in different cutaneous human melanocytic lesions including common and atypical nevi, in situ primary melanoma, superficial spreading melanoma, nodular melanoma, cutaneous, and lymph-nodal metastatic melanoma. Although β3-AR was expressed in all examined melanocytic lesions, its expression level, taking into account both staining intensity and percentage of positive cells, was significantly higher in malignant compared to benign lesions [89]. Importantly, in these biopsies β3-AR was found expressed also in stromal, endothelial and inflammatory cells of the TME, in accordance with the data obtained at preclinical level. These data clearly suggested that β3-AR expression correlates with melanoma malignancy features in human melanocytic lesions.Therapeutic options for patients with advanced-stage melanoma have been increased in the last years, and approval of new therapeutic agents such as the checkpoint inhibitors (anti-CTLA4 and anti-PD1 antibodies) and BRAF/MEK inhibitors has opened new expectations for survival. Moreover, early clinical data in a small patient population suggests that targeted therapy with BRAF/MEK inhibitors may work in synergy with checkpoint inhibitors and this triplet therapy may improve survival in patients with metastatic melanoma [100]. Despite the therapeutic improvement, some melanomas still remain unresponsive. Therefore, the discovery of new therapeutic strategies, especially for advanced melanoma, is still necessary not only to improve the cure rate but also the quality of life of these patients.During the years, several studies have accumulated evidence that, in melanoma, both tumor cells and non-tumor stromal cells cooperate in a complex signaling network to sustain tumor growth and progression. This meticulously orchestrated arrangement often underlies the onset of resistance mechanisms or the recurrence of the disease in some therapeutic regimens, especially in those that hit targets involved in a single pathway of the entire network. Accordingly, identifying a therapeutic target involved in regulation of multiple pro-tumor signaling pathways could represent a successful approach.In light of this, preclinical data summarized in this review have clearly suggested that β3-AR is able to modulate the activity of different cells in the melanoma microenvironment and, consequently, its blockade exerts an important anti-tumor action by affecting multiple pro-tumor signaling pathways. Even though further investigations are needed, especially at clinical level, these first experimental evidences highlight the functional role of the β3-AR subtype in melanoma malignancy, and suggest β3-AR as a therapeutically valid target to counteract melanoma progression.Conceptualization, L.F., G.B. and M.C.; writing—original draft preparation, L.F., G.B. and V.D.; writing—review and editing: C.F. and M.C. All authors have read and agreed to the published version of the manuscript.Fondazione Meyer provided financial resources.The authors declare no conflict of interest.Schematic representation of β3-AR-regulated processes in tumor and stromal cells of melanoma microenvironment (TME). A complex network of interaction and crosstalk between tumor and stromal cells of the TME, sustained through β3-AR-enanched processes, promotes melanoma progression. (CAFs = cancer-associated fibroblast; ECM = extracellular matrix; HSCs = hematopoietic stem cells; M2 = macrophage type 2; MSCs = mesenchymal stem cells; SNS = sympathetic nervous system). Created with BioRender.Studies Describing β3-AR Involvement in the Development of Different Cancers.AR = adrenergic receptor; iNOS = inducible NO synthase; NB = neuroblastoma; NO = nitric oxide; TME = tumor microenvironment.

Med-MDPI/cancers/cancers-12-06-01416.txt

@@ -0,0 +1 @@
+Oncolytic virotherapy is a promising antitumor therapeutic strategy. It is based on the ability of viruses to selectively kill cancer cells and induce host antitumor immune responses. However, the clinical outcomes of oncolytic viruses (OVs) vary widely. Therefore, we performed a meta-analysis to illustrate the efficacy and safety of oncolytic viruses. The Cochrane Library, PubMed, and EMBASE databases were searched for randomized controlled trials (RCTs) published up to 31 January 2020. The data for objective response rate (ORR), overall survival (OS), progression-free survival (PFS), and adverse events (AEs) were independently extracted by two investigators from 11 studies that met the inclusion criteria. In subgroup analyses, the objective response rate benefit was observed in patients treated with oncolytic DNA viruses (odds ratio (OR) = 4.05; 95% confidence interval (CI): 1.96–8.33; p = 0.0002), but not in those treated with oncolytic RNA viruses (OR = 1.00, 95% CI: 0.66–1.52, p = 0.99). Moreover, the intratumoral injection arm yielded a statistically significant improvement (OR = 4.05, 95% CI: 1.96–8.33, p = 0.0002), but no such improvement was observed for the intravenous injection arm (OR = 1.00, 95% CI: 0.66–1.52, p = 0.99). Among the five OVs investigated in RCTs, only talimogene laherparepvec (T-VEC) effectively prolonged the OS of patients (hazard ratio (HR), 0.79; 95% CI: 0.63–0.99; p = 0.04). None of the oncolytic virotherapies improved the PFS (HR = 1.00, 95% CI: 0.85–1.19, p = 0.96). Notably, the pooled rate of severe AEs (grade ≥3) was higher for the oncolytic virotherapy group (39%) compared with the control group (27%) (risk difference (RD), 12%; risk ratio (RR), 1.44; 95% CI: 1.17–1.78; p = 0.0006). This review offers a reference for fundamental research and clinical treatment of oncolytic viruses. Further randomized controlled trials are needed to verify these results.Cancer is a common disease globally that seriously affects human health. The USA, for instance, projects to have 1,806,590 and 606,520 new cancer cases and cancer deaths, respectively, in 2020 [1]. Although traditional treatment methods such as radiotherapy, chemotherapy, and targeted drugs are preferred in cancer treatment, their disadvantages include severe adverse events, development of drug resistance, and cross-resistance [2,3]. Therefore, the development of more effective cancer treatment strategies is urgently needed. Oncolytic viruses (OVs) are natural or artificially modified viruses that selectively replicate in and destroy cancer cells; hence, they represent a promising approach for antitumor therapy [4,5]. Oncolytic viruses generally exert antitumor effects by two mechanisms, namely, the selective killing of tumor cells, and induction of antitumor immunity [6]. To achieve specificity for tumor cells, key proteins required by OVs to infect the host are first modified to reduce infection of normal tissues [7,8,9]. Besides, oncolytic viruses utilize signaling pathways such as p53, epidermal growth factor receptor (EGFR)/Ras, and protein kinase R (PKR) to target tumor cells for selective expansion [10,11,12,13]. OVs can also kill tumor cells by triggering the expression of the suicide gene [14,15]. The key steps employed by OVs to transform “cold tumors” into “hot tumors” and activate antitumor immune responses include targeted replication, the release of tumor-associated antigens through oncolysis, upregulation of chemokines and danger signals, recruitment of dendritic cells and lymphoid cells, and upregulation of immune checkpoint molecules [16,17,18].Oncolytic viruses are either RNA or DNA viruses. RNA viruses such as reoviruses, paramyxoviruses, and picornaviruses, which encode only a few genes, often undergo rapid proliferation and lysis of tumor cells [5,18,19,20]. On the other hand, oncolytic DNA viruses such as herpes viruses, adenovirus, or poxviruses allow for the insertion of multiple foreign genes but are slower in replication and amplification [5,21,22]. The structure, gene components, expression strategies, and antineoplastic mechanisms are therefore different between the two types [23]. Talimogene laherparepvec (T-VEC), which is an oncolytic herpes virus type I, is presently the only oncolytic virus approved by the Food and Drug Administration. The success of T-VEC in the treatment of melanoma has further promoted the research of oncolytic viruses. With the increased number of clinical studies on oncolytic viruses, the efficacy and safety of oncolytic viruses have drawn much attention. Clinical trials of oncolytic viruses in combination with chemotherapeutic drugs, radiotherapy, and immune checkpoint inhibitors have shown massive progress in cancer treatment [5,16,24]. In particular, the combination of oncolytic virus and immune checkpoint inhibitors has yielded good results in melanoma [25]. Although many oncolytic viruses exist, a real champion among the oncolytic viruses has not yet emerged. In addition, no systematic review has been conducted on the efficacy and safety of oncolytic viruses in randomized controlled trials.In this meta-analysis, we included the following viruses: T-VEC (herpes virus) [26,27], pelareorep (reovirus) [28,29,30,31,32,33], NTX-010 (seneca valley virus; picornavirus) [19], Ad5-yCD/mutTKSR39rep-ADP (adenovirus) [34], and pexastimogene devacirepvec (Pexa-Vec; poxvirus) [35]. We first evaluated the efficacy of oncolytic virus from objective response rate (ORR), overall survival (OS), and progression-free survival (PFS); then we analyzed severe adverse events (grade ≥3) and detailed adverse events (AEs). Overall, we conducted this meta-analysis to investigate the effectiveness and safety of oncolytic viruses in randomized controlled trials to provide insights for fundamental research and clinical treatment.A systematic search was conducted in EMBASE, PubMed, and Cochrane databases for studies published up to 30/1/2020. The search terms included: “oncolytic viruses”, or “viruses, oncolytic”, or “oncolytic virus”, or “virus, oncolytic”, or “oncolytic virotherapy”, or “oncolytic virotherapies”, or “virotherapies, oncolytic”, or “virotherapy, oncolytic”, or “oncolytic virus therapy”, or “oncolytic virus therapies”, or “therapies, oncolytic virus”, or “therapy, oncolytic virus”, or “virus therapies, oncolytic”, or “virus therapy, oncolytic”. There was a language restriction of English in the search, and we followed the PRISMA guidelines for randomized controlled trials (RCTs) to conduct the meta-analysis [36].We included studies in the meta-analysis if they met the following inclusion criteria: (1) the studies were randomized controlled trials in cancer patients treated with an oncolytic virus; (2) the articles had at least one of the following outcomes: objective response rate (ORR), overall survival (OS), progression-free survival (PFS), or adverse events (AEs); (3) cancer patients in the control group received the control regimen without oncolytic virus. However, articles were excluded if: (1) they were conference abstracts, case reports, letters, meta-analyses, cohort studies, single-arm studies, reviews, animal studies, or in vitro studies; (2) patients in the control group received oncolytic virotherapy; (3) they included literatures with overlapping patients. Two independent investigators screened the potentially eligible articles by reading the titles and abstracts. Thereafter, the full text of all remaining studies was read to determine if they met the set eligibility criteria. Disagreements on study selection were resolved by discussion with other investigators.Two investigators independently read full texts of the included literatures and extracted the data. Any divergence of opinions concerning the extracted data was resolved through consultation. The extracted data included first author, publication, year, country, treatment, injection mode of OVs, types of cancer, the total number of patients, and clinical endpoints. The primary endpoints were ORR, OS, and PFS, while secondary endpoints included adverse events, which were evaluated using the National Cancer Institute—Common Terminology Criteria for Adverse Events (version 3.0 or 4.0). In addition, we carefully read supplementary materials of the included literatures to prevent any loss of information.Quality assessment was done by two independent investigators using the Cochrane risk of bias tool. The risk of bias parameters included the random sequence generation (selection bias), allocation concealment (selection bias), blinding of participants and personnel (performance bias), blinding of outcome assessment (detection bias), incomplete outcome data (attrition bias), selective reporting (reporting bias), and other bias. Each entry was determined as high-risk, low-risk, or unclear. If an item could not be assessed due to lack of information, it was considered as having an unclear risk of bias. Disagreements on quality assessment were resolved by consensus.Statistical analyses were performed using Review Manager (RevMan) 5.3 software and STATA 12.0. Results were presented as hazard ratios (HRs), risk ratios (RRs), or odds ratios (ORs) with 95% CI (confidence interval). Heterogeneity among RCTs was assessed by the Chi-square test and index of heterogeneity (I2). A mixed-effects model was used when heterogeneity was not significant (I2 < 50% or p-value > 0.1); otherwise, the random-effects model was performed. Publication bias was evaluated statistically via funnel plots, Begg’s test, and Egger’s test. Statistical significance was set at p < 0.05. A total of 9269 records were retrieved from PubMed, EMBASE, and Cochrane Library. A flow chart of study screenings and the election process is shown in Figure 1. From the remaining 6283 references screened after removing duplicates, 385 potentially eligible references were identified. Eventually, 11 RCTs that met the inclusion criteria were selected for full-text review. The risk of bias for the 11 included RCTs is shown in Figure 2. All the included RCTs were open-label trials. Most RCTs mentioned random allocation performed without using the random sequence generation method. Blinding was not performed because of the moral risk associated with the sham injection. In some RCTs [19,29,30,31,32,33,34,35], non-blinding had no significant effect on the efficacy or safety of oncolytic viruses; hence, they were judged as a low-risk factor.We included eleven studies with a total of 1452 patients in this meta-analysis. The characteristics and outcomes of RCTs are presented in Table 1 and Table 2. The OVs used in the included trials were T-VEC (n = 2), pelareorep (n = 6), NTX-010 (n = 1), Ad5-yCD/mutTKSR39rep-ADP (n = 1), and Pexa-Vec (n = 1). The types of tumors included melanoma, breast cancer, lung cancer, prostate cancer, hepatocellular carcinoma, colorectal cancer, pancreatic adenocarcinoma, and ovarian, tubal, or peritoneal cancer. The injection methods were either intratumoral or intravenous. Eleven included clinical trials of oncolytic viruses were conducted in the United States and Canada. Oncolytic DNA viruses include T-VEC, Pexa-Vec, and Ad5-yCD/mutTKSR39rep-ADP, and they all carry transgenes. T-VEC is modified by deleting the ICP47 gene and ICP34.5 gene (the herpes virus neurovirulence factor) to reduce viral pathogenicity and enhance selective tumor replication [37,38]. In addition, T-VEC could elicit human granulocyte macrophage colony-stimulating factor (GM-CSF) to recruit and activate antigen-presenting cells with subsequent induction of tumor-specific T-cell responses [13]. Pexa-Vec (JX-594) is a thymidine kinase gene-inactivated vaccinia virus engineered by expressing the transgenes, including GM-CSF and β-galactosidase; it selectively targets tumor cells with activation of the Ras/MAPK signaling pathway [35,39]. Ad5-yCD/mutTKSR39rep-ADP is adenovirus carrying two cytotoxic gene systems, cytosine deaminase (cytosine deaminase (CD)/5-fluorocytosine (5-FC) and herpes simplex virus thymidine kinase (HSV-1 TK)/valganciclovir (vGCV), and it can enhance the sensitivity of tumor cells to specific drugs and radiation [34].Oncolytic RNA viruses include pelareorep and NTX-010. Pelareorep is a human reovirus type 3 Dearing strain, which contains live, replication-competent reovirus, and has specific oncolysis with an activated Ras pathway [31,33]. Direct oncolysis of pelareorep led to release of “danger signals”, such as soluble tumor-associated antigens, viral pathogen-associated molecular patterns, and cell-derived damage-associated molecular patterns [16,40]. Therefore, direct oncolysis could result in generating innate and adaptive immune response to the tumor microenvironment and induces the antitumor immune response. Besides, NTX-010 (seneca valley virus) was a novel oncolytic picornavirus, which could target and lyse tumor cells [19,41].Ten RCTs reported objective response rate (ORR). Since differences were observed in efficacy among various OVs; we performed subgroup analysis on the ORR based on species, oncolytic DNA/RNA viruses, and injection mode. There was a statistically significant difference in ORRs between patients that received T-VEC (n = 2, OR = 4.05, 95% CI: 1.96–8.33, I2 = 52%, p = 0.0002). However, there was no significant difference in ORRs between patients treated with pelareorep (n = 6, OR = 1.06, 95% CI: 0.70–1.58, I2 = 6%, p = 0.79), NTX-010 (n = 1, OR = 0.25, 95% CI: 0.03–2.38, p = 0.23), and Pexa-Vec (n = 1, not estimable) (Figure 3). Objective response rate benefit was observed in patients that received oncolytic DNA viruses (n = 3, OR = 4.05, 95% CI: 1.96–8.33, I2 = 52%, p = 0.0002) but not in those treated with oncolytic RNA viruses (n = 7, OR = 1.00, 95% CI: 0.66–1.52, I2 = 13%, p = 0.99) (Figure 4). In the subgroup analysis for injection methods, results showed that the intratumoral injection arm produced significant improvement (n = 2, OR = 4.05, 95% CI: 1.96–8.33, I2 = 52%, p = 0.0002), but no significant improvement was found for the intravenous injection arm (n = 7, OR = 1.00, 95% CI: 0.66–1.52, I2 = 13%, p = 0.99) (Figure 5).Data regarding overall survival (OS) were available in ten RCTs, seven of which provided data for progression-free survival (PFS). Compared with the control group, patients treated with T-VEC had better OS (n = 2, HR = 0.79, 95% CI: 0.63–0.99, p = 0.04). However, treatment with pelareorep (n = 6, HR = 1.05, 95% CI: 0.84–1.31, p = 0.67), Pexa-Vec (n = 1, HR = 1.19, 95% CI: 0.77–1.83, p = 0.43), and NTX-010 (n = 1, HR = 1.49, 95% CI: 0.77–2.87, p = 0.24) did not improve the OS significantly compared to the control group (Figure 6). In addition, none of the patients benefited from T-VEC (n = 1, HR = 0.83, 95% CI: 0.56–1.23, p = 0.35), pelareorep (n = 5, HR = 1.07, 95% CI: 0.85–1.34, p = 0.59), and NTX-010 treatment (n = 1, HR = 1.03, 95% CI: 0.58–1.83, p = 0.92) in terms of PFS (Figure 7).Safety of oncolytic viruses remains a concern and most trials evaluate the safety aspect. The pooled risk ratio (RR) of severe adverse events (grade ≥3) was 1.44 (95% CI: 1.17–1.78, p = 0.0006, I2 = 13%) as shown in Figure 8a. The incidence of severe adverse events (AEs) in the oncolytic virus treatment group was higher than the control group (39% vs. 27%), with a pooled risk difference (RD) of severe AEs recorded at 0.12 (95% CI: 0.06–0.18, p = 0.0002, I2 = 37%) (Figure 8b); RD represents the rate of severe AEs attributed to oncolytic virotherapy. Furthermore, we analyzed detailed adverse events that may be associated with oncolytic virus treatment (Table 3). Patients treated with OVs had a higher risk for all-grade AEs such as fever (RR = 3.87, 95% CI: 2.15–6.69, p < 0.00001), neutropenia (RR = 1.66, 95% CI:1.21–2.29, p = 0.002), diarrhea (RR = 1.56, 95% CI:1.26–1.95, p < 0.0001), nausea (RR = 1.49, 95% CI: 1.28–1.74, p < 0.00001), vomiting (RR = 1.65, 95% CI: 1.27–2.14, p = 0.0002), chills (RR = 7.04, 95% CI: 4.64–10.66, p < 0.00001), flu-like symptoms (RR = 4.13, 95% CI:2.15–7.94, p < 0.0001), arthralgia (RR = 1.51, 95% CI: 1.09–2.12, p = 0.01), myalgia (RR = 1.97, 95% CI: 1.32–2.96, p = 0.001), extreme pain (RR = 1.50, 95% CI: 1.06–2.11, p = 0.02), headache (RR = 1.90, 95% CI: 1.42–2.53, p < 0.0001), and thrombocytopenia (RR = 2.74, 95% CI: 1.65–4.57, p = 0.0001). However, only neutropenia treatment yielded statistically significant severe adverse events (RR = 1.36, 95% CI: 1.03–1.80, p = 0.03).Publication bias was formally assessed using Begg’s test and Egger’s test. OS (Begg’s test, p = 0.283; Egger’s test, p = 0.126), PFS (Begg’s test, p = 0.548; Egger’s test, p = 0.307), and severe AEs (Begg’s test, p = 0.707; Egger’s test, p = 0.966) did not reveal any significant publication bias, but ORR (Begg’s test, p = 0.118; Egger’s test, p = 0.046 <0.1) had significant differences of publication bias. We made a sensitivity analysis by omitting a study to estimate meta-analysis of ORR. It suggested that omitting any one study had little effect on the overall result (each offset is minimal and between the upper CL limit and lower CL limits) (Figure 9). Therefore, the publication bias of ORR had limited impact on our conclusions.Oncolytic viruses possess the potential to kill cancerous cells (oncolysis); they also induce antitumor immune response through multiple mechanisms [42,43]. Such characteristics have made oncolytic virotherapy a promising immunotherapeutic approach for cancer patients. However, clinical trials have revealed that the presence of neutralizing antibodies in the blood prevents the oncolytic viruses (except reovirus) from replicating; activation of the immune system leads to rapid elimination of oncolytic viruses, and oncolytic viruses cannot target tumors due to physical parameters [5,44,45]. Furthermore, the best oncolytic virus, route of administration, prognosis of patients, and adverse reactions remain controversial. In this study, we extracted data for objective response rate (ORR), overall survival (OS), and progression-free survival (PFS) for in-depth analysis of the effectiveness of oncolytic virotherapy. Generally, T-VEC (OR = 4.05, 95% CI: 1.96–8.33) showed remarkable clinical efficacy of ORR. Interestingly, the objective response rate benefit was observed in patients treated with oncolytic DNA viruses (OR = 4.05, 95% CI: 1.96–8.33) but not in those treated with oncolytic RNA viruses (OR = 1.00, 95% CI: 0.66–1.52). This may be because DNA viruses carry many external genes with important immunomodulatory effects. In addition, DNA viruses express high fidelity DNA polymerases, which maintain the integrity of the viral genome and sufficient amplification [16,43]. Increasing evidence suggests that the antitumor effect of oncolytic viruses is not only dependent on pure oncolysis but also virus-induced antitumor immunity [16,46,47]. The three mechanisms in which oncolytic virus breaks the immune tolerance include: (1) after the virus infects tumor cells, it induces antigen-presenting cells (APCs) to infiltrate the tumor infection site; (2) the tumor antigen released after the virus lyses tumor cells and enhances the antigen presentation ability of APCs, thereby generating a specific immune response against the tumor antigen, forming a long-term antitumor immune response; (3) while OVs replicate in the tumor, they also express immunomodulatory factors, and they jointly participate in further amplification of antitumor immunity [48,49]. Since RNA viruses often replicate quickly and only possess few foreign genes [16,23], their antitumor effect is mainly dependent on oncolysis than immune activation. In respect of injection mode, cancer patients gained a significant objective response rate benefit from intratumoral injection (OR = 4.05, 95% CI: 1.96–8.33). Due to physical parameters and virus dilution, the targeting and effect of intravenous injection were unsatisfactory [5]. Although intratumoral injection can circumvent the above-mentioned problems, it is also limited by tumor type. From the survival data, only T-VEC (HR = 0.79, 95% CI: 0.63–0.99, p = 0.04) could effectively prolong overall survival (OS) of cancer patients. Pelareorep, Pexa-Vec, and NTX-010 were not statistically significant for OS. Moreover, no oncolytic virus affected progression-free survival (PFS) (HR = 1.00, 95% CI: 0.85–1.19). In patients with metastatic breast cancer, the median survival time of the experimental group (17.4 months) treated with pelareorep was remarkably longer than that of the control group (10.4 months). The HR of overall survival was 0.65 (80% CI: 0.46–0.91, p = 0.10). This suggests that pelareorep may be a new promising drug for metastatic breast cancer; more RCTs are, however, needed to validate it. Oncolytic viruses are generally considered safe. However, the oncolytic virotherapies were associated with specific risks in this meta-analysis. The pooled risk ratios (RR) and risk difference (RD) of severe adverse events (AEs) were 1.44 (95% CI: 1.17–1.78, p = 0.0006) and 0.12 (95% CI: 0.06–0.18, p = 0.0002), respectively, indicating such therapies carry risks that should not be ignored. Any-grade AEs with an incidence greater than 10% included fever (48.90%), neutropenia (63.01%), febrile neutropenia (25.18%), leukopenia (71.23%), diarrhea (28.78%), nausea (45.24%), vomiting (27.84%), chills (45.84%), fatigue (55.35%), flu-like symptoms (31.29%), decreased appetite/anorexia (25.91%), arthralgia (19.01%), myalgia (18.42%), extreme pain (20.98%), headache (24.11%), cough (21.66%), and thrombocytopenia (54.79%). Severe AEs with an incidence greater than 5% included neutropenia (40.36%), febrile neutropenia (15.52%) leukopenia (26.61%), fatigue (6.836%), and thrombocytopenia (10.09%). In the one-sided test, statistically significance of high-grade flu-like symptoms (1.23%), cellulitis (5.822%) of any-grade, and decreased appetite/anorexia (25.91%) of any-grade were observed. Detailed severe AEs have not been reported yet, and may be due to the loss of follow up, leading to underestimation.Our meta-analysis had the following limitations. First, we did not consider tumor types because of the insufficient number of RCTs to analyze same cancer. Secondly, in the subgroup analysis of objective response rate, there were few RCTs about oncolytic DNA viruses and intratumoral injection, and the conclusion needs more research to verify. Besides, the effective oncolytic virus was T-VEC. Therefore, the analysis results of the objective response rate may be affected by it. Finally, the heterogeneity of adverse events was biased upward since a wide range of oncolytic viruses was included. This review may provide new ideas for further research on oncolytic viruses to address the remaining challenges. We believe that oncolytic virotherapy will play an increasingly important role in cancer therapy with the increase of number of studies conducted.In conclusion, the results of our meta-analysis showed that the objective response rate benefit was observed in oncolytic DNA viruses and intratumoral injections. Currently, only patients treated with T-VEC can prolong overall survival. Besides, our meta-analysis revealed that occurrence of severe adverse events associated with oncolytic virotherapy cannot be ignored. More qualitative RCTs are needed to test the efficacy and safety of oncolytic viruses.Study design: Z.L. and Q.L.; data extraction, quality assessment, and data analysis: Z.L., Z.J., Y.Z., and Q.L.; manuscript writing and edition: Z.L. and Y.Z., Q.L. and X.H. revised the manuscript for its integrity and accuracy. Q.L. and Z.L. approved the final version of this manuscript and take responsibility for its content. All authors have read and agree to the published version of the manuscript.This study was supported by the National Natural Science Foundation of China (31760261 and 31660035), the Science and Technology Research Project of Jiangxi Provincial Education Department (60224), the Key Research and Development Projects of Jiangxi Natural Science Foundation (20192BBG70067), National Innovation and Entrepreneurship Program for College Students (20190403070), and the Key Projects of Jiangxi Natural Science Foundation (20171ACB20003).The authors declare that they have no conflict of interestPRISMA flow diagram of randomized controlled trials (RCTs) of patients treated with oncolytic virus.Assessment of risk of bias for 11 included randomized controlled trials.Forest plot of the pooled odds ratios (ORs) for objective response rate (ORR) in different oncolytic virus species.Forest plot of the pooled odds ratios (ORs) for objective response rate (ORR) of oncolytic DNA viruses and oncolytic RNA viruses.Forest plot of the pooled odds ratios (ORs) for objective response rate (ORR) of intratumoral and intravenous injections.Forest plot of the pooled hazard ratios (HR) for overall survival (OS).Forest plot of the pooled hazard ratios (HR) for progression-free survival (PFS).Forest plot of severe adverse events (grade ≥3): (a) the pooled risk ratios (RR); (b) the pooled risk difference (RD).Sensitivity analysis of ORR.Characteristics of the RCTs included in this meta-analysis.EG, experimental group; CG, control group; NR, not reported; BSC, best supportive care; IMRT, intensity modulated radiation therapy; IT, intratumoral; IV, intravenous.Summary of outcomes in the selected RCTs.EG, experimental group; CG, control group; HR, hazard ratio; OS, overall survival; PFS, progression-free survival; NR, not reported; CI, confidence interval.Adverse events of interest.*, statistically significant value; 95% CI, 95% confidence interval; RR, risk ratio; NA, not available; I2, index of heterogeneity; EG, experimental group.

Med-MDPI/cancers/cancers-12-06-01417.txt

@@ -0,0 +1 @@
+Although the 5-year survival rate of patients diagnosed with nonmuscle invasive bladder cancer (NMIBC) has reached 85%, more than 50% of patients suffer from frequent recurrences. To identify molecular targets associated with recurrence of NMIBC, we analyzed gene expression data and found that FOXM1 and FANCD2 were involved in recurrence. Therefore, we investigated how these genes were involved in the mechanism of recurrence and confirmed their usefulness as biomarkers. Investigation have shown that FOXM1 directly regulated the transcription of FANCD2, which is the key gene of the Fanconi anemia (FA) pathway. Depletion of FOXM1 resulted in DNA repair defects in the FA pathway and in decreased resistance to chemotherapy. Thus, the FANCD2-associated FA pathway activated by FOXM1 is an important mechanism involved in chemotherapy-related recurrence. In conclusion, FOXM1 and FANCD2 can be used as prognostic factors that are associated with high risk of recurrence and with anticancer drug resistance properties in NMIBC patients.Bladder cancer (BC) is the fifth most frequent cancer among men, with an estimated 549,393 new diagnoses and 199,922 BC deaths per year worldwide (2018) [1]. BC is classified into categories as nonmuscle invasive bladder cancer (NMIBC) or muscle invasive bladder cancer (MIBC) according to the presence or absence of muscle layer invasion [2]. More than 70% of patients are diagnosed with NMIBC and are grouped into low, intermediate, and high risk according to the EORTC (The European Organization for Research and Treatment of Cancer) risk stratification table [3,4]. Among the three groups, high-risk NMIBC (T1, with high grade/G3, and/or carcinoma in situ (CIS)) results in poor prognosis and additional intravesical bacillus Calmette-Guerin (BCG) and anticancer drug treatments (doxorubicin (DOX) and mitomycin C (MMC)) are recommended after surgery [5]. Despite these treatments, more than 50% of patients with NMIBC develop recurrence, and 10% to 30% of recurrent patients progress to MIBC [6]. These reports indicate that traditional indicators such as tumor grade, T stage, and CIS used in the diagnosis of BC patients are not sufficient [7]. It is essential that a new method be developed to complement the present diagnostic methods and provide insight into the molecular mechanism of recurrence. Yet, the precise mechanism of recurrence has not been clarified.A recent hypothesis in recurrence is that small numbers of anticancer drug-resistant cancer cells in a tissue survive and proliferate to cause recurrence after chemotherapy [8,9]. To explore the biomarkers of recurrence in bladder cancer cell lines and bladder cancer tissues, we wanted to find the relationship between drug resistance and recurrence of bladder cancer. According to recent reports, anticancer drug resistance is caused by an increase in drug efflux, anti-apoptosis activity, and DNA damage repair [10,11]. Most of the current anticancer drugs used to treat cancer patients are designed to cause DNA damage and cell death [12]. However, because cancer cells use various DNA damage repair mechanisms to repair DNA damage and prevent cell death, the inhibition of DNA damage repair in cancer cells has emerged as an important issue for effective chemotherapy [12].In previous studies, we found that patients with high expression of FOXM1 have a high risk of recurrence, suggesting FOXM1 could act as a recurrent biomarker [13]. FOXM1 is a transcription factor known to regulate the progression of the G2/M phase [14]. However, recent studies have shown abnormal overexpression of FOXM1 in various cancer tissues, such as those of the bladder, liver, prostate, brain, breast, lung, colon, pancreas, skin, cervix, ovary, and mouth [15]. FOXM1 affects the abnormal functions of cancer cells, including proliferation, cell cycle progression, apoptosis, angiogenesis, and DNA damage repair [16,17,18,19,20]. FOXM1 is associated with cancer metastasis, recurrence, and resistance to various chemotherapeutic drugs, such as DOX, epirubicin, and MMC [21,22]. FOXM1 is known to regulate the transcription of various DNA repair factors to increase homologous recombination (HR), nonhomologous end joining (NHEJ), base excision repair (BER), and mismatch repair [23]. These reports suggest FOXM1 as a DNA repair regulator associated with resistance to various chemotherapeutic agents and proliferation.DNA interstrand crosslinking (ICL) is known to be repaired by the FA pathway, which is one of the DNA repair pathways [24]. MMC, which is used in chemotherapy for high-risk NMIBC patients, induces ICL and apoptosis, thereby eliminating cancer cells [25,26]. The FA pathway is a complex DNA recovery pathway known to recruit various DNA repair pathways, such as nucleotide excision repair (NER), homologous recombination repair (HRR), and translation synthesis (TLS) to restore ICLs [24]. FOXM1 has been reported to be involved in many DNA repair processes, but there have been few direct reports of the FA pathway. However, our previous study of gene expression profile analysis identified FOXM1 and FANCD2 as recurrent biomarkers of bladder cancer [13]. Among the FA pathway factors, FANCD2 is ubiquitinated by the FA core complex and is known as a key factor in completing ICL repair [24]. During the FA pathway, a total of 19 FA proteins work together, and the FANCD2–FANCI protein complex restores ICL in the cell cycle S phase [27]. Therefore, the expression of FANCD2 is expected to contribute to the resistance to MMC by ICL repair. However, the association of FANCD2 with FOXM1 and its transcriptional control mechanism is unclear.Herein, we analyzed whether the transcription factor FOXM1 directly regulates the expression of FANCD2 and whether the increased expression of FOXM1 affects ICL repair by the FA pathway. We analyzed whether the transcription factor FOXM1 directly regulates the expression of FANCD2 and whether the expression of FOXM1 affects ICL recovery by the FA pathway. Through this study, we identified genes associated with NMIBC recurrence and chemotherapy resistance. These genes can be used as prognostic biomarkers for recurrence and anticancer drug resistance and have also revealed potential molecular mechanisms that could be the basis for developing new therapies.In our previous studies, we showed that FOXM1 is highly associated with recurrence of NMIBC [13]. Therefore, we confirmed the link between recurrences in high and low FOXM1 expression groups. We found that recurrence was significantly increased in the high expression group of FOXM1 compared to the low expression group of FOXM1 in both Korean cohorts (GSE13507) and European cohorts (GSE5479) (Figure 1A). Next, we identified the genes associated with FOXM1 to determine the pathway through which the expression of FOXM1 affects recurrence. As a result, 509 genes were found to be associated with FOXM1 (Figure 1B). Hierarchical clustering analysis with 509 genes divided the NMIBC patients into two subgroups based on FOXM1 expression: a FOXM1-low cluster (Cluster 1) and a FOXM1-high cluster (Cluster 2) (Figure 1B). In addition, we divided the two groups according to the expression of FOXM1 into clinical factors to identify the recurrence prognosis. When the stage was divided into Ta and T1, recurrence significantly increased in the FOXM1-high group compared with the FOXM1-low group. Similarly, the same results were obtained when the grade was divided (Figure S1A,B).We conducted a gene-to-gene network analysis based on the 509 genes associated with the expression of FOXM1. Then we performed a functional enrichment test and assessed the genes with Ingenuity Pathway Analysis. (Figure S2 and Table S1). As a result, FANCD2, which is known as a key factor in the FA pathway, was confirmed to be related to FOXM1 and is associated with its expression (Figure 1B,C). It was also confirmed that there was a significant positive correlation between FOXM1 and FANCD2 genes (Figure 1D). Risk scores of two patient subgroups were calculated using expression levels of FOXM1 and FANCD2 (Figure 1E). The area under the curve (AUC) by receiver operating characteristic (ROC) analysis was performed, and the proportion of recurrence-free survival in the good- and poor-prognosis groups was estimated in GSE13507 (Figure 1F). These results indicate that FOXM1 and FANCD2 may interact with each other and affect the recurrence of bladder cancer.We investigated the modulation of FANCD2 expression by FOXM1 to determine the relationship between the two proteins. Inhibition of FOXM1 expression using siFOXM1 in both 5637 and KU7 bladder cancer cell lines significantly reduced FANCD2 expression at both the mRNA and protein levels (Figure 2A,B and Figure S3). Lentiviral particles containing a shRNA expression vector for FOXM1 or control NTS were transduced into KU7 and 5637 cells to produce stable FOXM1 knockdown (shFOXM1) and control (shNTS) cell lines. FANCD2 levels were reduced by knockdown of FOXM1, which was confirmed in both cell lines by qRT-PCR and Western blotting analysis (Figure S4A,B and Figure S5).Next, we analyzed the sequence of the FANCD2 promoter region to determine whether the transcription factor FOXM1 binds to the promoter region of FANCD2 and directly regulates transcriptional activity. We identified one putative FOXM1 binding site in the promoter region of FANCD2 and constructed a luciferase vector with the FANCD2 promoter (Figure 2C). When this vector was transfected into the 5637 and KU7 bladder cancer cell lines, expression of the reporter vector was induced. However, additional siFOXM1 treatment significantly reduced the transcriptional activity of FANCD2 (Figure 2D). To further confirm that FANCD2 regulates the expression of FOXM1, FANCD2 was overexpressed or knocked out in the 5637 and KU7 cell lines. The results showed that FOXM1 expression was not altered, indicating that FANCD2 does not significantly affect FOXM1 (Figure S6A,B and Figure S7). These results show that the expression of FANCD2 is regulated by FOXM1. Immunoprecipitation was used to examine the FOXM1 binding affinity for the FOXM1 promoter region, which would reveal whether FOXM1 directly regulates FANCD2. Control siRNA or siFOXM1 were transfected into KU7 cells, and immunoprecipitation was performed using control IgG and FOXM1 antibodies. The result is shown in Figure 2E; we found that FOXM1 binds directly to the locus I region in the promoter of FANCD2. These results demonstrate that FOXM1 binds directly to the promoter region of FANCD2 and regulates its expression.Many anticancer drugs, including MMC, kill cancer cells by causing DNA damage. The ability of cancer cells to exhibit excessive activity of DNA repair mechanisms is considered to be a cause of cancer chemoresistance [10]. Therefore, we examined cell survival in cells transfected with control scRNA or siFOXM1 and in cells stably expressing shNTS or shFOXM1, and we also examined the correlation between MMC resistance and FOXM1 expression (Figure 3A,B). 5637 and KU7 cell lines were treated with various concentrations of MMC (0, 50, and 500 nM), and cell viability was examined by colony formation assay and MTT assay. The results revealed that when FOXM1 was inhibited, the survival rate of cells was remarkably decreased (Figure 3A,B).FOXM1 suppression has been confirmed to reduce resistance to MMC, suggesting that FOXM1 is associated with DNA repair activities such as the FA pathway. From the above results, we confirmed that FOXM1 regulates the expression of FANCD2 (Figure 2) and that the expression of these two genes influences recurrence (Figure 1). We therefore investigated whether FOXM1 is involved in anticancer drug resistance through the direct control FANCD2, which is involved in the FA pathway. We conducted experiments to determine whether there is a relationship between the expression of these two genes and resistance to anticancer drugs. First, we investigated whether the expression of FANCD2 is regulated by siFOXM1 treatment. Both cell lines were transfected with control scRNA or siFOXM1 and then were treated with various concentrations of MMC to confirm the reduction of FANCD2 expression by siFOXM1 (Figure 4). The expression of FANCD2 mRNA and protein decreased significantly following inhibition of FOXM1 in 5637 and KU7 cells that were treated with MMC (Figure 4A,B and Figure S8). Furthermore, an increase in cell viability due to FANCD2 overexpression was identified at 48 hours after the treatment. Cell viability increased when FANCD2 was overexpressed in the shFOXM1 cell line even after treatment with 50 and 500 nM MMC for 48 hours. (Figure S9A–C). FANCD2 is a key gene in the FA pathway, and the deletion of this gene affects the activity of DNA repair by the FA pathway [27,28]. Therefore, we determined how FOXM1, which directly affects DNA repair activity, regulated the expression of FANCD2. After we transfected bladder cancer cell lines with control scRNA or siFOXM1, MMC was used to treat cells, and FANCD2-foci were observed during the DNA repair process. The results showed that the formation of FANCD2-foci was significantly reduced by suppression of FOXM1 expression in 5637 and KU7 bladder cancer cell lines (Figure 4C). These results suggest that depletion of FOXM1 expression leads to a decrease in the expression of FANCD2, and consequently decreases the DNA repair response of FANCD2 to MMC. The Fanconi anemia pathway is a mechanism to complete DNA repair by recruiting various DNA repair mechanisms such as nucleotide cleavage repair, mutagenesis, and homologous recombination. Therefore, to investigate whether FOXM1 directly affects the FA pathway, the activity of these various DNA repair mechanisms was measured after suppression of FOXM1 expression.We conducted a comet assay to investigate whether the inhibition of FOXM1 expression affects the DNA repair process of single strand breaks (SSBs) and double strand breaks (DSBs). DNA damage was induced by treating 5637 and KU7 BC cell lines with MMC, and the degree of DNA repair was measured by the tail length in a comet assay. Inhibition of FOXM1 expression increased comet tail length in both alkaline gels (DSBs and SSB-detectable) and neutral gels (DSB-detectable) (Figure 5A,B). These results show that the repair of SSBs and DSBs caused by treatment with MMC is decreased by suppression of FOXM1 expression (Figure 5A,B).We also performed γ-H2AX staining to confirm that the suppression of FOXM1 reduces DNA damage repair. Cells were exposed to MMC for 4 hours, and γ-H2AX foci were measured after 0 hours (treatment) and 24 hours (release) of repair time (Figure 5C). DNA damage caused by MMC treatment was identified as γ-H2AX foci, and these foci were reduced by DNA repair after a 24-hour repair period in the scRNA control-treated cells (Figure 5C). On the other hand, it was confirmed that the siFOXM1 group remained, the number of γ-H2AX foci was over the same period of repair in 5637 and KU7 bladder cancer cell lines (Figure 5C). Additionally, we measured the homologous recombination repair (HRR) activity to determine whether the HRR involved in the last step of the Fanconi anemia pathway was regulated by the inhibition of FOXM1 expression. We found that the activity of HRR significantly decreased by inhibiting the expression of FOXM1 (Figure 5D).Next, we measured chromosomal aberrations known to occur when there is a repair problem with the FA pathway. The effect of FOXM1 inhibition on chromosomal aberrations induced by MMC treatment was examined. The scRNA and siFOXM1 groups showed a similarly low number of chromosomal aberrations (Figure 5E). However, when treated with MMC, there was a significant increase in chromosomal aberrations in both groups, especially in the siFOXM1 group, compared with that of the scRNA control group (Figure 5E). This suggests that depletion of FOXM1 results in an inability to repair chromosomal aberrations to the level of the control.Taken together, these results suggest that FOXM1 is involved in various DNA damage repair pathways associated with the FA pathway that are induced by MMC treatment. We then assessed whether FOXM1 and FANCD2 affect the recurrence of bladder cancer by analyzing their expression in clinical tissues. We examined the gene expression levels of FOXM1 and FANCD2 in primary and recurrent tumor tissues in the NMIBC patient group (Figure 6). Figure 6A shows the comparison between the primary and recurrent tumor groups in the NMIBC patient groups. The results showed that the expression of these genes was significantly higher in recurrent cancer tissues than it was in primary cancer tissues (p = 0.004; FOXM1 and p = 0.001; FANCD2 by two sample t-test, Figure 6A). To determine whether FOXM1 and FANCD2 were associated with recurrence according to the expression in primary cancer tissue, primary cancer tissue was classified into three NMIBC patient groups (no-recurrent primary, recurrent primary, and recurrent tumor groups), and the mRNA expression patterns were determined (Figure 6B). The results revealed that FOXM1 expression was not significantly different between the two primary tumor tissues, but FANCD2 expression was significantly higher in recurrent tumor tissues than it was in nonrecurrent tumor tissues (p = 0.486; FOXM1 and p = 0.002; FANCD2, Figure 6B). In the three-group classification, the expression of these two genes was significantly higher in recurrent cancer tissues than it was in primary cancer tissues (p = 0.003; FOXM1 and p = 0.02; FANCD2, Figure 6B).Moreover, we determined the protein levels of FOXM1 and FANCD2 by IHC (Immunohistochemistry) in 57 bladder cancer samples (30 nonrecurrent primary tumors and 27 recurrent primary tumors) using a tissue microarray (Figure 6C). We identified 13 cases (44.8%) with low intensity FOXM1 or/and FANCD2 staining (IRS ≤ 3) and 16 cases (55.2%) with high intensity FOXM1 and FANCD2 staining (IRS > 3) in tissue microarray (TMA) samples of nonrecurrent tumors. In addition, there were 4 cases (15.4%) with low intensity FANCD2 or/and FANCD2 staining and 22 cases (84.6%) with high intensity FOXM1 and FANCD2 staining in the TMA of recurrent tumors (p = 0.018) (Figure 6C). Thus, the IHC results were correlated with clinicopathological and recurrence data. Although NMIBC is known to have a relatively high survival rate compared to MIBC, the incidence rate is expected to significantly increase with the increase in the elderly population [25]. However, NMIBC has a high recurrence rate of over 50% and requires repeated treatment [7]. Therefore, the cost of the treatment is inevitably high, which may result in a negative effect on the quality of life of the elderly population [25]. Therefore, if biomarkers could be used to predict the recurrence of BC, accurate diagnosis and effective treatment could be conducted, which would prevent predicted recurrence problems.Our previous studies have shown that recurrence significantly increases in NMIBC patients with high CCNB1 expression [13]. In addition, BCG, which is currently used in high risk NMIBC treatment, requires a definite treatment indicator, according to the current treatment standards, because of the risk of adverse effects in some patients with no therapeutic responses or serious side effects [2]. Interestingly, our results showed that treatment with BCG can be more effective in patients with high CCNB1 expression [13]. These results indicate that CCNB1 is a suitable biomarker since CCNB1 expression can also be used as a criterion to clearly indicate whether therapy should be performed [13]. However, the risk of side effects that may be caused by BCG still remain. The increased effect of BCG in the high CCNB1 expression group was found as a result of demonstrating that immunotherapy is effective in a group of patients not responding to conventional chemotherapy [13]. Therefore, if this group can increase the effectiveness of chemotherapy, then safe and effective treatment of BCG is possible.In this study, we identified FOXM1 as a driver gene that is strongly associated with bladder cancer recurrence and is known to regulate expression upstream of CCNB1. In addition, we could determine the recurrence rate of patients classified according to the expression of FOXM1 and FANCD2 (Figure 1 and Figure 2) and by using these genes as biomarkers in combination with CCNB1 [13], we could predict the prognosis of patients with NMIBC after the first surgery. We also investigated the pathway through which FOXM1 acts to interfere with anticancer drug therapy and the mechanism of cancer recurrence. We examined the mechanism of the FANCD2 transcriptional regulation by FOXM1 (Figure 3) and demonstrated that MMC-induced DNA damage is reduced by the inhibition of FOXM1 (Figure 4 and Figure 5). The transcriptional regulatory mechanism of FANCD2, a key element of the FA pathway, was confirmed. It is expected that the effect of MMC on chemotherapy can be increased by confirming decreased DNA repair activity as a result of decreased FANCD2 expression. Currently, the expression of FANCD2 did not increase by MMC treatment because MMC repressed DNA replication by inhibiting DNA replication and transcription to eliminate cancer cells. Our results demonstrate that the inhibition of the expression of FOXM1, the driver gene, in NMIBC recurrence can inhibit different anticancer resistance pathways. Although new drugs targeting FOXM1 have not yet been developed, natural substances that restrict FOXM1 and inhibitors have been identified [26,29]. Therefore, it is expected that more effective chemotherapy will be achieved if these substances are used in concert with anticancer drugs. In addition, NMIBC treatment will be more effective if anticancer drugs targeting FOXM1 are developed.Previous studies have shown that FOXM1 can regulate CCNB1 expression to induce cell proliferation and recurrence [30]. The study also showed that FOXM1 can increase anticancer drug resistance by regulating the expression of FANCD2, which regulates the Fanconi anemia pathway and consequently increases DNA repair. Through this, we propose that FOXM1 directly regulates the expression of CCNB1 and FANCD2 as an important pathway for recurrence and thereby is involved in cell proliferation and anticancer drug resistance.In summary, we confirmed that FOXM1 can modulate the DNA repair pathway by directly regulating FANCD2 transcription and that it regulates resistance to MMC. Finally, in the case of FOXM1 and FANCD2, we confirmed that the expression could be used as a biomarker to predict recurrence and survival rates.The 5637 and HEK-293T cells originated from the American Type Culture Collection (ATCC). KU7 and U2OS-DRGFP cells were provided by Ju-Seog Lee (The University of Texas MD Anderson Cancer Center, Houston, TX, USA). The 5637 cells were cultured in RPMI 1640 medium (HyClone, UT, USA) supplemented with 10% FBS (Fetal Bovine Serum, HyClone, UT, USA) and 1% penicillin/streptomycin (P/S, Gibco, NY, USA). HEK293T and KU7 cells were cultured in H-DMEM (HyClone) supplemented with 10% FBS and 1% P/S. The U2OS-DRGFP cells were grown in McCoy’s 5A (Modified) medium (Gibco, NY, USA), plus 10% FBS and 1% P/S. All cells were incubated at 37 °C under 5% CO2 in a humidified incubator. MMC, puromycin, and hexadimethrine bromide (polybrene) were purchased from Sigma (MO, USA) and were dissolved in sterile dH2O. pGL3-Basic and pRL-Renilla luciferase reporter plasmids were purchased from Promega (WI, USA). To generate the pGL3 Basic-FANCD2 promoter (from −3347 to –1), human genomic DNA was amplified by PCR using the indicated primer sets (Table S2). We purchased shRNA for FOXM1 from Sigma. The plasmids that stably expressed a shRNA against FOXM1 (shFOXM1) were established in a pLKO.1-TRC cloning vector (Addgene, MA, USA). We used a pLKO.1-puro nontarget shRNA plasmid (Sigma) for the negative control. pMD2.G and psPAX2 plasmids were purchased from Addgene (MA, USA). To construct pFANCD2 (FANCD2 overexpression vector), the coding sequence (CDS) of FANCD2 was inserted into the NheI/XhoI restriction enzyme sites of a pcDNATM6/V5-His A plasmid (Invitrogen, MA, USA). All constructs were verified using a DNA sequencing. Scrambled RNA (scRNA) was purchased from Shanghai GenePharma (Shanghai, China). siFOXM1 (5’-GGACCACUUUCCCUACUUU-3’) was synthesized by the ST Pham Oligo Center (Korea). Plasmid and siRNA transfection were conducted using jetPRIME reagent (Polyplus, NY, USA) according to the manufacturer’s protocol.All RNA was obtained from BC cells using RNAiso (TAKARA, Shiga, Japan) according to the standard protocol, and the synthesis of complementary DNA (cDNA) and qRT-PCR were performed using a PrimeScript RT reagent kit (TAKARA) according to the manufacturer’s instructions. cDNA was amplified using qRT-PCR with the indicated primer sets (Table S2) and CFX960 Optics Module (Bio-Rad, CA, USA). Amounts of mRNA were determined from the threshold cycle number with the expression of L19 as an endogenous control. All experiments were performed in triplicate and the values were averaged.Luciferase assays were performed as described previously [31]. They were carried out using a Dual-Luciferase Reporter Assay System (Promega). To measure luciferase activity, a Wallac Victor 1420 Multilabel counter (PerkinElmer, MA, USA) was used. All firefly luciferase data were normalized to Renilla luciferase activities, each experiment was replicated three times, and the values were averaged.We analyzed a gene expression dataset from a previous study (GSE13507) that involved 165 primary Korean BC tissues. Among the 165 tissues, 102 tissues were identified as primary NMIBC. Clinical data including recurrence-free survival were acquired from the Chungbuk National University Hospital (Cheongju, South Korea). To estimate recurrence values of a signature combined with FOXM1 and FANCD2 genes, we selected a strategy that, for the genes in its signature, used the Cox regression coefficient (prognostic index (PI)). Ingenuity Pathway Analysis (IPA) was used to analyze the gene network-based activation regulator. Other gene expression datasets of BC patients with NMIBC from hospitals in the Swedish southern healthcare region (GSE32894; the SSH cohort, n = 213), Skane University Hospital (GSE32549; the SUH cohort, n = 92) and RNA sequencing data set (TCGA-BLCA; n = 412) were used to identify FOXM1 expressed genes. All gene expression datasets were opened at the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database.HEK-293T cells were transfected with pMD2.G (envelope plasmid), psPAX2 (packaging plasmid), and shRNA expressing plasmid (shNTS or shFOXM1) using jetPRIME reagent. The medium was changed 12 hours later and was harvested from the cells after 24 hours. The 5637 and KU7 cell lines were transduced with medium containing lentiviral particles and added polybrene was added. The cells stably expressing shRNA against NTS (shNTS) or FOXM1 (shFOXM1) were selected puromycin (5 μg/mL). FOXM1 mRNA and protein expression in cell lines was confirmed using qRT-PCR and Western blot assays.Western blotting was performed as described previously [31]. Blots were conducted using the following antibodies: mouse-anti-β-Actin (A5441, Sigma), rabbit-anti-FOXM1 (Cat.A301-533A, Bethyl Laboratories, TX, USA), rabbit-anti-FANCD2 (Cat.NB100-182, Novus Biologicals, CO, USA). Horseradish peroxidase-linked goat anti-rabbit IgG polyclonal antibody (Cat.ADI-SAB-300-J, Enzo Life Sciences, NY, USA), and goat anti-mouse IgG polyclonal antibody (Cat.ADI-SAB-100-J, Enzo Life Sciences). Equal protein loading verified by detection of β-actin expression.ChIP assays were performed as described previously [32]. Immunoprecipitation was performed using a rabbit-anti-FOXM1 antibody (Cat.A301-532A, Bethyl Laboratories, TX, USA). The indicated primer sets used for PCR amplification were a primer set for the FOXM1 site at position (Table S2).siRNA-transfected BC cells were seeded in 6-well plates at 1000 cells/well and then were incubated for 24 hours. Afterwards, the cells were treated with the indicated concentration of MMC for 24 hours. Then, the cells were harvested and reseeded in 6-well plates at 100 cells/well. After 2 weeks of incubation, colonies were fixed with 4% paraformaldehyde for 15 minutes at room temperature before being washed with PBS (Phosphate-buffered saline). Crystal violet (0.5%; Sigma) was used to stain the fixed cells for 30 minutes, which was followed by washing the plates with dH2O. The plates were then left to dry overnight. Colonies were counted using Carl Zeiss Axiovert 40 CFL microscopy (Göttingen, Germany).MTT assays were performed as described previously [31]. Absorbance for each well was determined at 540 nm with a Wallac Vector 1420 Multilabel Counter (PerkinElmer, MA, USA). For each experimental condition, 3 wells were used.BC cells were seeded on coverslips coated with collagen (Sigma). Cells were treated both with 0.5 μM MMC, and control cells were untreated. After treatment, cells were washed with PBS and then fixed with 2% formaldehyde in PBS at room temperature (RT) for 20 minutes. After being washed with PBS, the process was followed by permeabilization with 0.5% Triton X-100 (Fluka, Buchs, Switzerland) in PBS at RT for 30 minutes. Then, cells blocked with 20% FBS were probed with the following antibodies in 5% FBS of PBS for 2 hours at RT: FOXM1 antibody (Cat.A301-533A, Bethyl Laboratories), FANCD2 antibody (Cat.NB100-182, Novus Biologicals, CO, USA), and phosphor-histone H2AX antibody (Cat.04-636, Millipore, MA, USA). After being washed with PBS, the cells were incubated with mouse and rabbit IgG-heavy and light chain antibodies (Cat.A90-116F, A120-101D4, Bethyl Laboratories) for 1 hour at RT and then were washed with PBS. Cells were costained with 100 μg/mL Hoechst 33342 and then were mounted with Vector Vectashield mounting media (Vector Laboratories, CA, USA). After fluorescence images were acquired using an LSM 700 confocal microscope (Carl Zeiss, Göttingen, Germany), the cells were counted in at least four randomly selected fields.BC cells (1 × 106 cells) were harvested after treatment with 0.5 μM MMC for 24 hours and combined with molten LMAgarose (Trevigen, MD, USA). Before being incubated at 4 °C in lysis solution (Trevigen) overnight, cells were embedded in low melting agarose on a glass slide. Under alkaline conditions, slides were then immersed in an alkaline unwinding solution (200 mM NaOH, 1 mM EDTA, pH >13) in the dark at room temperature for 40 minutes before being placed in an electrophoresis slide tray. In addition, 850 mL of cold alkaline electrophoresis solution (200 mM NaOH, 1 mM EDTA, pH >13) was added, and a set power of 20 volts was applied for 30 minutes. Slides were first washed in dH2O twice for 5 minutes each and then in 70% EtOH for 5 minutes. Under neutral conditions, the slides were submerged in 1 × neutral electrophoresis buffer (10 × neutral electrophoresis buffer diluted 1:10 in dH2O 60.57 g Tris-Base, 204.12 g sodium acetate, pH = 9.0 with glacial acid) for 30 minutes. The slides were placed in an electrophoresis slide tray with 850 mL of cold 1 × neutral electrophoresis buffer, and a set power of 20 volts was applied for 45 minutes. The slides were then submerged in DNA precipitation solution (7.5 M NH4AC, 95% EtOH) for 30 minutes at room temperature. After they were washed with 70% EtOH for 30 minutes, they were dried at 37 °C for 15 minutes. Slides were then stained with SYBR Green for 30 minutes in the dark. Slides were rinsed in dH2O and finally were mounted using Vectashield mounting medium. Nuclei were visualized by fluorescence microscopy. The percentage of DNA was quantitated for 50 cells by using a microscope.U2OS cells stably transfected with a DR-GFP plasmid vector were used for the HR assay. A total of 1 × 105 siRNA-transfected cells were plated in a 6-well plate. Twenty-four hours later, the I-SceI endonuclease plasmid was delivered into the cells by transfection for 24 hours. Then, the cells were washed with PBS and harvested using trypsin. Then, cells expressing GFP were sorted by flow cytometry FC500 (Beckman Coulter, Krefeld, Germany) at 520 nm to analyze the efficiency of HR repair using the CXP v2.1 program.For the detection of chromosomal aberration, siRNA transfected BC cells were treated with 0.5 μM MMC. Twenty-two hours later, the cells were exposed to 100 ng/mL colcemid (Sigma) for 2 hours. Then, the cells were treated with a hypotonic solution (75 mM KCl) for 20 minutes and then were fixed with 3:1 methanol/acetic acid. Slides were stained with a Giemsa solution (Sigma), and over 50 metaphase spreads were counted to detect aberrations. The relative number of chromosomal breaks and radials was calculated relative to scRNA or siFOXM1.IHC was performed on a subset of 57 BC tissues from nonrecurrent primary cancer patients and recurrent primary cancer patients. A tissue microarray (TMA) was created from 30 nonrecurrent tumors, and 27 recurrent tumors were used. IHC was performed with a panel of antibodies against 2 markers (FOXM1 and FANCD2). All stained slides were digitalized with an SL801 autoloader and a Leica SCN400 scanning system (Leica Microsystems; Concord, Ontario, Canada) at a magnification equivalent to 20×. The images were subsequently stored in a Slide Path digital imaging hub (Leica Microsystems) at the Vancouver Prostate Centre. Values were assigned on a 4-point scale for each image. Descriptively, 0 represented no staining, 1 represented a low but detectable degree of staining, 2 represented a low detectable degree of staining, 3 represented clearly positive cases, and 4 represented strong expression. IHC was quantified for staining intensity (0–4).Data are represented as the mean ± SEM of three independent experiments. Unpaired Student’s t-tests were used to analyze the dissimilarities between the groups. Categorical data were analyzed by Fisher’s exact test. The cumulative recurrence was calculated by the Kaplan–Meier method and the log-rank test. Analyses were performed using GraphPad Prism 7 software (GraphPad Software, Inc., CA, USA). Asterisks, as described in the figure legends (ns, not significant; *, p < 0.05; **, p < 0.01; and ***, p < 0.001), were used to illuminate the statistically significant p-values that were less than 0.05. In this study, we confirmed the direct relevance of FOXM1 to the major FA pathway for anticancer drug resistance. These findings raised the possibility of using FOXM1-FANCD2 expression level as a prognostic factor when considering anticancer drug treatment and risk of recurrence in NMIBC patients.The following are available online https://www.mdpi.com/2072-6694/12/6/1417/s1, Figure S1: Prognosis according to tumor stage and grade of two subgroups divided by FOXM1 expression in NMIBC patient cohort, Figure S2: Function enrichment test and network analysis, Figure S3: Uncropped blots showing all the western bands in 5637 and KU7 cell lines transfected with siFOXM1, Figure S4: The expression of FOXM1 and FANCD2 on knockdown stable cell lines, Figure S5: Uncropped blots showing all bands for stable knockdown of FOXM1 in KU7 and 5637 cell lines, Figure S6: No significance of the expression of FOXM1 by overexpressed FANCD2, Figure S7: Uncropped blots showing all the western bands in 5637 and KU7 cell lines transfected with pFANCD2 vector or siFANCD2, Figure S8: Uncropped blots showing all the western bands in 5637 and KU7 cell lines transfected with siFOXM1, treated to MMC, Figure S9: Increased cell viability due to overexpression of pFANCD2 in shFOXM1 cells, Table S1: Significantly enriched functions and their involved genes, Table S2: qPCR and plasmid construction primer set.Y.-G.R. and S.-H.L. Conceptualization and Writing the manuscript. Y.-G.R., J.-Y.M., W.Y.P., M.-S.J., T.N.K., W.-T.K. and Y.H.C. Data curation, Methodology. S.-K.K. and I.-S.C. Data curation, Formal analysis and Investigation. I.-S.C. Project administration and Supervision. S.-H.L. Funding acquisition, Project administration and Supervision. All authors have read and agreed to the published version of the manuscript.This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science, ICT & Future Planning [2017R1A2B2007836]. The authors declare no conflict of interest.Recurrent prognosis and identification of genes associated with FOXM1 in NMIBC patients. (A) Recurrence-free survival (RFS) in the high and low FOXM1 expression groups in the Korean and Denmark cohorts. (B) Expression patterns of genes highly associated with FOXM1. A total of 509 genes that underwent cluster analysis had a high expression pattern with FOXM1 (r > 0.5) in the Korean cohort (n = 102). (C) Gene to gene network analysis of FOXM1-correlated genes in NMIBC using IPA (Ingenuity pathway analysis). (D) Correlation analysis of FOXM1 and FANCD2 in NMIBC in NMIBC patient gene expression data. (E) Prognosis of the combination of FOXM1 and FANCD2. Risk scores of two patient subgroups were calculated by expression levels of FOXM1 and FANCD2 with GSE13507 (Korean cohort). Prognosis indicates the RFS of NMIBC. (F) ROC (receiver operating characteristic) curve for prediction of recurrence using the FOXM1 and FANCD2 signatures.Direct transcriptional regulation of FANCD2 by FOXM1. (A) scRNA and siFOXM1 were transfected into 5637 and KU7 cells. mRNA expression was measured using qRT-PCR. (B) FOXM1 and FANCD2 protein expression was measured by Western blot in scRNA- and siFOXM1-transfected 5637 and KU7 cells. (C) A schematic diagram of the FANCD2 promoter region (−3347/−1). The black bar represents the putative FOXM1 binding site (−2239/−2234) and the qChIP amplification locus (−2333/-2198). The white bar represents the nontarget sequence of the qChIP amplification locus (−1492/−1323). (D) FANCD2 promoter activity was measured using a luciferase assay in siRNA (scRNA or siFOXM1)-transfected BC (Bladder cancer) cells. pGL3 vector was used as a control. (E) FOXM1 binding affinity of the FANCD2 promoter region. siRNA or siFOXM1 was transiently transfected into KU7 cells and immunoprecipitated using FOXM1 antibody and rabbit IgG (control). The amount of chromatin was measured using qRT-PCR along with a target site primer (I) and a nontarget site (NTS) primer. (*, p < 0.05; **, p < 0.01; and ns, not significant).Analysis of MMC resistance following suppression of FOXM1 expression. (A) siRNA (scRNA or siFOXM1)-transfected 5637 and KU7 cells exposed to the indicated concentrations of MMC for 24 hours and seeded at 500 cells/well. After 10 days, colonies were stained with crystal violet and counted using a microscope. (B) The survival rate of 5637 and KU7 cells stably expressing shNTS or shFOXM1 was measured by MTT assay after exposure to the indicated concentrations of MMC (Mitomycin C) at the indicated times. (*, p < 0.05; **, p < 0.01; and *** p < 0.001).Control of FANCD2 expression and repair activity of FANCD2 following inhibition of FOXM1 and MMC treatment. siRNA (scRNA or siFOXM1)-transfected 5637 and KU7 cells were exposed to MMC for 24 hours. (A) FOXM1 and FANCD2 mRNA expression was measured by qRT-PCR. (B) FOXM1 and FANCD2 protein levels were analyzed by Western blot. (C) Nuclear foci formation of FANCD2 measured by immunofluorescence. Hoechst 33342 stained nuclear region. (*, p < 0.05; **, p < 0.01; and ***, p < 0.001).Measurement of DNA damage and repair activity in FOXM1-inhibited bladder cancer cells. To determine DNA damage and repair activity, siRNA (scRNA or siFOXM1)-transfected 5637 and KU7 cells were exposed to 0.5 μM MMC for 24 hours. A comet assay was performed to measure DNA damage. (A) The alkaline condition was used to measure single strand breaks. (B) The neutral condition was used to measure double strand breaks. γ-H2AX staining was performed to measure DNA recovery activity. (C) siRNA (scRNA or siFOXM1)-transfected 5637 and KU7 cells were exposed to 0.5 µM MMC for 4 hours. Then, the cells were incubated in fresh media for 24 hours. An HR assay was performed to measure HRR activity. (D) GFP (Green Fluorescent Protein)-positive cells were counted to measure HRR activity using FACS (Fluorescence-activated cell sorting) in siRNA (scRNA or siFOXM1) and I-sceI co-transfected U2OS-DRGFP cells. Chromosomal aberrations were analyzed to detect chromosomal abnormalities. (E) siRNA (scRNA or siFOXM1)-transfected 5637 and KU7 cells were exposed to 0.5 µM MMC for 24 hours, and chromosomal aberrations were measured by Giemsa staining. (*, p < 0.05; **, p < 0.01; *** p < 0.001; and ns, not significant).Expression of FOXM1 in NMIBC tissues. (A) Boxplot for two NMIBC patient groups (primary and recurrent tumor groups). (B) Boxplot for three NMIBC patient groups (nonrecurrent primary, recurrent primary, and recurrent tumor groups). (C) RFS (Recurrence free survival) of FOXM1 expression levels in two independent NMIBC cohorts. (**, p < 0.01; and ns, not significant).

Med-MDPI/cancers/cancers-12-06-01418.txt

@@ -0,0 +1 @@
+Authors share co-first authorship.Despite recent innovations and advances in early diagnosis, the prognosis of advanced gastric cancer remains poor due to a limited number of available therapeutics. Here, we employed pharmacogenomic analysis of 37 gastric cancer cell lines and 1345 small-molecule pharmacological compounds to investigate biomarkers predictive of cytotoxicity among gastric cancer cells to the tested drugs. We discovered that expression of CCNA2, encoding cyclin A2, was commonly associated with responses to polo-like kinase 1 (PLK1) inhibitors (BI-2536 and volasertib). We also found that elevated CCNA2 expression is required to confer sensitivity to PLK1 inhibitors through increased mitotic catastrophe and apoptosis. Further, we demonstrated that CCNA2 expression is elevated in KRAS mutant gastric cancer cell lines and primary tumors, resulting in an increased sensitivity to PLK1 inhibitors. Our study suggests that CCNA2 is a novel biomarker predictive of sensitivity to PLK1 inhibitors for the treatment of advanced gastric cancer, particularly cases carrying KRAS mutation.Gastric cancer is one of the most common malignant tumors of the gastrointestinal track [1]. Due to complex molecular mechanisms and clinical heterogeneity, clinical outcomes for patients with advanced gastric cancer remain poor, with a 5-year survival of 5–20% and a median overall survival (OS) of 10 months [2].To date, only two targeted therapies, treatment with trastuzumab (HER2 inhibitor) or ramucirumab (VEGFR2 inhibitor), have been approved for the treatment of advanced gastric cancer in patients carrying relevant biomarkers and development of more targeted therapeutic strategies for gastric cancer is needed.Polo-like kinase 1 (PLK1), a mitotic serine/threonine protein kinase, regulates various cellular events throughout the cell cycle and has been shown to potentially be a new target in cancer treatment [3]. Increased expression of PLK1 has been observed in several types of malignant tumors and has been shown to be correlated with lower survival rates among solid tumor patients [4,5]. Meanwhile, several PLK1 kinase inhibitors have been developed as anticancer drugs and are currently being evaluated in clinical trials [6]: BI-2536, a dihydropteridinone compound and potent ATP-competitive PLK1 inhibitor [7], was found to inhibit cell proliferation in several human cancer cells, including breast, colon, lung, pancreas and prostate cancer [8]. Building on these results, BI-2536 became the first selective PLK1 inhibitor investigated in clinical trials of patients with solid tumors [9] and exhibited an acceptable safety profile in phase I clinical trials. In phase II study, however, BI-2536 showed relatively poor clinical efficacy, with only 4.2% of patients achieving a partial response in treatment of stage IIIB/IV non-small cell lung cancer [10]. Similar clinical data for BI-2536 were observed in another study of advanced solid tumors [11]. In light of these reports, further clinical studies of BI-2536 as a monotherapy have garnered little interest [10,11]. However, identifying a patient selection biomarker may help to overcome the inefficiency associated with BI-2536 monotherapy and assist with identifying patients who may better respond to treatment with PLK1 inhibitors. Indeed, research has shown that KRAS mutant cancer cells are highly sensitive to PLK1 inhibition [12], wherein cancer cells carrying the oncogenic mutation KRAS were sensitive to PLK1 depletion by shRNA or to treatment with PLK1 inhibitors. However, detailed mechanisms of the actions of PLK1 inhibitors on KRAS mutant cancers are largely unknown.While several drugs targeting KRAS G12C mutant cancer sare under clinical trials [13], the KRASG12C mutation is very rare in gastric cancer: only 3.6% of KRAS mutant gastric cancer patients have the mutation according to combined cohort datasets in the cBioPortal (http://www.cbioportal.org). Therefore, development of alternative therapies will be significant for treatment of KRAS mutant gastric cancers.In this study, we reviewed toxicity screens of 1345 FDA-approved, small-molecule pharmacological compounds and investigational anticancer compounds against a panel of 37 gastric cancer cell lines. Using elastic net regularization, we generated statistical models that predicted the sensitivity of gastric cancer cells to each of the tested drugs based on mRNA expression features, which allowed us to identify distinct drug–biomarker relationships. By focusing on an observed relationship between PLK1 inhibitors and CCNA2, we discovered that oncogenic KRAS mutation drives CCNA2 upregulation and consequent mitotic catastrophe and apoptosis in the presence of PLK1 inhibitors.We previously screened seven gastric cancer cell lines against 1345 pharmaceutical compounds and selected 63 compounds that induced a greater than 50% decrease in cell viability in at least four of the seven cell lines after 72 h of exposure [14]. In this study, we expanded this to 37 gastric cancer cell lines and to 75 compounds targeting cell proliferation, survival and signal transduction pathways (Figure 1a,b). Cell line-specific responses to each of the 75 drugs were calculated by estimating the mean area under survival curves in duplicate (Figure 1c and Table S1). Using elastic net regularization, we generated statistical models that predicted the sensitivity of gastric cancer cells to each of the tested drugs according to mRNA-based gene expression features. In result, we found 23 biomarkers that predicted sensitivity among gastric cancer cells to nine drugs under bootstrapping (random sampling of cell lines with replacement) and a frequency threshold of 75% (Figure 1d and Figure S1). Intriguingly, CCNA2, encoding cyclin A2, which regulates cell cycle progression during the S phase and in G2/M transition, was commonly associated with responses to PLK1 inhibitors BI-2536 and volasertib (BI-6727) (Figure 1d). The concordant associations with CCNA2 expression (i.e., elevated CCNA2 predicts hypersensitivity) with two structurally distinct PLK1 inhibitors, but not with other drugs, were suggestive a biologically meaningful relationship. Therefore, we decided to further investigate whether CCNA2 may be a functional of differential responses to PLK1 inhibitors in gastric cancer.First, we sought to validate differential expression of cyclin A2 protein in gastric cancer cell lines selected from both sides of the drug response profile for PLK1 inhibitors. Compared to resistant cells, gastric cancer cells sensitive to PLK1 inhibitors showed increased expression of cyclin A2 (Figure 2a). MKN28 (sensitive) and SNU719 (resistant) cells were further evaluated in regards to multi-point dose responses to BI-2536. As expected, MKN28 cells exhibited greater sensitivity to BI-2536 than SNU719 cells (Figure 2b). In MKN28 and other sensitive cancer cell lines (AGS and SNU601), but not in SNU719 cells, BI-2536 elicited PARP1 cleavage, JNK phosphorylation and caspase-3 cleavage, all of which are indicative of apoptosis induction (Figure 2c and Figure S2a). To determine if elevated CCNA2 is required to confer sensitivity to BI-2536 in gastric cancer cell lines, CCNA2 was transiently overexpressed in SNU719 cells and knocked down in MKN28 cells (Figure 2d). MKN28 cells transfected with CCNA2 siRNAs gained resistance to BI-2536, compared to cells transfected with negative control siRNA (siNC) (Figure 2e). Meanwhile, however, SNU719 cells transfected with CCNA2 cDNA exhibited greater sensitivity to BI-2536 than control cells (Figure 2f). To confirm that BI-2536 sensitivity indeed acts in relation to CCNA2 expression, we stably knocked down CCNA2 in MKN28 cells using viral transduction of shRNAs and tested the resultant cell viability. Therein, MKN28 isogenic cells, in which CCNA2 was knocked down by five shRNA clones, showed decreased cyclin A2 expression (Figure 2g) and increased viability against BI-2536 (Figure 2h). Similarly, we also observed the shCCNA2-mediated reversal of cytotoxicity to BI-2536 in AGS and SNU601 (Figure S2b,c). Taken together, we deemed that elevated CCNA2 expression is required to confer sensitivity to PLK1inhibitors in gastric cancer cell lines.Cyclin A2 regulates cell cycle progression by promoting S phase entry upon forming a complex with CDK2, as well as by facilitating mitosis through cooperation with the cyclin B1-CDK1 complex [15]. In contrast, PLK1 primarily functions in the M phase of the cell cycle [16,17,18,19] and inhibition of PLK1 causes cell cycle arrest at the G2/M-phase, followed by mitotic catastrophe, a type of apoptosis that occurs during mitosis, in cells with higher mitotic index [20,21]. Therefore, we hypothesized that aberrant upregulation of cyclin A2 in gastric cancer cells may elicit synthetic lethal vulnerability to PLK1 inhibition through failed cell cycle progression, particularly at the M phase. To test this, we performed immunocytochemistry using phospho-histone H3 antibody to detect mitotic cells and found that cyclin A2-knockdown cells show lower mitotic index values than control cells (Figure 3a). To investigate whether elevated cyclin A2 induces sensitivity to BI-2536 due to impaired mitotic progression, we assessed changes in cell numbers in each phase of the cell cycle at 24, 48 and 72 h post treatment with BI-2536 and with or without CCNA2 knockdown in MKN28 cells. Control cells showed marked cell cycle arrest at the G2/M-phase at 24 h posttreatment with BI-2536; however, shCCNA2 cells slipped over from the G2/M-arrest and showed accumulation of polyploidy at 48 h and 72 h post BI-2536-treatment in a dose-dependent manner (Figure 3b), indicating that cancer cells characterized by high expression of cyclin A2 undergo less mitotic slippage and more apoptosis in response to BI-2536 treatment than cyclin A2 knockdown cells. We confirmed this through flow cytometricanalysis and subsequent immunoblot analysis of control MKN28 cells, which showed early apoptotic cell populations within 24 h of BI-2536 treatment (Figure S3) and accumulation of apoptotic marker proteins (e.g., 89-kDa cleaved PARP1, JNK phosphorylation and cleaved caspase-3) upon exposure to BI-2536 (Figure 3c). shRNA-mediated knockdown of cyclin A2 significantly reduced apoptotic marker proteins (Figure 3c), cell-fractions under mitotic catastrophe (Figure 3d) and dead cell populations (Figure 3e) induced by BI-2536.These observations indicated that PLK1 inhibition in the context of elevated CCNA2 leads to mitotic catastrophe and apoptosis rather than to cell survival through mitotic slippage. Meanwhile, research has indicated a potential direct regulatory mechanism for cyclin A2 on PLK1 activation and phosphorylation [22]. To investigate if cyclin A2 functions through PLK1 activity, we developedphosphomimetic mutant PLK1 (T210D) and non-phosphorylatable mutant PLK1 (T210A) proteins (Figure S4a,b) and observed their effects on BI-2536 sensitivity. Interestingly, overexpression of neither of these mutant PLK1 proteins nor wild-type PLK1 altered sensitivity to BI-2536 (Figure S4c,d), indicating that CCNA2-induced BI-2536 sensitivity is independent of PLK1 and phospho-PLK1 levels.While it was reported that KRAS mutant cancer cells are highly sensitive to PLK1 inhibitors [12], detailed mechanisms underlying the sensitivity are largely unknown. Here, we hypothesized that oncogenic KRAS would drive aberrant upregulation of CCNA2. To test this, we compared CCNA2 expression levels between wild-type and mutant KRAS or pan-RAS (KRAS, HRAS and NRAS) tumor samples in The Cancer Genome Atlas (TCGA) cohort. CCNA2 expression was significantly higher in pan-RAS and KRAS mutant tumors than wild-type controls (p =1.01 × 10−31 and 4.27 × 10−20 by Wilcoxon test, respectively). KRAS mutant gastric tumor samples (TCGA-STAD) also showed significantly higher expression of CCNA2 (p = 4.76 × 10−4 by Wilcoxon test) than wild-type gastric tumors (Figure 4a). The 37 gastric cancer cell lines, which includedeight KRAS mutant cell lines (AGS, SNU601, SK4, SNU1, NCC24, SNU668, YCC2 and NCC59) in this study also showed similar results in that KRAS mutant cell lines had higher sensitivity to BI-2536 and volasertib (Figure 4b). To test whether KRAS affects CCNA2 expression or vice versa, KRAS mutant gastric cancer cell lines were transfected with siRNAs targeting KRAS or CCNA2. Therein, the KRAS mutant gastric cancer cell lines (AGS, SK4 and SNU601) showed decreased CCNA2 expression after depletion of mutant KRAS, whereas CCNA2 knockdown did not affect expression of KRAS (Figure 4c). In addition, while depletion of mutant KRAS reversed cytotoxicity to BI-2536 and volasertib (Figure 4d), co-expression of CCNA2 in this context was sufficient to reintroduce sensitivity (Figure 4e), indicating that sensitivity to PLK1 inhibitors in KRAS mutant gastric cancer cells is mediated by CCNA2 upregulation.Taken together, these data demonstrated that oncogenic KRAS-driven CCNA2 upregulation confers hypersensitivity to PLK1 inhibition through mitotic catastrophe and apoptosis (Figure 4f).The cell cycle is tightly regulated by cyclins, cyclin-dependent kinases (CDKs) and various checkpoint kinases [23]. Deregulation of the cell cycle is a hallmark of tumorigenesis [24]. Given its importance in tumorigenesis, several cell cycle inhibitors have emerged as potential therapeutic drugs for the treatment of cancers, both as single and combination therapies with traditional cytotoxic or molecular targeting agents. Currently, the most promising cell cycle inhibitors in anticancer therapeutics are orally bioavailable CDK4/6 inhibitors, which have received regulatory approval in combination with hormonal therapy for treatment of patients with metastatic hormone receptor (HR)-positive, Her2-negative breast cancer [25,26,27,28,29]. Many other compounds designed to interrupt cell cycle progression or checkpoint control have problems in demonstrating sufficient antitumor efficacy. Thus, to facilitate clinical development of this target class, proper patient selection biomarkers are urgently needed.PLK1 is a serine/threonine kinase that regulates various cellular events during cell cycle progression, including centrosome maturation, DNA checkpoint activation, mitotic entry, spindle assembly and cytokinesis [3]. Because hyper-activation of PLK1 causes overriding checkpoints, which leads to immature cell division, several PLK1 kinase inhibitors have been developed as anticancer drugs and are currently being evaluated in clinical trials [6]. Several challenges limiting the success of these drugs are as follows: First, most PLK1 inhibitors exhibit a narrow therapeutic window, with their therapeutic effects often coupled with normal toxicity. Second, responses to PLK1 inhibitors are inconsistent. Last is the emergence of drug resistance [6]. Discovery of patient selection biomarkers can help overcome this challenge by identifying correct tumor types and by facilitating patient stratification to increase response rates to these drugs. Indeed, some studies have reported that TP53 mutation is a biomarker of sensitivity to PLK1 inhibitors [30,31]. This suggests a compensatory mechanism mediated by p53 that rescues cancer cells from mitotic arrest and subsequent apoptosis caused by PLK1 inhibition. Another potential biomarker predicting sensitivity to PLK1 inhibitors is the oncogenic KRAS mutation. Research has been shown that cancer cells with KRAS mutation are more sensitive to PLK1 inhibitors than KRAS wild-type cancers [32], suggesting that KRAS mutation induces mitotic stress in tumor cells and may underlie tumor sensitivity to anti-mitotic agents.Cyclin A2 belongs to the highly conserved cyclin family and is expressed in almost all tissues in the human body. It plays critical roles in control of the cell cycle at G1/S and in G2/M transition. Data from the Human Protein Atlas show that CCNA2 is overexpressed in dozens of cancer types, suggesting a potential role in tumorigenesis. In this study, we demonstrated that, compared to cells with basal CCNA2 expression, cancer cells highly expressing CCNA2 are more addicted to PLK1 activity and show increased mitotic index values, leading to G2/M arrest and mitotic catastrophe, followed by apoptosis, in response to PLK1 inhibitors. We observed neither PLK1 nor phospho-PLK1 affects sensitivity to PLK1 inhibitor, suggesting that the increased sensitivity of CCNA2-elevated cancer cells to PLK1 inhibition is not due to direct perturbation of the cyclin A2-PLK1 axis, but rather, likely due to a different mechanism that needs to be further elucidated. One possibility may be that increased cyclin A2 facilitates G2/M transition by activation of the cyclin B1/CDK1 complex [33], resulting in increased mitotic cell populations that more greatly rely on proper spindle assembly checkpoint, wherein PLK1 kinase activity plays an essential role [34].We also discovered a causal relationship between oncogenic KRAS mutation and CCNA2 upregulation. Our data suggest that aberrations in CCNA2 expression are a consequence of oncogenic KRAS mutation potentially contributing to cell cycle progression, while at the same time, conferring dependence on PLK1 function for productive mitotic progression. Although it is beyond the scope of this manuscript, several mechanistic hypotheses may explain the connections between KRAS mutation, cyclin A2 elevation and PLK1 inhibitor sensitivity: Potentially, upon glutamine (Gln) deprivation, KRAS-driven cancer cells bypass a late G1 Gln-dependent cell cycle checkpoint and enter S-phase, followed by cell cycle arrest due to insufficient nucleotide biosynthesis [35]. Meanwhile, cyclin A2 forms a complex with CDK2 at the S phase of the cell cycle to initiate and progress DNA synthesis. Thus, elevated cyclin A2 in KRAS mutant cancers may reflect anadaptation mechanism from this cell cycle stress at the S phase. Alternatively, as cyclin A2 directly phosphorylates and activates protein kinase B, also known as Akt [36,37], elevated cyclin A2 may contribute to Akt-driven tumorigenesis as well. Either way, as a consequence of CCNA2 upregulation, cancer cells mayface an unavoidable dependence on PLK1 to prevent mitotic catastrophe by disrupted spindle assembly. Thus, we considerthis relationship as a “synthetic lethality.”Accordingly, synthetic lethal approaches targeting cell cycle progression with PLK1 inhibitors may prove to be effective in treating tumors characterized by increased CCNA2 expression.This study analyzed data for 37 gastric cancer cell lines treated with 75 small-molecule compounds selected from libraries of FDA-approved small-molecule pharmacological compounds (#L1300, Selleckchem, Houston, TX, USA) and investigational anti-cancer compounds (#L2000, Selleckchem). The pharmacological profiles of 29 of the 37 cell lines have previously been reported [14]. For cell-based drug assay, 5000 cells were seeded onto individual wells of 96-well plates. After 24 h of incubation, half-log 12-serial dilutions of pharmacological compounds in DMSO were added using a BiomekFXp liquid handler (Beckman Coulter, Brea, CA, USA), resulting in final concentrations of 50 μM to 0.5 nM. The cells were then incubated for 72 h and cell viability was measured with CellTiter-Glo assay kits (Promega, Madison, WI, USA). In each cell line, DMSO (0.5%) controls were used for normalization. Finally, we calculated area under the viability curve (AUC) values from 12-point dose–response curves for each pharmacological compound. All gastric cancer cell lines, except SK4 and the Yonsei Cancer Center (YCC)-series cell lines, were purchased from the Korea Cell Line Bank. SK4 cells were kindly provided from Dr. Julie Izzo (MD Anderson Cancer Center, Houston, TX, USA). YCC-series cell lines were obtained from the Song–Dang Institute for Cancer Research (Yonsei University College of Medicine, Seoul, Korea). The cell lines were maintained in RPMI-1640 medium supplemented with 10% fetal bovine serum (Gibco/Thermo Fisher Scientific, Waltham, MA, USA) and 1% penicillin–streptomycin (Invitrogen, Carlsbad, CA, USA) in mycoplasma-free condition. All gastric cancer cell lines have been authenticated using STR profiling within the last three years.RNA sequencing (RNA-seq) data for 29 of the 37 gastric cancer cell lines were previously reported [14]. Total RNA from the eight remaining gastric cancer cell lines were extracted with a RNeasy Plus Mini Kit (Qiagen, Hilden, Germany). RNA-seq libraries were then generated with a TruSeq RNA Sample Prep kit v2 (Illumina, San Diego, CA, USA) and sequencing with the HiSeq 2500 platform. The TopHat-Cufflinks pipeline was used to align reads to the reference genome and to calculate normalized values in FPKM (fragments per kilobase of exon per million fragments mapped).We established elastic net models using a previously described method with some modifications [38]. Briefly, transcriptome and drug response data (n = 75) for the 37 gastric cancer cell lines were used to build the model. To do so, gene expression values were converted into Z-scores. Optimal values of α and λ were determined by 10-fold cross validation from 100 iterations. Bootstrapping (200×) was applied to estimate average weights (β) and selection frequency of features by the model. For drug marker selection, we chose features occurring at a frequency > 75%. Next, we applied different weight cutoffs to individual drugs because their weight spectrumsvaried greatly, which made it difficult to apply a single weight cutoff. The feature selection process was conducted using the GlmnetR package (version 2.0–8) and R (version 3.3.3).In each well of 96-well plates, 30 µL of 333 nM siRNA solution was mixed with 10 µL of 2% RNAiMAX (Invitrogen) solution and incubated for 15 min. Subsequently, 7000 cells in 100 µL of growth medium were added to the mixture. Culture medium was replaced 24 h post transfection. siRNA oligonucleotides were custom synthesized (Genolution, Seoul, Korea) with the sequences: 5′-GAUAUACCCUGGAAAGUCUUU-3′ (siCCNA2-1), 5′-GGAUGGUAGUUUUGAGUCAUU-3′ (siCCNA2-2), 5′-CUAUGGACAUGUCAAUUGUUU-3′ (siCCNA2-3), 5′-CGAAUAUGAUCCAACAAUAUU-3′ (siKRAS-1), 5′-GACAAAGUGUGUAAUUAUGUU-3′ (siKRAS-2), 5′-GCAUGGGACAUUUGUGAUUUU-3′ (siNC).The Myc-DDK-tagged human CCNA2 plasmid (#RC211148L1) was purchased from OriGene (Rockville, MD, USA). Two million cells were grown in 60-mm dishes for 24 h. On the day of transfection, 5 µg of plasmids and 15 µL of Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA) were diluted in 500 µL of Opti-MEM and added onto the culture dishes. Twenty-four hourspost transfection, cell lines transfected with either target plasmid or empty vector were trypsinized and re-plated into 96-well plates (8000 cells per well) for drug toxicity assay.Total cell extracts were prepared by dissolving cells with 1×Laemmli SDS reducing buffer (50 mM Tris-HCL [pH 6.8], 2% SDS and 10% glycerol) and boiled for denaturation. Equal amounts of protein sample were separated on 4-15% Mini-PROTEAN TGXTM Precast Gel (Bio-Rad, Hercules, CA, USA). Anti-Cyclin A2 (#4656S, Cell Signaling, Danvers, MA, USA), anti-β-actin (#sc-47778, Santa Cruz Biotechnology, Dallas, TX, USA), anti-KRAS (#sc-30, Santa Cruz Biotechnology), anti-cleaved PARP (#9541S, Cell Signaling Technology), anti-phospho-JNK (#4668S, Cell Signaling Technology), anti-cleaved Caspase-3 (#9661S, Cell Signaling Technology), anti-PLK1 (#4513S, Cell Signaling Technology), anti-FLAG (DYKDDDDK)(#2368S, Cell Signaling Technology) and anti-GAPDH (#60004-1-Ig, Proteintech Group, Rosemont, IL) antibodies were used as primary antibodies. Peroxidase-AffiniPure Goat Anti-Rabbit IgG (#111-035-144) and Anti-Mouse IgG (#115–035–003, Jackson ImmunoResearch, West Grove, RA, USA) were used as secondary antibodies. Antibody binding was visualized by SuperSignal West Pico/FemtoChemiluminescent Substrate (Thermo Fisher Scientific, Waltham, MA, USA) and X-ray films (AGFA-Gevaert, Mortsel, Belgium).Gene expression and mutation data for 16 cancer types were downloaded from the Genomic Data Commons data portal (https://portal.gdc.cancer.gov). Gene expression data in FPKM were transformed to log2 scale values and quantile normalization was performed to remove technical biases. Samples bearing any mutations in KRAS, HRAS or NRAS were considered as RAS mutation samples. A two-sided Wilcoxon rank-sum test was used to test the difference in CCNA2 expression levels between RNA-mutant and wild-type samples.For cell cycle analysis using flow cytometry (FACSVerseTM, BD Biosciences, Franklin Lakes, NJ), MKN28 cells were plated at a density of 1 × 103 cells in 60mm culture dishes and then treated with BI-2536 (200 and 500 nM) for 72 h. Cells were then harvested and fixed in ice-cold 70% ethanol overnight at −20°C. Afterwards, cells were centrifuged at 300× g for 5 min, re-suspended in PBS containing 10 µg/mL of propidium iodide (PI) (P4170, Sigma-Aldrich, St. Louis, MO, USA), 100 µg/mL RNase A and 0.1% (v/v) Triton X-100 and incubated at 37 °C for 10 min. For cell apoptosis assay of MKN28 cells treated with BI-2536 (200 and 500 nM), we used the Annexin V-FITC Apoptosis Kit (#640914, BioLegend, San Diego, CA, USA) according to the manufacturer’s protocol. Briefly, a total of 1 × 103 cells was seeded in 60mm culture dishes for 24 h and then treated with BI-2536 for 72 h. Adherent and floating cells were then harvested, stained with Annexin V and PI for 15 min at room temperature and subjected to flow cytometry. The data were analyzed with FlowJo software.Cells were washed with ice-cold PBS and fixed with 3.7% paraformaldehyde (PFA) for 10 min, permeabilized with 0.5% Triton X-100 in PBS (PBS-T) for 5 min and incubated with blocking solution (0.1% BSA + 10% goat serum in 0.1% PBS-T) for 30 min. Cells were then incubated with primary antibodies diluted in 0.1% PBS-T for an hour. After incubation with secondary antibodies labeled with Alexa Fluor-488 or Alexa Fluor-568 (Invitrogen), coverslips were mounted on slide glasses using Prolong Gold Antifade mounting solution (Thermo Fisher Scientific) and the slides were allowed to dry at room temperature.We performed pharmacogenomics analysis using a gastric cancer cell line panel and discovered a causal linkage cascade of oncogenic KRAS mutation, aberrant CCNA2 upregulation and hypersensitivity to PLK1 inhibitors. Our findings hold translational implications for the treatment of gastric cancer patients with aberrant upregulation of CCNA2 via synthetic lethal approaches targeting cell cycle progression.The following are available online at https://www.mdpi.com/2072-6694/12/6/1418/s1, Table S1: List of 75 compounds and their cell line-specific responses, Figure S1: Identified biomarker and drug response relationships by the elastic net regularization method, Figure S2: BI-2536 sensitivity in AGS and SNU601 cells, Figure S3: Apoptosis analysis using Annexin V-FITC/PI dual staining and flow cytometric analysis after BI-2536 (200 nM) treatment for 24 h are presented in scatter plots, Figure S4: BI-2536 sensitivity with wild-type and mutant PLK1 levels.Conceptualization, S.B.K. and H.S.K.; experiments and data analysis, Y.L. and C.E.L.; bio-informatics analysis and visualization, S.O. and H.K.; initial screening, J.L.; writing—original draft preparation, Y.L. and S.B.K.; writing—review and editing, S.B.K. and H.S.K.; visualization, Y.L. and S.B.K.; supervision, H.S.K.; funding acquisition, H.S.K. All authors have read and agreed to the published version of the manuscript.This study was supported by grants from the Korea Health Technology R & D project through the Korea Health Industry Development Institute (HI14C1324) and the National Research Foundation of Korea (NRF) (2017R1A2B2006777, 2020R1A2C3007792).S.B.K. was supported by the Brain Pool program funded by the Ministry of Science and ICT through the NRF (2019H1D3A2A01050712). H.K was supported by the Global Ph.D. fellowship program funded by the NRF (2019H1A2A1075632).The authors have no competing interests related to this work.Pharmacogenomic analysis identifies biomarker–drug response relationships. (a) Flowchart of overall screening strategy; (b) classification of the 75 compounds according to their target pathways; (c) sensitivities (area under the viability curve (AUC)) of the 37 gastric cancer cell lines to 75 compounds are ordered by row. Rank-ordered original AUC values are indicated as a heat map. Heat mapsare colored on a blue (sensitive) to white to red (resistant) gradient scale of original AUC values. Target pathways for each compound are annotated by the same color code as in b; (d) representative biomarker and drug response relationships by elastic net regularization method. The average weights of features are displayed in bar plots and their frequencies are shown in parenthesis.Bar plots on the left are colored in red when the expression level of a biomarker is positively correlated with the resistance of the given drugs and colored in blue when negatively correlated. Heat mapsaredepicted on a blue–white–red gradient scale of median-centered AUC values and expression levels (FPKM) of genes, respectively.Elevated CCNA2 is required to confer sensitivity to BI-2536. (a) Expression levels of endogenous cyclin A2 were assessed by immunoblotting whole cell lysates from the indicated “resistant” and “sensitive” gastric cancer cell lines; (b) dose–response curves of cell viability for the indicated gastric cancer cell lines after 72 h of exposure to BI-2536; (c) Induction of apoptotic markers were assessed by immunoblotting. Whole cell lysates were prepared post 72 h of BI-2536 or vehicle (DMSO) treatment with indicated concentrations; (d) ectopic expression of CCNA2 in SNU719 cells and knockdown of CCNA2 in MKN28 cells were demonstrated by immunoblotting; (e) dose–response curves of MKN28 cells expressing non-silencing siRNA (siNC) or siRNA against CCNA2 (siCCNA2); (f) Dose–response curves of SNU719 cells transfected with empty pCMV6 plasmid or CCNA2 cDNA plasmid; (g) immunoblot shows depletion of cyclin A2 in MKN28 cells expressing shRNA clones against CCNA2; (h) relative viability of MKN28 cells at 72 h post BI-2536 (300 nM) treatment; (b,e,f) * p <0.05, ** p < 0.01, *** p <0.001; one-way ANOVA; (h) * p <0.05, ** p< 0.01, *** p <0.001; Wilcoxon rank-sum test.Elevated CCNA2 is required for BI-2536-induced mitotic catastrophe and apoptosis. (a) Mitotic cells were visualized by immunostaining using anti-phospho-Histone H3 antibody (red) in MKN28 cell lines expressing shCCNA2 or shCTRL (left). DAPI was counterstained to detect nuclei (blue). Scale bar: 200 µm. Mitotic index (right), calculated by dividing the number of mitotic cells by the total number of cells. * p < 0.05; Student’s t-test; (b) evaluation of cell cycle by propidium iodide (PI) staining and flow cytometer analysis after BI-2536 treatment for the indicated time periods. Percentages of cells in each cell cycle are presented in bar plots; (c) effects of BI-2536 on the expression of apoptotic markers (e.g., cleaved PARP, phospho-JNK and cleaved caspase-3) 72 h post treatment. The amount of β-actin was measured as an internal control; (d) evaluation of mitotic catastrophe by DAPI staining and fluorescence microscopy (Axio Imager M2, ZEISS) at 72 h post BI-2536 (300 nM) treatment (Scale bar, 25µm); (e) apoptosis analysis using Annexin V-FITC/PI dual staining and flow cytometer analysis after BI-2536 (200 nM and 500 nM) treatment for 72 h are presented in both scatter plots (left) and bar plots (right). * p < 0.05; Wilcoxon rank-sum test.Oncogenic KRAS driven CCNA2 upregulation confers sensitivity of KRAS mutant cancer to PLK1 inhibitors.(a) Comparison of CCNA2 expression levels between wild-type (WT) and pan-RAS (KRAS, HRAS and NRAS) or mutant KRAS tumor samples (pan-cancer and gastric cancer) in the Cancer Genome Atlas (TCGA) cohort; (b) cumulative distribution fraction plots of drug response (median-centered AUC) in the 37 gastric cancer cell lines show that KRAS mutant cell lines had higher sensitivity to BI-2536 and volasertib. p values were calculated by two-sided Kolmogorov–Smirnov tests (KS-test); (c) evaluation of cyclin A2 and KRAS expression by immunoblotting post knockdown of CCNA2 and KRAS in KRAS mutant gastric cancer cell lines (AGS, SK4 and SNU601). GAPDH was measured as an internal control; (d) relative viability of AGS cells expressing siKRASoligos at 72 h post treatment with BI-2536 (0.16μM) and volasertib (0.16μM). Expression changes of KRAS and cyclin A2 by expression of siKRASoligos were observed by immunoblotting. * p <0.05, ** p <0.01; Student’s t-test; (e) relative viability of AGS cells expressing siKRAS with or without CCNA2 cDNA at 72 h post BI-2536 (0.16μM and 0.5μM) treatment. Expression changes of KRAS and cyclin A2 were assessed by immunoblotting. ** p <0.01, *** p <0.001; two-way ANOVA; (f) Hypothetical model of selective toxicity to PLK1 inhibitor in KRAS mutant cancer.

Med-MDPI/cancers/cancers-12-06-01419.txt

@@ -0,0 +1 @@
+Metabolic reprogramming is critically involved in the development and progression of cancer. In particular, lipid metabolism has been investigated as a source of energy, micro-environmental adaptation, and cell signalling in neoplastic cells. However, the specific role of lipid metabolism dysregulation in hepatocellular carcinoma (HCC) has not been widely described yet. Alterations in fatty acid synthesis, β-oxidation, and cellular lipidic composition contribute to initiation and progression of HCC. The aim of this review is to elucidate the mechanisms by which lipid metabolism is involved in hepatocarcinogenesis and tumour adaptation to different conditions, focusing on the transcriptional aberrations with new insights in lipidomics and lipid zonation. This will help detect new putative therapeutic approaches in the second most frequent cause of cancer-related death.In the last few decades, the incidence of hepatocellular carcinoma (HCC) has rapidly increased and HCC has become the second most frequent cause of cancer-related death worldwide [1,2]. The short-term prognosis has improved because of new advances in therapy and early diagnosis, although the long term prognosis remains poor, with a 5-year survival rate of 17% [3,4,5]. The surgical resection is still the option with the highest recovery rate; however, only 15% of patients are eligible and the 5-year recurrence rate is about 70% [6,7]. Therefore, new pharmacological treatment options are needed. More than 90% of liver cancers develop in chronic liver disease, typically associated with viral infections, such as hepatitis B virus (HBV) and hepatitis C virus (HCV) [8]. However, the incidence of non-viral HCC is increasing, since obesity, diabetes, and alcohol drinking in young people are a real outbreak [9,10,11]. It is indeed estimated that, by 2030, in US, the number of non-alcoholic fatty liver disease (NAFLD) and non-alcoholic steato-hepatitis (NASH) cases will increase about 21% and 63%, respectively, while HCC cases will increase by 137% [9]. Viral and non-viral aetiologies normally encompass a spectrum of histological alterations from simple steatosis to inflammation, fibrosis with cirrhosis, and complications, such as HCC. However, obesity, diabetes, and steatosis account for independent risk factors of HCC [12,13,14], which can indeed occur in early stages of liver diseases, especially in metabolic syndrome. Recently, cancer research has been focused on the role of metabolic reprogramming in tumours, particularly on the role of lipid metabolism. The recent advances in the understanding of the role of lipids in HCC provide a new point of view for the interpretation of tumour development and progression. HCC pathogenesis is very complex as several factors are involved, such as cytokine pattern, ER stress, insulin resistance, oxidative stress, gut microbiota [15,16,17,18,19,20,21], and genetics. In the present review, we shed light on metabolic lipid alterations as a cornerstone in hepatocellular tumour initiation and adaptation.A milestone paper that first established the association between HBV infection and HCC in the Taiwanese population was published in 1981 [22]. Over 90% of HBV cases resolve spontaneously, but 25% of chronic infections can evolve into HCC [23,24]. In East Asian countries, the HCC incidence rates were estimated in a systematic review, reporting 0.2 person-years per 100 person-years in inactive carriers, 0.6 person-years in chronic HBV infection without cirrhosis, and 3.7 person-years in compensated cirrhosis [25]. Interestingly, although 70–90% of HBV-related HCC occurs in cirrhotic patients, cases in absence of cirrhosis are also possible [26]. Several studies have described lipid metabolism alterations in HBV infections [27,28,29]. Park et al. reported the hepatic suppression of choline-phosphate cytidyltransferase A (PCYT1A) expression in HBV infected mice, with significant differences in phosphatidylcholine composition [30]. Accordingly, Li et al. showed the up-regulation of phosphatidylcholine biosynthesis promoted by choline kinase alpha (CHKA) in HepG2 cells, a pathway required for HBV replication [31]. In metabolomic and gene expression analysis, Teng et al. used transgenic mice expressing the HBx gene of HBV (HBx mice) to investigate the lipid profile in serum and liver during tumorigenesis [32]. They identified two peaks, one related to generic inflammation and oxidative stress and the second corresponding to the tumour phase after a resolution phase. The tumour phase correlated with the expression of five genes implied in lipid metabolism: arachidonate 5-lipoxygenase, lipoprotein lipase, fatty acid binding protein 4, 1-acylglycerol-3-phosphate O-acyltransferase 9, and apolipoprotein A-IV. The same data were later validated both in vitro and in human HBV-related HCC. Overall, these studies suggest a potential role of lipid metabolism as a driving force in HBV-related HCC.HCV has been so far the most frequent cause of virus related HCC [6]. Recent direct-acting antivirals cure almost 100% of HCV infections, predicting an important reduction of future infections and HCV-related liver disease. The high cost of such drugs will impact the world epidemiology of HCV-related HCC with significant reduction in western countries, but not in developing areas. Moreover, infected people still have risk the development of HCC, despite the availability of an effective therapy [33,34]. In a recent study, we showed that during viral eradication with direct-acting antivirals (DAAs), vascular endothelial growth factor (VEGF) circulating levels increase until the end of the treatment in a mutating inflammatory background, accounting VEGF as a reasonable tumoral risk factor [35]. In accordance with these observations, HCV-related HCC incidence is predicted to increase until 2030, despite the new therapeutic approaches [36]. It has been widely proved that HCV replication interferes with cell survival and proliferation and with gene expression [37,38]; nevertheless, the pathogenesis of HCV-induced HCC is still largely unclear. The association of HCV infection with steatosis and diabetes has emerged powerfully in the last few years. HCV replication induces hepatic steatosis through insulin resistance or by direct metabolic interference of core proteins [39]. HCV proteins can inhibit microsomal transfer protein activity, an enzyme involved in the formation and secretion of very low density lipoprotein (VLDL) and, therefore, with consequent accumulation of triglycerides (TAG) [40,41]. Moreover, as shown in human tissue, animal model, and cell lines, the regulatory element binding protein (SREBP-1c) is up-regulated with consequent transcriptional activation of enzymes involved in fatty acid (FA) synthesis, such as acetyl-CoA carboxylase (ACC), sterol CoA dehydrogenase 4 (SCD4), and fatty acid synthase (FASN) [42,43,44]. Lerat et al. reported that transgenic mice with hepatic expression of HCV proteins are characterised by high lipogenesis and defective TAG exportation with final micro- and macro-vesicular steatosis [45]. Furthermore, males are particularly inclined to develop HCC. Several of these mechanisms are typically involved in lipid metabolic reprogramming of liver carcinogenesis, as discussed below. Therefore, we can assume that HCV-related lipid perturbations might play a key role in HCC development.The incidence of non-viral related HCC is increasing, particularly in developed countries where vaccination for HBV and DAAs for HCV are constantly changing the scenario. Most of these patients are affected by steatohepatitis or cirrhosis [19,46]. However, even NAFLD can directly evolve to HCC independently of NASH [13,47,48]. In fact, being overweight alone increases the risk of HCC by 17%, while obesity increases the risk by 89% [49]. In addition, in patients with viral hepatitis (HBV and HCV), obesity was associated with a higher risk of HCC, highlighting a synergistic pro-tumoral effect provoked by dysregulation of lipid metabolism [46,50]. Even if the mechanisms by which obesity and steatosis promote hepatic carcinogenesis remain quite unclear, the role of lipid dysregulation in this process is well recognized. Since the pathogenesis of HCC in metabolic syndrome is complex, we should consider interaction among multiple patterns, as recently defined in the multiple hit model [51]. For example, during obesity development, the liver accumulates fat to counteract excess of free FAs [52,53], which in turn induces the release of gut-derived endotoxins and the production of adipose tissue cytokines, alongside the worsening of liver inflammation and damage [20,54,55,56,57,58].Alcohol-related liver disease (ALD) includes a wide spectrum of histological hepatic alterations, such as steatosis, alcoholic hepatitis (AH), cirrhosis, and HCC [59]. In western countries, ethanol still represents the main cause of liver disease [60,61]. Ethanol provokes liver damage by multiple mechanisms, which include oxidative stress, lipids accumulation, inflammation, and leaky gut [62]. Interestingly, histological features typically observed in ALD depend on lipid accumulation; in fact, ethanol promotes FA synthesis and suppresses oxidation. On the other hand, ethanol increases hepatic uptake of FAs, which will be incorporated in TAG [63,64,65] by the up-regulation of de novo lipogenesis genes, such as ACC1, FASN, and SCD1 via sterol regulatory element-binding protein1 (SREBP1c), as described in different murine models [66,67,68]. Several findings in cultured hepatocytes and murine livers report that ethanol is also able to over-express Lipin-1, a protein with Mg2+-dependent phosphatidic acid phosphohydrolase (PAP) activity, involved in the penultimate step of triglyceride synthesis [69,70,71,72,73,74]. Despite the high availability of FAs in alcohol drinking, β-oxidation is dampened. This effect seems to depend on peroxisome proliferator activated receptor alpha (PPAR-α) inhibition, as observed in vitro and in animal models [75]. Accordingly, in ethanol-fed mice, PPARα DNA binding activity decreases with consequent lower expression of enzymes essential in fatty acids oxidation (FAO) [76].Carcinogenesis and tumour adaptation to a local microenvironment are fuelled by metabolic alterations, which promote the survival of cancer cells. This process is known as metabolic reprogramming, a set of adaptations clonally selected during tumorigenesis by generating metabolites with different functions at different levels [77]. The two most studied examples of tumoral metabolic reprogramming are the Warburg effect and glutaminolysis [78]. Reprogramming cell energetic functions is an important survival strategy in cancer, as shown by the enhancement of aerobic glycolysis instead of mitochondrial oxidative phosphorylation (the Warburg effect) [79]. Similarly, the increased glutamine metabolism, glutaminolysis, sustains the mitochondrial tricarboxylic acid (TCA) cycle by generation of citrate and αketoglutarate [80]. Moreover, the role of lipid metabolism in cancer is receiving more and more attention, and thus, several studies show that FAs and/or cholesterols promote tumoral activity [81,82,83,84,85,86]. The multiple functions of lipids in cell signalling, membrane components, and sources of energy are essential in cancer cells [87]. The high “starvation” of lipids in tumours is fulfilled by external uptake and de novo lipogenesis [88], and, consequently, FAO is also increased in several tumour types [89,90].HCC is typically characterised by up-regulation of genes involved in FA synthesis, such as ATP-citrate lyase (ACLY), acetyl-CoA carboxylase (ACC), and fatty acid synthase (FASN), which induce conversion of citrate to acetyl-CoA, malonyl-CoA, and FA, respectively [91,92,93,94,95,96,97,98]. FAs are converted to monounsaturated fatty acids (MUFA); sources for the synthesis of TAG. Stearoyl-CoA desaturase (SCD) is responsible for MUFA generation, and its up-regulation has been associated with HCC in humans [88]. Moreover, the expression of these genes involved in FA synthesis can be regulated by the transcription factor SREBP-1c, whose lipogenic pathway is upregulated in human HCC [99]. Interestingly, hepatic lipogenesis is further modulated by peroxisome proliferator-activated receptor-γ coactivator β (PGC-1β), a transcriptional cofactor which interacts with SREB1c, inducing the transcription of genes, such as as FASN and SCD1 [100] (Figure 1). PGC-1β overexpression in mice promotes HCC development [101]. However, the de novo lipogenesis is not a unique source of FAs in HCC. In obesity- and NASH-related HCC, the persistent lipolysis of the adipose tissue provides an enormous amount of FAs (in particular non-esterified FAs) to the liver, with consequent adaptation to this stress [102]. To do this, the liver increases aerobic glycolysis and glutamine synthesis with the enhancement of TCA, promoting hepatocarcinogenesis [103]. However, the mechanism to explain how malignant cells escape lipotoxicity is not yet understood. Recent studies described a new histological variant of HCC, steatohepatitic HCC (SH-HCC), whose peculiarity is macrovesicular steatosis with underlying viral or non-viral steatohepatitis [104,105]. Accordingly, most obesity-related HCC models in mice are characterised by a prominent lipid accumulation in tumoral tissue than in non-tumoral tissue [19,20,106]. Fujiwara et al. suggested that this is plausible because, in the diethylnitrosamine (DEN)-induced HCC model, high fat diet (HFD)-fed mice showed an up-regulation of CPT1A, which converts FA-derived acyl-CoA to Acylcarnitine, while CPT2 (which reconverts acylcarnitine to AcylCoA) results in down-regulated CPT2 [107]. The consequence is the accumulation of Acylcarnitine with oncogenic effects and lower availability of acyl-CoA for β oxidation, resulting in lipids storage. In human SH-HCC, CPT2 expression is also downregulated and high circulating levels of acylcarnitine are detectable in subjects with NASH or HCC [107,108]. Similar results in terms of gene expression profile are given in other murine models, such as major urinary protein (MUP)-urokinase-type plasminogen activator (MUP-uPA) mice [19] and phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA) transgenic mice [106]. Collectively, these works show how CPT2 down-regulation promotes FAO low activity and, therefore, protects lipotoxic cell death. Moreover, HCC cells knocked down for CPT2 acquired resistance to lipotoxicity cell death by inhibiting FAO and Src-mediated c-jun NH-2-terminal kinase (JNK) activation [109,110,111,112]. Therefore, although FAO activity is essential to maintain the NADPH amount and ATP generation, suppling energy in several tumours [90,113,114], an excessive activity of electron transport chain may produce reactive oxygen species (ROS) and oxidative damage, leading to cancer cell death [115,116]. However, the studies regarding the role of FAO in HCC progression are extremely variable, depending on different conditions that are explained below. While the implication of FA metabolism has been widely described, the direct effect of oxidised sterols in HCC development remains unknown. However, it is known that patients with HCV hepatitis and NAFLD present higher serum concentrations of oxysterols [117,118]. Accordingly, in a previous study, we reported that supplementation with cholesterol in HFD-fed mice determine the hepatic accumulation of specific oxysterols (e.g., 7β-hydroxycholesterol, 7-Ketocholesterol, and 5α-cholestane-3β,5,6β-triol), which impair mitochondrial function, acting synergistically with FAs and therefore facilitating NAFLD progression [119]. However, some typical oxysterols associated with NAFLD (e.g., 25-hydroxycholesterol) are known ligands of liver X receptor (LXR), whose low tumoral expression has been recently suggested as a negative prognostic marker in patients operated for HCC [120]. In fact, LXR enhances, at a transcriptional level, the cytostatic and pro-apoptotic effects of transforming growth factor beta (TGFβ-1) [121] and inhibits HCC cell proliferation by through activation of suppressor of cytokine signalling 3 (SOCS3) [122].HCC is characterised by several histological features with different underlying diseases, and, therefore, studies are not conclusive on the role of FAO. Lu et al. demonstrated that activation of FAO stimulates HCC cells to survive energy deprivation via expression of CCAAT/enhancer binding protein alpha (C/EBPα), which in turn induces autophagy [123]. Accordingly, other reports highlighted the importance of 5′ adenosine monophosphate-activated protein kinase (AMPK) in FAO enhancement in HCC when nutrients are scarce [113,124]. Moreover, ACC can modulate FAO activity in human and murine HCC as it forms a complex with carnitine palmitoyltransferase 1A (CPT1A). In a nutrient deficiency condition, AMPK phosphorylates ACC, which dissociates from CPT1A, permitting its translocation to the mitochondrial membrane to sustain FAO by FAs transport [98]. Cassim et al. proved that HCC cells show a prevalent glycolytic metabolism, but under glucose deprivation FAO is activated, supplying energy and facilitating proliferation [125]. In contrast, in the condition of hypoxia, HCC cells react differently, as FAO is repressed. When the tumoral growth rate accelerates, the blood oxygen supply could be inadequate, thereby provoking hypoxia [126]. Two recent studies propose that hypoxia inducible factor 1-α (HIF-1α) induces FAO inhibition, protecting HCC from the excessive production of ROS under hypoxic conditions [115,127]. HIF-1α reduces the expression of medium- and long-chain acyl-CoA dehydrogenases (MCAD and LCAD), two rate-limiting enzymes involved in mitochondrial FAO initiation, as demonstrated in vitro in a rodent model and human specimens [115]. Another proposed pathway of HCC adaptation during hypoxia involves the up-regulation of mitochondrial acetyl-CoAsynthetase 1 (ACSS1), which converts acetate to acetyl-CoA [91]. In this gene expression study of 361 HCC tissue, the authors structured a model to identify metabolic processes involved in tumour proliferation, predicting that in hypoxic conditions, FA synthesis is enhanced, while FAO is repressed. Furthermore, ACSS1 high expression was associated with malignancy [91]. Despite these reports, Iwamoto et al. described that oxygen and nutrient depletion by antiangiogenetic drugs provokes a metabolic switch to lipid-dependent metabolism by raising FA uptake and FAO [128]. Notably, in murine models of tumour implantation in a lipid-rich environment (i.e., adipose tissue and steatotic liver), the authors showed that metastatic cells can easily proliferate because of increased FA uptake and catabolism.Another condition in which enhanced FAO permits tumoral progression is βcatenin-activated HCC. A quote of HCC develops in non-cirrhotic livers and in patients with metabolic syndrome without fibrosis [129,130,131], partially explained by an alternative mechanism for HCC development, which is called β-catenin-activated signalling. Senni et al. showed a different lipid metabolic reprogramming in βcatenin-activated HCC as the expression of PPARα and CPT2 is increased with a consequent enhancement of FAO in humans and mice [132]. PPARα is also able to up-regulate the expression of MCAD and LCAD, thereby fuelling FAO initial steps. Here, the authors corroborate data by PPARα genetic ablation and CPT1 inhibition with etomoxir, blocking development and progression in βcatenin-activated HCC in mice. In contrast, we have discussed above that in non-mutated βcatenin HCC, CPT2 is normally downregulated [91], as well as in SH-HCC and HFD-fed MUP-uPA mice [133,134]. However, studies also recently reported βcatenin activation in cases of NASH [97,135,136,137]. Finally, it is intriguing to observe that, although FAO activity suppression is beneficial for HCC development, especially in obesity and NASH, increasing β-oxidation is fundamental in hypoxia, nutrient deficiency, and β-catenin-activated HCC (Figure 2). Moreover, enhancement of FAO seems to be essential in cancer stem cells (CSCs), a subpopulation of cells generally resistant to chemotherapy and able to rapidly regenerate tumours [138,139,140,141]. Although the existence of CSC in HCC is still debated, Chen et al. demonstrated that, in murine liver, CSCs, with the cell marker NANOG, modulate the expression of several mitochondrial genes (Acadvl, Echs1, and Acads), supplying energy by FAO [142].The enzymes involved in FA de novo synthesis and mentioned in the previous paragraph can show a role in tumoral progression as well. For instance, ACC1, the first rate-limiting enzyme in de novo lipogenesis, has been associated with HCC cell survival under metabolic stress, such as glucose deprivation, accounting it as an independent predictor of poor HCC prognosis [98]. FASN is responsible for palmitate (C16:0) synthesis from malonyl-CoA and cetyl-CoA [143]. Its fundamental role in HCC has been demonstrated by genetic ablation and pharmacological inhibition, both in vitro and in vivo [92,94,144]. Moreover, in human HCC, FASN acetylation is reduced and thereby protected by proteasomal degradation [145]. Destabilization of FASN by acetylation suppresses the growth of HCC. Gene expression profile studies in patients showed that higher expression of SCD-1 and SREBP1 are associated with poorer prognosis [92,94,144], underling a role in HCC progression and resistance. Budhu et al. found that palmitoleate (C16:1), the biological product of SCD1, facilitates HCC cells migration, while the ablation of SCD1 in HCC cells diminishes migration and xenograft development [88]. Furthermore, SCD1 overexpression confers sorafenib resistance in HCC, while SCD1 knockdown makes HCC tumour initiating cells more sensitive to sorafenib via ER-stress-induced unfolded protein response [146]. Similarly, down-regulation of CPT2 has been associated not only with hepatocarcinogenesis, but also with cisplatin chemoresistance of HCC cells [147].As already shown, several studies tried to deepen the mechanisms by which lipid metabolism rearrangement is involved in development and progression of HCC, mostly by gene expression analysis. However, as HCC cells display a different metabolic behaviour in vitro, rather than in vivo [148], this HCC heterogeneity imposes a profound understanding of the lipidic profile of individual cells. This is possible thanks to a branch of metabolomics, the Lipidomics, which was introduced in 2003, permitting researchers to characterise the diversity of FAs and other lipids in cells, tissues, organs, and organic fluids. Some studies already showed the importance of hepatic lipid composition in NAFLD and particularly in changes of free FAs ratios [149,150]. Patterson et al. described that serum of HCC patients is enriched in glycodeoxycholate, deoxycholate 3-sulfate, bilirubin, bliverdin, and other fetal bile acids, while lignoceric acid and nervonic acid, two very long-chain fatty acids (VLCFAs), were particularly increased in plasma of HCC subjects compared to cirrhosis [151]. Accordingly, VLCFAs are involved in inflammation as lipid mediators, showing a putative role in hepatocarcinogenesis [152]. Other studies have been subsequently published some common or controversial data, which we try to elucidate in this review. Some reports agree with the observation of increased saturated and monounsaturated FAs (SFAs and MUFAs) and with the contemporary reduction of polyunsaturated FAs (PUFAs) in HCC [153,154]. This was also related to HCC severity [153] and NASH [150]. Moreover, in Pten-null mice with NASH or HCC, an increased ratio of long n6-polyunsaturated FAs to n3-polyunsaturated FAs was reported [155], while in fat-1 transgenic mouse omega-3, polyunsaturated FAs dampened inflammation and tumorigenesis [156]. Phosphatidylcholine, containing palmitoleic acid or oleic acid, was reported as elevated in HCC by Morita et al. [157], while krautbauer described a reduction of this molecule [158]. Lu et al. also found high levels of phosphatidylcholine in HCC tissue, but the aim of their research was to find different lipidomic profiles between tissue and plasma in HCC patients, proposing plasmalogens (36:4) and (40:6) as potential biomarkers of diagnosis and tumoral progression [159]. Moreover, they described higher hepatic levels of six TAGs, one sphingomyelin (SM), and one ceramide (CM). In contrast, two works reported an important reduction of CM levels in HCC tissues with a concomitant increase of SMs [153,158]. This underlines an impaired activity of sphingomyelinase with a consequent lower conversion of SM to ceramide. The alteration of sphingolipid metabolism is associated with cancer, because levels of pro-apoptotic lipid ceramide are reduced, while levels of proliferative lipids SMs are elevated [160]. Another important lipidic product is Acylcarnitine, which has been proposed as a marker of HCC diagnosis and prognosis. Particularly, two studies showed higher levels of long-chain acylcarnitines and lower levels of medium and short-chain acyl-carnitines in HCC and serum of patients [154,161]. Acylcarnitine plays a central role in cellular lipid metabolism, as it is involved in the transport of activated LCFAs into mitochondria to sustain FAO [162]. The authors explain that the high ratio long-chain acylcarnitine/short-chain acylcarnitine is likely due to the high request of β-oxidation [154]. However, we have already discussed above that the altered expression of CPT1A (up-regulated) and CPT2 (down-regulated) could be responsible of low βoxidation and accumulation of acylcarnitine. Moreover, acetylcarnitine can be converted to malonyl-CoA with consequent inhibition of CPT1 activity and reduction of βoxidation [163]. However, we could assume that lipidomics of serum/plasma is a candidate to be used for detection of biomarkers, as blood withdrawal is easy and not invasive, thus studies are arising [154,159,161,164,165,166,167]. Chen and Passos-Castilho used serum lipidomics to differentiate patients with HBV-related HCC from HBV chronic hepatitis [164,166]. Similar analysis was performed on serum with discrimination of HCV-related HCC from patients with chronic HCV [167]. Fages et al. collected blood samples from HCC patients before and after diagnosis and identified 16 metabolites, involved in lipid and aminoacids metabolism and ammonium detoxification, which can differentiate HCC patients from controls and, most notably, can predict HCC development [165]. Recently, lipid profiling in HCC cells revealed that low levels of acyl-based glycerophospholipids, an important component of the cell membrane, were associated with metastatic activity [168].Collectively, these studies (Table 1) show that lipidomics contributes to the understanding of metabolic alteration in HCC, especially if combined with transcriptional studies, and provides potential new biomarkers of disease diagnosis and progression.In the adult liver, the hepatocyte is a highly specialized cell polarized with three different membrane domains: sinusoidal (basal), lateral (or inter-hepatocytic), and canalicular (apical). Moreover, the hepatocyte function varies widely within the lobule localization, as shown by different gene expression profile or biochemical activity [169,170,171]. Three zones in the liver sinusoid with different functions may be recognized: Zone 1 has the higher oxygen tension, as it is extended around the portal tract, receiving blood from hepatic artery and from the portal vein; zone 3 is around the portal vein, thus with a very low oxygen tension, while zone 2 is between zone 1 and zone 3. The lobular structure exposes groups of hepatocytes to different concentration of oxygen, nutrients, toxins, and intestinal-derived molecules, and, therefore, hepatocytes exhibit different metabolic functions, such as glycolysis, gluconeogenesis, and FA metabolism [172,173]. It is known that about 50% of the expression of liver genes are zonated, revealing, for example, that FAO and gluconeogenesis increase in the portal side, while lipogenesis is preponderant in the central side [170,174,175]. The zonal distribution of steatosis is not well understood yet, although it is known that steatosis is preponderant in the pericentral area and this heterogeneity is visible across the entire liver tissue [176,177]. Moreover, the severity and the localization of lipid accumulation have been correlated to NASH [178]. In this work, the authors report that, in liver biopsies from 500 patients with NAFLD, the severity of steatosis was positively associated with lobular inflammation and fibrosis in zone 3, whereas around the central veins, where oxygenation is low, a worse steatosis presented advanced fibrosis with Mallory body and ballooning. Accordingly, an association between phospholipids distribution and pro-inflammatory hepatic phenotype has been described [179]. Along these lines, it is plausible that distribution, storage, and metabolism of lipids change in a specific manner and in specific lobular zones, driving the progression of steatosis to NASH, cirrhosis, and cancer. The analysis of human specimens from simple steatosis revealed a zonation of expression of enzymes involved in phosphatidylcholine synthesis, confirmed by the zonation of specific phosphatidylcholines [179]. Intriguingly, in NASH biopsies, this lipid zonation is lost, identifying a potential mechanism of disease progression. Recently, Hall et al., analysing human and murine livers with advanced mass spectrometry imaging, identified several lipids with different zonations between the controls and NAFL, and notably they also observed the loss of zonation in NASH [180]. Furthermore, they asserted that the distribution of arachidonic acid-containing lipids drive inflammation in NASH pericentral hepatocytes, by releasing arachidonic acid from membranes with eicosanoids production [180]. However, HCC can occur in early stages of NAFLD without the development of cirrhosis or fibrosis, but the mechanism remains elusive [47]. One potential mechanism, described as the driver of liver tumorigenesis without fibrosis, is the activation of the β-catenin pathway [130].Wnt/βcatenin signalling has been recently indicated as a regulator of liver function and development and is responsible for hepatic zonation [172,173]. This pathway is suppressed in the periportal area and preponderant in the pericentral area, where it modulates the expression of several genes. As already mentioned, the aberrant expression of β-catenin drives a subset of hepatic tumours [172,173], thus, 25–40% of HCC presents mutation in genes with β-catenin activation. Knockout (KO) mice and expressing transgenic (TG) mice have been used to study the implication of β-catenin signalling in lipid metabolism [181]. Here, HFD-fed TG mice showed the obesogenic phenotype with predominant pericentral steatosis. In addition, expression of glycolytic and lipogenic genes was higher in HFD-fed TG mice than in KO mice [181]. In another study, mice with hepatocyte overexpression of HCV proteins (FL-N/35 model) presented a zonated pattern of lipids, as the lipid accumulation occurred in a couple of hepatocyte rows in the middle zone of the hepatic lobule [182]. In addition, the expression of FASN and other enzymes, such as SCD-1 and acyl-CoA synthase long-chain family member 3 (ACSL3), were increased. This particular zonation profile was validated in 50 human biopsies of HCV-infected patients. Notably, the authors found the up-regulation of genes involved in the wnt/β-catenin pathway in mice and in higher levels of nuclear β-catenin in human livers. Accordingly, Edamoto et al. documented wnt/β-catenin signalling involvement in HCV-related HCC [183]. Overall, wnt/β-catenin plays a key-role in lipid metabolism and zonation; however, the mechanism by which this is linked to tumorigenesis in the liver is not clear and further studies are eagerly needed.HCC is a very heterogenous tumour with many factors involved in pathogenesis and different local environmental conditions; therefore, finding new and safe molecular therapy is complicated. Although new antiangiogenetic drugs and checkpoint inhibitors have been approved as molecular treatments in HCC [184], the benefits in survival rate are limited. In the present review, we showed the centrality of lipid metabolic reprogramming in HCC, indicating potential therapeutic targets. Most of the described pathways are universally altered in HCC, while other genes might be targeted in specific environmental conditions, in a personalised manner. For instance, we have shown how FA synthesis is typically enhanced, while β-oxidation is heterogeneously modulated. Moreover, the lipidomics approach can integrate gene expression analysis, providing better knowledge of tumoral metabolic alterations with identification of potential biomarkers for diagnosis and prognosis. Lipidomics and gene expression data can be further combined with new technology imaging in order to define the zonation of lipid metabolic alterations. We believe that this multiple approach is essential for developing new therapies and efficient strategies in prevention and early detection of HCC.M.S. wrote the manuscript and prepared figures; R.V., F.C.; A.R. and D.L. revised the manuscript and figures; G.S. drafted the paper and supervised the work. All authors have read and agreed to the published version of the manuscript.Bando pubblicazioni scientifiche 2019–2020.This work has been published with a contribution from a 5 x 1000 IRPEF funds in favor of the University of Foggia, in memory of Gianluca Montel.The authors declare no conflict of interest.Hepatocellular carcinoma (HCC); hepatitis b virus (HBV); hepatitis c virus (HCV); non-alcoholic fatty liver disease (NAFLD); non-alcoholic steato-hepatitis (NASH); phosphate cytidylyltransferase 1 choline alpha (PCYT1A); choline kinase alpha (CHKA); vascular endothelial growth factor (VEGF); very low density lipoprotein (VLDL); triglycerides (TAG); sterol regulatory element-binding protein1 (SREBP1c); acetyl-CoA carboxylase (ACC); stearoyl-CoA desaturase (SCD); fatty acid synthase (FASN); direct-acting antivirals (DAAs); fatty acid (FA); alcohol related liver disease (ALD); alcoholic hepatitis (AH); phosphatidic acid phosphohydrolase (PAP); peroxisome proliferator activated receptor alpha (PPARα); fatty acid β-oxidation (FAO); tricarboxylic acid (TCA); ATP citrate lyase (ACLY); monounsaturated fatty acid (MUFA); peroxisome proliferator-activated receptor-γ coactivator beta (PGC-1β); steatohepatitic HCC (SH-HCC); diethylnitrosamine (DEN); high fat diet (HFD); carnitine palmitoyltransferase 1A (CPT1A); carnitine palmitoyltransferase 2 (CPT2); phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA);  Src-mediated c-jun NH-2-terminal kinase (JNK); reactive oxygen species (ROS); liver X receptor (LXR); transforming growth factor beta (TGFβ); suppressor of cytokine signalling 3 (SOCS3); CCAAT/enhancer binding protein alpha (C/EBPα); 5′ adenosine monophosphate-activated protein kinase (AMPK); hypoxia inducible factor 1 subunit alpha (HIF-1α); medium-chain acyl-CoA dehydrogenases (MCAD); long-chain acyl-CoA dehydrogenases (LCAD); mitochondrial acetyl-CoAsynthetase 1 (ACSS1); cancer stem cell (CSC); very long-chain acyl-CoA dehydrogenase (Acadvl); enoyl-CoA hydratase short chain 1 (Echs1); short-chain acyl-CoA dehydrogenase (Acads); saturated fatty acid (SFA); polyunsaturated fatty acid (PUFA); ceramide (CM); sphingomyeline (SM); very long chain fatty acid (VLCFA); long chain fatty acid (LCFA).Enhancement of fatty acid synthesis in hepatocellular carcinoma (HCC). ATP citrate lyase (ACLY); acetyl-CoA carboxylase (ACC); fatty acid synthase (FASN); stearoyl-CoA desaturase (SCD); fatty acid (FA); monounsaturated fatty acid (MUFA); tricarboxylic acid cycle (TCA cycle); sterol regulatory element-binding protein1 (SREBP1c); peroxisome proliferator-activated receptor-γ coactivator beta (PGC-1β).Different conditions influencing β-oxidation in HCC and proposed mechanisms. In a “lipid-rich condition” (e.g., obesity, non-alcoholic steato-hepatitis (NASH)) carnitine palmitoyltransferase 1A (CPT1A) is up-regulated, while CPT2 is down-regulated with consequent accumulation of pro-carcinogenic acylcarnitine and lower availability of acyl-CoA to sustain β-oxidation. Hypoxia has also been associated with β-oxidation suppression, as hypoxia inducible factor 1-α (HIF-1α) is induced and inhibits the expression of medium- and long-chain acyl-CoA dehydrogenases (MCAD and LCAD), two rate-limiting enzymes involved in the first β-oxidation steps. In contrast, in βcatenin-activated HCC, β-oxidation is fuelled by CPT2 activity, while peroxisome proliferator activated receptor alpha (PPARα) is up-regulated and induces the expression of LCAD and MCAD. During nutrient deficiency, 5′ adenosine monophosphate-activated protein kinase (AMPK) phosphorylates ACCα, permits CPT1A migration to the mitochondrial membrane to transport FAs and sustain β-oxidation. Fatty acid (FA); carnitine palmitoyltransferase 1A (CPT1A); carnitine palmitoyltransferase 2 (CPT2); medium-chain acyl-CoA dehydrogenases (MCAD); long-chain acyl-CoA dehydrogenases (LCAD); hypoxia inducible factor 1-α (HIF-1α); peroxisome proliferator activated receptor alpha (PPARα); 5′ adenosine monophosphate-activated protein kinase (AMPK); acetyl-CoA carboxylase alpha (ACCα).Main lipid composition alterations in HCC revealed by Lipidomics studies. Saturated fatty acid (SFA); monounsaturated fatty acid (MUFA); polyunsaturated fatty acid (PUFA).

Med-MDPI/cancers/cancers-12-06-01420.txt

@@ -0,0 +1 @@
+Concurrent activation of voltage-gated sodium channels (VGSCs) and blockade of Na+ pumps causes a targeted osmotic lysis (TOL) of carcinomas that over-express the VGSCs. Unfortunately, electrical current bypasses tumors or tumor sections because of the variable resistance of the extracellular microenvironment. This study assesses pulsed magnetic fields (PMFs) as a potential source for activating VGSCs to initiate TOL in vitro and in vivo as PMFs are unaffected by nonconductive tissues. In vitro, PMFs (0–80 mT, 10 msec pulses, 15 pps for 10 min) combined with digoxin-lysed (500 nM) MDA-MB-231 breast cancer cells stimulus-dependently. Untreated, stimulation-only, and digoxin-only control cells did not lyse. MCF-10a normal breast cells were also unaffected. MDA-MB-231 cells did not lyse in a Na+-free buffer. In vivo, 30 min of PMF stimulation of MDA-MB-231 xenografts in J/Nu mice or 4T1 homografts in BALB/c mice, concurrently treated with 7 mg/kg digoxin reduced tumor size by 60–100%. Kidney, spleen, skin and muscle from these animals were unaffected. Stimulation-only and digoxin-only controls were similar to untreated tumors. BALB/C mice with 4T1 homografts survived significantly longer than mice in the three control groups. The data presented is evidence that the PMFs to activate VGSCs in TOL provide sufficient energy to lyse highly malignant cells in vitro and to reduce tumor growth of highly malignant grafts and improve host survival in vivo, thus supporting targeted osmotic lysis of cancer as a possible method for treating late-stage carcinomas without compromising noncancerous tissues.Metastatic carcinomas express high levels of voltage-gated sodium channels (VGSCs), a feature that imparts an increased ability to invade normal tissue and to metastasize [1,2,3]. This feature of epithelium-derived cancer cells provides the basis for proposed treatments using cytotoxic agents that target VGSCs with the goal to destroy or to block the function of these channels in an attempt to eliminate the cancer [3,4,5,6,7,8]. This approach has been able to slow tumor growth and metastasis by negatively modulating VGSC expression or function, but unfortunately, it does not kill the cancer cells [4,5,6,7,8,9,10,11,12,13,14,15].By contrast, we have shown that the simultaneous administration of electrical or physiological stimulation of VGSCs and a pharmacological blockade of sodium, potassium-ATPase (Na+, K+-ATPase; sodium pumps) causes lysis of primary afferent neurons that over-express VGSCs without damaging neurons that express VGSCs at normal physiological levels [16]. We have thus proposed that, as in primary peripheral afferent neurons, the simultaneous activation of VGSCs and pharmacological blockade of Na+, K+-ATPase in cancer cells will result in an excess of intracellular sodium and increased osmotic pressure. In advanced-stage cancers that greatly over-express VGSCs, the rise in osmotic pressure is sufficient to cause osmotic lysis of the cancer cells, leaving unaffected cells that express VGSCs normally. When this “targeted osmotic lysis” (TOL) is administered to several lines of malignant cells in vitro using Na+, K+-ATPase-blocking cardiac glycosides in combination with electrical stimulation, we have been able to affect 100% lysis of the malignant cells. However, when TOL is similarly applied to ectopic xenografts of malignant breast cancer MDA-MB-231 cells in vivo, the response to treatment has been less complete [17].Two features of the TOL treatment as previously applied may contribute to limiting TOL efficacy and precluding optimum treatment in vivo. The first is the form of stimulation that we have used to activate VGSCs. Direct contact electrical stimulation introduces electrons into body tissues that then flow along the shortest path between the positively charged anodal source and the negatively charged cathode. The path of electron flow is determined by the degree of resistance imposed by the tissues that support the cancer cells. In biology, the tissues that comprise the extracellular tumor environment and the interstitial/extracellular tumor matrix provide a heterogeneous source of resistance that, as in electronics, can alter or inhibit the degree of stimulation that any given cell will receive [18,19,20]. Thus, as is evident in Figure 1, electrical current is not the ideal method for stimulating VGSCs especially in solid and deeply seeded tumors.The second feature that may limit the efficacy of TOL in vivo is the ability to deliver sufficient cardiac glycoside to effectively block all of the sodium pumps in the cancer cells within a solid tumor. Variable vascularization, refractory drug transporters, and the composition of the extracellular tumor matrix consisting of collagen, elastin, fibronectin and laminin and the resulting intratumoral hydrostatic pressure, can limit the distribution of many chemotherapeutic agents by impeding blood supply, thus diminishing the efficacy of promising treatment options [20]. To this point, a number of agents designed to degrade the extracellular tumor matrix, such as losartan, an anti-hypertensive and anti-fibrotic drug that inhibits collagen I synthesis, and sildenafil, an agent that is used to treat erectile dysfunction by modulating the peripheral vascular tone and the distribution of blood supply through the inhibition of phosphodiesterase type 5 (PDE5), have been used to enhance drug penetration and the efficacy of chemotherapeutic agents in the treatment of cancer [19,21,22]. We therefore hypothesized that it might be possible to improve the efficacy of TOL for destroying advanced stage cancer cells by utilizing a more effective method for stimulating VGSC activation and by improving the delivery of the cardiac glycoside that is used to block sodium pumps.It is well known that pulsed magnetic fields (PMFs) can be used to activate action potentials in excitable tissues [23,24,25,26]. Transcranial magnetic stimulation of motor cortex has been used clinically for the treatment of depression, movement disorders, post-traumatic stress disorder, migraine and chronic pain and has been shown to produce evoked potentials and movement in contralateral limbs [27]. As the delivery of magnetic field stimulation is much less influenced by tissue resistance than electric current stimulation [23,24,25,26], we hypothesized that PMFs would be a more efficient way to stimulate all cells in a body and improve lysis of invasive cancers. We therefore assessed the effect of TOL for treating highly malignant human (MDA-MB-231) and murine (4T1) breast cancer cells in vitro using PMF as the stimulus modality. PMF stimulation was also used in vivo to test the efficacy of TOL for treating ectopic xenografts and homografts of highly malignant murine breast cancer cells with and without losartan or sildenafil pretreatment to enhance the blood supply and delivery of cardiac glycoside.MDA-MB-231 cells suspended in Dulbecco’s modified Eagle’s medium (DMEM) with 500 nM digoxin lysed within 15 min in a dose-dependent fashion when stimulated with an 80-mT PMF (Figure 2). Maximum lysis achieved was 95–100% at the 500 nM dose compared to the 3–5% that was observed in the three control groups (drug alone, stimulation alone, neither drug nor stimulation).To assess the effect of PMF-induced TOL on normal cells, we used MCF-10a cells that minimally express VGSCs and do not metastasize. Using the 80 mT stimulus and 500 nM digoxin concentration, TOL had no effect on MCF-10a cells, with 96.5% of the cells appearing viable after treatment compared with 97.6% of the cells remaining viable after being treated with drug or stimulation alone (p > 0.8).To demonstrate that Na+ entry mediates this osmotic lysis, we assessed the effect of TOL in MDA-MB-231 cells that were incubated in Ringer’s solution with and without 500 nM digoxin and stimulated with 80 mT PMF. The 500 nM concentration is ½ log units below the minimally toxic concentration for digoxin alone. Sixty-seven percent of the cells that were suspended in Ringer’s solution with 500 nM digoxin and treated with PMF were lysed, compared to 15–22% lysis for the controls (p < 0.001). MDA-MB-231 cells that were suspended in Na+-free Ringer’s media and similarly treated were indistinguishable from controls (Figure 3).To establish a basis for comparing the efficacy of TOL for treating triple-negative breast cancer in an in vivo immune competent model, VGSC expression was evaluated in 4T1 murine breast cancer cells using immunocytochemical staining and flow cytometry and compared with similarly treated MDA-MB-231 cells. Figure 4 depicts similar patterns of labeling of VGSCs in MDA-MB-231 and 4T1 cells using a pan-specific antibody of a conserved portion of the VGSC. Flow cytometry evaluation of the two cell populations revealed protein expression of VGSCs in 4T1 cells was similar to the expression in MDA-MB-231 cells (4.1 + 0.09 vs. 3.9 + 0.12–fold greater than autofluorescence). The MDA-MB-231 cells had been previously shown to express VGSC protein about 7.5-fold greater than “normal” MCF-10a cells [17]. Therefore, 4T1 cells also over-express these channels.Using stimulus and dose parameters determined with MDA-MB-231 cells, 4T1 murine breast cancer cells were incubated in 500 nM digoxin and exposed to 10 min of an 80-mT PMF in order to compare the efficacy of TOL for treating triple-negative breast cancer cells that could establish a homograft tumor in an in vivo immune-competent (BALB/c) model. This TOL treatment significantly decreased the cells’ viability (Figure 5).To determine the effect of TOL on tumor destruction, ectopic xenografts of MDA-MB-231 cells were established in immune-compromised J/Nu mice. The mice were injected with digoxin (7 mg/kg), a dose that is 30% less than the lethal dose in 1% of mice, then stimulated with PMF for 30 min on days 1, 3 and 5. The mice were euthanized 24 h after the last treatment and the tumors were removed, sectioned and stained with hematoxylin and eosin. Tumors from mice treated with digoxin and PMF (TOL) showed 80–100% tumor lysis (tumors were not found in 2 treated mice). Of the harvested tumors, tumor viability was assessed by a veterinary pathologist who was blinded to the treatments and rated tissue damage on a 1–5 scale: 1 indicated no damage, 2.5 indicated moderate damage and 5 indicated complete destruction of the tumor (Figure 6). A part of necrosis observed can be attributed to damage seen during the natural history of a rapidly growing tumor and on average is assumed to be similar across all xenografted tumors. Tumors treated with TOL averaged 20–40% viability (60–80% necrosis) compared to 50–60% viability (40–50% necrosis) in control tumors (Figure 7). Drug-only and stimulation-only controls did not differ from untreated controls and no damage was observed in normal tissues following TOL treatment (Figure 8).To assess the effect of TOL using PMF on growth and survival, ectopic xenografts of either MDA-MB-231 cells or homografts of the highly malignant 4T1 murine breast cancer cells were established in immune-incompetent J/Nu nude or immune-competent BALB/c mice, respectively. The mice were treated as before and stimulated for 30 min on days 0, 2 and 4, then observed for 60 days. In both cases, survival was longer and tumor growth was slower in TOL-treated mice than in controls. Figure 9A shows the rate of tumor growth seen when treating MDA-MB-231 xenografts in nude mice. None of the TOL-treated mice met Nathional Institutes of Health – National Cancer Institute (NIH) criteria for humane endpoint euthanasia, but 3 mice in the group that received drug-only, 2 mice in the group that received only stimulation and 2 mice in the group that received the vehicle alone had to be sacrificed. Similarly, the rate of tumor growth seen when treating 4T1 xenografts in BALB/c mice is significantly slower (Figure 9B) and survival is significantly longer (Figure 10) when compared to controls. TOL treatment extended the time it took for 50% of the mice to reach NIH criteria for humane endpoint euthanasia by approximately 1 week (Figure 11).In an attempt to improve the delivery of digoxin and enhance the efficacy of TOL, pretreatment with losartan, a drug that inhibits collagen I synthesis, or concurrent treatment with sildenafil, a modulator of peripheral vascular tone, was added to the treatment protocol. Tumor growth or host survival did not affect the TOL-treated or the control mice (data not shown).The results of the present study are in vitro and in vivo evidence that support our earlier observation that targeted osmotic lysis, the simultaneous stimulation of VGSCs and Na+, K+-ATPase blockade, is able to kill highly malignant MDA-MB-231 human and 4T1 mouse breast cancer cells while preserving similarly treated normal breast cells [17]. In the in vitro experiments, the lysis of the cancer cells was dependent upon the intensity of the PMF, with greater than 95% of the digoxin treated cancer cells lysing within 15 min, with as little as 80 mT PMF. As with our previous report, when Na+ was eliminated from the incubation buffer, lysis of the cancer cells was no different than controls. Thus, TOL is Na+-dependent.We also demonstrated that the delivery of TOL using PMFs slows the growth of ectopic xenografts of both highly malignant human and homografts of murine breast cancer cells and increases survival of the hosts. Importantly, noncancerous organs in these animals were unaffected by the TOL treatment. Therefore, the cytotoxicity in tumors was not due to Rho/Rho kinase mediated apoptosis as has been seen with long-term treatment with cardiac glycosides [28]. The cells of normal tissues have only a fraction of the number of VGSCs compared to highly malignant cancer cells. Consequently, less Na+ and thus less water enters the cells, and they do not lyse [17].The efficacy of treating xenografts with TOL using PMFs was comparable to treating with TOL using electrical current for slowing the growth of both the xenograft and homograft tumors, and increasing the survival of the hosts than when the tumors were treated with PMF alone, digoxin alone or vehicle, but the effect of TOL seemed to be less focused when PMFs were used. Moreover in neither case did TOL negatively affect noncancerous organs. Thus, this treatment is likely to be selective for cancerous tissue, sparing healthy tissue and reducing morbidity compared to current cancer treatments.To assess whether our digoxin administration protocol was insufficient to perfuse tumors sufficiently, we used two treatments shown to increase chemotherapeutic pharmacokinetics [19,21,22]. Neither losartan, which degrades the extracellular tumor matrix [21], nor sildenafil, which increases vascular permeability [23], increased survival beyond our standard treatment protocol. Therefore, digoxin levels had achieved steady-state in all tissues of the body based on the drug’s pharmacokinetics, i.e., five doses separated by the digoxin t1/2.PMF stimulation has the added benefits of providing a noncontact mode of stimulation and enabling the simultaneous targeting of all tumors in metastatic disease, even when collections of metastatic cells are too small to be imaged. This is because the neoplastic cells that over-express VGSC carry both the targeting feature and the mechanism necessary for producing a therapeutic effect.In summary, we have shown in mouse models of breast cancer that PMF is a sufficient stimulus method to lyse carcinomas in a sodium-mediated TOL paradigm. Together, these results are strong evidence that PMF stimulation of VGSCs combined with blockade of Na+, K+-ATPase is a potentially efficacious treatment for late stage, highly malignant breast carcinomas.Digoxin was purchased from Sigma-Aldrich (St. Louis, MO, USA). For in vitro experiments, the drugs were diluted in the appropriate medium to a concentration twice that of the final concentration for that experiment. For in vivo experiments, digoxin was diluted in 10% dimethylsulfoxide (DMSO)/saline. Losartan and sildenafil were dissolved in saline.MDA-MB-231 human breast cancer cells were purchased from ATCC (Manassas, VA, USA) and cultured in Dulbecco’s modified eagle’s medium (DMEM; Gibco, Grand Island, NY, USA) supplemented with 10% fetal bovine serum (FBS; Gibco, Grand Island, NY, USA) and penicillin/streptomycin (pen/strep; Gibco, Grand Island, NY, USA). Insulin (10 µM; Sigma-Aldrich, St. Louis, MO, USA) was added 48 h before testing to assure sodium pump expression and activity. Mouse 4T1 breast cancer cells were also purchased from ATCC and cultured in RPMI-1640 media supplemented with FBS and pen/strep. MCF-10a normal breast epithelial cells were cultured in complete mammillary epithelial growth medium (MEGM; Lonza, Basel, Switzerland), which contains 10 µM insulin, supplemented with pen/strep.All of our procedures were approved by the Louisiana State University Health Sciences Center Institutional Animal Care and Use Committee (IACUC), which is OLAW approved (Assurance #D16-00058), AAALAC accredited (#000037), and US Dept. of Agriculture certified (#72-R-0003). For MDA-MB-231 cell xenografts, 6-week-old, 15–20 g, female J/Nu mice were purchased from Jackson Laboratories (Bar Harbor, ME, USA) and housed in a specific pathogen-free, climate-controlled colony room with a 12/12 h light/dark cycle with free access to water and lab chow. To establish xenografts, approximately 4 million cells were injected in a 50% matrigel/saline suspension. Tumors formed to 0.7–1.2 cm diameter in 3–4 weeks.For 4T1 homografts, 6-week-old, 15–20 g, female BALB/c mice were purchased from Jackson Laboratories and housed as with the J/Nu mice, but in a room with modified barrier controls. To establish homografts, approximately 500,000 cells were injected subcutaneously between the scapulae in a saline suspension. Tumors formed to 0.7 to 1.2 cm in 7 days. Treatment day 1 was initiated when average tumor size attained 5–10 mm in length as measured with a digital caliper by at least 2 investigators, one of which was not blinded to the treatment groups.MDA-MB-231 and 4T1 cells were dissociated using Cellstripper (Corning Life Science, Corning, NY, USA) centrifuged, decanted and resuspended in RPMI. 4% paraformaldehyde was added for 10 min. The cells were then centrifuged, decanted and washed with saline. The cells were resuspended in 5% goat serum + 3% bovine serum albumin for 1 hr. After centrifugation and aspiration of the supernatant, the cells were resuspended in PBS and incubated overnight with a pan-specific VGSC antibody (Alomone Labs; Jerusalem, Israel) that was pre-conjugated to an Allophycocyanin (APC) fluorophore, excitation 594 and 633 nm/emission 660 nm. The cells were centrifuged and resuspended in PBS and analyzed using a FACSCanto II flow cytometer (BD Biosciences, San Jose, CA, USA).MDA-MB-231 and 4T1 cells were incubated in RPMI for 2 days on a 22 × 22 mm slide cover. Media was aspirated and the cells were fixed with 0.5% paraformaldehyde for 10 min, rinsed 3 times for 5 min with 1 × PBS and then blocked with 5% goat serum + 3% bovine serum albumin for 1 h. Cells were then incubated overnight with the pan-specific VGSC antibody (Alomone Labs), 1:200 dilution. After overnight incubation with the primary antibody, cells were washed 3 times for 5 min with 1 × PBS. Cells were then incubated at room temperature for 45 min with a goat-anti-rabbit 488 Alexa fluorophore secondary, 1:800 dilution. Cells were washed 3 times for 5 min with 1 × PBS. Cells were incubated 10 min with DRAQ5 as a nuclear counterstain, dilution 1:600. Cells were rinsed and set with ProLong Gold antifade mountant. Images were taken on a Leica DMi8 confocal microscope (Leica Microsystems, Wetzlar, Germany) at 40 × oil magnification. As a control for nonspecific fluorescence, the anti-sodium channel antibody was pre-blocked with a 500-fold excess of the peptide antigen to which it was raised.Before testing, cells were dissociated with Cellstripper and approximately 150,000 cells resuspended in each 1.5 mL microfuge tubes using DMEM with or without digoxin (Sigma-Aldrich). Tubes were placed in an 8” length, 2.5” inner diameter solenoid with 697 turns (500 ft.) of 12 ga copper wire that gives an expected peak field strength of 4.4 mT/A with an expected resistance of 0.825 Ω. Based on measurements performed by driving 0.76 A at 0.60 V, the measured peak field strength using a F.W. Bell (Bedford, MA, USA) Model 4048 Handheld Gauss/Tesla Meter was 3.99 mT/A. The calculated resistance was 0.79 Ω. For stimulus/response experiments, 1–5 VDC current from an AE Techron (Elkhart, IN, USA) 7224 DC-extended AC amplifier was pulsed at 25 Hz, with a 10 msec ramp and fall, controlled by a Tektronix (Beaverton, OR, USA) AFG 3021B waveform generator, which produced 35–80 mT magnetic pulses in the solenoid. Cells were stimulated for 10 min. For dose/response experiments, cells were resuspended in 500 nM digoxin and stimulated for 10 min at 80 mT. Following each experiment, cells were fixed with 0.5% paraformaldehyde in phosphate buffer, pH 7.6, then stained with Cresyl violet. An evaluator blind to the treatment counted live vs. dead cells in at least 10 fields and a total of more than 100 cells for each treatment. Cells with a centrally located nucleus were scored as viable. Cells that were pyknotic or had no nucleus were considered nonviable.For Na+ dependency studies, cells were resuspended in normal Ringer’s solution [125 mM NaCl, 5.0 mM KCl, 2.0 mM CaCl2, 1.0 mM MgSO4, 10.0 mM glucose, 10.0 mM HEPES ] or Na+-free Ringer’s [NaCl was replaced with 250 mM sucrose to maintain osmotic pressure] with or without 500 nM digoxin. Approximately 150,000 cells were distributed to each 1.5 mL microfuge tube. Tubes were treated with 0 or 80 mT PMF, then evaluated as with the stimulus-response curve.For in vivo validation, groups of J/Nu mice (n = 8) hosting MDA-MB-231 xenografts were injected subcutaneously between the scapulae, five times at 1 h intervals with 7 mg/kg digoxin or an equal volume of the 10% DMSO/saline vehicle to bring digoxin or vehicle levels to steady-state for testing. The mice were treated on day 1 or on days 1, 3 and 5. On each of these days, were exposed to the PMF for 15 min starting 30 min after the last injection. For pathological analysis, mice were sacrificed by cervical dislocation and fixed with 0.5% paraformaldehyde in phosphate buffer 24 h after the last treatment. Samples of tissue from the grafted tumors as well as representative samples of kidney, spleen, skin, and muscle tissues were harvested, embedded in paraffin, sectioned and stained with hematoxylin and eosin. The tissue sections were then evaluated with light microscopy by a veterinary pathologist who is board certified by the American College of Veterinary Pathologists and was blinded to the treatments provided. Because the ranking was not continuous, the effect of TOL on tumors was compared using a Χ2 test.Groups of BALB/c mice with 4T1 homografts were injected with 7 mg/kg digoxin or the vehicle and tested as with the J/Nu mice.Post-treatment survival for MDA-MB-231-J/Nu and 4T1-BALB/c mice were evaluated using the same treatment parameters as above. The cross-sectional area of the xenografts were measured every third day for up to 33 days. Cross-sectional area was calculated by measuring the length and width of the tumors (d1 and d2, respectively) and calculating the assumed oval area using the formula: A = (d1/2) × (d2/2) × π) Mice were sacrificed when they met the NIH criteria for humane endpoint euthanasia of laboratory animals in cancer studies. In two experiments, one using MDA-MB-231 xenografts and another using 4T1 homografts, the tumor cross-sections were measured every other day and the mice were allowed to survive until humane endpoint criteria were met.Ectopic homografts of 4T1 murine breast cancer cells were established in female BALB/c mice. For TOL, mice were injected with digoxin (7 mg/kg), then stimulated with PMF (80 mT at 25 Hz) for 30 min on days 1, 3 and 5 and compared with controls. Groups of TOL-treated or control mice received either losartan (20 mg/kg/d; Sigma-Aldrich) beginning 2 days before TOL or sildenafil (1 mg/kg; Sigma-Aldrich) 15 min prior to each digoxin dosing and were compared to groups of mice that did not receive supplementary drug. Tumor growth was measured every other day. Animals were euthanized when they met NIH criteria for humane endpoint euthanasia.We have provided evidence that TOL can kill both highly malignant human (MDA-MB-231) and murine (4T1) breast cancer cells in vitro and can reduce growth of xenografts and increase the survival of mice in vivo when compared with controls while sparing normal cells and tissues. In addition, because of the highly conserved and essential nature of dynamic sodium channel–sodium pump relationship for intercellular communication and maintenance of intracellular homeostasis, TOL may be effective in treating a wide range of malignant carcinomas. We conclude that the novel, targeted osmotic lysis approach warrants further study as a potential option for treating advanced stage carcinomas.A patent for the technology described in this manuscript entitled, Targeted Osmotic Lysis of Cancer Cells—File No. 11M01 (Serial No. 13/552,909) Paul DJ and Gould HJ III was allowed on 12/30/2014.Conceptualization: D.P., H.J.G.III; Data curation: D.P., H.J.G.III; Formal analysis: D.P., F.D.P. and H.J.G.III; Funding acquisition: D.P. and H.J.G.III; Investigation: D.P., P.M., F.D.P., S.D.S., K.J.S., S.E. and H.J.G.III; Methodology: D.P., H.J.G.III; Project administration: D.P., H.J.G.III; Resources: D.P., H.J.G.III; Supervision: D.P., H.J.G.III; Validation: D.P., H.J.G.III; Visualization: D.P., H.J.G.III; Writing–original draft: D.P., H.J.G.III; Writing—review & editing: D.P., P.M., F.D.P., S.D.S., K.J.S., S.E. and H.J.G.III. All authors have read and agreed to the published version of the manuscript.This research was funded by intramural funds from the Departments of Pharmacology (D.P.) and Neurology (H.J.G.III), by Oleander Medical Technologies, LLC, and by a grant to the LSU Health Foundation from the Joe W. and Dorothy Dorsett Brown Foundation to support research. S.C.S. is supported by a Louisiana Board of Regents Fellowship for Graduate Training in Integrative Pharmacological Sciences.We would like to thank Charles Nichols for his expert advice and equipment support (NIH grant: P30GM106392 (CN)) during the conduct of these experiments. We would also like to thank the staff of the Guitar Center, New Orleans for their advice on amplifiers.D.P. and H.J.G.III are co-founders and managing members of Oleander Medical Technologies, LLC. Each holds an equity stake in this company, as does HJG’s spouse.Evidence of tumor lysis following treatment with TOL using electric current stimulation. The photomicrograph depicts an ectopic, MDA-MB-231 xenograft that was removed 24 h after 3 treatments with TOL (10 mg/kg ouabain; 10 V, 1 msec pulses, 15 pps, DC electric current for 5 min). Due to the size and configuration of the stimulating electrodes, the stimulus had been delivered to only the portion of the tumor between the arrows. Note the obvious difference in gross appearance between the stimulated and unstimulated portions of the tumor. The stimulated area is necrotic. Calibration bar = 5 mm.Stimulus-response curve for pulsed magnetic fields. MDA-MB-231 cells incubated in DMEM with or without 500 nM digoxin were treated with the indicated magnetic field stimulus intensity for 15 min. Subsequent cell counts revealed the stimulus dependence of TOL, with a maximum lysis rate of 95–100% cell death following TOL compared to 3–5% in controls. A two-way ANOVA revealed a main effect for treatment and a treatment X group interaction. * p < 0.05 by Tukey II post-hoc analysis.TOL dependence on Na+. The lysis of cells incubated in normal Ringer’s solution with 500 nM digoxin was 67% compared to the 15–22% lysis of cells in normal Ringer’s solution without digoxin when exposed to an 80 mT PMF. Counts of lysed cells incubated in Na+-free Ringer’s with digoxin and stimulated with PMF was comparable to controls. * p < 0.001 by planned orthogonal t-test.Sodium Channel Expression in MDA-MB-231 and 4T1 Breast Cancer Cells. Cultured cells were stained with a pan-specific anti-body for a conserved segment of the VGSC protein. Immunocytochemical imaging reveals labeling in cells of both MDA-MB-231 (A) and 4T1 (B) cell lines. The labeling depicted in the photomicrographs in Figure 4 was not observed in cells in which the anti-sodium channel antibody was pre-blocked with a 500-fold excess of the peptide antigen to which it was raised. The relative expression of VGSC proteins was confirmed quantitatively with flow cytometric analysis of the cell populations. Calibration bars = 15 µm).Targeted Osmotic Lysis of 4T1 Mouse Breast Cancer Cells. Cultured 4T1 cells in suspension were incubated for 15 min in DMEM + 500 nM digoxin or in DMEM alone, then stimulated with the pulsed magnetic field for 15 min. Assessment of viability was as with the previous experiment. (A): TOL-treated cells; (B): Vehicle-treated cells (calibration bars = 30 µm); (C): Cell counts comparing relative viability in samples that were treated with digoxin or vehicle and then exposed concurrently to an 80 mT PMF. *: p < 0.01 by t-test.Tumor viability rating. The photomicrographs depict representative tissue sections selected by the blinded veterinary pathologist to illustrate the morphologic features of tissue samples taken from MDA-MB-231 tumors rated at 1 (A; no damage (there is no necrosis), ×10), 2.5 (B; moderate damage (necrosis with pyknotic cells), ×20) and 4 (C; significant damage (necrosis with cavitations), ×10). To date, we have been unable to reliably achieve complete tumor destruction, a Grade 5 response. The rating system was applied to the evaluation of samples from >200 mice.Relative viability of homografts. The average viability scores of TOL-treated xenografts compared to controls as determined by a veterinary pathologist who was blinded to the treatment provided are illustrated. Sections taken from 35 mice were used to determine the mean ratings of tumor viability for TOL-treated and control mice. Note that the TOL-treated tumors averaged 20–40% viability compared to 50–60% viability in control tumors. The three control groups were collapsed and compared to the tumors treated with both digoxin and PMF. * Χ2 p < 0.05. The necrosis observed in control can be largely attribute to damage seen during the natural history of a rapidly growing tumor.Morphology of representative tissues taken from the kidney (A; ×10), spleen (B; ×10), skin (C; ×10) and skeletal muscle (‡) adjacent to a homograft tumor (D (*); ×1.25) treated with TOL. The morphology of these tissues were determined to be normal, (although the skin has a nonsignificant lesion) indicating that it is unlikely that they were affected by treatment with TOL.Growth of MDA-MB-231 xenografts (A) and 4T1 homografts (B) treated with TOL compared to the growth of xenografts that received drug or stimulation alone or vehicle. A. Groups of mice (n = 8) were treated as indicated. None of the mice that were treated with TOL (red curve) met NIH criteria for humane endpoint euthanasia. Three mice in the drug-only group (brown curve), 2 in the stim-only group (blue curve) and 2 in the vehicle-only group (green curve) met humane endpoint criteria and had to be sacrificed. B. The graph shows that the rate of growth of ectopic 4T1 homografts treated with TOL is significantly slower when compared to the growth of homografts that received drug or stimulation alone or vehicle. A one-way ANOVA revealed a main effect. * p < 0.05 compared to each of the controls by Tukey II post-hoc comparisons.Post-treatment survival. Treating ectopic 4T1 homografts with TOL significantly increased survival when compared to the growth of homografts that received drug or stimulation alone or vehicle. n = 12/group.Time to 50% mortality (meeting NIH humane endpoint criteria for euthanasia). Treatment with TOL consistently extended the time to 50% survival for mice with 4T1 breast cancer homografts by approximately 1 week when compared to controls. One-way ANOVA revealed a main effect. * = p < 0.01 by Tukey II post-hoc comparisons.

Med-MDPI/cancers/cancers-12-06-01421.txt

@@ -0,0 +1 @@
+All authors contributed equally to this work.Ovarian cancer is known for its aggressive pathological features, including the capacity to undergo epithelial to mesenchymal transition, promoting angiogenesis, metastatic potential, chemoresistance, inhibiting apoptosis, immunosuppression and promoting stem-like features. Galectins, a family of glycan-binding proteins defined by a conserved carbohydrate recognition domain, can modulate many of these processes, enabling them to contribute to the pathology of ovarian cancer. Our goal herein was to review specific galectin members identified in the context of ovarian cancer, with emphasis on their association with clinical and pathological features, implied functions, diagnostic or prognostic potential and strategies being developed to disrupt their negative actions.Recognizing the importance of the biological information dictated by cell surface carbohydrates, investigators found it necessary to isolate and characterize the glycoconjugates responsible for complex cell-surface glycan interactions. Galectins are a family of endogenous lectins defined as small, soluble β-galactoside binding proteins [1], which were first discovered in 1975 in the electric eel [2,3]. As of today, there are 16 galectin family members in mammals, which are numbered based on their order of discovery. However, some galectins identified in mammals were not found in humans. These include Galectin-5 (Gal-5), which has only been identified in rats, Galectin-6 (Gal-6) in mouse, Galectin-11 (Gal-11), and Galectin-15 (Gal-15) in sheep and goat [4,5]. The galectin family members are classified into three subfamilies, based on the structure and number of carbohydrate recognition domains (CRD) that they possess (Figure 1A). The CRD is a stretch of ~130 conserved amino acids [6]. The largest subfamily are homodimers, which include Gal-1,-2,-5,-7,-10,-11,-13, -14,-15, and contain a single CRD. The second subfamily are galectins with tandem-repeats, which include Gal-4,-6,-8,-9,-12. These galectins, in contrast to the homodimeric subfamily, contain double and non-identical CRD in a single polypeptide chain. Lastly, one of the galectins, Gal-3, is under the chimera subfamily. Galectin-3 contains one CRD connected to a non-lectin domain, capable of forming a multimeric structure [7].Galectins are commonly synthesized on free polyribosomes and primarily present in the cytoplasm, but are often shuttled between the cytoplasm and nucleus. Galectins are also expressed on the cell surface and in the extracellular matrix. Intracellularly, galectins are involved in the regulation of apoptosis and proliferation, as well as mediate pre-mRNA splicing [8]. The presence of galectins within and on the extracellular matrix of the cell allows them to mediate cell-to-cell and cell-to-matrix interactions, which depend on the glycosylation patterns of the cells. Those glycosylation patterns, in turn, reflect the underlying nutritional state of the microenvironment and availability of carbohydrate precursors. Although this review will focus primarily on the role of Gal-1, Gal-3, and Gal-7 in ovarian cancer, a brief description of some of the nonmalignant functions of these and other galectins is included below.Galectins are expressed on cells of the innate immunity (dendritic cell, macrophages, mast cells, natural killer cells, gamma/delta T cells, and B-1 cells). They are also expressed on members of the adaptive immunity (activated B and T cells) [9,10]. Their ability to regulate the innate and adaptive immune system has been extensively described in the literature [11,12,13]. Galectins appear to have a key role in the cell-cell interactions in inflammation. As an example, Gal-1 can act as a potential immunosuppressive agent, leading to a restoration of immune cell homeostasis in settings of autoimmunity and inflammation [14,15,16].Additionally, Toscano et al. reported that recombinant Gal-1 suppressed retinal inflammatory disease by promoting Treg cell-mediated anti-inflammatory response [17]. Other examples include the role of Gal-3 in IL-2-dependent T cell growth [18] and Gal-9 activating inflammatory cytokine gene expression by interacting with NF-IL6 in THP-1 cells [19]. In an animal model of arthritis, Lgals3-/- mice were found to develop less joint inflammation, less bone erosion, and lower amounts of IL-17 producing cells in the spleen, compared to their wild type counterparts [20].Galectins are distributed in a cell-specific manner and are often differentially expressed in tumor cells relative to the normal cells [21]. The aberrant regulation of galectins have been implicated in several cancer types, including head and neck [22], gastric [23,24], colorectal [25,26], bladder [27], melanoma [28,29], and gynecological cancers [30,31,32]. The function of galectins (Figure 1B) differs depending on the type of cancer, as well as whether the specific galectins are intercellular or extracellular. Functions such as apoptosis have been linked to Gal-1 [33], Gal-3 [34,35], Gal-7 [36,37] and Gal-9 [38]. However, galectins have had opposing functions described in different cancers. For example, Gal-1 induced apoptosis in prostate [33] and colon cancer [39], but had anti-apoptotic effects in cervical [40] and lung cancers [41].Similarly, the differential regulation of apoptosis was seen with Gal-3 [35,42,43,44] and Gal-7 [36,45]. Because galectins are found in the extracellular space, the cellular surface, the cytoplasm, and even in the nucleus, the location of Galectin proteins drives aspects of biological function. For instance, Gal-3 induced tumor growth, angiogenesis, and reduced apoptosis when present in the cytoplasm, and had anti-tumor effects when it is nuclear [46].Ovarian cancer is the most lethal of the gynecologic cancers in the United States [47]. Its lethality is due in a large part to its aggressive features and inability to be diagnosed at an early stage. Hence, the bulk of patients present, at a late stage, with a highly metastatic, invasive disease. Ovarian cancer is highly heterogeneous, with adenocarcinoma making up the bulk of the malignancies [48,49]. The most common histophenotype is high-grade serous ovarian cancer with the endometrioid, clear cell and mucinous subtypes being less prevalent. The current treatment strategy includes surgical debulking followed by a regimen of a platinum and taxane based chemotherapy, or alternatively, neoadjuvant chemotherapy with interim cytoreduction followed by additional cycles of chemotherapy. Unfortunately, the greater majority of the patients will present with recurrent platinum resistant disease in 36 months [50,51,52,53,54]. Many of the aggressive features of ovarian cancer are likely attributed, at least in part, to one or more of the pro-tumor galectin family members. Herein, we provide an overview of LGALS gene expression, the levels and locality of specific galectin proteins, their regulation at the mRNA and/or protein levels, and their potential functional role as it pertains to the genesis, progression, and overall pathology in ovarian cancer.Galectin-1 is encoded by the human LGALS1 gene, and similar to the other galectins, is characterized by its affinity for β-galactoside-containing glycans [55,56]. Galectin-1 exists in both a homodimeric and monomeric form. The different forms are associated with different functions, which have been reviewed in detail by others [57,58]. Galectin-1 has been shown to mediate cellular functions, including cell-cell interactions, cell proliferation, cell migration, adhesion, immune cell function, cell signaling, and apoptosis [57,59,60,61,62,63]. The full spectrum of intracellular and extracellular functions of Gal-1, like many other galectins, remains to be completely defined in its diverse cell type. Moreover, the recognized functions may differ during cell proliferation, differentiation, transformation to a benign hyperplastic state, or a malignant transformation. It is also possible that aberrant galectin expression, whether it be a result of a stress-related stimulus, where it is localized within the cell or on a neighboring cell, or in response to a gain or loss of function, may be a primary driver of malignant transformation or serve to promote of a more aggressive phenotype.Galectin-1 levels have been assessed in multiple tumor types [22,30,63,64,65]. In general, the majority of the studies described have shown a positive correlation with Gal-1 levels and increased tumor invasiveness and/or metastasis in areas including prostate, lung, and neuroblastoma [33,41,66,67,68].One of the first studies describing Gal-1 in ovarian cancer demonstrated an increased Gal-1 expression relative to the normal ovary [69]. The Gal-1 positive staining was reported to be heterogenous from sample to sample. Both tumor and stromal cells showed positive staining. It was noted that in those areas where there was invasive carcinoma, the stroma was more likely to be positive for Gal-1 than in the stromal cells distal from the tumor. The authors speculated that the presence of Gal-1 in the stroma was due to the release of the lectin by the neighboring tumor cells. In situ hybridization was performed to ascertain which cells expressed Gal-1 mRNA on an additional subset of tumors [69]. Albeit a small cohort, it revealed that Gal-1 mRNA expression was found predominantly in the stroma with moderate positivity in the cancer cells themselves. Interestingly, there was no correlation with Gal-1 protein levels (based on immunohistochemistry) and their corresponding clinical pathologic features [69].Van den Brûle et al. [69] extended their findings by evaluating the levels of Gal-1 in several ovarian cancer cell lines (AZ364, SKOV3, AZ224 OVCAR-3, AZ419, and AZ382), by Western blot analysis. Of the lines assessed, Gal-1 was only detected in the AZ364, SKOV3, and AZ224 and not in the OVCAR-3, AZ419, and AZ382 cell lines. The investigators utilized flow cytometry to assess the cell membrane levels of Gal-1, in an attempt to distinguish between intra and extracellular levels. Of those that they could assess, only the AZ364, SK-OV-3, and AZ224 cell lines had evidence of Gal-1 on the surface of the cell membrane, although it lacks a transmembrane domain and is likely bound to the glycosylation structures of cell surface proteins. Given that these investigators postulated that the Gal-1 was secreted, they also assessed the conditioned medium from their lines shown to be positive for Gal-1, suggesting that other, nonclassical mechanisms are responsible for its extracellular location. Thus, despite concentrating the media, they were unable to show evidence of Gal-1 by Western blotting techniques. This finding suggests that free Gal-1 was not passively or actively being released into the medium under the culture conditions described. Whether this would change under other conditions remains to be determined. Using the fibroblast line, 84BR, which expresses Gal-1, the authors investigated whether the secretion of Gal-1 could be modulated by conditioned medium from 6 ovarian cancer cell lines. Conditioned media from four of the six lines cultured with 84BR resulted in an increase in secreted Gal-1 [69].To determine whether Gal-1 influenced cell proliferation, they utilized bromodeoxyuridine (BrdU) incorporation as a readout. Recombinant Gal-1 was added to the cultures at increasing concentrations. No increase in cell proliferation was observed at the lower concentrations, and at the highest concentrations, there was a decrease in cell number in all the lines tested. It was suggested that maybe physiologically elevated Gal-1 in the stroma might serve as a defense against the invading tumor, although this explanation was not well justified. Finally, they demonstrated that the coculture with recombinant Gal-1 could enhance cancer cell binding to either fibronectin or laminin in three of the six lines, which matches the idea that Gal-1 can contribute to the invasive or migratory properties of a cell.In a second study by Kim et al. [70], the investigators assessed a small cohort of primary tumor samples from patients diagnosed with ovarian cancer by immunohistochemistry. Of the samples evaluated, the greater majority were of the serous subtype, with only a few endometrioid and mucinous subtypes. Although Gal-1 positive staining was present in the peritumoral stroma of all tumor samples, there was no evidence of Gal-1 in the normal ovarian tissues evaluated. Of the tumor samples assessed, they were divided into two groups; a low to moderate and high expression group. After determining the clinical correlates, it was determined that the high expressors were associated with advanced-stage disease, a serous histology, and the patients had evidence of a greater residual tumor volume after initial debulking surgery. Those tumors that had lower levels of Gal-1 in the peritumoral stroma were more sensitive to fist line taxane-carboplatin treatment, whereas those that had higher levels in the peritumoral stroma were more likely to be resistant and had a poor PFS (22.5 vs. 48 months) [70]. These findings were later corroborated by Zhang and colleagues [71] who used quantitative RT-PCR, Western blot, and immunohistochemistry analyses on fresh frozen or paraffin-embedded samples to demonstrate that high Gal-1 levels were positively correlated with advanced stage and poor prognosis. When broken down into stages, the investigators found that normal ovarian tissue had little to no Gal-1 mRNA or protein. The levels of Gal-1 mRNA and protein were greater in stage III-IV than in stage I-II samples.To assess the tumor samples via immunohistochemistry, the tumor sections were stained, scored, and divided up into low and high expressors. Based on this differentiation, they determined that the high expressors had a shorter PFS time than the low expression group.In the Kim et al. study, the investigators assessed the functional contributions of Gal-1 in ovarian cancer, utilizing the HeyA8, A2780-CP20, and SKOV3ip1 cell lines. Galectin-1 levels were shown to be most prominent in the HeyA8 and the SKOV3ip1 lines, as evidenced by Western blot analysis. Given that Gal-1 was present in the peritumoral region of the primary tumors, human endometrial fibroblast line T HESCs were also evaluated for Gal-1 presence. Western blot analysis revealed that these cells were also positive for Gal-1. The HeyA8 and SKOV3ip1 cells were subsequently transfected with a Gal-1 siRNA. After determining that the siRNA knockdown of Gal-1 was successful, 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) based assays were performed with the Gal-1 siRNA and Gal-1 control cells. The reduction in Gal-1 was concurrent with the decrease in metabolic activity, which was interpreted to be a decrease in cell proliferation. The A2780-CP20 line, which was initially determined to have a low level of Gal-1, was transfected with a recombinant Gal-1 protein, resulting in an increase in cell proliferation. The HeyA8 and SKOV3ip1 Gal-1 siRNA and the A2780-CP20 recombinant cells and their respective controls were then tested for their differential invasive properties in a Matrigel transwell system. The Gal-1 siRNA cells had a reduced invasion compared to their control. In contrast, the Gal-1 overexpressing cells displayed an increased rate of invasion relative to its control.In response to previous findings by others, suggesting that cancer cell-associated fibroblasts promoted the invasive and migratory properties of malignant cells [72], Kim et al. assessed Gal-1 levels in the conditioned media of the T-HESCs. The levels found in the conditioned media of the siRNA cells were less than their respective controls. Although not reported, it would have been interesting to assess the effect of the Gal-1 knockdown and overexpression T-HESCs when co-cultured with the HeyA8, SKOV3ip1, and A2780-CP20 cell lines. To begin to evaluate Gal-1’s role in platinum resistance, a siRNA strategy was used to reduced Gal-1 levels in A2780CP cells. The reduced Gal-1 levels corresponded with an increase in cells undergoing apoptosis. In contrast, forced expression of Gal-1 in Hey cells made them more resistant to cisplatin.Similar functional studies were performed by Zhang et al., whereby they also utilized siRNA techniques to assess the impact of reduced Gal-1 on cell proliferation, migration, and invasion in vitro. Of interest, they evaluated the levels of Gal-1 in SKOV3, CAOV3, SKOV3ip1, Hey, and A2780cp cell lines. In contrast to the study by Kim et al., the A2780cp Gal-1 levels were similar to SKOV3, CAOV3, and SKOV3ip1, and the Gal-1 levels in the Hey cells were low. Nevertheless, they showed that the siRNA induced reduction of Gal-1 in SKOV3ip1 cells corresponded with reduced cell proliferation and invasion. Unlike the transient transfection of recombinant Gal-1 in the Kim study, Zhang and colleagues used a lentiviral-based strategy to overexpress Gal-1 in the Hey cell line. The increase in Gal-1 was concurrent with increased cell proliferation rate, as determined by CCK-8 assay. Using the transwell invasion/migration assay, they further demonstrated that the rise in Gal-1 expression increased Hey cell proliferation, migration, and invasion, relative to their corresponding controls. Given the previously described role that Gal-1 had in facilitating membrane-associated Ras [40,41], Zhang et al. demonstrated that the downregulation of Gal-1 decreased H-Ras, p-Raf-1, and p-ERK expression. In contrast, increased Gal-1 expression corresponded with increased H-Ras, p-Raf-1, and p-ERK expression. This suggests a sustained activation of theses pathways, as depicted in Figure 2. Furthermore, using co-immunoprecipitation, they demonstrated that Gal-1 and H-Ras interacted; whether it was direct or indirect interaction, was not shown. Therefore, Gal-1 may interact with H-Ras to activate the ERK pathway and promote epithelial ovarian cancer pathology by promoting cell invasion and proliferation [71].Galectin-1 knockdown was previously reported to sensitize lung cancer cells to platinum-based chemotherapy [41]. There have since been similar reports in in vitro and in vivo models of ovarian cancer [71,75]. Zhang and colleagues used an siRNA-based strategy to down-regulate Gal-1 levels in cisplatin-resistant ovarian cancer cells (A2780CP cells) [71]. Following treatment with cisplatin, they found that cell proliferation was inhibited, and levels of apoptosis increased in a dose-dependent manner, relative to the controls. Interestingly, the cisplatin-resistant ovarian cancer cells transfected with Gal-1 siRNA were more responsive to the negative effects of cisplatin than those transfected with the control siRNA. These findings [71], and those of others [55,69,70,76], suggest that if a targeted disruption of Gal-1 expression/function in tumor tissue or the immediate tumor microenvironment could be obtained, it may be a plausible therapeutic option for cisplatin-resistant ovarian cancer.As discussed, previous studies of non-ovarian tumor types provided evidence to support the concept that Gal-1 promotes cancer cell invasion and/or metastasis, processes mediated, at least in part, by c-jun-NH2-terminal kinase 1 (JNK1) signaling [77,78]. Moreover, there is data that supports the idea that JNK1 signaling is associated with an epithelial-mesenchymal transition (EMT) [79,80,81]. Epithelial-mesenchymal transition is a critical process that often contributes to the invasive and metastatic potential of tumor cells [82]. Consequently, Zhu and colleagues investigated the relationship between Gal-1 and EMT in ovarian cancer. To accomplish this, they began by comparing the levels of Gal-1 and E-cadherin in a cohort of 107 samples, from patients diagnosed with ovarian cancer of varied histologies. It is important to note that this cohort had high- and low-grade serous cancer samples, as well as other low and high-grade histologies. The investigators reported that the higher levels of Gal-1 were closely associated with higher grade, more lymphatic metastases, and advanced stage. In addition, they determined that there was a negative correlation between Gal-1 and E-cadherin.Given the limited knowledge of how Gal-1 might mediate the enhancement of EMT in ovarian cancer, Zhu et al. [56] utilized the SKOV3ip and SKOV3 cell lines that were transfected with Gal-1 siRNAs or transduced with a Gal-1 lentivirus. Similar to other studies described above [70,71,75], they assessed basal Gal-1 levels in several ovarian cancer cell lines, A2780cp, A2780, SKOV3, SKOV3ip, and Hey cells. The SKOV3ip cell line demonstrated the highest levels of Gal-1, while their SKOV3 cells displayed the lowest level of the group. Armed with this information and the knowledge that galectin mediated functions are often dictated by their cellular location (cytosolic, nuclear, etc.) the investigators took the initiative to define the location of the Gal-1 in the two cell lines. In this case, both lines demonstrated a cytosolic presence. The subsequent decrease in Gal-1 in the SKOV3ip induced by the siRNAs corresponded with the reduction in invasive and migratory abilities of these cells, in a similar way to that seen by others [69,70,71]. Similarly, the expression increased as a result of the lentiviral infection of Gal-1 enhanced invasive and migratory activity of the SKOV3 control cells. To begin to explore whether these changes were the result of intracellular or secreted Gal-1, Zhu et al. used a Gal-1 antibody to ascertain if it would block the invasive or migration. Upon seeing no change, they speculated that the effect of Gal-1 was the result of intracellular mediated action.They further demonstrated that the downregulation of Gal-1 with a siRNA-based strategy resulted in an increase in levels of mRNA for E-cadherin and decreased levels of N cadherin, MMP7, uPA, and fibronectin snail and slug. The changes in E-cadherin and N-cadherin at the mRNA level corresponded to what was observed at the protein level. As anticipated, the opposite was found in the SKOV3 Gal-1, whereby overexpressing cells demonstrated decreased levels of mRNA encoding E-cadherin with an increase in mRNA for N-cadherin, MMP7, uPA and fibronectin Snail and Slug. Again, the changes in E- and N- cadherins reflected that of the changes observed at the mRNA level. Collectively, these results led the investigators to conclude that Gal-1 played a significant role in the EMT- mesenchymal epithelial transition (MET) plasticity of ovarian carcinoma cells [56].Based on previous studies suggesting that activated MAPK JNK/p38 signaling pathway could contribute to EMT in malignant tumors [83], Zhu et al. [56] assessed whether this pathway was involved in the regulation of EMT by Gal-1 in their siRNA Gal-1 knocked down SKOV3ip cells and Gal-1 overexpressing SKOV3 cells. The Gal-1 siRNA cells demonstrated a decrease in the basal phosphorylation levels of phosphorylated JNK and p38 in the siRNA Gal-1 knocked SKOV3ip cell line. Conversely, the lentivirus transduction of Gal-1 in SKOV3ip1 cells decreased the basal phosphorylation levels of MAPK JNK/p38. These data support the concept that elevated Gal-1 levels can promote EMT in ovarian cancer cells via MAPK JNK/p38 signaling.To determine if the activation of the MAPK JNK/p38 signaling pathway was concurrent with the regulation of Gal-1 on EMT in ovarian cancer cells, they utilized a pharmacologic approach. Specifically, they used the JNK antagonist (SB203580), the JNK/p38 antagonist (SP600125), and used a MAPK JNK/p38 agonist, anisomycin. Both the antagonists reduced N-cadherin and vimentin expression in the SKOV3-Gal-1 overexpressing cells, which corresponded with an increase in E-cadherin levels. However, the Gal-1 agonist, anisomycin, decreased E- cadherin expression and upregulated N-cadherin and vimentin expression in Gal-1 siRNA- transfected SKOV3ip cells. Treatment with anisomycin enhanced cell migration and invasion properties of the Gal-1 siRNA-transfected SKOV3ip cells. In addition, both SB203580 and SP- 600125 decreased migration and invasion properties in the SKOV3 Gal-overexpressing cells. Interestingly, these effects are evident in their respective controls. Nevertheless, their overall response to the agonists and antagonists provides evidence to support the potential importance of the MAPK JNK/p38 signaling in Gal-1 mediated EMT and metastasis in ovarian cancer.The contribution of MAPK JNK/p38 signaling to metastasis was affirmed in an in vivo model of ovarian cancer. Using the nude mouse tumor model, Gal-1 upregulation promoted the metastasis of SKOV3 cells. Importantly, treatment with antagonists of the MAPK JNK/p38 signaling pathway reduced the metastatic potential of Gal-1, overexpressing SKOV3 cells in mice [56].At this point, it is important to note that when comparing the various studies described above or to follow, there are contrasting data related to the basal or induced galectin levels in the multiple cell lines. Whether this is a result of different cell densities, variations in cell culture conditions, or detection methods used, is not yet known. Nevertheless, the investigators involved in the different studies have, for the most part, triaged them to high and low expressors and used them accordingly in their individual model systems. Therefore, rather than focus on the contrasting aspects, we attempted to provide a conceptual perspective.In a study by Park et al. [76], they were investigating the differences in toll-like receptor (TLR) mediated phosphoinositol 3 kinase (PI3K) signaling activity in CAOV3 and SKOV3 ovarian cancer cell lines. The investigators had postulated that Gal-1 might be a promising candidate for downstream targeting of the TLR/PI3k mediated signaling pathway in metastatic ovarian cancer. For their purposes, the CAOV3 line was supposed to represent a line from a primary tumor, whereas the SKOV3 was representative of a metastatic lineage. Of interest herein, Gal-1 levels were found to be elevated in the SKOV3 cell line when treated with a TLR4 agonist, LPS, a TLR3 agonist (poly I:C), and a TLR2/6 agonist MALP2, relative to the vehicle control. No effect was observed in the CAOV3 cell line. The TLR4-mediated Gal-1 also regulated migration and invasion in SKOV3 cells. Interestingly, pharmacologic inhibition of PI3K signaling blocked Gal-1 secretion in LPS stimulated SKOV3 cells. Based on these data, the investigators speculated that TLR/PI3K induced Gal-1 promoted the migratory and invasive capacity of ovarian cancer cells and potentially important to mediating ovarian cancer metastasis [76].Chen et al. [84] attempted to discern whether serum levels of Gal-1 could be used as a diagnostic for high-grade epithelial ovarian cancer. Differences in Gal-1 levels were detected via enzyme-linked immunosorbent assay (ELISA) in a pilot study assessing serum from healthy volunteers, patients with benign gynecologic tumors, patients diagnosed with epithelial ovarian cancer, or patients known to have another type of gynecologic cancer. Unfortunately, there was no difference in Gal-1 levels between healthy normal individuals, benign gynecologic tumor patients, and patients with high-grade epithelial cancer. Moreover, there was no difference in the levels of Gal-1 among the different histologies. There was, however, a difference among patients with non-metastatic epithelial cancer compared to metastatic disease [84]. Of interest, a subset of patients diagnosed with epithelial ovarian cancer had their blood drawn two days before and after debulking surgery. Nine of the 10 cases demonstrated a decrease in Gal-1 levels.These same investigators compared Gal-1 and CA-125 levels. In their cohort, 98 of the 140 patients identified as positive by CA125 were also positive for using Gal-1. A subsequent co-immunoprecipitation study found that CA125 was associated with a labeled Gal-1, supporting a previous report that CA125 might serve as a receptor for Gal-1 in HeLa cells [85].A more detailed analysis of the stroma associated with the ovarian carcinoma cells revealed that it was positive for Gal-1 in a majority of the samples scored. This is in contrast to the low to no positive Gal-1 staining in stroma found in normal ovarian tissues. It was further determined that the cancer-associated stromal staining was much higher in the invasive carcinoma when compared to non-invasive carcinoma. This was not the case when they assessed the Gal-1 in the cancer cells alone. In addition, the investigators revealed that the levels of Gal-1 in the cancer-associated stroma were positively correlated with stage. Furthermore, those patients with lymph node metastasis were found to have higher levels of Gal-1 stromal staining. If the cohorts were divided into weak and strong stromal staining, there was a correlation with the strong staining stroma and a recurrence rate in three years.Although overall survival (OS) did not show a correlation with the levels of Gal-1 in the nucleus of ovarian cancer cells, it was reported that the levels of Gal-1 in the cytoplasm were closely related [86]. In addition, they showed, via a multivariate analysis, that interstitial Gal-1 was an independent prognostic factor in ovarian cancer patients. Together, these results suggest Gal-1 has the potential to be prognostic or a marker of disease progression in epithelial ovarian cancer.The expression, localization, cellular distribution, proposed function and biomarker relevance for Gal-1 and other galectins in ovarian cancer discussed herein are summarized in Table 1.The LGALS3 protein is one of the more well studied of the galectin family members. It is the only chimera galectin found in vertebrates [2,95,96,97,98,99]. Similar to other galectins, its expression is found among several types of cells. It is involved in a broad range of physiological and pathological processes, including, but not limited to, cell adhesion, invasion, proliferation, cell cycle, metastasis, and apoptosis [8,96,100,101,102,103].Galectin-3 has a unique complex structure that contributes to its versatile roles. In the absence of ligand, the N terminus and C terminus are loosely bound via low-affinity binding block the carbohydrate-binding domain. A high-affinity ligand (i.e., lactosamine) can outcompete the low-affinity binding of the N terminal domain and bind to the CBD. This enables the N terminal domain to become available for homo-pentamerization. These pentameric structures allow for the binding of other glycosylated proteins (see Figure 3 [104]). The ovarian cancer effects of Gal-3 are dependent on common glycovariants. The affinity of the CRD for Gal-3 favors binding to complex, tri and tetra-antennary forms found on complex N-glycans, dependent on MGAT-5 action [97]. Recent observations have identified high affinity binding between the common cancer antigen Gal(β1–3)GalNAc(α1-O-Ser/Thr (Thomsen–Friedenreich antigen,) and Galectin-3, which appears to enhance metastatic behaviors [105].Intracellularly, Gal-3 is primarily found in the cytosol, but can also be found perinuclear, in the nucleus and near the mitochondria membranes [106,107]. Galectin-3 is also found in the extracellular space [96,108,109,110], whereby it can interact with a multitude of different binding partners. Its exit from the cell is believed to be due in part to exosomes and not via the more classic method of secretion from the endoplasmic reticulum or Golgi apparatus [111]. Once at the membrane surface, Gal-3 binds primarily polylactosamine-rich molecules located within the extracellular matrix or on the cell surface, where it can modulate functional properties (i.e., invasion, migration, and metastatic potential, etc.) that contribute to the tumor pathology/progression.The forms Gal-3 can take are dependent on its location. Extracellularly, Gal-3 binds glycans of glycoproteins or glycolipids, via its carbohydrate recognition site at the C terminal domain, resulting in cross-linking, which is thought to be mediated by its N-terminal noncarbohydrate-binding domain [96]. This self-association can result in the establishment of pentameric structures [104]. Through the complex binding of the Gal-3 pentamers, lattices are formed that can regulate the special position and interactions of growth factor receptors, including epidermal growth factor receptor (EGFR), 1-integrin receptors, N-cadherin and cytotoxic T-lymphocyte associated factor-4 (CTLA-4), among others, in a variety of cell types [66,109,112,113,114]. Consequently, these complexes are postulated to regulate complex cell functions, including cell signaling, glycoprotein trafficking and cell–cell adhesion [97,98,115].Apoptosis, one of many forms of programmed cell death, is mediated in large part by B cell lymphoma-2 (Bcl-2) protein family members [116,117,118]. The Bcl-2 protein family harbors Bcl-2 homology (BH) domains that mediate their interactions and execution of pro- or anti-apoptosis activity [116,117,118]. It is well understood that anti-apoptotic protein family members bind the BH3 domain of pro-apoptotic proteins like Bax and promote their oligomerization and subsequent pro-death activity at the mitochondrial level. In response to overwhelming cellular stress, whether it be endogenous or exogenous, the balance can shift in favor of the oligomerization of pro-apoptotic family members (i.e., Bax/Bax homodimers), leading to permeabilization of the outer mitochondria membrane, resulting in the release of cytochrome C (Cyt C). The release of Cyt C from the mitochondria triggers an enzymatic cascade orchestrated by members of the Caspase family, resulting in the ordered cleavage of DNA [116,117,118].Galectin-3 has been proposed as an inhibitor of the apoptotic response. Galectin-3 can translocate from the cytosol and/or the nucleus to the mitochondria, inhibiting the stressor induced disruption of the mitochondrial membrane potential and subsequent release of Cyt C [107,119]. Galectin-3 contains the anti-death aspartate-tryptophan-glycine-arginine (NWGR) motif, which is conserved in the BH1 domain of the Bcl-2 protein family. The NWGR motif within Gal-3 is located within the carbohydrate recognition domain (CRD). This motif is conserved in Bcl-2 protein family members, promoting balance by fostering the heterodimerization of pro and anti-apoptotic players (i.e., Bcl-2/Bax). In response to apoptotic stimuli, it has been shown that nuclear or cytoplasmic Gal3 can translocate to the mitochondria and interact with Bax in human thyroid carcinoma cells [102] (see Figure 4). This interaction of Bax and Gal-3 was confirmed by immunoprecipitation experiments and was only evident in response to doxorubicin [102]. The Gal-3 Bax interaction could be disrupted by GCS 100 (modified citrus pectin, a Gal-3 antagonist. These findings provide additional support for the concept that mitochondrial translocation of Gal-3 serves to tie up Bax and possibly other pro-apoptotic Bcl-2 family members, preventing them from forming pro-death promoting homodimers that initiate the cell death cascade.Using immunofluorescence and a TMA, Labrie et al. [87] showed that Galectin-3 was present in both the epithelial and stromal cells of the normal ovary. Again, using another TMA, they showed that Gal-3 was commonly found in all the histological subtypes assessed, including serous, endometrioid, mucinous, and clear cell. When focusing on the serous subtype alone, there was Gal-3 positive staining present in epithelial cells (~60%) and stromal cells around the tumor (~40%) in the samples assessed. They found no association between epithelial or stromal Gal-3 with stage, recurrence, or death.Wang et al. [120] conducted a meta-analysis of 36 eligible studies, looking at the prognostic role of Gal-3 expression in patients with solid tumors. The result of their meta-analysis was that there was a significant association of Gal-3 expression with OS in ovarian cancer.There is sufficient data to support the role of ovarian cancer stem cells in the repopulation of the recurrent tumor, contributing to the heterogeneity of the disease, and chemoresistance [121,122]. More recently, there have been reports implicating that Gal-3 supports the stemness phenotype in different malignancies [103,123]. This is not surprising, given that overexpression of Gal-3 can promote properties associated with stemness, including chemoresistance [31,89,90,123], enhanced sphere or colony-forming capacity [88], enhanced DNA damage response [124], increased tumorigenesis [125], etc. Consequently, Kang and colleagues [88] assessed multiple ovarian cancer cell lines for their capacity to form spheres. Of the cell lines tested, A2780, OVCAR3, OVCAR429, SNU-251, and SKOV3 were proficient in forming tumorspheres. The SKOV3 and OVCAR429 lines had high levels of Gal-3, and the A2780 and OVCAR3 were considered to have low levels of Gal-3. Using a shRNA strategy, they depleted Gal-3 in the high Gal-3 cells SKOV3 and OVCAR429. This resulted in a reduced number of spheres, and those spheres that developed were smaller in size when compared to the control shRNA treated cells. Moreover, they reported that the total number of cells that could form spheres was reduced as well. To assess the impact of increased Gal-3 levels, they transformed A2780 and OVCAR3 (low Gal-3 level lines) with Gal-3 containing plasmids. This resulted in an increase in the number of spheres, as well as the number of cells within a sphere. Likewise, the rise in Gal-3 levels corresponded with an increase in the number of cells that could form spheres. Of interest, the levels of CD133, which is a cell surface stem cell marker in ovarian cancer, increased in Gal-3 overexpressing A2780 cells. In addition, the levels of CD133 and Gal-3 were increased after OVCAR3 cells were cultured in sphere-forming conditions. Based on these data, the investigators surmised that Gal-3 contributed to the stem-like properties of ovarian cancer cells [88].The shRNA induced reduction of Gal-3 corresponded with a decrease in cell viability, and the increase level of Gal-3 was associated with an increase in cell viability. The A2780 cell line with elevated levels of Gal-3 was more resistant to cisplatin than their controls. The effect of increased Gal-3 levels on the OVCAR3 cells was not as pronounced, but there remained some level of resistance to cisplatin or paclitaxel, albeit at specific concentrations. This finding was similar to previous studies which demonstrated that treatment resistance was concurrent with an increase in Gal-3 levels in ovarian cancer cells [31,123].The shift in Gal-3 levels also corresponded with the cells’ invasive and migratory properties. The elevated Gal-3 levels equated to increased invasion and migration relative to their respective control cells, as determined by transwell and migration assays. The inverse was seen with the cells that underwent shRNA reduction in Gal-3 levels. To extend their in vitro findings, utilizing a xenograft nude mouse model of ovarian cancer, Kang et al. [88] demonstrated that the A2780 cells overexpressing Gal-3 had a larger tumor volume than their control counterparts, further supporting the argument that Gal-3 contributes to the overall pathology of ovarian cancer [88].It has been reported that Gal-3 exerts its pro-tumor promoting properties through the Thomsen–Friedenreich (TF) antigen [126,127], which occurs in 90% of all human cancer cells [128]. The TF antigen is clearly expressed in most ovarian carcinomas, with minimal expression in benign and normal ovarian tissues [129]. In ovarian cancer, mucins MUC1, MUC5AC, MUC6 and MUC16 are carriers of TF antigen [130,131,132]; among them, MUC1 is a natural ligand for Gal-3. The binding of Gal-3 to MUC1 can trigger a cascade of transmembrane signaling events. For example, in SKOV3 cells, Gal-3 binding activates the MAPK and PI3K/Akt signaling pathways, leading to the enhancement of cell proliferation and motility [133]. In non-ovarian tumor cell types, the binding of Gal-3 to cancer associated TF/MUC1 induces redistribution of MUC1 on the cell surface, increasing cancer cell homotypic aggregation and the formation of tumor micro-emboli [134]. It also promotes metastasis by enhancing cancer cell endothelial adhesion [135,136]. These latter properties have yet to be shown in ovarian cancer.Galectin-7 is a prototype member of the galectin family that is encoded by the gene LGALS7 on chromosome 19q13.2 [137]. Galectin-7, acting intracellularly or extracellularly, is involved in a variety of processes, including epithelial maintenance, cell adhesion, cell migration, and apoptosis [137,138,139].Kim et al. [91], evaluated the levels of Gal-7 and compared those levels with clinicopathological variables and survival outcomes, in a small cohort of paraffin-embedded samples derived from patients that were diagnosed with ovarian cancer. There were 63 samples of ovarian cancer with mixed histologies and five normal ovarian samples obtained from patients diagnosed with benign gynecologic disease. The authors reported week positive cytoplasmic and nuclear staining evident in the epithelial cells of the normal ovary. In contrast, much higher levels of Gal-7 were observed in the malignant samples. The cohort of malignant samples was divided into two groups (high vs. low), based on their immunohistochemical score. There was no difference with respect to stage, grade, histological type, and sensitivity to treatment among the two groups. However, patients with higher levels of Gal-7 had a more inferior OS; 72 months for patients with low levels of Gal-7 vs. 56 months for those with high levels of Gal-7. Using a multivariate analysis, they determined that advanced stage, platinum resistance, and high Gal-7 expression was consistently an independent prognostic factor for poor OS in their cancer patients. Labrie et al. [92] also evaluated Gal-7 in a TMA, which contained a total of 112 patient samples, including normal, benign serous and mucinous, borderline serous, and mucinous, serous, endometrioid, transitional cell, clear cell, and granular tumors. They found no detectable levels of Gal-7 in the normal ovarian tissue. However, they did find Gal-7 positive staining in the epithelial cells of all the ovarian cancer subtypes. Within this TMA, Gal-7 was found mostly in the cytoplasm of the tumor cells. They found no correlation with levels of Gal-7 and the age of the patients of stage of disease. They did find that Gal-7 was present more frequently in metastatic samples compared to non-metastatic samples.Moreover, Gal-7 levels were elevated in high-grade, borderline, and metastatic samples relative to benign tumors. The low-grade tumors also had lower levels than samples representing metastatic tumors. These same investigators also analyzed the public RNAseq datasets obtained from the cBio Cancer Genomics Portal (http://cbioportal.org), which revealed a correlation between LGALS-7 mRNA and a lower OS with ovarian serous cystadenocarcinoma.In the subsequent report by Schulz et al. [86], Gal-7 was found to be primarily present in the cytoplasm of the tumor cells, with just a few cases showing nuclear positivity. The level of expression varied among the 129 cases showing positivity, out of a total of 149 specimens. There were 15 samples considered to have high Gal-7 levels, 114 considered to have low levels of Gal-7, and 20 that were negative. Within the different subtype, the serous ovarian histological type had the highest levels of Gal-7, whereas the endometrioid had the weakest staining. There was a reduced OS for cases with high positivity for Gal-7 and a better OS rate for Gal-7 negative cases, when compared to cases with low expression of Gal-7. Their multivariate analysis supported the concept that higher Gal-7 levels could be an independent prognostic factor for OS in ovarian cancer. In general, these studies collectively imply that increased levels of Gal-7, as determined by immunohistochemistry, imply a poor outcome.Regarding the localization of Gal-7, there are reports that it is detected mainly in the nucleus of tumor cells [91], reports that it is localized in the cytoplasm and extracellular compartment [92], and reports that it is present only in the cytoplasm in a cohort of primary ovarian cancer samples [86], as evidenced by immunohistochemistry.Kim et al. [70] evaluated Gal-7 levels in Hey8A, Hey8A-MDR, SKOV3ip1, SKOVTR, A2780-PAR, and A2780-CP20 cell lines by Western blot analysis. Although not quantified, qualitatively, the Hey8A, A2780-PAR, and SKOV3ip1 demonstrated higher levels when compared to the SKOVTR, Hey8A-MDR, and A2780-CP20 cell lines. Examining the differential proliferation rates between the A2780-PAR line and the A2780-CP20 line, the A2780-PAR line, with higher levels of Gal-7, proliferated at a much higher rate than the A2780-CP line, which displayed lower levels of Gal-7. The reduction of LGALS-7 by siRNA strategy resulted in a decrease in mRNA and protein, which was concurrent with a decreased cell proliferation.Labrie et al. [92] highlighted the contrasting roles of p53 mediated Gal-7 effects in different subtypes. Given these deferential effects, they assessed Gal-7 levels in cell lines with differential p53 status. They showed that OVCAR-3 cells, which harbor a mutated p53, express Gal-7. However, they did not detect Gal-7 in A2780 and COV434, which have wild type p53, or in SKOV3 cells with a p53 null genotype. To test whether forced expression of a mutant p53 could alter LGLAS-7 mRNA and protein levels, they transfected SKOV3 and OVCAR3 cells and found that there is a concurrent increase in LGLAS-7 mRNA and protein levels, with the increase in p53 mRNA and protein levels. Likewise, the suppression of an endogenous p53 mutant by a siRNA diminished LGALS7 levels, as determined by RT-PCR.While Labrie and colleagues [92] found most of the immunohistochemistry based Gal-7 primarily in the cytoplasm of the tumor cells, they also found Gal-7 in the supernatants derived from both OVCAR-3 and in SKOV3 cells transfected with human Gal-7 expression vector, suggesting that it is released from the cell, either actively or passively. The levels of Gal-7 in the supernatant were determined by both Western blots and ELISA. Moreover, they used confocal microscopy to show that Gal-7 was present and that Gal-7 was bound to the cell surface and in the cytoplasm in OVCAR3 and SKOV3 cells. The investigators added -lactose to the culture media, which resulted in the decreased binding of Gal-7 to SKOV-3 cells, suggesting that the binding might be CRD-dependent. They further assessed Gal-7 binding activity in the A2780 and SKOV3 cells using a FITC-labeled recombinant Gal-7 and flow cytometry. Again, the addition of -lactose decreased the binding of Gal-7. Collectively, they surmised that Gal-7 was present intra- and extracellularly.Based on the finding that Gal-7 was found extracellularly, they went on to demonstrate that Gal-7 was released from culture ovarian cancer cells. For this, they again used the FITC- labeled recombinant Gal-7 and found it bound to the surface of Jurkat T cells in a dose-dependent manner. The binding was inhibited by the addition of -lactose, as well as an unlabeled Gal-7. The increase in binding was found to be associated with an increase in apoptosis of the T cells. They confirmed this effect in human PBMC, as evidenced by an increase in apoptosis in CD-14-positive monocytes and CD4- or CD8-positive T cells. Together, these findings support the idea that Gal-7 may have immunosuppressive effects.Galectin-8 is one of the tandem repeat galectins that contain two CRDs connected by a linker chain [140,141]. Galectin-8 is expressed in several tissues, including the lung, liver, kidney, brain, and myocardium [142]. It has been suggested that it mediates angiogenesis-related disorders of myocardial infarction, diabetic retinopathy, and various cancers, by promoting endothelial cell migration and tubular formation and regulating angiogenesis [143]. It has also been shown to induce cell arrest and apoptosis in Jurkat T cells [144].Given that Gal-8 has received relatively little attention as a prognostic marker in ovarian cancer, Schulz and their team [93] used the human protein atlas (www.proteinatlas.org) to conduct an in silico analysis of Gal-8 expression in normal and malignant ovarian tissue. Galectin-8 was not detected via antibody staining in ovarian stromal cells. However, 8 of 12 ovarian cancer tissues had what was deemed medium Gal-8 expression. Based on these preliminary findings, Schulz and colleagues [93] chose to evaluate 156 ovarian cancer samples using immunohistochemistry and compared the Gal-8 levels and its localization with clinical and pathological outcomes. Of the 156 ovarian cancer samples, they were able to evaluate 143 samples for Gal-8. These samples were of mixed histologies, with serous making up the bulk of the population (n=102). The remainder were of clear cell, endometrioid and mucinous subtypes. Galectin-8 positive staining was predominantly in the cytoplasm and nuclei of ovarian tumor cells. Galectin-8 positivity was not evident in the peritumoral stroma. High Gal-8 staining in the cytoplasm was evident in 67% of the samples, while 32% of the samples had low levels. They found that low Gal-8 expression in the cytoplasm correlated with lymph node metastasis, as well as higher stage. Half of the samples had Gal-8 positive nuclei. It was determined that positive nuclear Gal-8 staining was seen more often in stage I and II disease. Based on Kaplan–Meier analysis, those patients with high Gal-8 levels had a better DFS and overall survival. There was no difference in DFS and overall survival when comparing nuclear Gal-8 levels. Finally, a multivariate analysis revealed that Gal-8 positive staining served as a prognostic factor, independent of clinical and pathological variables.Labrie et al. also utilized a TMA for analysis of galectin expression in normal and diseased tissue [87]. Using immunofluorescence in a TMA constructed from normal ovarian and fallopian tubes tissue samples, they found that Gal-8 was present in both the epithelial and stroma of normal ovarian and fallopian tube tissue samples. The level of Gal-8 positive staining was much more evident in the tube than in the stroma. For the most part, the Gal-8 positive staining was present in the cytoplasm.In a second TMA described above, they evaluated the samples from 63 patients, representing serous, endometrioid, mucinous, and clear cell histology, and Gal-8 was commonly found in all histological subtypes. In yet another TMA representative of 209 specimens of ovarian high-grade serous cancer, approximately 55% of the samples stained positive for Gal-8 in the epithelial portion of the tumor, and roughly 30% stained positive for Gal-8 in the stromal area immediately adjacent to the tumor. Of interest, a univariate Cox analysis of the Gal-8 positive staining revealed that epithelial Gal-8 was correlated with chemoresistance [87], suggesting that the presence of Gal-8 may have some prognostic value.The differing results observed in the studies describing the presence and locality of Gal-8 could be attributed, in part, to the technique or the affinity of the different antibodies used. Moreover, there are reportedly seven different isoforms of Gal-8 encoded by the LGALS8 gene as a result of alternative splicing [145]. There are no reports of specific antibodies targeting any particular isoform of Gal-8. Therefore, only the total expression of the reactive Gal-8 isoforms is observed by immunohistochemistry. It is also not clear whether different anti-Gal-8 antibodies have a high affinity for a particular Gal-8 isoform [93]. Future functional studies may discern whether the isoforms display different actions.Labrie et al. [87] assessed the plasma levels of Gal-8 in 160 samples from healthy controls and 145 from patients diagnosed with ovarian cancer. They observed that Gal-8 plasma levels were significantly higher in patients with high-grade serous cancer (HGSC), as compared to the plasma from healthy controls. High Gal-8 plasma levels were associated with a lower 5-year disease-free interval and OS. High plasma Gal-8 levels were predictive for 5-year DFS and 5-year OS in patients with low cancer antigen 125 (CA125). The authors concluded that high Gal-8 might be a powerful molecular marker as a prognostic predictor of HGSC.Galectin-9 protein is transcribed from the LGALS9 gene, which is encoded on the short arm of chromosome 17 [146]. Gal-9 has been reported to have a number of biological functions, including contributing to innate and adaptive immunity [147]. Moreover, it has been shown to contribute to tumorigenesis and tumor pathology, by promoting cell transformation, cell cycle regulation, cell adhesion and angiogenesis [21,148,149]. Galectin-9 has been evaluated histologically and as a prognostic marker for several cancer types, including gastric cancer, non-small cell lung cancer, hepatocellular carcinoma, melanoma, and breast cancer [29,149,150,151,152].Similar to Gal-8, Gal-9 was found to be present in a TMA containing normal ovary and fallopian tube samples [87]. Compared to Gal-8 positive staining, the cytosolic staining for Gal-9 was weaker. Again, using immunofluorescence and another TMA representing serous, endometrioid, clear cell, and mucinous subtypes, it was observed that unlike Gal-8, which was found in most ovarian cancer subtypes, Gal-9 was not as readily prevalent in all subtypes. Interestingly, Gal-9 was found to be present in more than 80% of all clear cell and mucinous samples; whereas, Gal-9 was evident in less than 40% of the HGSC samples in that TMA. Using a larger TMA, which had 209 samples of HGSC, they showed that roughly 65% of the tumors had Gal-9 positive staining in the stromal compartment of the tumor. In contrast, only about 45% of the tumor cells stained positive for Gal-9. The staining was reportedly strong in the cytosol. More interesting, however, was that the tumor cells that stained positive for cytosolic Gal-9 often showed evidence of positive staining in aggregates in the perinuclear region. Based on other immunohistochemical stains, they determined that these aggregates were not autophagosomes of mitochondria. Galectin-9 was commonly found in the stroma immediately adjacent to tumor cells.With regards to clinical correlates, a univariate Cox analysis showed that epithelial Gal-9 punctum staining correlated with a lower 5-year survival rate. The Cox proportional hazards model showed that epithelial Gal-9 in the punctum was independently linked to a poor 5-year OS. There was no association when assessing stromal Gal-9. These same investigators [87] also showed that epithelial Gal-9 had a significant predictive value of 5-year OS in patients with low plasma levels of CA125, but not with high plasma levels of CA125.In contrast to Labrie and colleagues, Schulz et al. [93] reported that Gal-9 positive staining was primarily evident in the cytoplasm of ovarian cancer cells, but not in the nucleus or peritumoral stroma. In their cohort of 147 ovarian cancer samples, the high level of Gal-9 was related to a lower grade of the tumor, lower stage, and patients of a younger age. Furthermore, the majority of Gal-9 negative cases showed high-grade, advanced stage, and older age. In addition, cases with moderate Gal-9 expression showed a reduced PFS and decreased OS compared to Gal-9 negative cases.The differences observed in the studies described, with respect to levels and expression of Gal-9 and their clinical correlates, do not provide a great deal of confidence in their overall value. The functional data is likely more reliable as a predictor of outcomes.Functionally, Galectin-9 (Gal-9) has been implicated in regulating apoptosis in various types of cancer cells [153,154,155]. LALS9 was first described as being present in OVCAR3 cells by RT-PCR, albeit inconsistently [156]. Jafari et al. [94] went on to assess the anti-tumor effect of Gal-9 in the OVCAR3 cell line. They demonstrated that Gal-9 inhibited cell proliferation in a dose-dependent manner, as evidenced by an MTT assay. There was also a Gal-9 dose-dependent increase in intracellular levels of reactive oxygen species, which was concurrent with an increase in Caspase 3 and Caspase 6 activity. The addition of Gal-9 also resulted in a reduction in mitochondrial membrane potential (ΔΨm) in the OVCAR-3 cells. Moreover, there was a decrease in Bcl-2, which was inversely associated with a reduction of Bax. Collectively, these findings support their idea that an increase in Gal-9 would push OVCAR3 cells towards cell death via apoptosis. Furthermore, this effect is mediated by mitochondria. Overall, the functional data would suggest Gal-9 would likely serve to oppose Gal-3’s anti-apoptotic.Labrie et al. initially assessed the plasma levels of Gal-9 in a small cohort of ovarian cancer patients (n = 35) by ELISA. Gal-9 plasma levels were elevated in blood from malignant cases compared to the healthy controls. Therefore, as described above for Gal-8, they went on to measure Gal-9 levels in a larger cohort by ELISA. Their first finding was that plasma levels did not necessarily correlate with tissue levels in primary tumors. However, high plasma levels of Gal-9 were associated with a lower 5-year DFS and 5-year OS, and in the case of low plasma levels of CA125, Gal-9 was only associated with 5-year DFS [87]. A multivariate analysis showed that plasma levels of Gal-9 was an independent predictor of 5-year DFS and 5-year OS. The authors suggested that galectin levels in plasma, as well as the expression in tumor and peritumoral stromal cells, could potentially be used to predict 5-year DFS, chemotherapy response, and 5-year OS in HGSC patients and to potentiate the predictive value of CA125 [87].Collectively, these studies share disparate findings with respect to levels, distribution, and outcomes; whether or not this is attributed to the different several splice variants that were previously reported [26,157] or differences in technique, antibodies or types of analyses remains to be determined.As galectins play a pivotal role in the development of a variety of human diseases, blocking the effects of tumor promoting galectins has become a rational target for therapeutic applications, particularly Gal-1 and Gal-3. Novel inhibitors of both Gal-1 and Gal-3 have been developed, utilizing a variety of strategies, including blocking the CRD using oligosaccharides, small molecules, short hairpin RNA (shRNA), or monoclonal antibodies, with a varying degree of clinical success. The various sites of galectin action make the strategies for targeting potentially very different. The inhibition of protein glycosylation by inhibition of the glycosylation enzymes (such as inhibition of the key N-glycosylation enzyme MGAT V) can be highly effective, involving the reduction of both intracellular (apoptosis, nuclear effects, etc.) and extracellular functions (binding to N-glycosylation sites on key proteins) [158,159]. Such an intervention would necessarily affect all galectin functions and might lack specificity. In contrast, long-acting protein inhibition by truncated binding domains or monoclonal antibodies would be restricted to extracellular functions and might miss key functions. Small molecule inhibitors may have variable cellular uptake and could also penetrate the central nervous system, with potential effects that would be unlikely in antibody therapeutics [160].Exploiting the CRD of galectins and their preference for binding to galactose residue, the oligosaccharides GM-CT-01, GR-MD-02 (both are galactomannan polysaccharide that present N-terminal galactose residue), GCS-100 (a modified citrus pectic carbohydrate), and TD139 (bis-3-deoxy-3-[4-(3-fluorophenyl)-1H-1,2,3-triazol-1-yl]-b-d-galactopyranosyl[1]sulfane) have shown pre-clinical efficacy in diseases that are galectin mediated, including liver fibrosis, melanoma, myeloma, and lung fibrosis [161,162,163,164]. While GM-CT-01 has shown promise in early phase clinical trials in colon cancer [165], GR-MD-02 failed to improve fibrosis or liver-related outcomes in patients with nonalcoholic steatohepatitis [166]. TD139 has also shown promise in the treatment of idiopathic pulmonary fibrosis through its suppression of Gal-3 levels [167]. While these early results are promising, oligosaccharide derivates are broad-spectrum, often targeting more than one galectin, and can potentially result in undesirable side effects. Small molecule galectin inhibitors have been employed as treatment strategies against galectin-mediated tumorigenesis. OTX008 is a phenyl-based molecule that binds the CRD of Gal-1 and has been shown to inhibit tumor cell survival and angiogenesis in ovarian cancer cell lines and synergizes with chemo- and immunotherapies in vitro [168,169]. Other galectin inhibiting small molecules include 6DBF7, a dibenzofuran (DBF)-based peptidomimetic of the Gal-1, which has been shown to block tumor angiogenesis and tumor growth in melanoma, lung, and ovarian cancer mouse models [170]. GB1107 (3,4-dichlorophenyl 3-deoxy3-[4(3,4,5-trifluorophenyl)-1H-1,2,3-triazol-1-yl]-1-thio-a-Dgalactopyranoside) is a small molecule Gal-3 inhibitor that has been shown to reduce human and mouse lung adenocarcinoma cell growth [171]. Clinical data for either 6DBF7 or GB1101 have not been reported to date.Other galectin inhibiting strategies employed include shRNA and anti-galectin monoclonal antibodies. Li et al. achieved Gal-3 knockdown in vivo using a Gal-3 short hairpin RNA expressed in the pLKO.1 lentiviral vector [172]. The survival of animals was significantly improved with the Gal-3 inhibition treatment. To date, there are no reports of utilizing this strategy in large mammals. Monoclonal anti-galectin-1 antibody (F8.G7) was shown to decrease tumor angiogenesis and promote tumor regression in a mouse model of Kaposi’s sarcoma [173].Galectins have been recognized to play a vital role in an array of human diseases, and as such, have become the subject of interest as therapeutic targets in cancer therapies, as well as lung and heart diseases. A multitude of pharmacologic strategies have been undertaken to block the interactions between galectins and their saccharide partners. Some have shown promise in early clinical trials. Further research will help elucidate which galectin inhibition strategy works best in a clinical setting. Of importance herein there is sufficient evidence to suggest that disruption of pro tumor galectins may well be a potential strategy to explore for the treatment of highly aggressive ovarian cancer.Our objective herein was to review what is known about distinct galectin family members in the context of ovarian adenocarcinoma, with emphasis on their levels, extra and intracellular localization, association with clinical and pathological features, implied functions, diagnostic or prognostic potential and strategies being developed to disrupt their negative actions. Of all the galectin family members, Gal 1, 3, 7, and 9 appear to be the most relevant to the invasive, metastatic, chemoresistance and immunosuppressive properties of ovarian tumors. While the functional significance in the different studies, for the most part, aligned with each other, their prognostic or diagnostic potential remains uncertain. It is likely that as more specific antibodies covering the various isoforms are developed, some of the discrepancies will be resolved. Nevertheless, the evidence provided thus far by the shRNA, siRNA and pharmacologic inhibitors suggests that targeting galectin mediated effects could serve to augment standard of care approaches.Conceptualization, B.R.R. Writing–Original draft preparation, C.S., R.X., L.A.-A., M.S., D.R.S., and B.R.R; Review and Editing, L.A.-A., D.R.S., and B.R.R: Illustrations L.A.-A. All authors have read and agreed to the published version of the manuscript.This work was supported in part by the Advanced Medical Research Foundation (BRR), Vincent Memorial Hospital Foundation (BRR), China Scholarship Council (Grant Number 201908610120, (RZ)) and National Cancer Institute 1 P01 CA190174-01A1 (DS).The authors declare no conflict of interest.A schematic showing the different galectin structures and members of each (A). Galectins have been shown to play a role in altering many functions in cancer, including angiogenesis, apoptosis, tumor growth, immune escape, immune cell adhesion, cell transformation and metastasis/invasion (B).Oncogenic H-Ras plays a major role in tumor transformation via two major pathways, PI3K/AKT and MEK/ERK [73,74]. H-Ras recruits intracellular Gal-1 from the cytosol. This interaction enhances H-Ras-mediated cell transformation. Since Gal-1 has no effect on the membrane localization of inactive H-Ras, Ras activation, via GTP binding is needed for the H-Ras/Gal-1 interaction. Gal-1 is then able to enhance H-Ras-GTP, leading to an increase in Raf-1 recruitment, which culminates in a sustained activation of the MEK-ERK pathway and enhanced cell transformation [63].Gal-3 is involved in cell adhesion regulation, migration, invasion, angiogenesis, and metastasis. The specific extracellular Gal-3 function depends on the polymerization of Gal-3 into pentameric complexes. The action of Galectin 3 depends on glycan binding partners. These complexes link to glycans of high complexity (e.g., N-glycosylation lactosamine tetra-antennary forms and the Thomsen–Fredenreich antigen on O-glycans, especially in cancer. Through carbohydrate binding and polymerization to pentamers, Gal-3 forms a lattice and regulates the position of growth factor receptors, including EGFR, integrins and proteins like MUC16.Galectin-3 as an inhibitor of the apoptotic response. Gal-3 can translocate from the cytosol and/or the nucleus to the mitochondria, inhibiting stressors of the mitochondrial membrane potential and subsequent release of Cyt C. In response to apoptotic stimuli, Gal-3 can translocate to the mitochondria and interact with Bax and prevent its function, as well as other pro-apoptotic Bcl-2 family members, subsequently preventing the formation of pro-death promoting homodimers.Expression, localization, cellular distribution, proposed function, and biomarker relevance for galectins in ovarian cancer.

Med-MDPI/cancers/cancers-12-06-01422.txt

@@ -0,0 +1 @@
+Hepatocellular carcinoma (HCC), the most frequent primary liver cancer, is the sixth most common cancer, the fourth leading cause of cancer-related deaths worldwide, and accounts globally for about 800,000 deaths/year. Early detection of HCC is of pivotal importance as it is associated with improved survival and the ability to apply curative treatments. Chronic liver diseases, and in particular cirrhosis, are the main risk factors for HCC, but the etiology of liver disease is rapidly changing due to improvements in the prevention and treatment of HBV (Hepatitis B virus) and HCV (Hepatitis C virus) infections and to the rising incidence of the metabolic syndrome, of which non-alcoholic fatty liver (NAFLD) is a manifestation. NAFLD is now a recognized and rapidly increasing cause of cirrhosis and HCC. Indeed, the most recent guidelines for NAFLD management recommend screening for HCC in patients with established cirrhosis. Screening in NAFLD patients without cirrhosis is not recommended; however, the prevalence of HCC in this group of NAFLD patients has been reported to be as high as 38%, a proportion significantly higher than the one observed in the general population and in non-cirrhotic subjects with other causes of liver disease. Unfortunately, solid data regarding the risk stratification of patients with non-cirrhotic NAFLD who might best benefit from HCC surveillance are scarce, and specific recommendations in this field are urgently needed due to the increasing NAFLD epidemic, at least in Western countries. To further complicate matters, liver ultrasonography, which represents the current standard for HCC surveillance, has a decreased diagnostic accuracy in patients with NAFLD, and therefore disease-specific surveillance tools will be required for the early identification of HCC in this population. In this review, we summarize the most recent evidence on the epidemiology and risk factors for HCC in patients with NAFLD, with and without cirrhosis, and the evidence supporting surveillance for early HCC detection in these patients, reviewing the potential limitations of currently recommended surveillance strategies, and assessing data on the accuracy of potential new screening tools. At this stage it is difficult to propose general recommendations, and best clinical judgement should be exercised, based on the profile of risk factors specific to each patient.Non-alcoholic fatty liver disease (NAFLD) is considered to be the hepatic manifestation of the Metabolic Syndrome (MetS) and is closely related to obesity and insulin-resistance [1]. The spectrum of disease encompasses two phenotypes: simple fatty liver, defined by the accumulation of triglycerides in >5% of the hepatocytes, and non-alcoholic steatohepatitis (NASH), characterized by the presence of steatosis, ballooning and lobular inflammation at histology. NASH is considered the progressive form of the disease, eventually leading to fibrosis, cirrhosis and its complications, including hepatocellular carcinoma (HCC) [2].The prevalence of NAFLD has shown an increase in the last decades in parallel with that of obesity and diabetes mellitus [3]. Indeed, around a third of the global population is estimated to have NAFLD, with a prevalence that varies widely in different geographical regions being the highest in South America (30%), followed by Asia (27%), North America (24%), Europe (23%) and Africa (13%) [4]. In recent years, due to improvements in the prevention and treatment of chronic hepatitis C (HCV) and hepatitis B virus (HBV) infections, NAFLD become—proportionally—a major cause of liver disease and also one of the leading etiologies for end-stage liver disease and HCC. In addition, the burden of disease is expected to further increase [5]. Indeed, Estes et al. forecasted that by 2030 the prevalence of NAFLD in the United States will increase by 21%, from 83.1 million (2015) to 100.9 million (2030) cases, while the prevalence of NASH will increase by 63% from 16.5 million to 27.0 million cases. Accordingly, due to both disease progression and ageing of the population, the global incidence of decompensated NAFLD-cirrhosis is estimated to increase by 168% in 2030, while that of HCC is expected to increase by 137% [5]. Thus, due to the large global prevalence of NAFLD in the general population, the burden of advanced liver disease secondary to NALFD and NAFLD-associated HCC will soon be heavily felt, not only in Western countries, but reasonably also in Eastern ones, as in Asia, together with a growing industrialization and increase in Western diet pattern and metabolic diseases, the prevalence of NAFLD has worryingly increased in the last 20 years, being nowadays around 20% in Japan, about 30% in China, and as high as 51% in Indonesia [6]. Interestingly, in comparison with other Asian countries, the incidence of HCC among NAFLD patients in Japan seems to be 4-fold (incidence 4.8 per 1000 person years in Japan versus 0.3, 0.2 and 0.5 in Taiwan, South Korea and Hong Kong, respectively) [6]. Altogether, these data suggest that ethnicity, lifestyle, and social and economic conditions might contribute to the wide variation of the prevalence, phenotype and incidence of NAFLD and NAFLD-related HCC.Whereas the evidence for a high risk of HCC in NAFLD patient with cirrhosis is substantial and bi-annual surveillance with ultrasound (US) is universally recommended in these patients [2,7], there is increasing evidence that also NAFLD patients without cirrhosis can develop HCC, with a reported proportion of non-cirrhotic NAFLD among NAFLD-related HCC cases, as high as 50% [8,9,10,11,12,13]. However, data on this topic are scant and highly heterogeneous, as different definitions of NAFLD, NASH and stages of fibrosis have been used in different series. Furthermore, the majority of the available studies are retrospective, often including small cohorts of patients, and therefore underpowered and ultimately unable to provide solid evidence in favor or against surveillance in this population. On the other hand, a prospective analysis would require a very large number of cases and a prolonged follow-up. Therefore, the HCC risk-assessment among patients with NAFLD remains an unmet need, and it is currently unclear whether surveillance for HCC should be universally offered or only be recommended in a subset of patients carrying a clinically meaningful risk of developing primary liver cancer, where early identification of HCC is cost-effective. Decision-analysis studies have shown that in general an intervention can be considered effective when it is associated with an increase in life-expectancy of approximately 3 months, and cost-effective when it can be achieved at a cost of approximately 50,000 USD per year of life gained [14,15]. In patients with compensated cirrhosis surveillance is considered cost-effective when the annual incidence of HCC is ≥1.5%, and therefore this value is considered the threshold above which surveillance should be offered [16,17]. On the contrary, there are no published studies evaluating the cost-effectiveness of surveillance in non-cirrhotic patients. Therefore, the yearly incidence of HCC above which surveillance is cost-effective in the population of patients with chronic liver disease without liver cirrhosis is actually unknown, although it is certainly lower than in cirrhotic patients. Moreover, non-cirrhotic NAFLD patients with HCC can benefit more from early diagnosis, thus suggesting that the threshold for cost-effective surveillance should be placed at an annual HCC incidence <1.5%.In this review, we will summarize the most recent evidence on the epidemiology of HCC in patients with NAFLD and on the risk factors for HCC in patients with NAFLD. Based on these risk factors, we will highlight the sub-populations of patients with NAFLD where HCC surveillance is indicated or should be taken into consideration. We will also discuss the potential limitations of currently recommended screening and surveillance strategies, and the accuracy of potential new screening tools.Hepatocellular carcinoma is the sixth most common cancer, the fourth leading cause of cancer-related deaths globally, and the most frequent primary liver cancer, accounting for around 800,000 deaths/year worldwide [18,19]. Unlike most solid cancers, patients diagnosed with HCC are frequently not eligible for curative treatments and show high mortality rates with an incidence/mortality ratio close to one. In 2015, indeed, 854,000 incident liver cancer cases and 810,000 HCC-related deaths were reported [19]. Several factors are responsible for the low applicability of curative treatments in HCC patients, including late diagnosis, presence of comorbidities, older age, decompensated liver disease and/or poor liver function the low efficacy of systemic therapies. Furthermore, decompensated liver disease often limits even the application of loco-regional treatments with palliative intents [2]. In this context, the implementation of surveillance strategies for early detection of HCC nodules is fundamental, in order to increase the probability of access to curative treatments for these patients.Currently, chronic HBV infection accounts for 33% of liver-cancer deaths, followed by alcohol (30%), chronic HCV infection (21%), and other causes including NAFLD (16%) [19]. Whereas clear indications have been given for the surveillance of HCC in viral- and alcohol-related liver diseases, there is still controversy as to which NAFLD patients best benefit from HCC surveillance [20,21,22,23]. Patients with NAFLD have a 7-fold increased risk of HCC in comparison to the general population and, and among NAFLD patients, those with cirrhosis carry the highest risk, with an annual HCC incidence of around 10.6/1000 person-years (PY) of follow up [9,24]. Although this risk is lower than that for HCV-infected patients, the high prevalence of NAFLD raises cause for concern. In a recent analysis using steady state prevalence models, it was estimated that there are 64 million people in the US and 52 million people in Germany, France, Italy, and United Kingdom with NAFLD [25]. Furthermore, prevalent NAFLD cases are forecast to increase up to 101 million in the US by 2030, with NASH cases increasing from 1.5 million to 2.7 million [5]. Considering these estimates, the overall contribution of NAFLD to global liver cancer burden becomes comparable with that, or even greater of the other more established causes of HCC.Consistent with the above, Baffy et al. in 2013 showed that based on the estimates of the prevalence of HCV, HBV, alcoholic liver disease and NAFLD, as well as on the estimated incidence of HCC for each etiology, NAFLD may represent the relatively major contributor to the burden of patients with HCC (Figure 1), just behind HCV [26]. Taking these estimates into account, it is not surprising that NAFLD and NASH are the underlying cause of HCC in up to 59% cases in the US [27]. Moreover, a retrospective study conducted in the United Kingdom in the period between 2000 and 2010, found NAFLD as the underlying etiology for liver disease in nearly a fifth of HCC cases and was the etiology that showed the greatest increase in prevalence, registering a 35% increase during the study period [28]. Globally, age standardized death rate due to NAFLD-related liver cancer has increased annually by 1.42% since 2012, whereas there were no increases for viral hepatitis etiologies [29]. These data indicate that NAFLD is the most rapidly growing contributor to liver-related morbidity and mortality in the Western world. Not surprisingly, NAFLD is a growing indication for liver transplantation in industrialized countries, while NASH is the fastest growing cause of HCC in candidates for liver transplantation in the United States [30,31,32,33,34].Population attributable fraction (PAF) is the proportion of cases with disease that can be avoided by removing the underlying risk factor (for liver diseases HBV, HCV and NAFLD are examples). It is calculated using the prevalence (how common) and risk estimate (how strong) of the diseases. HCV and HBV are uncommon but strong HCC risk factors in the general population; however, their PAFs are less than that of NAFLD, as the latter is a weak but highly prevalent risk factor [26,35]. Therefore, increasing the awareness of the global burden, clinical manifestations and complications of NAFLD and implementing strategies for a correct and accurate estimation of the risk of HCC across the spectrum of disease is essential. Furthermore, surveillance should be implemented in sub-populations of patients where application of this standard of care results clinically meaningful and cost-effective.The 2016 EASL/EASD/EASO Clinical Practice Guidelines recommended HCC surveillance program with a 6-month interval US for patients with NASH-cirrhosis, and the same indication was very recently confirmed also by the AGA Clinical Practice recommendations [2,7]. The evidence for this recommendation is strong, as the association between NASH-cirrhosis and a significant increase in the risk of HCC has been extensively described. However, the evidence that surveillance for HCC in this population is above the threshold where this recommendation is cost-effective (i.e., an incident rate ≥ 1.5% per year) is not so solid [16].Available studies show contrasting results. In fact, in a prospective global study, conducted in patients with NAFLD and compensated cirrhosis at inclusion who were followed for a median mean of 85.6 months (range: 6–297 months), the annual incidence of HCC was reported to be 0.5%, which is well below the minimum 1.5% incidence rate threshold above which surveillance is considered cost-effective [36]. In contrast, a pivotal, although retrospective, study by Ascha et al. comparing 195 patients with NASH-related cirrhosis with 315 patients with HCV-related cirrhosis without a previous history of HCC found that, within a median follow-up of 3.2 years, NASH patients had a yearly cumulative incidence of HCC of 2.6%, which was lower than the 4% yearly incidence rate observed in HCV patients, but still higher than the threshold above which HCC surveillance is considered cost-effective and therefore recommended [37,38]. Similar results were obtained in a study conducted in Japan, where the 5-year incidence of HCC among 54 patients with biopsy-proven NASH-cirrhosis without HCC at inception was 11.3% [39].More recently, Kanwal et al. estimated the risk of HCC among patients with NAFLD seen in the United States National Veterans Health Administration system. The investigators included 296,707 NAFLD patients and 296,707 matched controls without any history of liver disease. NAFLD patients had a 7-fold increased risk of developing HCC (Hazard Ratio (HR) (95% Confidence Interval (CI)): 7.62 (5.76–10.09)) compared to controls, and among patients with NAFLD, those with cirrhosis had the highest annual incidence of HCC (10.6/1000 PY). Among this subgroup of patients, HCC risk ranged from 1.6 to 23.7 per 1000 PY based on other demographic characteristics. In more detail, men (HR: 11.05 per 1000 PY (9.83–12.39)) but not women (HR: 1.62 per 1000 PY (0.20–5.85)) had an increased risk of HCC, and a gradient towards greater risk was observed in patients aged ≥ 65 years [HR: 13.43 per 1000 PY (10.82–16.49) versus those aged < 65 years (HR: 9.74 per 1000 PY (8.46–11.17)) and in diabetics [HR: 12.36 per 1000 PY (10.67–14.24) versus non-diabetics (HR: 8.51 per 1000 PY (6.96–10.29)), with the highest risk of HCC observed in older Hispanics with cirrhosis [9]. These findings suggest that even among the cirrhotic NAFLD population, the risk of incident HCC is not the same for all patients but increases in patients with specific co-morbidities and/or demographic characteristics.Lastly, additional evidence supporting the increased risk of HCC among NAFLD patients with cirrhosis comes from a meta-analysis by White et al. that included 61 studies and was specifically aimed at identifying the risk of HCC in the NAFLD population. The results showed that NASH-cirrhosis was consistently related to an increased risk of HCC, with a cumulative incidence ranging between 2.4% over 7 years to 12.8% over 3 years in different series [40]. Again, even among the clinic-based studies included, the cohorts of patients with biopsy-proven non-cirrhotic NAFLD or NASH, showed a negligible risk of HCC, being 0% over an average of 21 years in a Danish cohort of NAFLD subjects without significant fibrosis, while a Swedish study reported an HCC-related cumulative mortality of 3% and 6% in subjects with NAFLD and NASH, respectively, followed for two decades [41].On the basis of this evidence, the AGA Clinical Practice Update on Screening and Surveillance for HCC in patients with NAFLD, confirms the indication to offer HCC screening to all patients with NAFLD-cirrhosis [7]. Noteworthy, the authors also recommend to enroll into screening programs those patients without a clinical diagnosis of cirrhosis but with at least two non-invasive markers suggestive for the presence of cirrhosis, such as FIB-4 (point of care) > 2.67 or Enhanced Liver Fibrosis Panel (serum-based specialized test) ≥ 11.3, and an elastography examination suggestive for cirrhosis (stiffness value ≥ 16.1 kPa) [42,43]. This recommendation is based on the evidence provided by Kanwal et al. that that a FIB-4 score ≥ 2.67 is associated with an increased risk of HCC, that is the highest in subjects with known cirrhosis (incidence rate (IR) 1.36% per year), while it is lower in patients with cirrhosis and low FIB-4 scores (IR 0.5% per year), in those without cirrhosis and FIB-4 ≥ 2.67 (IR 0.04% per year) and, as expected, the lowest in patients with low FIB-4 scores and no history of cirrhosis (0.004% per year) [9]. This finding further emphasizes the evidence that the main risk factor for HCC in NAFLD is cirrhosis, that surveillance in cirrhotic patients is justified and that non-invasive markers of advanced fibrosis, such as FIB-4 ≥ 2.67, even if associated with an increased risk of HCC, are not a reliable measure of HCC risk if used alone, since the associated incidence of HCC in patients with high FIB-4 and no other evidence suggestive for cirrhosis is negligible and well below the threshold for which surveillance would be cost-effective. Therefore, in the risk stratification of patients for HCC development, non-invasive markers need to be combined with coherent radiological or clinical parameters suggestive of cirrhosis.More recently, a study performed among 354 Mayo Clinic patients with NAFLD-cirrhosis, showed that diabetes (HR: 4.2, 95% CI 1.2–14.2, p = 0.02), age and low albumin significantly predicted the development of HCC, whereas other metabolic risk factors, such as increased Body Mass Index (BMI), hyperlipidemia and hypertension, did not [44]. This finding adds more evidence to the hypothesis that even among cirrhotic patients, it is possible to identify sub-groups at higher risk requiring a stringent surveillance strategy. These data may help narrowing the population of patients with cirrhotic NAFLD in which surveillance for HCC can be offered thereby improving its cost-effectiveness. Consistent with this view, although we do agree that on the basis of the current evidence surveillance should be offered to all cirrhotic patients with NAFLD, we feel that there may be sub-populations at higher risk (e.g., older male patients with diabetes) where the risk is greater and who can be the subject of enhanced surveillance. It is, however, important to note that, patients with NAFLD-cirrhosis and those with advanced fibrosis are a minority among the patients with NAFLD, and there is evidence that HCC might also appear in NAFLD patients with milder disease. Given the prevalence of NAFLD, these cases provide a relative greater contribution to the burden of disease due to the widespread prevalence of these milder liver conditions in the general population [26,45,46,47].Unlike patients with NAFLD and cirrhosis, the evidence supporting the cost-effectiveness of surveillance for HCC in NAFLD patients without cirrhosis is controversial. Indeed, while it has been repeatedly reported that, in comparison to the general population, patients with NAFLD without cirrhosis are at increased risk for HCC, the estimated incidence of HCC in non-cirrhotic NASH seems to be too low to justify screening [10,14,17]. In the VHA study by Kanwal et al., 20% of NAFLD-related HCC cases occurred in the absence of cirrhosis [8]. However, the annual incidence rate of HCC in NAFLD patients who had neither a diagnosis of cirrhosis nor a FIB-4 score ≥ 2.67 was too low to justify surveillance (0.04 per 1000 PY, 95% CI (0.04–0.05)) even if this population represented 87% of the at-risk study population [9].As shown in the meta-analysis by White et al., most of the studies addressing the issue of HCC risk among non-cirrhotic NAFLD patients had several limitations: They were retrospective, under-powered or heterogeneous with regards to the definition of NAFLD and NASH [40]. Moreover, in the studies included in this meta-analysis, the follow-up was too short for the studied endpoint (HCC incidence), ranging from a mean follow-up of 3 years to a mean follow-up of 13 years in just one 1 natural history cohort study [48]. Lastly and noteworthy, as most of the studies were underpowered, the authors could not perform a multivariate analysis aimed at defining the risk factors for HCC in non-cirrhotic NAFLD. It is worth mentioning though, that numerous case-control and cross-sectional studies showed a higher prevalence of diabetes and obesity among patients with NAFLD or cryptogenic liver disease with metabolic comorbidities, in comparison with controls with other causes of chronic liver disease.Additional evidence for the risk of HCC development in patients with NAFLD with no signs of advanced fibrosis or cirrhosis comes from small series [49,50,51,52]. Notably, in all these studies a highest prevalence of features of the MetS characterized patients who developed HCC as compared to patients where liver cancer did not arise, further supporting the idea that particular attention should be paid to patients with NAFLD and multiple additional risk factors. Mittal et al. also confirmed this hypothesis in a retrospective analysis of data from 1500 HCC cases from Veterans Health administration hospitals [8]. In this study, only 58% of NAFLD-related HCC cases arose in the context of cirrhosis, compared to patients with alcohol- or HCV-related HCC (72.4% and 85.6%, respectively; p < 0.05). Furthermore, patients with NAFLD-related HCC and MetS had a more than 5-fold higher risk of having HCC in the absence of cirrhosis compared to patients with HCV-related HCC [8]. Similarly, Dyson et al. analyzed the characteristics of 632 HCC patients referred for multidisciplinary meeting in the United Kingdom—in the period between 2000 and 2010—and showed that as much as 23% of NASH-related HCC cases were non-cirrhotic [28]. Interestingly, 31% of the patients with cirrhosis were classified as cryptogenic cirrhosis, but in this group of patients a higher prevalence of diabetes, obesity, hypertension and dyslipidemia was found, suggesting that this might have been cases of the so called “burn out NASH”, i.e., advanced cases of NASH-cirrhosis in which the histological hallmarks of NASH are no longer identifiable [28,53].Therefore, one could argue that in the context of non-cirrhotic NAFLD, patients with metabolic comorbidities are at higher risk of HCC. However, most likely, these triggers need to interact with other factors, such as ongoing inflammation and fibrosis, to promote carcinogenesis. Indeed, a recent Swedish cohort study, including 229 patients with biopsy-proven NAFLD, showed fibrosis as the main risk factor for overall mortality, cardiovascular mortality and liver-related events, including HCC [47]. In this study, during a mean follow-up of 26.4 years (range, 6–33 years), NAFLD patients had a 7-fold increased risk of HCC (HR: 6.55, 95% CI: 2.14–20.03; p < 0.001) as compared with the general population, similarly to the results by Kanwal et al. [9]. Interestingly, overall mortality was not related to NASH but only to the fibrosis stage as shown by the finding that mortality did not significantly increase in patients with NAFLD Activity Score between 5 and 8 and fibrosis stage 0–2 (HR: 1.41, 95%CI: 0.97–2.06; p = 0.07) but increased 3-fold (HR: 3.3, 95%CI: 2.27–4.76; p < 0.001) in patients with fibrosis stage 3–4, irrespective of NAFLD Activity Score. Thus, in non-cirrhotic NAFLD, fibrosis stage could be a useful parameter to stratify patients in different risk categories for liver-related events, including HCC.Consistent with these findings, Yasui et al. showed that among 87 biopsy-proven NASH patients with HCC, the risk of liver cancer tended to increase as fibrosis stage increased with a prevalence of advanced grades of fibrosis (3 or 4) in 72% of HCC cases. Another relevant finding was that, apparently and consistently with previous findings, male patients were at higher risk of HCC and tended to develop HCC at earlier stages of fibrosis [51].Recently, another systematic review with meta-analysis aimed at determining the pooled risk of HCC in patients with NASH, both in the presence and absence of cirrhosis, was published by Stine et al. [10]. The results of this analysis confirmed the indication to screen all patients with NASH-cirrhosis for HCC. Furthermore, the overall pooled estimate from the studies included in the analysis, accounting for 3567 HCC cases in 23,059 patients, indicated that not only the overall prevalence of HCC was higher but also that non-cirrhotic NASH patients had a near 3-fold increased risk of HCC in comparison to non-cirrhotic patients with other etiologies of liver disease (Odds Ratio: 2.61, 95%CI: 1.27–5.35; p = 0.009). Unfortunately, data on the fibrosis stage could not be extracted, and therefore the effect of fibrosis on HCC risk could not be calculated.Overall, these findings suggest that the risk of HCC in patients with simple fatty liver is negligible and that for patients with steatosis as the only risk factor for HCC, universal HCC surveillance may not represent a cost-effective strategy. However, there is increasing evidence that the presence of NASH and advanced fibrosis, male gender and metabolic comorbidities such as diabetes and obesity may identify sub-groups of patients for which surveillance might be cost-effective, as summarized in Table 1. Therefore, further research is needed to adequately identify those factors that are independently associated with an increased risk of HCC in patients with non-cirrhotic NAFLD. This will improve the identification of sub-categories of NAFLD patients that best benefit from surveillance and allow the implementation of treatment strategies aimed at modifying preventable risk factors for HCC development.Besides NASH and the presence of advanced fibrosis or cirrhosis, other independent risk factors for HCC have been recently described, with progressively increasing evidence (Figure 2). Among them, obesity, diabetes, male gender, older age, alcohol consumption and smoking each seem to be independent risk factors for HCC. It is well known that most patients with HCC have several risk factors and the HCC risk increases almost exponentially with the number of risk factors.Notably, insulin-resistance and obesity seem to play a pivotal role in HCC development in NAFLD, independently from the progression to cirrhosis [60,61,62,63], and might partially explain the high incidence of liver cancer in the non-cirrhotic NAFLD population. Indeed, in patients with features of the MetS as the only risk factor for liver disease HCC seems to have distinct morphological characteristics and mainly occur in the absence of significant fibrosis of the background liver [50]. Furthermore, the presence of multiple features of the MetS may act synergistically further increasing the risk of liver cancer up to 6-fold if two or more features of the MetS co-exist [64]; accordingly, there is evidence that the presence of diabetes (OR: 3.5; 95%CI: 1.3–9.2), obesity (OR: 3.5; 95%CI: 1.6–7.7), both conditions (OR: 5.2; 95%CI: 1.2–22.0) or of the MetS (OR: 2.13; 95%CI: 1.96–2.31, p < 0.0001), as defined by the US National Cholesterol Education Program Adult Treatment Panel III, is significantly associated with a higher risk of HCC [65,66,67].The association between diabetes and HCC is dated as far as 1986 and in the past two decades increasing evidence from large cohort studies has contributed to establishing diabetes as an independent risk factor for HCC [68]. Indeed, a population-based cohort study including 153,852 diabetic patients, showed that during 1,037,417 person-years of follow-up, patients with diabetes had a 4-fold increased risk of HCC (standardized incidence rate (SIR): 4.1; 95% CI: 3.8–4.5). The risk was higher in males (SIR: 4.7; 95%CI: 4.2–5.2) than in females (SIR: 3.4; 95%CI: 2.9–3.9), and was independent from other risk factors such as alcoholism, cirrhosis and viral hepatitis [69]. Similarly, in a more recent series from the Department of Veterans Affairs in the United States, including 173,643 patients with diabetes and 650,620 controls, the risk of HCC was significantly higher among diabetics (OR 2.39 versus 0.87 per 10,000 person-years, p < 0.0001) [70]. A 2- to 3-fold increased risk of HCC among patients with diabetes was also reported from a population-based study including patients from the Surveillance Epidemiology and End-Results Program—Medicare linked database; notably, the higher HCC risk in diabetics persisted even after the exclusion of other major risk factors (i.e., HBV, HCV, alcoholism) in this study as well [71]. Additionally, a high prevalence of non-cirrhotic cases of HCC was found in a multi-center observational prospective study performed in Italy, comparing 145 NAFLD-related HCC cases with 611 HCV-related HCC cases [13]. Cirrhosis was present in only about 50% of NAFLD-HCC patients, in contrast to the near totality of HCV-related HCC subjects but, as expected, also in this cohort the metabolic risk factors were more often present in NAFLD patients than in controls, although an analysis aimed at assessing the causality of this finding was not performed [13]. Similar results were described in Japan in a cross-sectional multi-center study including 87 histologically proven NASH patients with HCC of whom only approximately half had cirrhosis and, notably, most patients (62%) were male, obese (62%) and had diabetes (59%) [51].Of note, diabetes was not only found as an independent risk factor for HCC in Western series but also in studies performed in Eastern countries, with similar odds ratios [56,72,73]. Indeed, Kawamura et al. showed that in a cohort of 6508 NAFLD patients, those with diabetes had a 3-fold increased risk of HCC (HR: 3.21; 95%CI: 1.09–9.50; p = 0.035), independently of other risk factors [56]. Lastly, two meta-analyses have confirmed the increased incidence of HCC in diabetics independently from geographic location, alcohol consumption, history of cirrhosis, or viral hepatitis [74,75].Overall, the data presented provide strong evidence that diabetes increases the risk for HCC regardless of the presence of cirrhosis. Furthermore, diabetes has been shown to increase the risk of hepatic decompensation as well and to be related with poorer outcome after curative treatments for HCC [75,76].The link between obesity and cancer has been extensively reported in studies performed in Western and Eastern countries, with an observed positive linear trend in death rates from all cancers with increasing BMI [77,78,79,80]. The underlying pathogenic mechanisms are not fully understood but a direct effect of obesity on insulin resistance and the perpetuation of a pro-inflammatory milieu associated with obesity play a key role in DNA damage. Particularly, men with a BMI higher than 35.0 have a nearly 5-fold increased relative risk (RR) of death from HCC (RR: 4.52, 95%CI 2.94–6.94) as compared with men of normal weight [80].Even though the independent association of obesity with HCC in NAFLD is difficult to ascertain as this condition is frequently accompanied by diabetes and other features of the MetS, obesity has been accepted as an independent risk factor for liver cancer. A meta-analysis by Larsson et al studied the risk of HCC in a cohort of patients with normal BMI with that of a cohort of overweight and obese patients. The analysis showed that, compared to individuals with normal weight, those who were overweight or obese had respectively a 17% and 89% increased risk of liver cancer (RR: 1.17, 95%CI: 1.02–1.34 and 1.89, 95%CI: 1.51–2.36, respectively). Moreover, the RR for obesity was significantly higher for men (RR: 2.42, 95%CI: 1.83–3.20) than for women [81]. Further evidence shows that alternatively to BMI, other parameters reflective of body fat may be useful estimates of the risk of HCC. Indeed, in a study including 359,525 cases from the European Prospective Investigation into Cancer and Nutrition study, all anthropometric measures were positively associated with the risk of HCC [82]. Particularly, waist-to-hip and waist-to-height ratio showed the strongest association with HCC (RR comparing extreme tertiles 3.51, 95%CI: 2.09–5.87; p < 0.0001). Furthermore, weight gain during adulthood was positively associated with HCC as well (RR: 2.48, 95%CI: 1.49–4.13; p < 0.001) [82].In line with these findings, there is evidence that the risk for HCC can arise as far from childhood in overweight children as shown by a Danish registry including 372,636 children born between 1930 and 1989. The hazard ratio (95% CI) of adult liver cancer was 1.20 (1.07–1.33) and 1.30 (1.16–1.46) per 1-unit BMI z-score increase at 7 years and 13 years of age, respectively. This means that the association between childhood BMI and HCC slightly increased with the child’s age and by the age of 13, the risk had increased to 30% for each unit increase in BMI z-score. Of note, similar associations were found in boys and girls across years of birth and after accounting for diagnoses of viral hepatitis, alcohol-related disorders, and biliary cirrhosis [83].In the last 8 years three meta-analyses have addressed the role of obesity in the risk of primary liver cancer and have shown similar results with an overall agreement of an increased risk of HCC among overweight and obese patients, independently from other risk factors [84,85,86]. Abdominal obesity has also been related to a higher recurrence rate of HCC after radio-frequency ablation (RFA) in patients with suspected NASH [87]. Moreover, the presence of obesity has been reported to be associated with an increased risk of cancer-related mortality [80,88].Another widely accepted risk factor for HCC in all etiologies of liver disease, including NAFLD and NASH, is male gender [23]. The male-to-female ratio usually ranges between 3:1 and 5:1, but in selected series ratios between 7:1 and 9:1 have been reported [89,90]. Indeed, in all of the above mentioned studies on the incidence and risk factors for HCC in NAFLD, a significant male preponderance of HCC is reported and several studies have shown that males have an up to 4-fold increased risk of HCC as compared with females [13,23,37,56,91,92,93,94]. In the study by Yasui et al., interestingly, male subjects developed HCC at a younger age and at earlier stages of fibrosis than females [51]. This finding might be related to a protective effect of estrogen hormones against HCC development [95], to a potential favoring effect of androgens on HCC [23,96] and to the higher prevalence of MetS in the male population with NAFLD [51].It is well known that the incidence of HCC increases with advancing age, with a peak in the seventh decade [16,23,27,40]. In line with this evidence, the vast majority of studies have shown that patients with NAFLD-related HCC are usually senior [8,13,39,49,57] and significantly older than patients with HCCs occurring on the background of other chronic liver diseases [8,50]. Moreover, patients with non-cirrhotic NAFLD-related HCC are frequently older than patients with NAFLD- related HCCs and cirrhosis [97]. Finally, two studies reported that older age was an independent risk factor for NAFLD-related HCC [11,57]. These findings might reflect a longer exposure to liver damage and, possibly, the onset of more severe degrees of liver fibrosis; unfortunately, a limitation of most studies addressing the issue of HCC in NAFLD is their retrospective design and the low rate of liver biopsies, therefore sub-analysis taking into account liver fibrosis among different age strata were impossible to perform.Another demographic factor that has shown a significant preponderance in Western studies among NAFLD patients developing HCC is Hispanic ethnicity. Couto et al., in a cohort of 1266 liver transplant recipients between 2000 and 2010, reported that Hispanics were more likely to develop NAFLD-related HCC as compared with non-Hispanics (58% vs. 30%; p = 0.018) [98]. This evidence was later confirmed by an analysis from the data collected in the Surveillance Epidemiology and End-Results registries and from the liver cancer mortality data from the National Center for Health Statistics in the United States, where Hispanics aged over 50 years had a significantly higher incidence of HCC [99]. In addition, Hispanics born in the United States were found to have a higher incidence of HCC when compared to foreign-born Hispanics, suggesting that environmental, socio-economic, and cultural differences may be contributing factors as well [100]. Probably, the higher prevalence of NAFLD-related HCC among Hispanics is related to the higher prevalence of features of the MetS, NAFLD and NASH in this population, as described by earlier studies and by more recent literature where higher rates of NAFLD in Hispanics were observed, with a highest risk of HCC in older Hispanics with NAFLD-cirrhosis [8,9,100,101]. Moreover, a genetic predisposition favoring HCC onset among Hispanics has been described [102].Genetic predisposition may play a role in HCC development in NAFLD patients. Indeed, homozygous carriers of the I148M variant protein of PNPLA3 have a 2-fold higher hepatic fat content than non-carriers and are at higher risk of NAFLD [102]. A study by Liu et al. demonstrated that the PNPLA3 rs738409 C→→G single nucleotide polymorphism, which encodes the I148M variant protein, had a gene-dosage effect for which an increased number of G alleles (i.e., homozygous G allele) was associated with an increased incidence of NAFLD-HCC, independently from the presence of cirrhosis, with an odds ratio as high as 12.19 when compared with the general United Kingdom population [58]. Two meta-analyses have confirmed these findings showing that cirrhotic carriers of the PNPLA3 variant protein had an increased risk of HCC (OR: 1.40, 95%CI: 1.12–1.75) [103] and that this higher HCC risk persisted even after adjusting for age, sex, and BMI [104]. More recently, Grimaudo et al. performed a prospective study on 471 patients with histologically proven NAFLD which demonstrated an independent association between PNPLA3 C > G variant and HCC (HR: 2.10, 95%CI: 1.03–4.29; p = 0.04 [59]. These results suggest that PNPLA3 genotype could be a useful tool to stratify the risk for HCC in the NAFLD population, of course, used in combination with other accepted risk factors.Additional genetic variants that have been associated with an increased risk of hepatic steatosis and of progressive liver disease and fibrosis are the rs58542926 C > T variant in the TM6SF2 gene and the MBOAT7, particularly in non-cirrhotic patients [105,106] but the incidence rate, determinants of risk and direct role of these genetic variants in HCC predisposition requires further investigation. Although the definition of NAFLD excludes the consumption of significant amounts of alcohol, the assessment of alcohol consumption is not easy to perform with precision in clinical practice as it depends on the reliability of patients’ history and is extremely difficult to ascertain in retrospective research databases. However, Ascha et al. reported that, in the context of NAFLD, any amount of alcohol consumption increases the risk of HCC and, noteworthy, even a history of social alcohol intake was associated with an increased risk of HCC as compared to non-drinkers and alcohol consumption was the strongest independent risk factor for the development of HCC (HR: 3.8, 95%CI: 1.6–8.9, p = 0.002) [37].Another risk factor for HCC is tobacco smoking [107]. Even though evidence in circumscribed NAFLD cohorts is limited, an International cohort study showed that HCC incidence among NAFLD patients with advanced fibrosis who were cigarette smokers was approximately twice as higher than that of non-smokers (HR: 2.11) [93]. Furthermore, smoking accounts for 13% and 9% of HCC globally and in North America, respectively, as shown by the analysis from The Liver Cancer Pooling Project, a consortium of 14 United States-based prospective cohort studies that includes data from 1,518,741 individuals and 1423 cases of HCC. Based on these data, current smokers had increased risk of HCC (HR: 1.86, 95%CI: 1.57–2.20), while individuals who had quit for more than 30 years had a risk near equivalent to never smokers. Importantly, smoking is also associated with lower survival rates in HCC [108,109]. Therefore, complete abstinence from any alcohol as well as from tobacco smoking should be recommended in patients with NAFLD, and these recommendations have recently been endorsed by the American Gastroenterological Association [7].Based on these studies, it is reasonable to assume that greater public health efforts are needed to implement the treatment of modifiable HCC risk factors, such as increased body weight and tobacco and alcohol consumption, in the NAFLD population with the aim of possibly reducing the risk and incidence of primary liver cancer. On the other hand, from a surveillance perspective, the development of an algorithm including some or all of the abovementioned risk factors for HCC in NAFLD (i.e., age, fibrosis stage as assessed with biopsy or non-invasive fibrosis assessment tools, presence of diabetes, BMI, presence of obesity from child age, smoking status, alcohol assumption, and possibly genetic predisposition) is needed to stratify the risk of HCC at the individual patients’ level, through the identification of clusters of non-cirrhotic NAFLD patients for which surveillance for HCC would be cost-effective. Indeed, recently, the usefulness of a risk stratification based on age and alanine aminotransferase levels (ALT) levels in NAFLD patients without cirrhosis has been suggested by a Japanese study including 18,080 patients with NAFLD and no evidence of cirrhosis [11]. In this study, the 10-year cumulative incidence of HCC was found to be highest in older patients (age > 55 years) with increased ALT (12.41%, 95%CI: 5.99–18.83). These initial findings may suggest that HCC surveillance in non-cirrhotic NAFLD could be initiated at an older age, and likely further stratified according to the presence of ALT increase. Of course, we are well aware that further investigation is needed before such a risk-stratified surveillance model can be generalized. Similarly, Ioannou et al. created an HCC risk estimation model aimed at identifying those cirrhotic NAFLD patients at higher risk of HCC with the objective of stratifying patients into risk categories and further improving the cost-effectiveness of screening [54]. The models were developed separately for NAFLD-cirrhosis and ALD-cirrhosis and included seven predictors: age, gender, diabetes, BMI, platelet count, serum albumin and aspartate aminotransferase to ALT ratio. The models showed an area under the receiver operating characteristic curve (AUROC) of 0.75 for NAFLD-cirrhosis but showed poor accuracy in the prediction of the risk at the single patient level. Nevertheless, this study showed that a risk-based screening according to this prediction model was associated with a higher standardized net benefit in comparison with an approach to screen all cirrhotic patients.To date, screening for HCC with US at 6-month intervals in NAFLD is recommended for cirrhotic patients [23,110]. However, this imaging technique presents challenges in the NAFLD population as whether the accuracy of ultrasound in patients who have a good acoustic window is adequate for the identification of early HCC, its accuracy is significantly lower in obese patients, due to the thickness of subcutaneous fat and its sound-attenuating properties, especially at increased depths, thus resulting in low imaging definition [111,112,113]. Therefore, it has been questioned whether for these subjects the use of different imaging techniques would be a more appropriate screening and surveillance tool for the early detection of HCC. Technically, Computed Tomography (CT) scan and Magnetic Resonance Imaging (MRI) have higher sensitivity than US, but these imaging techniques also have lower specificity, which would carry a higher rate of false positive results, thus triggering further investigation with repeated scans and referrals [114,115]. Moreover, CT scan and MRI are expensive and, at least for CT, expose patients to radiation risks, and therefore their routine use for the surveillance of HCC is currently not recommended, and US remains the recommended method for HCC surveillance in NAFLD patients [7,115]. However, when the quality of US images is unacceptably low, CT scan or MRI may represent a valid alternative [7,23].To increase the detectability rate of HCC at a screening level, it has been questioned if adding alpha-fetoprotein (AFP) to routine bi-annual US would be of help. However, among patients with NAFLD-related HCC, a higher proportion of AFP non-secretors has been reported and in most HCC cases occurring on a background of fatty liver disease or steatohepatitis, the serum levels of AFP were is low [13,50], an observation often suggestive of a less aggressive tumor biology [116,117]. In fact, AFP is normal in 76%–82.5% of patients with very early HCC without viral etiology, normal aminotransferases and absence of cirrhosis, features that are commonly observed in patients with NAFLD-related HCC [117]. Therefore, the usefulness of AFP testing to increase the detectability rate of HCC during screening in NAFLD patients is universally considered to be low. Moreover, no specific recommendation for HCC screening in the NAFLD population has been developed so far. As previously reported, it is widely accepted that all NASH-cirrhosis patients should be screened for HCC, but it is also evident that the risk of HCC for these subjects is lower than that for other CLD. In this context, non-invasive accurate biomarkers are needed combined with available imaging techniques in order to improve the quality of HCC surveillance in NAFLD high-risk patients.PIVKA-II, glypican-3 (GP3) and Squamous Cell Carcinoma Antigen -1 (SCCA-1) had been proposed as new HCC-biomarkers in a study including 19 patients with histologically proven NAFLD-related HCC and 31 patients with alcohol-related HCC [118]. The authors showed that the sensitivity and specificity of GP3 was poor (sensitivity = 68%, specificity = 46.3%), whereas the serum levels of SCCA-1 were not statistically different in patients with or without HCC. In contrast, the combination of AFP and PIVKA-II testing had higher sensitivity (94% versus 58%) than AFP alone, at a modest expense of specificity (80.5% versus 100%). However, the added benefit was only limited to the detection of more advanced HCC, therefore questioning the advantage of this test in terms of curative treatment candidacy and survival benefit.Gray et al. performed a pilot proteomic study [119] aimed at identifying novel non-invasive markers of HCC in the setting of NAFLD with. CD5L, a novel serum protein, was identified in the sera from cirrhotic individuals with and without HCC. Even if the performance of CD5L alone as a surveillance marker for HCC was poor (AUROC = 0.495), the authors argued that, given its increasing trend in cirrhosis but not per each level or stage of fibrosis, it may be reflective of hepatocyte regeneration rather than fibrosis per se. Hence, if used in combination with other serological markers it might be useful to identify patients at higher risk of HCC, although we do feel that further research in this field is needed.Recently, novel potential non-invasive biomarkers of HCC have been proposed by researchers in the field of metabolomic, who identified acylcarnitine species as metabolites that accumulate specifically in obesity- and NASH-related HCC tissues in mouse HCC models [120]. Specifically, Enooku et al. showed that long-chain acyl-carnitines AC14:1 and AC18:1 gradually increase with the progression of fibrosis and further increase in patients with HCC, whereas the middle-chain acylcarnitine AC5:0 exhibited the opposite trend [121].Additionally, the possible role of serum testing of micro-RNA (mi-RNA) is under investigation, as these proteins play a role in the pathogenesis and prognosis of NAFLD. In particular, among miRNAs, the liver-specific miR-122, miR-34a and miR-16 were recently found increased in the serum of NAFLD patients and their expression was associated with liver enzymes, inflammation and fibrosis [122,123,124]. Moreover, HCC patients also present elevated plasmatic levels of miR-106b-3p, miR-101-3p and miR-1246 when compared to healthy control subjects [124]. However, despite being promising, the results of these cross-sectional studies do not currently support the use of any additional biomarker for HCC surveillance in NAFLD patients.Current literature demonstrates that the risk of HCC among NAFLD patients is significantly higher than that of the general population and that in consideration of the high prevalence of the disease, NAFLD-related HCC will become a leading cause of morbidity, mortality and liver transplantation in the near future. The phenotype of HCCs emerging in the context of NAFLD seems to be distinct, as it can develop not only upon a cirrhotic liver but also at earlier stages of fibrosis. The reasons behind this are not completely understood, but probably the pathology of HCC in steatosis is unique and related to steatosis, insulin resistance and to a pro-inflammatory status supported by both diabetes and obesity and other conditions inherent to the MetS as well. However, cirrhosis remains the main risk factor and in patients with NAFLD-related cirrhosis surveillance for HCC by means of bi-annual US is recommended. For non-cirrhotic patients, the presence of advanced fibrosis seems to remain an important risk factor for the occurrence of HCC, as in patients with other etiologies of liver disease, but other co-factors such as diabetes, obesity, older age, and possibly a genetic predisposition, may further enhance the risk of HCC development. As a fact, elderly males with diabetes and obesity seem to represent the cohort of patients at highest risk of HCC among non-cirrhotic NAFLD patients, and Hispanics seem to be at an even higher risk. However, we still do not have a clear perception of how these factors may be used to stratify patients’ risk and to build a risk-score to help identify the sub-population(s) with non-cirrhotic NAFLD where surveillance for HCC may be cost-effective. The currently observed low (and mostly unknown) incidence of HCC in non-cirrhotic NAFLD patients does not justify systematic surveillance in this population but suggests that a stratification by fibrosis scores and additional clinical and biochemical markers of HCC risk are warranted (see above). Therefore, future clinical research should focus on defining clusters of non-cirrhotic NAFLD patients for which systematic surveillance would be beneficial. Optimally, study design would be prospective, international, multicenter, with large cohorts and strict definitions of steatosis and NASH. The formulation of dedicated scores and the investigation of the role of non-invasive markers of fibrosis would be extremely useful, as well as the definition, for the latter, of specific cut-offs aimed at identifying those patients at higher risk of HCC. Finally, novel non-invasive markers of HCC are needed in order to improve the detectability rate at a screening level since the current recommended imaging technique (US) for HCC screening and surveillance presents challenges in the NAFLD population and might therefore cause an under-detection of the tumor at early, curative stages.Search strategy and selection criteria: we identified references for this review through a search of PubMed with the terms “hepatocellular carcinoma”, “steatosis”, “non-alcoholic fatty liver disease”, and “steatohepatitis” from Jan 1, 2000, to Feb 29, 2020. Only papers published in English were reviewed.M.C.P.T.: literature research, drafting of the manuscript; G.B., M.F., E.M., P.Z., M.S. and E.G.G.: drafting of the manuscript, manuscript revision; M.S. and E.G.G.: critical revision of the manuscript. All authors have read and agreed to the published version of the manuscript.M.S. gratefully acknowledges the support of the Yale University Liver Center (DK034989 Silvio O. Conte Digestive Diseases Research Core Centers, Clinical Core).Edoardo G. Giannini wishes to acknowledge the help of the Yale University Liver Center and Clinical Translational Core. The lively discussion and interactions during my guest professorship at Yale inspired a great deal of this manuscript.M.C.P.T., G.B., E.M., M.F., P.Z. have no conflict of interest to declare. E.G.G. reports advisory boards for MSD and Eisai. M.S. reports advisory boards for Bayer and Engitix.Relative contributions for HCC incidence from the most frequent causes of liver disease. HCC = hepatocellular carcinoma; NASH = non-alcoholic steato-hepatitis; HBV = hepatitis B virus; HCV = hepatitis C virus.Main risk factors for hepatocellular carcinoma (HCC) among NAFLD patients.HCC incidence and main risk factors among the NAFLD population.NAFLD, non-alcoholic fatty liver disease; NASH, non-alcoholic steato-hepatitis; ALD, alcoholic liver disease; HCV, hepatitis C virus; HCC, hepatocellular carcinoma; NR, not reported; PY, person-years; GGT, gamma-glutamyl transpeptidase; ALT, alanine aminotransferase; AST, aspartate aminotransferase; MetS, metabolic syndrome; HR, hazard ratio; aHR, adjusted hazard ratio; CI, confidence interval.

Med-MDPI/cancers/cancers-12-06-01423.txt

@@ -0,0 +1 @@
+Lack of relevant preclinical models that reliably recapitulate the complexity and heterogeneity of human cancer has slowed down the development and approval of new anti-cancer therapies. Even though two-dimensional in vitro culture models remain widely used, they allow only partial cell-to-cell and cell-to-matrix interactions and therefore do not represent the complex nature of the tumor microenvironment. Therefore, better models reflecting intra-tumor heterogeneity need to be incorporated in the drug screening process to more reliably predict the efficacy of drug candidates. Classic methods of modelling colorectal carcinoma (CRC), while useful for many applications, carry numerous limitations. In this review, we address the recent advances in in vitro CRC model systems, ranging from conventional CRC patient-derived models, such as conditional reprogramming-based cell cultures, to more experimental and state-of-the-art models, such as cancer-on-chip platforms or liquid biopsy.Colorectal carcinoma (CRC) is the third most commonly diagnosed form of cancer in the world, with an estimated incidence of 1.8 million cases in 2018, and is expected to reach 2.2 million by 2030 [1,2]. Although cancer treatment in general has improved over the past few decades, the need for more personalized targeted therapies remains present, specifically for late-stage metastatic CRC (mCRC) patients for whom treatment options—and consequently overall survival rates—are limited [3].The attrition rate of anticancer drugs candidates is very high, and only approximately 5% of drugs successfully complete phase III clinical trials [4,5]. One of the problems that might impair the development and approval of new anticancer therapies is the lack of relevant models that recapitulate the complexity of human cancer nature.The main traits of an “ideal” CRC model for testing new treatment options reside in its capacity to resemble the in vivo conditions. This includes characteristics such as the genetic-, functional- and histological features of the patient’s tumor, along with sequential mutagenesis (i.e., loss of adenomatous polyposis coli, APC), followed by activating Kirsten rat sarcoma viral oncogene homolog (KRAS) mutations and loss of TP53, the presence of stromal- and immune cells, as well as the presence and composition of tumor stroma.In patient-derived xenograft (PDX) models, small pieces of surgical patient tumor tissue are used for implantation into an immunodeficient mouse. Detailed protocols of the engraftment and propagation procedure for CRC PDX were described by several groups [6,7,8]. Different studies have demonstrated the potential use of PDX as a preclinical model in the drug screening cascade, as it can reliably predict and recapitulate CRC patient drug responses in colorectal cancer [9,10,11,12,13,14,15]. Although the tumors grow in a biologically more relevant microenvironment than can be provided in vitro, the mice are immunocompromised, therefore some of the complex interactions between the host and the tumor might not be preserved. Furthermore, some genetic and epigenetic changes may occur in the tumor cells during manipulations, such as resection, culture or implantation. Among several factors, the quality of the patient specimen, tumor type and stage, technique and time to implantation may affect the engraftment rates [16,17,18,19,20,21].Multiple reports have shown that, after engraftment, the human tumor stroma is preserved, but is slowly replaced by murine stroma over time throughout the consequent passages [22]. In addition, in PDX models the microenvironment of the subcutaneous space differs greatly from that of the colon. The latter led scientists to develop orthotopic mouse models, where the tumor is directly implanted into the caecum. The main goal was to create an in vivo model that would allow tumor development locally in the colon, allowing all stages progression for CRC [23,24].Genetically engineered mice (e.g., germline APC mutant models [25] or models presenting mismatch repair-deficiency [26]) and carcinogen-induced models [27] are widely used to investigate CRC and its treatment screenings. These models remain the most developed CRC in vivo models, due to their genetic controllability. Several of them are elegantly reviewed by others [28,29] and they are not discussed in this review.Since the use of laboratory animals is laborious, time-consuming and expensive, in vitro models would greatly contribute to higher efficiency in drug screening. In addition, given that animal experimentation [30] is being widely discussed, the development of alternative in vitro models is needed to support the idea of reduction, refinement and replacement of laboratory animals. Even though drug discovery cannot be based purely on in vitro models, they can deliver important results that may, in turn, help in the reduction of further in vivo experiments.In this review we discuss a broad spectrum of in vitro CRC model systems, ranging from recent advances in conventional CRC patient-derived models, to more experimental and state-of-the-art technology models (Figure 1). We also suggest potentially attractive models used in other cancer types that would need further validation for CRC.Recently developed CRC patient-derived models are usually established from freshly resected tumor tissue that undergoes enzymatic and mechanical digestion. Patient-derived models have emerged from liquid biopsies i.e., blood containing circulating tumor cells. Current advances in tissue engineering allowed patient-derived cells to be incorporated into a bioink and bioprinted into 3D constructs.For decades, preclinical cancer research has been widely based on the use of cancer cell lines cultured in vitro and xenografts derived from these, grown in vivo. The culture of cells in vitro lead to the acquisition of multiple genetic and epigenetic alterations that diverge drastically from the original tumor they were derived from. Expanding and maintaining tissue-derived cell lines in culture often implicates the use of exogenous immortalization methods to keep cells in culture. These cells exhibit a similar gene expression and epigenetic profile, and can be propagated in vitro, into several germ layers, providing great potential for disease modelling such as cancer [31,32,33,34].Recently, conditional reprogramming became widely used as a preclinical model in cancer research [35]. It is a co-culture based technology that makes it possible to efficiently expand patient-derived cells in culture medium supplemented with Rho kinase inhibitor (ROCK inhibitor, Y27632) and irradiated feeder fibroblasts [36]. Y27632 has been shown to induce the proliferative capacity of primary tumor cells resulting in efficient immortalization of cells into stratified epithelium without any DNA damage [37,38]. Several groups have elaborated specific protocols for cell isolation from various tumor types, including hepatocellular carcinoma [39], prostate cancer [40], tongue squamous cell carcinoma [41] and non-small lung cancers [42]. Liu et al. described a detailed protocol of CRC patient-derived cell cultures establishment from both cancerous and non-cancerous tissue biopsies that had the capacity to grow indefinitely in vitro, while maintaining phenotypic and genotypic features of the original tissue [36]. The study included freshly resected CRC tumors that generated approximately 10,000 cells after four weeks being in culture. This was done using conditional reprogramming, i.e., a novel next generation tool for long-term culture of primary epithelium cells derived from almost all origins without alteration of genetic background of primary cells. Moreover, Kodack et al. reported on a platform for functional testing on tumor-biopsy-derived cultures [43]. The criteria of a successful generation of colon cancer cell lines were defined as the point where the cells no longer required feeder cells to grow, could be cryopreserved and thawed and regrown, while maintaining the genotypic and phenotypic features. The authors also elaborated an immunofluorescence-based assay using a cocktail of monoclonal antibodies targeting cytokeratin (CK) 8 and cytokeratin 18 to specifically identify cancer cells, since both CK8/CK18 are nearly present in all tumors of epithelial origin.The induction of conditional reprogramming in cancer cells is fast, and, unlike in the case of conventional cell lines, results in the generation of whole cell populations without clone selection. In addition, this technology makes it possible to maintain the phenotypic features of the primary tumor in culture. Future studies are still needed to confirm the genetic diversity within the isolated tumor cells.In general, 2D cell cultures lack in vivo characteristics, such as tissue specific architecture, which can affect the proliferation of the cells and their response to external stimuli. This lack of complex cellular interactions fails to replicate the aggressiveness and heterogeneity of the disease, making 2D cultures poor models to predict drug response for complex diseases such as cancer. Two-dimensional models are used in a relatively high-throughput in vitro screening, but are constrained by the limited viability and the low and/or short proliferative capacity of the cultivated cells [20]. Moreover, the result highly depends on the isolation and culturing conditions, e.g., composition of the cell culture medium, seeding density and the addition of supplements or cellular matrixes [44,45,46,47].Thus, traditional cell lines cultured in monolayers are not perfectly suited for complex CRC research and its further development led to creation of three-dimensional cell culture systems. To better recapitulate the organ and tumor complexity, researchers expanded technology of cell cultivation using the spheroid and organoid technology [46].Spheroids are spherical cellular constructs, consisting of an external proliferating zone surrounding an internal quiescent zone [48]. The co-existence of these multilayers makes it possible to mimic the cellular heterogeneity observed in solid tumors [49,50]. We have recently developed a robust short-term 3D spheroid model, where human CRC cells were simultaneously co-cultured with human fibroblasts and human endothelial cells in a clinical relevant ratio [51].Jeppesen et al. established short-term spheroids cultures obtained with a high success rate of over 80% from freshly-derived primary CRC tumors [52]. They show that the initial tumor fragment size does not affect the success rate of spheroid formation or the cellular characteristics of the spheroid, which preserve both the molecular and histological characteristics of CRC, while maintaining inter-patient sensitivity towards a given treatment.The cell culture media composition has a major impact on the success rate of maintaining high viability of tumor-derived spheroids in culture. Available protocols remain inconsistent, as the report on various combinations of cell medium supplements and their positive effect on cell viability [47,53,54,55,56].To date, the 3D CRC spheroids rarely contain immune cells [57,58]. Integrating a potential immune response in the patient-derived spheroid to a treatment might represent an important factor in the treatment design. Recently, CRC patient-derived spheroids were co-cultured with tumor infiltrating lymphocytes from the same patient [59] to study the infiltration, activation and function of immune cells in tumors. This study proved that both activated natural killer cells and activated T cells infiltrated the spheroids induced the death of cancer cells and disrupted the 3D spheroid structure. Heterotypic co-cultures of tumor spheroids with other immune cells types could further expand our knowledge of human anti-tumor immune responses.Each of the above-mentioned models has its own advantages and drawbacks in terms of replicating the in vivo physiology of original tumor architecture, TME, cellular composition, as well as response to different exogenous stimuli. This is very often highly dependent on the initial patient specimen. These models are being constantly improved, to better recapitulate complex cancer biology. Further research regarding the automation, miniaturization and adaptation of spheroid co-culture models to human tumor types will make it possible to dynamically study the anti-tumor immune response. Several critical aspects of cell isolation and culture conditions need to be carefully considered when handling patient-derived in vitro material. The type of dissociation method (mechanical vs. enzymatic) used might influence the number and quality of isolated cells. In addition, the choice of an adequate antibiotic or a mixture of thereof, as well as the concentration of this in the culture medium, is critical to avoiding microbial contamination during transport and culture of the CRC cells [60]. Furthermore, evaluation of the purity and tumoral nature of the isolated cells, e.g., by flow cytometric analysis or immunohistochemistry, is essential for preserving a representative ratio of the different cell populations in the tumor of origin [60,61]. Patient-derived intestinal crypts (see Section 3.4) require the presence of Matrigel, and a defined mixture of the Lgr-5-ligand R-spondin, epidermal growth factors, and Noggin, that are known to be the most essential stem cell maintenance factors [62].Three-dimensional culture systems have been developed to mimic in vivo tumors as closely as possible by considering two aspects of cancer: heterogeneity and stromal interactions. It is important to note that most of the models take only partially into account tumor complexity, and most of the components of the stroma are absent [63]. Patient-derived tissue slice culture model is a tumor slice of approximately 200–300 µm thick, which is enough to preserve histological features of the original tumor, as well as important cellular components such as immune, vascular and mesenchymal cells [64]. The latter are important key features of this culture model [65]. Patient-derived tissue slice culture models have already been established for various cancer types, such as prostate [66,67], breast [68] and lung [64]. Until now, only a few reports included this model to study CRC.Sönnichsen et al. showed that slice culture from patient-derived colorectal tumor tissue represented similar morphological features to the original tumor over the observed cultivation period of 3 days. Persisting tumor cell proliferation in tissue slice culture under treatment with 5-FU, as highlighted in the study, can help identify a non-responding patients to a treatment, and therefore may help in preventing administration of ineffective treatment in clinically applicable timeframe [69]. Unlike 3D culture models, the initial step for this technique does not include an enzymatic digestion step of the tumor-tissue before the cells are stimulated to grow in 3D, which, in turn, makes it possible to maintain the complexity of the tumor without extra manipulation of the tissue [63]. Ironically, a key advantage of this model can also be a limiting factor, as the normalization of tumor cell fractions is a major inconvenience. The exact number of tumor cells in the tissue slice culture is not determined prior to the initiation step, which can greatly impact the reproducibility of the results [65]. While this model represents a promising technology to assess drug sensitivity in individual colorectal tumors, further correlations with clinical outcomes in larger cohorts of patients to validate the clinical application of the technique, are to be considered [69].Tumor tissue slices of hepatic CRC metastases were used for the first time to evaluate the response to oxaliplatin, cetuximab and pembrolizumab and measure anti-proliferative and pro-apoptotic features of the tumor and morphometric changes [70]. Moreover, the RAS mutation status, as well as the immunohistochemical evaluation of microsatellite stability and checkpoint protein (PD1) expression, was assessed. This study identified non-responders and responding patients. Moreover, the original tumor sections showed moderate to high infiltrates of PD1 positive tumor-associated immune cells, indicating susceptibility to selected treatments.One of the obstacles of the tissue-slice technique is the lack of long-term tissue preservation methods. In order to address this issue, Zhang et al. developed a new method of vitrification-based cryopreservation of tumor biopsies [71]. The patient-derived xenograft models were then successfully established. The observed drug responses in the xenograft model were consistent with those in tissue slice cultures performed in vitro. Importantly, the cultivation retained the heterogeneous architecture of the original tumor giving opportunity to further analysis of tumor biology.Liquid biopsy refers to the analysis of biomarkers in any body fluid [72]. In oncology, liquid biopsy represents a non-invasive test using blood to analyze tumor-derived genetic materials (DNA, RNA and miRNA) and proteins that either can be circulating freely in the blood or incorporated in circulating tumor cells (CTCs) [73]. CTCs play an important key role in the understanding of the biology of metastasis in patients with cancer, as recent studies have shown that they are found in the blood of cancer patients, as single CTC or CTC clusters with a strong ability to seed metastasis [74]. CTCs are new potential biomarkers that have been recently employed as diagnostic, prognostic and predictive to a wide range of cancer type including breast, lung, prostate and colorectal cancers [75,76]. The detection and study of CTCs in peripheral blood have caught the attention of scientists over the past decade, mainly for their promising clinical implication. A recent study showed that the disruption of CTC clusters, which have been linked to high metastatic potential [77], increases the proportions of single CTCs in the blood stream, but suppresses overall metastasis formation, highlighting the importance of CTC clusters as potential therapeutic targets in cancer treatment [74].The CellSearch® platform, approved by the Food and Drug Administration (FDA), is currently used to identify, isolate and enumerate the CTC in the blood samples [78,79]. This technique consists mainly of the antibodies attached to magnetic beads against epithelial cell-adhesion molecule that are present on the surface of the CTC, and not on the healthy blood cells, separating magnetically the CTC from the bulk of other cells in the blood sample [80]. The low number of CTCs in peripheral blood makes it very challenging to establish their in vitro cultures. Recently, however, several groups managed to obtain CTC cell lines from patients with prostate [81] and breast cancer [82], two tumor types known to have a higher number of CTCs.Cayrefourcq et al. reported for the first time the establishment of a permanent cell line from approximately 300 CTCs of one CRC patient, using the CellSearch® platform [83]. This cell line has been cultured for more than one year. Thorough analysis of the cells at the genomic, transcriptomic, proteomic and secretomic levels showed high similarity to the tumor of the patient with colon cancer that they were derived from. This approach opened a new avenue for a potential platform for novel drug screening in CRC, by eventually generating single CTC or CTC spheres from patient-derived blood samples.Another protocol that can be used to establish a CTC colorectal cancer patient-derived cell line was described by Grillet et al. The authors generated sufficient cellular material (5 million cells) within three weeks after sample collection, and then used it to perform cytotoxicity assays. The study offered preliminary clinical data suggesting that toxicity assays on CTC might predict patient response to drugs. A patient, from whom a CTC line was established, died after being treated with FOLFIRI (FOL = Folinic acid (Leucovorin) + F = 5-fluorouracil + IRI = irinotecan), a first line treatment in patients diagnosed with mCRC [84]. The CTC line was shown in this study to be resistant to this chemotherapy combination in vitro [85].A potential future application of CTC in personalized medicine would be to develop CTC-derived organoids and CTC-derived xenografts that could be used in drug screening for CRC treatment, using minimally invasive methods, while reflecting to a high extent tumor heterogeneity ex vivo.A major inconvenience of the use of liquid biopsy and CTC is the low concentration and yield of CTC extraction, especially in the blood of patients with CRC [83]. Moreover, major discrepancies have been highlighted, depending on the technique used for CTC detection, i.e., EpCAM antigen detection-based (CellSearch®) or cell size-based (ISET assay, based on the use of specific designed filtration membranes that allow size based exclusion of blood components, in different tumor types) [86]. The limited number of FDA approved technologies available for CTC detection and extraction makes the technology less accessible. Its wide application in clinical practice is also limited by its high costs. Lately, the development of microfluidic technology (see Section 3.3) is considered a potential alternative solution for CTC isolation [87]. Nevertheless, liquid biopsy holds great promise for revolutionizing cancer diagnostics in the future, to enable early detection.The development of organ-on-chip (OOC) technology has made it possible to bridge the gap between conventional cell cultures, preclinical animal models and clinical trials in patients. In addition to recapitulating to a high extent the biology and physiology of the organ of origin, the OOC allow high-resolution-real-time imaging of living human cells in a functional human tissue and organ context [88].OOC are microfluidic culture devices consisting essentially of flexible polymers, such as polydimethylsiloxane, containing perfused micro-channels harboring living cells that mimic in vivo organ architecture and physiology. The viability of cells can be maintained over an extended period (weeks to months) by flowing medium through the micro-channels. When the system is stabilized, medium can be replaced by whole blood perfusion for a couple of hours [89].Multiple research groups managed to establish OOC platforms by replacing healthy cells and associated extracellular matrixes with those of cancers [90]. The concept behind the technology is to model cancer cell behavior within human-relevant tissue and organ microenvironments in vitro. OOC enable researchers to vary local cellular, molecular, chemical and biophysical parameters in a controlled manner, both individually and in precise combinations, while analyzing how they contribute to human cancer formation, progression and response to therapy. OOC have been developed for a wide range of solid tumors like brain [91], bladder [92], breast [93] and non-small lung cancer [94]. Not only has this technology been used to create organs and solid cancer-on-chip, but some research groups, like Zhou et al., have also employed it to isolate CTCs in cancer patients [87], or to model bone marrow angiogenesis [95]. Traditional static intestine models containing human epithelial cells (e.g., Cako-2 or HT-29 cell lines) cultured on extra-cellular matrix-coated membranes within the trans-well devices, could not support several intestine properties or its functions. Over the last few years, the intestine chips have been developed with increased complexity that include channels lined to human microvascular endothelium, immune cells or pathogenic bacteria, and allow interaction between them [96]. That, in turn, enabled studying physiology, as well as pathology of the intestine. A good example of such a device is the microfluidic two-channel gut chip, which contains human epithelial, endothelial, immune and microbial cells, co-existing on ECM-coated transparent silicon polymer [97]. In this model, the pneumatic application of suctions applied in cycles enabled device deformations, resembling the movements of intestine during the peristalsis. Importantly, under these conditions, epithelial cells that, in conventional 2D conditions, grow in monolayers, spontaneously undergo villus morphogenesis. This is an important improvement, as compared to organoids that do not experience physical stretching resulting from peristaltic contractions. Those villi are lined in a columnar manner similar to that in a real intestine [97]. Whether the human gut chip might be a potentially important application in CRC treatment remains to be demonstrated, but it is highly possible given its successful use in other complex disorders [96].Along with an understanding of colonic epithelial cell behavior in the presence of microfabrication substrates, improved crypt isolation and 3D culture was an important step in the development of ‘organ-on-chip’ approaches for studies using primary colonic epithelium. Ahmad et al. reported on a protocol to standardize the isolation of intact murine colonic crypts by optimizing matrix concentrations on different surfaces i.e., PDMSs. The authors presented a reproducible low-cost crypt culture protocol, which may pave the way for further intestinal studies on patient-derived material using “organ-on-chip” platforms [98].Concerning tumor-on-chip platforms, only a few studies are available, leaving great possibilities for further development. Carvalho et al. employed OOC technology to recapitulate the human colorectal tumor microenvironment, and assess the efficacy of gemcitabine loaded nanoparticles for the treatment of CRC using a microfluidic gradient [99]. In this device, the human CRC-like core, containing HCT-116 and HCoMECS cells, is surrounded by a vascularized microtissue, and serves as a tool to study radial drug penetration and efficacy of the microvascular network into a cancer-mimicking tissue. Although the oxygen gradient is not present in this device, its potential application is vast in CRC, or in solid tumors in general. This 3D microfluidic cell culture seems to be an extremely useful tool in the study of various phenomena, such as vascularization and oncogenesis under dynamic conditions. In the development of CRC, or its dissemination during the metastasis, the organ-on-chip-like microfluidic device has been developed [100]. This device, which merges microfluidics and photoconvertible protein technology, enables tracing the velocity of the circulating cells in the selectin-regulated process of adhesion and metastasis in a spatiotemporal manner.Effectively, the OOC can be considered as a reliable platform to evaluate drug toxicity, given their high capacity to mimic in vivo like structures and functions. However, with these models, only the tissue or organ responses are considered without taking into account multi-organ interactions, which is crucial to evaluating the pharmacodynamics and the pharmacokinetics parameters of a drug [101]. To overcome this challenge, so called multiple-organs-on-chip (body-on-chip) were developed [102]. The latter technology recapitulates numerous organ interactions on a limited surface, while maintaining the highest degree of similarity to the in vivo situation. These systems usually do not require the use of pumps, using gravity to drive fluid flow to better replicate the physiologic flow. Oleaga et al. developed a system consisting of four different 2D tissue cultures (i.e., liver, cardiac, skeletal muscle and neuronal components), which were integrated within a single device to evaluate the toxicity of doxorubicin, valproic acid, acetaminophen and atorvastatin [103]. This phenotypic culture model exhibited a multi-organ toxicity response, representing the next generation of in vitro systems.Esch et al. integrated liver and gastro-intestinal tract tissues within their device to evaluate intestinal barrier functions and metabolic rates of orally administered drugs and nutrients [104]. Fluidic flow through the organ chips was maintained via gravity and controlled passively via hydraulic resistances of the microfluidic channel network.Another improvement of OOC was reported by Kassendra et al. on a generation of a “hybrid model” of OOC that integrated intestinal organoids, resulting in a higher similarity to the in vivo situation. They were able to recapitulate “normal” intestinal functions by integrating fluid flow and peristalsis-like motions, along with immune cells to a vascular compartment, which are all key factors to the normal intestinal physiology. In these culture conditions, biopsy-derived epithelial cells used were differentiated into villus-like epithelium (thin brush border of the colon epithelium). The primary human intestine chip model can be adapted for a wide range of applications, particularly in personalized medicine, by establishing a platform that could help investigate patient-specific disease mechanisms, as well as novel drug screening and anti-cancer therapy response [105].Another recent development in human 3D in vitro technologies are the constructs derived from self-organizing stem cells that mimic the architecture, functionality and genetic feature of their corresponding organ [106]. The introduction of human patient-derived organoids (PDO) has enabled disease modelling, highlighting their great potential in biomedical applications, translational medicine and personalized therapy [107,108,109].Moreover, PDO established from metastases taken by sequential biopsies at multiple time points, and multiple regions of heavily pre-treated CRC patients were used as pre-clinical models in co-clinical trials [110,111,112]. Those organoids were exposed to anti-cancer drugs, and the results were compared to patients’ responses in clinical trials. The findings revealed the capacity of PDO to mimic in vivo tumor organization, at histopathological, molecular and functional levels, and to predict patient’s treatment response [111].Organoids can also be used to analyze mechanisms of drug resistance. Cancer stem cells expressing specific surface markers, such as CD44 and LGR5, known to be strongly associated with therapeutic drug resistance were co-cultured with epithelial and stromal cells to simulate the in vivo TME, with the use of an air-liquid interface (ALI) method [112,113]. ALI does not require exogenous growth factor supplementation and enables multilineage differentiation and sustained growth for over 60 days [112,114]. ALI patient-derived organoids presented higher resistance than CRC cell lines exposed to 5-fluorouracil and irinotecan (standard-of-care treatment of advanced CRC) [115].Despite their advantages, PDO possess also shortcomings. Their self-organizing structure leads to different phenotypes between organoids, and might induce high background noise during drug-screening. The use of scaffolds and other bioengineering methods could help standardize cancer stem cell development into a specific and robust organoid morphology [116]. Lab-grown organoids showed some abnormalities, such as lack of cellular diversity and altered gene expression patterns [117]. Another important drawback of organoid development is time, as it takes several weeks to form a relevant organoid [111,118]. Moreover, the lack of some epithelial components, tumor stroma and microbiome remain major limitations of the PDO model [119]. Only very recent studies reported the incorporation of immune cells derived from CRC patient biopsies. The infiltration of immune cells was found to correlate with tumor growth and drug response. The ALI organoids could therefore be a promising approach to better understanding the tumor immune microenvironment and its impact on treatment response [120].The PDO is an interesting in vitro model for preclinical drug development, due to its ability to mimic human physiopathology. As this still needs further technical and cost-effective improvements, it is unlikely that PDOs will fully replace existing drug development models.One other way to leverage tissue to predict drug sensitivity is to interrogate the tissue directly. For certain tumors, this approach has been a routine part of pathological assessment of patient samples. Breast cancers that express the receptor for estrogens and for progesterone are therefore rapidly and cheaply detected by immunohistochemistry, and can be treated effectively with one of the range of anti-estrogens available [121]. Similarly, B-cell lymphomas that express CD20 have had their prognosis transformed by the introduction of anti-CD20 therapeutic antibodies [122].A recent extension of this approach has been an FDA-approved tumor treatment based on the tumor molecular characteristics, and not on the tissue origin or pathohistological type. This was based on the observation that patients, whose tumors have lost their mismatch repair machinery respond better to immunotherapy than patients with tumors, in which the machinery is intact [123]. The identification of this phenotype is routinely detected by immunohistochemistry. Other approaches, such as the evaluation of the immune response, tumor mutation burden and expression of programmed death-ligand (PD-L1) are also aimed at identifying patients whose tumors are likely to respond to immunotherapy [124].With the democratization of molecular analysis of tumors many other anomalies have been and are being identified that can be targeted by specific therapies. The most established are the EGFR mutations in lung cancer [125] and BRAF mutations in melanoma with loss of the homology directed repair mechanism through loss of BRCA1/2 or other associated genes being a more recent example [126]. However, at the moment, we are experiencing an explosion of new molecules that target different molecular abnormalities, a detailed review of which is beyond the scope of this review.It is expected that, in the near future, we will witness a further expansion of our ability to characterize the phenotype of tumors, probably in the domain of expression analysis and proteomics through tissue analysis by mass cytometry [127]. These will allow the characterization of the pathways activated in different tumors, allowing the development of pathway instead of mutation directed therapies [128]. However, the high cost of the mass spectrometry remains the major constraint.Moreover, Coppe et al. reported on a high-throughput kinase-activity mapping (HT-KAM) assay, which makes it possible to reveal the phosphor-catalytic signatures of tumors [129]. The HT-KAM is based on identifying catalytically hyper-active kinases in cell models or tissue, in order to highlight drug resistance and identify potential new drug targets. The authors synthesized a 228-peptide library that include 151 biological substrate protein regions phosphorylated by oncogenic kinases, and serve as combinatorial sensors of kinases phosphor-catalytic activity. Peptide phosphorylation signatures can be converted in kinase activity profiles, which will make it possible to identify the activity of druggable proto-oncogenic kinases in these models. This platform was tested to determine the new mechanism/targets of drug resistance in BRAFV600E CRC. In CRC cells (WiDr), which were exposed to treatment with a BRAF inhibitor vemurafenib [130], downregulation of the phospho-proteins MEK1/2 and ERK1/2 and upregulation in phospho-EGFR were observed, which was in line with the previously reported literature findings. The authors further investigated the kinase signatures and identified additional targets such as an increased activation of AKT1, PDPK1 and PRKCA kinases. Synergy was observed when inhibitors of AKT1, PDPk1 and PRKCA were paired with BRAFV600E targeting agents. This example of the screening platform introduces a new versatile approach of target-based drug discovery, eventually to be implemented alongside other strategies to improve precision medicine.In the current unmet need of closer cellular and spatial complexity of a tumor in in vitro conditions, Boland et al. first deposed a patent for ink-jet printing of viable cells in 2006 [131]. During the last decade, tissue engineering has known significant advances with the emergence of 3D bioprinting. The latter showed potential to recreate tailored in vitro 3D heterocellular complex structures to replicate the heterogeneity of the native in vivo tissue [132]. High anatomic precise positioning of living cells embedded in a scaffold or scaffold-free based support, make it possible to obtain 3D structures [133]. A primordial component of the bioprinting procedure is the bioink. It consists of decellularized matrix, microcarriers, hydrogels and cells. Scaffold based approach consists in using bioink where cells are loaded in hydrogels (i.e., agarose, alginate, matrigel, etc.) that differ by their crosslinking properties and construct size they can create. Whereas, in scaffold-free models cell density is higher, cells self-organize and deposit an extracellular-matrix, which allows superior cell-cell interactions [134,135].To date, there are no reports available on the use of 3D bioprinting to mimic intestine models. This could be probably explained by the complex intestine functions, containing absorption and secretion functions. Currently, only pharmacokinetics and toxicity studies have been reported using such technology with the use of CRC cell lines (i.e., Caco-2) [136]. Madden et al. established a 3D in vitro model based on 3D bilayered bioprinitng of human primary intestinal epithelium for the evaluation of pharmacokinetic parameters, i.e., absorption, distribution, metabolism and elimination). In this study, human intestinal epithelial cells (hIEC) were cultured for 21 days with human intestinal myofibroblast and printed on cell culture inserts allowing easy passage of compounds between apical and basolateral surfaces. Tissue architecture obtained with the 3D bioprinted model was compared to monolayer culture of Caco-2 cells, the gold standard cell line model of the intestinal barrier [137]. Cell-specific markers were identified such as CK19 (epithelial) or vimentin (myofibroblasts) and allowed to distinct both compartments; protein involved in tight junction (E-cadherin) and brush border formation (villus). Immunohistochemical staining for mucin-2 confirmed mucous secretion, which indicates normal intestinal function. Furthermore, genomic analysis of this 3D intestinal tissue showed that main intestinal phase I Cytochrome p450, which is the main family enzyme implicated in the metabolism of drugs and xenobiotics [138], especially CYP3A4, and phase II metabolic enzymes, as well as efflux transporters, were expressed at similar levels compared to the native intestine. The same enzymes in Caco-2 monolayers were upregulated, downregulated or undetectable. To confirm these results, CYP450 metabolism in 3D conditions was further evaluated using an inductor of CYP3A4 rifampicin. Its activity significantly induced the metabolization of its substrate Midazolam in the 3D intestine [136]. Those findings create a new venue for 3D bioprinting as preclinical model for evaluation and prediction of drug efficacy in drug development.Langer et al. reported a study on patient-derived material integrated in the 3D bioprinting platform. Their approach was based on incorporating multiple cell types (fibroblasts, cancer cells including patient-derived cells or endothelial cells) into scaffold-free in vitro tissue of breast or pancreatic cancers. Various parameters including cell signaling, proliferation, and response to therapies were assessed. The 3D bioprinted model was used to create breast tumors using MCF-7 cell line. Ten-day-old tissues containing breast cancer cells were implanted into immunodeficient mice, and xenografts were treated using doxorubicin and targeted therapies (i.e., BEZ235, an mTOR inhibitor and sunitinib, a multi-target inhibitor). Sunitinib reduced significantly the vasculature density in the stromal compartment and increased collagen deposition.The authors further created the 3D bioprinted pancreatic tumor model containing a PDX cell line surrounded by endothelial cells and pancreatic stellate cells (PSCs). The tumor tissue was treated for 6 days with gemcitabine (standard of care chemotherapy for pancreatic cancer), and results showed a dose-dependent response of cancer cell death. Furthermore, a patient-derived pancreatic tumor tissue, after enzymatic digestion, into the bioink surrounded by endothelial cells and PSCs. Bioprinted patient-derived cells maintained their tumorigenic properties, as confirmed by proliferation marker Ki67 staining, and showed similar morphological properties when compared to matching in vivo PDX or primary patient tumor [139].Among the most common challenges faced by 3D bioprinting reside maintaining high cell viability and functionality, establishing constructs harboring in vivo like vascularization, obtaining resolution and reproducibility. Therefore, in order to improve cell viability during the bioprinting procedure, Colosi et al., employed a bioink that consists of a blend of alginate and gelatin methacroyl [140]. The authors optimized the formulation of a low viscosity bioink, which matches the physiological pH and osmolarity, and promotes cells adhesion as well as cellular migration, resulting in 80% of cell viability.Furthermore, scientists from Rice University have recently developed a new bioprinting model that allows to create highly complex vascular networks, which recapitulate the body’s natural passageways for blood, air, lymph and other vital fluids. Grigoryan et al. underlined the fact that their bioprint of a tissue was not only phenotypically similar to its healthy counterpart in the organism but also able to “breath” like the organism, through the oxygenation and flow of red blood cells through a complex distensible vascular network model [141].The clinical management of CRC has not majorly changed over the last two decades, as compared to other tumor types. The need for more representative preclinical models in the drug screening cascade is essential, as the attrition rate for anti-cancer drugs is very high especially for CRC. Most therapeutic agents developed mainly target VEGF (bevacizumab, aflibercept) or its receptors (regorafenib) or EGFR (cetuximab, panitumab). These discoveries have been made using cell lines and xenografts [142].Each of the presented platforms possesses its own strengths and drawbacks in terms of study design and expected outcome (see Figure 2). Patient-derived cell lines cultured in 2D monolayers are simple to manipulate, and usually allow for large high-throughput screenings. The lack of tumor microenvironment (TME) is improved in 3D organoids/spheroids, and they often retain the characteristics of the original tumor including tumor heterogeneity and complexity. In contrast to in vitro platforms, patient-derived xenograft models tumor microenvironment, but ethical limitations and host background represent main drawbacks of these models. Cancer-on-chip models have been recently developed as more physiologically relevant platforms [93,119,143,144].Interestingly, in the case of the BRAFV600E, single or double inhibition (BRAF inhibitor and/or MEK inhibitor) has been unsuccessful in CRC treatment [130]. It was later shown that insensitivity to the double inhibition was due to a feedback activation of EGFR [145]. Current standard of care for BRAFV600E-mutated CRC involve the triple inhibition of BRAF, MEK and EGFR [146]. This remains a major consideration that needs to be integrated through patient-derived models for drug discovery in CRC. This said, it is fair to assume that the potential treatment or combination of treatment options have been missed, due to the model selection. The determinants of such paucity are the choice of the model and lack of integration into the model of tumor heterogeneity.There are several expectations to be addressed from the clinician point of view regarding a patient-derived preclinical model in CRC. First, the treatment resulting from this process has to be more efficient than the current standard-of-care with similar or inferior toxicity. Second, when dealing with de novo CRC diagnosis or metastatic disease, two weeks is an acceptable turnaround time [147]. From a clinical point of view, from the diagnosis to the initial treatment, time should be as short as possible. However, preoperative workup and pre-habilitation are often time-consuming, but remain mandatory.An adequate preoperative staging is important when considering neoadjuvant treatment. Microscopic tissue assessment of CRC by a pathologist aims at describing the complex composition and architecture of the tumor. The recent development of computer-aided approaches helped in advances also in this discipline. A machine learning-based approaches for automated analysis of digitized microscopic images of CRC samples with the goal to improve prognostic stratification of patients are already available [148]. For colon cancer, preoperative chemotherapy or even radio-chemotherapy showed interesting results with less surgical complications, but no impact on disease relapsing [149]. On the other hand, the assessment of the resected surgical specimen is the cornerstone before starting any adjuvant treatment. The importance to evaluate precisely the tumor, node, metastasis (TNM) stage is obvious, as it determines whether the patient should receive adjuvant treatment or not. Tissue availability is thus limited in localized disease, compared to the metastatic setting. A good collaboration between the pathologist and the researcher, dealing with the presented models, is fundamental, in order to maintain a high level of quality for the tumor staging. The part devoted for the research should not alter this quality.It is important to underline that we are entering a new era of data-driven medicine. This is what offers the next-generation DNA sequencing (NGS). NGS describes the high-throughput technology that allows the sequencing of the entire human genome within a single day [150]. NGS-based diagnostic assays have achieved clinical utility, on one hand, by being a solid platform for direct therapeutic decision-making [151]. Today, the NGS enables to cluster patients’ tumor cases, based on their genomic profile, in virtual cohorts, in a way to determine whether the cancer of a specific patient is similar to that of another patient on the globe who received treatment A or B that saved them, matching genomic alterations with curated databases of evidence-based associations. Such platforms are being used in treatment of glioblastomas, lung, colorectal or gastrointestinal tumors, and can also be applied on liquid biopsy samples to help better diagnose, treat and monitor cancer in a less invasive manner.Another important technological tool are the imaging techniques that allow 3D visualization of the patient-specific tumor phenotype with prognostic pre- and post-treatment relevance. New non-invasive imaging techniques, apart from structural evaluation, help in assessment of TME and certain hallmarks of cancer, as elegantly reviewed by García-Figueiras et al. [152].Last but not least, the CRC in vitro model development might be supported by computer-aided approaches that facilitate experimental testing per se by guiding the researchers in the choice of tested conditions. It is extremely important especially in the context of CRC, where combinatory treatment approaches are mostly applied. Testing all combination options with multiple drugs is not trivial and requires a high time- and cost-effort. We have used the learning algorithm or statistical methods to lead experimental testing of multidrug combination candidates [153,154,155,156]. The generated data we modelled, made it possible to generate regression models that, in turn, enabled the elimination of ineffective and/or antagonistic compounds from the initial drug pool and led to identification of the most effective synergistic multidrug mixtures [155,156]. This approach brings the possibility of rapid patient-specific treatment optimization.Despite the absence of an “ideal” preclinical model that would completely recapitulate the complexity of colorectal cancer with its stages and heterogeneity, as well as genomic signature, the choice of available models is wide-ranging. A careful decision on which model to employ should be taken, depending on a specific scientific question prior experimentation. A careful consideration of the advantages and shortcomings of each model, as presented above, should help in the most optimal model selection.It is particularly important to mention that fundamental researchers should readily discuss the model choice with their clinical partners, i.e., oncologists, surgeons and pathologists, in order to secure the optimal conditions from tissue resection till its use in selected models. Already available in vitro models might provide very valuable information on several treatment aspects that can be further verified in more complex in vivo conditions. Model improvement should involve tumor phenotyping and genotyping (e.g., consensus molecular subtypes classifications), as well as a better representation of the TME [157]. Through the combination of multiple imaging along with biological and clinical information, the computer-aided-based platforms process the data using statistical models and the result is an accurate prediction of tumor growth and evolution. With the information gathered, eventual in vitro model can be further optimized through the characterization of a patient-specific tumor peculiarity.Conceptualization, P.N.-S.; Methodology, G.M.R.; Writing, G.M.R., T.K., P.N.-S.; Writing—Review & Editing, G.M.R., T.K., E.D., T.M., F.R., N.B., L.R.-B., P.-Y.D., P.N.-S.; Visualization, G.M.R.; Supervision, P.N.-S.; Project Administration, P.N.-S.; Funding Acquisition, P.N.-S. All authors have read and agreed to the published version of the manuscript.We acknowledge the European Research Council (ERC-StG-680209 to P.N.-S.) and Foundation for the fight against cancer and for medico-biological research (to P.N.-S.) for funding.The authors declare no competing interests.Overview of colorectal carcinoma (CRC) patient-derived preclinical models.Advantages and weakness of CRC patient-derived models in preclinical studies.

Med-MDPI/cancers/cancers-12-06-01424.txt

@@ -0,0 +1 @@
+Long noncoding RNAs (lncRNAs) are defined as RNAs longer than 200 nucleotides that do not encode proteins. Recent studies have demonstrated that numerous lncRNAs are expressed in humans and play key roles in the development of various types of cancers. Intriguingly, some lncRNAs have been demonstrated to be involved in endocrine therapy resistance for breast cancer through their own mechanisms, suggesting that lncRNAs could be promising new biomarkers and therapeutic targets of breast cancer. Here, we summarize the functions and mechanisms of lncRNAs related to the endocrine therapy resistance of breast cancer.For women worldwide, breast cancer is the most common cancer, and one in eight to ten women will develop breast cancer during their lifetime [1,2]. Although the endocrine therapies that target sex hormone receptor signaling pathways are effective treatment for breast cancer, therapy resistance and cancer recurrence remain important issues, and novel therapeutic strategies are required. Recent transcriptome analyses have revealed that a large number of long noncoding RNAs (lncRNAs), which are RNAs that are longer than 200 nucleotides in length and do not encode proteins, are expressed in humans [3,4,5]. LncRNAs play key roles in various biological process and diseases, including cancers [6,7,8,9,10]. In breast cancer, some lncRNAs exert oncogenic or tumor-suppressive functions to control breast cancer pathophysiology, such as invasion and metastasis, and drug resistance; these findings are summarized in a recent review article [11]. In terms of endocrine therapy, selective estrogen receptor modulators (SERMs), selective estrogen receptor degraders or downregulators (SERDs), and aromatase inhibitors, are mainly used as drugs to suppress estrogen signaling. For an experimental model of aromatase inhibitor-resistant breast cancer, cells that can obtain the ability to grow under long-term estrogen deprivation (LTED) conditions are preferentially used. Here, we extensively focus on endocrine therapy resistance-associated lncRNAs in breast cancer by including these drugs and experimental models, and describe the recent findings on their functions and mechanisms.Breast cancer is the most commonly diagnosed cancer and the leading cause of cancer death in women worldwide [1,2]. According to the GLOBOCAN 2018 database of the International Agency for Research on Cancer, which estimates the incidence and mortality of several cancers, the number of new cases of breast cancer in 2018 was estimated at 2,088,849, and those of deaths due to breast cancer are estimated as 1,276,106 [2]. Breast cancer is classified into at least four subtypes (luminal A, luminal B, human epidermal growth factor receptor 2 (HER2)/erythroblastic oncogene B 2 (ErbB2)-enriched, and basal-like) based on gene expression patterns [1,12,13]. The luminal subtypes are sex hormone receptor-positive [estrogen receptor (ER) or progesterone receptor (PR)-positive] and HER2-negative, and the HER2-enriched subtype is HER2-positive, while the basal-like subtype is ER-, PR-, and HER2-negative. The majority of breast cancers belong to luminal subtypes and are primarily sensitive to estrogen and progesterone [14,15,16]. The receptors of these hormones, ER and PR, respectively, function as ligand-dependent transcription factors. After binding to their ligands, these hormone receptors dimerize and associate with DNA through their DNA-binding domains. These hormone receptors form complexes with other transcription factors and co-regulators, such as the steroid receptor coactivator (SRC)/p160 family proteins and CREB-binding protein (CBP)/p300, and control the transcription of their target genes [17,18,19]. As sex hormone signaling pathways are essential for breast cancer pathophysiology, therapies targeting the hormones and their receptors, or endocrine therapies, remain the standard treatment for breast cancer [20,21]. For instance, drugs that suppress estrogen signaling or estrogen production are used for endocrine therapies. To suppress estrogen-mediated ER activation, drugs such as SERMs and SERDs are used. Although both SERMs and SERDs compete with estrogen, their mechanisms for the regulation of ER signaling are different. SERMs affect the interaction between the ER and co-factors, leading to changes in ER-targeted gene expression. Thus, SERMs, such as tamoxifen and raloxifene, act as ER antagonists in breast cancer and are used for breast cancer therapy or prevention. In contrast, SERDs mediate the destabilization of the ER to abolish ER signaling [21]. In addition to these modulators of the ER, drugs that block estrogen synthesis, such as aromatase inhibitors and luteinizing hormone-releasing agonists, are used for breast cancer treatment [20]. Although endocrine therapies are initially successful, breast cancers eventually acquire resistance to these therapies [22,23]. Moreover, patients with basal-like or triple-negative breast cancer (TNBC) exhibit poor outcomes, because this subtype lacks the expression of ER, PR, and HER2, and its effective therapeutic targets remain unidentified. Furthermore, metastatic breast cancer is considered incurable with the therapies available currently [1,24]. Thus, novel therapeutic targets and biomarkers are urgently needed. Recent studies have shown that lncRNAs play important roles in the pathophysiology of various cancers, including breast cancer, suggesting the potential of lncRNAs in developing novel strategies of cancer treatment [9,10].Together with the advancement of technologies of cDNA cloning and RNA sequencing, ~70–90% of mammalian genomes are shown to be transcribed to produce huge numbers of noncoding RNAs (ncRNAs), while less than 3% of these genomes are translated to proteins, suggesting the importance of ncRNAs in biological processes [25,26,27]. ncRNAs are classified by their length, i.e., ncRNAs shorter than 200 nucleotides are classified as small ncRNAs, while longer ncRNAs are defined as lncRNAs. MicroRNAs (miRNAs) belong to the small ncRNA category and are involved in translational repression and mRNA destabilization in cooperation with various proteins, including argonaute (AGO) proteins [28]. As it has been shown that miRNAs play key roles in numerous biological processes and diseases, including various types of cancers, their clinical application has been studied [10,29]. Moreover, lncRNAs have been suggested to be essential for cell physiology. Previous studies have identified a large number of lncRNA genes in mammals. For example, the GENCODE project, which is part of the ENCODE project and aims to annotate all gene features in the mouse and human genomes, has identified 13,197 and 17,952 lncRNA genes in mice and humans, respectively [30]. Moreover, a previous transcriptome study reported 58,648 lncRNA genes in humans [5]. Although most lncRNAs remain to be studied, it has been gradually elucidated that some lncRNAs play important roles in multiple biological phenomena, such as cell differentiation and organogenesis and diseases [6,7,8]. The expression of lncRNAs tends to be highly cell type- and tissue-specific [3], implying that lncRNAs are good candidate biomarkers and therapeutic targets for diseases. Intriguingly, the expression of some lncRNAs is deregulated in cancers, and these lncRNAs exert oncogenic or tumor-suppressive functions via various mechanisms, such as regulating the transcription or translation of target genes and modulating signal transduction [9,10]. Furthermore, some lncRNAs are involved in breast cancer progression via controlling some processes of breast cancer pathophysiologies, such as invasion and metastasis, and drug resistance (reviewed in [11]). Thus, lncRNAs may be promising biomarkers and therapeutic targets of cancers, including breast cancer. As mentioned above, endocrine therapy resistance is one of the major therapeutic problems for breast cancer. Intriguingly, some lncRNAs regulate the endocrine therapy resistance of breast cancer, and may be key factors for the treatment of breast cancer with endocrine therapy resistance. In the following sections, we introduce the studies of lncRNAs using models of breast cancer cells that are sensitive or resistant to drugs used for endocrine therapy (e.g., tamoxifen as the SERM, ICI182,780 as the SERD, and anastrozole as the aromatase inhibitor), models of breast cancer cells under LTED conditions, and clinical specimens of breast tumors, and intensely describe the functions and mechanisms of lncRNAs in the endocrine therapy resistance of breast cancer, as revealed by these studies.LncRNAs can be classified into one or more of five categories: (A) sense lncRNAs that overlap the neighboring protein-coding gene on the same strand; (B) antisense lncRNAs that overlap the neighboring protein-coding gene on the opposite strand; (C) bidirectional lncRNAs, which are transcribed from the opposite strand within 1 kb from the nearest protein-coding gene; (D) intronic lncRNAs that are derived wholly from intronic regions of protein-coding genes; or (E) intergenic lncRNAs, or long intergenic noncoding RNAs (lincRNAs) that are transcribed from the genomic interval between two genes (Figure 1) [31]. For example, among the lncRNAs related to the endocrine therapy of breast cancer, HOX transcript antisense RNA (HOTAIR) is classified as an antisense lncRNA, while urothelial cancer associated 1 (UCA1) is a lincRNA. In addition, several endocrine therapy resistance-related lncRNAs belong to multiple categories. In the following section, we introduce lncRNAs involved in the endocrine therapy resistance of breast cancer based on these categories.The HOTAIR lncRNA enhances resistance to tamoxifen [32]. HOTAIR is a ~2.2 kb lncRNA, and its gene overlaps the homeobox C11 (HOXC11) gene on the opposite strand [33]. A previous study using an ER-positive breast cancer cell line, MCF7, showed that HOTAIR binds to the estrogen receptor α (ERα), and the overexpression of HOTAIR enhances ER signaling by upregulating ERα expression levels and promoting the chromatin binding of the ERα, even under hormone-starved conditions. These results suggest that HOTAIR activates ligand-independent ER signaling, which may contribute to tamoxifen resistance (Figure 2) [32]. Moreover, HOTAIR has been demonstrated to promote breast cancer progression via transcriptional regulation. HOTAIR binds to the polycomb repressive complex 2 (PRC2) with its 5′ side, and regulates the PRC2-mediated trimethylation of H3K27 in trans at the HOXD locus on chromosome 2, which leads to transcriptional repression of the HOXD locus [33]. Furthermore, HOTAIR induces selective re-targeting of PRC2 and trimethylated H3K27 genome-wide, thus promoting the invasion of breast cancer cells [34]. Consistently, high expression of HOTAIR is associated with a short duration of metastasis-free and overall survival in patients with breast cancer [34,35]. Moreover, the 3′ side of HOTAIR binds to corepressor for element-1-silencing transcription factor (CoREST)/repressor element-1 silencing transcription factor (REST) complexes, including lysine-specific demethylase 1 (LSD1), which mediates the demethylation of dimethylated H3K4 (H3K4me2). HOTAIR can bind simultaneously to PRC2 and the LSD1/CoREST/REST complexes, to coordinate the targeting of these complexes to hundreds of genes for coupled H3K27 methylation and H3K4 demethylation [36]. Furthermore, the HOTAIR–LSD1 complex is involved in transcriptional activation mediated by c-Myc. In addition, the hepatitis B X-interacting protein (HBXIP) binds directly to c-Myc on target genes and recruits LSD1 via interaction with HOTAIR, which enhances the transcription of c-Myc target genes, possibly through the LSD1-mediated demethylation of H3K4me2 [37]. HOTAIR, HBXIP, and LSD1 promote breast cancer proliferation, highlighting the function of HOTAIR as a critical effector of c-Myc in cooperation with HBXIP and LSD1 [37]. Thus, HOTAIR plays important roles in the epigenetic regulation of gene expression. However, a recent study has proposed that PRC2 is dispensable for the HOTAIR-mediated trimethylation of H3K27 and gene silencing [38], suggesting that further studies are necessary for elucidating the precise mechanisms by which HOTAIR epigenetically controls gene expression. In addition, HOTAIR acts as a competing endogenous RNA (ceRNA) that specifically sponges the target miRNAs and inhibits their activities. For example, HOTAIR is a ceRNA for miR-206 and increases the expression of a miR-206 target gene, the Bcl-w/Bcl-2 like protein 2 (BCL2L2) gene, thus promoting breast cancer proliferation [39]. Moreover, HOTAIR functions as a ceRNA for miR-20a-5p and upregulates a miR-20a-5p target, the high mobility group AT-hook 2 (HMGA2) gene, which enhances the proliferation, survival, migration, and invasion of breast cancer cells, and the growth of breast tumors [40]. Thus, HOTAIR controls breast cancer progression via multiple pathways of regulation of gene expression.Thymopoietin antisense transcript 1 (TMPO-AS1) is an lncRNA that was demonstrated recently to enhance tamoxifen resistance in breast cancer [41]. TMPO-AS1 is a ~3.2 kb lncRNA, and its gene overlaps the thymopoietin (TMPO) gene on the opposite strand. High expression of TMPO-AS1 is associated with short distant-metastasis-free and overall survival in patients with breast cancer [41]. In addition, the upregulation of TMPO-AS1 is observed in MCF7-derived, tamoxifen-resistant cells (OHTR) and MCF7-derived LTED cells, which is associated with poor relapse-free survival in patients with breast cancer treated with tamoxifen [41]. TMPO-AS1 is induced by estrogen in MCF7 cells and another ER-positive breast cancer cell line, T47D. The purification experiments of TMPO-AS1 from these cell lines by using its antisense oligonucleotide probes suggest that TMPO-AS1 binds to the 3′ untranslated region (UTR) of the estrogen receptor 1 (ESR1) mRNA, which encodes the ERα protein, through an RNA–RNA interaction. Moreover, this RNA–RNA interaction results in the stabilization of the ESR1 mRNA. Thus, TMPO-AS1 upregulates ESR1 expression and ER signaling pathways, contributing to cell proliferation and tamoxifen resistance (Figure 2) [41]. Therefore, TMPO-AS1 is a promising biomarker and therapeutic target for ER-positive breast cancer. In other cancers, it has been suggested that TMPO-AS1 interacts with RNAs other than the ESR1 mRNA and promotes disease progression [42,43,44]. For example, in cervical cancer, TMPO-AS1 functions as a ceRNA via the sponging of miR-577 and upregulates a miR-577 target, RAB14, to promote cell proliferation, survival, and migration [42]. In osteosarcoma, TMPO-AS1 increases the expression of WNT7B by sponging miR-199a-5p, which promotes cell proliferation and survival [43]. In addition, TMPO-AS1 stabilizes the TMPO mRNA to promote the proliferation, survival, migration, and invasion of non-small cell lung cancer (NSCLC) cells. Mechanistically, TMPO-AS1 may stabilize the TMPO mRNA through their interaction [44]. Thus, TMPO-AS1 exerts oncogenic effects in various cancers by forming RNA duplexes with some target RNAs.Conversely, the ADAM metallopeptidase with thrombospondin type 1 motif 9 (ADAMTS9) antisense RNA 2 (ADAMTS9-AS2) lncRNA decreases tamoxifen resistance [45]. ADAMTS9-AS2 is an antisense transcript of the tumor-suppressor ADAMTS9 gene. It has been suggested that low expression of ADAMTS9-AS2 is associated with poor prognosis in patients with several types of cancers, such as lung, colorectal, and gastric cancers, while high ADAMTS9-AS2 expression is associated with poor prognosis in patients with some cancers, such as bladder cancer and salivary adenoid cystic carcinoma [46,47,48,49,50]. ADAMTS9-AS2 is downregulated in tamoxifen-resistant cells derived from MCF7, and downregulation of ADAMTS9-AS2 is also observed in breast cancer tissues, especially in breast tumors with grade III–IV or a tumor size larger than 2 cm [45]. From the knockdown or overexpression experiments of ADAMTS9-AS2 in an MCF7-derived tamoxifen-resistant cell line, it is shown that ADAMTS9-AS2 acts as a ceRNA by sponging miR-130a-5p to promote the expression of a miR-130a-5p target gene, phosphatase and tensin homolog (PTEN), which is a well-known tumor-suppressor and enhances tamoxifen sensitivity (Figure 2) [45].UCA1 is a lincRNA that was originally identified as a transcript that is upregulated in bladder transitional cell carcinoma [51]. UCA1 is downregulated in breast cancer and promotes disease progression [52]. From UCA1 knockdown and overexpression experiments in breast cancer cells, such as MCF7, T47D, and tamoxifen-resistant cells derived from these cells, it has been suggested that UCA1 enhances tamoxifen resistance by activating the mammalian target of rapamycin (mTOR), Wnt/β-catenin, and PI3K/AKT signaling pathways (Figure 3A) [53,54,55]. Moreover, UCA1 is shown to interact with the enhancer of zeste homolog 2 (EZH2), which is a catalytic subunit of the PRC2, and suppress the expression of cell-cycle regulator p21, by promoting the trimethylation of H3K27 on the p21 promoter, thus contributing to tamoxifen resistance (Figure 3A) [55]. In addition, from a previous study using breast cancer cell lines, such as MCF7 and BT474, it is shown that UCA1 acts as a ceRNA by sponging miR-18a to upregulate a target of miR-18a, the hypoxia-inducible factor 1α (HIF1α). As HIF1α activates the transcription of UCA1, UCA1 and HIF1α form a feedback regulatory loop that strengthens tamoxifen resistance (Figure 3A) [56]. Intriguingly, it has been reported that UCA1 is secreted from an MCF7-derived, tamoxifen-resistant cell line, LCC2, by exosomes, and that exosome-mediated UCA1 transfer enhances the tamoxifen resistance of MCF7 cells [57].Another lincRNA, breast cancer anti-estrogen resistance 4 (BCAR4), was identified in a functional screening of genes regulating the tamoxifen resistance of an ER-positive breast cancer cell line, ZR-75-1 [58]. Further studies suggest that the BCAR4-mediated tamoxifen resistance of ZR-75-1 depends on HER2/ErbB2, ErbB3, and ErbB4, but not ERα, and that BCAR4 overexpression enhances the resistance of MCF7 to antiestrogen ICI182,780 in a HER2/ErbB2-, ErbB3-, and ErbB4-dependent manner. [59,60,61]. BCAR4 interacts with several proteins, such as glioma-associated oncogene family zinc finger 2 (GLI2), smad nuclear interacting protein 1 (SNIP1), and phosphatase 1 nuclear targeting subunit (PNUTS), and regulates C–C motif chemokine ligand 21 (CCL21)-stimulated noncanonical hedgehog signaling pathway [62]. Although this activity of BCAR4 contributes to breast cancer metastasis [62], whether this mechanism is involved in the resistance to tamoxifen and ICI182,780 remains unknown.Metastasis-associated lung adenocarcinoma transcript 1 (MALAT1) is a lincRNA that has been suggested to be involved in the tamoxifen resistance of breast cancer [63]. MALAT1 was initially reported as an lncRNA that is highly expressed in stage I NSCLC tumors that subsequently metastasize, and high expression of MALAT1 is associated with short overall survival in patients with NSCLC [64]. Moreover, dysregulation of MALAT1 expression has been indicated in various cancers, including breast cancer [65]. MALAT1 is an ~8.7 kb lincRNA, and its gene is located on human chromosome 11q13.1. The MALAT1 primary transcript contains a tRNA-like structure at the 3′ end [66]. RNase P and RNase Z, which are endonucleases that cleave the 5′ or 3′ side of a tRNA precursor [67,68], cleave both sides of this tRNA-like structure, resulting in the 3′-end maturation of MALAT1 [66]. The excised tRNA-like RNA (MALAT1-associated small cytoplasmic RNA (mascRNA) is added with CCA trinucleotides at the 3′ end and exported to the cytoplasm [66].Although the function of mascRNA is not well understood, it is suggested that mascRNA is an immune regulator in monocytes that is involved in innate immunity in cardiomyocytes [69]. Moreover, the 3′ end of mature MALAT1 contains two U-rich sequences and the associated A-rich sequences, and these sequences form a triple-helical structure that enhances the stability of MALAT1 [70,71]. MALAT1 is localized in the nucleus, especially in nuclear bodies, which are termed nuclear speckles or SC35 domain and are enriched for splicing factors [72]. Regarding tamoxifen resistance, high expression of MALAT1 is associated with a short recurrence-free survival in patients with ER-positive breast cancer treated with tamoxifen [63]. In addition, high MALAT1 expression is associated with poor recurrence-free survival in patients with ER-negative breast cancer, indicating the importance of the ER-independent functions of MALAT1 [63]. The roles of MALAT1 in breast cancer are complicated, because both oncogenic and tumor-suppressive roles of MALAT1 in breast cancer have been reported. For example, MALAT1 acts as a ceRNA for some miRNAs, such as miR-124, miR-1, miR-129-5p, miR-204, and miR-339-5p, thus promoting breast cancer progression [73,74,75,76,77]. In contrast, MALAT1 functions as a ceRNA for miR-20a to inhibit the growth and metastasis of breast cancer [78]. Moreover, MALAT1 regulates transcriptional and posttranscriptional events in ways other than sponging miRNAs. For instance, MALAT1 interacts with the promoter of the eukaryotic translation elongation factor 1 alpha 1 (EEF1A1) gene and upregulates EEF1A1 expression by enhancing the trimethylation of H3K4, which promotes the proliferation and invasion of breast cancer [79]. In addition, MALAT1 forms a complex with the serine/arginine-rich splicing factor 1 (SRSF1), the inhibitor of the DNA binding 4, HLH protein (ID4), and mutant p53, and regulates the alternative splicing of the vascular endothelial growth factor A (VEGFA) mRNA precursor (pre-mRNA), which increases the angiogenic potential of breast cancer cells [80]. Conversely, MALAT1 binds to an RNA-binding protein, Hu antigen R (HuR), and interacts with the CD133 gene to downregulate CD133, thus suppressing the epithelial-to-mesenchymal transition (EMT) and migration activity of breast cancer cells [81]. Although the mechanisms via which MALAT1 exerts both oncogenic and tumor-suppressive functions are not well understood, its functions may depend on context, such as cell type and environment. Furthermore, the dual roles of MALAT1 in cancer progression have been suggested by studies using Malat1 knockout (KO) mice [82,83]. Arun et al. reported that Malat1 KO suppresses the lung metastasis of mammary tumors generated in mouse mammary tumor virus (MMTV)-polyomavirus middle T antigen (PyMT) mice [82]. Inversely, Kim et al. later demonstrated that Malat1 KO enhances the dissemination and lung metastasis of mammary tumors in MMTV-PyMT mice, and that this phenotype can be rescued by the transgenic expression of Malat1 from the ROSA26 locus [83]. Although it is not clear why there is a discrepancy between those results, it may be partly attributed to differences in the methodology for Malat1 KO. In the former study, Malat1 KO was accomplished by deletion of a ~3 kb genomic region containing the 5′ end of the Malat1 gene and its promoter using Cre-mediated recombination technology, while in the latter study, Malat1 was depleted by inserting the lacZ gene and polyadenylation sequences 69 bp downstream of the transcriptional start site of Malat1. These genomic rearrangements in Malat1 KO mice might affect the chromosomal conformation and some nuclear events of gene expression differently, resulting in differential phenotypes. Based on their findings, the manner in which the expression of lncRNAs is suppressed may be important for elucidating lncRNA functions.The large intergenic noncoding RNA-regulator of reprogramming (lincRNA-ROR) also upregulates tamoxifen resistance [84,85,86]. LincRNA-ROR was originally identified as an lncRNA that is upregulated in induced pluripotent stem cells (iPSCs) compared with embryonic stem cells (ESCs), and has been shown to modulate reprogramming [87]. LincRNA-ROR promotes the proliferation and invasion of MCF7 and a TNBC cell line, MDA-MB-231, by regulating the TGF-β signaling pathway, and high expression of lincRNA-ROR is associated with a poor prognosis in patients with breast cancer [88]. Regarding the mechanisms by which lincRNA-ROR regulates tamoxifen resistance, lincRNA-ROR knockdown experiments in BT474 suggest that lincRNA-ROR enhances tamoxifen resistance by inhibiting autophagy (Figure 3B) [85]. In addition, a previous study using lincRNA-ROR-KO MCF7 cells suggests that lincRNA-ROR promotes the degradation of an extracellular signal-regulated kinase (ERK)-specific phosphatase—the dual specificity phosphatase 7 (DUSP7)—under estrogen-deprived conditions, which results in the ligand-independent activation of ERα mediated by the mitogen-activated protein kinase (MAPK)/ERK signaling pathway. As a result, lincRNA-ROR promotes estrogen-independent growth and tamoxifen resistance in breast cancer (Figure 3B) [86]. Moreover, from a previous study using an MCF7-derived tamoxifen-resistant cell line, lincRNA-ROR is suggested to act as a ceRNA by sponging miR-205-5p to upregulate the miR-205-5p target zinc finger E-box binding homeobox 1/2 (ZEB1/2), thus enhancing EMT and tamoxifen resistance (Figure 3B) [84]. In MDA-MB-231, lincRNA-ROR acts as a ceRNA for another miRNA, miR-145, to upregulate its targets, i.e., ADP ribosylation factor 6 (ARF6) and mucin 1, which control the subcellular localization of E-cadherin and the metastasis of TNBC [89,90]. In addition to these findings, the genotypes of the rs4801078 SNP in lincRNA-ROR are associated with the risk of breast cancer [91], suggesting that lincRNA-ROR is both a promising biomarker and a therapeutic target of breast cancer.Furthermore, the lincRNA termed lncRNA in non-homologous end joining (NHEJ) pathway 1 (LINP1) enhances tamoxifen resistance [92]. LINP1 was initially identified as an lncRNA that is highly expressed in TNBC. In TNBC, LINP1 forms a complex with Ku autoantigen, 80kDa (Ku80) and the DNA-dependent protein kinase catalytic subunit (DNA-PKcs) and activates the NHEJ pathway, which repairs double-stranded breaks in DNA [93]. Consistent with these findings, LINP1 enhances the resistance to radiation and chemotherapeutic drugs that cause DNA damage, such as 5-fluorouracil and doxorubicin [93,94]. LINP1 also promotes the proliferation of ER-positive MCF7 breast cancer cells [94]. In addition, LINP1 expression is negatively regulated by estrogen, and is upregulated in ER-positive breast cancer cell lines, MCF7 and T47D, under estrogen-deprived or tamoxifen-treated conditions, as well as in tamoxifen-resistant breast cancer cells derived from these cell lines [92]. From the knockdown and overexpression experiments of LINP1 in MCF7 and T47D, it is suggested that LINP1 inhibits the ER signaling pathway by downregulating ERα, which may be involved in tamoxifen resistance [92].Recently, it has been reported that the cytoskeleton regulator (CYTOR)/LINC00152 lincRNA is involved in tamoxifen resistance [95]. CYTOR promotes the proliferation and migration of breast cancer cells, and high expression of CYTOR is associated with relapse-free survival in patients with breast cancer. CYTOR regulates the epidermal growth factor receptor and mTOR signaling pathways and control the organization of the filamentous actin cytoskeleton [96]. Moreover, CYTOR is upregulated in tamoxifen-resistant breast cancer cell lines derived from MCF7, and CYTOR functions as a ceRNA by sponging miR-125a-5p, and upregulates a target of miR-125a-5p, the serum response factor (SRF), which promotes the tamoxifen resistance and cell proliferation [95]. Consistent with these data, CYTOR expression is higher in breast tumors from tamoxifen-resistant patients [95]. In addition, CYTOR is associated with poor prognosis in patients with TNBC, and induces the ubiquitination-mediated degradation of PTEN in TNBC [97].Although there few studies have addressed lncRNAs that regulate the resistance to aromatase inhibitors, there is some evidence of this phenomenon. The lncRNA MIR2052HG lincRNA enhances the resistance to aromatase inhibitors [98,99]. LncRNA MIR2052HG is a ~2 kb lncRNA, and its gene is located on human chromosome 8q21.11–q21.13. This lncRNA upregulates the expression of ERα. From the knockdown experiments of MIR2052HG in ER-positive CAMA-1 cells, MCF7 cells stably transfected with the cytochrome P450 family 19 subfamily A member 1 (CYP19A1) gene that is an aromatase inhibitor target, and MCF7-derived cells resistant to an aromatase inhibitor anastrozole, it has been shown that MIR2052HG increases the expression level of the lemur tyrosine kinase 3 (LMTK3), which in turn regulates the expression of both ESR1 mRNA and ERα protein, and contributes to the resistance to anastrozole (Figure 3C) [99,100]. For regulating the ESR1 mRNA, LMTK3 decreases the activity of protein kinase C (PKC), which suppresses Ser 473 phosphorylation and the activation of AKT mediated by PKC. As AKT phosphorylates and induces the proteasome-mediated degradation of forkhead box O3 (FOXO3), a transcription factor that controls ESR1 expression, the MIR2052HG/LMTK3/PKC/AKT axis stabilizes FOXO3, thus upregulating ESR1 transcription [99,100]. Conversely, LMKT3 suppresses the activity of the mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (ERK) kinase (MEK)/ERK/ribosomal S6 kinase 1 (RSK1) signaling pathway through the downregulation of PKC activity, which results in a decrease in the Ser 167 phosphorylation level of ERα as well as its stabilization (Figure 3C) [99,100]. In addition, single-nucleotide polymorphisms (SNPs) located near or within the MIR2052HG gene locus are associated with the recurrence of breast cancer in patients treated with aromatase inhibitors as adjuvant therapy, suggesting that these SNPs in MIR2052HG are promising biomarkers that can be used to identify patients in whom aromatase inhibitors would be an appropriate therapy [98].Moreover, a recent study indicated that the lncRNA LINC00309 is associated with poor disease-free survival in patients with breast cancer treated with endocrine therapy using aromatase inhibitors, which suggests that LINC00309 plays important roles in the acquisition of resistance to these therapeutic agents [101].The growth-arrest specific 5 (GAS5) lncRNA is downregulated in tamoxifen-resistant breast cancer cells, and low GAS5 expression enhances resistance to tamoxifen [102]. GAS5 was originally isolated as a gene that is preferentially expressed in growth-arrested NIH3T3 cells [103]. The GAS5 gene has two alternative promoters, as well as multiple exons and introns. As alternative choices of these exons and alternative promoter usage produce multiple GAS5 variants, GAS5 can be defined as an antisense lncRNA overlapping a protein-coding gene, the zinc finger and BTB domain containing 37 (ZBTB37) gene, on the opposite strand or as a bidirectional lncRNA [104]. GAS5 is a host gene of a class of small　noncoding RNAs termed box C/D small nucleolar RNAs (SNORDs); 10 SNORDs are encoded within the GAS5 intronic regions [104]. These snoRNAs are transcribed as part of the GAS5 primary transcript, and are then excised and matured. Regarding endocrine therapy resistance, a previous study using an MCF7-derived tamoxifen-resistant cell line suggests that GAS5 acts as a ceRNA by sponging miR-222 and upregulates PTEN, which is a target of miR-222 and weakens the tamoxifen resistance of breast cancer cells (Figure 4) [102].In addition to downregulating tamoxifen resistance, GAS5 exerts tumor-suppressive effects in breast cancer via several pathways. For example, GAS5 acts as a ceRNA by sponging miR-21 and upregulates the expression of miR-21 targets programmed cell death 4 (PCDC4) and PTEN, both of which are tumor-suppressor genes [105]. Moreover, GAS5 acts as a ceRNA for miR-196a-5p and downregulates the forkhead box O1 (FOXO1)/phosphoinositide 3-kinase (PI3K)/AKT pathway, thus suppressing the invasion of TNBC cells [106]. In addition to its functions as a ceRNA, GAS5 is involved in transcriptional regulation. GAS5 suppresses glucocorticoid-induced transcription and sensitizes breast cancer cells to apoptosis [107,108]. Exon 12 of GAS5 contains a hairpin structure with two sequences similar to the GR target sequence, termed glucocorticoid response element (GRE). This hairpin structure is called the GAS5 GRE-mimic, and it interacts directly with the DNA-binding domain of GR and suppresses the transcriptional activation of GR target genes, including antiapoptotic genes like cellular inhibitor of apoptosis 2 (cIAP2) and serum- and glucocorticoid-regulated kinase 1 (SGK1), which facilitate stress-inducible apoptosis [107]. In addition to the GR, GAS5 and the GAS5 GRE-mimic bind to other 3-keto steroid receptors, such as the mineralocorticoid, progesterone, and androgen receptors, and inhibit their transcriptional activities [107,108]. Interestingly, the GAS5 GRE-mimic alone can increase apoptosis in breast cancer cells, suggesting that the oligonucleotides of the GRE-mimic may be applicable to breast cancer therapy [108]. Consistent with these findings, GAS5 is downregulated in breast tumors compared with normal tissues, and low expression of GAS5 is associated with poor overall survival in patients with breast cancer and TNBC [102,104,106]. Moreover, a recent study showed that an insertion (ins)/deletion (del) polymorphism located within the GAS5 promoter (rs145204276 AGGCA/–) affects the risk of breast cancer [109]. In that study, the rs145204276 ins/del and del/del genotypes, as well as the del allele, were associated with a reduced risk of breast cancer [109]. As GAS5 expression is significantly higher in patients with breast cancer carrying the rs145204276 ins/del and del/del genotypes versus the rs145204276 ins/ins genotype carriers, and since the rs145204276 del allele increases the transcription of GAS5, this polymorphism may affect the risk of breast cancer by modulating GAS5 expression levels [109].The Down syndrome cell adhesion molecule antisense RNA 1 (DSCAM-AS1) is an intronic antisense lncRNA that is transcribed from the opposite strand of the Down syndrome cell adhesion molecule (DSCAM) gene, and is wholly derived from the intronic region of DSCAM. DSCAM-AS1 promotes tamoxifen resistance [110,111] and is upregulated in breast cancer tissues compared with normal tissues [110,112]. Moreover, DSCAM-AS1 expression is higher in luminal and HER2-positive breast cancers, and particularly in the luminal B subtype [110,112]. Importantly, previous studies have demonstrated that DSCAM-AS1 is an ERα target gene [110,112], and is important for cell proliferation and the invasion of MCF7 and T47D cells [110,112,113], as well as for the growth and liver metastasis of T47D cells xenografted into immunodeficient mice [110]. Moreover, DSCAM-AS1 expression is elevated in tamoxifen-resistant breast cancer tissues, and the knockdown and overexpression experiments of DSCAM-AS1 in breast cancer cell lines, such as MCF7 and T47D, suggest that DSCAM-AS1 promotes tamoxifen resistance [110,111]. Consistent with these results, a high expression of DSCAM-AS1 has been associated with a short disease-free survival for patients with luminal breast cancer and those with luminal breast cancer treated with endocrine therapy [113]. Although the manner in which DSCAM-AS1 functions in breast cancers remains unclear, the RNA-binding protein heterogeneous nuclear ribonucleoprotein L (hnRNPL) is required for DSCAM-AS1 activity in MCF7 and T47D cells [110]. DSCAM-AS1 interacts with hnRNPL via its 3′ region, which contains CACA-rich RNA sequences [110]. Furthermore, a previous study using MCF7-derived, tamoxifen-resistant cells suggests that DSCAM-AS1 acts as a ceRNA by sponging miR-137, which increases the expression of epidermal growth factor receptor pathway substrate 8 (EPS8), thus contributing to tamoxifen resistance (Figure 4) [111]. In addition, it was reported recently that DSCAM-AS1 functions as a ceRNA for miR-204-5p in the breast cancer susceptibility gene 1 (BRCA1)-mutated TNBC cell line HCC1937, to promote tumor growth via the upregulation of ribonucleotide reductase M2 (RRM2) [114].ESR1 locus enhancing and activating noncoding RNAs (Eleanors) were identified as a group of lncRNAs that are transcribed from inside and around the ESR1 locus, and could consist of lncRNAs of all categories [115]. Previous studies have shown that Eleanors play important roles in ER-positive breast cancer progression under estrogen-deprived conditions. Eleanors are specifically expressed in ER-positive breast cancer tissues and MCF7 cells, and are increased in MCF7 cells cultured under LTED conditions [115]. u-Eleanor is an Eleanor that is transcribed from ~40 kb upstream of the canonical promoter of ESR1 and upregulates the transcription of the ESR1 mRNA and other Eleanors to promote the proliferation of LTED cells (Figure 4) [115]. A chromatin immunoprecipitation-sequencing (ChIP-seq) analysis showed that the u-Eleanor locus in LTED cells is enriched for monomethylated H3K4, rather than trimethylated H3K4, suggesting that the u-Eleanor locus functions as an enhancer. Clinically, the upregulation of u-Eleanor has been reported to be negatively associated with increasing breastfeeding duration [116]. u-Eleanor tends to be upregulated in healthy women without a history of breastfeeding and women with a breastfeeding duration of 1–6 months. Epidemiological studies have demonstrated that breastfeeding experiences play a protective role against breast cancer in women, while a lack or a short duration of breastfeeding increases breast cancer risk [117,118,119]. Therefore, u-Eleanor may be used as a biomarker of breast cancer at early stages [116]. Furthermore, a recent study has revealed the function of another Eleanor called promoter-associated Eleanor (pa-Eleanor), which is transcribed from the region proximal to the transcriptional start site of ESR1 [120]. In the nucleus, chromosomes fold into domains called topologically associating domains (TADs), which exhibit intra-chromatin interactions [121]. The chromosome conformation capture combined with high-throughput sequencing (4C-seq) analysis, reported in a recent study, reveals that a TAD including Eleanor-expressing regions (Eleanor TAD) resides on human chromosome 6q25.1, and that Eleanor TAD contains the ESR1 gene and three other genes: coiled-coil domain containing 170 (CCDC170), chromosome 6 open reading frame 211 (C6orf211), and required for meiotic nuclear division 1 homolog (RMND1) [120]. pa-Eleanor upregulates genes within Eleanor TAD and promotes the proliferation of LTED cells (Figure 4). In addition, pa-Eleanor upregulates u-Eleanor, whereas u-Eleanor does not affect pa-Eleanor expression, suggesting that pa-Eleanor upregulates the transcription of the ESR1 mRNA through u-Eleanor [120]. Moreover, pa-Eleanor enhances an intra-chromosomal interaction between the ESR1 promoter region and the region near the FOXO3 locus on human chromosome 6q21 [120]. This chromosomal interaction may affect the expression of genes within Eleanor TAD. FOXO3 is a transcription factor that induces apoptosis through the transcriptional regulation of apoptosis-associated genes, and its expression is elevated in LTED cells. The knockdown of pa-Eleanor decreases ESR1 expression levels (but does not affect the expression of FOXO3) and induces the apoptosis of LTED cells. Therefore, pa-Eleanor may promote the survival of LTED cells by regulating the balanced expression of ESR1 and FOXO3 [120]. Thus, the functions of u-Eleanor and pa-Eleanor suggest that the regulation of Eleanor expression may represent a new treatment strategy for breast cancer adapted to estrogen-deprived conditions. Consistent with this idea, resveratrol and glyceollin I, phytoalexins that are synthesized in plants under stress conditions, decrease the expression of Eleanors to induce apoptotic death in LTED cells [122].In addition to these lncRNAs, recent gene expression analysis in patients with ER-positive breast cancer, who were primarily treated with tamoxifen, identified 11 lncRNAs, belonging to multiple categories (PINK1-AS, RP11-259N19.1, KLF3-AS1, LINC00339, LINC00472, RP11-351I21.11, KB-1460A1.5, PKD1P6-NPIPP1, PDCD4-AS1, PP14571, and RP11-69E11.4), as prognostic lncRNAs that predict the risk of systemic relapse [123]. PINK1-AS, RP11-259N19.1, KLF3-AS1, PDCD4-AS1, PP14571, and RP11-69E11.4 are antisense lncRNAs, while LINC00339, LINC00472, RP11-351I21.11, and KB-1460A1.5 are lincRNAs. PKD1P6-NPIPP1 is a read-through transcript derived from two pseudogenes, polycystin 1, transient receptor potential channel interacting pseudogene 6 (PKD1P6) and nuclear pore complex interacting protein pseudogene 1 (NPIPP1), and classified as an intronic antisense lncRNA, because PKD1P6-NPIPP1 is wholly derived from the opposite strand of the intronic region of the pyridoxal dependent decarboxylase domain containing 1 (PDXDC1) gene. Although the mechanisms by which these 11 lncRNAs are involved in tamoxifen resistance and systemic relapse are unclear, several relapse- or metastasis-related pathways, such as the PI3K/AKT and Wnt signaling pathways, are upregulated in patients with breast cancer who have a high relapse risk predicted by the expression levels of these lncRNAs. Thus, it suggests that these signaling pathways may play important roles in the functions of the 11 prognostic lncRNAs [123].Considering that lncRNAs play essential roles in endocrine therapy resistance, intervention against lncRNAs may be promising for breast cancer treatment. Antisense oligonucleotides (ASOs) are used for regulating the stability and activity of RNAs. Some chemically modified ASOs targeting transcripts of protein-coding genes have been approved for clinical use by the U.S. Food and Drug Administration (FDA) [124]. For example, a 2′-O-(2-methoxyethyl) (MOE) phosphorothioate (PS) ASO called nusinersen is used for the treatment of spinal muscular atrophy (SMA). [125,126]. SMA is an autosomal-recessive neuromuscular disorder with degeneration of the motor neurons in the anterior horn of the spinal cord, leading to atrophy of the voluntary muscles of the limbs and trunk [125]. SMA is caused by deletions or loss-of-function mutations of survival of motor neuron 1, telomeric (SMN1) gene and the consequent reduced expression of survival of motor neuron (SMN) proteins from SMN1 transcripts. Although there is a homologue of SMN1 gene called survival of motor neuron 2, centromeric (SMN2), SMN proteins are not efficiently produced from SMN2 transcripts. The SMN2 gene has an identical coding sequence but differs from SMN1 gene by 11 nucleotides. The different sequences between these genes contain a C-to-T mutation on exon 7, which is a synonymous mutation but affects SMN protein expression by promoting the skipping of exon 7. Due to this mutation, 80%–90% of SMN2 mRNAs lack exon 7 and are translated into truncated SMN proteins, which are rapidly degraded (Figure 5A). Therefore, the SMN2 gene does not fully compensate for the loss-of-function of the SMN1 gene [125], and modulating the splicing pattern of SMN2 pre-mRNA to produce the full-length SMN proteins is one of therapeutic strategies of SMA. Nusinersen is an ASO complementary to a site within intron 7 of the SMN2 pre-mRNA called intronic splicing silencer-N1 (ISS-N1), which is involved in the skipping of exon 7 and blocks the activity of ISS-N1 to facilitate the inclusion of the exon 7, resulting in the synthesis of the functional, full-length SMN proteins and the rescue of motor neurons (Figure 5A) [124,125]. Another oligonucleotide drug, mipomersen, is used to treat homozygous familial hypercholesterolemia, an autosomal disorder of the lipid metabolism characterized by elevated levels of low-density lipoprotein (LDL) cholesterol [124,127]. Mipomersen targets the transcripts of the apolipoprotein B (APOB) gene. The middle region of mipomersen shows DNA-like properties and induces the cleavage of these transcripts mediated by ribonuclease H (RNase H), which cleaves RNAs that form heteroduplexes with DNA. As the apoB-100 protein, encoded by APOB gene, is a component of LDL cholesterol, mipomersen-mediated downregulation of APOB decreases the circulating levels of LDL cholesterol (Figure 5B) [123,126,128]. In addition to ASOs, a small interfering RNA (siRNA)-based drug, patisiran, was recently approved by the FDA [124]. Therefore, the targeting lncRNAs with ASOs and siRNAs may be translated into new therapies for breast cancer.In this review, we describe the functions and mechanisms of lncRNAs related to the endocrine therapy resistance of breast cancer (Table 1), and their potential as therapeutic targets. Additionally, LncRNAs may hold promise as biomarkers of breast cancer. Importantly, the quantification of the prostate cancer antigen 3 (PCA3) lncRNA in urine samples has been developed as a diagnostic test for prostate cancer [129], suggesting that lncRNAs may be applicable to the analysis of non-invasive liquid biopsies for the diagnosis of cancers, including breast cancer. Thus, lncRNAs are potential key factors in the development of new strategies of breast cancer treatment, and further studies of lncRNAs in the context of breast cancer are required.Conceptualization, T.T. and S.I.; writing—original draft preparation, T.T.; writing—review and editing, K.I., Y.M., K.H.-I., and S.I.; supervision, S.I. All authors have read and agreed to the published version of the manuscript.This study was supported by the Support Project of Strategic Research Center in Private Universities from the Ministry of Education, Culture, Sports, Science and Technology (MEXT); grants from the Japan Society for the Promotion of Science (15K15353 to S.I. and 17H04205 to K.H.-I.); and Practical Research for Innovative Cancer Control (JP18ck0106194 to K.I.), and Project for Cancer Research and Therapeutic Evolution (P-CREATE, JP18cm0106144 to S.I.) from the Japan Agency for Medical Research and Development (AMED).The authors declare no conflict of interest.Classification of long noncoding RNAs (lncRNAs). Based on the positions of their loci on the genome, lncRNAs are classified into one or more of five categories: (A) sense, (B) antisense, (C) bidirectional, (D) intronic, and (E) intergenic.Schematic representation of the functions of HOTAIR, TMPO-AS1, and ADAMTS9-AS2 in the tamoxifen resistance of breast cancer. HOTAIR enhances tamoxifen resistance by regulating the expression and activity of ERα. TMPO-AS1 binds and stabilizes ESR1 mRNA to enhance tamoxifen resistance. On the other hand, ADAMTS9-AS2 downregulates tamoxifen resistance by competing with miR-130a-5p to increase PTEN expression. HOTAIR: HOX transcript antisense RNA; TMPO-AS1: thymopoietin antisense transcript 1; ADAMTS9-AS2: ADAM metallopeptidase with thrombospondin type 1 motif 9 (ADAMTS9) antisense RNA 2; ERα: estrogen receptor α; PTEN: phosphatase and tensin homolog.Schematic representation of the functions of UCA1, lincRNA-ROR, and lncRNA MIR2052HG in the tamoxifen resistance of breast cancer. (A) UCA1 promotes the tamoxifen resistance by several mechanisms. UCA1 activates mTOR, Wnt/β-catenin, and PI3K/AKT signaling pathways to enhance tamoxifen resistance. In addition, UCA1 binds to EZH2 and epigenetically suppresses p21 expression. Moreover, UCA1 sponges miR-18 to upregulate HIF1α expression. Since HIF1α induces UCA1 expression, UCA1 and HIF1α form a feedback regulatory loop to strengthen tamoxifen resistance. (B) LincRNA-ROR enhances tamoxifen resistance by inhibiting autophagy. Moreover, lincRNA-ROR induces the degradation of an ERK-specific phosphatase, DUSP7, resulting in ERα activation mediated by the MAPK/ERK signaling pathway. LincRNA-ROR also acts as a competing endogenous RNA (ceRNA), which sponges miR-205-5p to upregulate the expression of EMT-related genes ZEB1/2 and contributes to tamoxifen resistance. (C) MIR2052HG increases the expression of LMTK3. LMTK3 suppresses the activity of PKC, which increases the expression of ESR1 mRNA and ERα protein through the inactivation of AKT and MEK/ERK/RSK1 signaling pathway, respectively. UCA1: urothelial cancer associated 1; lincRNA-ROR: large intergenic noncoding RNA-regulator of reprogramming; MIR2052HG: miR2052 host gene; mTOR: mammalian target of rapamycin; PI3K: phosphoinositide 3-kinase; EZH2: enhancer of zeste homolog 2; HIF1α: hypoxia-inducible factor 1α; DUSP7: dual specificity phosphatase 7; MAPK/ERK: mitogen-activated protein kinase (MAPK)/extracellular signal-regulated kinase (ERK); ZEB1/2: zinc finger E-box binding homeobox 1/2; LMTK3: lemur tyrosine kinase 3; PKC: protein kinase C; ESR1: estrogen receptor 1; MEK: MAPK/ERK kinase; RSK1: ribosomal S6 kinase 1.Schematic representation of the functions of GAS5, DSCAM-AS1, and Eleanors in the tamoxifen resistance of breast cancer. GAS5 sponges miR-222 and upregulates PTEN expression to enhance tamoxifen resistance. DSCAM-AS1 also sponges miR-137 to increase EPS8, which contributes to tamoxifen resistance. On the other hand, Eleanors promotes tamoxifen resistance by upregulating ESR1 expression. GAS5: growth-arrest specific 5; DSCAM-AS1: Down syndrome cell adhesion molecule antisense RNA 1; Eleanors: ESR1 locus enhancing and activating noncoding RNAs; PTEN: phosphatase and tensin homolog; EPS8: epidermal growth factor receptor pathway substrate 8.Antisense oligonucleotides (ASOs) in clinical use. (A) Nusinersen binds to a splicing regulatory sequence called intronic splicing silencer-N1 (ISS-N1) within intron 7 of SMN2 pre-mRNA, and enhances the inclusion of exon 7, resulting in the production of SMN2 mRNA coding the full-length SMN protein. (B) Mipomersen binds to ApoB-100 mRNA and causes its degradation, mediated by ribonuclease H (RNaseH). SMN2: survival of motor neuron 2: centromeric; SMN: survival of motor neuron; ApoB-100: apolipoprotein B-100.LncRNAs regulating endocrine therapy resistance in breast cancer.Activating mTOR, Wnt/β-catenin, and PI3K/AKT signaling pathways [53,54,55];Promoting EZH2 mediated repression of p21 [55];Inhibiting miR-18a activity to increase HIF1α expression [56]Inhibiting autophagy [85];Promoting ligand-independent activation of ERα and estrogen-independent growth [86];Inhibiting miR-205-5p activity to increase the expression of ZEB1/2 [84]

Med-MDPI/cancers/cancers-12-06-01425.txt

@@ -0,0 +1 @@
+Current address: SocraMetrics GmbH, Mainzerhofpl. 14, 99084 Erfurt, Germany.Current address: AstraZeneca, Granta Park, Great Abington, Cambridge CB21 6GP, UK.This phase 1 trial (NCT01938846) determined the maximum tolerated dose (MTD) of the mTOR serine/threonine kinase inhibitor, BI 860585, as monotherapy and with exemestane or paclitaxel in patients with advanced solid tumors. This 3+3 dose-escalation study assessed BI 860585 monotherapy (5–300 mg/day; Arm A), BI 860585 (40–220 mg/day; Arm B) with 25 mg/day exemestane, and BI 860585 (80–220 mg/day; Arm C) with 60–80 mg/m2/week paclitaxel, in 28-day cycles. Primary endpoints were the number of patients with dose-limiting toxicities (DLTs) in cycle 1 and the MTD. Forty-one, 25, and 24 patients were treated (Arms A, B, and C). DLTs were observed in four (rash (n = 2), elevated alanine aminotransferase/aspartate aminotransferase, diarrhea), four (rash (n = 3), stomatitis, and increased gamma-glutamyl transferase), and two (diarrhea, increased blood creatine phosphokinase) patients in cycle 1. The BI 860585 MTD was 220 mg/day (Arm A) and 160 mg/day (Arms B and C). Nine patients achieved an objective response (Arm B: Four partial responses (PRs); Arm C: Four PRs; one complete response). The disease control rate was 20%, 28%, and 58% (Arms A, B, and C). The most frequent treatment-related adverse events (AEs) were hyperglycemia (54%) and diarrhea (39%) (Arm A); diarrhea (40%) and stomatitis (40%) (Arm B); fatigue (58%) and diarrhea (58%) (Arm C). The MTD was determined in all arms. Antitumor activity was observed with BI 860585 monotherapy and in combination with exemestane or paclitaxel.The PI3K/AKT/mTOR pathway plays an important role in the regulation of metabolism, survival, and proliferation of mammalian cells [1,2,3]. This pathway is often hyperactivated in human cancers and has been associated with resistance to conventional therapy [4].mTOR comprises at least two different complexes, the mTOR complex 1 (rapamycin-sensitive mTORC1) and complex 2 (mTORC2) [5]. As the main downstream effectors in the PI3K/AKT pathway, the mTOR complexes are central factors in the regulation of cell growth, proliferation, metabolism, angiogenesis, and cell survival processes [3]. Thus, targeting the mTOR complexes could play an important role in cancer therapy [6].First-generation mTOR inhibitors (rapalogs), derived from the immunosuppressive drug rapamycin, demonstrated limited therapeutic success [7], exhibiting only partial inhibition of mTORC1 signaling, and no inhibition of mTORC2 or feedback-activation of PI3K/AKT signaling via S6K and mTORC2 activity [8]. A new generation of ATP-competitive mTOR inhibitors currently in clinical development were designed to target the kinase domains of mTOR and fully inhibit both mTORC1 and mTORC2 [7]. BI 860585 is a potent, selective, ATP-competitive mTORC1 and mTORC2 serine/threonine kinase inhibitor showing strong preclinical efficacy against various sarcoma types [9].Although data from early clinical trials with rapalogs and other mTOR inhibitors have demonstrated modest response rates as single agents [10,11], mTOR inhibition in combination with chemotherapy, hormone therapy, and other targeted therapies may be more effective than monotherapy. These combinations are hypothesized to induce tumor regression and could circumvent acquired resistance. This hypothesis is supported by preclinical and clinical studies with different rapalogs (e.g., temsirolimus and everolimus) in advanced cancers [4,12,13]. Of note, in the phase 3 BOLERO-2 study, everolimus in combination with the aromatase inactivator, exemestane, significantly improved PFS versus exemestane alone in postmenopausal women with hormone-receptor (HR)-positive advanced breast cancer who had relapsed or progressed on nonsteroidal aromatase inhibitors [4]. This treatment combination has since been approved by the US Food and Drug Administration (FDA) and European Medicines Agency (EMA) for use in this setting.To investigate the potential for combining mTOR-complex inhibition with current standard therapies, this phase 1, dose finding, first-in-human study evaluated BI 860585 as monotherapy and in combination with either exemestane or paclitaxel in heavily pretreated patients with advanced and/or metastatic solid tumors. Maximum tolerated doses (MTDs) of BI 860585 as monotherapy, and combined with exemestane or paclitaxel, were identified; adverse events (AEs) were generally predictable and manageable. Further, antitumor activity was seen in all treatment arms.The first patient was enrolled on 10 September, 2013 and the last patient was enrolled on 02 June, 2016. A total of 41 patients were treated in Arm A, 25 in Arm B, and 24 in Arm C. Patient baseline characteristics are shown in Table 1. The median age was 61 years (range 20–79 years) and most patients (95.6%) had an Eastern Cooperative Oncology Group performance status (ECOG PS) < 2. The majority of patients for whom data were available had ≥3 metastatic sites at screening and had received at least one prior line of either chemotherapy, radiotherapy, and/or prior surgery.Patients received BI 860585 5–300 mg/day as a single agent (Arm A), 40–220 mg/day in combination with exemestane (Arm B), and 80–220 mg/day in combination with paclitaxel (Arm C). At the time of data analysis (27 November, 2017), all patients had discontinued trial medication. Reasons for discontinuation included progressive disease (66.7%), AEs (16.7%), and patient withdrawal (5.6%). The median (range) treatment exposure to BI 860585 was 56 (17–561) days in Arm A, 56 (7–784) days in Arm B, and 125 (13–448) days in Arm C. The median (range) treatment exposure to exemestane (Arm B), and paclitaxel (Arm C) was 63 (14–791) days and 14.5 (3–32) infusions, respectively.Four evaluable patients treated in Arm A experienced a dose-limiting toxicity (DLT) during cycle 1 (all grade 3; Table 2); rash (120 mg (n = 1)), elevated alanine aminotransferase/aspartate aminotransferase (160 mg (n = 1)), diarrhea, and rash (both 300 mg (n = 2)). The MTD of BI 860585 was 220 mg/day (Table 2).Four treated and evaluable patients in Arm B experienced DLTs; grade 3 rash (120 mg (n = 1) and 220 mg (n = 2)), and grade 3 stomatitis and increased gamma-glutamyl transferase (160 mg; n = 1). The MTD of BI 860585 in combination with exemestane was 160 mg/day.Two evaluable patients treated in Arm C experienced a DLT during cycle 1 (both grade 3); diarrhea and increased blood creatine phosphokinase (both at 220 mg). The MTD of BI 860585 in combination with paclitaxel was 160 mg/day.All patients experienced at least one AE, including laboratory abnormalities, and most patients experienced a treatment-related AE (88% in Arm A, 96% in Arm B, and 100% in Arm C (Table 3)). The most frequent treatment-related AEs were hyperglycemia (54%; n = 22), diarrhea (39%; n = 16), and nausea (37%; n = 15) in Arm A; diarrhea (40%; n = 10), stomatitis (40%; n = 10), hyperglycemia (36%; n = 9), and fatigue (36%; n = 9) in Arm B; fatigue (58%; n = 14), diarrhea (58%; n = 14), hyperglycemia (54%; n = 13), anemia (50%; n = 12), and decreased appetite (46%; n = 11) in Arm C. The majority of treatment-related AEs were of grade ≤3 in severity; there were no grade 5 treatment-related AEs. One patient experienced a transient grade 3 elevation of blood creatine phosphokinase (DLT), but there were no clinical manifestations suggestive of adverse musculoskeletal effects in this patient.A total of 22 patients (54%) in Arm A, 10 patients (40%) in arm B, and 13 patients (54%) in Arm C had serious AEs. Seven (17%) patients in Arm A, nine (36%) patients in Arm B, and six (25%) patients in Arm C experienced AEs leading to dose reduction of BI 860585. Thirteen (32%) patients in Arm A, seven (28%) patients in Arm B, and five (21%) patients in Arm C experienced AEs leading to discontinuation of BI 860585.Best overall tumor responses for each treatment arm are shown in Table 4 and Supplementary Figure S1, including the percentage change from baseline in target lesions. The objective response rate (ORR) was 0%, 16% (four partial responses (PRs) including an estrogen-receptor-positive breast cancer patient who received a first-line combination of ridaforolimus, dalotuzumab, and exemestane) and 21% (one complete response (CR) and four PRs) in Arms A, B, and C, respectively (Table 4). The patient achieving a CR was a 69-year-old female with breast cancer who had received approximately 10 prior lines of therapy including: Epirubicin, CMF (5-FU, methotrexate, and capecitabine) carboplatin and paclitaxel; capecitabine and vinorelbine, nabpaclitaxel, exemestane plus everolimus; and multiple lines of hormone therapies. At screening for this trial, she presented with skin metastases and soft tissue involvement. At baseline; this patient did not have measurable disease/target lesions and only one nontarget lesion was documented. The patient received BI 860585 220 mg plus 80 mg paclitaxel for six cycles and achieved a CR, measured in the nontarget lesion, on day 120, which was maintained for 63 days. The median (range) duration of objective response was eight months (0–23) for Arm B and five months (2–9) for Arm C. One and three of the responding patients in Arm B and Arm C, respectively, had received ≥3 lines of previous chemotherapy. The disease control rate (DCR; CR + PR + stable disease (SD)) was 20% in Arm A, 28% in Arm B, and 58% in Arm C. The median (range) duration of disease control was eleven (4–17), nine (2–25), and seven (4–15) months for Arms A, B, and C, respectively (Table 4).Peak plasma concentrations (Cmax) occurred 2–6 h following once-daily oral administration of BI 860585 (5–300 mg) in Arm A. The trough BI 860585 plasma concentration did not increase after day 8, indicating that a steady state had been reached (Figure S2). Repeated daily dosing resulted in 1.89-fold accumulation of BI 860585 area under the curve (AUC) values at steady state. Inter-individual variability of the AUC at steady state at the MTD in Arm A was moderate (geometric coefficient of variation 24.3%). BI 860585 plasma exposure was almost dose-proportional following a single-dose administration across the complete dose range tested (5–300 mg), and after multiple dosing between 40 and 220 mg (Table S1; Figure S2). See Table S2 for the pharmacokinetic characteristics at the MTD of the three arms. Pharmacokinetics were determined in preselected patients enrolled in Arm A (120, 160, and 220 mg dose cohorts) with and without food (see Table S3 for proposed composition of standard continental breakfast) on day 1 and day 2 of the first treatment cycle. The presence of food had an influence on the rate of absorption, i.e., caused a delay and a reduction of less than 20% in Cmax. The effect of food on the extent of absorption was minimal, with the AUC0–24 being reduced by less than 10%.In Arm A, dose proportionality was confirmed after single administration. Exposure at steady state increased with the dose in an almost proportional manner over the dose range tested (5–300 mg). In Arms B and C, dose proportionality at steady state could be established for the dose ranges tested (40–220 mg BI 860585 in Arm B, and 80–220 mg BI 860585 in Arm C). In all three treatment arms, median reductions in AKT phosphorylation at Ser473 (pAKT)/AKT ratios of 45–54% were observed in platelet-rich plasma blood samples within 3 h following BI 860585 doses of 120 mg or greater (Figure S3). At steady state, the inhibition persisted for all three treatment arms. There was no clear indication of a pharmacokinetic/pharmacodynamic correlation, or a correlation between pAKT/AKT ratio reductions and disease control. Data for BI 860585 in combination with exemestane and paclitaxel were consistent with monotherapy.This phase 1 study demonstrated that the mTOR inhibitor, BI 860585, is tolerable as monotherapy or combined with standard treatments. The MTDs established in this study were BI 860585 220 mg/day as monotherapy, BI 860585 160 mg/day plus exemestane 25 mg/day, and BI 860585 160 mg/day plus paclitaxel 80 mg/m2/week. Safety was similar across the three treatment arms, with AEs that were generally manageable and consistent with the mechanisms of action of the study drugs. The most common DLTs were rash and diarrhea (Table 2), and the most frequently reported treatment-related AEs included hyperglycemia, diarrhea, nausea, rash, fatigue, and stomatitis (Table 3). These are consistent with previous reports of mTOR-inhibitor treatment of advanced solid tumors [14,15]. Higher incidences of anemia and neutropenia were observed in patients receiving BI 860585 combined with paclitaxel compared with those receiving BI 860585 alone or in combination with exemestane. These AEs are consistent with the known safety profile of paclitaxel [16].Signs of antitumor activity across various tumor types were observed with BI 860585 monotherapy and with the combination therapies (Table 4; Figure S1). Objective responses were observed with BI 860585 in combination with exemestane (16%) or paclitaxel (21%); in the monotherapy arm, 20% of patients had SD. These results are particularly encouraging considering that most patients who responded had been heavily pretreated with chemotherapy or hormonal cancer treatment. Given the small patient numbers, it is challenging to suggest a patient population that could derive a particular benefit from BI 860585 in combination with exemestane or paclitaxel; however, responses were seen in two patients with breast cancer, a patient population for which the combination of the mTOR inhibitor everolimus is approved in combination with exemestane. It is also encouraging that one of the patients responding had progressed on prior mTORC1-based therapy, suggesting that BI 860585 was able to resensitize the tumor through dual mTORC1/2 inhibition; however, this would need to be confirmed in larger studies.Pharmacokinetic analyses demonstrated almost dose-proportional plasma exposure during BI 860585 monotherapy following single administration and after multiple dosing. The rate of absorption of BI 860585 was rapid, and although the presence of food had a slight effect on the rate of absorption, the extent of absorption was similar in the fasting or fed state (Table S1). A reduction in Cmax after food has been described for other mTOR targeted agents [17]. Taken together, the safety and pharmacokinetic data suggest that treatment with BI 860585 as monotherapy or in combination with exemestane or paclitaxel is feasible and tolerable for patients with advanced solid tumors.Biomarker analyses were conducted to evaluate the biological activity of BI 860585. Decreased AKT phosphorylation at Ser473 has been identified as a pharmacodynamic biomarker for mTORC2 inhibition [18,19]. In this study, a 45%–54% reduction in pAKT/AKT ratios was observed in platelet-rich plasma blood samples within 3 h of dosing in all treatment arms (Figure S3), suggesting that treatment with BI 860585 has an impact on the intended molecular target. The reduction of pAKT persisted at steady-state concentrations of BI 860585. There was no clear correlation between pharmacokinetics/pharmacodynamics or between reductions in pAKT/AKT ratios and disease control, likely due to the high level of variability in the data, together with the small sample sizes. Biomarker findings with BI 860585 in combination with exemestane and paclitaxel were consistent with monotherapy.A dose-expansion stage was originally planned, but the development of BI 860585 was discontinued during this study due to a strategic decision made by the sponsor, and consequently, the dose-expansion stage was cancelled. Nevertheless, the results of this trial will contribute to informed decision-making about the clinical development of other mTOR inhibitors. Given the mode of action and available clinical data for compounds targeting the PI3K/AKT/mTOR pathway, including the preliminary antitumor activity shown in this trial, it is expected that a combination strategy either with endocrine, chemotherapy, and/or other targeted therapies may result in the most efficacious use of this class of drug. Further studies of mTOR inhibitors are warranted to evaluate these regimens in relevant patient populations, particularly if a predictive marker can be defined to select patients who would gain the most from this treatment approach. Eligible patients were aged ≥18 years and had advanced nonresectable and/or metastatic solid tumors, disease progression, and an ECOG PS ≤ 2. Full patient eligibility criteria for all treatment arms are included in the  Supplementary Methods and Table S4.The study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines and the study protocol (study protocol code 1325.1) was approved by the Institutional Review Boards of the participating institutions. Written informed consent was obtained from all patients.This multicenter, open-label, phase 1 trial (NCT01938846) employed a 3+3 dose escalation design to determine the MTD, safety, pharmacokinetics, and antitumor activity of BI 860585 as monotherapy and in combination with exemestane or paclitaxel in patients with advanced and/or metastatic solid tumors. The study was conducted at four study sites in Italy and Belgium and recruitment was staggered across three distinct treatment arms running in parallel (Figure 1). In the monotherapy arm (Arm A), patients received oral continuous daily dosing of BI 860585 at a starting dose of 5 mg/day over a 28-day cycle. Enrolment in the combination Arm B opened when the first treatment-related AE of grade ≥2 had been observed in Arm A, and the starting dose was determined by the Safety Monitoring Committee (SMC) as 40 mg, i.e., the dose level at which no drug-related grade ≥2 AE had been observed in Arm A. Enrolment in the combination Arm C opened when the SMC had determined a safe starting dose of 80 mg, i.e., the dose level at which no drug-related grade ≥3 AE had been observed in the dose escalation of Arm B. Patients were assigned to the individual combination arms (Arm B and Arm C) based on the investigators’ clinical judgment. In Arm B, patients received oral exemestane 25 mg/day (standard fixed dose) continuously on a 28-day cycle. In Arm C, patients received an intravenous infusion of paclitaxel 60 mg/m2 in the first dose level cohort, then escalated to 80 mg/m2 (standard dose) once weekly on a 28-day cycle. Treatment with exemestane or paclitaxel began on day 7 of cycle 1. Patients were eligible for repeated cycles until disease progression or intolerable AEs.The primary endpoints for this study were the MTD and the number of DLTs in each treatment arm. The MTD was based on the number of patients with a DLT during the first cycle, and was defined as the dose at which no more than one of six patients experienced a DLT during cycle 1 or the dose level below which ≥2 of six patients experienced a DLT during cycle 1. A DLT was defined as any of the following: Grade 4 neutropenia lasting ≥7 days; grade ≥3 febrile neutropenia/neutropenia with documented infection; grade 3 thrombocytopenia associated with bleeding requiring transfusion; grade 4 thrombocytopenia or anemia; any grade ≥3 nonhematologic toxicity that persisted despite adequate medical intervention or prophylaxis; any grade 3 hyperglycemia that did not recover to grade ≤1 within two weeks of adequate therapy; and any toxicity resulting in a >14 day delay in starting cycle 2.Other endpoints included ORR and DCR, both per Response Evaluation Criteria in Solid Tumors (RECIST) criteria version 1.1 [20], duration of objective response/clinical benefit, safety, and pharmacokinetic parameters of BI 860585, administered as a single agent and in combination regimens, with or without food, at the MTD for each arm. Detailed definitions and methodology for secondary assessments and pharmacokinetic parameters are described in the Supplementary Methods.Safety was assessed by monitoring AEs (National Cancer Institute Common Terminology Criteria for Adverse Events version 4.03), clinical laboratory parameters, electrocardiograms, and vital signs. Patients were included in the safety analysis if they had taken at least one dose of any trial medication. Patients were evaluable for DLTs if they had been observed for at least the first treatment cycle and had undergone all of the safety assessments.Plasma concentration-time profiles and pharmacokinetic parameters of BI 860585 were determined in all patients who received oral doses of 5–220 mg. In addition, an exploratory analysis of the effect of food was conducted for the BI 860585 monotherapy arm. The following definition of a standard continental breakfast was given in order to standardize the food intake prior to assessing food effects on BI 860585: One egg, two bread rolls, 20 g butter, 25 g cheese, 25 g ham/saveloy, 25 g jam, and one cup (~250 mL) of decaffeinated tea or coffee (average energy value per breakfast: 688 kcal or 2880 kJ). Alternative food components/quantity could be proposed by the investigator but the caloric breakdown of the test meal had to be consistent with the one indicated. See Supplementary Table S3 for the proposed composition of the standard continental breakfast. Pharmacokinetic parameters of interest were half-life (T1/2), time to reach maximum plasma concentration (Tmax), Cmax, and AUC for BI 860585 after single dosing and at steady state, when administered as a single agent and in combination with exemestane or paclitaxel.Decreased pAKT is a pharmacodynamic biomarker for mTORC2 inhibition [18,19]. As such, exploratory analyses of pAKT were conducted using platelet-rich plasma blood samples from patients in the different treatment arms (Supplementary Methods; Table S5).All patients who were treated with at least one dose of BI 860585 were included in the analyses of safety and efficacy. All statistical analyses were descriptive and exploratory, and no formal statistical tests were performed.The current phase 1 study identified the MTDs of BI 860585 as monotherapy (220 mg/day), and combined with exemestane (BI 860585 160 mg/day plus exemestane 25 mg/day) or paclitaxel (BI 860585 160 mg/day plus paclitaxel 80 mg/m2/week). These treatments had manageable safety profiles, and DLTs and treatment-related AEs were consistent with previous reports of mTOR-inhibitor treatment of advanced solid tumors [14,15]. Further, both monotherapy and combination regimens showed evidence of antitumor activity in this heavily pretreated patient population, with objective responses observed in the combination arms.The following are available online at https://www.mdpi.com/2072-6694/12/6/1425/s1, Figure S1: (a) Best percentage change from baseline in target lesions (Arm A: Monotherapy); (b) best percentage change from baseline in target lesions (Arm B: Combination with exemestane); (c) best percentage change from baseline in target lesions (Arm C: Combination with paclitaxel); Figure S2: Plasma concentration-time profiles after once-daily administration of 5 to 300 mg BI 860585 (semi-logarithmic scale); Table S1: Statistical evaluation of the effect of food on exposure to BI 860585 monotherapy; Table S2: Steady-state pharmacokinetic characteristics obtained at day 22 at the MTD in all arms; Table S3: Proposed composition of the standard continental breakfast; Table S4: Key inclusion and exclusion criteria; and Table S5: Blood sampling schedule for pharmacokinetic analyses during cycle 1.Conceptualization, F.d.B., M.O.-K., J.B., D.F., and G.L.M.; methodology, G.L.M.; validation, D.B., S.W.H.R., and J.B.; formal analysis, J.B., J.R., D.F., and J.H.; investigation, F.d.B., J.-P.H.M., S.W.H.R., M.D., M.L., S.S., S.D., A.D., J.B., and S.C.; resources, F.d.B., J.-P.H.M., M.D., L.D.F.L., M.T., and S.C.; data curation, D.B.; writing—original draft preparation, F.d.B., J.-P.H.M., and M.O.-K.; writing—review and editing, F.d.B., J-P.H.M., D.B., S.W.H.R., M.D., M.L., S.S., S.D., L.D.F.L., M.T., A.D., M.O.-K., J.B., J.R., D.F., J.H., G.L.M., and S.C.; visualization, F.d.B., J.-P.H.M., D.B., M.O.-K., and J.B.; supervision, F.d.B., S.W.H.R., S.S., A.D., and M.O.-K.; project administration, F.d.B. and S.W.H.R. All authors have read and agreed to the published version of the manuscript. To ensure independent interpretation of clinical study results, Boehringer Ingelheim grants all external authors access to all relevant material, including participant-level clinical study data, and relevant material as needed by them to fulfill their role and obligations as authors under the ICMJE criteria. Furthermore, clinical study documents (e.g., study report, study protocol, statistical analysis plan) and participant clinical study data are available to be shared after publication of the primary manuscript in a peer-reviewed journal and if regulatory activities are complete and other criteria met per the BI Policy on Transparency and Publication of Clinical Study Data: https://trials.boehringer-ingelheim.com/transparency_policy.html. Prior to providing access, documents will be examined, and, if necessary, redacted and the data will be de-identified, to protect the personal data of study participants and personnel, and to respect the boundaries of the informed consent of the study participants. Clinical Study Reports and Related Clinical Documents can be requested via this link: https://trials.boehringer-ingelheim.com/trial_results/clinical_submission_documents.html. All such requests will be governed by a Document Sharing Agreement. Bona fide, qualified scientific and medical researchers may request access to de-identified, analyzable participant clinical study data with corresponding documentation describing the structure and content of the datasets. Upon approval, and governed by a Data Sharing Agreement, data are shared in a secured data-access system for a limited period of one year, which may be extended upon request. Researchers should use https://trials.boehringer-ingelheim.com to request access to study data.This research was funded by Boehringer Ingelheim, grant number N/A.The authors would like to thank the members of the laboratory of Kristiane Wetzel, for their assistance with assay development, validation, and sample analysis for biomarkers. Medical writing assistance, supported financially by Boehringer Ingelheim, was provided by Hannah Simmons of GeoMed, an Ashfield company, part of UDG Healthcare plc, during the preparation of this article. Data were presented in part at the American Society of Clinical Oncology (ASCO) Annual Meeting, Chicago IL, USA, 3–7 June, 2016; and the 14th International Congress on Targeted Anticancer Therapies (TAT), Washington DC, USA, 21–23 March, 2016.Filippo de Braud reports membership of an advisory board or committee for TizianaLife Sciences, BMS, Celgene, Novartis, Servier, Pharm Research Associated, Daiichi Sankyo, Ignyta, Amgen, Pfizer, Octimet Oncology, Incyte, Teofarma, Pierre Fabre, Roche, and EMD Serono; and consultancy for or receipt of speaker fees from BMS, Eli Lilly, Roche, Amgen, AstraZeneca, Gentili, Fondazione Menarini, Novartis, MSD, Ignyta, Bayer, Noema S.r.l., ACCMED, Dephaforum S.r.l., Nadirex, Roche, Biotechspert Ltd., PriME Oncology, and Pfizer. Jean-Pascal H. Machiels reports membership of an advisory board or committee for Pfizer, Roche, AstraZeneca, Bayer, Innate, Merck Serono, Boehringer Ingelheim, BMS, Novartis, Janssen, Incyte, Cue Biopharma, ALX Oncology, MSD, Debio, and Nanobiotix (managed by institution); receipt of travel grants from Amgen, BMS, Pfizer, and MSD; and receipt of a grant and a refund for work performed during the trial from Boehringer Ingelheim. Marcello Tiseo reports membership of an advisory board or committee for Boehringer Ingelheim; consultancy for/receipt of speaker fees from Boehringer Ingelheim; research grants from AstraZeneca, and Boehringer Ingelheim; receipt of advisory board and/or speaker fees from AstraZeneca, BMS, MSD, Boehringer Ingelheim, and Takeda. Mahmoud Ould-Kaci, Juergen Braunger, and Daniela Fischer report employment by Boehringer Ingelheim. Juliane Rascher reports prior employment by Boehringer Ingelheim, and current employment by SocraMetrics GmbH. Josef Hoefler reports membership of Staburo GmbH board of directors; consultancy for/receipt of speaker fees from Boehringer Ingelheim, and receipt of other financial support from Boehringer Ingelheim. Gabriella L. Mariani reports prior employment by Boehringer Ingelheim, and current employment by AstraZeneca. Daniela Boggiani, Sylvie W.H. Rottey, Matteo Duca, Marie Laruelle, Stefania Salvagni, Silvia Damian, Lore D.F. Lapeire, Alexandre Dermine, and Sara Cresta report no conflicts of interest. Study design. Abbreviations: MTD: Maximum tolerated dose; PD: Pharmacodynamics; PK: Pharmacokinetics. † Including food effects.Baseline characteristics.Abbreviations: ECOG PS: Eastern Cooperative Oncology Group performance status. † Tumor type specified if n ≥ 3 patients. ‡  Other tumor classifications were, in Arm A: Anal region (1), biliary tree (2), bladder (2), gastrointestinal tract (1), lung (2), non-small cell lung cancer (1), not specified (2), pancreas (2), prostate (1), small intestine (1), stomach (1), testis (1); in Arm B: Cervix (1), carcinoma of unknown primary site (1), endometrial cancer (2), genitourinary system (1), gynecologic (1), ureter (1), uterine malignancy (1); in Arm C: Adrenal (1), bladder (1), cervix (2), carcinoma of unknown primary site (1), gynecologic (1), lung (1), pancreas (2), prostate (1), stomach (1).Dose-limiting toxicity (DLT) in cycle 1.Abbreviations: DLT: Dose-limiting toxicity; ALT: Alanine aminotransferase; AST: Aspartate aminotransferase; CP: Creatine phosphokinase; GGT: Gamma-glutamyl transferase; MTD: Maximum tolerated dose. † All DLTs among MTD-evaluable patients were grade 3.Treatment-related adverse events (AEs) occurring in ≥10% of patients in any one treatment arm.Abbreviations: AEs: Adverse events; ALT: Alanine aminotransferase; AST: Aspartate aminotransferase. † One grade 4 event.Best overall tumor response during treatment.† Including an estrogen-receptor-positive breast cancer patient pretreated with a first-line combination of ridaforolimus (mTORC1 inhibitor) + dalotuzumab (anti-IGF1R (insulin-like growth factor 1 receptor)) + exemestane. ‡ Three patients had <3 lines of prior chemotherapy and one patient had ≥3 lines of prior chemotherapy. § Two patients had <3 lines of prior chemotherapy and three patients had ≥3 lines of prior chemotherapy.

Med-MDPI/cancers/cancers-12-06-01426.txt

@@ -0,0 +1 @@
+The natural history of non-optic central nervous system (CNS) tumors in neurofibromatosis type 1 (NF1) is largely unknown. Here, we describe prevalence, clinical presentation, treatment, and outcome of 49 non-optic CNS tumors observed in 35 pediatric patients (0–18 years). Patient- and tumor-related data were recorded. Overall survival (OS) and progression-free survival (PFS) were evaluated. Eighteen patients (51%) harbored an optic pathway glioma (OPG) and eight (23%) had multiple non-optic CNS lesions. The majority of lesions (37/49) were managed with a wait-and-see strategy, with one regression and five reductions observed. Twenty-one lesions (42.9%) required surgical treatment. Five-year OS was 85.3%. Twenty-four patients progressed with a 5-year PFS of 41.4%. Patients with multiple low-grade gliomas progressed earlier and had a lower 5-year PFS than those with one lesion only (14.3% vs. 57.9%), irrespective of OPG co-presence. Non-optic CNS tumors are common in young patients with NF1. Neither age and symptoms at diagnosis nor tumor location influenced time to progression in our series. Patients with multiple lesions tended to have a lower age at onset and to progress earlier, but with a good OS.Neurofibromatosis type 1 (NF1) is a common autosomal-dominant condition with a worldwide prevalence of 1 in 3000 and an estimated incidence of 1 in 2500–3300 [1]. Neurofibromin, the protein encoded by NF1 gene, is a negative regulator of the RAS signal transduction pathway. The haploinsufficiency of NF1 leads to increased cell survival and proliferation mediated by hyperactivation of the Ras/PI3K signaling axis and secondary increased mTOR activation [2]. Both adults and children with NF1 are prone to developing central nervous system (CNS) tumors. Up to 15% of NF1 patients develop a brain tumor within the first two decades of life [3]. The risk of new tumor development was recently estimated to be 0.19% per year of follow-up for each NF1 patient older than 10 years [4].CNS tumors in NF1 children are typically grade 1 pilocytic astrocytoma (PA). The most common CNS tumor is optic pathway glioma (OPG), which is also a hallmark and diagnostic criteria of NF1 [5]. A low-grade glioma (LGG) is usually indolent and is only rarely life-threatening, but may also present more aggressive behavior and be responsible for significant morbidity [6,7]. In contrast, CNS tumors in adults with NF1 tend to be of higher histological grade and have a worse prognosis than those occurring in children [8,9,10]. These observations suggest that the natural history of CNS tumors may differ depending on the age of onset. Despite the abundant body of literature on OPGs in NF1, very few studies have focused on other CNS tumors [3,4,5,8,9,10,11,12,13,14,15,16,17,18,19,20,21].We retrospectively reviewed children and adolescents with NF1 and non-optic pathway CNS tumors followed-up and treated at two centers over a 20-year period in order to report their prevalence, clinical presentation, therapeutic approach, and outcome. We also evaluated overall survival (OS), progression-free survival (PFS), and risk factors for tumor progression.Thirty-five patients were enrolled in the study. Twenty-eight patients were followed at the Neurofibromatosis Referral Center of the “Luigi Vanvitelli” University of Campania, Italy, and seven at Dana Children’s Hospital, Tel Aviv Sourasky Medical Center, Israel. Some of the patients described in the present report were included in other previously published studies [6,22,23,24]. Patient demographics and clinical data are summarized in Table 1. An MRI scan was obtained for 12 patients (34.2%) with symptoms of intracranial hypertension (4 cases); persistent headache (3 cases); neurological signs, including hemiparesis, right-side pyramidal signs, cerebellar signs, absence, and dizziness (4 cases); and endocrine disorder (growth hormone hypersecretion; 1 case). MRI was performed as a follow-up of eight patients with OPG and one with moyamoya syndrome, and to screen 13 patients for OPG and one for intellectual disability (Table 1). In addition, eight out of 31 patients with histological or radiological diagnosis of LGG (25.8%) had multiple metachronous lesions other than OPG: four patients had two lesions, two had three lesions, and two had four lesions (Table 2). Age at diagnosis of CNS tumor was lower in patients with multiple non-optic lesions than in patients with one lesion only (median of 7 years, interquartile range (IQR) of 5.4–10.6, range 3.2–15.8 vs. median 10.8, IQR 9.3–14.6, range 3.8–18, p = 0.048). There was no difference in age at diagnosis of CNS tumor in patients with or without OPG (median 9.5 years, IQR 6.4–14.3, range 3.8–16.2 vs. median 11.1 years, IQR 9.4–14.4, range 3.2–18, p = 0.186).Four patients died (8.6%) at a median age of 14.25 years, IQR 8–18.56, range 6.83–19.08, at a median time from diagnosis of 0.7 years, IQR 0.11–1.81, range 0.01–2.08. Two patients with high-grade glioma (HGG) died because of tumor progression, occurring 4 months and 2.1 years from diagnosis. A third patient with intracerebral malignant peripheral nerve sheath tumor (MPNST) died as a result of post-surgical complications shortly after diagnosis, while a fourth affected by diffuse astrocytoma of corpus callosum died three years after diagnosis from another tumor, a progressive metastatic MPNST involving thoracic nerve roots from T4 to T12.We recorded a total of 49 tumors. Histology was available for 23 lesions (19 patients; Table 1), 18 of which (15 patients) were World Health Organization (WHO) grade 1 gliomas (14 PA and 4 gangliogliomas), one was a WHO grade 2 glioma (low-grade diffuse astrocytoma), two (2 patients) were grade 4 gliomas (glioblastoma), one was an intracranial MPNST, and one was an intracerebral schwannoma (Figure 1). The remaining 26 lesions were considered LGGs based on MRI characteristics.Thirty-six lesions (73.4%) showed enhancement after contrast injection on T1-weighted MRI sequences, while 13 lesions did not show any enhancement in post-gadolinium MRI images.Anatomical distribution of tumors according to histology are reported in Table 3. Most lesions (18, 36.7%) were located in the brainstem. The cerebellum was the second most common tumor location (10, 20.4%), followed by cerebral hemispheres (7, 14.3%), with other locations accounting for the remaining 28.6%. Seven out of 12 symptomatic patients (58.3%) had a brainstem tumor. However, brainstem lesions were more often asymptomatic (11/18). The initial approaches to lesions based on histology and tumor location are summarized in Table 3. A total of 25 lesions (21 patients) were treated at diagnosis or because of clinical or radiological progression. Specifically, 21 out of 49 tumors (42.9%) were treated with surgery (19 patients): six at diagnosis (12.2%), 12 during wait-and-see management (24.5%), and three after initial chemotherapy (6.1%). Thermal ablation of a PA of the right frontal lobe was performed in an 11-year-old girl.The two HGG lesions were treated with surgery, radiotherapy (60 Gy), and temozolamide (75 mg/m2/d). The first patient was a 6-year-old boy with HGG located in the right cerebellar hemisphere, while the second was a 17-year-old girl with HGG in the basal ganglia. The former died four months after diagnosis, while the latter 2.1 years after diagnosis, after having received cranial irradiation for an OPG 14 years previously.Initial approaches adopted for suspected LGG lesions are listed in Table 3. Thirty-seven lesions were treated with a wait-and-see approach for a median time of 5.68 years, IQR 3.26–7.71 (range 0.14–16.27). The evolution of these patients and lesions, together with their outcome, is schematically reported in Figure 2. Specifically, 12 tumors (12 patients) required surgery after an initial observation and were located in cerebellar hemispheres (4), cerebral hemispheres (4), the basal ganglia (1), brainstem (1), corpus callosum (1), and ventricle (1). Five spontaneous reductions were observed after a median time of 2.58 years, IQR 1.49–4.54 (range 0.83–5.00), and one spontaneous resolution occurred 1.16 years after diagnosis (Figure 3).One patient with OPG and multiple brain lesions developed a third PA within a lesion initially considered an unidentified bright object (UBO) of the occipital hemisphere (Figure 4).Three patients required chemotherapy as a first-line treatment because of symptomatic, unresectable tumors [25], with a total of six lesions, three of which were located in the brainstem, two in the thalamus, and one in the right temporal lobe. Patients were treated with vincristine and carboplatin according to International Society of Pediatric Oncology (SIOP) LGG 2004 protocol [25]. Three lesions (2 patients) subsequently required surgery because of progression; a subtotal resection was achieved for a lesion of the temporal lobe, whereas lesions of the brainstem and thalamus were partially removed. All tumors remained stable following surgery.Two lesions (2 patients) were surgically treated at diagnosis. One was a brainstem glioma causing hydrocephalus and the other was located in an anatomically critical site (cerebellar hemisphere). Both patients underwent subtotal resection (STR). The residual brainstem lesion progressed 11 months after surgery and was therefore resected again; the residual disease has remained stable for eight years. The cerebellar lesion progressed three months after the first surgery and was eventually completely resected, with no further relapse for three years.A total of 16 LGGs were treated by surgery, with a curative gross total resection (GTR) performed in eight of them. Of the remaining eight lesions that were not completely removed at the first attempt, three required further surgery and five remained stable.The intracerebral parietal schwannoma was asymptomatic and diagnosed by screening MRI (Figure 1). The lesion was completely removed and did not show any signs of relapse during a follow-up of 6.75 years.The MPNST, arising from the brainstem and occupying the occipital foramen, was surgically treated at diagnosis because of intracranial hypertension and severe brainstem compression. The patient died following a stormy post-operative period complicated by hydrocephalus, swallowing problems, tracheostomy, and cerebrospinal fluid mycotic infection. Whole population 5-year OS was 85.3%, irrespective of tumor histology and treatment (Figure 5). Twenty-four patients out of 35 progressed during follow-up with a 5-year PFS of 41.4% (Figure 5). No significant difference was found in median time to progression for the following binary variables: gender (p = 0.24), age at diagnosis less or more than 10 years (p = 0.92), tumor location (posterior fossa vs. others, p = 0.77; brainstem vs. others, p = 0.60; midline supratentorial vs. others, p = 0.84).When considering LGGs, patients with multiple lesions (outside the optic pathway) progressed earlier than those with one lesion only, with a median time to progression of 1.52 years (95% confidence interval (CI) 0.11–2.94) vs. 5.76 years (95% CI 2.5–9.0), respectively (p = 0.03). Five-year PFS was 14.3% for patients with multiple non-optic lesions and 57.9% for patients with one lesion only (Figure 6). No significant difference was found in median time to progression in patients with and without OPG (4.99 years (95% CI 0–10.23) vs. 3.66 years (95% CI 0–7.37), respectively, p = 0.63). A similar 5-year PFS was found in both groups (42.8% for patients with OPG vs. 48.2% for patients without OPG). None of the LGG patients received chemotherapy for OPG during the observation time.The presence of multiple non-optic LGG lesions proved to be the only independent predictor of progression in the univariate analysis (Hazard Ratio 2.81 (CI 95% 1.06–7.46), p = 0.038), while the other explored variables were not found statistically significant (Table S1).NF1 mutations were identified in 18 patients, of which 17 carried truncating mutations. The mutation type and effect together with the inheritance pattern are reported in Table 4 for each patient included in the study.To the best of our knowledge, our series represents the largest cohort of NF1 children and adolescents affected by extra-optic CNS tumors [4,5,8,12,26,27,28,29,30,31]. In our series, almost half of lesions were incidentally detected (23/49, 46.9%). MRI was requested mainly (13) for OPG screening. Although the efficacy of routine brain MRI screening is still debated in NF1, clinicians should be aware of the opportunity to detect and manage an incidental brain tumor in this population. However, a negative scan does not exclude the possibility that a patient may later develop a brain tumor.In our cohort, overall prognosis was good, with a 5-year OS of 85.3%, in line with previous reports [8]. Nevertheless, despite an excellent survival rate, 5-year PFS was 41.4% (Figure 5). Contrary to the findings of previous studies, neither age and symptoms at diagnosis nor tumor location influenced progression timing in our series [17,32,33,34].In terms of the location of lesions, the posterior fossa was confirmed to be the most commonly affected site, followed by the brainstem (18, 36.7%) and cerebellum (10, 20.4%) [17,26,27,28,29] (Table 3).Our population included a large number of LGGs (91.8% of all lesions). LGGs are in fact the most common CNS tumors in NF1, with PA representing the most prevalent histological subtype [5,30]. We also recorded rarer histologies in our population, including HGGs, MPNST, and schwannoma.We reported a prevalence of 26% of metachronous extra-optic LGGs in our pediatric and adolescent population with NF1 (8/31). In 2017, Sellmer et al. found 20% of patients with multiple non-optic gliomas in their series, including both adults and children [4]. In 2003, Guillamo et al. reported the same percentage of multiple CNS tumors, but they included OPGs [8]. Furthermore, they found that patients with multiple lesions were not affected by higher mortality. In agreement, none of our patients with metachronous lesions died. In our study, patients with multiple lesions had a greater risk of progression and a lower 5-year PFS than patients with a single lesion, whereas the co-presence of OPG was not a risk factor for progression (Table S1). Although nearly all multifocal lesions radiologically worsened, patients tended to remain asymptomatic. These findings support the evidence that multiple CNS tumors, especially LGGs, are a hallmark of NF1 [8,31,35], and even in the case of multiple lesions and radiological progressions, patients could still be asymptomatic [7,34]. Tumor molecular biology may be different in this category of patients. It was in fact speculated that a patient is more likely to develop additional non-optic tumors if they have already developed one [4]. Our patients with multiple non-optic lesions have a lower CNS tumor onset age than those with a single non-optic lesion, strengthening this hypothesis. Here, 13 out of 45 LGG lesions (28.9%) did not show any enhancement on post-contrast MRI images. LGG in NF1 children might be enhanced heterogeneously [36] or not at all [37,38,39]. This observation makes the differentiation between LGGs and UBOs even more challenging. UBOs are non-enhancing T2 hyperintensities typical of NF1 found in the brain and medulla [39,40], which are described as potential precursors of brain tumors [41]. In agreement, a boy with OPG and multiple lesions in our population developed a third PA within a lesion initially considered as a UBO of the occipital hemisphere (Figure 4). The absence of contrast enhancement could, therefore, be misleading in diagnosis of LGG [37].In terms of therapy, 60% of patients in our series received treatment for a CNS tumor. In the case of LGG, chemotherapy was the treatment option for symptomatic children with unresectable lesions [25], while GTR was confirmed curative in eight out of 16 tumors treated by surgery (50%), demonstrating its key role in managing LGG [25,42]. One patient previously treated for OPG in the first year of life and for aqueductal stenosis at the age of 7 developed a PA of the right frontal lobe at the age of 11 and underwent thermal ablation (LITT procedure) of this lesion. Although the long-term effects of laser ablation in patients with NF1 are not yet known, this technique is more precise, less invasive, does not involve ionizing radiations in terms of volume of tissue ablation compared to other methods, and allows complete resection, avoiding collateral damage [43,44,45,46].A wait-and-see approach was adopted as a first-line strategy for almost 75% of patients and lesions in our series. Although one-third of tumors required further surgery, 16% did not progress, 13% showed radiological reduction, and one disappeared (Figure 3). Cases of spontaneous resolution of brain gliomas in NF1 patients are described in the literature [47,48,49,50,51]. Therefore, when managing an asymptomatic lesion without radiological signs of high-grade histology, a wait-and-see policy should be taken [25]. Although NF1-associated gliomas are usually PA, patients with NF1 may present high-grade gliomas, especially in adulthood, with a poor prognosis [52,53,54]. In our series, the two patients with glioblastoma both died, despite surgery and adjuvant chemoradiotherapy [55]. One developed a basal ganglia glioblastoma 14 years after cranial irradiation for OPG. In the past, radiotherapy has been used to treat OPGs [56,57,58], achieving a high rate of tumor control [56,57,59]. It is now avoided in NF1-related LGGs, due to the known risk of secondary malignancies and radiation-induced vasculopathies [25,58,60]. We also recorded the case of an 11-year-old patient with intracerebral MPNST and a 4-year-old with schwannoma (Figure 1), which are both very rare in NF1 patients and exceptionally rare in childhood [61,62,63,64,65,66,67].From the genetic perspective, the number of genotype–phenotype associations in NF1 continues to increase. NF1 gene has one of the highest spontaneous mutation rates and more than 3000 mutations have been recognized to date [68]. The following variants are associated with specific phenotypes: missense variants at the NF1 codons p.Arg1809, p.Met1149, p.Arg1276, and p.Lys1423 and the in-frame deletion c.2970-2972 delAAT [69,70,71]. In our series, 19 out of 35 patients underwent molecular testing for NF1 and an NF1 mutation was identified in 18 cases (Table 4). Seventeen of these mutations were predicted to produce a truncated neurofibromin.NF1 germline mutations were shown to result in different levels of neurofibromin expression in NF1 patient fibroblasts, ranging from 25% to 75% [72]. This finding supports the hypothesis that not all NF1 mutations are equivalent and that residual amounts of functionally active neurofibromin might be linked to the phenotype [73]. Genotype–phenotype associations and the parent-of-origin effect overall or by patient sex were investigated and subsequently excluded for OPGs in NF1 [74,75,76]. It might be interesting to extend such studies and test this hypothesis in brain tumors. Any attempt to associate observed genotypes and brain tumors in our cohort of patients was limited by the very low number of cases examined. Although somatic NF1 gene inactivation is required for NF1-related tumorigenesis [77], we could speculate that: (1) the observed higher rate of truncating mutations in NF1 may suggest that at least 50% loss-of-function is necessary to initiate tumorigenesis; (2) additional genetic modifiers unlinked to the NF1 locus might play a role in brain tumors, as we proposed for NF1-related moyamoya vasculopathy [78].Major limitations of the present study are its retrospective design and the lack of a molecular tumor profile. Our molecular and immunological understanding of NF1-associated gliomas is rapidly evolving, and biopsy is increasingly indicated not only for histologic confirmation, but also for molecular-targeted therapy [42]. In terms of new therapeutic options, The MEK1/2 inhibitor selumetinib seems to be active in NF1-related pediatric LGG [79]. The efficacy of selumitinib versus standard chemotherapy in both newly diagnosed and progressive or recurrent tumors, irrespective of biopsy, is currently under evaluation in a phase III trial (clinical trial.gov number NCT03871257). This retrospective multicentric study involved patients (0–19 years of age) affected by NF1, according to NIH NF1 criteria (1988) [80] and diagnosed with non-optic CNS tumors from January 2000 to January 2019 at two centers. Brain MRI images of all enrolled patients were reviewed. Any measurable area in at least two dimensions of high signal intensity on T2-weighted images was considered a tumor if it had one of the following characteristics: gadolinium enhancement, mass effect, peripheral edema, or a mural nodule associated with a cystic or necrotic component [8,38,39].Histological confirmation of tumors was not mandatory for patient inclusion. Specifically, lesions with typical characteristics of LGGs did not undergo biopsy [7,37,42]. In contrast, histological examination was performed for lesions with radiological features of high-grade histology (high degree of tumor heterogeneity and contrast enhancement, restricted diffusion on diffusion-weighted MRI, and increased relative cerebral blood volume on perfusion-weighted MRI) [36,38,42]. Tumors with aggressive behavior and those requiring surgery were also subjected to histological examination.Collected data included demographic characteristics (gender, age at diagnosis, and inheritance of NF1), age at diagnosis of CNS tumors and NF1, and clinical signs or symptoms at diagnosis of CNS tumors. Indications for brain MRI scan at diagnosis were also recorded. When a patient presented multiple lesions, data on all lesions were collected. We documented the anatomical site, enhancement of all lesions, and histology when available [81], as well as the co-presence of OPG. We also recorded the management of tumors, including wait-and-see, surgery, chemotherapy, and radiotherapy approaches. We collected data on the type of surgical resection (gross total- and subtotal- resection (respectively GTR and STR) [82,83], duration of follow-up, recurrence, progression, or resolution. Ethical approval from local committees was obtained. Continuous non-parametric variables are presented as the median, IQR, and range, whereas categorical variables are expressed as number and percentage. Mann–Whitney U test was used to compare continuous non-parametric variables. Kaplan–Meier analysis was run to determine OS and PFS [84]. OS was calculated from date of diagnosis until death from any cause. PFS was measured from date of diagnosis until radiological or clinical progression date. Specifically, for multiple lesions, the first radiological progression of one lesion was considered as the progression date, irrespective of the evolution of any others. For both OS and PFS analyses, patients were censored at last available follow-up time if no event occurred. Factors that may influence time to progression (gender, age at tumor diagnosis, location of tumor, symptoms at diagnosis) were tested by comparing PFS curves with the log-rank test.Exclusively for patients with histological or radiological diagnosis of LGG, log-rank test was used to compare PFS curves of patients with and without associated OPG, and those of children with single and multiple lesions. The Cox regression model was used to explore predictors of progression in patients with LGG. For all analyses, p values < 0.05 were considered statistically significant. IBM SPSS Statistics 22 Software for Windows was used for statistical analysis.NF1 mutations were recorded in terms of the genomic and protein location and type of mutation. NF1 germline mutations were obtained from a review of clinical records and no further genetic analyses were performed during the study.Extra-OPG CNS tumors are a relatively common malignancy in children and adolescents with NF1. Although other histological types do rarely occur, LGGs are the most common CNS tumors in this population. In our experience, LGG lesions tend to be and remain asymptomatic in young patients, regardless of radiological progression. In addition, patients with multiple metachronous LGGs have a lower age at onset and tend to progress earlier than patients with one lesion only, irrespective of the co-presence of an OPG. Hence, in line with the consensus very recently published by Packer et al. in 2020 [42], our experience confirms that a wait-and-see approach is initially advised for asymptomatic lesions with radiological characteristics of low-grade tumors. Conversely, surgery remains the best therapeutic option in symptomatic LGGs, with the aim of achieving complete resection.The following are available online at https://www.mdpi.com/2072-6694/12/6/1426/s1, Table S1: Univariate Cox regression analysis of PFS in patients with LGGs.Conceptualization, C.S., S.P. (Silverio Perrotta), and G.C.; methodology, F.P. and S.P. (Stefania Picariello); brain MRI reviews and figure preparation, resources, A.D. and M.C.; formal analysis, F.P. and S.P. (Stefania Picariello); patient enrolment and data curation, L.Q., G.G., M.C.M., and U.F.; resources, C.S., D.M., G.C., J.R., and S.C.; original draft preparation, C.S., P.S., D.M., F.P., and S.P. (Stefania Picariello); writing—review and editing, J.R., G.C., and S.C.; supervision, S.C. and G.C. All authors have critically revised and approved the submitted version of the manuscript.The authors declare no conflict of interest.MRI images of the left parietal intracerebral schwannoma: (A) contrasted enhanced T1-weighted image; (B) FLAIR; (C) T2-weighted image.Evolution of the 37 lesions initially managed with a wait-and-see strategy, together with further approaches and outcome. LITT, laser interstitial thermal therapy; GTR, gross total resection; STR, subtotal resection; MPNST, malignant peripheral nerve sheath tumor; StereoRT, stereotactic radiotherapy.MRI images of an enhancing cerebellar lesion located in the right paravermian region (A). The lesion was radiologically followed every 6 months and was undetectable at 14 months of radiological follow-up (B).MRI images (contrast enhanced T1-weighted) demonstrating evolution of an occipital unidentified bright object (UBO) in an expansive lesion in a 9-year-old boy. The lesion appeared as a subcentimetric hypointense non-enhancing area in 2011 (A, arrow). In 2013, the hypointense lesion was surrounded by a ring of contrast enhancement (B). In 2014, the contrast enhancement appearance was more diffuse (C). Six months later, a mixed cystic–solid tumor was detected (D). Post-operative image (E). The tumor histology was confirmed as a pilocytic astrocytoma.Kaplan–Meier whole population overall survival (OS) and progression-free (PS) survival curves.Kaplan–Meier progression-free survival (PFS) curves by non-optic multiple lesions in patients with low-grade gliomas. Please note that just one patient was censured among those with multiple lesions and the relative hyphen is in light grey.Patient demographics and characteristics. Categorical variables are expressed as a number and percentage. Continuous non-parametric variables are reported as the median, interquartile range (IQR), and range. NF1, neurofibromatosis type 1, CNS; central nervous system, MRI; magnetic resonance imaging; OPG, optic pathway glioma; MPNST, malignant peripheral nerve sheath tumor.Tumor-related characteristics and treatments of the eight patients with multiple non-optic tumors. GTR: gross total resection; PA, pilocytic astrocytoma; LITT: laser interstitial thermal therapy; STR: subtotal resection.Brainstem (midbrain)ThalamusChemotherapyGTR of both lesionsBrainstem (midbrain)Right temporal lobeChemotherapyGTR of the right temporal lobe lesionBiopsy of the brainstem lesionBrainstem (midbrain)ThalamusChemotherapyVentriculo-Peritoneal shuntBrainstem (midbrain)Right frontal lobeLITT of the right frontal lobe lesionBrainstem (midbrain)Corpus callosumLeft occipital lobeGTR of the left occipital lobe lesionRight Cerebellar HemisphereLeft paravermianLeft Cerebellar HemisphereSTR and subsequent GTR of the right cerebellar hemisphere lesionGTR of the left paravermian lesionRight cerebellar hemisphereLeft fornixBrainstem (pons)Brainstem (medulla)GTR of the right cerebellar hemisphere lesionCorpus callosumRight cerebral hemisphere (rolandic)Left cerebellar hemisphereBrainstem (midbrain)GTR of the right cerebral hemisphere lesionThe initial approach to lesions based on histology and tumor location. CT, chemotherapy; RT, radiotherapy; HGGs high grade gliomas; MPNST, malignant peripheral sheet tumor; LGGs, low grade gliomas.NF1 molecular findings of all included patients. NF1 reference sequence: NM_000267.3.

Med-MDPI/cancers/cancers-12-06-01427.txt

@@ -0,0 +1 @@
+These authors share equal contribution.These authors share equal senior authorship.Breast cancer is still one of the most common cancers for women. Specified therapeutics are indispensable for optimal treatment. In previous studies, it has been shown that RL2, the recombinant fragment of human κ-Casein, induces cell death in breast cancer cells. However, the molecular mechanisms of RL2-induced cell death remain largely unknown. In this study, mechanisms of RL2-induced cell death in breast cancer cells were systematically investigated. In particular, we demonstrate that RL2 induces loss of mitochondrial membrane potential and cellular ATP loss followed by cell death in breast cancer cells. The mass spectrometry-based screen for RL2 interaction partners identified mitochondrial import protein TOM70 as a target of RL2, which was subsequently validated. Further to this, we show that RL2 is targeted to mitochondria after internalization into the cells, where it can also be found in the dimeric form. The importance of TOM70 and RL2 interaction in RL2-induced reduction in ATP levels was validated by siRNA-induced downregulation of TOM70, resulting in the partial rescue of ATP production. Taken together, this study demonstrates that RL2–TOM70 interaction plays a key role in RL2-mediated cell death and targeting this pathway may provide new therapeutic options for treating breast cancer. Breast cancer is still one of the most common tumorigenic diseases for women. The most common factors associated with breast cancer occurrence are obesity, hormone-associated reproductive factors and hyperplasia of the mammary gland [1]. Surgery, radiation- and chemotherapy remain, until today, the most common approaches for breast cancer treatment. For a more efficient treatment of breast cancer, it is highly necessary to discover novel drugs. To date, various naturally occurring proteins or chemical compounds resulting from pharmacodynamical studies have been used to develop specific drugs for breast cancer [2,3]. Particularly, human milk was found to be a source of multiple bioactive peptides for the treatment of breast cancer. Some of these peptides were recently been utilized to engineer new antitumor drugs such as HAMLET [4], Lactoferrin [5] or Lactaptin [6].Lactaptin is a proteolytic fragment of the human milk protein κ-Casein and comprises its amino acids 57 to 134 [6]. The recombinant analogue of Lactaptin RL2 (recombinant Lactaptin 2) is comprised of the amino acids 23–134 of human κ-Casein (Figure 1A). RL2 was shown to induce cell death in MDA-MB-231 and MCF-7 breast cancer cells [6]. Further to this, RL2 suppresses tumor growth and metastasis in mice [7,8,9]. Additionally, it has been reported that RL2 influences the expression of apoptotic proteins and induces autophagy in MDA-MB-231 cells [8]. Besides its tumor-suppressive actions in breast cancer, it was found that RL2 also acts on other cancer cell types such as endometrial cancer, lung cancer and hepatoma cells [6,10]. Importantly, RL2 was shown to spare normal tissue, such as non-malignant mesenchymal stem cells (MSCs) [6]. RL2 has been reported to exist as a mixture of both monomer and dimer forms, the latter are formed via formation of disulfide bonds [11,12]. However, the detailed molecular mechanisms of RL2 interference with cell death machinery are still not known.There are two ways of apoptosis induction: extrinsic and intrinsic. The extrinsic apoptotic signaling is triggered by ligand binding to the death receptors (DRs), e.g., CD95 (APO-1/Fas) [13,14] or TRAIL-R1/2 [15]. The specific ligand binding to the receptor results in formation of the death inducing signaling complex (DISC) and subsequent activation of the caspase cascade. [15,16,17]. The intrinsic apoptosis pathway is mediated via mitochondria. In particular, mitochondrial outer membrane permeabilization (MOMP) [18] leads to a release of cell death mediators [19,20], activation of effector caspases and apoptosis [21]. The release of other death-inducing factors from mitochondria such as endonuclease G (EndoG) and apoptosis-inducing factor (AIF) might lead to caspase-independent DNA fragmentation and apoptosis.Another important protein complex for mitochondrial signaling is the translocase of outer membrane (TOM) complex [22]. This complex is closely associated with the translocase of inner membrane (TIM) complex and enables mitochondrial import of proteins [23]. The TOM complex consists of multiple proteins such as TOM20, TOM22, TOM40 and TOM70, playing distinct functions [24,25,26]. TOM20 and TOM22 are receptors recognizing their substrates that are transported via a channel formed by TOM40. The TOM70 receptor recognizes a similar set of substrates as TOM20/TOM22, but is also suggested to have distinct functions [27]. In this study, we demonstrate that RL2 induces mitochondrial membrane potential loss, cellular ATP loss and cell death in breast cancer cells. The necrotic morphology of dying cells was observed. Furthermore, we uncovered dimerization processes of RL2 and localized RL2 dimers at mitochondria. The mass spectrometry analysis has further underlined the key role of mitochondria in RL2-induced signaling by identification of potential RL2-targets for cell death mediation including the mitochondrial import protein TOM70. The interaction with TOM70 provides further insights into the connection between RL2 and cell death.RL2 has been reported to induce cell death in breast cancer cells. To uncover the mechanisms of RL2-induced cell death, RL2-mediated signaling in breast cancer cells was systematically investigated. At the first step, it was analyzed whether RL2 is uptaken by cells over time. Breast carcinoma MDA-MB-231 and MCF-7 cells were treated in a time-dependent manner with 200 µg/mL of RL2. RL2 was detected in the cells shortly after stimulation (Figure 1B,C). Notably, an efficient dimerization of RL2 was observed in MDA-MB-231 cells as well as its time-dependent degradation (Figure 1B; Figure S1). The substantial degradation of RL2 was also observed in MCF-7 cells and was already detected after 4 h (Figure 1C; Figure S1). The dimers assemble via formation of disulfide bridges, and therefore, should mostly diminish after SDS-PAGE under reducing conditions [11]. This is in contrast to the analysis of RL2 via SDS-PAGE under non-reducing conditions, in which the formation of the homodimers can be efficiently detected [6]. Hence, apparently, we observe only a residual amount of RL2 dimers in our experiments. The intracellular localization of RL2 was also observed in single cells using Rhodamine-labeled RL2 [8] and Imaging Flow Cytometry in MDA-MB-231 and MCF-7 cell lines (Figure 1D). Taken together, it was shown that RL2 is internalized into the cells shortly after RL2 administration.To investigate whether RL2 treatment of MDA-MB-231 and MCF-7 cells results in a loss of cell viability, these cells were stimulated in a time- and dose-dependent manner with RL2 followed by measuring total cellular ATP amount (Figure 2A,B). MCF-7 cells showed a marginal reduction in ATP levels at 6 and 12 h after RL2 treatment, but a strong reduction after 24–48 h (Figure 2A). Incubation for 6 and 12 h led to the loss of cellular ATP in MDA-MB-231 cells, which was even more prominent 24 h after RL2 treatment (Figure 2B). Interestingly, MDA-MB-231 cells were more sensitive to RL2-induced loss of ATP compared to MCF-7 cells. Consistent with the drop of ATP levels, the cell viabilities of MCF-7 and MDA-MB-231 cells were reduced after RL2 treatment (Figure 2C). These results were in line with the cell death measurements on MDA-MB-231 cells, which were carried out using Propidium Iodide staining and Imaging Flow Cytometry (Figure 2D,E). MDA-MB-231 cells treated for 24 h with RL2 undergo cell death. RL2-treated cells showed a variety of morphologies which could not be correlated to one specific type of cell death (Figure 2E). The dying cells showed features of apoptotic as well as of necrotic cell death. Both, the typical blebbing of apoptotic cells as well as swollen necrotic cells were observed by bright field imaging. The latter had similar morphological features as control cells after heat shock administration. Taken together, it was shown that RL2 is internalized by breast cancer cells and triggers ATP loss and cell death in MCF-7 as well as MDA-MB-231 cells.To identify whether the cell death induced by RL2 indeed has apoptotic features, caspase-3/7 activity assays were carried out. RL2 treatment induces minor caspase-3/7 activity in MDA-MB-231 cells, which was completely blocked by pan-caspase inhibitor zVAD-fmk (Figure 3A). Interestingly, the strength of caspase-3/7 activation did not change over time and remained nearly constant from six to 22 h after RL2 treatment (Figure 3A). A dose-dependent analysis of caspase activity after RL2 stimulation also did not reveal a stronger increase in the amount of caspase-3/7 activity upon the increase in RL2 concentration. Interestingly, the treatment with pan-caspase inhibitor zVAD-fmk could not rescue the ATP loss after RL2 stimulation, indicating that this ATP drop induced by RL2 treatment was not dependent on caspase-3/7 activity (Figure 3B).The results of Western Blot analysis of caspase cleavage in MDA-MB-231 cells were consistent with the results of the caspase-3/7 activity assays. Even though the first cleavage of caspase-3 was monitored already one hour after RL2 stimulation, the amount of caspase-3 cleavage products did not strongly increase over time (Figure 3C; Figure S2). This was also in accordance with the weak increase in caspase-3/7 activity. Only a slight increase in the amount of p19/p17 was observed three hours after stimulation. Accordingly, a rather weak cleavage of caspase-8 was detected with induced processing to p43/p41, but not p18, after three hours of treatment. The processing of Bid was also detected three hours after RL2 stimulation. To compare the strength of caspase activation upon RL2 treatment with a well-established apoptosis induction, the stimulation of MDA-MB-231 cells with 75 ng/mL TRAIL was used (Figure 3C; Figure S2). Interestingly, in TRAIL-stimulated cells, a much more prominent level of caspase-3 processing was observed at earlier time points compared to RL2-stimulated cells. Accordingly, a stronger cleavage of caspase substrates Bid and PARP was detected after TRAIL as compared to RL2-stimulated cells. Moreover, the amount of TRAIL-induced caspase cleavage products increased over time.Taken together, the analysis of caspase activity shows that RL2 induces caspase-3/7 activation in breast cancer cells. However, the loss of ATP levels observed in Figure 3B is not dependent on caspase-3/7 activity. Additionally, in comparison to the induction of the extrinsic apoptotic pathway by TRAIL, the activation strength of effector caspases upon RL2 treatment seems to be relatively weaker.To identify the mediators of RL2-induced cell death, mass spectrometry analysis of the RL2 interactome was carried out in MDA-MB-231 cells. This was performed by immobilization of RL2 on protein A conjugated Sepharose beads followed by RL2-pulldown from MDA-MB-231 cell lysates and subsequent mass spectrometry analysis (‘RL2′-axis (right), Figure 4A). To control the specificity of RL2-pull-down, lysates from MDA-MB-231 cells were incubated with protein A Sepharose beads without prior RL2 immobilization and subsequently analyzed by mass spectrometry (‘control’, Figure 4A). Thereby, we could exclude unspecific off-target hits and search for the most prominent interactors of RL2 in vitro.The mass spectrometry screen with RL2-pulldown reproducibly revealed the mitochondrial import protein TOM70 as one of the most prominent hits (Figure 4B, Table S1). It was detected in RL2-pulldown with high scoring in three independent experiments. Furthermore, it was not found to be associated with the protein A Sepharose beads in the control pulldown in three independent experiments. Tubulin (TBB5, Figure 4B) and actin (ACTN1, Figure 4B), which have already been identified as interaction partners of RL2 in the previous studies [8], were also detected in this analysis with a relatively high score, indicating the correct performance of the entire pull-down assay. Other hits were also present both in the RL2 and control pulldowns, but in a lower abundance concluded by a lower number of identified unique peptides. Hence, we have concluded that TOM70 presents a promising hit in the mass spectrometry screening assay. Furthermore, an increased number of mitochondrial proteins were identified in the RL2-pulldown (Table S2). This list comprises other proteins of mitochondrial import/export such as TOM22, TIM8A and TIM8B detected by mass spectrometry, albeit with a similar or lower scoring than TOM70 (Figure 4B, Tables S1 and S2).The proteomics screening was followed by validation of the interaction of TOM70 with RL2 via pulldown assays in combination with Western Blot analysis. The interaction of RL2 with TOM70 was confirmed in MDA-MB-231 cells (Figure 4C; Figure S3A). Furthermore, Bcl-2 family protein members present at the mitochondria such as tBid were detected in RL2-pulldown (Figure 4C). Finally, the RL2–TOM70 interaction was validated by the so-called reverse approach of TOM70 immunoprecipitation from RL2 treated MDA-MB-231 or MCF-7 cells (Figure 4D,E; Figure S3B,C). These results provide strong evidence on RL2–TOM70 interaction and indicate that RL2 is targeted to the mitochondria after penetration into the cell.To consider possible RL2 localization within the cell, we have analyzed its properties and identified RL2 as a positively charged peptide. It has a pI value of 9.85 and a hydrophobicity of −0.9, which was calculated with the ExPASy ProtParam tool. Hence, it was suggested that it might be targeted to the mitochondrial membrane, and thereby, causes events leading to alterations in ATP production and cell death. To check this hypothesis, a cellular fractionation of RL2-treated MDA-MB-231 cells was carried out. RL2 was prominently detected in the mitochondrial fraction along with the mitochondrial marker SOD2 as well as TOM70 (Figure 5; Figure S4). Remarkably, the dimers of RL2 in the mitochondrial fraction were also observed. Interestingly, RL2-mediated release of caspase-independent mediators of cell death from mitochondria like AIF was not observed upon RL2 treatment (Figure 5; Figure S4). This indicates that these factors might not be involved in RL2-mediated cell death.We have demonstrated that RL2 treatment causes ATP-loss followed by cell death (Figure 1). Because mitochondria are the main source of ATP production, it was suggested that RL2 translocation to mitochondria alters the mitochondrial membrane potential, and thereby, the function of the respiratory chain followed by the loss of ATP production and cell death.Hence, mitochondrial membrane potential was analyzed by quantitative single cell analysis of MDA-MB-231 and MCF-7 cells using our FlowSightTM imaging flow cytometer (Figure 6A,B). Both cell lines were treated for 6 or 24 h with RL2 and then subjected to mitochondrial TMRM staining. Reduced TMRM signals, observed in MDA-MB-231 and MCF-7 cells treated with RL2, correlate with loss of mitochondrial membrane potential in these cells upon RL2 treatment. CCCP treatment was used as a positive control for the loss of mitochondrial potential (Figure 6A,B). Additionally, MDA-MB-231 cells were analyzed by confocal laser scanning microscopy using CellMaskTM Deep red plasma membrane staining dye and mitochondrial TMRM staining. Cells positive for TMRM and plasma membrane staining were analyzed for correlation of both fluorescence signals. Here, loss of TMRM intensity was also observed upon 6 or 24 h of RL2 treatment (Figure 6C,D; Figure S5A–E), confirming the results observed with imaging Flow Cytometry assay in Figure 6A. Taken together, these results demonstrate that RL2 treatment induces loss of mitochondrial membrane potential in MDA-MB-231 and MCF-7 cells (Figure 6A–D; Figure S5A–E).Translocation of RL2 to mitochondria and the impact on the mitochondrial membrane potential supported the role of mitochondria in RL2-induced cell death and the interaction with TOM70 at the mitochondrial membrane. Moreover, the TOM complex has been reported to be crucially involved in ATP production and the function of the respiratory chain as a main macromolecular transport complex at the mitochondrial membrane. To delineate the role of TOM70 in RL2-mediated ATP loss, TOM70 was downregulated in MDA-MB-231 and MCF-7 cells using siRNA (Figure 7C; Figure S6). This was followed by measuring the ATP levels of RL2-treated cells. Importantly, the downregulation of TOM70 partially rescued RL2-induced ATP loss, underlining the importance of TOM70 in RL2-mediated cell death (Figure 7A,B). Taken together, these data indicate that RL2 targets TOM70 at the mitochondria, which might block the function of the TOM complex, leading to the impairment of ATP production that is followed by the demolition of the cell.Contemporary anticancer research strongly requires the establishment of antitumor therapies that are specific for a particular cancer type. In this regard, bioactive peptides, obtained from human milk, are promising antitumor therapeutics for the treatment of breast cancer. One of these peptides is Lactaptin, which is a fragment of proteolytically cleaved κ-Casein [6]. Previous studies have shown that Lactaptin and its recombinant analogue RL2 have strong antitumor effects in breast cancer, endometrial cancer, lung cancer and hepatoma cells [6,10]. Interestingly, there is strong evidence that RL2 specifically affects cancer cells and does not exert any suppressive action on normal cells or non-malignant mesenchymal stem cells [6]. Besides the clinical implications of RL2 in cancer treatment, the molecular mechanisms of RL2-induced antitumor effects remain elusive. Here, further insights into RL2-mediated cell death signaling were delineated, which plays an indispensable role for the follow-up clinical development.Efficient internalization of RL2 by cancer cells provides the essential prerequisite for its action [28]. The positively charged RL2 has a pI value of 9.85 and a hydrophobicity of −0.9, enabling RL2 to pass the plasma membrane of the cells [11,12]. Due to these properties, there seems to be no need for a specific RL2 receptor or an active transport through the cell membrane. Another key feature for the efficient action of the anticancer drug is its bioavailability. Interestingly, it was observed that the degradation of RL2 occurs four to eight hours following the RL2 penetration into the cells. The potential of cell death induction by RL2 could be limited to its relatively short half-life (Figure 1). This is corroborated by the observation that RL2 degradation in MCF-7 cells occurs faster than in MDA-MB-231 cells. Faster RL2 degradation in MCF-7 cells correlates with less ATP downmodulation compared to MDA-MB-231 cells. Thus, reduced bioavailability of RL2 might be the explanation for reduced loss of ATP and cell death of MCF-7 cells. Hence, increasing the bioavailability of RL2 in vivo is an approach which could further enhance the RL2-induced cell death of cancer cells [29,30].In accordance with recent reports, we have observed the dimerization of RL2 in breast cancer cells [11]. In addition, we have found out that RL2 co-localizes with mitochondria in breast cancer cells. The detection of the RL2 dimers at the mitochondria allows us to suggest that dimers of RL2 might have a higher efficiency for binding to TOM70. Subsequently, one can suggest that the dimerization of RL2 might play a crucial role in its binding to TOM70 and cell death induction.According to our results, RL2 is directly targeted to mitochondria. We uncovered that RL2 is localized to the mitochondria by cellular fractionation of MDA-MB-231 cells and further identified mitochondrial outer membrane protein TOM70 as a major target of RL2 (Figure 8). TOM70 was first identified as a receptor and a regulator of transmembrane transit for the precursor of the ADP/ATP carrier (AAC) [22,31,32]. Moreover, we have shown that RL2 treatment induces the loss of mitochondrial membrane potential and thereby might lead to the decreased ATP production, reduced cell viability and increased cell death (Figure 8). The mitochondrial localization of RL2 is consistent with its biochemical properties and previous reports on RL2-mediated MOMP induction in HA-1 hepatoma cells [7]. Additionally, it has been reported that TOM70 serves as a link for Ca2+ transfer from ER to the mitochondria by recruiting IP3 receptors to mitochondria [33]. This leads to inhibition of mitochondrial respiration, induction of autophagy, and inhibition of cell proliferation. Therefore, it might be suggested that RL2 binding to TOM70 impairs Ca2+ transfer between ER and mitochondria, that contributes to cell death [33]. The downregulation of TOM70 rescued the RL2-induced reduction in ATP levels, indicating that TOM70 is one of the potential key players in RL2-induced ATP loss. The ATP levels of unstimulated cells after TOM70 downregulation were not affected (Figure 7). This could be explained by the fact, that siRNA based TOM70 downregulation does not completely block protein expression, and therefore, minor levels of TOM70 remain. TOM complex could potentially function normally without TOM70, which is not a part of the TOM complex core [34]. This and the functional analysis of the RL2–TOM-complex interaction remain unclear and could be further addressed in future studies.Our results provide evidence for and against the involvement of intrinsic apoptosis in RL2-mediated cell death [18,19]. TMRM loss cannot be considered as a direct readout of MOMP. Moreover, we observe rather weak caspase-3/7 activity after RL2 treatment as well as procaspase-3 processing, mostly to caspase-3 cleavage product p19. The latter is a precursor of the active p17 subunit, which indicates the full maturation of caspase-3 and serves as a marker of caspase-3 activation in apoptosis [21]. Moreover, the substrates of caspase-3, PARP1 and BID, were only marginally processed upon RL2 treatment, which differs from the signaling pattern observed upon intrinsic apoptosis induction [21]. Taken together, these results indicate very little contribution of caspase activity to RL2-induced cell death. Moreover, the experiments with zVAD-fmk indicate that RL2 can act in a caspase-independent manner. Considering that RL2 treatment might lead to cell death independent of caspase activity, which is accompanied by the loss of MMP and ATP levels, it might be suggested that RL2 induces one of the programs of regulated necrosis, such as MPT-induced or energy crisis-induced necrosis [35]. Interestingly, imaging flow cytometry has revealed the evidence for both types of cell death: apoptosis and necrosis. In particular, a high number of cells with necrotic morphologies as well as the presence of apoptotic cells after RL2 treatment were observed (Figure 2; Figure 6). Thus, the mechanisms of caspase activity induction, as well as the crosstalk of apoptotic and necrotic cell death, upon RL2 administration, have to be addressed in future studies.Our findings indicate that RL2 treatment downmodulates the intracellular ATP production, possibly leading to the metabolic reprogramming of the cells. Tumor cells have different metabolic properties [35,36] and, accordingly, this might explain the different sensitivity of cancer cells to RL2. Likewise, in this study, we used two breast cancer cell lines: a highly metastatic MDA-MB-231 and a non-metastatic MCF7. The MDA-MB-231 were more sensitive to RL2 treatment compared to MCF-7, which might be explained by their different metabolic status. In this regard, RL2 might be used as a sensitive tool to uncover the metabolic status of a particular cancer cell line in future studies.Furthermore, it would be very interesting to investigate whether a combinatorial treatment based on RL2 in combination with other chemotherapeutics would have a clinical implication and has advantages over conventional breast cancer therapeutics. Moreover, as we have shown that RL2 leads to the metabolic reprogramming of the cancer cells via reducing the ATP production, its applications in low concentrations in combinatorial treatments might provide further therapeutic benefits [35,36].Taken together, we have identified TOM70 as one major target of RL2 at the mitochondrial membrane and suggested that RL2 specifically acts at the mitochondria, leading to the drop of ATP production and cell death (Figure 8). Our results provide new perspectives in targeting mitochondria in breast cancer development and progression via a new RL2-based anticancer treatment.All chemicals were of analytical grade and purchased from AppliChem (Darmstadt, Germany), Carl Roth (Karlsruhe, Germany), Merck (Darmstadt, Germany) or Sigma-Aldrich (Taufkirchen, Germany). RL2 was purified as described previously [6] and applied to cells in 137 mM NaCl. Z-VAD-FMK (N-1510, Bachem, Bubendorf, Switzerland) and recombinant TRAIL (KillerTRAIL™, Enzo Life Sciences, Farmingdale, NY, USA) were applied to cells in indicated concentrations. The following antibodies were used for Western Blot analysis: polyclonal anti-AIF antibody (#5318), polyclonal anti-BID antibody (#2002), polyclonal anti-caspase-3 antibody (#9662), polyclonal anti-PARP antibody (#9542), and polyclonal anti-SOD2 antibody (#13194) from Cell Signaling Technology, Danvers, MA, USA); polyclonal anti-actin antibody (A2103, Sigma-Aldrich, St Louis, MO, USA); polyclonal anti-κ-Casein antibody (#ab111406, Abcam, Cambridge, United Kingdom); monoclonal anti-TOMM70A antibody (sc-390545n Santa Cruz Biotechnology, Dallas, TX, USA); and monoclonal anti-caspase-8 antibody (kindly provided by Prof. P. H. Krammer, DKFZ, Heidelberg, Germany). Horseradish peroxidase-conjugated goat anti-mouse IgG1, IgG2b, goat anti-rabbit and rabbit anti-goat were from Santa Cruz (CA, USA).Human adenocarcinoma cells MDA-MB-231 (purchased: #ACC 732, DSMZ, Braunschweig, Germany) were maintained in Leibovitz L15 media (GibcoTM), supplemented with 10% heat-inactivated fetal calf serum and 1% Penicillin-Streptomycin. Human adenocarcinoma cells MCF-7 (purchased: #ACC 115, DSMZ, Braunschweig, Germany) were maintained in RPMI 1640 Phenol Red (Thermo Fisher Scientific Inc., Waltham, MA, USA), supplemented with 10% heat-inactivated fetal calf serum, 1% Penicillin-Streptomycin, 1 mM sodium pyruvate and 1 x MEM non-essential amino acids in 5% CO2.1.2 × 104 (MDA-MB-231) or 2 × 104 (MCF-7) cells were seeded in 96-well plates. Cells were stimulated in a volume of 50 µL. Measurements were performed according to the manufacturer’s instructions (CellTiter-Glo® Luminescent Cell Viability Assay, Promega, Germany) with the addition of 50 µL CellTiter-Glo® solution to each well. The luminescence intensity was analyzed in duplicate using the microplate reader Infinite M200pro (Tecan, Männedorf, Switzerland). The values were normalized against the viability of untreated cells, which was set as one relative unit (RU).In total, 1.2 × 104 (MDA-MB-231) or 2 × 104 (MCF-7) cells were seeded in 96-well plates. Cells were stimulated in a volume of 100 µL mastermix including 1× concentrated ‘MT Cell Viability Substrate’ and 1× concentrated NanoLuc® luciferase. Measurements were performed according to the manufacturer’s instructions (RealTime-Glo™ MT Cell Viability Assay, Promega, Germany). The luminescence intensity was analyzed in duplicate using the microplate reader Infinite M200pro (Tecan, Männedorf, Switzerland). The values were normalized against the viability of untreated cells that was set as one relative unit (RU).In total, 1.2 × 104 MDA-MB-231 cells were seeded in 96-well plates. Cells were stimulated in a volume of 50 µL. Measurements were performed according to the manufacturer’s instructions (Caspase-Glo® 3/7 Assay, PromegaMadison, WI, USA) with addition of 50 µL of the Caspase-Glo® 3/7 solution to each well. The luminescence intensity was analyzed in duplicate by the microplate reader Infinite M200pro (Tecan, Männedorf, Switzerland). The values were normalized against caspase activity of non-treated cells and set as one relative unit (RU).Analysis of RL2 internalization and cell death induction was performed with FlowSight® Imaging Flow Cytometer (Amnis/Merck Millipore, Darmstadt, Germany). MDA-MB-231 or MCF-7 cells were treated with RL2 or Rhodamine-labelled RL2 [8] for measuring cell death or internalization, respectively. Samples for measuring cell death were additionally stained with propidium iodide (PI). The samples were excited with a 488 nm laser. Emission was detected in channel 4 and bright field images were acquired in channels 1 and 9. For every sample, 10,000 events were recorded. Data were analyzed with IDEAS software version 6.2 (Amnis/Merck Millipore, Darmstadt, Germany). For internalization, Rhodamine-positive cells were taken as RL2-positive cells. RL2-induced cell death was calculated via the percentage of PI positive cells.In total, 1.25 to 2.5 × 105 cells were seeded in 6-well plates. Cells were harvested, washed with PBS and lysed for 30 min on ice in lysis buffer (20 mM Tris HCl, Ph 7.4, 137 mM NaCl, 2 mM EDTA, 10% glycerine, 1% Triton X-100, Protease Inhibitor mix (Roche, Mannheim, Germany)) and subjected to Western blot analysis. SDS-PAGE was performed with 12% SDS gels. The TransBlot Turbo system (Biorad, Hercules, CA, USA) was used to blot the gels to nitrocellulose membranes. Blots were blocked in 5% non-fat dried milk (SantaCruz, Dallas, TX, USA) in PBS with 0.05% Tween 20 for one hour. Washing steps were performed with PBS-Tween threefold for 5 min. Incubation with primary antibodies was performed overnight at 4 °C in PBS-T. HRP-coupled isotype-specific secondary antibodies were incubated for 1 h at room temperature in 5% non-fat dried milk. Chemiluminescence signal was produced with Luminata Forte (Merck Millipore, Darmstadt, Germany) and detected with a ChemiDoc imaging system (Biorad, Hercules, CA, USA). TOM70 immunoprecipitation (IP) was performed with MDA-MB-231 and MCF-7 cells after RL2 stimulation. Stimulation was stopped by adding 10 mL cold PBS and scrubbing of the cells. Cells were centrifuged for 5 min at 500× g and washed once with cold PBS. Cells were lysed in 500 µL lysis buffer for 30 min on ice and subsequently centrifuged for 15 min at 14,600× g. An amount of 50 µL supernatant was used as input control. The remaining supernatant of all samples was adjusted to the same protein concentration and used for immunoprecipitation. TOM70 was immunoprecipitated by mixing 10 µL protein A Sepharose (GE Healthcare, Chicago, IL, USA) and anti-TOM70 antibody (Abcam #ab89624). TOM70 IP samples were rotated overnight at 4 °C, washed three times with lysis buffer and once with PBS. Samples were subjected to Western Blot analysis.In total, 2.5 × 106 MDA-MB-231 cells were stimulated with RL2 or left untreated. Subsequently, cells were trypsinated and washed in PBS. All centrifugation steps in the following fractionation were performed at 17,000× g. A swelling step was performed in swelling buffer (10 mM HEPES Ph 7.6, 10 mM KCl, 2mM MgCl2, 0.1 mM EDTA, Protease Inhibitor mix) for 5 min followed by addition of 0.3% NP-40 (Thermo Fisher Scientific Inc., Waltham, MA, USA) for 1 min. Centrifugation for 1 min resulted in a separation of the cytoplasmic fraction. The remaining pellet was resuspended in 500 µL swelling buffer and centrifuged for 15 s. The pellet was incubated for 30 min with 40 µL ‘nucleus buffer’ (50 mM HEPES pH 7.8, 50 mM KCl, 300 mM NaCl, 0.1 mM EDTA, 10% Glycerol, Protease Inhibitor mix), resuspended every 10 min and centrifuged for 5 min. The nuclei-containing supernatant and mitochondria-containing pellet were separated, and the pellet was washed twice with PBS. The samples were subjected to Western blot analysis.Silencing of TOM70 was performed with TOM70 siRNA (#AM16708, Thermo Fisher Scientific Inc., Waltham, MA, USA) and AllStars Negative Control siRNA (SI03650318, Qiagen, Hilden, Germany). A total of 1.25 × 105 cells was seeded in antibiotic-free medium in 6-well plates and transfected with DharmaFECT1 transfection reagent (T-2001, Dharmacon, Lafayette, CO, USA). Medium was changed after 24 h. Then, 48 h after, the transfection cells were used for Western Blot or cell viability assay analysis.Analysis of RL2-binding partners was performed by protein pull-down. RL2 was covalently coupled to sepharose beads using Pierce™ Co-Immunoprecipitation Kit according to the manufacturer’s instructions (Thermo Fisher Scientific Inc., Waltham, MA, USA) and incubated with total cell lysates of MDA-MB-231 cells. The precipitates were analyzed by nanoLC- tandem mass spectrometry and Western Blot. Sample preparation for mass spectrometry was performed via on-beads digestion. In brief, beads were rehydrated in 50 mM NH4HCO3, pH 8.0, and subsequently incubated with 1 mM DTT at 56 °C for 45 min. Afterwards, reduced cysteine residues were β-methylthiolated via the addition of 5 mM methyl methanethiosulfonate at room temperature for 30 min. Proteins were digested by adding 0.5 μg trypsin (TrypsinGold, Promega, Madison, WI, USA) and incubated at 37 °C overnight. The generated tryptic peptides were eluted from the beads with two washing steps using 50 μL of 25 mM NH4HCO3 for each wash. The washes corresponding to one sample were pooled together and dried in a vacuum centrifuge. The peptides were dissolved in 5 μL 0.1% trifluoroacetic acid (TFA) and purified on ZIP-TIP, C18-nanocolumns (Millipore, Billerica, MA, USA). The peptides were eluted in 7 μL of 70% (v/v) acetonitrile (ACN) and subsequently dried in a vacuum centrifuge. LC-MS/MS was performed on a hybrid dual pressure linear ion trap/orbitrap mass spectrometer (LTQ Orbitrap Velos Pro, Thermo Scientific, San Jose, CA, USA) equipped with an EASY-nLC Ultra HPLC (Thermo Scientific, San Jose, CA, USA). The peptide samples were dissolved in 10 μL of 2% ACN/0.1% TFA and fractionated on a 75 μm I.D., 25 cm PepMap C18-column, packed with 2 μm resin (Dionex, Germany). Separation was achieved by applying a gradient from 2% ACN to 35% ACN in 0.1% formic acid (FA) over a 150 min gradient at a flow rate of 300 nL/min. The LTQ Orbitrap Velos Pro MS exclusively used CID-fragmentation when acquiring MS/MS spectra, consisting of an orbitrap full MS scan followed by up to 15 LTQ MS/MS experiments (TOP15) on the most abundant ions detected in the full MS scan. The essential MS settings were as follows: full MS (FTMS; resolution 60,000; m/z range 400–2000); MS/MS (Linear Trap; minimum signal threshold 500; isolation width 2 Da; dynamic exclusion time setting 30 s; singly charged ions were excluded from selection). Normalized collision energy was set to 35%, and the activation time was set to 10 ms. Raw data processing and protein identification of the high resolution orbitrap datasets were performed with de novo sequencing algorithms of PEAKS Studio 8.0 (Bioinformatics Solutions Inc.,Waterloo, ON, Canada). The false discovery rate was set to <1%.A total of 1 × 106 MDA-MB-231 cells was seeded on poly-L-lysine-coated glass cover slips and cultivated for 24 h. Subsequently, cells were stimulated with 300 µg/mL RL2 for 6 or 24 h. As the positive control for TMRM indicated mitochondrial membrane potential loss, the cells were stimulated with CCCP (Carbonyl cyanide-m-chlorophenyl hydrazone) for 5 min. Sequentially, cells were stained with 20 µM TMRM (Tetramethylrhodamine-methyl ester) (MitoProbe™ TMRM Kit for Flow Cytometry, M20036, Invitrogen, Carlsbad, CA, USA) and 5 mg/mL cell membrane stain (CellMask™ Deep Red Plasma membrane Stain, C10046, Thermo Fisher). Afterwards, cells were placed into the pre-heated incubation chamber of the laser scanning microscope and left for 30 min. Imaging was performed with a confocal laser scanning microscope (LSM700, Zeiss, Jena, Germany). The temperature of the incubation chamber (Pecon, Erbach, Germany) and of the microscope’s objective was adjusted to 37 °C. Images of TMRM fluorescence were obtained using a 63× objective lens with an excitation at 555 nm and an emission of 585–620 nm. Cell Mask Membrane stain fluorescence was excited with laser light of λ = 639 nm and emission was detected at 685–800 nm. Obtained cell images were quantitatively analyzed for changes in mitochondrial membrane potential by correlation of red and green fluorescence signals in each cell. Analysis and calculation of several correlation factors (RWC [37]) was obtained with cell image analysis software (CellProfiler, Stable 3.1.8)For imaging flow cytometry analysis, 6.5 × 105 MDA-MB-231 or MCF-7 cells were treated with RL2 or CCCP (positive control) for indicated time points. After harvesting the cells by trypsination, the cells were stained with 20 µM TMRM (MitoProbe™ TMRM Kit for Flow Cytometry, M20036, Invitrogen). The samples were loaded to the FlowSight® Imaging Flow Cytometer (Amnis/Merck Millipore, Darmstadt, Germany) and excited with a 581 nm laser. Emission was detected in channel 3 and bright field images were acquired in channel 1. For every sample, 10,000 events were recorded. Data were analyzed with IDEAS software version 6.2 (Amnis/Merck Millipore, Darmstadt, Germany). RL2-induced loss of mitochondrial membrane potential was calculated via the percentage of TMRM negative cells.The shown statistical analysis was performed with GraphPad Prism (Version 8.3.0 [38]). The paired Student t-test and one-way ANOVA were performed as indicated. The shown p-values indicate the calculated significance: ns (not significant; p > 0.05), * (significant; p < 0.05), ** (significant; p < 0.01), *** (significant; p < 0.005), **** (significant; p < 0.001).Contemporary anticancer research strongly requires the establishment of antitumor therapies that are specific for a particular cancer type. Previous studies have shown that Lactaptin and especially its recombinant analogue RL2 have strong antitumor effects. In this study, we have identified the cell death network of RL2. In particular, we have identified the mitochondrial protein TOM70 crucial for RL2-induced cell death. Thus, TOM70 is a promising target for combinatorial therapeutic strategies based on RL2-derived peptides obtained in future clinical development.The following are available online at https://www.mdpi.com/2072-6694/12/6/1427/s1, Figure S1: Relative Western Blot quantifications of Figure 1B,C, Figure S2: Relative Western Blot quantifications of Figure 3, Figure S3: Relative Western Blot quantifications of Figure 4A,B,C, Figure S4: Relative Western Blot quantifications of Figure 5, Figure S5: TMRM-indicated mitochondrial membrane potential loss of MDA-MB-231 cells, Figure S6: Relative Western Blot quantifications of Figure 7, Table S1: Absolute values of unique peptides and coverages of potential RL2-interacting proteins identified by mass spectrometry. The analysis of potential RL2-interacting proteins in Figure 4B, are based on the absolute values shown here. The abbreviations n1, n2 and n3 indicate absolute values for every experimental replicate, Table S2: Absolute values of unique peptides and coverages of all mitochondrial proteins identified by mass spectrometry. Abbreviations n1, n2 and n3 indicate absolute values for every experimental replicate.Experiments, M.R. and F.W.; manuscript writing, M.R.; experiments and methodology, T.K., K.S., H.B., O.C. and Y.K.; manuscript writing and supervision, V.A.R., O.A.K. and I.N.L. All authors have read and agreed to the published version of the manuscript.We acknowledge Volkswagen Foundation (VW 90315), Wilhelm Sander-Stiftung (2017.008.01), Center of dynamic systems (CDS), funded by the EU-program ERDF (European Regional Development Fund) and DFG (LA 2386) for supporting our work.Additionally, we thank Yvonne Ducho, Corinna König, Nikita Ivanisenko and Nadine Köhler for supporting our experiments and advice.The authors declare no conflict of interests.MDA-MB-231 and MCF-7 cells internalize RL2 and degrade it in a time dependent manner. (A) Scheme of RL2, Lactaptin and κ-Casein proteins. (B) MDA-MB-231 and (C) MCF-7 cells were treated with 200 µg/mL RL2 for indicated time intervals. Extracellular RL2 was removed by washing during cell harvest. RL2 was detected in the cellular lysates by Western Blot using anti-κ-Casein antibody. One representative Western Blot out of three independent experiments is shown. Quantification was done for three independent experiments. The statistical analysis was performed by one-way ANOVA test. (D) MDA-MB-231 and MCF-7 cells were treated with 50 µg/mL Rhodamine-labelled RL2 for four hours and analyzed with FlowSight® for RL2 internalization. Bottom: At the left column single cells are shown in bright field (BF) channel, at the right column merging of Rhodamine signal and bright field image is performed for representative single cells. Top: The amount of Rhodamine-RL2 positive cells are presented from three independent experiments. The statistical analysis was performed by paired student t-test. Relative Western Blot quantifications of Figure 1B,C are shown in Figure S1. ns (not significant; p > 0.05), ** (significant; p < 0.01), *** (significant; p < 0.005), **** (significant; p < 0.001).Treatment of RL2 promotes loss of ATP levels and cell death in breast cancer cells. (A) MCF-7 and (B) MDA-MB-231 cells were treated with indicated concentrations of RL2 for indicated time intervals. Cellular ATP levels were measured using the Cell Titer-Glo®-Luminescent Cell Viability Assay. ATP levels are presented in relative units (RU) and normalized to the non-treated cells (norm.). (C) MCF-7 and MDA-MB-231 cells were treated with indicated concentrations of RL2 for 24 h. Cell viability was measured using the metabolic RealTime-Glo™ MT Cell Viability Assay. (A–C) Means and standard deviations are shown for three independent experiments. The statistical analysis was performed by one-way ANOVA test. (D,E) MDA-MB-231 cells were treated with 200 µg/mL RL2 for 24 h. Cells were stained with PI and analyzed with FlowSight®. (D) Top: The amount of PI positive cells from three independent experiments are shown in percentage. The statistical analysis was performed by paired student t-test. Bottom: The morphological features of apoptotic and necrotic cell death are shown exemplarily. Apoptotic cell can be characterized by membrane blebbing, cell shrinkage and nucleus fragmentation, whereas necrotic cell shows enlarged cell body and swollen nucleus. (E) Representative pictures of cells are shown. Heat shock treatment serves as positive necrosis control. ns (not significant; p > 0.05), * (significant; p < 0.05), *** (significant; p < 0.005), **** (significant; p < 0.001).RL2 induces marginal caspase-3 activity and BID processing. (A,B) MDA-MB-231 cells were pre-treated for one hour with 50 µM pan-caspase inhibitor zVAD-fmk and stimulated with 200 µg/mL RL2 as indicated. (A) Caspase-3/-7-activity was determined by Caspase-Glo3/7® Assay. Means and standard deviations of caspase activity are shown in relative units (RU) for three independent experiments and statistically analyzed by paired student t-test. (B) ATP levels were measured using the Cell Titer-Glo®-Luminescent Cell Viability Assay. Mean and standard deviation are shown for three independent experiments. The statistical analysis was carried out by paired student t-test. (C) MDA-MB-231 cells were stimulated with 200 µg/mL RL2 or 75 ng/mL TRAIL (positive control) for indicated periods of time, and subjected to Western Blot analysis with the indicated antibodies. One representative Western Blot out of three independent experiments along with its quantification is shown in Figure S2. ns (not significant; p > 0.05), * (significant; p < 0.05).RL2 interacts with mitochondrial membrane protein TOM70. (A) Workflow for RL2-based mass spectrometry analysis. MDA-MB-231 cell lysates were incubated with RL2-coupled Protein A Sepharose beads (RL2-pulldown) or Protein A Sepharose beads (control). Precipitated proteins were identified by nanoLC-tandem mass spectrometry. (B) Mass spectrometry identified unique peptides for each protein. The median of the unique peptides from three independent experiments is shown for each protein and compared for RL2- and control pulldown. The scoring indicates the increase in unique peptides identified. (C) RL2-pulldown precipitates were analyzed by Western Blot and probed for indicated proteins. (D) MDA-MB-231 and (E) MCF-7 cells were treated with 200 µg/mL RL2 as indicated. TOM70 immunoprecipitation (TOM70 IP) was performed with TOMM70A antibody or bead-only control and probed for indicated proteins by Western Blot. One representative experiment out of three is shown for all Western Blots and immunoprecipitations are shown in Figure S3.Internalized RL2 is localized at mitochondria. MDA-MB-231 cells were stimulated with 100 µg/mL RL2 for indicated time points. Cellular fractionation for cytoplasm (‘C’), nucleus (‘N’) and mitochondria (‘M’) was performed. The fractions were subjected to Western Blot analysis with indicated antibodies. SOD2, Actin, GAPDH and PARP were used as fraction control. The bands corresponding to RL2-Monomer and -Dimer were quantified by ImageLab 5.1 beta (Bio-Rad) from three independent experiments. The statistical analysis was performed by paired student t-test (bottom panel). One representative Western Blot out of three is shown in Figure S4. ns (not significant; p > 0.05), ** (significant; p < 0.01).RL2 induces loss of mitochondrial membrane potential. (A–D) MDA-MB-231 and MCF-7 cells were stimulated with 300 µg/mL RL2 for 6 and 24 h. Cells were subsequently stained with 20 µM TMRM. The reduced TMRM signal indicated the loss of mitochondrial membrane potential. CCCP treatment served as positive control for TMRM reduction. (A) Cells were analyzed with FlowSight® for TMRM-signal. Representative cells from three independent experiments are shown. (B) Mean and standard deviations for the amounts of TMRM negative cells (in %) from three independent experiments are shown. The statistical analysis was performed by paired student t-tests. (C,D) RL2-treated MDA-MB-231 cells were stained with 5 mg/mL cell membrane stain (CellMask™ Deep Red Plasma membrane Stain, C10046, Thermo Fisher, Walham, MA, USA) and introduced to confocal laser scanning microscopy. Membrane- and TMRM-stained cells are shown in merge for single cells (C, scale: 10 µM) and cell populations (D, scale: 100 µM). The red color corresponds to the plasma membrane staining. The green (C) and yellow (A,B) colors correspond to TMRM staining. * (significant; p < 0.05), ** (significant; p < 0.01), *** (significant; p < 0.005).TOM70 knock-down rescues RL2-induced drop of ATP levels in MDA-MB-231 and MCF-7 cells. (A) MDA-MB-231 and (B) MCF-7 cells were transfected with TOM70 siRNA (#AM16708, Thermo Fisher) or scrambled control siRNA. Subsequently, cells were stimulated for indicated times and concentrations of RL2. ATP levels were measured using the Cell Titer-Glo®-Luminescent Cell Viability Assay. Means and standard deviations are shown for three independent experiments. The statistical analysis was performed using paired student t-test. (C) The transfection efficiency for TOM70 downregulation was controlled by Western Blot analysis (left). Protein quantification was performed with ImageLab 5.1 beta (Bio-Rad) for three independent experiments. Actin was used as a loading control. The statistical analysis was performed by paired student t-test (right). Relative Western Blot quantifications of Figure 7C are shown in Figure S6. * (significant; p < 0.05), ** (significant; p < 0.01), *** (significant; p < 0.005).RL2-mediated cell death pathways. RL2 translocates into the cytosol and localizes at the mitochondrial membrane by interaction with TOM70. This interaction leads to a drop of mitochondrial membrane potential and reduces ATP production. The latter might be due to the impairment of the TOM complex function, which is required for the transport of the substrates for ATP synthesis or conformational changes at the outer mitochondrial membrane caused by RL2–TOM70 interaction. The reduced ATP levels lead to cell death.

Med-MDPI/cancers/cancers-12-06-01428.txt

@@ -0,0 +1 @@
+Background: Traditionally, the treatment options for unresectable locally advanced (UR-LA) and metastatic (UR-M) pancreatic ductal adenocarcinoma (PDAC) are palliative chemotherapy or chemoradiotherapy. The benefits of surgery for such patients remains unknown. The present study investigated clinical outcomes of patients undergoing conversion surgery (CS) after chemo(radiation)therapy for initially UR-PDAC. Methods: We recruited patients with UR-PDAC who underwent chemo(radiation)therapy for initially UR-PDAC between April 2006 and September 2017. We analyzed resectability of CS, predictive parameters for overall survival, and early recurrence (within six months). Results: A total of 468 patients (108 with UR-LA and 360 with UR-M PDAC) were enrolled in this study, of whom, 17 (15.7%) with UR-LA and 15 (4.2%) with UR-M underwent CS. The median survival time (MST) and five-year survival of patients who underwent CS was 37.2 months and 34%, respectively; significantly better than non-resected patients (nine months and 1%, respectively, p < 0.0001). MST did not differ according to UR-LA or UR-M (50.5 vs. 29.0 months, respectively, p = 0.53). Early recurrence after CS occurred in eight patients (18.8%). Lymph node metastasis, positive washing cytology, large tumor size (>35 mm), and lack of postoperative adjuvant chemotherapy were statistically significant predictive factors for early recurrence. Moreover, the site of pancreatic lesion and administration of postoperative adjuvant chemotherapy were statistically significant prognostic factors for overall survival in the patients undergoing CS. Conclusion: Conversion surgery offers benefits in terms of increase survival for initially UR-PDAC for patients who responded favorably to chemo(radiation)therapy when combined with postoperative adjuvant chemotherapy.In the present-day situation, successful treatment of pancreatic ductal adenocarcinoma (PDAC) remains a therapeutic challenge, and the prognosis is generally poor [1]. Approximately 70% of patients with PDAC are not eligible for surgery, due to locally advanced or metastatic disease at the time of diagnosis [2]. Current guidelines of the National Comprehensive Cancer Network (NCCN) recommend nab-paclitaxel combined with gemcitabine (GnP) or FOLFIRINOX regimens as standard treatments for unresectable (UR) PDAC [3,4]. However, the median survival time (MST) for UR-PDAC remains low (9.2–13.5 months) [5,6,7]. The result of remarkable therapeutic response may occasionally become an indication for conversion surgery (CS) [8,9], which is defined as additional surgery for patients with UR-PDAC who responded favorably to multimodal treatment. However, the incidence and clinical effects are unknown at present. In the present study, we evaluated the clinical outcomes of CS after chemo(radiation)therapy for UR-PDAC, predictive parameters for early recurrence (within six months after CS) and prognostic parameters for overall survival (OS).This retrospective study was conducted by using data from a prospective database. We recruited all consecutive patients undergoing chemotherapy or chemoradiotherapy for UR-PDAC who were to the Department of Surgery, Kansai Medical University, for any treatment between April 2006 and September 2017. All patients were diagnosed with PDAC by cytology or pathology through endoscopic retrograde cholangiopancreatography or endoscopic ultrasound-guided fine-needle aspiration. We have previously reported the details of multidetector-raw computed tomography (MDCT) imaging for the diagnosis of PDAC and to rule out distant metastasis, as well as staging laparoscopy techniques [10,11]. Moreover, multimodal image findings such as contrast-enhanced ultrasonography (CE-US), gadoxetic acid–enhanced magnetic resonance imaging (EOB-MRI), and positron emission tomography (PET) were considered, and we certainly confirmed that all patients had UR-PC initially, according to the National Comprehensive Cancer Network (NCCN) guideline version 2.2017 [3,4].Ethical approval: All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional research committee and with the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards. Informed consent: Written informed consent was obtained from all study participants.The following data were collected: clinicopathological characteristics, type of chemotherapy or chemoradiotherapy, frequency of CS, rates of peri-operative morbidity and mortality, predictive parameters for early recurrence (defined as within 6 months after CS), and prognostic parameters for OS.Data are presented as median (range). Continuous or categorical variables were compared by using the Mann–Whitney U, chi-square, or Fisher’s exact tests as appropriate. The OS and recurrence-free survival curves were estimated by using the Kaplan–Meier method and compared by using the log-rank test. Predictive factors identified by the univariate analysis were further examined by multivariate logistic regression analysis, to determine significant factors for OS and early recurrence among patients undergoing CS. The hazard ratio and 95% confidence intervals were calculated for all estimates. A two-tailed p-value of <0.05 was considered to be statistically significant. Calculations were performed by using JMP software, version 10 (SAS Inc., Cary, NC, USA).Between April 2006 and September 2017, a total of 758 patients received treatment at our department; 290 of those patients underwent surgical resection. The remaining 468 patients with unresectable (UR) PDAC were finally enrolled in this study. Diagnoses were confirmed by using MDCT for 189 patients (40.4%) with unresectable locally advanced (UR-LA) PDAC and 279 (59.6%) with unresectable metastatic (UR-M) PDAC. We performed staging laparoscopy for 133 patients (28.4%) and palliative gastrojejunostomy for 20 patients (4.3%) with radiologically defined locally advanced disease. Positive peritoneal lavage cytology was identified in 30 patients (6.4%), peritoneal dissemination in 25 (5.3%), liver metastasis in 20 (4.3%), and other metastases in six (1.3%). In total, we treated 108 patients (23%) with UR-LA and 360 patients (77%) with UR-M (Figure 1).Baseline characteristics of the study population and regimens that were selected as first-line treatment are listed in Table 1. The most frequently used regimen was gemcitabine (GEM), followed by GEM combined with S-1 and GEM combined with nab-paclitaxel.The standard treatment for advanced pancreatic cancer has changed to gemcitabine since 2001, FOLFIRINOX since 2010, and GnP since 2013 in Japan. Moreover, gemcitabine combined with S-1 was often used as a treatment option. There was liver metastasis in 193 patients, peritoneal metastasis in 123 patients, and LA in 108 patients, respectively. Standardized regimen of chemotherapy in each time has been used in patients with UR-M PDAC. Patients with peritoneal metastasis were treated with S-1 + intravenous and intraperitoneal paclitaxel [12]. Moreover, we have implemented additional radiation therapy in UR-LA patients who still had the low-density area around celiac artery or superior mesenteric artery just before the planned conversion surgery for expecting the margin-negative resection. Positive peritoneal washing cytology was not defined as M1 at that time. Therefore, chemoradiation therapy was implemented for UR-LA with positive cytology.Radiographic partial responses (PR) according to Response Evaluation Criteria in Solid Tumors (RECIST) criteria were observed in 45 patients (42%) with UR-LA and 86 (24%) with UR-M. Stable disease (SD) was observed in 38 patients (35%) with UR-LA and 119 (34%) with UR-M, and disease progression observed in 25 patients (23%) and 155 (42%), respectively. Disease control was achieved in 83 patients (77%) with UR-LA and 205 (58%) with UR-M. Furthermore, patients who could maintain PR or SD for more than eight months were shown in 44 patients (40.7%) with UR-LA and in 85 patients (23.6%) with UR-M, respectively.The major eligibility criteria for surgical exploration were as follows: clinical response (PR/CR) on CT imaging, reduction of tumor markers, fine performance status with patient’s willingness for surgery, and an interval of at least eight months since initial treatment [13]. In patients with peritoneal metastasis, disappearance of occult distant organ metastasis was confirmed by second-look staging laparoscopy in the context of the above criteria. In patients with liver metastasis, a maximum of three occult metastases on the liver surface were resected. In cases where tumor extension to the major vessels with attachment was observed, these patients were indicated for resection. Clinical staging and surgical exploration were re-evaluated at multidisciplinary team meetings.During the study period, 36 patients were planned to undergo CS, and four underwent exploratory laparotomy for occult distant organ metastasis. Finally, CS was performed on 17 patients (15.7%) with UR-LA and 15 (4.2%) with UR-M. Some reasons were raised in 99 patients who had PR but did not undergo conversion surgery due to still UR-LA status on CT imaging and poor performance status. We performed subtotal stomach-preserving pancreaticoduodenectomy for 13 patients (40.6%), distal pancreatectomy for 11 (34.4%), total pancreatectomy for four (12.5%), and distal pancreatectomy with en-bloc celiac axis resection (DP-CAR) on four patients (12.5%) (Table 2). Concomitant CHA resection was done in four patients (12.5%), and concomitant portal vein resection was in 15 patients (46.9%). R0 resection was achieved in 29 patients (90.6%). The median operative time for the total study population was 441 (range 223–866) min, and the median intraoperative blood loss was 1250 (range 207–6301) mL. Although the complication of Clavien–Dindo classification ≥IIIa [14] was reported for eight patients (25.0%), there was no mortality. The median postoperative hospital stay was 14 (range 7–116) days. Histopathologically, Evans grade ≥III was noted in nine patients (28.1%), one of whom exhibited pathological complete response (pCR). The 23 patients (71.9%) received postoperative adjuvant chemotherapy; S-1 was administered to 13 patients (40.6%), GEM to three (9.4%), GEM plus S-1 to one (3.1%), and intraperitoneal infusion and intravenous administration of paclitaxel combined with S-1 to six (18.8%). Twenty-two patients (68.8%) completed adjuvant chemotherapy. The nine patients (28.1%) did not receive postoperative adjuvant chemotherapy, because of our policy of non-adjuvant chemotherapy in the first four patients, patient’s willingness (n = 3), or insufficient nutritional condition (n = 2).The MST of the entire study population was 10 months, and the one- and two-year survival rates were 39% and 12%, respectively (Figure 2). Patients who achieved PR (n = 99) and did not undergo CS exhibited significantly increased survival in comparison with other patients (15 vs. 7.5 months, p < 0.0001; Figure 2). The MST following initial treatment of patients who underwent CS (n = 32) was 37.2 months, and the one-, three-, and five-year survival rates were 100%, 51%, and 34%, respectively. These patients also exhibited significantly increased survival than those who achieved PR (37.2 vs. 18 months, p < 0.0001; Figure 2).When long PR/SD was defined as PR/SD persisting for eight months or more, survival was significantly better among patients who underwent CS compared with those with long PR/SD who did not undergo CS (n = 97) (37.2 vs. 19.5 months, p < 0.0001).Age, gender, tumor location, tumor diameter, tumor markers, pretreatment period to operation, postoperative complications, mortality, and length of hospital stay were not significantly different patients with UR-LA who underwent CS and those with UR-M who underwent CS (Table 2). Significant differences were identified in metastatic site and requirement of additional radiation therapy. There was no significant difference in survival from the time of initial treatment or from the time of CS between patients who underwent CS with UR-LA and those with UR-M (50.5 vs. 29.0 months, p = 0.53; 25.0 vs. 21.0 months, p = 0.61, respectively; Figure 3).The MST from CS was 23 months, and the median recurrence-free survival time was 13 months (Figure 4).Recurrence was confirmed in 20 (62.5%) of 32 patients who underwent CS, presenting as peritoneal dissemination in seven patients, locoregional recurrence in six, liver metastasis in five, and lung metastasis in two. Recurrence within six months after CS was observed in six patients (18.8%), presenting as liver metastasis in three patients, peritoneal dissemination in two, and local recurrence in one. One of those patients received GEM, and four patients received S-1 as adjuvant chemotherapy after CS. After relapse was confirmed, two of the six patients received the same regimen as was administered for initial treatment; these patients survived 23 and 32 months after CS. Patients who suffered recurrence within six months after CS had relatively poorer prognoses than non-recurrent patients (25.5 vs. 50.5 months, p = 0.22). Multivariate logistic regression analyses revealed that lymph node metastasis, washing cytology positive, large tumor (>35 mm), and lack of postoperative adjuvant chemotherapy were predictive factors for early recurrence (Table 3).The multivariate analysis revealed the site of pancreatic lesion and postoperative adjuvant chemotherapy to be statistically significant prognostic factors for OS among patients undergoing CS (p = 0.0092 and p < 0.0001, respectively). Other parameters, including reduction of tumor markers and Evans grading, were not significantly risk factors (Table 4).Despite recent advances in diagnostic medicine, detection of pancreatic cancer while it is within the resectable stage remains a clinical challenge. According to systematic reviews, the condition is not detected until it has reached the locally advanced or metastatic stage in 30–40% and 40–50% of patients, respectively [15,16,17]. Thus, despite the development of chemotherapy, the prognosis of patients with UR-PDAC remains poor, with a median survival of 9.2–13.5 months and low rates of long-term survival. [5,6,7]Favorable outcomes may be achieved for a certain period of time, through the use of chemo(radiation)therapy for patients with unresectable malignancies, and this treatment can be converted to surgical resection, as required. Conversion surgery represents a new therapeutic strategy which may improve short- and long-term outcomes of patients with UR-PDAC. Several articles have reported the utility of CS in such patients, as well as the positive effects on prognosis [17,18,19,20,21,22,23,24,25,26]. In the present study, the rate of CS among patients with UR-LA and UR-M was similar to that reported previously [25]. We found the long-term prognosis; one-, three-, and five-year OS rates from initial treatment; and MST were significantly better among patients with long PR/SD who did not undergo CS, although there were no significant differences in survival with relation to UR-LA or UR-M. Therefore, CS should be considered even for patients initially diagnosed with UR-M if they exhibit surgical indicators. Considering the favorable long-term survival of patients who underwent CS in the present study, our suggestion of tumor extension to the major vessels with attachment as an indication for surgery appears reasonable. However, early recurrence was observed in almost 20% of patients, in line with the findings of Wright et al., who reported that seven out of 23 patients (30.4%) with metastatic PDAC who underwent CS experienced early recurrence. Other studies have also reported early recurrence rates after conversion surgery of approximately 30% [27,28,29]. This would suggest that patients cannot be expected to survive longer than patients who receive non-surgical treatment, and conversion surgery may be harmful to patients because of the high risk of mortality and morbidity associated with extensive pancreatectomy. The early recurrence rate should be decreased as much as possible for patients undergoing CS [30]. Thus, although CS can prolong OS, early recurrence remains a considerable risk. Appropriate preoperative selection of patients for CS is absolutely necessary in order to improve prognosis. The relatively strict surgical indication employed in the present study resulted in prolonged survival and a reduced incidence of early recurrence. In contrast, a review article reported that some authors recommend patients with UR-PDAC who did not experience progression after chemo(radiation) therapy should be offered surgical exploration [30]. The resectability and MST of patients in these studies who underwent CS ranged from 20% to 69% (median, 52%) and from 19.5 to 33 months (median, 21.9 months), respectively. Strict criteria may lead to lower resectability but longer OS, as a result of patient selection. Broad criteria may be associated with higher resectability but shorter OS, due to the risk of early recurrence after conversion surgery. Surgical indications for CS should be carefully decided through discussion in a multidisciplinary meeting.To the best of our knowledge, there have been no previous studies on predictive factors of early recurrence after CS. The present study demonstrates that lymph node metastasis, positive washing cytology, large tumor size (>35 mm), and lack of postoperative adjuvant chemotherapy are significant predictive factors for early recurrence after CS. Thus, tumor size and washing cytology may be important preoperative factors which should be considered during patient selection for CS. Staging laparoscopy should be routinely performed before proceeding with CS in order to exclude patients with positive washing cytology. Metastatic site, decreased CA19-9 level, and performance status are not significant predictive factors for early recurrence. Several articles have reported that decreased CA19-9 levels after multimodal therapy represent a reliable predictive factor for resectability, OS, and DFS [21,29,30,31,32,33,34,35]. In most patients of the present study, CA19-9 decreased to within normal limits after multimodal treatment. Although the optimal selection criteria for surgical exploration or resection remain controversial for patients with initially UR-PDAC, it may be appropriate to base decision-making for CS on clinical response (defined by RECIST criteria) and decreased CA19-9 level after multimodal therapy [30].Regarding pathological examination, the utility of Evans classification reflecting the extent of tumor degeneration or necrosis has been extensively studied as a prognostic factor after preoperative treatment [24,35,36,37,38]. There have been reports of the association between histopathological responses to chemo(radiation)therapy and the prognosis of patients with PDAC [24,35,36,37,38]. Chaterjee et al [36]. reported that 42 (18.8%) of 223 patients with resectable PDAC who received neoadjuvant chemotherapy were classified as Evans grade ≥III and had better survival rates than patients classified as Evans grade <III. Moreover, White et al. [37] suggested histologic response to be a useful surrogate marker for treatment efficacy, but Evans grade was not found to be a prognostic factor of CS in the present study.The present study has some limitations which should be acknowledged. Firstly, it is a single-institute and retrospective study involving a small number of patients. All studies on this subject, to date, are retrospective studies, and so we believe that a prospective study is necessary to define the efficacy of CS. In Japan, the results of the PREP-04 trial (UMIN000017793)—a multi-institutional prospective cohort study investigating clinical outcomes of CS on patients with initially UR-PDAC— will be published in the near future. Given that only patients who responded favorably to chemo(radiation)therapy were analyzed among all patients with UR-PDAC, a selection bias exists. The development of an effective therapeutic strategy involving combined multimodal treatment with surgical resection is critical.In conclusion, CS can provide clinical benefits, including increased survival for patients with initially UR-PDAC who have responded favorably to chemo(radiation)therapy. In addition to CS, postoperative adjuvant chemotherapy is necessary to prolong survival. It is essential that efforts are made to reduce early recurrence and to investigate surrogate markers in order to determine appropriate indications for surgery.Data curation, S.Y., S.H., M.K., and H.R.; methodology, M.I.; supervision, Y.M. and M.S.; writing—original draft, H.Y.; writing—review and editing, S.S., T.Y., and M.S. All authors have read and agreed to the published version of the manuscript.No funding was received for this work.All authors declare that they have no conflict of interest. Human and animal rights: This article does not contain any studies with animals performed by any of the authors.Study flow diagram. Diagnoses were made, using multidetector-raw computed tomography. In total, 189 patients were diagnosed with unresectable locally advanced pancreatic ductal adenocarcinoma (UR-LA PDAC), and 279 patients were diagnosed with unresectable metastatic (UR-M) PDAC. We performed staging laparoscopy for 133 patients with radiologically defined locally advanced disease. We finally enrolled 108 patients with UR-LA PDAC and 360 patients with UR-M PDAC in the present study. Abbreviations: KMU, Kansai Medical University; MDCT, multidetector-raw computed tomography; UR-LA, unresectable locally advanced; UR-M, unresectable metastatic.Overall survival of all patients, patients with radiographic partial response, and patients who underwent conversion surgery. The median overall survival (OS) for the study population (solid line, n = 468) was 10 months. Survival was significantly better among patients with partial response (dashed line, n = 99) compared with other cases (p < 0.0001). Survival of patients who underwent conversion surgery (dotted line, n = 32) was significantly better than those with partial response (p < 0.0001). Abbreviations: CS, conversion surgery; Pts, patients; PR, partial response.Overall survival of patients with unresectable locally advanced or metastatic disease who underwent conversion surgery. There was no significant difference in overall survival between patients with unresectable locally advanced (solid line, n = 17) and unresectable-metastatic disease (dashed line, n = 15) (p = 0.53). Abbreviations: UR-LA, unresectable locally advanced; UR-M, unresectable metastatic.Recurrence-free survival of patients who underwent conversion surgery. The median recurrence-free survival time of patients who underwent conversion surgery was 13 months. Abbreviations: CS, conversion surgery.Baseline patient characteristics.UR-LA: unresectable locally advanced pancreatic cancer, UR-M: metastatic pancreatic cancer, PS: performance status, GEM: gemcitabine, GS: S-1 combined with gemcitabine, GnP: nab-paclitaxel combined with gemcitabine, PTX: paclitaxel.Patient characteristics of conversion surgery.Ph: pancreas head, Pbt: pancreas body and tail, Mets: metastasis, L: liver, P: peritoneum, GEM: gemcitabine, GS: S-1 combined with gemcitabine, GnP: nab-paclitaxel combined with gemcitabine, PTX: paclitaxel, RT: radiation, PD: pancreaticoduodenectomy, DP: distal pancreatectomy, DP-CAR: distal pancreatectomy with en-bloc celiac axis resection, CHA: common hepatic artery, CA: celiac artery, PV: portal vein.Predictive factor for the recurrence within six months after CS (Univariate and multivariate logistic regression analyses).CS: conversion surgery, HR: hazard ratio, CI: confidential interval, LN: lymph node, R: residual tumor, CY: washing cytology, Tx: chemotherapy.Univariate and multivariate analysis of prognostic factor of overall survival in CS group.CS: conversion surgery, HR: hazard ratio, CI: confidential interval, LN: lymph node, R: residual tumor, CY: washing cytology, Tx: chemotherapy.

Med-MDPI/cancers/cancers-12-06-01429.txt

@@ -0,0 +1 @@
+Background: Predictive biomarkers of response to chemotherapy plus antiangiogenic for metastatic colorectal cancer (mCRC) are lacking. The objective of this study was to test the prognostic role of splenomegaly on baseline CT scan. Methods: This study is a sub-study of PRODIGE-9 study, which included 488 mCRC patients treated by 5-fluorouracil, leucovorin and irinotecan (FOLFIRI) and bevacizumab in first line. The association between splenic volume, and PFS and OS was evaluated by univariate and multivariable Cox analyses. The relation between circulating monocytic Myeloid derived suppressor cells (mMDSC) and splenomegaly was also determined. Results: Baseline splenic volume > 180 mL was associated with poor PFS (median PFS = 9.2 versus 11.1 months; log-rank p = 0.0125), but was not statistically associated with OS (median OS = 22.6 versus 28.5 months; log-rank p = 0.1643). The increase in splenic volume at 3 months had no impact on PFS (HR 0.928; log-rank p = 0.56) or on OS (HR 0.843; log-rank p = 0.21). Baseline splenic volume was positively correlated with the level of baseline circulating mMDSC (r = 0.48, p-value = 0.031). Conclusion: Baseline splenomegaly is a prognostic biomarker in patients with mCRC treated with FOLFIRI and bevacizumab, and a surrogate marker of MDSC accumulation.Colorectal cancer (CRC) is the second most commonly diagnosed cancer in Europe and a leading cause of death both in Europe and worldwide [1,2]. In 2012, there were 447,000 new cases of CRC in Europe, with 215,000 deaths, and worldwide there were 1.4 million new cases, with 694,000 deaths. About 20–25% of patients with CRC present with metastatic disease at the time of diagnosis, and a further 20–25% of patients will develop metastases later [3]. For metastatic CRC (mCRC), the typical chemotherapy backbone comprises a fluoropyrimidine (intravenous 5-FU or oral capecitabine) used in various combinations and schedules with irinotecan or oxaliplatin [4]. The VEGF antibody (bevacizumab) and EGFR antibodies (cetuximab and panitumumab) are frequently used in combination with chemotherapy [5]. Currently, the classical first line chemotherapies consist in bi-chemotherapy, comprising an injection of irinotecan, fluorouracil and leucovorin (FOLFIRI) or the combination of an injection of oxaliplatin, fluorouracil and leucovorin (FOLFOX) with a biotherapy targeting VEGF or EGFR. More recently, tri-chemotherapy has come to be used in first line for aggressive disease [6,7]. The concept of treatment discontinuation has also recently been introduced, and after “induction” treatment, an active maintenance treatment is seen as a possible option [8,9]. Today, the median overall survival for patients with mCRC is about 30 months [5]. In mCRC, predictive and prognostic biomarkers are used to guide the therapeutic strategy. RAS mutational status predicts the absence of efficacy of EGFR antibody therapies in mCRC [10,11]. BRAF mutations are a significant negative prognostic marker [12], and are predictive of the efficacy of combined EGFR, BRAF and MEK tyrosine kinase inhibition [13,14]. Microsatellite instability (MSI) is another negative prognostic marker in the metastatic setting [15] and a strong predictive marker of the efficacy of immune checkpoint inhibitors [16]. Dihydropyrimidine dehydrogenase (DPD) activity has been shown to predict potential toxicity when using 5-FU and capecitabine [17], while UDP glucuronosyltransferase 1 family polypeptide A1 (UGT1A1) gene polymorphisms are predictive of irinotecan-related side-effects (diarrhea, neutropenia and vomiting) [18,19]. Tumor sidedness and consensus molecular classification could also be used to predict prognostic and response to biochemotherapies [20,21,22,23,24]. All these biomarkers are used, in combination with clinical markers such as the patient’s performance status, tumor burden and comorbidity, to stratify patients and determine the optimal therapeutic strategy.Currently, biomarkers to predict response to the combination chemotherapy FOLFIRI plus bevacizumab in mCRC and help with clinical decision-making are needed. The factors associate with shorter overall survival (OS) in PRODIGE 9 trial were WHO performance status > 2, unresected primary tumor, age over 65 were and BRAF mutant tumor [25]. Another analysis revealed that high baseline leukocytes count and the lack of carcino-embryonic antigen (CEA) decrease level at first evaluation were associated with early progression [26]. Moreover, a radiomic signature at baseline and 2-month CT was able to predict OS [27]. Myeloid-derived suppressor cells (MDSC) can support tumor progression and have been shown to accumulate in the blood and peripheral lymphoid organs, such as the spleen, in animal models of cancer, leading to splenomegaly [28]. We and others described an accumulation of circulating MDSC in patients with mCRC [29,30,31] and in those with pancreatic cancer [32,33,34] as compared to healthy donors. A high level of circulating MDSC at baseline is significantly associated with poor progression-free survival (PFS) and poor OS in mCRC and pancreatic cancer [29,33,34,35]. In addition, our group [36] observed that baseline splenomegaly is a predictive marker of poor response to FOLFIRINOX in advanced pancreatic carcinoma. Together, these data suggest that MDSC level could be a surrogate marker associated with better PFS. The assessment of circulating MDSC levels is not routinely performed, but splenomegaly could be a surrogate marker of MDSC levels.In this prospective cohort study based on the PRODIGE 9 population, we first aimed to determine the prognostic role of baseline splenomegaly and chemotherapy-induced splenomegaly in mCRC patients treated with first line FOLFIRI. Secondly, we aimed to determine whether splenic volume is correlated with the rate of circulating MDSC.Out of the 488 patients included in the modified intention-to-treat population of the PRODIGE-9 study, 266 with available CT scan and written informed consent for sub-studies were eligible for the present analysis. Of these, 14 were excluded because their CT scan was not amenable to measurement of the spleen because of technical problems or splenectomy (Figure 1). Two hundred and fifty-two patients were thus included in the present analysis. The characteristics of the cohort were representative of those of the PRODIGE-9 patients (Table 1). Only 55 patients (21.8%) had received adjuvant chemotherapy for localized CRC before inclusion in PRODIGE-9.Overall, at the last follow-up, 248 patients progressed or died; only 4 patients did not progress and were censored. We first analyzed baseline splenic volume as a continuous variable. Univariate Cox analyses identified nine factors significantly (p < 0.05) associated with PFS (Table 2), including baseline splenic volume (HR 1.001; 95CI% [1–1.002]; log-rank p = 0.05). Multivariate Cox analyses identified four significant prognostic factors for PFS (Table 2), including baseline splenic volume (HR 1.001; 95%CI [1.000–1.0003]; log-rank p = 0.01). Next, we analyzed baseline splenic volume as a binary variable. Using the median of baseline splenic volume (214 mL) as the cut-off, no difference in PFS was demonstrated. We then sought to identify the threshold with the best discriminatory power, using Cutoff Finder software. The baseline splenic volume threshold retained was 180 mL with a sensitivity of 67.3% (61.3–72.9%) and a specificity of 50% (15–85%). Using this threshold in univariate Cox analysis, baseline splenic volume was identified as a prognostic factor for PFS (HR 1.403; 95%CI [1.073–1.834]; log-rank p = 0.01). By multivariate Cox analysis, the baseline splenic volume threshold of 180 mL was found to be significantly associated with PFS (HR 1.362; 95%CI bootstrapped [1.338–1.370]; p = 0.0249) (Table 3), and the stability of the model was verified using a bootstrap with 500 replications. Harrell’s C-index for the baseline splenic volume threshold at 180 mL was 0.84 (95%CI [0.75–0.93]). In our cohort, baseline splenic volume was ≤180 mL in 83 patients (32.9%), and > 180 mL in 169 patients (67.1%). The Kaplan–Meier curves showed that baseline splenic volume > 180 mL was associated with poor PFS (median PFS = 9.2 vs. 11.1 months, >180 mL vs. ≤180 mL respectively; log-rank p = 0.0125) (Figure 2A).In our cohort, 216 patients died. We analyzed baseline splenic volume as a binary variable, and tested the previously identified threshold of 180 mL. Using this threshold, baseline splenic volume was not significantly associated with OS by univariate Cox analyses (HR 1.223; 95%CI [0.920–1.624]; log-rank p = 0.1654) or by multivariate Cox analyses (HR 1.094; 95%CI [0.819–1.461]; log-rank p = 0.5419) (Table S1). Baseline splenic volume > 180 mL was not statistically significantly associated with OS (median OS = 22.6 vs. 28.5 months, >180 mL vs. ≤180 mL respectively; log-rank p = 0.1643) (Figure 2B). Association between the splenic volume at 3 months and patient’s prognosis in our cohort, after 3 months of treatment, we observed a decrease in splenic volume in 110 patients (43.8%), and an increase in 141 patients (56.2%). The increase in the splenic volume at 3 months was not associated with PFS (HR 0.928; log-rank p = 0.56) nor OS (HR 0.843; log-rank p = 0.21).In a cohort of 19 still untreated mCRC patients from another cohort (clinical trial MEDITREME), we evaluated splenic volume at baseline, as well as the levels of various blood immune cell populations. We used 10 age- and sex- matched healthy volunteers as controls. We observed that patients with mCRC had baseline higher myeloid cell counts, and lower lymphoid cell counts than healthy controls (Figure 3A,B). Regarding the myeloid subset, mCRC patients had significantly higher levels of granulocytes, monocytic MDSC (mMDSC) and lower levels of activated monocytes than healthy donors (Figure 3C). Baseline splenic volume was positively correlated with the level of baseline circulating neutrophils and mMDSC (Figure 3D,E). Moreover, baseline splenic volume was inversely correlated with the number of lymphocytes and activated monocytes (Figure 3F,G), but not with other immune parameters, thus suggesting that splenic volume is a surrogate marker of an unfavourable peripheral immune response with accumulation of immunosuppressive myeloid cells.This study underlines that splenic volume could be a predictive marker of PFS in patients treated with FOLFIRI plus bevacizumab. As splenic volume appears to be a surrogate marker of mMDSC levels, this study raises the hypothesis that mMDSC could modulate the efficacy of FOLFIRI plus bevacizumab in mCRC.In line with other studies [3,37,38], we observed that Köhne criteria, number of metastatic sites, resection of primary tumor, and baseline level of alcaline phosphatase, platelets, leukocytes, and LDH were prognostic factors for survival. Previous ancillary studies in PRODIGE9 study confirmed the poor prognostic role of high leukocytes count and show that lack of CEA decrease is also an important factor of poor prognosis [26]. Imaging-based predictive factor was also tested in PRODIGE 9 cohort and a radiomic signature at baseline and 2-month CT was able to predict OS and identify good responders better than RECIST1.1 [27]. However, our study is the first to observed that splenic volume is associated with PFS in mCRC, and it confirms our previous findings in patients treated with FOLFIRINOX for advanced pancreatic cancer [36]. Our work suggests that baseline splenic volume, used as either a continuous or binary variable, could be a new prognostic marker of FOLFIRI plus bevacizumab efficacy in first-line treatment for mCRC, insofar as we show that baseline splenic volume >180 mL is associated with poor PFS. Additional studies are required to validate this cut-off, and also to determine whether it is a new prognostic marker or predictor of the efficacy of FOLFIRI plus bevacizumab. We expect that new studies in patients treated with anti-EGFR or FOLFOX-based chemotherapy will improve the comprehension of our results. Since normal splenic volume ranges between 110 and 340 mL [39], we cannot rule out the possibility that a splenic volume threshold of 180 mL as selected in this study may be suboptimal, and this warrants confirmation in another study. In addition, the cut off determine is this study differ from our study in pancreatic cancer which select a cut off of 340 mL. We believe that such data could be explained by an anatomic and a biological hypothesis. First, in pancreatic cancer, many patients have portal hypertension which could lead to splenic enlargement. Secondly neutrophilia and accumulation of MDSC is more frequent in pancreatic disease than colorectal cancer. Abnormal differentiation and function of myeloid cells is a hallmark of cancer. MDSC can support tumor progression by promoting tumor cell survival, angiogenesis, invasion of healthy tissue by tumor cells, and metastases [40]. MDSC are a heterogeneous population of myeloid cells with either a granulocyte or monocytic phenotype and characterized by their capacity to exert immunosuppressive functions [41]. They are also suspected of exerting their deleterious effect via their capacity to blunt T dependent antitumor immune response [42]. In tumor-bearing hosts, MDSC accumulate in peripheral lymphoid organs, such as the spleen and tumor tissues [43]. Marked splenomegaly related to MDSC accumulation is a classical observation in many experimental models of transplantable cancer in MDSCs. In humans, MDSC correlates with clinical outcomes and is an independent prognostic indicator of clinical disease progression in patients with pancreatic cancer, esophageal cancer, gastric cancer, and melanoma [34,43]. In mCRC, we previously demonstrated that there is an accumulation of circulating MDSC compared to healthy donors, and high levels of circulating MDSC at baseline are associated with poor PFS and OS [29]. Moreover, in mCRC, treatment with FOLFOX plus bevacizumab induces a decrease in MDSC, which in turn is associated with better PFS [29]. Likewise, Tada et al. showed that mCRC patients with high mMDSC and low blood T cell levels had significantly shorter PFS [35]. Our results are consistent with previous studies, as we observed a higher rate of circulating MDSC in patients with mCRC than in healthy donors. The circulating MDSC level could represent an early marker of disease progression [44], but analysis of MDSC levels is complex and not yet possible in routine. In animal models, MDSC accumulation is associated with splenomegaly [28]. To the best of our knowledge, this is the first time that a positive correlation has been demonstrated between splenic volume and circulating MSDC levels in humans. Thus, we believe that baseline splenic volume could be used as a surrogate marker of MDSC accumulation. As MDSC level is associated with poor prognosis in mCRC, cancer therapy targeting this immunosuppressive cell type could be of therapeutic interest. Inhibitors of CXCR4, CCL2 and VEGF are known to affect MDSC levels and are currently in development to combat MDSC dependent immunosuppression [45]. Splenomegaly may be a surrogate marker of the efficacy of such therapies.Our study has several strengths. We evaluated a large, multicenter cohort of patients with mCRC generated from a multicenter phase III trial. Thus, clinical patient characteristics were homogeneous and limit potential for bias. In addition, bootstrapping enabled us to assign measures of accuracy and internal validation of the prognostic role of spleen volume (defined in terms of bias, variance, confidence intervals, and prediction error). However, a validation cohort in another prospective cohort of patients treated with FOLFIRI bevacizumab before using spleen volume as a prognostic marker in mCRC. In addition, further studies with different treatment strategies are warranted to determine the prognostic versus predictive role of this new marker.PRODIGE 9 was a cooperative, multicenter, open-label, phase III randomized controlled trial, conducted by the Fédération Francophone de Cancérologie Digestive and the PRODIGE (Partenariat de Recherche en Oncologie DIGEstive) group in 66 French centers [25,46]. From March 2010 to July 2013, 491 patients were included. Main eligible criteria were histologically proven, unresectable mCRC, WHO status ≤ 2, life-expectancy ≥ 3 months, and absence of previous chemotherapy or antiangiogenic therapy for metastatic disease [25]. All patients provided written informed consent, and the study was approved by the Committee for the Protection of Persons Ile de France VIII. The trial was registered on ClinicalTrials.gov (NCT00952029). All patients were treated with FOLFIRI and bevacizumab for 6 months. After this induction treatment the patients were randomized in a bevacizumab maintenance versus no treatment (observation) arm during the chemotherapy free interval (CFI). After disease progression during the maintenance or pause, the induction regimen was repeated for eight cycles, followed by a new CFI. In the present cohort, we included all PRODIGE 9-patients for whom computed tomography (CT) scan was available for central review at baseline, 3 months and 6 months. Only patients with available CT scan and who provided written informed consent for substudies were included in the present analysis. Patients with splenectomy were excluded. To investigate the correlation between MDSC level and splenomegaly, we used a subset of patients include in another first line clinical trial MEDITREME (NCT03202758). Blood from age- and sex-matched healthy donors was obtained from Fench Blood Transfusion Service.Spleen volume was measured by CT scan as previously described [46]. The maximal width (W) of the spleen, determined as the largest diameter on any transverse section, the thickness at the hilum (Th), determined as the distance between the inner and outer borders of the spleen on a plane perpendicular to the splenic width and through the hilum, and length (L) were obtained from abdominal CT examinations. Spleen volume was calculated using the formula: Spleen volume = 30 + 0.58 (W × L × Th.). A value between 110 and 340 mL is considered normal [46]. Spleen measurements and volume calculations were performed by the same investigator, who was blinded to clinical outcomes. Splenomegaly was defined as spleen volume > 340 mL.Myeloid subpopulations were identified using custom Duraclone tubes (Beckman Coulter), using anti-CD15-Pacific Blue, anti-CD33-FITC, anti-CD3-PC5.5, anti-CD19-PC5.5, anti-CD20-PC5.5, anti-CD56-PC5.5, anti-HLA-DR-AA750 and anti-CD14-Krome Orange dry antibodies (all obtained from Beckman Coulter). We added in liquid form anti-CD11b-BV605 antibody (Biolegend) and Draq7 viability dye (Beckman Coulter). Whole blood was removed to heparinized tubes (100 µL) and stained in Duraclone tubes for 15 min at room temperature. After surface staining, 2 mL of Versa Lyse/IOTest 3 solution (Beckman Coulter) was added for 10 min. Cells were then centrifuged at 350g for 5 min). All events were acquired on a BD CANTO2 cytometer equipped with BD FACSDiva software (BD Pharmingen, Le Pont De Claix, France), and data were analyzed using Kaluza software (Beckman Coulter). The gating strategy used to analyze myeloid cell subsets has previously been described elsewhere [29]. Briefly, singlet live blood CD11b+ Lineage− leukocytes were considered as myeloid cells and CD11b− Lineage+ as total lymphocytes. Then, we identified granulocytes as CD15+ CD33−, granulocytic MDSC (gMDSC) as CD15+ CD33+, monocytic MDSC (mMDSC) as CD15− CD33+ CD14+ HLA-DRlow/neg and mature monocytes as CD15− CD33+ CD14+ HLA-DRhigh. For HLA-DR positivity identification, we used FMO controls. The gating strategy is presented in the supplementary data (Figure S1).The primary objective was to determine whether baseline splenic volume is an independent prognostic factor of PFS in patients with mCRC treated by FOLFIRI plus bevacizumab, and if so, to determine the cut-off of splenic volume associated with PFS.Secondary objectives were to determine whether baseline splenic volume is a prognostic factor of OS in patients with mCRC treated by FOLFIRI plus bevacizumab in first line; whether splenic volume size modification at 3 months was associated with PFS or OS, and to examine the correlation between splenic volume and circulating MDSC levels.Categorical variables are described as number and percentage, excluding missing data. Continuous variables are described as median and range. Median follow-up was calculated according to the reverse Kaplan–Meier method. The response to treatment was determined on CT scan using RECIST version 1.1 criteria. PFS was defined as the time between randomization and first progression or death from any cause. OS was defined as the time between randomization and death from any cause. Patients without event were censored at the date of last follow-up. Survival curves and median survival (with 95% confidence interval) were estimated using the Kaplan–Meier method. The log rank test was used to compare survival curves. Baseline splenic volume was analyzed as a continuous variable and its prognostic role on OS and PFS was determined by univariate and multivariate Cox models. Variables with a p-value < 0.10 by univariate analysis were entered into the multivariate model. The multivariate model was adjusted for spleen volume at baseline. The correlation between variables included in the model was tested, and if the correlation coefficient between two variables was >0.5, only the more clinically relevant of the two, or the variable with the most significant Hazard Ratio (HR) was entered into the model. In case of non-compliance with the log linearity condition of the variable “spleen volume”, a logarithmic transformation could be used. A bootstrap approach with 500 replications was used to check for the stability of the model. The predictive power of the model was assessed using Harrell’s C index.Analyses were repeated including the variable “splenic volume” as a binary variable. The threshold was determined using the best cut-off method using Cutoff Finder software (Charité-Universitätsmedizin Berlin, Berlin Germany) [47] and analyzed in a multivariate model.Correlations between blood immune parameters and spleen volume were performed using the non-parametric Spearman correlation test.Tests were two-sided and a p-value < 0.05 was considered significant. All analyses were performed using SAS software version 9.4 (SAS Institute Inc., Cary, NC, USA).In conclusion, we propose that baseline splenomegaly could be a prognostic marker of FOLFIRI plus bevacizumab efficacy in mCRC, and a surrogate marker of MDSC accumulation. These results warrant confirmation in other studies with FOLFIRI plus bevacizumab treatment, and of course, in studies using other drugs.The following are available online at https://www.mdpi.com/2072-6694/12/6/1429/s1, Figure S1: Identification of blood leukocyte subsets by flow cytometry, Table S1: Factors associated with overall survival using baseline splenic volume as a binary variable by multivariate Cox analyses.Conceptualization F.G.; Formal analysis, E.L.; J.B.; and A.B.; Investigation, J.N.; M.T.; Y.R.; F.-X.C.-B.; F.A.; S.N.; C.S.; C.L.-B.; C.L.; M.C.; J.-L.L.; P.-L.E.; M.B.; M.P.; and T.A.; Methodology, E.L.; J.B.; A.B.; and F.G.; Project administration, F.G.; Resources, K.L.M.; Supervision, F.G.; Visualization, J.-N.; and E.L.; Writing—original draft, J N; Writing—review & editing, J.N.; E.L.; J.B.; Y.R.; F.-X.C.-B.; F.A.; S.N.; C.S.; C.L.-B.; C.L.; M.C.; J.-L.L.; P.-L.E.; M.B.; M.P.; T.A.; and F.G. All authors have read and agreed to the published version of the manuscript.This research received no external funding.We thank Fiona Caulfield for English editing.The authors declare no conflict of interest.Flow chart of the study population.Association between spleen volume and progression-free (PFS) and overall survival (OS). (A) Kaplan–Meier survival curves for PFS according to spleen volume (B) Kaplan–Meier survival curves for OS according to spleen volume.Association between blood parameters and spleen volume. (A,B) Comparison of myeloid and lymphocyte cell proportions in mCRC and peripheral blood leukocytes from matched healthy donors (HD). (C) Myeloid subset/lymphocyte ratio in mCRC and matched healthy donors. (D–G) Correlation between spleen volume and the proportion of granulocytes, mMDSC, lymphocytes and activated HLA-DR+ monocytes respectively.Baseline characteristics of the modified intention to treat population of the overall PRODIGE 9 trial, and of the cohort included in the present analysis.Factors associated with Progression-Free Survival using baseline splenic volume as a continuous variable, in univariate and multivariate Cox analyses. HR, hazard ratio; CI, confidence interval, LDH, Lactate deshydrogenase.Factors associated with Progression-Free Survival using baseline splenic volume as a vinary variable, in multivariate Cox analysis with bootstrapping. HR, hazard ratio; CI, confidence interval.

Med-MDPI/cancers/cancers-12-06-01430.txt

@@ -0,0 +1 @@
+Background: Protein O-fucosyltransferase 1 (POFUT1) overexpression, which is observed in many cancers such as colorectal cancer (CRC), leads to a NOTCH signaling dysregulation associated with the tumoral process. In rare CRC cases, with no POFUT1 overexpression, seven missense mutations were found in human POFUT1. Methods: Recombinant secreted forms of human WT POFUT1 and its seven mutated counterparts were produced and purified. Their O-fucosyltransferase activities were assayed in vitro using a chemo-enzymatic approach with azido-labeled GDP-fucose as a donor substrate and NOTCH1 EGF-LD26, produced in E. coli periplasm, as a relevant acceptor substrate. Targeted mass spectrometry (MS) was carried out to quantify the O-fucosyltransferase ability of all POFUT1 proteins. Findings: MS analyses showed a significantly higher O-fucosyltransferase activity of six POFUT1 variants (R43H, Y73C, T115A, I343V, D348N, and R364W) compared to WT POFUT1. Interpretation: This study provides insights on the possible involvement of these seven missense mutations in colorectal tumors. The hyperactive forms could lead to an increased O-fucosylation of POFUT1 protein targets such as NOTCH receptors in CRC patients, thereby leading to a NOTCH signaling dysregulation. It is the first demonstration of gain-of-function mutations for this crucial glycosyltransferase, modulating NOTCH activity, as well as that of other potential glycoproteins.Colorectal cancer (CRC) is the third most diagnosed cancer worldwide (second in females and third in males), with 1.8 million cases in 2018 and 880,792 deaths according to the World Health Organization. CRC is characterized by heterogeneous solid tumors, caused by the accumulation of numerous genetic alterations within cells, as well as epigenetic ones. Among the main causes identified in the progression of benign adenoma to malignant carcinoma, chromosomal instability was incriminated in more than 80% of sporadic cases of CRC [1]. The long arm of chromosome 20 was shown as a frequently dysregulated unstable region in several cancers such as CRC [2]. The region 20q11.21 was found to contain only four genes, including pleomorphic adenomagene-like 2 (PLAGL2) and protein O-fucosyltransferase 1 (POFUT1) [3]. A recent study showed that these two genes (POFUT1 and PLAG2) share a bidirectional promoter and, associated with a copy number amplification, promote colorectal cancer through dysregulation of both Notch and Wnt/ß-catenin signaling pathways [4,5]. The increased proliferation in CRC was attributed to PLAGL2 overexpression impacting the Wnt/ß-catenin pathway [4], while the colorectal tumor progression was correlated to dysregulation of Notch signaling caused by POFUT1 overexpression [6,7].POFUT1 is an endoplasmic reticulum (ER)-resident glycosyltransferase allowing O-fucosylation of the membrane and secreted glycoproteins within their EGF-like domains (EGF-LDs). This 388 amino-acid-long enzyme (EC 2.4.1.221) in humans, which comprises four disulfide bridges and a terminal KDEL-like ER-retention sequence (RDEF), is an invertase with a GT-B 3D structure characterized by two Rossmann-fold domains [8]. The interface between these two conserved domains forms the catalytic center comprising conserved residues interacting with the GDP-fucose donor substrate [9], and also with acceptor substrates, namely h-type EGF-LDs (hEGF-LDs) [10]. Interestingly, to be soluble and fully functional, POFUT1 must be modified by N-glycosylation at its two conserved consensus sites, as was shown for bovine POFUT1 [11]. POFUT1-mediated O-fucosylation is a rare post-translational modification occurring in ER, which leads to fucose transfer to serine or threonine residues of hEGF-LDs within the consensus sequence C2-XXXX-(S/T) -C3, where C2 and C3 are the second and third conserved cysteines. Among the potentially POFUT1-modified proteins, the most documented were Notch receptors which are extensively modified with O-fucoses in their extracellular domains [12,13,14]. Interestingly, Notch receptors O-fucosylation was widely reported to be involved in the modulation of their interaction with ligands, thereby controlling the activation of the Notch signaling pathway [15]. Consequently, O-fucosylation was shown to be required for regulation of Notch signaling in several physiologic processes, and its alteration can induce pathological situations [16]. Indeed, the knockout of Pofut1 in mice is lethal at midgestation [17], with embryos displaying the same severe phenotype as that of KO mice for Notch1 or Rbp-jk, the main transcriptional repressor of Notch signaling [18,19]. POFUT1 overexpression was observed in many cancers affecting different organs such as the liver [20], stomach [21], oral cavity [22], or breast [23]. The increased quantity of POFUT1 in hepatocellular carcinomas was associated with a poor prognosis. This was due to an abnormal NOTCH activation that led to in vitro increase of both proliferation and migration [20]. More recently, POFUT1 overexpression, mainly due to a 20q11.21 amplification and a subsequent increase of the POFUT1 copy number, was also detected in CRC from the stage I leading to a dysregulation of Notch signaling [6]. Remarkably, the previous study showed that among tumors with amplification of the unstable 20q.11.21 region (76%), 90% of tumors from CRC patients exhibited POFUT1 overexpression. However, in less frequent CRC cases (19%), with no chromosomal amplification and no change in POFUT1 quantity, seven missense mutations were found for POFUT1 with no consequences demonstrated on its enzymatic activity. These seven point-mutations (R43H, Y73C, T115A, S300L, I343V, D348N, and R364W), found in CRC using databases reporting single-nucleotide mutations in cancers, were predicted to be associated or not with a malignant prognosis.Given the importance of the O-fucosylation in the regulation of Notch receptors-ligands interactions and its consequences on Notch signaling activation, we chose to investigate the functional significance of these seven POFUT1 variants found in CRC. For this purpose, we first produced and purified soluble forms of WT and mutated POFUT1 variants in stable CHO cell lines and a correctly folded and non-glycosylated EGF-LD, as a substrate, in the bacterial periplasm. More precisely, the ability of fucose transfer was determined in vitro for each POFUT1 variant using GDP-fucose (azido-labeled or not) and NOTCH1 EGF-LD 26, known to be modified on its highly conserved O-fucose site [24]. After in vitro O-fucosylation reactions, copper-catalyzed azide-alkyne cycloaddition (CuAAC) referred to as click chemistry [25], and multiple reaction monitoring-mass spectrometry (MRM-MS), were performed, as previously described in Reference [25]. We did this provide evidence and quantify the O-fucose transfer activity of each POFUT1 variant compared to the WT one. The prediction of GDP-fucose binding to a mutated POFUT1 variant (R43H) by automatic molecular docking, as well as the current knowledge about interactions of POFUT1 with its donor and acceptor substrates, allowed us to discuss the impact of mutations on activities of POFUT1 variants, especially those with mutations close to the GDP-fucose binding cavity. In addition to investigations about the functional state of mutated POFUT1 proteins carried by CRC patients, this study also provides additional information on the structure–function relationships of this glycosyltransferase.According to the BioMuta database [26,27,28], reporting non-synonymous single-nucleotide mutations associated with cancers, the mutations R43H, Y73C, and T115A were found twice in POFUT1 of CRC patients, while the others (S300L, I343V, D348N, R364W) were observed once. Interestingly, one CRC patient was found to carry both mutations R43H and Y73C (TCGA-D5-6540-01) using the Firebrowse database. Our study could be helpful to speculate about the potential cumulative or compensatory effect of these two mutations on POFUT1 activity in this patient.We aligned peptide sequences encoding POFUT1 of several species (Figure 1). The human sequence, which is very similar to that of the mouse, exhibited many differences to nematode C. elegans. However, alignments showed several conserved regions, where some were known to be involved in binding to donor substrate, namely the regions R43-N46, H238-R240, and S357-F358 [29]. Interestingly, the residue R43, mutated to histidine in CRC (R43H), and located in the first substrate-binding region, was known to directly interact with the fucose moiety of the donor substrate (GDP-fucose) [30]. This crucial region also contains three highly conserved residues, namely M41, G42, and N46 in humans, known to establish important links through sulphur-hydrogen or hydrogen bonds with the C2-C3 subdomain of hEGF-LDs from mouse proteins [10]. The highly conserved Y73, mutated to cysteine in CRC (Y73C), was also demonstrated to be involved in binding to hEGF-LD [10]. For the three other highly conserved residues mutated in CRC (I343V, D348N, and R364W) and located close to the third substrate-binding region S357-F358, their contribution to POFUT1 activity was unknown yet. On the contrary, the two last mutations found in CRC, namely S300L and T115A, were considered to be non-conserved residues among species. Overall, the high conservation of at least five residues, among the seven mutated ones in CRC patients, as well as the implication of three of them in other types of cancers suggested functional consequences on POFUT1 activity.When structure-function studies should be performed on POFUT1, major drawbacks related to the isolation of this glycosyltransferase should be considered due to its status of ER-resident protein. To overcome this issue, we produced recombinant secreted proteins for human WT and mutated POFUT1 variants after establishing stable transfected CHO Flp-InTM cell lines. For this purpose, a modified expression vector named pSec-NtermHis6 [25] was used to allow secretion of recombinant POFUT1, with an N-terminal polyhistidine tag and devoided of its C-terminal ER-retention signal RDEF.To verify secretion of all recombinant POFUT1 variants, we performed western blot analyses with equal volumes of supernatants from CHO conditioned culture media using a specific anti-POFUT1 antibody. As shown in Figure 2, recombinant proteins were secreted unequally. Using medium supplemented with 10% Fetal Bovine Serum (FBS), only the secreted variants I343V and R364W were weakly detected in the supernatant. However, strong signals of equivalent intensity were seen for the other variants (Figure 2, upper panel). Similar results were obtained with 0% FBS medium (the condition used for protein production before purification), except for I343V, where its increased amount became comparable to that of other variants exhibiting strong signals (Figure 2, lower panel).All the recombinant histidine-tagged proteins, produced in serum-free media, were then purified using nickel-affinity chromatography. Regarding R364W variant, its secretion was very low in the supernatant of stable CHO cells; thus, higher amounts of cells were required to obtain enough purified protein in sufficient amount to perform functional experiments. The O-fucosyltransferase activity of WT protein and each POFUT1 variant was assessed in vitro, following incubation with an acceptor substrate (EGF-LD) and a chemically modified donor substrate (GDP-azido-fucose). As previously described [25], click chemistry reaction (CuAAC) coupling alkynyl-biotin to the azido-modified O-fucose, followed by blotting techniques using labeled peroxidase-streptavidin, were then performed to reveal fucose transfer to EGF-LDs (Figure 3).We first tested WT POFUT1 activity with the two NOTCH1 EGF-LDs, namely EGF-LD 12 and EGF-LD 26, known to be modified with O-fucose by POFUT1 [10]. Using the same procedure as previously reported, they were produced as isolated proteins in E. coli BL21 strain and purified. As expected, they were recognized previously as correctly folded and able to receive O-fucose by POFUT1 [25]. As negative controls, we used T466A EGF-LD 12 and T997A EGF-LD 26, having their O-fucosylation sites mutated.As expected, specific intense bands were observed at the proper size for both NOTCH1 WT EGF-LDs 12 and 26 incubated in vitro with human WT POFUT1 while no signal was obtained with their T/A mutated counterparts (Figure 3a). This result confirmed the successful transfer of azido-fucose to threonine residues of both O-fucosylation consensus sites of each EGF-LD. Similarly to mouse POFUT1 [25], the human counterpart was able to transfer fucose more efficiently to EGF-LD 26 than to EGF-LD 12 in vitro, since signals were obtained after 1 h and 20 h incubation with POFUT1 respectively. This led us to only perform in vitro O-fucosylation assays with EGF-LD 26. All recombinant POFUT1 variants were assessed in the same way with WT EGD-LD 26 and its negative control T997A (Figure 3b). Surprisingly, specific and strong signals were obtained in all cases, meaning that no mutation prevented WT EGF-LD 26 from being specifically modified by O-fucose at T997. However, signals with variable intensities were clearly observed, strongly suggesting that mutations affected POFUT1 O-fucosyltransferase activity. Indeed, all in vitro O-fucosylation assays were carried out with equal quantities of EGF-LD and POFUT1 variants, determined following protein quantification using the BCA method.MRM-MS is a highly sensitive method of targeted mass spectrometry (MS) which is used to selectively detect and quantify peptides, obtained after protein reduction, alkylation, and digestion [31]. MRM-MS was performed to quantify the ability of each POFUT1 variant to transfer fucose compared to WT POFUT1 as a reference, in the same way as previously described in Reference [25].For the in vitro O-fucosylation assays, the same amount of each POFUT1 variant was mixed with the same quantities of WT EGF-LD 26 (as a relevant acceptor substrate) and unlabeled GDP-fucose (as the donor substrate). After 20 h of incubation at 37 °C followed by trypsin digestion, the MRM-MS method was performed to confirm our previous results using click chemistry and to quantify the O-fucosyltransferase capacity of POFUT1 variants. Spectra obtained showed that human WT POFUT1 was indeed able to transfer specifically O-fucose to WT EGF-LD 26 (Figure 4) but with less efficiency (25.00 ± 20.68%) (Figure 5) than mouse recombinant POFUT1 (about 50%) [25]. All POFUT1 variants can partially modify at different degrees EGF-LD 26 with O-fucose since both non-modified and O-fucosylated peptides were detected. Regarding the mutants R364W, I343V, and D348N, peaks with more intensity were clearly obtained for the O-fucosylated forms than for the non-modified ones. This was consistent with an increased ability of these POFUT1 variants to transfer O-fucose on this EGF-LD (Figure 4). The variant I343V exhibited the best rate of O-fucosylation (86.45 ± 9.59%) (Figure 5), close to the rates obtained for the POFUT1 variants D348N, R364W, and surprisingly T115A. To a lesser extent, the R43H and Y73C mutations also led to a significantly increased O-fucosyltransferase activity (Figure 5). Finally, POFUT1 activity was not significantly affected when the non-conserved S300 residue was mutated to leucine.POFUT1 was recently shown to exert an essential role in the colorectal progression from precancerous lesions (adenomas) to carcinoma [32]. The overexpression of this O-fucosyltransferase, which is an enzyme involved in the modulation of NOTCH signaling activation, was found in CRC from stage I and was associated with tumor progression and metastasis [6]. On the contrary, its silencing in CRC cells led to inhibition of cell proliferation and diminished cell invasion and migration in vitro and in vivo [7]. Besides, POFUT1 overexpression was found in many other cancers and notably in hepatocellular carcinomas, where it was associated with a poor prognosis [20]. Overall, these results highlighted an undeniable oncogenic activity of POFUT1 in cancer such as CRC.POFUT1 overexpression was mainly associated with chromosomal amplification, leading to an increased amount of POFUT1 [3,6]. However, little is known about the modulation of POFUT1 activity, potentially caused by missense mutations related to human diseases, including CRC. A recent approach was conducted to determine how POFUT1 missense mutations could impact NOTCH1 signaling (R240A, M262T, S356F and R366W) found in Dowling-Degos disease (DDD), an autosomal dominant genodermatosis [30]. Three of these mutations located at highly conserved positions within or close to the substrate-binding regions (R240A, S356F, and R366W) were deleterious for Notch activity. However, the mutation of the residue M262 (M262T), which is not conserved in all species had no effect. To date, no other such point mutations were reported for POFUT1 in human diseases.In this study, structure-function studies were carried out for seven POFUT1 mutated variants (R43H, Y73C, T115A, S300L, I343V, D348N, R364W), resulting from point mutations and found in CRC patients according to BioMuta database. According to information found in this database regarding POFUT1, some of these seven residues (R43, T115, D348) were also mutated in other cancers such as uterine cancer (R43C), melanoma (T115A, D348N), and lung cancer (D348Y). These data could emphasize that these residues, including the less conserved residue (T115) among species, are critical for POFUT1 activity and that their mutation could impair its O-fucosyltransferase activity.Using the chemo-enzymatic approach to reveal the O-fucosyltransferase activity, we first showed that no mutation found in POFUT1 from CRC patients leads to a loss of enzymatic activity. On the contrary, all POFUT1 variants allowed an efficient azido-fucose transfer to NOTCH1 EGF-LD26. Indeed, the quantification of EGF-LD26 O-fucosylation using MRM-MS even highlighted a significant gain-of-function for the mutated POFUT1 variants, except for S300L. This mutation S300L is located in a region with low levels of conservation among species, namely the 299-304 region corresponding to the end of an α-helix of the Rossmann domain followed by a turn, far away from the catalytic center, as shown in Figure 6a.This study describes, for the first time, some mutated human POFUT1 variants as more active than WT POFUT1. To progress in understanding the effects of mutations on POFUT1 activity, especially those close to the GDP-fucose binding cavity (R43H, Y73C), automatic modelling by docking was performed using the COACH-D Server [33,34]. Possible explanations were proposed based on knowledge of interactions between the complex POFUT1/GDP-fucose with EGF-LD [10]. Since POFUT1 R43 residue was known to interact with the fucose moiety of GDP-fucose directly, the meta-server COACH-D was used to predict using molecular docking how the rearrangement of the catalytic cavity of mutated R43H POFUT1 could still allow binding to GDP-fucose without generating steric clashes. Interestingly, the proposed rearrangement (with a confident score of prediction (C-Score) of 0.79) of the binding cavity with a much more deeply buried H43 than R43 in the WT counterpart might result in a more exposed fucose moiety (Figure 6b,c). This led us to think that this new position of the donor substrate could facilitate fucose transfer to EGF-LD. However, a more complex docking would be required if considering the interaction of three molecules, namely POFUT1, GDP-fucose, and EGF-LD. In addition to a potential effect on GDP-fucose binding, the Y73C mutation is likely to affect EGF-LD binding since the equivalent residue in mice (Y78) was known to be involved in the interaction with EGF-LD [10,30]. Indeed, a stacking interaction was highlighted between Y78 of POFUT1 and the Gly residue of EGF-LD. This effect was observed frequently at the position C2+3 of O-fucosylable hEGFs [10]. While Asp or Glu at this position was shown in the latter study to induce a steric clash and to lead to a weakened interaction, a cysteine at this position (as for the mutation Y73C found in CRC) could establish a hydrogen bond with a EGF-LD reinforcing interaction, thereby leading to increased POFUT1 activity.We noticed that three mutations (I343V, D348N, R364W) leading to the most increased in vitro POFUT1 activity, with similar O-fucosylation rates ranging from 83.08 ± 7.17% to 86.45 ± 9.59% (compared to that of WT POFUT1 of 25.00 ± 20.68%), were all located in the highly-conserved region L341-D367 comprising the substrate-binding region S357-F358. Indeed, these three mutations are in the same Rossmann domain, relatively close to the binding pocket of GDP-fucose compared to the mutation T115A. Surprisingly, T115A mutation also gave rise to a significant increase of POFUT1 activity towards NOTCH1 EGF-LD26 while S300L did not have any effect. Indeed, T115 appeared to be less conserved among species than S300. We can hypothesize that mutations leading to a gain-of-function for POFUT1 induced a global conformation change favoring the accessibility of binding pocket to GDP-fucose, thereby leading to a more efficient transfer to EGF-LD 26. Additional experiments should be carried out to understand better the molecular interactions between mutated POFUT1 proteins and different EGF-LDs, whose interaction with WT POFUT1 is known. This could shed light on the gain-of-function induced by missense mutations found in POFUT1 from CRC patients. Recently, an increased amount of WT POFUT1, resulting from a chromosomal amplification of the region 20q11.21, was shown in CRC and correlated to tumor progression through NOTCH signaling activation [6]. A gain-of-function for POFUT1 could have similar effects in CRC patients carrying the missense mutations described in this study. In both cases, protein targets of POFUT1 such as NOTCH receptors and their ligands might be affected by POFUT1 overexpression or hyperactivity. This could have led to more O-fucose transferred to these protein partners, influencing receptor-ligand interactions, and subsequently, NOTCH signaling activation. However, Notch receptors and ligands are not the only targets of POFUT1. Indeed, they belong to the 87 human proteins identified to have at least one EGF-LD with a potential O-fucosylation consensus sequence [35]. Thus, it is likely that other membrane or secreted EGF-LD-containing proteins, predicted to be modified with O-fucose, are affected by overexpressed or hyperactive POFUT1 in CRC. It could be interesting to identify specifically, in CRC samples, the proteins affected by increased expression or activity of this ER-resident O-fucosyltransferase 1. The identification of these proteins would be helpful to provide more insights into the contribution of POFUT1 to the tumor process.In view of this study, the presence of these mutations in new CRC patients could be associated with colorectal tumor aggressiveness. Therefore, these mutations support previous studies [6,22] considering POFUT1 as a potential biomarker for CRC and other cancers.Plasmid constructs using pET-25b(+) vector (Novagen, Millipore, MA, USA) for bacterial production of WT NOTCH1 EGF-LDs 12 and 26, as well as their counterparts T/A mutated at their O-fucosylation site (T466A and T997A, respectively), were previously described in the Supplementary Materials and Methods [25]. The modified vector named pSec-NtermHis6, used in the latter study and derived from the commercial pSecTag/FRT/V5-His-TOPOR vector (Thermo Fisher Scientific, Waltham, MA, USA) was used to express a secreted form of recombinant human POFUT1 with an N-terminal polyhistidine tag. Indeed, human POFUT1 (NP_056167.1) (residues 27-384) cDNA without its signal peptide sequence and its C-terminal KDEL-like motif replaced by a stop codon, was cloned after PCR amplification between KpnI and BamHI sites in the modified vector named pSec-NtermHis6. The resulting plasmid construct referred to as pSec-NtermHis6Pofut1 was used as a template to generate the seven mutated plasmid constructs each carrying a missense mutation found in CRC, using the GENEART® Site-Directed Mutagenesis System (Invitrogen, Carlsbad, USA) according to the manufacturer’s protocol. After nucleotide sequence verification, each plasmid construct was subjected to a cotransfection with pOG44 vector (Thermo Fisher Scientific, Waltham, MA, USA) expressing the Flp recombinase to produce stably transfected Flp-InTM CHO cells (Thermo Fisher Scientific, Waltham, MA, USA).CHO Flp-InTM cells were cultured in an F12 medium (Thermofisher Scientific, Waltham, MA, USA) supplemented with 10% FBS (Biowest, EuroBio, Courtaboeuf, France) and 0.5% penicillin/streptomycin (Gibco, Carlsbad, CA, USA) and 100 µg×mL−1 of ZeocinTM (Thermo Fisher Scientific, Waltham, MA, USA), in a humidified atmosphere with 5% CO2. CHO Flp-InTM cells were cotransfected with 1 µg of plasmid DNA and 10 µg of pOG44 vector by lipofection with X-tremeGENE™ DNA Transfection Reagent (Sigma-Aldrich, Saint Louis, MO, USA) according to the manufacturer’s protocol. After transfection, the selection was initiated by changing medium with complete F12 medium containing 500 µg × mL−1 of Hygromycin B (Thermofisher Scientific, Waltham, MA, USA). Hygromycin B-resistant cells were then amplified and controlled for production of recombinant POFUT1 variants in culture medium by Western blot.In this study, we used the E.coli BL21 strain to produce recombinant peptides EGF-LDs as described in Reference [25]. We also used Flp-InTM CHO cells (Thermo Fisher Scientific, Waltham, MA, USA) and eight established cell lines corresponding to Flp-InTM CHO cells, stably expressing WT POFUT1 or one of the seven recombinant mutated (R43H, Y73C, T115A, S300L, I343V, D348N, and R364W) POFUT1 variants.Recombinant human WT and mutated POFUT1 variants with an N-terminal polyhistidine tag were produced as secreted proteins by stable Flp-InTM CHO cells. After production during 72 h in serum-free F12 medium, proteins were recovered by centrifugation from cell culture supernatants, concentrated in binding buffer (25 mM Tris-HCl, 500 mM NaCl, 5 mM CaCl2, 20 mM imidazole, pH 7.5) and purified on a Ni-NTA column by imidazole gradient using AKTA prime system (GE Healthcare, Piscataway, NJ, USA). Recombinant WT and T/A mutated EGF-LDs of mouse NOTCH1 were produced in BL21 and purified as previously described in Reference [25]. All purified recombinant proteins used in this study were concentrated through Amicon ultra centrifugal filters 3K or 10K in Tris-CaCl2 (25 mM Tris, 5 m CaCl2, pH 7.5) and quantified using a bicinchoninic acid (BCA) protein assay (Sigma-Aldrich, Saint Louis, MO, USA) with bovine serum albumin as a standard.For POFUT1 variants, purified proteins were resolved by SDS-PAGE using 12% polyacrylamide gels and transferred to 0.45 µm nitrocellulose membrane (GE Healthcare, Buckinghamshire, UK), for 90 min at 0.8 mA per cm2. Membranes were blocked with TBS-T (50 mM Tris, 150 mM NaCl, pH 7.6 (TBS), supplemented with 0.1% Tween-20 (v/v)) and 5% (w/v) fat-free milk, for 1 h at room temperature. Membranes were then incubated with anti-POFUT1 antibody [36], diluted at 1:2000 in TBS with 0.5% Tween-20 (TBS-T0,5) containing 5% (w/v) of fat free milk overnight at 4 °C. After three washes with TBS-T0.5, the membranes were incubated with anti-rabbit HRP-conjugated IgG (Dako, Glostrup, Denmark), diluted at 1:3000 in TBS-T0.5 containing 2% (w/v) non-fat dry milk for 1 h at room temperature. After three washes in TBS-T0.5, proteins were revealed after addition of ECLTM Prime Western blotting detection reagent (GE Healthcare, Uppsala, Sweden) and visualized using an Amersham Imager 600 device (GE Healthcare, Uppsala, Sweden). After reactions by click chemistry, EGF-LDs were separated on 15% polyacrylamide gel and transferred to 0.2 µm nitrocellulose (GE Healthcare, Buckinghamshire, UK), for 30 min at 0.8 mA per cm2. Membranes were blocked with 10% fat-free milk-TBST for 10 min, before being incubated with streptavidin-HRP in TBS-T0.5 at 25 ng.mL−1 for 30 min. The membranes were washed three-times before and after streptavidin-HRP incubation with TBST, for 15 min per wash. The membranes were revealed as described above.Before the CuAAC experiments, glycosyltransferase reactions were carried out with 1µg of WT POFUT1 or one mutated POFUT1 variant, mixed with 2 nmoles of GDP-azido-fucose (R&D Systems Inc., Minneapolis, MN, USA) as recommended by the manufacturer (R&D Systems) and 2 µg of purified EGF-LD in 25 µL of reaction buffer (25 mM Tris, 5 mM CaCl2, 10 mM MnCl2, pH 7.5), and then incubated for 1 h or 20 h at 37 °C.For mass spectrometry analysis, 1 µg of WT or mutated POFUT1 was incubated with 2 µg of purified EGF-LD and 2 nmoles of GDP-fucose in 20 µl of reaction buffer and incubated for 20 h at 37 °C. CuAAC was performed using 1.25 mM CuCl2, 2.5 mM ascorbic acid, and 0.125 mM alkynyl biotin (R&D Systems Inc., MN, USA), directly added to the glycosyltransferase reactions. The mixture was incubated in the dark for 1 h at room temperature.Proteins were reduced, alkylated, and digested by Glu-C. Resulting peptides were analyzed with a nanoLC 425 in micro-flow mode (Eksigent, Dubli, CA, USA) coupled to a TTOF5600+ mass spectrometer (SCIEX, Framingham, USA) in Information-Dependent Acquisition mode. ProteinPilot 5.0 (SCIEX) was applied to search against the recombinant protein sequence database, and the MRM transition list was established using Skyline 3.5.0 (MacCoss Lab, University of Washington, Seattle, WA, USA). The O-fucose was added in silico at the expected position with PeakView software (SCIEX, Framingham, MA, USA), and m/z of precursor and fragments were calculated. The data were acquired in high-Resolution MRM mode and processed with MultiQuant Software 3.0.1 (SCIEX, Framingham, MA, USA). Areas were collected for the same most intense fragment of O-fucosylated and non-modified peptides, and a ratio of O-fucosylation was calculated.Alignments of POFUT1 protein sequences were performed with FASTA sequences of POFUT1 from different species, using the Multalin server (http://multalin.toulouse.inra.fr) [37]. Using MatchMaker of UCSF CHIMERA (http://www.rbvi.ucsf.edu/chimera) (Resource or Biocomputing, Visualization, and Informatics at the University of California, San Francisco, CA, USA, with support from NIH P41-GM103311) [38], mouse NOTCH1 EGF-LD 26, co-crystallized with mouse POFUT1 (PDB 5KY4) [10], was superimposed with co-crystallized human POFUT1 with GDP-fucose (PDB 5UXH) [30]. Docking experiments were made using the COACH-D server (https://yanglab.nankai.edu.cn/COACH-D/), FASTA sequences of WT or R43H POFUT1 variant, and GDP-fucose (in sdf format file) as a donor substrate. The proposed model for R43H POFUT1 interacting with GDP-Fucose as a substrate was selected according the best C-score, in PoseU, and visualized using UCSF CHIMERA.All the experiments were performed at least three times independently. Statistical comparisons were achieved using the t-Student test implemented in GraphPad Prism 7 (GraphPad Software Inc, San Diego, CA, USA). Results were considered statistically significant if the p-value was less than 0.05.In conclusion, our findings indicate that among the rare missense mutations affecting POFUT1 and found in patients with colorectal cancer, six of them induced an increase of O-fucosyltransferase activity compared to WT POFUT1. These mutated variants could modify the O-fucosylation status of POFUT1 protein targets such as NOTCH1 receptor and its ligands, and subsequently promote colorectal cancer.The following are available online at https://www.mdpi.com/2072-6694/12/6/1430/s1, Figure S1: Uncropped blots of Figure 2, Figure S2: Uncropped blots of Figure 3a, Figure S3: Uncropped blots of Figure 3b. Conceptualization, M.D. and A.M.; Funding acquisition, A.M.; Investigation, M.D., F.P., and S.L.; Methodology, E.P.; Supervision, A.M.; Writing original draft, M.D.; Writing review & editing, F.P., S.L., and A.M. All authors have read and agreed to the published version of the manuscript.This work was supported by a French Ministry of Higher Education and Research doctoral fellowship to MD. The “Ligue contre le cancer” partly funded the work.We gratefully acknowledge Julien Chabanais for his help with the Firebrowse database. We are grateful to Layla Haymour for her help in English editing. The authors declare no conflict of interest.Alignment of POFUT1 protein sequences. Using the Multalin server, multiple sequence alignment for POFUT1 was done for different species such as Homo sapiens, Mus musculus, Gallus gallus, Xenopus laevis, Danio rerio, and Caenorhabditis elegans. Among the seven missense mutations found in POFUT1 from CRC patients, five of them corresponded to mutations of highly conserved residues (R43, Y73, I343, D348, and R364), even located in a substrate-binding region for some residues such as R43. However, POFUT1 T115 and S300, mutated in CRC patients, were not conserved among species. The three conserved regions known to be involved in binding to GDP-fucose are indicated with black arrows and specific sequences (signal peptide and KDEL-like ER-retention signal) with green arrows. Asterisks indicate the positions of conserved cysteines forming disulfide bridges.Western blot analysis, using anti-POFUT1 antibody, of recombinant POFUT1 variants secreted by stable CHO cell lines. Supernatants from each stable cell line were recovered by centrifugation from culture media after three days of protein production in F12 media with 10% (upper panel) or 0% (lower panel) fetal bovine serum (FBS). Same volumes of crude supernatants were loaded for WT and mutated POFUT1 proteins. Uncropped blots are shown in Figure S1.In vitro O-fucosyltransferase assay for activity of WT and mutated POFUT1 variants using click chemistry. (a) After 1 h or 20 h incubation at 37 °C of recombinant WT human POFUT1 with NOTCH1 EGF-LD 26 (WT and T997A) or EGF-LD 12 (WT and T466A) and GDP-azido-fucose, click chemistry (CuAAC) with alkynyl biotin was performed to link biotin to transferred O-linked fucose covalently. After SDS-PAGE and blotting techniques, EGF-LDs modified with O-fucose were revealed using HRP-streptavidin. (b) After 1 h independent incubations of WT POFUT1 and each mutated POFUT1 variant with WT or T997A EGF-LD 26 at 37 °C, the same procedure as described in (a) was carried out. A low exposure did not allow us to visualize WT POFUT1 labelling. After high exposure, the signal for WT POFUT1 was detected, and quantifications were performed for each POFUT1 variant compared to WT. Uncropped blots are shown in Figures S2 and S3.Multiple reaction monitoring mass spectrometry (MRM-MS) analyses of trypsin-digested WT EGF-LD26, after incubations with GDP-fucose and either WT POFUT1 or mutated POFUT1 variants independently. Peaks corresponding to non-modified peptides (upper panels) and peptides modified with O-fucose (lower panels) were shown for WT POFUT1 and all POFUT1 variants. Peaks correspond to the most intense MS2 fragment signal, obtained during the run time.Quantification of the O-fucosyltransferase activity of WT POFUT1 and its seven mutated counterparts found in CRC patients. The percentage of O-fucosylation was determined from previous MRM-MS analyses based on ratios of peak areas of O-fucosylated peptides reported to those of total peptides. Bar graph represented mean of percentage ± SD. Each experiment was performed three times independently. Statistical significance was assessed using a two-tailed Student test; * p < 0.05, ** p < 0.01 vs. WT.Location of POFUT1 residues mutated in CRC and automatic docking models for WT POFUT1 and R43H variant with GDP-fucose as a ligand. (a) Using Matchmaker of CHIMERA, X-ray structure of human POFUT1 (grey) co-crystallized with GDP-fucose (black) (PDB 5UXH) was superimposed with mouse POFUT1 co-crystallized with EGF-LD 26 (blue) (PDB 5KY4). After removal of mouse POFUT1, residues implicated in CRC mutations were shown with their sidechains (red). (b) Using the COACH-D docking server, we performed automatic molecular docking using the FASTA sequence of WT human POFUT1 as a template and GDP-fucose (in red) as a ligand (donor substrate). The residue R43, located in front of the binding pocket to GDP-fucose, was colored in cyan. (c) In the same way, a structural model was generated for mutated POFUT1 in which the H43 moved to a more deeply buried position in the binding pocket of GDP-fucose, compared to R43 in WT POFUT1. This could lead to a more exposed fucose moiety of the donor substrate to improve fucose transfer to an EGF-LD.

Med-MDPI/cancers/cancers-12-06-01431.txt

@@ -0,0 +1 @@
+Understanding the cellular interactions within the tumor microenvironment (TME) of melanoma paved the way for novel therapeutic modalities, such as T cell-targeted immune checkpoint inhibitors (ICI). However, only a limited fraction of patients benefits from such therapeutic modalities, highlighting the need for novel predictive and prognostic biomarkers. As myeloid cells orchestrate the tumor-specific immune response and influence the efficacy of ICI, assessing their activation state within the TME is of clinical relevance. Here, we characterized a myeloid activation (MA) signature, comprising the three genes Cxcl11, Gbp1, and Ido1, from gene expression data of human myeloid cells stimulated with poly(I:C) or cGAMP. This MA signature positively correlated to overall survival in melanoma. In addition, increased expression of the MA signature was observed in melanoma patients responding to ICI (anti-PD-1), as compared to non-responders. Furthermore, the MA signature was validated in the murine B16F10 melanoma model where it was induced and associated with decreased tumor growth upon intratumoral administration of poly(I:C) and cGAMP. Finally, we were able to visualize co-expression of the MA signature genes in myeloid cells of human melanoma tissues using RNAscope in situ hybridization. In conclusion, the MA signature indicates the activation state of myeloid cells and represents a prognostic biomarker for the overall survival in melanoma patients.Recent clinical success using immune checkpoint inhibitors (ICI) has demonstrated the therapeutic potential of harnessing T cells to treat various cancer types, including cutaneous melanoma [1,2]. However, only a proportion of patients exhibit durable responses to ICI and certain cancer types remain largely refractory to this line of treatment [1,3].While it is well known that cytotoxic T cells are a primary effector population in tumor-specific immune responses, the roles of myeloid cells in the tumor microenvironment (TME) are less clear and remain a major focus of investigation [4,5,6]. Tumor-associated myeloid cells represent a vast majority of leukocytes in most tumors and regulate tumor-specific immune responses as well as responses to cancer therapies [5]. Moreover, tumor-associated myeloid cells display significant phenotypic and functional heterogeneity and can either promote or suppress tumor immunity [5,6]. In particular, dendritic cells (DC) and tumor-associated macrophages (TAM) are considered to be key regulators of tumor-specific immune responses due to their function in the priming and recruitment of tumor-infiltrating lymphocytes (TIL) and their ability to modulate tumor stromal and cancer cells [7,8]. As such, a number of therapeutic options such as DC-vaccines and CAR-T cells have been focused on reprogramming intratumoral DC and TAM towards an immunogenic phenotype [7,9,10].The fact that certain types of human cancer show evidence for spontaneous tumor immunity, prompted investigations to identify innate immune pathways that regulate these responses [11]. Preclinical studies revealed a major role of the stimulator of interferon genes (STING) pathway, a cytosolic DNA sensor, as a crucial event required for optimal type I interferon production, dendritic cell activation, and priming of CD8+ T cells against tumor-associated antigens [11,12]. Notably, there is a substantial interest in the investigation of molecules which mimic microbial infection or cellular damage by binding to pattern recognition receptors (PRR) expressed by innate immune cells [13,14]. While a number of PRR agonists are currently undergoing pre-clinical investigations in murine models and some have entered clinical trials, cytosolic PRR for nucleic acids are considered highly attractive targets for the activation of innate immune cells [13,14]. Of particular interest to modulate tumor-specific immune responses are the two PRR, Toll-like receptor 3 (TLR3), which recognizes double stranded RNA and the cytosolic DNA sensor STING [12,14]. TLR3 and STING agonists mimic viral infections, thereby resulting in similar immunological responses such as the production of Type I interferon, the skewing of TAM to an immunogenic phenotype, and promoting maturation and cross-presentation function of DC [4,13]. A number of TLR3 and STING agonists in combination with ICI or other treatments have entered clinical trials to promote an immunogenic TME [14].In order to select the optimal myeloid cell targeted agents, it is essential to define biomarkers or signatures of activated myeloid cells that are associated with therapeutic response, improved survival, or other clinical parameters. These biomarkers or signatures can then be used in clinic to determine the activation state of myeloid cells and to tailor the treatment regimen. In the context of both DC and TAM, phenotyping requires the assessment of numerous surface and intracellular markers (e.g., CD80, CD86, HLA-DR, CD206, CD163, iNOS, Arg1), to distinguish between pro-inflammatory, immunogenic, and non-inflammatory, immature or immune-suppressive subsets [10,15]. In this study, we describe a 3-gene myeloid activation (MA) signature derived from DC and monocytes treated with the TLR3 agonist poly(I:C) and the bacterial cyclic dinucleotide 3′3′-cGAMP, a STING agonist. The MA signature holds potential for the use in both clinical and pre-clinical applications to determine the presence of activated myeloid cells, such as immunogenic M1 macrophages and to predict the overall survival of melanoma patients.To define a human myeloid activation (MA) gene signature, peripheral blood-derived dendritic cells (DC) and CD14+ monocytes were isolated from healthy individuals and stimulated with poly(I:C) or cGAMP in vitro. Gene expression profiling was performed using the NanoString nCounter Immunology Panel containing 594 immune-related genes. Unsupervised t-Distributed Stochastic Neighbor Embedding (t-SNE) plot showed significant clustering between unstimulated and stimulated cells for both monocytes and DC (p-value: <0.001, Figure 1A).Further analysis revealed ten differentially expressed genes (DEG) that were shared between myeloid cells stimulated with either poly(I:C) or cGAMP (Table 1).In order to validate whether these shared genes denote a general state of myeloid cell activation, we analyzed two publicly available microarrays (GSE2706, GSE1925) and two RNA-sequencing (RNA-seq) datasets (GSE57494, GSE82227) that involved human myeloid-derived cells activated with classical innate immune cell stimuli such as LPS and IFNγ. We identified three genes, Cxcl11, Gbp1, and Ido1, that were significantly upregulated in all datasets (Table 2).Furthermore, 3D scatterplots with Cxc11, Gbp1, and Ido1 expression data showed clustering between unstimulated and stimulated cells (Figure 1B), indicating the utility of these three genes as a signature to assess myeloid cell activation.To investigate whether the gene signature (Cxcl11, Gbp1, and Ido1) is specific to activated myeloid cells, we tested whether these genes were upregulated in activated non-myeloid cells (NK cells, B cells, CD8, and CD4 T cells) (GSE63038, GSE85543, GSE79626, GSE60235, respectively). None of the three signature genes was upregulated in activated non-myeloid immune cells, suggesting a selective induction in activated myeloid cells (Supplementary Table S1). Considering the complexity of the immune cell composition in blood and tissue, which can impede the detection of gene signatures, we investigated the expression of our MA signature in RNA-seq data from peripheral blood mononuclear cells (PBMC), comprising 29 distinct immune cells [16]. We found that the baseline expression of Cxcl11, Gbp1, and Ido1 in lymphoid cells was lower than in steady-state myeloid cells. Furthermore, the expression of the signature genes was highest in myeloid cells as indicated by a non-parametric, rank-based score (Figure 2A).Using bulk tumor RNA-seq data from The Cancer Genome Atlas Project (TCGA), we correlated our MA signature to previously described M1 macrophage and mature DC signatures in various cancer types (Supplementary Table S2) [17,18]. The MA signature showed a strong correlation to both the M1 macrophages, as well as to mature DC in all cancer types of the TCGA cohort (Figure 2B).Given the important function of M1 macrophages and mature DC within the tumor microenvironment (TME), we investigated the direct association of the MA signature with overall survival (OS) in all cancer types available in TCGA [19]. Survival analysis using log-rank test showed strong association between MA signature and overall survival (OS) in various cancer types (Figure 3A).In particular, in the skin cutaneous melanoma samples (SKCM), the Kaplan–Meier curve shows that those with higher MA signature scores have consistently higher OS throughout their follow-up (Figure 3B). To further analyze the correlation between the MA signature gene expression and immune cell populations within SKCM, CIBERSORT (Cell-Type Identification By Estimating Relative Subsets of RNA Transcripts) was used to estimate the relative proportions of various immune cell types from bulk RNA-seq data [20]. Tumors with high MA signature expression displayed increased proportions of M1 macrophages, CD8+ T cells, and activated CD4+ T cells, but reduced M2 and M0 (unpolarized) macrophages (Figure 3C). No additional associations were observed between the MA signature and any of the other immune cell populations investigated (Figure S1). We further examined gene expression data from biopsies of advanced melanoma patients receiving PD-1 blockade therapy. On-treatment biopsies taken from patients with a beneficial response to PD-1 blockade showed increased expression of the MA signature compared to patients with no response to therapy (Figure 3D). Gene expression data from melanoma patients before ICI therapy show a trend of increased signature expression in ICI responders as compared to non-responders (Figure S2). Therefore, the predictive value of this signature for the response to ICI therapy requires further investigation.To investigate if the MA gene signature is induced in activated murine macrophages, we used bone-marrow derived macrophages (BMDM) stimulated with poly(I:C) or cGAMP for 6 and 24 h in vitro. We detected a significantly enhanced expression of the signature genes Cxcl11, Gbp2b (murine homologue to hGbp1), and Ido1 in stimulated BMDM at both timepoints compared to untreated cells (Figure 4A,B).To further validate the use of the MA signature in vivo, we used the well-established murine B16F10 melanoma model. Initially, B16F10 melanoma cells were analysed for the expression of the MA signature genes after stimulation with poly(I:C) or cGAMP for 24 h in vitro. The only signature gene that was marginally upregulated in B16F10 cells upon stimulation with poly(I:C) or cGAMP was Cxcl11 (Figure 4C). However, the magnitude of induction was approximately five-fold lower as observed in BMDM (Figure 4A,B). In addition, poly(I:C) and cGAMP were investigated for their cytotoxic effect on B16F10 melanoma cells and the human melanoma cell lines SK-Mel-37 and D10 using an AlamarBlue assay. However, cell viability was not affected in any of the tested cell lines upon incubation with poly(I:C) or cGAMP (Figure 4D).To further explore the utility of the MA signature as a biomarker for myeloid cell activation in melanoma, B16F10 melanoma bearing C57BL/6J mice were treated with intratumoral injections of poly(I:C) or cGAMP. A significant reduction of tumor growth was observed in both treatment groups compared to control treated mice (Figure 5A).RNA was extracted from whole tumor tissue and analyzed for the expression of Cxcl11, Gbp2b, and Ido1. Each of the three MA signature genes showed a significant increased expression in both poly(I:C) and cGAMP treated mice, compared to PBS treated controls (Figure 5B). Taken together, our findings suggest a potential use of the three genes Cxcl11, Gbp2b, and Ido1 as a signature for myeloid cell activation in whole tumor tissue and will be further investigated as a clinical prognostic biomarker for melanoma patients.By studying the gene expression profiles of human myeloid cells upon stimulation with poly(I:C) or cGAMP, we found a gene signature associated with their activation, that comprises the three genes Cxcl11, Gbp1, and Ido1 and allows to assess the activation state of myeloid cells in bulk tumor tissue. The myeloid activation (MA) signature positively correlated with the presence of M1 macrophages and CD8+ T cells within melanoma tumors, as well as improved overall survival of melanoma patients. Moreover, significantly increased expression of the MA signature was observed in on-treatment biopsies from patients with a beneficial response to PD-1 blockade.While an in-depth study of the function of these genes and their proteins is beyond the scope of this study, we will briefly discuss the functional role of each gene. CXCL11 is a C-X-C motif chemokine that binds to the CXCR3 receptor and is involved in the recruitment, differentiation and activation of various immune cells such as monocytes, NK cells, T cells and DC [22]. In melanoma, the presence of CXCR3 ligands such as CXCL9, CXCL10, and CXCL11 has been associated with enhanced CD8+ T cell infiltration and hence, better survival prognosis [23].GBP1 belongs to the GBP family of IFN-induced GTPases which play essential roles in orchestrating protective immunity to a wide range of microbes [24]. Down-regulation of GBP1 is reported to cause mitochondrial dysfunction, which eventually induces cellular senescence in inflammatory macrophages [25]. The enzyme indoleamine 2,3-dioxigenase 1 (IDO1) is a protein that catabolizes tryptophan and produces a range of kynurenine metabolites [26]. In cancers, increased IDO1 activity has been shown to promote the development of an immunosuppressive microenvironment and inhibit anti-tumor immune responses [27]. In fact, IDO1 plays a role in the suppression of effector T and NK cells and differentiation and activation of regulatory T (Treg) cells and myeloid-derived suppressor cells (MDSCs) [28,29,30]. Although IDO expression is induced by IFN-γ, TNF-α, or prostaglandins, macrophages are driven toward an immunosuppressive M2 phenotype when IDO is overexpressed [31]. In spite of the immunosuppressive action of IDO1, it has also been shown to induce inflammatory responses within the tumor microenvironment [32]. IDO1 is also involved in promoting tumor neovascularization by modulating the expression of interferon-γ (IFN-γ) and IL-6 [33,34]. However, the complete immunomodulatory effects of IDO have not yet been characterized fully. Clinical trials have targeted IDO1 with inhibitory molecules as a strategy to counteract intratumoral immunosuppression but the results from these trials have been largely inconclusive [26]. In fact, IDO1 was described as an independent prognostic marker for increased disease-free survival in patients with colorectal cancer [35]. Notably, each of the three genes constituting the MA signature has non-redundant biological functions.In our study, we characterized a MA signature consisting of three distinct genes (Cxcl11, Gbp1, Ido1), which correlates with the presence of M1 macrophages and mature DC in various cancer cohorts available in TCGA. Moreover, the MA signature is positively associated with increased overall survival and beneficial treatment response in cutaneous melanoma patients.Using the TLR3 and STING agonists poly(I:C) and cGAMP, respectively, the expression of the MA signature was induced in murine BMDM, but not in B16F10 melanoma cells. The MA signature expression was significantly elevated in murine B16F10 melanomas treated with intratumoral injection of poly(I:C) or cGAMP. Myeloid cells play a crucial role in regulating tumor immunity by altering the tumor microenvironment and regulating tumor-specific T cell recruitment [4,5,7,8]. Accordingly, increased densities of mature DC and M1 macrophages are associated with improved patient survival in a number of cancer types [36,37,38,39].Currently, reprogramming tumor-associated myeloid cells, especially macrophages and DC, towards an immunostimulatory phenotype is a major area of investigation in the field of tumor immunology [10,40]. TLR3 targeting ligands and STING agonists are currently being tested in clinical trials as adjuvant therapy, or along with other drugs and vaccines, against a variety of cancers such as melanoma, bladder cancer, and lymphomas [11,41]. Evidence from murine models suggests that both TLR3 and STING agonists can induce macrophage polarization towards an inflammatory, anti-tumorigenic phenotype [42,43]. At the present, only a limited number of intracellular or surface markers exist for the assessment of M1 activated macrophages (e.g., iNOS, CD68, CD80) [15,44], however, as recent single-cell profiling approaches have demonstrated, canonical M1 and M2 macrophage markers are co-expressed on many subsets of tumor-associated macrophages [45]. Similarly, in-depth dissection of the tumor-infiltrating DC landscape has also revealed significant heterogeneity in phenotype and function resulting in the delineation of distinct subsets [8]. Therefore, assessing the activation state of myeloid cells remains a challenging albeit essential component of tumor immunology research both at the clinical and pre-clinical stage [46,47]. While it is possible to infer the abundance or presence of activated and resting immune cell subsets using immune deconvolution algorithms (such as CIBERSORT) in microarray or RNA-Seq datasets, these algorithms employ multigene signatures which are not feasible for the application in routine clinical tests [48]. In contrast, the MA signature described here comprises three genes that, in combination, are highly specific for activated myeloid cells and readily assessable via qPCR in tumors and by in situ hybridization of human melanoma tissues. Importantly, none of the MA signature genes were shown to be expressed in activated lymphoid cells such as T cells, B cells, or NK cells.Recently, myeloid cell targeted therapies have entered clinical trials not only as monotherapies, but also in combination with immune checkpoint antibodies [14,40,49]. Thus, further research is required to dissect the clinical utility of myeloid cell activation therapies. While, enhanced expression of the MA signature is associated with activated M1 macrophages and improved survival in melanoma patients, it may also serve to predict the necessity and success of myeloid cell-targeted therapies under pre-clinical investigation.The MA signature therefore offers a prognostic and diagnostic tool for both clinical researchers and basic scientists investigating the roles of myeloid cells in the immune response to tumors. In our report, we demonstrated in situ and in silico, that the MA signature is a marker for activated M1 macrophages in melanoma and other cancer types. However, whether the MA signature can also be used as a predictive marker for therapy efficacy in melanoma as well as in other malignancies is subject to further investigation.Murine B16F10 melanoma cells (ATCC) were cultured in complete RPMI-1640 medium (Sigma Aldrich, Darmstadt, Germany) supplemented with 10% FBS, 100 μg/mL streptomycin, 100 U/mL penicillin, 1 mM sodium pyruvate, and 2 mM L-glutamine. Cultured B16F10 melanoma cells were plated at a concentration of 0.1–0.2 × 106 cells/mL and stimulated for 24 h with poly(I:C) (Tocris Bioscience, Bristol, United Kingdom, 10 μg/mL),) cGAMP (InvivoGen, San Diego, CA, USA 10 μg/mL), or left unstimulated. BMDM were cultured using complete RPMI-1640 medium as described above. The human melanoma cell lines SK-Mel-37 and D10 were a gift from P. Zajac (University of Basel). Melanoma cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 100 U/mL penicillin, 1 mM sodium pyruvate, and 2 mM L-glutamine. All cells were maintained at 37 °C under 5% CO2 atmosphere.PBMC from healthy human donors (Interregionale Blutspende SRK) were isolated using Ficoll (GE Healthcare, Chicago, IL, USA) density gradient centrifugation. Monocytes were enriched using the EasySepTM Human Monocyte Enrichment Kit w/o CD16 Depletion (STEMCELL Technologies).DC were isolated using the MACS-based Pan-DC Enrichment Kit (Miltenyi Biotec, Bergisch Gladbach, Germany) followed by Fc-receptor blocking with anti-mouse CD16/32 (2.4G2, generated in house) for 15 min. Subsequently, cell surface markers were stained using anti-CD11c (3.9), anti-CD123 (6H6), anti-human lineage cocktail (CD3/14/16/19/20/56), anti-CD141 (M80), anti-HLA-DR (L243) (Biolegend, San Diego, CA, USA) in FACS buffer (PBS with 2% FBS and 1 mM EDTA) for 45 min on ice. Cells were then purified by FACS using the Moflo Astrios EQ cell sorter (Beckman Coulter, Nyon, Switzerland). Ethical approval provided by the Kantonale Ethikkommission Bern (KEK), Switzerland. Code: 2017-02246.Purified monocytes and DC were stimulated with poly(I:C) (Tocris Bioscience, 10 μg/mL), cGAMP (InvivoGen, 10 μg/mL) complexed with lipofectamine (3 μg/mL, Invitrogen), or left untreated for 6 h. Cells were then lysed in Buffer RLT (Qiagen, Hilden, Germany) containing 1% β-mercaptoethanol (Sigma Aldrich) and diluted 1:2 in PBS. Subsequently, 5 µL of the lysate was directly used for mRNA profiling using the NanoString Human Immunology Panel V2. Transcriptional profiling was performed using the nCounter™ Digital Analyzer (NanoString Technologies, Seattle, WA, USA). Data quality control and normalization was performed using NanoStringQCPro and NanoStringNorm. Differential gene expression was performed using edgeR package. T-distributed Stochastic Neighbor Embedding (t-SNE) was used for the dimensional reduction using the R tsne package. Permutational analysis of variance (PERMANOVA) paired test was used to test for the effect of poly(I:C) and cGAMP in both monocytes and DC cell-types.To differentiate murine BMDM, bone marrow was flushed from femurs and tibias of C57BL/6J mice and cultured in petri dishes at 5 × 106 cells per dish using complete RPMI-1640 (as described above) including 20% L929 supernatant and the medium was replaced on day 4. After 7 days, the BMDM were harvested and collected with a cell scraper and seeded at a concentration of 0.5 × 106 cells/mL, followed by stimulation with poly(I:C) (Tocris Bioscience, 10 μg/mL), cGAMP (InvivoGen, 10 μg/mL), or left untreated for 6 and 24 h, respectively.Cells were cultured and stimulated as mentioned before and then lysed with RLT buffer (RNeasy Mini Kit, Qiagen) supplemented with 1% β-mercaptoethanol (Sigma Aldrich). Column purification was performed according to the manufacturer’s protocol (RNeasy Mini Kit, Qiagen).For the bulk tumor analysis, total RNA was isolated in TRI Reagent (Sigma Aldrich), followed by column purification. RNA quantity was assessed using NanoDrop (ND-1000 Spectrophotometer, Thermo Scientific), followed by DNase digestion (DNase, Invitrogen) and reverse transcription of the cDNA with random primers (High Capacity cDNA Reverse Transcription Kit, Applied Biosystems) at a total RNA concentration range of 0.7–2 mg. Using the SYBR Green PCR Master Mix, 2–3 technical replicates were performed on the StepOnePlus Real-Time PCR System (ThermoFischer, Waltham, MA, USA). Applied Biosystems software provided the raw data (Ct). Relative mRNA levels were calculated by the 2−ΔCt method, using Rplp0 as a housekeeping gene [21]. Primer list: Supplementary Table S3.Melanoma cells (SK-Mel-37, D10, B16F10) were stimulated with poly(I:C) (Tocris Bioscience, 10 μg/mL), cGAMP (InvivoGen, 10 μg/mL), or left untreated for 24 h. Following incubation of cells (B16F10, SK-Mel-37, D10) in wells (200 μL of culture medium), 10% v/v AlamarBlue (ThermoFischer) was added. After 6 h of incubation at 37 °C, fluorescence intensity was measured at 560 nm excitation and 590 nm emission using the Infinite 200 Pro Plate Reader (Tecan Life Sciences, Zürich, Switzerland). Fluorescence intensity was plotted using Graph Pad Prism 8.0.C57BL/6 wt mice were purchased from Janvier Labs (France) and age-matched (8–10 weeks), as well as sex-matched for each experiment. Tumor inoculation was performed by subcutaneous injection of 2 × 105 B16F10 melanoma cells into the left flank (day 0). After randomization, mice were treated with intratumoral (i.t.) injection of poly(I:C) (Tocris Bioscience, 50 μg/mouse) or cGAMP + LF (InvivoGen, 10 μg/mouse) and PBS as a control on day 7 and 11. Two dimensions of the tumor size were measured with a digital caliper in a blinded manner. Tumor volume was calculated as specified in the formula V=length* width22 [50]. Mice were euthanized on day 15, followed by tumor isolation and qPCR analysis. Mice were housed in specific-pathogen free (SPF) conditions in the Central Animal Facility of the University of Bern. All animal experiments were performed according to institutional guidelines and approved by the Cantonal Veterinary Office. Ethical approval provided by the LANAT Amt für Landwirtschaft und Natur, Bern, Switzerland. Code: BE137/16.The mRNA expression levels for CXCL11, GBP1, and IDO1 were investigated using the RNAscope Multiplex Fluorescent Reagent Kit v2 and the RNAscope 4-Plex Ancillary Kit for Multiplex Fluorescent Kit v2, Advanced Cell Diagnostics, Inc. (ACD, Hayward, CA). The following probes were used: Hs-CD68-C4, Hs-CXCL11-C3, Hs-GBP1, Hs-IDO1-C2. For fluorescent detection, the label probes were conjugated to Opal 520 (CD68, 1:800), Opal 570 (CXCL11, 1:1500), Opal 620 (GBP1, 1:1500), and Opal 690 (IDO1, 1:800) from Akoya Biosciences/PerkinElmer (Waltham, MA, USA). The assay was performed according to the manufacturer’s instructions. Briefly, FFPE tissue slides were deparaffinizing and pre-treated before progressing onto hybridization with target probes. For the hybridization, slides were covered in a HybEZ™ Humidity Control Tray and placed in the HybEZ™ Oven and incubated at 40 °C for 2 h. Hybridization with target probes, preamplifier, amplifier, and label probes were performed according to manufacturer’s instructions. Nuclear staining was performed using DAPI. Images were acquired using the Akoya Vectra 3.0 Automated Quantitative Pathology Imaging System (P/N CLS142568) and analyzed using with Phenochart™ and inForm (Akoya Biosciences, Menlo Park, CA, USA). Adjustment of brightness and color merging of TIFF format files was performed using ImageJ.TCGA RNA-seq gene expression data as well as clinical information from all available cancer cohorts were obtained using the package GDCRNATools in R [51]. Raw counts were transformed into transcripts per million (TPM), which normalizes by gene lengths. Gene expression of Cxcl11, Gbp1, and Ido1 was used to assign a score to every patient based on a non-parametric, rank-based method implemented in the singscore R package. The score allowed for patients’ stratification into two groups according to their MA signature expression (low or high split by the sample median).Kaplan–Meier curves were plotted using the R survminer package. The survival curves were compared using the log-rank test. The M1 signature and mature DC signature scores were calculated using the singscore R package [17,18]. Correlation between the MA signature score and the M1 signature and mature DC based scores were tested using Pearson correlation. The relative proportions of 22 types of infiltrating immune cell subsets in melanoma with high vs. low MA signature expression were determined via the LM22 leukocyte signature matrix using CIBERSORT [52].Publicly available datasets (GSE2706, GSE57494, GSE1925, GSE82227, GSE60235, GSE79828, GSE85543, GSE63038, GSE107011) were obtained from gene expression omnibus (GEO) using GEOquery package. Microarrays were individually background corrected and normalized using R limma package. RNA-seq data used log2 FPKM values. Genes with multiple probes or transcripts were aggregated to the sample median. Ultimately, linear models were used to assess the association between gene expression and classical innate immune stimuli such as LPS and IFNγ, adjusting for “time” and “sample pairing” if relevant.Statistical analyses were performed using GraphPad Prism 8.0 (GraphPad Software) or R (R-project CRAN). Statistical significance was represented as described in the figure legends.In this study, we characterized a myeloid activation (MA) signature comprising the three genes Cxcl11, Gbp1, and Ido1. The MA signature indicates the activation state of myeloid cells and represents a prognostic biomarker for the overall survival in melanoma patients.The following are available online at https://www.mdpi.com/2072-6694/12/6/1431/s1, Figure S1: Correlation between CXCL11, GBP1, and IDO1 signature score and abundance of immune cells deconvoluted with CIBERSOR., Figure S2: Signature Score in melanoma patients before anti-PD-1 treatment, Table S1: Expression of MA signature genes Cxcl11, Gbp1 and Ido1 in B cells, NK cell, CD4 and CD8 T cells upon activation., Table S2: TCGA study abbreviations and corresponding study names, Table S3: Primer list.All authors have read and agree to the published version of the manuscript. Conceptualization, M.S. and N.R.; methodology, N.R. and T.G.; formal analysis, N.R., M.K. and A.A.C.; investigation, N.R., M.K. and L.B.; resources, M.S.; data curation, N.R., M.K. and A.A.C.; writing—original draft preparation, M.K. and M.S.; writing—review and editing, M.S. and D.J.L.; visualization, M.K. and A.A.C..; supervision, M.S.; project administration, M.S.; funding acquisition, M.S. This research was funded in part by the Foundation for Experimental Biomedicine Zürich, Switzerland (M.S.), the Helmut Horten Foundation (M.S.), and the Swiss National Science Foundation, SNSF 320030_176083 (M.S.).The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.Identification of a three gene signature characteristic to activated myeloid cells. (A) Dendritic cells (DC) and CD14+ monocytes were isolated from healthy donors and stimulated with poly(I:C) (10 µg/mL) or cGAMP (10 μg/mL) for 6 h in vitro. Gene expression counts were quantified using the NanoString nCounter Immunology Panel. Unsupervised t-Distributed Stochastic Neighbor Embedding (t-SNE) analysis showed significant clustering between unstimulated and stimulated cells for both monocytes and DC. (B) IDO1, CXCL11, and GBP1 expression values (RNA-seq, microarray) of unstimulated and stimulated myeloid cells using public Gene Expression Omnibus (GEO) datasets: GSE57494 (LPS + IFNγ), GSE82227 (IFNγ), GSE2706 (LPS), and GSE1925 (IFNγ) plotted as 3D scatter plots using the R package plotly and show clustering between unstimulated and stimulated cells.Expression of the myeloid activation (MA) signature genes positively correlated with M1 macrophage and ‘mature DC’ signatures in various cancer types. (A) RNA-seq data from peripheral blood mononuclear cells (PBMC) were obtained from GEO (GSE107011) and analyzed for the expression of the signature genes for every cell type available. Transcripts per million (TPM) values were scored with a non-parametric, rank-based method using the R package singscore based on the co-expression of Cxcl11, Gbp1, and Ido1. (B) Bulk tumor RNA-seq gene expression data was obtained from The Cancer Genome Atlas (TCGA ) cohorts using GDCRNATools in R. The MA signature score was compared to previously described M1 macrophage (upper panel) and mature DC (lower panel) signatures by Pearson correlation using R. X-axis represents TCGA study abbreviations.MA signature positively correlates with increased overall survival and the presence of M1 macrophages and CD8+ T cells in melanoma. (A) RNA-seq data from various tumor types were obtained from TCGA and the MA signature score was assessed with a non-parametric, rank-based method in R using the singscore package. The horizontal line indicates a p value of 0.05. (B) Kaplan–Meier survival curves of patients with high and low MA signature expression plotted as -log10 p values of log-rank tests of survival data for skin cutaneous melanoma (SKCM) with high vs. low MA signature gene expression (split by median) performed in R using the survminer package (n = 118). (C) Pearson correlation of the MA signature score and the presence of immune cell subsets in SKCM patient data obtained from TCGA. The estimated abundance of various immune cells was determined by Cell-type Identification By Estimating Relative Subsets Of RNA Transcripts (CIBERSORT). Each dot represents an individual patient. (D) Signature score in melanoma patients receiving anti-PD-1 (Nivolumab, Pembrolizumab) treatment. Patients were stratified into responders (complete response and partial response, n = 13) and non-responders (progressive disease, n = 14). Each dot represents an individual patient and only patients sampled pre- and post-treatment were included in the analysis. Dataset was obtained from GSE91061. Box plot defines the maximum, third quartile, first quartile, and minimum values. p-values were determined by two-sided Welch’s t-test (* p < 0.0332; ** p < 0.0021; *** p < 0.0002; **** p < 0.0001).The MA signature expression was significantly induced in bone marrow-derived macrophages (BMDM) but not in melanoma cells upon stimulation with poly(I:C) or cGAMP in vitro. BMDM from three biological replicates were stimulated with poly(I:C) (10 μg/mL) or cGAMP (10 μg/mL) for 6 h (A) and 24 h (B), following gene expression analysis for Cxcl11, Gbp1, and Ido1 by qPCR (n = 6). (C) B16F10 melanoma cells were cultured with poly(I:C) (10 µg/mL) or cGAMP (10 μg/mL) for 24 h following gene expression analysis by qPCR (n = 6). Data are represented as mean ± standard error of log2 transformed values. Data were normalized to the housekeeping gene Rplp0 [21]. (D) Cell viability of B16F10 and human melanoma cell lines SK-Mel-37 and D10 upon stimulation with poly(I:C) and cGAMP was assessed by AlamarBlue assay. Cell lines were incubated with poly(I:C) (10 µg/mL), cGAMP (10 µg/mL) for 24 h. After 6 h of incubation with AlamarBlue (10% v/v), fluorescence intensity was measured. Statistical analysis was performed using an unpaired, two-tailed Students t-tests and the p values are indicated as follows: p > 0.05 (ns), p ≤ 0.05 (*), p ≤ 0.01 (**), p ≤ 0.001 (***), p ≤ 0.0001 (****).Intratumoral poly(I:C) and cGAMP significantly reduced tumor growth and induced the expression of the MA signature genes in vivo. C57BL/6J wt mice were injected with 2 × 105 B16F10 melanoma cells and treated with poly(I:C) (50 μg/mouse), cGAMP in lipofectamine (10 μg/mouse), or PBS on day 7 and 11 after tumor injection. (A) Tumor size was measured using a caliper and tumor volume was calculated using the following formula: V = (length × width2)/2 (n = 12 per group). Statistical significance was calculated by a two-way ANOVA followed by Šidák’s multiple comparisons test. Data are represented as mean ± SEM and the p values are indicated as follows: p > 0.05 (ns), p ≤ 0.05 (*), p ≤ 0.01 (**), p ≤ 0.001 (***), p ≤ 0.0001 (****). (B) On day 12, tumors were isolated and analyzed for the MA signature expression by qPCR. Data were normalized to the housekeeping gene Rplp0 and are represented as mean ± SE of log2 transformed values (n = 12 per group) [21]. Statistical analysis was performed using an unpaired, two-tailed Students t-tests, and the p values are indicated as follows: p > 0.05 (ns), p ≤ 0.05 (*), p ≤ 0.01 (**), p ≤ 0.001 (***), p ≤ 0.0001 (****). (C) Fluorescent detection of RNA transcripts in human melanoma tissue. FFPE tissue section was hybridized with Opal-labeled probes for CD68 (Opal 520), CXCL11 (Opal 570), GBP1 (Opal 620), and IDO1 (Opal 690). Nuclei were counterstained with DAPI (blue). Adjustment of brightness and color merging was performed using ImageJ. Scale bar = 30 μm.Signed fold-change table of common genes that were differentially expressed upon cGAMP or poly(I:C) stimulation in both monocytes and dendritic cells.Signed fold-change table of differentially expressed genes (DEG) that were common between human DC (n = 4–5) and CD14+ monocytes (n = 4–5) upon stimulation with cGAMP or poly(I:C). Gene expression was quantified using the NanoString nCounter Immunology.p-value table from differential gene expression tests between unstimulated versus stimulated myeloid cells in RNA-seq and microarray datasets.p-value table from differentially expressed genes between stimulated and unstimulated myeloid cells derived from GEO datasets comprising RNA-seq or microarray gene expression data. Data sets include circulating myeloid DC (GSE2706), primary human CD14+/CD16+ monocytes (GSE57494), CD14+ monocytes (GSE1925), and CD14+ cells (GSE82227). Data were normalized and p-value was computed by regression analysis. Not available (NA).

Med-MDPI/cancers/cancers-12-06-01432.txt

@@ -0,0 +1 @@
+Background: To evaluate whether a model based on radiomic and clinical features may be associated with lymph node (LN) status and overall survival (OS) in lung cancer (LC) patients; to evaluate whether CT reconstruction algorithms may influence the model performance. Methods: patients operated on for LC with a pathological stage up to T3N1 were retrospectively selected and divided into training and validation sets. For the prediction of positive LNs and OS, the Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression model was used; univariable and multivariable logistic regression analysis assessed the association of clinical-radiomic variables and endpoints. All tests were repeated after dividing the groups according to the CT reconstruction algorithm. p-values < 0.05 were considered significant. Results: 270 patients were included and divided into training (n = 180) and validation sets (n = 90). Transfissural extension was significantly associated with positive LNs. For OS prediction, high- and low-risk groups were different according to the radiomics score, also after dividing the two groups according to reconstruction algorithms. Conclusions: a combined clinical–radiomics model was not superior to a single clinical or single radiomics model to predict positive LNs. A radiomics model was able to separate high-risk and low-risk patients for OS; CTs reconstructed with Iterative Reconstructions (IR) algorithm showed the best model performance.The lung represents the second most frequent site of cancer each year, and the first cause of death from cancer [1]. For early stages, when surgery is the primary option, complete lymph node (LN) excision with a microscopic evaluation is the most accurate method for determining LN metastasis. However, the best local control and decreased risk of residual lesions guaranteed by the dissection of a high number of LNs is associated with greater trauma for patients, such as prolonged air leaks, excessive chest tube drainage and prolonged hospitalization [2,3,4]. A method that could non-invasively predict the presence of LN metastasis would therefore be helpful to guide systematic dissection and might be considered, for example, in patients with small tumours and no apparent enlarged LNs [5], or in the presence of many co-morbidities.Radiomics is an emerging translational field of research, aiming to extract mineable high-dimensional data from clinical images, containing information that may reflect the underlying patho-physiology of a tissue [6]. Many recent studies have evaluated the associations between radiomics and the prognosis of various tumours, such as lung adenocarcinoma [7,8,9], renal cancer [10], hepatocellular carcinoma [11,12], glioblastoma [13], and ovarian cancer [14].Some methodological issues have been raised regarding radiomics studies, including the possibility that differences in quantitative data may depend on different parameters of acquisition and reconstruction [15]. It has also been widely demonstrated that at least internal, possibly external, and independent validation of the model is always needed [16].The purpose of this study was to evaluate whether a model based on quantitative CT radiomic and clinical features of lung cancer patients may be associated with LN status and with overall survival (OS); a secondary purpose was to assess the influence of CT reconstruction algorithms on the quantitative parameters and on the performance of the predictive models.All patients underwent anatomical resection with radical lymphadenectomy; no cases of intraoperative nodal upstaging to pN2 from cN0 or cN1 were observed.A total of 270 patients were enrolled and randomly divided into training (n = 180) and validation datasets (n = 90). The histological type was adenocarcinoma or squamous cell carcinoma in 254/270 patients; other types were present in 16/270 patients. The distribution of histological type between the training set and the validation set did not show significant differences (p = 0.16). Other baseline characteristics of the cohort are summarized in Table 1.The clinical and radiological parameters are comparable between the two datasets, although a slightly higher percentage of patients with pT > 1 was randomly allocated in the validation set (standardized mean difference = 0.21).A total of 881 radiomic features was calculated for each patient. The numbers of radiomic features which survived after elimination of identical features and selected as the most stable and reproducible after phantom analysis and Analysis of Variance (ANOVA) are reported in Figure S1 for the different cases (analysis performed on the entire Filtered Back Projection (FBP) + Iterative Reconstructions (IR) dataset and analysis performed separating the patients according to FBP or IR reconstructing algorithm).The radiomics score obtained for the prediction of positive lymph nodes (LNs) on the entire training dataset consisted of three radiomic features: ClusterShade from GLCM25 category calculated along 135° direction with four voxels offset (GLCM25135, 4_CIusterShade), 70th percentile of the intensity values in the cumulative histogram (IH_70PercentileArea), and the maximum diameter evaluated on the 3D lesion volume (Shape_Max3DDiameter). Coefficients of the Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression model for the calculation of the radiomics score are shown in Figure 1.Among clinical variables, univariable analysis in the training set showed that mixed site (meaning that the tumour showed a transfissural growth into two lobes) and nodule size (the larger, the worse) were significantly associated with positive LNs. Multivariable analysis confirmed the importance of the site (Table 2).Receiving Operating Characteristic (ROC) curves for prediction of positive LNs showed no significant differences in performance of the combined clinical–radiomics model, as compared to a single radiomics or single clinical model in the training and validation sets (Figure 2).Figure 3 shows coefficients of the LASSO logistic regression model for the radiomics score prediction of positive LNs, separately for FBP and IR (training set).In both cases, a prevalent role of the textural features is evident, whereas this was not observed when considering the entire dataset including both FBP and IR. ROC curves in the training and validation sets acquired with FBP and IR were separately evaluated (Figure S2).The overall performance of the combined model to predict LN positivity was higher for IR than FBP algorithms, with an Area Under the Curve (AUC) of 0.76 for IR compared to 0.61 for FBP in the validation set.Looking at OS prediction, in the entire training set (including both FBP and IR), the Cox regression LASSO model selected six radiomic features for the radiomics score (Figure 4).Table 3 shows parameters associated with OS in high-risk and low-risk patients, defined according to the third quartile of the radiomics score.The percentage of deaths at the second and third years in the validation set for high-risk patients was almost double that of the low-risk patients (respectively, 35% vs. 19% and 45% vs. 23%). As shown in Figure 5, a significant difference in OS for the high- and low-risk groups according to the radiomics score was observed in the training set and confirmed in the validation set.The above-described results were further confirmed after dividing the two groups according to the reconstruction algorithms (Figure 6).Clinical variables associated with OS in the model including both FBP and IR algorithms were tumour side, site and pT—tumour site and pT were associated with OS for the subgroup of patients with FBP algorithm, while only tumour site was associated with OS for the subgroup of patients with IR algorithm.Table 4 summarizes the performance of the models including radiomic only, clinical only, and clinical–radiomic parameters, on both the entire FBP+IR datasets and on datasets separated according to reconstruction algorithm.The best model performance for prediction of OS was obtained with the combined clinical–radiomic model for CT reconstructed with the IR algorithm.In this series, selected clinical and radiomic features showed association with the positivity of LN, although the combined radiomics–clinical model did not perform better than a single clinical or radiomics model. Among the clinical features, the mixed site was significant in the univariate and multivariate analysis. The importance of this feature is recognized by its role in Tumor-Nodes-Metastasis (TNM) staging, where the invasion of the visceral pleura makes the tumour belong to the T2 category, despite its size [17]. This result is concordant also with Li et al., who demonstrated a significant association between adjacent lobe invasion through pleural fissure and LN positivity [18]. The size of the tumour, measured on an axial image, was borderline-significant in the univariable and multivariable analysis. Size is a well-known feature for prognosis, as demonstrated by the sub-division of tumours in the T category according to size [17]. Its importance is confirmed in our series by the better performance of the single clinical model, after the inclusion of size as a feature. Likewise, the importance of size as a prognostic factor for LN positivity is demonstrated by one radiomic feature included in the score, the Shape_Max3DDiameter. This feature represents a measure of the maximum dimension of the lesion, evaluated in 3D, not directly related to the volume and is more precise than the maximum axial diameter. This may also account for the slightly better performance of the radiomics model compared to the clinical model, where the Max3DDiameter radiomic feature may provide a more precise definition of size than the maximum axial diameter. Other significant features in the radiomics score for LN positivity prediction were IH_70PercentileArea and GLCM25135, 4_ClusterShadeClShade. IH_70PercentileArea belongs to the Intensity Histogram category and roughly indicates that the mean Hounsfield Units (HU) number within the lesion is associated with LN status. GLCM25135, 4_ClusterShade, belonging to the GLCM category, is a measure of the global skewness. When the Cluster Shade parameter is high, the distribution of HU values is asymmetric. In our series, low values of this feature, that may be associated with intratumoral necrosis, were more frequently associated with positive LNs, whereas high values, encountered in lesions with calcifications and ground-glass opacities, were associated with negative LNs.Despite the abovementioned associations, the combined clinical and radiomics model did not perform better than a single clinical or radiomics model. This result is discordant with Tan et al., who demonstrated that, in patients with resectable oesophageal carcinoma, a radiomics nomogram provided a good risk estimation of LN metastasis and outperformed size criteria [19]. However, unlike these authors, we did not evaluate LN radiomic features, because our purpose was to predict the presence of positive LNs according to the characteristics of the primary tumour. Conversely, Yang et al. [20] demonstrated a good performance of a radiomics-based nomogram extracted from lung tumour to predict LN metastases. Although we used a similar method, by performing the radiomics analysis on the lung tumour volume, our results may be different because Yang et al. included patients with CT examinations acquired with the same parameters and reconstructed with the same algorithm [20]. In this regard, our entire sample is less homogeneous, because we included CTs obtained from different scanners and reconstructed with two different algorithms. However, it must be pointed out that from a methodological point of view, we tried to take this into account by performing a preliminary selection with ANOVA to choose features that were not significantly affected by the use of different scanners or reconstruction algorithms. This is important because reproducibility and differences in acquisitions and reconstructions are frequent issues in retrospective studies, which currently represent the majority of radiomics studies.As a consequence, the poor performance of our predictive model, including radiomic and clinical parameters, as compared to the model based on clinical variables alone, could be due to the fact that feature selection with ANOVA may have eliminated features which potentially carry relevant predictive information, but are significantly affected by the scanner or reconstruction algorithm. It is possible that, in a more homogeneous dataset, such features would survive the feature selection process and would enter the predictive model with a significant improvement in terms of performance.In order to verify this hypothesis, we performed a separate analysis, according to reconstruction algorithms: FBP, the most common analytical reconstruction method, which uses a 1D filter on the projection data before back-projecting them onto the image space [20,21], and IR, which optimizes the reconstruction with multiple iterations of forward and back projection between image and projection space. With the advances in computing technology, IR has progressively replaced FBP in routine CT practice because it makes it possible to reduce artefacts [22]. Therefore, while retrospective databases are likely to include CT images reconstructed with either FBP or IR, future studies will mainly involve IR, unless FBP is proved to be more indicated for quantitative analysis and expressly added to the clinical IR reconstruction. In this series, for association with LN status, the radiomics score associated with FBP included five quantitative features, where the most important coefficients were attributed to two different features of Correlation type, belonging to the GLCM25 category. Correlation features measure the linear dependency of voxel intensities at each point with those of neighbouring pixels at fixed distances along different directions. In our series, this feature was high for lung tumours with air bronchograms and with high heterogeneity due to contrast-enhancement (particularly in the presence of subtle vascularization within the tumour). For CT examinations reconstructed with IR, the most important coefficients were attributed to the InverseVariance feature from the GLCM2.5 category, calculated along 180° direction with one voxel offset, and to the LocalEntropyMax feature from ID category. The larger the changes in grey values, the higher the GLCM contrast; the lower the values, corresponding to an inhomogeneous texture, the higher the association with LN positivity. LocalEntropyMax (Maximum of Local Entropy) is a primary-order statistics feature measuring the entropy within the Region of Interest (ROI), and shows high values even for small areas of entropy within a single voxel. Higher local entropies were associated with LN positivity. These results confirmed that focusing on a more homogeneous dataset (as far as CT reconstruction algorithm is concerned) may make it possible to identify texture features significantly associated with LN status, not revealed in a heterogeneous dataset. We were still not able to demonstrate that, in a homogeneous dataset, the performance of a combined clinical–radiomics model is superior to a model based on clinical or radiomic data only, probably due to the fact that dividing the dataset according to a CT reconstruction algorithm considerably reduced the size of each training and validation set. Nonetheless, promising results were found in the IR dataset (Figure S2), which may deserve to be further investigated on a larger sample.The analysis of association between clinical and radiomic features with OS showed a very good performance of the radiomics score, that significantly separated patients into high-risk and low-risk groups (p < 0.0001 in the training set; p = 0.012 in the validation set). Among radiomic features, GLCM25180, 1_InformationMeasureCorrelation1, GLRLM25_LongRunLowGrayLevel Emphasis, GOH_Kurtosis and Shape_Max 3DDiameter were included in the score. GLCM25180, 1_InformationMeasureCorrelation1 belongs to the GLCM category and quantifies the degree of randomness within the tumour, in terms of entropy and statistical disorder. GLRLM25_LongRunLowGrayLevel Emphasis belongs to the GLRLM category, and quantifies runs, intended as consecutive pixels with the same grey level. GOH_Kurtosis indicates the flatness of the curve of values, without concern for spatial relationships. A prevalent direction for the voxel intensity gradient indicates that structures within the 3D volume of interest (VOI) develop along a precise direction (e.g., an intratumoral vessel). Shape_Max 3DDiameter (see features of the radiomics score for LN) was associated with OS, with high values (large lesions) associated with worse OS.This study has some limitations. The number of patients with malignant lymph nodes (pN1) in our group was relatively small (71/270; 26%) and this may account for different results compared to previous papers. However, the inclusion of patients with pN2 introduced a bias related to the neo-adjuvant chemotherapy. The CT examinations included in this study were performed over quite a long period of time (four years), and this may have affected the acquisition protocols. However, we specifically performed a preliminary ANOVA test to select stable and reproducible radiomic features. Furthermore, since the most important change in CT acquisition during the period selected was the reconstruction algorithm, we performed a separate analysis according to the algorithm used, as described. These subgroup analyses were performed on even smaller samples, which may have generated incidental results; they could therefore be considered as exploratory analyses that need to be validated in future, larger studies. Finally, segmentation was carried out by a single operator. Although this has been considered a limitation in previous studies [23], the inclusion of multiple slices to obtain a volume, instead of a single ROI, as well as the methods of extraction of the features by dedicated software may overcome this limitation.Patient selection. The Institutional Review Board approved this retrospective study (UID 2172), waiving the need for informed consent. The study population was retrospectively selected from a database of patients with lung cancer staged up to T3 N1, operated on between 01/01/2012 and 01/08/2016. Preoperative staging was performed by whole body CT and Fluorodeoxyglucose Positron Emission Tomography (FDG PET) scan; in the event of suspected cN2 disease, preoperative endobronchial ultrasound trans bronchial needle aspiration (EBUS TBNA) was performed to rule out or to confirm lymph node involvement. Inclusion criteria were: availability of pre-surgical CT at our Institution after contrast medium injection, with helical mode, 120 kVp, 2.5 mm slice thickness, 2.5 mm spacing; reconstruction performed with “Body” filter and “Standard” convolution kernel; surgery performed at our Institution; availability of histology, pathological node status (pN0; pN1), grading. Exclusion criteria were: CT performed with parameters different to those specified above; pre-operative chemotherapy.CT imaging. Examinations were randomly performed on the following CT scanners: Lightspeed Ultra, Lightspeed 16; Optima 660; Discovery CT750 HD (all GE Healthcare, Milwaukee, WI, USA). All scans were acquired in the portal venous phase and segmentations were performed on that series. All scanners implemented current modulation; Light speed 16 and Light Speed Ultra were equipped only with longitudinal z-axis modulation, while Optima 660 and Discovery CT750 also had angular xy modulation.Clinical and radiological data recording. The following data were recorded: age; gender; grading; side; site (upper, medium, lower, mixed); axial nodule size; pT; pN; CT scanner; CT reconstruction algorithm (Filtered Back Projection, FBP, or Iterative Reconstruction, IR); exposure (mAs, defined as mA_central slice * RevolutionTime * SliceThickness)/(Pitch*TotalCollimationWidth); contrast medium (Visipaque® 320, GE Healthcare; Ultravist® 370, Bayer; Iomeron 350, Bracco Imaging; Xenetix 350, Guerbet); status (alive or deceased); date of last contact or death.Lesion segmentation: On each axial CT image including the lung nodule, a radiologist traced free-hand 2D regions of interest, resulting in a 3D volume of interest (VOI) that was converted into a DICOM RT Structure format (AW Server 2.0 workstation, GE Healthcare).Radiomic feature extraction: CT images and VOI were imported into the IBEX V 1.0 β tool (Imaging Biomarker Explorer Software [24]) for the extraction of radiomic features. The “Resample_VoxelSize” pre-processing was used to resample images to the same pixel size, chosen as the value most frequently observed in the dataset (0.7 × 0.7 mm2, the values in patient images ranging from 0.59 × 0.59 mm2 to 0.98 × 0.98mm2). “Threshold_Image_MaskXF” pre-processing, that excludes voxels on the ROI edge having intensity outside a user-defined range, was used to exclude parenchyma voxels erroneously included during manual delineation (threshold: −400 Hounsfield Units). All the features available in IBEX for the following categories were extracted: Shape, Intensity Histogram (IH), Intensity Direct (ID), Grey Level Co-occurrence Matrix (GLCM) 2.5, Grey Level Run Length Matrix (GLRLM) 2.5, Neighbour Intensity Difference (NID) 2.5, and Gradient Orient Histogram (GOH). GLCM and GLRLM were calculated comparing voxel intensities along 8 different directions (0°, 45°, 90°, 135°, 180°, 225°, 270°, 315°) and with three different offsets between voxels (1, 4 and 7 voxels). The full list of features calculated by IBEX for each category is reported in [24], also including references for feature definition and formula. The [0–4096] HU interval was discretized in 256 bins for GLCM2.5 calculation, and 64 bins for the other categories.Repeatability and reproducibility: In order to select the most stable and reproducible radiomic features, we first selected stable features with Overall Concordance Correlation Coefficient (OCCC) >0.95 based on the test-retest experiments on a phantom, where identical measures in different tests are expected. We then used one-way ANOVA to assess the features’ reproducibility according to contrast medium, scanner, reconstruction algorithm, and exposure. Features with significantly different means according to at least one of the abovementioned parameters were considered not robust and excluded.Training and validation datasets: We randomly selected 2/3 of the patients as a training dataset, and 1/3 as a validation dataset.Positive LN prediction: The Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression model was used to select radiomic features associated with positive LNs. We combined selected features into a radiomics score. We assessed the predictive accuracy of the radiomics score for positive LNs by calculating the Area Under the Curve (AUC). The association of clinical variables (age, gender, side, site, and nodule size) with positive LNs was assessed with univariable and multivariable logistic regression analysis. A clinical score was obtained as a linear combination of the selected clinical variables weighted by their respective coefficients; the corresponding AUC was then calculated for both datasets. Finally, a radiomics–clinical score was obtained by applying a logistic regression multivariable model to the above-mentioned scores, and the corresponding AUC was calculated for both datasets. The AUC for the radiomics, clinical and clinical–radiomics models were compared with the DeLong test [25]. We replicated all the analyses separately on the two subgroups of patients with CT images reconstructed with FBP or IR algorithms. For each subgroup, AUCs for the radiomic, clinical and clinical–radiomics models were compared with the DeLong test [25].OS prediction: Overall survival was calculated from the date of CT to the date of death or last follow-up, whichever occurred first. The LASSO Cox regression model was used to select the radiomic features predicting OS. We combined the selected features into a radiomics score as a linear combination of the selected features weighted by their respective coefficients. The association of the radiomics score with OS was assessed in the training and validation datasets by using Kaplan–Meier survival analysis. For this, the patients were classified into high-risk or low-risk groups according to the radiomics score, by using the third quartile as the threshold. The difference in the survival curves of the high-risk and low-risk groups was evaluated by using the Log–Rank test. The predictive accuracy of the radiomics score for OS was assessed in both datasets [26]. The association of clinical variables (age, gender, side, site, nodule size, histological type, grading, pT and pN) with OS was assessed with univariable and multivariable analysis. A clinical score was obtained as a linear combination of the clinical variables weighted by their respective coefficients. Finally, a combined radiomics–clinical score was obtained by applying a Cox regression multivariable model to the above-mentioned scores, and the corresponding C-index was calculated for the clinical–radiomics model in both datasets.To assess the role of reconstruction algorithms on the models, we replicated all the analyses separately on the two subgroups, according to FBP and IR algorithms.p-values < 0.05 were considered statistically significant. The analyses were performed using SAS software (SAS Institute Inc., Cary, USA), version 9.4 and R software (http://www.Rproject.org), version 3.5.3. (R Core Team, R Foundation for Statistical Computing, Vienna, Austria). More details are provided in the Supplementary Materials.Training and validation datasets: The subjects were randomly split into training and validation groups so that 2/3rd of subjects were included in the training group and 1/3rd in the validation group. This allocation proportion is commonly used to ensure the model is trained on a sufficient number of patients, in order to obtain precise parameter estimates. From a previously published paper, this commonly used strategy was demonstrated to be close to optimal for reasonably sized datasets (n ≥ 100) with strong signals (i.e., 85% or greater full dataset accuracy) [27].Positive LN prediction: The Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression model was used to select radiomic features predicting positive LNs. First, a 10-fold cross-validation was computed to find the best lambda parameter. The regularization strength was selected as the minimum value that maximized the Area Under the receiver operating characteristics Curve (AUC). The R package “glmnet” was used to perform the LASSO logistic regression on the training dataset with standardized features. The coefficients were then returned to the original scale. We then combined the selected features into a radiomics score as a linear combination of the selected features weighted by their respective coefficients. We also assessed the predictive accuracy of the radiomics score for positive LNs both in the training and in the validation datasets by calculating the AUC. The association of clinical variables (age, gender, side, site, and nodule size) with positive LNs was assessed with univariable and multivariable logistic regression analysis. Multivariable analysis included the clinical variables with a p-value < 0.10 at univariable analysis. A clinical score was obtained as a linear combination of the selected clinical variables weighted by their respective coefficients, and the corresponding AUC was calculated for the clinical model for both the training and the validation datasets. Finally, a radiomics–clinical score for prediction of positive LNs was obtained by applying a logistic regression multivariable model to the radiomics score and the clinical score, and the corresponding AUC was calculated for the clinical–radiomics model for both the training and the validation datasets. The AUC for the radiomics, the clinical and the clinical–radiomics models were compared with the DeLong test [16].Since the measurements of most radiomic features showed dependency according to the different CT reconstruction algorithms and were excluded from the previous analysis, we replicated all the above-mentioned analyses separately on the two groups of subjects, according to FBP and IR reconstruction algorithms including all the features that proved to be repeatable after test-retest analysis. A leave-one-out cross validation was performed for both algorithms and a LASSO logistic model was implemented with the same settings as previously described. For each subgroup of patients, AUC for the radiomics, clinical and clinical–radiomics models were compared with the DeLong test [16].OS prediction: Overall survival was calculated from the date of CT to the date of death or last follow-up, whichever occurred first. We censored patients who had not died during the study period or until the date they were lost to follow-up, whichever occurred first, and we estimated their overall survival time. The LASSO Cox regression model was used to select the radiomic features that are most useful to predict OS. First a 10-fold cross-validation based on partial log-likelihood deviance was computed to find the best lambda parameter. The regularization strength was selected as the minimum value that minimizes the partial log-likelihood deviance. The R package “glmnet” was used to perform the LASSO Cox regression as previously described. We then combined the selected features into a radiomics score as a linear combination of the selected features weighted by their respective coefficients. The potential association of the radiomics score with OS was assessed in the training dataset and validated in the validation dataset by using Kaplan–Meier survival analysis. For this, the patients were classified into high-risk or low-risk groups according to the radiomics score, by using the third quartile as the threshold. The difference in the survival curves of the high-risk and low-risk groups was evaluated by using the Log–Rank test. The predictive accuracy of the radiomics score for OS was assessed in both datasets by calculating the Harrel concordance index (C-index) with 95% Confidence Intervals (CI) [17]. The association of clinical variables (age, gender, side, site, nodule size, histological type, grading, pT and pN) with OS was assessed with univariable and multivariable Cox regression analysis. Multivariable analysis included the clinical variables with a p-value < 0.10 at univariable analysis. A clinical score was then obtained as a linear combination of the selected clinical variables weighted by their respective coefficients and the corresponding C-index was calculated for the clinical model for both the training and the validation datasets. Finally, a radiomics–clinical score for prediction of OS was obtained by applying a Cox regression multivariable model to the radiomics score and clinical score, and the corresponding C-index was calculated for the clinical–radiomics model for both the training and the validation datasets.As described for positive LN prediction, we replicated all the above analyses separately on the two groups of subjects with the FBP and IR reconstruction algorithm.In conclusion, a combined clinical–radiomics model was not superior to a single clinical or radiomics model in predicting LN metastases in lung cancer patients, whereas a radiomics score was able to significantly separate high-risk and low-risk patients for OS.For the prediction of OS, the combined clinical–radiomics model demonstrated the best model performance for CT reconstructed with IR.Based on these results, in the near future, clinical prediction of OS might include a radiomics score for better precision; furthermore, in similar radiomics studies, the extraction of radiomic features should include CT reconstructed with IR.The following are available online at https://www.mdpi.com/2072-6694/12/6/1432/s1, Figure S1: Number of radiomic features selected as the most reproducible, robust and reliable, Figure S2: ROC curves for prediction of positive lymph nodes in (a) the training set with FBP algorithm; (b) the validation set with FBP algorithm; (c) the training set with IR algorithm; (d) the validation set with IR algorithm according to clinical *, radiomics and clinical–radiomics models ^. (* clinical model includes site and nodule size as independent variables; ^ For the FBP algorithm: the clinical–radiomics model includes site, nodule size and radiomics score. For the IR algorithm: the clinical–radiomics model includes site and radiomics score. Nodule size was not included because of overlapping with the radiomic feature Shape_Max3DDiameter).Conceptualization, S.R. (Stefania Rizzo), F.B. (Francesca Botta), D.O., L.R., F.P., L.S., F.D.G., M.B.; methodology, S.R. (Stefania Rizzo), F.B. (Francesca Botta), D.O., L.R., S.R. (Sara Raimondi), F.B. (Federica Bellerba), F.C., V.B.; software, F.B. (Francesca Botta); validation, F.B. (Francesca Botta), S.R. (Sara Raimondi), F.B. (Federica Bellerba), F.C., V.B., S.R. (Stefania Rizzo); formal analysis, F.B. (Francesca Botta), S.R. (Sara Raimondi), F.B. (Federica Bellerba), F.C., V.B.; investigation, S.R. (Stefania Rizzo), F.B. (Francesca Botta); resources, S.R. (Stefania Rizzo), F.B. (Francesca Botta), F.P., L.S., M.B., A.G.M.; data curation, S.R. (Stefania Rizzo), F.B. (Francesca Botta), R.M., G.P.; writing—original draft preparation, S.R. (Stefania Rizzo), F.B. (Francesca Botta), S.R. (Sara Raimondi); visualization, S.R. (Stefania Rizzo), F.B. (Francesca Botta), S.R. (Sara Raimondi), A.G.M., F.P.; supervision, S.R. (Stefania Rizzo); project administration, S.R. (Stefania Rizzo), F.B. (Francesca Botta); funding acquisition, S.R. (Stefania Rizzo), F.B. (Francesca Botta). All authors have read and agreed to the published version of the manuscript.This research received no external funding. The APC was funded by a grant from the Ministry of Health, number GR-2016-02362050.The English text was revised by Susan West.The authors declare no conflict of interest.Values of the coefficients of the Least Absolute Shrinkage and Selection Operator (LASSO) logistic regression model for the prediction of positive lymph nodes according to radiomic features (training set). The plot shows the model coefficients of the three radiomic features selected as significantly associated with lymph node status. These coefficients were used to calculate the radiomics score used to predict lymph node status in the validation set.ROC curves for prediction of positive lymph nodes in (a) the training set and (b) the validation set according to clinical, radiomics and clinical–radiomics models. The plots show the ROC curves of the three models and the associated values of the Area under the Curves (AUC).Values of the coefficients of the LASSO logistic regression model for the prediction of positive lymph nodes according to radiomic features for (a) FBP algorithm and (b) IR algorithm (training set). The plots show the model coefficients of the (a) five and (b) eight radiomic features selected as significantly associated with lymph nodes in the two subsets of patients with (a) FBP algorithm and (b) IR algorithm. These coefficients were used to calculate the corresponding radiomic scores used to predict lymph node status in the validation set for the two separate sub-samples.Values of the coefficients for Cox regression LASSO model for prediction of overall survival according to radiomic features (training set). The plot shows the model coefficients of the six radiomic features selected as significantly associated with overall survival. These coefficients were used to calculate the radiomic score used to predict overall survival in the validation set.Kaplan–Meier curves and Log–Rank test for high-* and low-risk groups according to the radiomics score (* Radiomics score higher than the third quartile (q3)).Kaplan–Meier curves and Log–Rank test for high * and low-risk groups according to the radiomics score for (a) FBP algorithm and (b) IR algorithm (* Radiomics score higher than the third quartile (q3)).Baseline characteristics of the study population.FBP = Filtered Back Projection; IR = Iterative Reconstructions; ^ Median (InterQuantile Range).Univariable and multivariable Odds Ratios for the association between clinical variables with positive lymph nodes (training set).CI = Confidence Intervals; OR = Odds Ratio; * includes radiomic score and clinical variables significantly associated with lymph nodal status at univariable analysis (p < 0.10). Significant values in bold.Overall Survival in high-risk and low-risk patients according to the radiomics score.* Radiomics score higher than the third quartile.Model performance statistics.

Med-MDPI/cancers/cancers-12-06-01433.txt

@@ -0,0 +1 @@
+These authors contributed equally to this work.The diagnosis of adenocarcinomas located in the pancreas head, i.e., distal cholangiocarcinoma (dCCA) and pancreatic ductal adenocarcinoma (PDAC), constitutes a clinical challenge because they share many symptoms, are not easily distinguishable using imaging techniques and accurate biomarkers are not available. Searching for biomarkers with potential usefulness in the differential diagnosis of these tumors, we have determined serum metabolomic profiles in healthy controls and patients with dCCA, PDAC or benign pancreatic diseases (BPD). Ultra-high-performance liquid chromatography coupled to mass spectrometry (UHPLC-MS) analysis was performed in serum samples from dCCA (n = 34), PDAC (n = 38), BPD (n = 42) and control (n = 25) individuals, divided into discovery and validation cohorts. This approach permitted 484 metabolites to be determined, mainly lipids and amino acids. The analysis of the results led to the proposal of a logistic regression model able to discriminate patients with dCCA and PDAC (AUC value of 0.888) based on the combination of serum levels of nine metabolites (acylcarnitine AC(16:0), ceramide Cer(d18:1/24:0), phosphatidylcholines PC(20:0/0:0) and PC(O-16:0/20:3), lysophosphatidylcholines PC(20:0/0:0) and PC(0:0/20:0), lysophosphatidylethanolamine PE(P-18:2/0:0), and sphingomyelins SM(d18:2/22:0) and SM(d18:2/23:0)) and CA 19-9. In conclusion, we propose a novel specific panel of serum metabolites that can help in the differential diagnosis of dCCA and PDAC. Further validation of their clinical usefulness in prospective studies is required.Although distal cholangiocarcinoma (dCCA) and pancreatic ductal adenocarcinoma (PDAC) share a close anatomical location, they are considered distinct entities and require specific management strategies [1]. Whereas dCCA is an aggressive malignancy that arises in the biliary tract below the cystic duct and represents approximately 20% of CCAs, PDAC derives from the epithelium of pancreatic ducts and is the fourth cause of cancer-related deaths [2,3]. Although dCCA has a poor clinical outcome [4] due to its late diagnosis and resistance to chemotherapy [5], in general, the prognosis is worse in the case of PDAC [3].Despite improvements in imaging techniques during recent years, the accurate diagnosis of adenocarcinomas located in the pancreas head area represents a clinical challenge in gastrointestinal oncology. Biopsy, either using cytologic brushing or fine-needle aspiration guided by endoscopic ultrasound, is mandatory to confirm the diagnosis. However, this has serious limitations: (i) repeat sampling is often required since the quality of the samples is not always sufficient to carry out the anatomopathological analysis, and (ii) the detection of malignant cells can confirm the diagnosis, but a negative result does not permit ruling it out [6]. To distinguish PDAC from benign pancreas diseases (BPD), such as chronic pancreatitis or pancreatic cysts, is also challenging, and the lack of accurate tumor biomarkers justifies that ≈ 5–10% of surgical removals of the head of the pancreas due to presumed malignancies are finally identified as benign lesions.Several non-invasive biomarkers have been evaluated for the diagnosis of PDAC [7] and CCA [8,9], but none of them are being used in the clinical setting. Serum carbohydrate antigen 19-9 (CA 19-9) is the only FDA-approved biomarker for PDAC for both the follow-up of the therapeutic response [10] and for the detection of recurrence after surgery. Nevertheless, owing to its low sensitivity and specificity, CA 19-9 is far from being considered an optimal biomarker. Serum CA 19-9 is also used clinically to help in diagnosis and to monitor the response to therapy in biliary cancers, usually in combination with another unspecific marker, i.e., carcinoembryonic antigen (CEA). However, its accuracy is low and is not suitable for early detection. In addition, CA 19-9 can be elevated in patients with obstructive cholestasis, chronic liver and pancreatic diseases, and premalignant pancreatic lesions. Moreover, ≈ 10% of the Caucasian population with Lewis-negative phenotype do not express this biomarker [11].Therefore, there is an urgent need to identify reliable minimally invasive biomarkers that can help in the differential diagnosis of dCCA and PDAC. An optimal biomarker would also be expected to contribute to the early detection of these cancers. The analysis of a large number of small metabolites in biological samples represents an interesting approach for identifying clinically relevant biomarkers for different diseases. In this context, the aim of the present study was to evaluate the usefulness of differences in serum metabolomic profiles between dCCA and PDAC, as well as between these severe malignancies and BPD and healthy individuals.The demographic and clinical features of individuals from both cohorts are shown in Table 1. The age was higher in patients with dCCA and PDAC than in patients with BPD and healthy individuals and only the latter group included a lower percentage of males. Most tumors included in the dCCA group were in early stage, while there was a similar distribution of tumors in early and advanced stage in the PDAC group. Regarding liver biochemical parameters (Table 1), a significant increase in ALT, GGT, alkaline phosphatase and total bilirubin was found in patients with dCCA and PDAC. Except for total bilirubin, these parameters were also found to be elevated in BPD, although the magnitude of changes was lower than that observed in patients with tumors. A significant increase in serum levels of CA 19-9 was found in both dCCA and PDAC, with a marked interindividual variability. Moreover, although CA 19-9 levels were also elevated in some patients with BPD, both with pancreatic cysts and with chronic pancreatitis, these were significantly lower than those found in patients with cancer.Any clustering of the different groups of samples according to the serum metabolome was evaluated using multivariate data analysis, unsupervised principal component analysis (PCA) and supervised orthogonal partial least-squares to latent structures discriminant analysis (OPLS-DA) approaches. As shown in Figure 1, no differences in serum metabolomic profiles were found between the hospitals of origin, discovery and validation cohorts, gender, and group of age or group of samples (Figure 1A–E, respectively). A random distribution of patients with cysts and pancreatitis, both included in the BPD group, was found (Figure 1F). The supervised OPLS-DA model showed a good predictive ability to discriminate patient groups from healthy individuals, since Q2X = 0.694 (Figure 2A), triglycerides and, to a lesser extent, oxidized fatty acids and bile acids (all of them increased) and sphingomyelins and glycerophosphatidylcholines (both decreased) being the main contributors to the differences found between patients and control individuals. However, the supervised OPLS-DA models to differentiate dCCA vs. BPD patients, PDAC vs. BPD and both types of tumors showed very low predictive ability (Figure 2B–D, respectively), since Q2X values were low, especially in the comparisons of PDAC with BPD (Q2X = 0.163) and dCCA (Q2X close to 0).During the discovery phase we were able to determine 484 metabolites in serum samples, which was confirmed in the validation cohort. Changes in the levels of molecules belonging to the different families of analyzed metabolites (lipids, amino acids and amino acids derivatives) were found. Figure 3 depicts the heatmaps showing the fold-changes and the p-values generated from different two-groups comparisons carried out in the discovery and validation cohorts, and considering all samples together.Figure 4 shows the volcano plots generated for each two-groups comparison, and the number of metabolites significantly changed in each comparison considering the full cohort (Figure 4G). When BDL was compared with control, altered serum concentrations of 268 metabolites (mainly phosphatidylcholines > triglycerides > sphingomyelins ≈ lysophosphatidylcholines) were found. The comparison of dCCA with control revealed altered serum levels of 236 metabolites (mainly triglycerides ≈ phosphatidylcholines > lysophosphatidylcholines > sphingomyelins). The highest number of metabolites affected by changes in their serum levels (n = 280; mainly triglycerides > phosphatidylcholines > lysophosphatidylcholines) was found in the PDAC group. Different serum levels of 111 metabolites were found when comparing dCCA with BPD (mainly phosphatidylcholines > lysophosphatidyletanolamines = sphingomyelins), whereas this number increased to 178 when comparing PDAC with BPD (mainly phosphatidylcholines > triglycerides). The number of serum metabolites altered when comparing dCCA vs. PDAC was 63 (mainly triglycerides > phosphatidyletanolamines > lysophosphatidyletanolamines), and most of them were higher in PDCA than in dCCA. The number of metabolites with a value of area under the receiver operating characteristic curve (AUC) ≥ 0.8 was 73 when comparing BPD vs. control, 63 when comparing dCCA vs. control and 72 when comparing PDAC vs. control. An important number of metabolites were found altered in the serum of more than one group of patients, although the magnitude of changes was higher in patients with cancer. Table 2 shows the 10 metabolites with the best diagnostic capacity (best values of AUC, sensitivity and specificity) for each disease vs. control. Complete panels are presented in Table S1A–C. Although fewer alterations in the circulating metabolomic profiles were observed when the different diseases were cross compared, we found changes with interest in diagnosis. Among 50 metabolites with significant AUC values in the comparison of dCCA vs. BPD, 6 showed AUC values of ≥ 0.8 (Table 3), while 2 among 61 in the comparison PDAC vs. BPD reached these AUC values. In the comparison dCCA vs. PDAC, 9 metabolites showed significant AUC values, although all with AUC < 0.8. In the last comparison serum concentrations of the 9 metabolites were lower in dCCA than in PDAC. Table 3 shows the 9–10 metabolites with the best diagnostic capacity in each two-groups comparison, and the complete panels are presented in Table S1D,E.In our study, with a cut-off fixed in 37 IU/mL, CA 19-9 showed a good diagnostic capacity to differentiate patients with tumors (dCCA+PDAC) from healthy individuals, with an AUC of 0.93 in both cohorts. However, as shown in Figure 5A, it was not so good in differentiating between dCCA+PDAC and patients without cancer (Control+BPD). AUC was 0.845, 0.820 and 0.828 in discovery, validation and the whole cohort, respectively (Figure 5B).As shown in Figure 6, a model including 10 metabolites [amino acids sarcosine, tryptophan and aspartic acid, lysophosphatidylethanolamine PE(0:0/16:0), phosphatidylinositol PI(18:0/18:2), diglycerides DG(38:4) and DG(34:0), sphingomyelin SM(42:1), N-acyl ethanolamine NA(16:0) and sterol pregnenolone sulfate] was generated to differentiate patients with tumors (dCCA+PDAC) and without malignancies (Control+BPD). Using this model, the probability of diagnosing patients with chronic pancreatitis or healthy subjects as individuals suffering from dCCA or PDAC is low. However, this risk is higher for patients with benign pancreatic cysts (Figure 6A). AUC was 0.93 in discovery, 0.86 in validation and 0.89 considering the whole cohort. Sensitivity was 73.6% and specificity 83.6% considering the whole cohort. We have evaluated the relationship between the age and the diagnostic error rate of the model. Based on a stratification of the patients in quantiles, the diagnostic error rate was constant and around 20% (average 21%, ranging from the 17% to 29%) and was not associated with the patient’s age.In our study, CA 19-9 showed a sensitivity of 71% and a specificity of 83% to differentiate patients with tumors from individuals without tumors (Controls+BPD).Since none of the individual circulating metabolites had a sufficient capability of distinguishing dCCA from PDAC (Table 3), our next goal was to obtain a predictive model for discriminating between both tumors. A logistic regression model was built with nine metabolites (Figure S1) (acylcarnitine AC(16:0), ceramide Cer(d18:1/24:0), phosphatidylcholines PC(20:0/0:0) and PC(O-16:0/20.3), lysophosphatidylcholines PC(20:0/0:0) and PC(0:0/20:0), lysophosphatidylethanolamine PE(P-18:2/0:0), and sphingomyelins SM(d18:2/22:0) and SM(d18:2/23:0)), with an AUC of 0.91 in discovery, 0.82 in validation and 0.85 considering the whole cohort; sensitivity was 55.9% and specificity 89.5% considering all the patients. The analysis of CA 19-9 showed a sensitivity of 77% and a specificity of 48% to differentiate patients with PDAC from those with dCCA (Figure S2).Another logistic regression model was built with the nine metabolites plus CA 19-9 (Figure 7), which improved the sensitivity. However, the specificity slightly decreased in the full cohort and especially in the validation cohort. Thus, AUC was 0.888, sensitivity 71.4% and specificity 89.2 considering the whole cohort.The lack of non-invasive biomarkers for the early diagnosis of PDAC and dCCA contributes to the bad prognosis of these tumors [12]. The anatomical difficulty in accessing the tumors to obtain good quality biopsies for diagnostic purposes makes it necessary to identify minimally invasive biomarkers that could help, not only in the early detection of these tumors to enable more patients to benefit from surgical treatment, but also in the prognosis and follow-up of these patients during treatment. However, although important efforts have been made during recent years, none of the identified markers have been validated and reached clinical practice. Despite their moderate clinical utility, only CA 19-9 and carcinoembryonic antigen (CEA) are currently used for PDAC and CCA diagnosis [13].Omics technologies are providing valuable information to understand cancer biology. Metabolic reprogramming is one hallmark of tumor cells [14,15]; thus, the analysis of the metabolome (hundreds of small molecules or metabolites) in body fluids of patients with cancer can give an indirect reflection of the metabolic behavior of the tumors and could be used to identify potential biomarkers. Several studies have been conducted to identify serum metabolomic profiles for the diagnosis of pancreatic or biliary cancers. Most of them included only patients with pancreatic cancer and healthy controls [16,17,18] or with biliary cancer and healthy individuals [19]. However, it is important to include clinically relevant controls since the metabolome can be affected by many factors, including gender, age, comorbidities, medication, life style, environment or circadian rhythms; in fact, important intra-day variations have been observed in serum levels of patients with advanced pancreatic cancer, which were further affected by cachexia [20].The use of metabolomics to discriminate between different types of tumors and between tumors and benign diseases has been less explored. Combinations of metabolites discriminating malignant from benign pancreaticobiliary diseases and from healthy controls have been reported, although the number of cases was low and most of the patients with tumors were in an advanced stage, for which their usefulness in early diagnosis cannot be guaranteed [21]. More recently, a biomarker signature for the differential diagnosis between PDAC and chronic pancreatitis was reported, consisting of nine metabolites, five of them lipids (two sphingomyelins, sphinganine 1-phosphate, one phosphatidylcholine and one ceramide), and proline, histidine, pyruvate and isocitrate plus CA 19-9, with a negative predictive value of 99.9% in patients with chronic pancreatitis [22].All these studies support the concept that the combination of several metabolite markers allows for a more accurate diagnosis. In this study, we have included patients with biopsy-proven tumors or cysts located in the head of the pancreas divided into two independent cohorts of PDAC, dCCA, BPD and controls. Although serum bile acids levels represented the most marked alteration in patients with cancer, this hypercholanemic condition occurs in different pathologies that are accompanied by cholestasis, in which compensatory mechanisms are developed to limit the accumulation and toxic effects of these compounds [23]. It has been demonstrated that obstructive jaundice impacts the performance of biomarkers for PDAC [24], and in our study, a certain degree of cholestasis was found in some patients with tumors, since serum bilirubin was elevated, and as a consequence, none of the bile acid species measured could be considered as a good biomarker.In the present study we have identified a multimarker signature for the differential diagnosis of adenocarcinomas located in the pancreas including nine metabolites plus CA 19-9 with better performance than serum CA 19-9 alone and another panel of ten metabolites (seven lipids and three amino acids) with similar performance to serum CA 19-9 to discriminate tumors from BPD but which are especially useful for chronic pancreatitis. Since this disease is a risk factor for the development of pancreatic cancer [25], these biomarkers could be useful for early detection of tumor development, for monitoring patients during treatment and for avoiding unnecessary pancreatic surgery and its complications. Interestingly, some of the metabolites included in the signature proposed here belonged to the same families of compounds (amino acids, sphingomyelins and ceramides) of a previously described model [22]. Changes in serum levels of certain amino acids have been described in other tumors, such as liver [26,27] and breast [28] cancer. In addition, sphingomyelins and ceramides have been found altered in the serum of patients with liver [27] and ovarian [29] cancer. Alterations in sphingolipid metabolism have been associated with cell proliferation [30]. Our model of changes in ten metabolites was less accurate than CA 19-9 levels in distinguishing pancreatic cysts from tumors in the head of pancreas, although the low number of cases of cystic lesions in our cohort can be considered a limitation. In recent years, several studies have proposed circulating microRNA (miRNA) signatures for early detection of pancreatic cancer [31] or for the differential diagnosis of PDAC and chronic pancreatitis with good sensitivity and specificity [32], although none of them included a group of patients with pancreatic cysts. A recent study proposed a two-miRNA panel of downregulated miR-16 and upregulated miR-877 to differentiate patients with dCCA from benign disease (AUC = 0.90) and from PDAC (AUC = 0.88) [33]. Serum proteins have also been investigated. The analysis of cell migration-inducing hyaluronan binding protein (CEMIP) plus CA 19-9 improved the diagnostic value compared to CA 19-9 alone for the diagnosis of pancreatic cancer [34]; the study included a small but very heterogeneous group of patients with BPD in the control cohort, but the results must be validated.In sum, in this study, using two independent cohorts of patients, we have identified a model consisting of 9 metabolites in serum with promising capability to differentiate both types of pancreatic head adenocarcinomas, with AUC = 0.854. Because accurate diagnosis of these tumors remains challenging, our results suggest that the analysis of multiple types of biomarkers could help in the early and differential diagnosis and in the follow-up of these aggressive tumors.Fasting serum samples from dCCA (n = 34), PDAC (n = 38), BPD (n = 42), and healthy subjects (n = 25) were obtained from two Spanish hospitals; University Hospital of Salamanca, National DNA Bank Carlos III, and Donostia University Hospital in San Sebastian. Samples were randomly divided in two cohorts, “discovery” and “validation”, with equal proportional representation of individuals belonging to each pathology as well as to each origin of samples. Inclusion criteria for patients with dCCA and PDAC were histopathologic confirmation of diagnosis by expert pathologists and serum obtained before any type of treatment. Exclusion criteria were other types of CCA or synchronous presence of another type of malignancy. The BPD group included 22 samples from patients with cysts and 20 from patients with chronic pancreatitis. Selected healthy individuals had no history of any type of malignancy and no clinical evidence of hepatopancreaticobiliary disease. Clinical and laboratory test values were collected from the patients’ records. The research protocol was approved by the Ethics Committee for Clinical Research of Salamanca (July 18, 2018) and San Sebastian (October 16, 2019), and informed written consent for the samples to be used for biomedical research was obtained from each patient.Serum metabolic profiles were analyzed as previously described [35]. Briefly, two ultrahigh-performance liquid chromatography (UHPLC)-time of flight-MS based platforms analyzing methanol and chloroform/methanol serum extracts were combined with the amino acid measurement using an UHPLC-single quadrupole-MS based analysis. Identified ion features in the methanol extract platform included amino acids and its derivatives and lipids.Metabolite extraction procedures, chromatographic separation conditions and mass spectrometric detection conditions have been previously described [35]. Metabolomics data were pre-processed using the TargetLynx application manager for MassLynx 4.1 (Waters Corp., Milford, MA, USA). Intra- and inter-batch normalization was performed by inclusion of multiple internal standards and pool calibration response correction, following a previously described procedure [36]. Data quality was assessed by the inclusion of quality control samples, including repeated injections of these samples to evaluate the reproducibility of the analysis process [36].Data are shown as mean ± SD. Differences between groups were determined using the Student´s t-test or the Bonferroni method of multiple range test, as appropriate. Calculations were performed using the statistical software package R v.3.4.0 (R Development Core Team, 2017; http://cran.r-project.org).Multivariate principal component analysis (PCA) [37] and orthogonal partial least squares discriminant analysis (OPLS-DA) [38] modeling were performed with the software SIMCA 14.1 (Umetrics, Malmo, Sweden). Model quality was assessed using R2 and Q2 values, which indicate the explained fraction of variance and the goodness of prediction, respectively. The Q2 parameter was calculated by sevenfold cross validation.To find statistical models to differentiate patients with tumors (dCCA or PDAC) and subjects without tumors (controls or BPD [chronic pancreatitis or pancreatic cysts]), as well as to differentiate each type of tumor, dCCA vs. PDAC, generalized linear models (GLM) were used and those selected were confirmed by leave-one-out cross validation (LOOCV). Box-Cox transformations were applied to the biomarker metabolite levels for correcting non-normally distributed data and used to calculate the classification algorithm. The diagnostic accuracy of the model to identify patients in each comparison was assessed using the AUC p < 0.05.Based on the results obtained in the present study, we propose novel specific panels of serum metabolites that can help in the early and differential diagnosis of dCCA and PDAC. Further validation of their clinical usefulness in prospective studies including other relevant controls and in combination with clinical features is required.The following are available online at https://www.mdpi.com/2072-6694/12/6/1433/s1, Figure S1: Diagnostic prediction capacity of the model of nine metabolites in dCCA vs. PDAC, Figure S2: Diagnostic prediction capacity of CA 19-9 in dCCA vs. PDAC, Table S1: Diagnostic capacity of the metabolites in the comparison.Conceptualization, R.I.R.M., J.J.G.M., J.M.B., L.M.-B. and L.B.; methodology, R.I.R.M., J.J.G.M., A.S.-M., E.A., I.M.-A., C.A.; software, E.A., I.M.-A., C.A.; data curation, R.I.R.M., L.M.-B., A.L., A.L.C., L.M.G., M.L.G.; writing—original draft preparation, R.I.R.M. and J.J.G.M.; writing—review and editing, all authors; visualization, all authors; project administration, R.I.R.M.; funding acquisition, R.I.R.M., J.J.G.M., J.M.B., M.A.A., M.L.M.-C. All authors have read and agreed to the published version of the manuscript.This study was supported by the Centro Internacional sobre el Envejecimiento, Spain (OLD-HEPAMARKER, 0348_CIE_6_E) co-financed with European Union ERDF funds; Carlos III Institute of Health, Spain (PI16/00598, PI16/01126, PI18/01075, PI19/00819) and Miguel Servet Program (CON14/00129) co-financed by European Regional Development Fund; Asociación Española Contra el Cancer, Spain (AECC-Cánceres raros 2017/2020); H2020 ESCALON project: H2020-SC1-BHC-2018-2020; Fundacion La Caixa (Hepacare Project); MCIU/AEI/FEDER, EU (SAF2017-87301-R); Severo Ochoa Excellence Accreditation (SEV-2016-0644). A. Sanchez-Martin and A. Lapitz were supported by pre-doctoral scholarships funded by the Ministry of Science, Innovation and Universities (FPU17/04027) and the Basque Government (PRE_2017_1_0345), respectively, and M.L. Gutiérrez is supported by the "Stop fuga de Cerebros" grant from ROCHE FARMA SA. This work was carried out in the framework of Working Group 5 of the COST Action CA18122, European Cholangiocarcinoma Network, EURO-CHOLANGIO-NET.Enara Arretxe and Cristina Alonso and Ibon Martínez-Arranz are employed by OWL Metabolomics and Jesus M. Banales is a member of the scientific advisory board of OWL Metabolomics.Principal component analysis (PCA) score plots of human serum samples. Colors represent (A) the origin of the samples, (B) discovery or validation cohort, (C) gender, (D) age range, (E) group of samples and (F) type of benign pancreatic disease (BPD). (A–E) Principal component 1 (t[1]) and principal component 2 (t[2]) explain 17.4% and 11.2% of the total variance, respectively. (F) t[1] and t[2] explain 19.8% and 13.1% of the total variance, respectively. Each dot represents one sample. The ellipse represents 95% confidence interval according to Hotelling’s T2 test.Score plots for the first predictive (t[1]) and orthogonal (to[1]) components of the supervised orthogonal partial least squares discriminant analysis (OPLS-DA) models for (A) disease vs. control samples; R2X = 0.445; R2Y = 0.835; Q2X = 0.694), (B) dCCA vs. BPD samples; R2X = 0.315; R2Y = 0.697; Q2X = 0.425, (C) PDAC vs. BPD samples; R2X = 0.265; R2Y = 0.471; Q2X = 0.163,and (D) dCCA vs. PDAC samples; R2X = 0.24; R2Y = 0.527; Q2X = 0.036. Each dot represents one sample. The ellipse represents 95% confidence interval according to Hotelling’s T2 test.Metabolomic signatures in serum of patients with dCCA, PDAC or BPD and healthy individuals (Control). Heatmaps show fold-changes and p-values in each two-group comparison of the relative metabolite levels in serum samples in the discovery cohort (Disc.), in the validation cohort (Val.) and considering all samples together (All). The log2 transformed metabolite abundance ratios are depicted for each comparison. In the scale, colors from green to red correspond to drop or elevation of metabolite levels and gray lines show significant fold-changes of individual metabolites; darker gray colors indicate higher significance. Metabolites are grouped by chemical group/subgroup: AA, amino acids; AC, acylcarnitines; BA, bile acids; Cer, ceramides; ChoE, cholesteryl esters; CMH, monohexosylceramides; DAPC, diacylglycerophosphocholines; DAPE, diacylglycerophosphoethanolamines; DG, diglycerides; FSB, free sphingoid bases; LPC, lysophosphatidylcholines; LPE, lysophosphatidylethanolamines; LPI, lysophosphatidylinositols MAPC, monoacylglycerophosphatidylcholines; MAPE, monoacylglycerophosphatidylethanolamines; MEMAPC, 1-ether, 2-acylglycerophosphatidylcholines; MEMAPE, 1-ether, 2-acylglycerophosphatidylethanolamines; MEPC; 1-monoetherglycerophosphatidylcholines; MEPE, 1-monoetherglycerophosphatidylethanolamines; MUFA, monounsaturated fatty acids; NAE, N-acyl ethanolamines; oxFA, oxidized fatty acids; PC, phosphatidylcholines; PE, phosphatidylethanolamines; PI, phosphatidylinositols; PUFA, polyunsaturated fatty acids; SFA, saturated fatty acids; SM, sphingomyelins; ST, steroids; TG, triglycerides.Volcano plots [-log10(p-value) and log2(fold-change)] considering the serum metabolite levels of the whole cohort of (A) BPD patients vs. controls, (B) dCCA vs. controls, (C) PDAC vs. controls, (D) dCCA vs. BPD, (E) PDAC vs. BPD and (F) dCCA vs. PDAC. (G) Number of metabolites and metabolite classes significantly different in each comparison. AA, amino acids; AC, acylcarnitines; BA, bile acids; Cer, ceramides; ChoE, cholesteryl esters; CMH, monohexosylceramides; DG, diglycerides; FSB, free sphingoid bases; LPC, lysophosphatidylcholines; LPE, lysophosphatidylethanolamines; LPI, lysophosphatidylinositols; MUFA, monounsaturated fatty acids; NAE, N-acyl ethanolamines; oxFA, oxidized fatty acids; PC, phosphatidylcholines; PE, phosphatidylethanolamines; PI, phosphatidylinositols; PUFA, polyunsaturated fatty acids; SFA, saturated fatty acids; SM, sphingomyelins; ST, steroids; TG, triglycerides.Diagnostic prediction capacity of CA 19-9 in tumors (dCCA+PDAC) vs. non tumors (Control+BPD). (A) Box plot diagrams show the log10 CA 19-9 (cut-off of 37 IU/mL). (B) Area under the receiver operating characteristic curve (AUC) in discovery and validation cohorts and considering all cohorts.Diagnostic prediction capacity of the logistic model in tumors (dCCA+PDAC) vs. non tumors (Control+BPD). (A) Box plot diagrams showing the probability to detect each group as tumors. (B) Area under the receiver operating characteristic curve (AUC) in discovery and validation cohorts and considering all cohorts. (C) AUC, sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of the algorithm to differentiate tumors vs. non tumors in each cohort. (D) Selected metabolites included in the model. AA, amino acids; DG, diglycerides; LPE, lysophosphatidylethanolamines; NAE, N-acyl ethanolamines; PI, phosphatidylinositols; SM, sphingomyelins; ST, steroids.Diagnostic prediction capacity of the logistic model in dCCA vs. PDAC. (A) Box plot diagrams showing the probability to detect each type of tumor. (B) Area under the receiver operating characteristic curve (AUC) in discovery and validation cohorts and considering the whole cohort. (C) AUC, sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of the algorithm to differentiate dCCA vs. PDAC in each cohort. (D) Selected metabolites included in the model: AC, acylcarnitine; Cer, ceramide; LPC, lysophosphatidylcholines; LPE, lysophosphatidylethanolamines; PC, phosphatidylcholines; SM, sphingomyelins plus CA 19-9. ***, p < 0.001.Demographic and clinical characteristics of the discovery and validation cohorts.a, p < 0.05 compared with control (in the same cohort) and b, p < 0.05 compared with BPD (in the same cohort) using the Bonferroni method of multiple range test. *, AJCC Cancer Staging Manual, 7th Edition. ALT; alanine aminotransferase, BPD, benign pancreatic disease; CA 19-9, carbohydrate antigen 19-9; dCCA, distal cholangiocarcinoma; GGT, gamma-glutamyl transpeptidase; PDAC, pancreatic ductal adenocarcinoma.Diagnostic capacity of the top 10 metabolites in the comparison of each disease vs. control considering the whole cohort.AUC, area under the receiver operating characteristic curve; FC, fold change. Colors from green to red correspond to drop or elevation of metabolite levels.Diagnostic capacity of the top 9-10 metabolites in each two-disease group comparison considering the whole cohort.AUC, area under the receiver operating characteristic curve; FC, fold change. Colors from green to red correspond to drop or elevation of metabolite levels.

Med-MDPI/cancers/cancers-12-06-01434.txt

@@ -0,0 +1 @@
+An important drawback in the management of glioblastoma (GBM) patients is the frequent relapse upon surgery and therapy. A likely explanation is that conventional therapies poorly affect a small population of stem-like cancer cells (glioblastoma stem cells, GSCs) that remain capable of repopulating the tumour mass. Indeed, the development of therapeutic strategies able to hit GSCs while reducing the tumour burden has become an important challenge to increase a patient’s survival. The signal transducer and activator of transcription-3 (STAT3) has been reported to play a pivotal role in maintaining the tumour initiating capacity of the GSC population. Therefore, in order to impair the renewal and propagation of the PDGFRβ-expressing GSC population, here we took advantage of the aptamer–siRNA chimera (AsiC), named Gint4.T-STAT3, that we previously have shown to efficiently antagonize STAT3 in subcutaneous PDGFRβ-positive GBM xenografts. We demonstrate that the aptamer conjugate is able to effectively and specifically prevent patient-derived GSC function and expansion. Moreover, because of the therapeutic potential of using miR-10b inhibitors and of the broad expression of the Axl receptor in GBM, we used the GL21.T anti-Axl aptamer as the targeting moiety for anti-miR-10b, showing that, in combination with the STAT3 AsiC, the aptamer–miR-10b antagonist treatment further enhances the inhibition of GSC sphere formation. Our results highlight the potential to use a combined approach with targeted RNA therapeutics to inhibit GBM tumour dissemination and relapse.Glioblastoma (GBM) is the most common primary brain tumour with a very dismal prognosis despite advances in surgical and medical neuro-oncology [1]. GBM is an infiltrating and highly heterogeneous tumour that contains a small population of stem-like cells with an undifferentiated phenotype that retains stemness properties, including the ability to undergo self-renewal by symmetric cell division and differentiate by asymmetric division, repopulating the bulk tumour mass [2,3]. This population is characterized by an enhanced capacity to initiate tumour formation in vivo and resistance to conventional therapies, being considered as mainly responsible for tumour propagation and recurrence [4,5,6,7]. Therefore, the development of therapeutic options able to target the resistant stem-like cells represents an important challenge to generate effective approaches able to render the tumours unable to maintain themselves or grow. Indeed, the deregulated activity and expression of few transcription factors and microRNAs (miRs), including STAT3 and miR-10b, is critical to initiate and sustain the GSC population [8,9].In the central nervous system, STAT3 is involved in several processes, such as early development and embryonic stem cell biology [10,11]. This factor is activated by various cytokines and growth factors’ signal cascades and, upon tyrosine phosphorylation, moves into the nuclei where it regulates the expression of a wide range of genes involved in the cell cycle, survival, angiogenesis and immune response. STAT3 abnormal activation has been reported to be involved in the progression of several cancer types, including GBM [12,13,14,15,16]. In addition, its inhibition resulted in an effective alteration of GSC sphere formation and stem-like growth potential [17,18,19], and the pathway has been showed to play a crucial role in GSC chemo and radio-resistance [20,21,22]. These studies indicate STAT3 as a highly promising therapeutic target for GBM able to affect both bulk tumour cells and resistant GSCs, enhancing the success of the treatment. Further, a critical role in GBM has been demonstrated for miR-10b, which is highly expressed in this tumour and acts as an oncomiR to promote cancer stem cell propagation [9,23,24]. Thus, its inhibition shows a powerful therapeutic potential. We and others have recently described the use of aptamer-based RNA molecules able to selectively drive a small interfering RNA (siRNA) against STAT3 to GBM cells [25,26]. Aptamers are short oligonucleotides able to bind with high affinity and specificity to their targets by acquiring a structured folding. They hold great promise as antagonists of cancer-associated proteins as well as delivery carriers of secondary reagents to target cells [27,28]. Indeed, aptamers against cell surface receptors may inhibit the receptor signalling and be internalized into the cell cytoplasm in a receptor-mediated manner. The last function permits their successful application as delivery vehicles of different therapeutic cargoes, including anti-cancer drugs, toxins, and siRNA or miRNA molecules [29,30,31]. This allows the cargos’ action to be restricted to receptor-expressing target tissues with a consequent reduction of unwanted off-target effects.In our previous report [25], we used a nuclease-resistant internalizing RNA aptamer, named Gint4.T, to bind and inhibit the platelet-derived growth factor β receptor (PDGFRβ) [32], and then designed an AsiC (Gint4.T-STAT3) for the delivery of a STAT3 siRNA to GBM cells, inhibiting tumour cell growth. Given the key role of STAT3 in GSC propagation and the importance in targeting this population for effective anti-cancer therapies, here our primary objective was to address the functional characterization of Gint4.T-STAT3 on GSCs. Indeed, PDGFRβ is frequently overexpressed in GBM and it is preferentially associated with the self-renewing GSCs [33,34]. We thus hypothesized that Gint4.T-STAT3 could have been used to reduce STAT3 levels in GSCs and alter their function. We demonstrate that the AsiC efficiently deliver STAT3 siRNA to PDGFRβ-positive patient-derived GSC primary cell lines, hampering cell survival and migration. Further, we explored the therapeutic potential on GSCs of STAT3 and miR-10b combined inhibition. We used as a targeting moiety for a miR-10b antagonist (anti-miR-10b), the GL21.T aptamer, an inhibitor ligand specific for the receptor tyrosine kinase (RTK) Axl [35], which is expressed in several tumours, including GBM, and also implicated in GSC maintenance [36]. We show that the combined treatment of Gint4.T-STAT3 and GL21.T–anti-miR-10b complexes drastically abrogates the propagation of GSCs.In order to determine whether the Gint4.T-STAT3 chimera (Figure S1a) inhibits the propagation of the stem-like cells population, as a first attempt we evaluated the targeting efficacy in patient-derived primary human GSCs. We selected three well-characterized primary GBM-derived neurospheres and cultured them as previously reported [37,38,39,40,41], named GSC#83, GSC#61, and GSC#1. All the three lines are positive for the PDGFRβ aptamer target (Figure S1b and [24]) and express comparable levels of STAT3 (Figure S1b,c). Even if standard procedures to culture GSCs using adherent conditions have been successfully described [3,42], we used free-floating neurospheres that are considered to be a specific growing characteristic of GSCs [43]. Further, the effects of aptamer/chimera treatments in a 3D cell environment would better reflect the physiological conditions for penetration of the molecules. Upon dissociation, we treated GSCs with Gint4.T-STAT3 AsiC (at 400 nmol/L) and the levels of STAT3 protein and mRNA were determined following 72 h by immunoblotting and quantitative reverse transcription polymerase chain reaction (RT-qPCR), respectively. As shown in Figure 1, the conjugate treatment resulted in an efficient reduction of STAT3 both at the protein (about 70%) and mRNA (about 40%) levels as compared to cells treated with a control conjugate containing a scrambled unrelated aptamer linked to siSTAT3 (CtrlApt-STAT3) in all three lines analysed. These results indicate that the AsiC effectively delivers a functional STAT3 siRNA into PDGFRβ positive GSCs. Next, we determined whether Gint4.T-STAT3 could antagonize stem-like GBM cell propagation. The three GSCs were dissociated, treated with the conjugate, and left to form clonal spheres for ten days. We found that the AsiC treatment effectively inhibits tumour sphere formation, reducing the number of spheres (>50 μm diameter) to approximately 50% (Figure 2a–c). In addition, the median size of the spheres (>25 μm diameter) was significantly reduced upon conjugate treatment (Figure 2a–c). Notably, the treatments with the control conjugate did not affect sphere formation, thus indicating that the functional effects depend on Gint4.T-mediated delivery of STAT3 siRNA. In addition, we found that the inhibition of the self-renewal potential of the AsiC-treated spheres correlates with the reduction of the stem-cell associated gene SRY-Box 2 (Sox-2), as detected by RT-qPCR and immunoblot (Figure 2d). Conversely, a clear increase in the differentiation marker Glial fibrillary acidic protein (GFAP) levels (Figure 2e) was detected upon Gint4.T-STAT3 treatment in all the three GSCs tested. To further investigate the Gint4.T-STAT3 functional effect, we measured the cell viability and cell count of treated GSCs. In all the three lines, the MTT analysis showed that the AsiC treatment induces a 20–30% reduction of cell viability (Figure 3a–c). In addition, the cell count reached about 60–70% compared to the untreated cells or control conjugate upon Gint4.T-STAT3 treatment (Figure 3d–f). Taken together, the results show the AsiC treatment effectively inhibits tumour sphere formation, reducing the cell number and viability.Stem-like cancer cells are endowed with a high motility potential and can migrate in vitro in the presence of a chemoattractant stimulus. Since we have previously reported that the Gint4.T aptamer hampers cell migration [32] and that the aptamer synergizes with STAT3 siRNA to interfere with cell migration of differentiated GBM cells [25], we then determined whether this function might be as well preserved on GSC motility. GSCs were left either untreated or treated with Gint4.T, CtrlApt, CtrlApt-STAT3, or Gint4.T-STAT3 for 24 h, and their migration ability was monitored by Boyden chamber assays. As shown in Figure 4a–c and Figure S2a, the treatment with AsiC reduced cell migration by approximately 60–65%, further enhancing the ability of Gin4.T to alter cell mobility (30% reduction). No reduction was found upon treatment with control aptamer or conjugate. Next, we analysed the AsiC ability to interfere with the invading capability of GSCs. To this end, pre-treated GSCs were plated on Matrigel-coated filters and allowed to migrate. We found that a cell’s ability to migrate through the Matrigel in the presence of 10% FBS was significantly prevented in the presence of Gint4.T (30–40% reduction) and further inhibited by STAT3 AsiC, reaching about 65% inhibition as compared to the control aptamer or conjugate (Figure 4d–f and Figure S2b).These data indicate that in the context of the AsiC, the Gint4.T aptamer and STAT3 siRNA synergize to hamper stem cell migration and invasion. One key aspect of targeted delivery strategies is their ability to specifically act only on cells recognized by the targeting moiety. We thus attempt to demonstrate that the AsiC action on the GSCs was mediated by the aptamer recognition of the PDGFRβ. To this end, we treated a patient-derived GSC line (GSC #144) showing high levels of STAT3 but expressing low/undetectable levels of PDGFRβ (Figure S3). Notably, the AsiC treatment of GSC#144 did not change the STAT3 mRNA or protein levels (Figure 5a). Both the protein and mRNA were instead reduced when cells were transfected with the STAT3 siRNA, indicating that the AsiC-mediated silencing requires the presence of the PDGFRβ aptamer target. Further, we confirmed that the functional inhibitory actions of the AsiC were highly dependent on the presence of PDGFRβ. Thus, we analysed the ability of AsiC to affect tumour sphere formation in PDGFRβ-negative GSCs. As shown in Figure 5b, no change in GSC#144 sphere number and diameter was found upon AsiC treatment. On the contrary, both aspects were impaired when STAT3 silencing was forced by the transfection with STAT3 siRNA. Accordingly, the same results were obtained analysing cell viability by MTT assay (Figure 5c) or cell migration by using the Boyden chamber assay (Figure 5d and Figure S4). Taken together, these data indicate that the AsiC functionally acts in a receptor-dependent manner, allowing a specific targeting and inhibition of GSCs. In our previous report, we described the ability of an miR-10b antagonist (antimiR-10b) conjugated to Gint4.T or GL21.T aptamers to selectively target miR-10b and inhibit GSC stem-like phenotype and tumour sphere formation [24]. miR-10b is a biomarker that is highly expressed in GBM and GSCs, acting as an oncomiR to promote cancer stem cell propagation [8,22]. Although, STAT3 and miR-10b contribute to overlapping regulatory pathways, there is no evidence of direct expression regulation. We thus determined whether Gint4.T-STAT3 might synergize with the targeting of miR-10b. Since the efficiency of aptamer delivery is limited by the amount of target receptors on the cell surface [38], we took advantage of the GL21.T aptamer against the Axl receptor [24,35] to deliver the antimiR-10b to the GSCs (Figure S5a). The generated conjugate (GL21.T-10b) was used in combination with Gint4.T-STAT3 on GCS#83 and 61 that are positive for both PDGFRβ (Figure S1b) and Axl (Figure S5b) receptors. By analysing cell viability, we found that both Gint4.T-STAT3 and GL21-10b give a significant reduction that is not further enhanced upon their combination (Figure S5c,d). Conversely, conjugates efficiently synergize to affect tumour sphere formation (Figure 6). As shown, both conjugates interfere independently with sphere formation with similar efficacy, getting about a 50% reduction in the sphere number in the GSCs analysed. Notably, the combined treatment with the conjugates further decreased the number of spheres to approximately 20% (Figure 6, left panels) and size of spheres (Figure 6, right panels), thus strongly reducing GSC self-renewal potential. On the contrary, the treatments with control conjugates containing the control aptamer either alone or in combination are unable to affect sphere formation. These data underline the potential to combine Gint4.T-STAT3 with GL21.T-10b to drastically abrogate GSCs, enhancing the efficacy and the specificity of the treatment. In the present study, we addressed the inhibition of GBM stem-like cells by an aptamer–siRNA molecule, combining the targeting of PDGFRβ and STAT3 gene silencing. We used a previously described nuclease resistant conjugate (Gint4.T-STAT3) [25] containing an aptamer (Gint4.T) that binds and inhibits the receptor tyrosine kinase PDGFRβ, [32], and a STAT3-specific siRNA. By using primary GSCs derived from patient with GBM tumours (WHO grade IV), we demonstrate the AsiC ability to alter self-renewal, viability, and migration.The high intra-tumour heterogeneity of GBM is a serious impediment to the effectiveness of conventional anticancer chemo and radiotherapies, which preferentially targets bulk tumour cells while sparing the more resistant cell populations, such as GSCs [2,3,4,5,6]. Therefore, development of therapeutic strategies aimed at targeting the resistant population has become an urgent need to enhance the responsiveness to treatments. The deregulated activity of STAT3 in cancer cells as well as in GSCs makes it a very promising therapeutic target for the development of an effective strategy for a complete tumour eradication. On the other hand, in normal cells, STAT3 activity is tightly regulated by extracellular signals, enabling cells to respond to the microenvironment and maintaining a momentary active state [44]. The use of a targeting moiety for the selective silencing of STAT3 in cancer cells is thus imperative because it would permit to avoid the occurrence of severe side effects in normal tissues. A growing body of literature is demonstrating the selectivity of aptamer-mediated delivery that allows affecting only cells that express the aptamer target, thus sparing healthy tissues [45]. Accordingly, here we found that, at the difference of STAT3 siRNA transfection, the Gint4.T-STAT3 AsiC treatment is selective for GSCs that express high levels of the PDGFRβ aptamer target (Figure 5). We have recently described that the Gint4.T-STAT3 AsiC shows enhanced serum stability (up to 24 h in 80% serum) and is able to selectively drive the STAT3 siRNA to GBM differentiated cell lines [25]. In the present study, we addressed the efficacy of the Gint4.T-STAT3 treatment to target GSCs and suppress their self-renewing potential. We took advantage of different patient-derived primary human GSCs and found that upon AsiC treatment, the STAT3 gene is silenced (Figure 1), altering the stem-like phenotype and GSC propagation. Indeed, the treatment (1) hampered the formation of tumour spheres, decreasing the levels of Sox-2 and sustaining GFAP levels (Figure 2); (2) reduced the number of viable GSCs (Figure 3); and (3) inhibited GSC migration and invasion (Figure 4), enhancing the functional effect of the unconjugated Gint4.T aptamer [32]. We also found that Gint4.T-STAT3 synergizes with a chimera containing the anti-Axl aptamer GL21.T linked to the single chain antagonist of miR-10b (GL21.T-10b) [24]. Several miRNAs have been reported to be deregulated in GBM, governing different aspects of this tumour, including the maintenance and propagation of the GSCs [46]. Among others, miR-10b acts as an oncomiR and is required for GSC self-renewal and proliferation [9,23], and the targeted delivery of a miR-10b antagonist reduces GSC propagation [24]. Here we found that the combined treatment of Gint4.T-STAT3 with GL21.T-10b resulted in a synergistic and drastic inhibition of GSC self-renewal (Figure 6).Numerous studies have shown that the STAT3 signalling pathway is required for the maintenance of the stem-like malignant glioma cells and that its inhibition by either chemical inhibitors, dominant-negative mutant protein, decoy oligodeoxynucleotides, or siRNAs can be a promising therapeutic strategy [47,48,49]. Our findings consolidate the therapeutic importance of STAT3 in GBM as a fundamental regulator of key tumour features. Most importantly, we provide a stable RNA-based molecule able to selectively inhibit STAT3, allowing impairing the maintenance of the GSC population. Despite many STAT3 inhibitors having been already developed, they still have a limited clinical use [50]. Peptide therapeutics are specific and potent but suffer from rapid degradation/instability and poor bioavailability (membrane permeability). Small molecule inhibitors, although being stable and able to cross the membrane efficiently, are less effective and specifically lead to unwanted side effects. In contrast, DNA and RNA oligonucleotides directed against STAT3 offer high specificity and efficacy, but membrane permeability and tissue-specific delivery remain limited. The multifunctional RNA bio-drug here described combines a cell-targeted inhibitory aptamer and a siRNA STAT3 antagonist offering the possibility to overcome the barriers to oligonucleotide therapies. Our results also show the possibility to design a combined approach targeting both STAT3 and miR-10b to affect the GSC population with high efficacy. In addition, the obtained data strongly support the idea of attacking GBM with innovative multiple therapies to enhance the success of the treatment; for example, associating STAT3 inhibitors with the drugs currently used in the clinic for GBM, such as temozolomide or bevacizumab [51]. Notably, recent evidence supports the ability of our targeting moieties to drive molecular carriers to the tumour site [24,52], thus sustaining the potential applicability of the described AsiC in GBM treatment, although its ability to successfully penetrate into intracranial GBM tumours remains to be determined. Collectively, our findings provide a proof-of-principle for the development of an AsiC-based therapeutic intervention with the potential to selectively target GSCs, potentially enhancing the current therapeutic treatment options. Gint4.T, 5′-UGUCGUGGGGCAUCGAGUAAAUGCAAUUCGACA-3′;Gint4.T stick, 5′-UGUCGUGGGGCAUCGAGUAAAUGCAAUUCGACAXXXXGUACAUUCUAGAUAGCC-3′;Aptamer used as control (CtrlApt),5′-UUCGUACCGGGUAGGUUGGCUUGCACAUAGAACGUGUCA-3′;Aptamer stick used in the control complexes,5′-GCCGCUAGAACCUUCUAAGCGAAUACAUUACCGCXXXXGUACAUUCUAGAUAGCC-3′;human STAT3 siRNA antisense (AS) strand stick,5′-UUAGCCCAUGUGAUCUGACACCCUGAAGGCUAUCUAGAAUGUAC-3′;human STAT3 siRNA sense strand (SS), 5′-CAGGGUGUCAGAUCACAUGGGCUAA-3′GL21.T, 5′-AUGAUCAAUCGCCUCAAUUCGACAGGAGGCUCAC-3′;GL21.T stick,5′-AUGAUCAAUCGCCUCAAUUCGACAGGAGGCUCACXXXXGUACAUUCUAGAUAGCC-3′;antimiR-10b stick (indicated as anti-10b),5′-CACAAAUUCGGUUCUACAGGGUAGGCUAUCUAGAAUGUAC-3′;The STAT3 siRNA duplex sequences were previously reported [53]. All RNAs were produced by the DNA/RNA Synthesis Laboratory, Beckman Research Institute of City of the Hope or by Tebu-bio srl (Magenta, Milan, Italy). RNAs were modified with 2′-F-Pyrimidines. A stick-based approach [54,55,56] was adopted for complex generation. Stick sequences (underlined) within the complexes contained both 2′-F-Py and 2′-oxygen-methyl purines. A three-carbon linker ((CH2)3) spacer, indicated with the italic X, was included in the stick aptamers. Aptamers were refolded before each use by the following temperature cycle: 5 min at 85 °C, 3 min on ice, and 10 min at 37 °C. Complexes were prepared by two-step annealing: (1) the annealing of the STAT3 AS stick, with STAT3 SS done in an annealing buffer (20 mM 2-(4-(2- hydroxyethyl)piperazin-1-yl), ethane sulfonic acid (HEPES; pH 7.5), and 150 mM NaCl, 2 mM CaCl2) by incubating at 95 °C for 10 min, 55 °C for 10 min, and 37 °C for 20 min; and (2) to obtain the final complex, the AS–SS duplex was subsequently annealed with a stick aptamer incubating them at 37 °C for 30 min. For the generation of anti-10b complexes, the stick antimiR-10b, previously denatured at 95 °C for 10 min, was annealed to a refolded stick aptamer at 37 °C for 30 min. The complex formation was checked on a 12% non-denaturing polyacrylamide as the appearance of a shifted band. The GSCs isolated from the tumour tissue of patients with GBM (WHO grade IV) were already published [38,39,40,41,57] and were provided by Dr Lucia Ricci-Vitiani. All the lines were grown in serum-free medium containing 20 ng/mL EGF and 10 ng/mL bFGF (Life technologies, Milan Italy), as described [37]. Before each treatment, tumour spheres were recovered by centrifugation (at 1000 rpm) and dissociated by using 0.25% trypsin. For treatments longer than 72 h, aptamers or conjugates were renewed three times a week at a 200 nmol/L concentration. For transfections, serum-free Opti-MEM and Lipofectamine 2000 reagent (Life technologies, Milan Italy) were used and the manufacturer’s protocol was followed with the annealed STAT3 siRNA (AS–SS duplex) at a 100 nmol/L concentration. Dissociated tumour spheres (1.4 × 105 cells/plate in 3.5-cm plates) were treated with 400 nmol/L aptamers or complexes or transfected, as indicated. JS buffer (50 mM Hepes (pH 7.5), 150 mM NaCl, 1% glycerol, 1% Triton X-100, 1.5 mM MgCl2, 5 mM EGTA, 1 mM Na3VO4, and protease inhibitors) was used for total cell lysates. Samples were prepared in sodium dodecyl sulfate/β-mercaptoethanol buffer and boiled before SDS-PAGE. The SDS-PAGE gels were blotted onto polyvinylidene difluoride membranes (Millipore, Billerica, MA, USA) by electrophoretic transfer. The following primary antibodies were used for immunoblots: anti-STAT3, anti-Sox-2, and anti-vinculin (Cell Signaling Technology Inc., Danvers, MA, USA); as well as anti-α-tubulin and anti-actin (used as a loading control) (Santa Cruz Biotechnology, CA, USA). Single bands at the expected molecular weight were considered (Figures S6–S9). Western blots were quantified by Image J NIH and band intensities (reported below the blots) were expressed as ratio normalized on the loading control signals. Gene mRNA levels were analysed by reverse transcription of 1 mg of total RNA with iScript cDNA Synthesis Kit followed by real-time PCR amplification with IQ-SYBR Green supermix (Bio- Rad, Hercules, CA, USA). The ΔΔCt method was used for relative mRNA quantization by applying the equation 2−ΔΔCt. Primers used were the following: STAT3: fw, 5′ ACCTGCAGCAATACCATTGAC 3′; rev, 5′ AAGGTGAGGGACTCAAACTGC 3′; GFAP: fw 5′ CTGCGGCTCGATCAACTCA 3′, rev. 5′ TCCAGCGACTCAATCTTCCTC 3′; Sox-2: fw 5′GCACATGAACGGCTGGAGCAAGC 3′, rev. 5′ TGCTGCGAGTAGGACATGCTGTAGG 3′; GAPDH (used as a housekeeping control): fw, 5′ CTTTGTCAAGCTCATTTCCTGG 3′; rev, 5′ TCTTCCTCTTGTGCTCTTGC 3′. Tumour spheres were dissociated and 3 × 103 cells/well (in 96-well plates) or 1.4 × 105 cells/plate (in 3.5-cm plates) were seeded for cell viability or cell count assays, respectively. Cells were grown without treatment, treated with 400 nmol/L aptamers or complexes, or transfected, as indicated. Following 72 h, a CellTiter 96 Proliferation Assay (Promega, Madison, WI, USA) was used to measure cell viability or cells were counted after gentle pipetting. Dissociated tumour spheres were counted and 1.4 × 105 cells/plate (in 3.5-cm plates) were seeded and grown with or without treatments for 24 h (400 nmol/L final concentration). Then, cells were recovered, washed, and 1 × 105 cells/point were suspended in serum-free medium and seeded into a 24-well transwell (Corning Incorporate, Corning, NY, USA) upper chamber. The lower chambers were filled with 10% FBS (0.6 mL) used as an inducer of migration. Cells were allowed to migrate for an additional 24 h. To assess invasion, cells were plated in the upper chambers of a 24-well Transwell, previously coated with a 20% Matrigel matrix (BD Biosciences, San Jose, CA, USA). Cell staining with 0.1% crystal violet (in 25% methanol) was used to visualize the migrated or invaded cells. Crystal violet was eluted with 1% sodium dodecyl sulphate and the absorbance at 570 nm was read to quantify cell migration or invasion. Dissociated tumour spheres were counted and 500 cells/well were seeded in a 96-well plate (in duplicate) and treated according to a previously published protocol [24]. Briefly, cells were untreated or treated with aptamers or complexes at 400 nmol/L concentrations. Treatments were renewed three times a week by the addition of 200 nmol/L aptamers or complexes. After 10 days of treatment, pictures of the spheres were acquired with a Leica Application Suite and the number of spheres (diameter > 25 μm) was counted. For statistics, one-way ANOVA with multiple comparison was performed with GraphPad Prism. Student’s t-test was used to compare two groups. Here we demonstrated the ability of an aptamer–STAT3 RNA bio-drug to selectively target GSCs and inhibit their propagation. The molecule combines the inhibitory functions of the anti-PDGFRβ aptamer used as a targeting moiety and of the STAT3 siRNA cargo. Our results indicate that the STAT3 AsiC has the potential to selectively target the heterogonous complexity of GBM stem-like cells by interfering with multiple processes, including cell survival, migration, and stemness phenotype. Furthermore, the conjugate synergizes with the therapeutic targeting of miR-10b in inhibiting the elusive tumour-initiating GSC population. The study provides a proof-of-concept study that paves the way to the rational application of AsiC-based therapeutics for effective GBM therapy. The following are available online at https://www.mdpi.com/2072-6694/12/6/1434/s1, Figure S1: Gint4.T-STAT3 and primary GSCs, Figure S2: Gint4.T-STAT3 effect on GSC migration and invasion, Figure S3: PDGFRβ, STAT3 expression in primary GSC#144, Figure S4: Gint4.T-STAT3 effect on GSC#144 migration, Figure S5: Cell viability with Gint4.T-STAT3 and GL21.T-10b combination, Figure S6: Whole blots of Figure 1, Figure S7: Whole blots of Figure 2d; Figure S8: Whole blots of Figure 5a, Figure S9: Whole blots of supplementary figures. Methodology, C.L.E., S.N., M.L.I. and S.C.; project administration, V.d.F.; resources, C.L.E., L.R.-V., R.P., G.C. and V.d.F.; supervision, C.L.E., S.C. and V.d.F.; writing—original draft, C.L.E.; writing—review and editing, G.C., S.C., L.R.-V., R.P. and V.d.F. All authors have read and agreed to the published version of the manuscript.This work was partially supported by Associazione Italiana Ricerca sul Cancro (AIRC) (IG 2013 N.14046, IG 2016 N. 18473, to GC; N. 9980 to VdF); H2020-MSCA-RISE-2019 cONCReTE 872391, H2020-MSCA-RISE-2019 PRISAR2 872860 to GC; Italian Ministry of Health, GR-2011-02352546 to CLE, and POR Campania FESR 2014-2020 “SATIN” to GC and CLE. We wish to thank L. Baraldi, F. Moscato and D. Rotoli for technical assistance.The authors declare no conflict of interest. Gint4.T-mediated delivery of STAT3 siRNA in primary GSCs. (a–c) Primary GSCs (PDGFRβ+) were treated with 400 nmol/L Gint4.T, Gint4.T-STAT3, control aptamer (CtrlApt), or control chimera (CtrlApt-STAT3) as indicated. After 72 h, the STAT3 protein (left panels) or mRNA (right panels) levels were analysed by immunoblotting or RT-qPCR, respectively. Anti-tubulin antibody was used as an immunoblot loading control. Values below the blots indicate quantization relative to the untreated (“−”) sample, labelled with an asterisk normalized on the loading control signals. Error bars depict the mean ± SD on two experimental replicates. Statistics were calculated using one-way ANOVA: *, p < 0.05; ***, p < 0.001, (Gint4.T-STAT3 versus control chimera). Whole blots of Figure 1 are shown in Figure S6.The effect of Gint4.T-STAT3 on primary GSC tumour sphere formation and stemness. (a–c) Sphere formation of indicated primary GSC-derived tumour spheres (PDGFRβ+) left untreated (−) or treated with Gint4.T, Gint4.T-STAT3, or CtrlApt-STAT3. Left panels are representative micrographs; middle panels are spheres with a diameter > 50 μm and were counted and expressed as percentage relative to the untreated samples (−), set to 100%. Vertical bars depict the mean ± SD; and the right panels are boxplot representations of the diameter measures (spheres with a diameter > 25 μm). Statistics of the conjugate treatment versus the control sample using one-way ANOVA: **, p < 0.01; ***, p < 0.001 ****; p < 0.0001. (d) Levels of Sox were measured by RT-qPCR (left) or immunoblot (right) in primary GSC-derived tumour spheres (PDGFRβ+) treated for 10 days with Gint4.T-STAT3 or control conjugates. Values below the blots indicate quantization relative to the controls, labelled with an asterisk normalized on anti-vinculin signals as a loading control. (e) GFAP levels by RT-qPCR after 10 days of GSC treatment with Gint4.T-STAT3 or control conjugates. In (d,e), statistics for the conjugate treatment versus the control sample were obtained by Student’s t-tests: *, p < 0.05; **; p < 0.01; ***; p < 0.001. Vertical bars depict the mean ± SD on replicates (n = 2). Whole blots of Figure 2d are shown in Figure S7.Gint4.T-STAT3 effect on GSC growth. (a–f) Indicated GSCs (PDGFRβ+) were left untreated (−) or treated with indicated aptamer or conjugates (400 nmol/L) for 72 h. (a–c) Cell viability was measured and expressed as the percentage of the viable cells with respect to the untreated cells. (d–f) The cell number was counted and expressed as the percentage relative to the untreated cells. In (a–f), statistics were obtained by one-way ANOVA: *, p < 0.05; **, p < 0.01 (Gint4.T-STAT3 versus control sample). Vertical bars depict the mean ± SD (n = 2).Gint4.T-STAT3 effect on GSC migration and invasion. Cell motility (a–c) or invasion (d–f) of indicated GSCs (PDGFRβ+) left untreated or treated with indicated aptamers or conjugates (400 nmol/L) for 24 h was analysed. The results are expressed as the percentage of the migrated/invaded cells with respect to the untreated cells. In (a–f), vertical bars indicate the standard deviation values (n = 3). Statistics were obtained by one-way ANOVA (versus control conjugate): ****, p < 0.0001.Gint4.T-STAT3 specificity. (a) GSC#144 (PDGFRβ−) were left untreated (−), treated with indicated aptamer or conjugates (400 nmol/L), or transfected with siSTAT3 (100 nmol/L) for 72 h, as indicated. Left panel: Cell lysates were immunoblotted with anti-STAT3 and anti-tubulin (used as a loading control) antibodies. Values below the blots indicate quantization relative to the control samples (labelled with asterisks) normalized on the loading control signals. Right panel: STAT3 mRNA levels were measured by RT-qPCR. Vertical bars indicate the mean ± SD (n = 2). (b) Sphere formation of GSC#144 (PDGFRβ−) left untreated (−), treated, or transfected, as indicated. Left panel: Spheres with a diameter > 50 μm were counted and expressed as the percentage relative to the untreated samples (−), set to 100%. Mean ± SD (n = 2) is reported. Right panel: Boxplot representation of diameter measures (spheres with a diameter > 25 μm). (c) Cell viability of GSC#144 (PDGFRβ−) following 72 h of indicated treatment (400 nmol/L) or transfection (100 nmol/L). Vertical bars: Mean ± SD (n = 2). (d) Cell migration of GSC#144 (PDGFRβ−) following 24 h of indicated treatment (400 nmol/L) or transfection (100 nmol/L). Vertical bars: Mean ± SD (n = 3). In (c,d), results are expressed as percentages with respect to the untreated cells. In (a–d), the statistics were obtained by one-way ANOVA: *, p < 0.05; **, p < 0.01; ****, p < 0.0001. Whole blots of Figure 5a are shown in Figure S8.Gint4.T-STAT3 and GL21.T-10b combined effect on primary GSC tumour sphere formation. (a,b) Sphere formation of indicated primary GSC-derived tumour spheres (PDGFRβ+) left untreated (−) or treated with Gint4.T; Gint4.T-STAT3 or CtrlApt-STAT3; GL21.T-10b; or control aptamers—alone or in combination, as indicated. Left panels: Spheres with a diameter >50 μm were counted and expressed as the percentage relative to the untreated samples (−), set to 100%. Vertical bars depict the mean ± SD (n = 2). Right panels: Boxplot representation of the diameter measures (spheres with a diameter > 25 μm). Statistics of the conjugate treatments versus the untreated samples were obtained using one-way ANOVA: **, p < 0.01; ***, p < 0.001; ****, p < 0.0001.

Med-MDPI/cancers/cancers-12-06-01435.txt

@@ -0,0 +1 @@
+An emerging theme for Wnt-addicted cancers is that the pathway is regulated at multiple steps via various mechanisms. Infection with hepatitis B virus (HBV) is a major risk factor for liver cancer, as is deregulated Wnt signaling, however, the interaction between these two causes is poorly understood. To investigate this interaction, we screened the effect of the various HBV proteins for their effect on Wnt/β-catenin signaling and identified the pre-core protein p22 as a novel and potent activator of TCF/β-catenin transcription. The effect of p22 on TCF/β-catenin transcription was dose dependent and inhibited by dominant-negative TCF4. HBV p22 activated synthetic and native Wnt target gene promoter reporters, and TCF/β-catenin target gene expression in vivo. Importantly, HBV p22 activated Wnt signaling on its own and in addition to Wnt or β-catenin induced Wnt signaling. Furthermore, HBV p22 elevated TCF/β-catenin transcription above constitutive activation in colon cancer cells due to mutations in downstream genes of the Wnt pathway, namely APC and CTNNB1. Collectively, our data identifies a previously unappreciated role for the HBV pre-core protein p22 in elevating Wnt signaling. Understanding the molecular mechanisms of p22 activity will provide insight into how Wnt signaling is fine-tuned in cancer.Liver cancer is the second most common cause of cancer deaths worldwide and is projected to increase by ~40% by 2030 [1]. The most common type of liver cancer is hepatocellular carcinoma (HCC), which has very limited treatment options and a poor prognosis because it is usually diagnosed at a late stage [2]. The Wnt signal transduction pathway is aberrantly activated in most cases of HCC and mutations to the catenin beta 1 (CTNNB1) gene, the gene that codes for β-catenin, occurs in up to 40% of cases making it the most frequent mutation in HCC [3,4]. β-Catenin is the main effector of the canonical Wnt signaling pathway [5] and these mutations to CTNNB1 lead to constitutive activation of Wnt signaling [6,7]. Liver cancer is also linked to chronic infection with the hepatitis B virus (HBV) that leads to cirrhosis and accounts for 50% of HCC cases [8]. Here, we investigated the oncogenic interplay between these two drivers of liver cancer, namely HBV and Wnt signaling.Wnt/β-catenin signaling is activated by the coupling of Wnt to its cognate receptor, Frizzled (FZD), which initiates a series of events in the cytoplasm that leads to the activation of (TCF)/lymphoid enhancer factor (LEF)/β-catenin (referred to as TCF/β-catenin for simplicity from here on) mediated gene transcription. In the absence of Wnt, β-catenin is primarily engaged at cell-cell adherens junctions and any free β-catenin is cleared by a cytoplasmic destruction complex that contains several proteins, including Axin, adenomatous polyposis coli (APC), glycogen synthase kinase 3 (GSK3) and casein kinase 1 (CK1) [5]. Free, cytoplasmic β-catenin associates with the destruction complex and is sequentially phosphorylated by CK1 and GSK3 at its N-terminus, a post-translational modification that targets it for ubiquitylation and proteasomal degradation. However, upon activation of Wnt-FZD signaling, GSK3 enzyme activity is inhibited and β-catenin escapes phosphorylation and subsequent degradation, accumulates in the cytoplasm and translocates into the nucleus where it complexes with the enhanceosome to initiate the TCF/β-catenin target gene transcription [9]. In liver cancer, the phosphorylation sites of β-catenin are absent due to mutations to the CTNNB1 gene, leading to the constitutive activation of Wnt/β-catenin signaling [3,4,10]. Another common etiologic factor in liver cancer is HBV infection [10,11]. HBV is an enveloped DNA virus whose genome codes for four overlapping genes, namely the envelope or surface (S) gene, the core (C) gene, the X gene and the polymerase (P) gene. The protein products include the surface antigens coded by the S gene, the capsid core proteins coded by the C gene and the HBx protein coded by the X gene. Post-translational processing of the HBV pre-core protein (p25) yields the HBV e antigen (HBeAg, p17) via a p22 intermediate [12]. The HBx protein has been extensively studied for its effects on Wnt/β-catenin signaling [13], however, much less is known about the potential oncogenic interplay with the other HBV proteins. Here, we performed a screen to determine the effects of HBV proteins on Wnt/β-catenin signaling and identified p22, the HBe precursor protein, as a potent activator on its own and in conjunction with active Wnt signaling. Importantly, p22 activated Wnt/β-catenin signaling in colon cancer cells that harbor mutations in intracellular components of the Wnt signaling cascade that result in constitutive activation of signaling. Concomitant regulation of Wnt signaling at multiple levels of the signaling cascade via various mechanisms (genetic, epigenetic, post-translational etc.) to achieve the “just right” level of Wnt signaling for a particular process is a common theme emerging for Wnt-addicted cancers [14,15,16] and here, we demonstrate that HBV p22 might contribute to our understanding of this fine tuning in cancer. To investigate novel mechanisms of oncogenic interaction between HBV and Wnt signaling we screened the ability of various HBV proteins (Figure S1) for their effect of TCF/β-catenin transcription in the presence of Wnt stimulation (Wnt3a conditioned medium). TCF/β-catenin transcription was detected using the TCF reporter, super TOPflash (sTOPflash), which contains eight TCF response elements upstream of a minimal TK (Thymidine Kinase) promoter and sFOPflash, which has the TCF sites mutated [17,18]. The HBx protein activated TCF/β-catenin transcription above Wnt stimulation, however, the pre-core protein p22 was able to increase Wnt activity to a level markedly greater than the HBx protein (Figure 1a). The HBV envelope proteins did not activate reporter activity, nor did the pre-core precursor p25 or core p21, despite significant overlap in the amino acid sequence between the core/precore proteins (Figure S1). The precore contains the genetic sequence of two different proteins, the core protein HBc (p21) (183 amino acids) and precore polypeptide p25 (212 amino acids). They differ only by 29 amino acids at the N-terminus as p25 retains the signal sequence. The cleavage of 19 amino acids from this signal sequence releases cytosolic p22. P22 is further truncated, losing the arginine-rich C-terminal domain, to yield HBe (p17), which is secreted [19]. Expression of p22 was confirmed by immunoblot on whole cell lysates prepared from transfected Huh7 cells using an anti-HBc antibody and, as shown by others [19], neither p17 nor p25 were detected by immunoblot (Figure 1b and Figure S2). HBV p17 and p25 were detected by confocal immunofluorescence in transfected Huh7 cells (Figure S3). Confocal microscopy of Huh7 cells transfected with pCI-p22 and the same anti-core antibody showed diffuse cytoplasmic, diffuse nuclear and, cytoplasmic puncta (Figure 1c and Figure S3) placing p22 in the cellular compartments where Wnt signaling components are found [20]. Next, we demonstrated that p22 activates Wnt signaling on its own and can increase Wnt signaling activity in cells, which are stimulated with either Wnt3a or ectopic over-expression of full length, wild type β-catenin (β-cat-WT) (Figure 2a). The stimulatory effect of p22 on reporter activity was dose-dependent (Figure 2b) and decreased at the higher levels of p22 in the presence of β-cat-WT (Figure 2c). Notably, the levels of transcriptionally active non-phosphorylated β-catenin (β-cat-ACT) [21,22] were increased above that seen with β-cat-WT when p22 was co-expressed (Figure 2d and Figure S4). In the presence of active Wnt signaling, β-catenin escapes phosphorylation and subsequent degradation, and the elevated levels of β-cat-ACT confirm this mechanism for p22 activation of TCF/β-catenin transcription. Data to illustrate the comparative reporter activity between the different conditions is shown in Figure S5.During natural HBV infection, p22 is processed to p17 or HBV e antigen (HBeAg) and secreted into the extracellular space [19]. We confirmed that the transfected p22 is processed to p17 by detecting and quantifying HBeAg in the supernatant of transfected Huh7 cells (Figure S6). Notably, ectopically expressed p17 or p25 did not activate sTOPflash reporter activity above activation by β-catenin (Figure S7).Next, we tested the ability of p22 to activate native TCF/β-catenin target gene promoters. First, we used our previously characterized Frizzled-7 (FZD7) promoter reporter, pFz7-prom [23]. FZD7 is a TCF/β-catenin target gene [23,24] and forms a positive feedback loop in various cancers, including HCC [25,26,27]. As shown above with the sTOPflash reporter (Figure 2a), HBV p22 activated the pFz7-prom on its own, and in the context of Wnt3a stimulation or β-cat-WT over-expression (Figure 3a).Secondly, given that Wnt signaling is dependent on a three-dimentional tissue context [28], we tested the ability of p22 to activate native TCF/β-catenin target gene promoters in the liver in vivo. HBV is an exquisitely human hepatotropic virus and does not infect mouse hepatocytes. However, using hydrodynamic tail vein injection (HDI) plasmids can be introduced into mouse hepatocytes in live animals [29]. A large volume of plasmid containing saline was intravenously injected into mice. This volume overwhelms the heart and is shunted into the hepatic vein and the hepatocytes take up the injected solution (Figure 3b). The mice were culled 6 days and 20 days post HDI and their livers processed for mRNA gene expression analyses using quantitative RT-PCR (qRT-PCR). Expression of Wnt target genes (e.g., Fzd7, Glul) and those that are not target genes (e.g., SOCS3) was determined. At 6 days post-HDI, cyclin D2 and SOCS3 were upregulated by p22 (Figure S8a). Cyclin D2 is upregulated upon activation of Wnt signaling via truncating the APC gene and regulates proliferation in this setting [30], suggesting it is a Wnt target gene, however this may be indirect. Fzd7, a Wnt target gene [23,24] shows a trend in upregulation in response to p22 at 6 days post HDI, which was significantly different by 20 days post-HDI (Figure 3c and Figure S8), whilst the expression of another TCF/β-catenin target gene glutamine synthetase (Glul, Figure 3c and Figure S8b) was only upregulated by p22 at day 20, suggesting early and late regulation or signaling thresholds. There were trends towards increased expression of other TCF/β-catenin target genes but these changes did not reach significance (full qRT-PCR gene analyses are shown in Figure S8 and primer sequences in Table S1). Collectively, these data show p22 activates natural promoters of TCF/β-catenin target genes in the context of a human liver cancer cell line Huh7 (Figure 3a) and normal liver hepatocytes in vivo (Figure 3c and Figure S8).Thus far, we have demonstrated that p22 activates TCF/β-catenin transcription on its own and in the context of Wnt stimulation and β-cat-WT over-expression. This mimics one scenario of additional Wnt signaling in cancer i.e., signaling from the ligand-receptor complex. Next, we investigated p22 activity in other cancer contexts, namely in the context of mutant intracellular components that constitutively activate the Wnt pathway i.e., truncated APC and stabilized, mutant β-catenin. The role of Wnt signaling in cancer has been most extensively studied in colon cancer as Wnt signaling is frequently deregulated in these cancers [32]. Thus, to investigate the effect of p22 in cancer cells with endogenous mutations to intracellular Wnt pathway components, we used colon cancer cell lines SW480 and HCT116 that harbor truncated APC and mutated β-catenin, respectively [18,33]. We also tested the effect of p22 in HEK293T cells that have no known mutations in the Wnt pathway and are known to respond to Wnt [34]. In each cell line (HEK293T, SW480 and HCT116) p22 activated TCF/β-catenin transcription (sTOPflash) above the basal level (Figure 4a). There are four mammalian TCF genes and TCF4 is known to be expressed by SW480 cells [18]. Thus, we tested the ability of a dominant negative form of TCF4 (dnTCF4) [18] to inhibit TCF/β-catenin transcription (sTOPflash) in this cell line. As expected, dnTcf4 decreased constitutive Wnt signaling in SW480 cells. HBV p22 increased Wnt signaling in SW480 cells and this increase was reduced by dnTcf4 (Figure 4b). Collectively, these data show p22 regulates Wnt/β-catenin signaling in the context of genetic mutations that initiate Wnt-addicted cancers. Next, to further test p22 activity in the context of mutant β-catenin compared to β-cat-WT, we used the N-terminally truncated, oncogenic form of β-catenin (ΔN-β-cat) that lacks the regulatory domains [33]. ΔN-β-Cat increased TCF/β-catenin transcription (sTOPflash) above β-cat-WT to a similar level as p22, while ΔN-β-cat and p22 together elevated reporter activity above either alone (Figure 4c). Data to illustrate comparative reporter activity between some of these different conditions is shown in Figure S5.The emerging theme for Wnt-addicted cancers is that the pathway is regulated via multiple mechanisms [16]. This has been extensively investigated in colon cancer. Colon cancers frequently harbor truncating mutations to APC that yield proteins with impeded function in degrading β-catenin; or oncogenic mutations to the CTNNB1 gene that remove the destruction complex phosphorylation sites in the N-terminus of β-catenin [35]. The end result of either mutation is the constitutive activation of Wnt signaling and adenoma formation [6,18,33,36,37]. However, Wnt signaling is also deregulated at the level of the ligand/receptor in colon cancer. Naturally occurring inhibitors of Wnt-FZD interaction are silenced by promoter hypermethylation, while Wnts and FZDs are over-expressed (reviewed in [15,25]). Thus, transcription of TCF/β-catenin target genes can be increased or decreased despite a mutation to downstream components of the pathway. Indeed, all Wnt-addicted cancers show concomitant deregulation to Wnt signaling via intracellular and cell surface mechanisms [16]. Consistent with this, a potent anti-tumor effect was demonstrated by blocking FZD7 function in gastric cancer cells with and without mutant APC [38].Notably, liver cancer displays similar Wnt-addicted mechanisms to colon and gastric cancer [16]. Constitutive activation of Wnt signaling in HCC is primarily via mutations to the CTNNB1 gene that remove the regulatory phosphorylation sites from the N-terminus of β-catenin [3]. However, as in colon and gastric cancer, there is additional regulation of the pathway via over-expression of Wnts and FZDs and epigenetic silencing of naturally occurring inhibitors of Wnt-FZD interaction, for example secreted frizzled related proteins (sFRP) [16,39]. Furthermore, most cases of HCC have a viral etiology and are the culmination of chronic infection with HBV leading to liver disease where HBV proteins, such as HBx, are hypothesized to exert their oncogenic activity, at least in part, through activation of Wnt/β-catenin signaling [8]. Here, we screened the various HBV proteins for their impact on Wnt signaling and demonstrated that another HBV protein, p22, was more potent than HBx. HBV surface proteins (small, middle or large) did not activate TCF/β-catenin transcription. Interestingly, the other pre-core/core proteins (p25, p21 or p17) also did not activate TCF/β-catenin transcription despite significant overlap in their amino acid sequence with p22. Clinical studies show HBe-positivity is a significant independent risk factor of HCC and fatality in chronic HBV-infected patients [40,41]. Furthermore, HBe is produced within the first week after HBV infection in experimental models [42], and thus p22 has the potential to contribute to early events in the transition to cancer. Here, we showed ectopically expressed p22 was localized diffusely in the cytoplasm and nucleus, and in cytoplasmic puncta, indicating potential co-localization with various levels of the Wnt signaling machinery [20]. We also demonstrated Fzd7 and GLUL are induced by p22 in vivo; this shows that genes associate with liver cancer (Fzd7 [39]) and β-catenin-mediated liver zonation and regeneration (GLUL, [43]) are induced by p22 in normal hepatocytes. Furthermore, we demonstrated that p22 can increase TCF/β-catenin transcription on its own and in conjunction with ectopically expressed wild-type or mutant β-catenin; and in colon cancer cells with endogenous mutant APC (SW480 cells) or CTNNB1 (HCT116 cells). Activation of TCF/β-catenin transcription in the SW480 cells by p22 was blocked by dnTCF4, confirming impact specifically on Wnt signaling. Collectively, our data identifies HBV p22 as a novel regulator of Wnt signaling in the context of cancer and provides insight into the mechanisms of ‘just right’ Wnt signaling in cancer. Identifying the molecular interactors of p22 will not only be relevant to HCC but to all Wnt-addicted cancers as it is a new tool to investigate context-dependent Wnt signaling. Immunohistochemical studies in colon cancer carcinomas show variable β-catenin staining where β-catenin is primarily membrane-bound in central areas of the tumor, and intense cytoplasmic and nuclear staining in localized regions that are referred to as the invasive front associated with metastasis [44,45]. This implies that Wnt signaling is constrained in cancer cells allowing for bursts of intense signaling for various processes such as metastasis. It remains to be determined if this localized hyperactive Wnt signaling is due to loss of transcriptional repression or activation of transcription. Further investigation of the p22 mechanism of action in ex vivo models systems for example that do not have the limitations of continuous, transformed cell lines and mouse models with respect to human disease [27], might reveal novel avenues of research to help identify new components to selectively harness different aspects of Wnt signaling; for example, blocking oncogenic Wnt signaling while preserving the critical role Wnt signaling provides to ensure the correct regulation of stem cells and homeostasis of many epithelial tissues. Selective regulation of Wnt signaling is at the core of identifying druggable Wnt pathway targets, as the desired outcome for a cancer specific drug that inhibits Wnt is for the drug to allow normal physiological processes to proceed thus reducing the toxicity of a blanket approach of inhibiting Wnt signaling.C57BL/6 mice used in experiments were between 6 and 10 weeks old, and age- and sex-matched (both sexes were used). Hydrodynamic injection (HDI) was performed as we previously described [29]. Briefly, unanesthetized mice were injected intravenously (iv) through the tail vein with 10 μg pCI-p22 or pCI-EV (pCI, Promega, Madison, WI, USA) in a volume of saline equivalent to 8% of the mouse body weight. The injection was performed within 5 s. Mice were killed 6- and 20-days post HDI, their liver resected and processed for analysis. The Walter and Eliza Hall Institute of Medical Research Animal Ethics Committee (AEC) reviewed and approved all animal experiments (AEC number 2017.016).Mouse liver tissues were homogenized in TRizol (Invitrogen, Carlsbad, CA, USA) and total RNA purified, DNAse treated and quantified as previously described [46]. cDNA was synthesized and subjected to qPCR using SYBR green (ABI). Gene expression was calculated relative to the housekeeping gene β2M (2−ΔΔCt) as described previously [46] and was expressed as fold change over empty vector (EV).The human cell lines (SW480, HCT116, HEK293T and Huh7) were purchased from ATCC. SW480, HCT116 and HEK293T were maintained in RPMI-1640 supplemented with 20 mM HEPES, 10% (v/v) heat-inactivated fetal bovine serum (FBS), L-glutamine and antibiotics (penicillin and streptomycin). Wnt3a producing L-cells (L-3a) and the parental L-cells (L) were a generous gift from Prof Karl Willert [34]. L-3a, L and the Huh7 cells were maintained in DMEM, 10% (v/v) heat-inactivated FBS, supplemented with L-glutamine and antibiotics. Conditioned medium was prepared from L-3a and L cells in parallel as previously described [34]. Cells were seeded into 24-well plates to reach 60–70% confluence overnight. Cells were transfected with 400 ng total plasmid (empty vector added to keep total plasmid constant) that included 100 ng sTOPflash or sFOPflash (generous gift from Prof Randall T Moon [17]); or 100 ng pGL or pGL-FZD7 promoter [23] and 2 ng Renilla luciferase plasmid (phRG-TK, Promega). The pDNA3.1 plasmids expressing β-catenin, ΔNβ-catenin and dnTCF4 were generous gifts from Professor Hans Clevers [18,22] and added at 100 ng/well. The pCI HBV protein expression plasmids were a generous gift from Professor Stephen Locarnini and added at 100 ng, unless indicated otherwise in the text. Cells were transfected using plasmids in OptiMEM (Life Technologies, Grand Island, NY, USA) and Lipofectamine LTX with Plus reagent (Invitrogen) according to the manufacturer’s instructions. Cells were harvested 48 h later and analyzed using the dual luciferase reporter assay system (Promega). For Wnt3a stimulation, cells were treated with 200 µl L-3a or L conditioned medium for 6 h before harvesting in passive lysis buffer. Luciferase activity with control reporters sFOPflash and pGL, and L conditioned medium were negligible. Reporter activity was expressed relative to Renilla to the control for transfection efficiency and plotted as fold change over empty vector (EV) as previously described [38].Pre-cast 4–20% polyacrylamide gels (Mini-Protein TGX, Biorad, Hercules, CA, USA were used to separate proteins (Mini-Protein Tetra Cell, Biorad) and transferred onto nitrocellulose membranes using the Transblot-Turbo instrument (Biorad). The membranes were air-dried and blocked overnight in 1% skim milk at 4 °C. The following day, the membranes were incubated in primary antibody for 1 h and bound antibody detected with secondary antibody and ECL (Western Lightening Plus ECL, Perkin Elmer, Waltham, MA, USA). Primary antibody used were mouse anti-HBcAg [C1] (1:1000, Abcam ab8637, Cambridge, UK), mouse anti-αTubulin (1:1000, Abcam ab7291), mouse anti-active β-catenin (1:1000, Merck Millipore 05-665) and mouse anti-β-actin (1:5000, ThermoFisher AM4302, Waltham, MA, USA). Secondary antibody was rabbit anti-mouse polyclonal antibody HRP (1:10,000 Dako P0260, Glostrup, Denmark). Cells were seeded into two-well Nunc Lab-Tek (Thermofisher) chamber slides to reach 60–70% confluence overnight. Cells were transfected with 200 ng plasmid as described above. After 48 h, cells were fixed with 4% paraformaldehyde, permeabilized with 1% Triton-X100 and blocked with 1% FBS and stained with control antibody or anti-HBcAg [C1] (1:400, Abcam ab8637) primary antibody and detected with goat anti-mouse alexa fluor 488 (1:1000 Invitrogen A11029). DAPI (1:2000) was used for nuclear staining and the cells analyzed using Zeiss LSM700 as previously described [38].The data represent mean ± SEM, where n is at least three independent experiments with cell lines or tissue from at least three mice per cohort, unless stated otherwise. The Student t-test was used for comparisons and significance was defined as p < 0.5.Mutations to APC and CTNNB1 are the most frequent mutations in colon and liver cancer, respectively, and are thought to initiate cancer. Here we demonstrate that the HBV precore protein p22 can activate Wnt signaling in these cancer contexts. The ability of p22 to additionally activate Wnt signaling in the context of these mutations indicates oncogenic interplay between HBV infection and Wnt signaling in liver cancer. Furthermore, it is now clear that Wnt-addicted cancers harbor aberrations to Wnt signaling via both intracellular and cell-surface mechanisms [16], thus our findings identify HBV p22 as a novel tool to understand “additional” regulation and “fine-tuning” of Wnt signaling in the context of cancer [14,25]. Understanding the mechanisms that underly normal, wanted Wnt signaling and pathological, unwanted Wnt signaling is an important step for exploiting the Wnt pathway for anti-cancer treatment. The following are available online at https://www.mdpi.com/2072-6694/12/6/1435/s1, Figure S1: Schematic of the HBV genome and the genes encoding various HBV proteins. The HBV genome, depicted as a long purple continuous strand, encodes 7 proteins from 4 open reading frames (ORFs) (surface [S], core [C], polymerase [P], and the X gene [X]), which are shown as large arrows in different colors, and 3 upstream regions [precore (preC), preS1, and preS2]. The transcripts, ORFs, gene regions and protein products are also shown on the right, Figure S2: Expression of protein from the indicated plasmids. Huh7 cells were transfected with the indicated plasmids and protein expression confirmed by immunoblot. Lysates prepared from Huh7 cells transfected with EV and the parental, un-transfected cells served as negative controls. Lysate from HBV core p21 transfected Huh7 cells was used as a positive control. (a) The membrane was stained with anti-HBc antibody first, then (b) re-probed with anti-tubulin antibody. The boxed areas were used for the cropped blots in Figure 1, Figure S3: Sub-cellular localization of HBV p22. The indicated expression plasmids were transfected into Huh7 and the cells subjected to confocal microscopy following staining with control anti-body and anti-HBV core antibody (red, while DAPI stained nuclei are blue). A higher magnification of the boxed area of the p22 transfected cells is also shown. Scale bars = 20 µM, Figure S4: HBV p22 stimulates Wnt signaling in Huh7 cells. Huh7 cells were co-transfected with 100 ng wild-type β-catenin and the indicated amounts of p22 plasmid and the cell lysates subjected to immunoblot for (a) active β-catenin. The membrane was stripped and re-probed with (b) anti-actin antibody. The boxed regions in (a) and (b) were used the cropped immunoblots in Figure 2d, Figure S5: Comparative reporter activity in Huh7 cells across the various conditions. The TOPflash and FOPflash reporter activities in Huh7 cells transfected with the indicated plasmids and treated with the indicated conditioned media [L-cell conditioned medium (CM) or L-cell-Wnt3a conditioned medium (Wnt3a CM)] are plotted on the same Y-axis to demonstrate the relative reporter activity between controls [(FOPflash, CM, empty vector (EV)] and test samples (TOPflash, expression plasmids, Wnt3a CM) and are shown as fold change reporter activity relative to FOPflash/EV (Mean ± SEM, Student t test, n = 3 experiments). Reporter activity in control samples was negligible, Figure S6: Quantitation of HBeAg levels in the supernatant of transfected Huh7 cells. HBeAg levels in the supernatant fluid of transfected cells were determined (a) two days and (b) three days after transfection using a commercial Roche anti-HBe kit and Cobas e411 instrument. Cells were transfected with increasing amounts of HBV p22-containing plasmid, from 0 - 200 ng per well, with or without co-transfected 100 ng wild type β-catenin (Mean ± SD, n = 3 replicate wells). Transfected p22 was processed to HBeAg and detected in the supernatant, confirming normal processing, Figure S7: Effect HBV p25 and p17 on Wnt signalling. Effect of increasing amounts of transfected HBV precore p17 (a) and p25 (b) expression plasmids on TCF/β-catenin transcription (sTOPflash reporter) in Huh7 cells co-transfected with 100 ng wild type β-catenin was determined and is shown relative to no p17 and p25, respectively (Mean ± SEM, * p < 0.05, Student t-test, n = 3 experiments), Figure S8: HBV p22 upregulates gene expression in vivo. (a) Quantitative RT-PCR analysis of gene expression in livers of mice tail-vein-injected with EV or HBV p22 containing plasmids at 6 days post injection (mean ± SEM, * p < 0.05, n = 7 and 8 for EV and p22 injected mice, respectively). (b) Quantitative RT-PCR analysis of gene expression in livers of mice tail-vein-injected with EV or p22 containing plasmids at 20 days post injection (mean ± SEM, * p < 0.05, n = 4 and 5 for EV and p22 injected mice, respectively), Table S1: qRT-PCR Primer sequences.Conceptualization, B.M.T.; T.J.P.; and E.V.; formal analysis, B.M.T.; D.J.F.; T.J.P.; and E.V.; funding acquisition, D.J.F.; G.E.; N.W.; H.T.; T.F.; C.C.; J.T.; and E.V.; investigation, B.M.T.; D.J.F.; and G.E.; methodology, B.M.T.; D.J.F.; G.E.; T.J.P.; and E.V.; resources, N.W.; H.T.; T.F.; G.K.; C.C.; M.P.; and J.T.; supervision, E.V.; visualization, B.M.T.; D.J.F.; and E.V.; writing—original draft, B.M.T.; and E.V.; writing—review and editing, D.J.F.; N.W.; H.T.; T.F.; C.C.; M.P.; and T.J.P. All authors have read and agreed to the published version of the manuscript. This research was funded by Melbourne Health through a project grant number PG-002-2016 awarded to E.V.; T.J.P.; G.E.; N.W.; T.F.; and C.C.; and a post-graduate scholarship to B.M.T.; E.V.; and T.J.P. were funded, in part, by grants from the National Health and Medical Research Council (NHMRC), project grant number APP1099302 and investigator grant number APP1181580. T.J.P. was funded by BLS/CMU Fellowship and MRC (MR/R026424/1). D.J.F.; was funded, in part, by a Cancer Council of Victoria fellowship and a Melbourne Health early career grant GIA-033-2016.We thank Damian Neate, Danni Colledge and Jean Moselen for technical assistance. We also thank Randall T. Moon, Hans Clevers, Thomas Brabletz, Peter Revill, Stephen Locarnini and Karl Willert for gifting cell lines and plasmids; and the staff at the Walter and Eliza Hall Institute Biological Resource Facility (mice) and the Biological Optical Microscopy Platform (BOMP), University of Melbourne for their assistance.The authors declare no conflicts of interest.Wnt signaling activation is induced by hepatitis B virus (HBV) precore protein p22. (a) Effect of various HBV proteins on TCF/β-catenin transcription activity in Huh7 cells, was determined by reporter activity (sTOPflash reporter) and is shown as fold change relative to empty vector (EV) (mean ± SEM, * p < 0.05, *** p < 0.0001 Student t-test, n ≥ 3 independent experiments for each data point) (b) Expression of protein from the indicated plasmids transfected in Huh7 cells was confirmed by immunoblot. Lysates prepared from Huh7 cells transfected with EV and the parental, un-transfected cells served as negative controls. Lysate from HBV core p21 transfected Huh7 cells was used as a positive control. The membrane was stained with anti-HBc antibody first, then re-probed with anti α-tubulin antibody. (c) Huh7 cells were transfected with p22 plasmid and p22 protein expression (red) and localization detected with anti-HBV core antibody and confocal microscopy (nuclei are blue). Scale bars = 20 µM.HBV p22 stimulates Wnt signaling in Huh7 cells. (a) The effect of HBV p22 alone or in addition to stimulation by Wnt 3a or wildtype β-catenin (β-cat-WT) on TCF/β-catenin transcription in Huh7 cells, was determined by reporter activity (sTOPflash reporter) and is shown as fold change relative to empty vector (EV) (mean ± SEM, *** p < 0.0001 Student t-test, n = 8 independent experiments). (b) Huh7 cells were transfected with the indicated amounts of p22 expression plasmid. The figure shows the dose-dependent effect of HBV p22 on TCF/β-catenin transcription activity (sTOPflash reporter) (mean ± SEM, * p < 0.05, ** p < 0.001 Student t-test, n = 4 independent experiments). (c) Huh7 cells were transfected with the indicated amounts of p22 and 100 ng of wild-type β-catenin expression plasmids. Co-expression of 5–50 ng p22 increased TCF/β-catenin transcription activity (sTOPflash reporter) mediated by wild-type β-catenin; reporter activity decreased when 100 or 200 ng p22 was co-transfected with wild-type β-catenin (mean ± SEM, ** p < 0.001, *** p < 0.0001 Student t-test, n = 3 independent experiments). (d) Immunoblot analysis for the transcriptionally active form of β-catenin (β-cat-ACT) on lysates prepared from Huh7 cells co-transfected with 100 ng wild-type β-catenin, 100 ng of p22 or equivalent EV expression plasmids. The membrane was stripped and re-probed with anti-actin antibody. The bar graph shows quantitative analysis for the levels of detected active β-catenin using Image Lab software and normalized for β-actin levels (mean ± SEM, ** p < 0.001 Student t-test, n = 3 samples).HBV p22 activates TCF/β-target gene native promoters. (a) Effect of HBV p22 on FZD7-native promoter reporter activity, with and without stimulation with Wnt3a or 100 ng wild-type β-catenin (β-cat-WT), in Huh7 cells was determined by luciferase activity (pFz7-prom reporter) and is shown as fold change relative to empty vector (EV) (mean ± SEM, ** p < 0.001, *** p < 0.0001 Student t-test, n = 6 independent experiments). (b) Schematic diagram of hydrodynamic tail-vein injection in mice (adapted from [31]). (c) Expression of TCF/β-target genes Fzd7 and glutamine synthase (Glul) was increased in mouse livers 20 days post HDI injection of p22. Gene expression was determined by qRT-PCR and is shown relative to empty vector (EV) (mean ± SEM, * p < 0.05 Student t-test, n ≥ 4 mice).HBV p22 increases TCF/β-catenin signaling in the context of oncogenic activation of the Wnt pathway. (a) Effect of 100 ng p22 expression plasmid on TCF/β-catenin transcription activity (sTOPflash reporter) in HEK293T cells with no known mutation or aberrant modulation of Wnt signaling; SW480 cells with truncated, mutant APC, rendering Wnt signaling constitutively active and HCT116 cells with mutation at the N-terminus of β-catenin, making Wnt signaling constitutively active (mean ± SEM, * p < 0.05, *** p < 0.0001 Student t-test, n = 3, 5 and 3 experiments, respectively). Reporter activity is expressed relative to empty vector (EV). (b) HBV p22 upregulates TCF/β-catenin transcription (sTOPflash reporter) in the context of truncated APC and this upregulation is blocked by dnTCF4. SW480 cells were co-transfected with 100 ng of p22 and dnTCF4 expression plasmids and the reporter activity is expressed relative to EV (mean ± SEM, *** p < 0.0001 Student t-test, n = 5 experiments). (c) HBV p22 upregulates TCF/β-catenin transcription (sTOPflash reporter) in the context of mutant, oncogenic β-catenin in Huh7 cells. The effect of co-transfection of 100 ng p22 expression plasmid with 100 ng of wild-type or mutant β-catenin on TCF/β-catenin is shown relative to EV (mean ± SEM, *** p < 0.0001 t-test, n = 4 experiments).

Med-MDPI/cancers/cancers-12-06-01436.txt

@@ -0,0 +1 @@
+These authors contributed equally to this work.Although improvement in early diagnosis and treatment ameliorated life expectancy of cancer patients, metastatic disease still lacks effective therapeutic approaches. Resistance to anticancer therapies stems from the refractoriness of a subpopulation of cancer cells—termed cancer stem cells (CSCs)—which is endowed with tumor initiation and metastasis formation potential. CSCs are heterogeneous and diverge by phenotypic, functional and metabolic perspectives. Intrinsic as well as extrinsic stimuli dictated by the tumor microenvironment (TME)have critical roles in determining cell metabolic reprogramming from glycolytic toward an oxidative phenotype and vice versa, allowing cancer cells to thrive in adverse milieus. Crosstalk between cancer cells and the surrounding microenvironment occurs through the interchange of metabolites, miRNAs and exosomes that drive cancer cells metabolic adaptation. Herein, we identify the metabolic nodes of CSCs and discuss the latest advances in targeting metabolic demands of both CSCs and stromal cells with the scope of improving current therapies and preventing cancer progression.Despite significant advances in cancer prevention and treatment, metastatic disease is mostly incurable and resistant to common therapeutics.The development of cancer depends on a small pool of tumor cells owing a phenotype comparable to normal adult stem cells, called cancer stem cells (CSCs), which are characterized by self-renewal and multilineage differentiation capabilities, as well as a great ability to initiate and promote tumorigenesis, metastasis formation and anticancer therapy resistance [1]. CSC features are also dictated by paracrine interactions occurring between tumor cells and their neighboring tumor microenvironment (TME), mainly composed by adipocytes, cancer-associated fibroblasts (CAFs), endothelial and immune cells. CAFs constitute the predominant cellular portion of TME thereby encouraging tumor expansion and progression by providing cytokines, growth factors, metabolites and extracellular matrix remodeling proteins [2,3].Altered metabolism is a common feature of cancer. Normally, non-cancerous cells mainly catabolize glucose by oxidative phosphorylation (OXPHOS) in the mitochondria to produce ATP [4]. However, proliferating cancer cells metabolize significant amounts of glucose into lactate, even in the presence of normal oxygen, a phenomenon which is known as “Warburg effect” [5]. The preference to utilize glycolysis offers several advantages to cancer cells including adaptation in hypoxic and acidic environments (lactate production), which help cancer cells to rapidly proliferate and invade and surrounding tissues [6].However, cancer cells are not characterized by a unique metabolic profile. Both intrinsic and extrinsic factors, such as a nutrient-poor microenvironment, have critical roles in determining cell metabolic phenotypes.Indeed, it is well-established that CSCs can make use of their fully proficient mitochondrial respiration capacity, to face scarce nutrients supply or a hostile microenvironment, with a process named “reverse Warburg effect” [7].CSCs rely on the release of nutrients, such as amino acids and fatty acids (FAs), by TME stromal cells, to accomplish the biosynthetic and energy requirements necessary to proliferate and metastasize even in a low oxygen microenvironment. One conceivable mechanism of interchange between neoplastic and stromal cells is represented by the release of miRNAs and exosomes, small particles containing nucleic acids (DNA, mRNA and miRNA), metabolites and inflammatory factors [8]. Moreover, exosomes and their cargo are in charge for the establishment of the pre-metastatic niche and, importantly, may contribute to cancer treatment failure [9,10].Increasing the knowledge about the heterogeneity of CSC’s metabolic demands may be an essential tool for designing novel personalized therapeutic approaches aimed at enhancing the efficacy of treatments by restricting tumor metabolic adaptation.Here, we will focus on the molecular mechanisms associated with both establishment and maintenance of CSC metabolic phenotypes listed in Table 1, and the biological processes that influence cell metabolic choices including hypoxia, reactive oxygen species (ROS), epithelial tomesenchymal transition (EMT) and the administration of anticancer therapies.One of the most appointed metabolic hallmark of cancer cells was postulated by Otto Warburg in the early twentieth century [11]. The ‘Warburg effect’ foresees the conversion of pyruvate to lactate, from glucose, under aerobic conditions with a consequent predominance of glycolysis on OXPHOS [12]. This metabolic reprogramming has been initially ascribed to defects on mitochondrial function [13]. According to Warburg an injury on mitochondria respiration is at the base of the oncogenic transformation and the shift from OXPHOS toward glycolysis is persistent in the cancer cell progeny [13]. Recently, in a model of colorectal cancer it has been proved that tumor initiation is sustained by an enhancement of the glycolytic program in a condition in which the mitochondrial pyruvate carrier is inactivated, causing low pyruvate import, necessary for oxidative metabolism, onto the mitochondria. In vivo loss of mitochondrial pyruvate carrier 1 increased high-grade adenoma formation and it was fundamental for the subsistence of tumor initiating cells at the base of colon crypt [14].Stem and cancer cells prefer glycolysis because it perfectly meets their metabolic requirements [15,16]. Notwithstanding glycolysis is an inefficient process with respect to ATP production, cancer cells uptake high quantity of glucose that enters glycolysis pathway and generates a comparable number of ATP molecules as OXPHOS [17]. Glycolysis is fundamental for cell bioenergetics as well as for the generation of metabolic intermediates used to produce NADPH, lipid or glycogen molecules. For instance, ribose-5-phosphate groups derived from glucose-6-phosphate are used for nucleotide biosynthesis [18]. In this context an enhanced nucleotide production, as it is the case in cancer cells, leads to a decrease in the NAD+/NADH ratio, which in turn encourages the conversion of glucose to lactate [19].Glycolysis counteracts the formation of reactive oxygen species (ROS) in proliferating cells [28]. Luo et al. demonstrated that the inhibition of glycolysis by 2-deoxyglucose (2-DG) induced the transition of mesenchymal breast CSCs harboring low levels of ROS toward a ROShigh phenotype of epithelial breast CSCs, which are sensitive to auranofin and L-buthionine-sulfoximine (BSO), two inhibitors of the antioxidant pathways thioredoxin (TXN) and glutathione (GSH), respectively [27,29]. The latter study clearly indicates that despite glycolysis plays a central metabolic role, cancer cells modulate their metabolic phenotype in response to metabolic stress, including the possibility to switch to OXPHOS and vice versa if needed [30].Indeed, pancreatic CSCs harboring high level of the peroxisome proliferator-activated receptor gamma co-activator 1-alpha (PGC-1α) rely on OXPHOS and diverge from the glycolytic phenotype of their differentiated counterpart. Mitochondrial inhibition by metformin causes the selection of a resistant subtype of MYC-expressing pancreatic CSCs showing an intermediate glycolytic/oxidative metabolic phenotype [24].A conspicuous entity of studies shows that cancer cells are characterized by high energy demand. According to the “Warburg effect” cancer cell sustenance is solely glycolytic.Nowadays we know that although the Warburg phenotype represents an undisputed feature of many cancer cells, most tumor cells possess intact tricarboxylic acid (TCA) cycle and OXPHOS [31,32].The oxidative metabolism allows CSCs to have selective advantages, greater resistance to the inhibition of glycolysis, enhanced degree of independence from the supply of macronutrients and also a more efficient source for energy production [20,33]. In order to emphasize the central role of mitochondria in the self-renewal of CSCs and in their resistance to differentiation, two years ago it was coined the term “mitostemness” [34,35].Several studies have highlighted that CSCs are endowed with a plastic metabolic phenotype, both glycolytic [36] and oxidative through the OXPHOS [37]. The latter process is mediated by electron transport chain (ETC) and involves supply of reducing equivalents from TCA cycle or fatty acid oxidation (FAO) for ATP and NADPH generation.Janiszewska M. et al. demonstrated that the subpopulation of tumor-initiating gliomaspheres depends on OXPHOS for energy production. In particular, the oncofetal insulin-like growth factor 2 mRNA-binding protein 2, which regulates the protein synthesis and contributes to assembly of the subunits of ETC, sustained glioblastoma cell clonogenicity [21].Mitochondrial metabolism is fundamental in CSCs because it constitutes a source of intermediate metabolites such as acetyl-coA, NAD+, alpha-ketoglutarate, succinate, FAD, S-adenosylmethionine. The variation in the concentration of these intermediate metabolites could lead to important phenotypic changes because they can be used by enzymes capable of modifying the state of the chromatin [38] and thus the transcriptomic profile of cancer cells. There are various ways to modify chromatin, for example, acetyl-coA is used to transfer acetyl group to lysine residues of histones [39] and S-adenosylmethionine is a donor of methyl group for the methylation of DNA and histones [40]. Histone acetylation, mediated by histone acetyl transferases, increases gene expression [41]. The proteins involved in the cytosine methylation of CpG dinucleotides are the DNA methyl transferases DNMT1, DNMT3A and DNMT3B and they have been identified as factors driving the CSCs formation and maintenance [42]. Histone methylation occurs predominantly on lysine (K) and arginine (R) residues. Such modifications are commonly associated with gene activation or repression, depending on the target histone modification [43]. Interestingly, the enhancer of zeste homolog 2 (EZH2), which is the catalytic subunit of polycomb repressive complex 2, has been demonstrated to downregulate gene transcription through trimethylation of histone H3 on lysine 27 [44], while it upregulates NOTCH expression, being crucial in the expansion of the stem pool in breast cancer [45].Given the discovery of functional mitochondria in CSCs, it was hypothesized that the capability to rapid switch from glycolytic metabolism toward oxidative metabolism [46] and vice versa, could represent an escape mechanism to standard anticancer therapies. Therefore, using drugs that selectively target mitochondria could make CSCs more sensitive to standard therapies.PGC-1α is a stress sensor, activated following the presence of limited amount of nutrients, oxidative stress and chemotherapy, to increase mitochondrial biogenesis and therefore all mitochondrial activities (OXPHOS, FAO and detoxification ROS) [47]. The XCT790 is an inhibitor of estrogen-related receptor α, the latter being a transcription factor coupled with PGC-1α. In this context, De Luca et al. have shown that the use of XCT790 reduces the proliferation of CSCs in breast cancer [48].One of the most studied ETC inhibitor to target CSCs is the antidiabetic drug metformin, which targets the complex I in ETC in mitochondria. Metformin acts as driver of apoptosis in CD133+ pancreatic ductal adenocarcinomas cells and CD44highCD24lowbreast cancer spheres [24,49]. Promising clinical data havebeen reported for the use of metformin in breast, endometrial, prostate and pancreatic cancers (NCT01266486; NCT02755844; NCT01620593; NCT02978547) [50,51]. For instance, a retrospective study conducted on 445 patients affected by neuroendocrine pancreatic tumors revealed that progression free survival was doubled in diabetic patients receiving metformin than patients without diabetes (median progression free survival, 32.0 months versus 15.1 months) [51].Another important mitochondrial target is represented by mitochondrial ribosomes, which are the site of protein synthesis. The homology existing between mammalian mitochondrial ribosomes and prokaryotic ribosomes is well established, for this reason it is possible to use a wide variety of antibiotics already approved by the FDA [52]. Ribosomes are targeted by the two main families of antibiotics, tetracycline and macrolide [53,54]. Among the wide range of antibiotics, tested by Lamb et al. doxycycline was found to inhibit mammosphere formation in breast cancer and many different tumor types [52]. Therefore, this therapeutic approach could be applicable to many cancer types.Notwithstanding inside the cell there are various production sites of ROS, mitochondrial ETC represents their primary endogenous source [55]. The intracellular increase of ROS may result from the activation of oncogenes, inactivation of tumor suppressor genes, enhanced oxidative metabolism and mitochondrial dysfunction [56]. The accumulation of ROS causes damage to proteins, lipids and DNA and therefore irrecoverable damage to the cell [27]. Notably, cells have protection mechanisms from oxidative stress, including enzymatic antioxidants such as superoxide dismutase (SOD), catalases, TXN, peroxiredoxins, glutathione (GSH) peroxidases, p38-MAPK and sirtuins [57,58,59] and non-enzymatic molecules such as GSH, vitamin C (ascorbate), vitamin E (tocopherols) and polyphenols [60]. Interestingly, the administration of SOD-1 inhibitor, disulfiram, strongly counteracted breast CSCs expansion by modulating MAPK, NF-kB and STAT3 pathways [61,62].Although it has been shown that there is an increase in mitochondrial functionality in CSCs compared to other cancer cells [48], it appears that there are low concentrations of ROS within CSCs [63,64] as there is a high production of antioxidant agents that maintain the disposal of ROS, protecting CSCs from oxidative damage [65]. The excessive increase in ROS levels reduced clonogenicity and mediated death of CSCs [66]. Moreover, the use of paraquat, an inducer of ROS, increased ROS production and significantly reduced tumor spheroid formation of CSCs [16].Albeit several authors have shown that CSCs harbor high ROS levels, further studies clarified their role in cancer onset and progression. In particular, ROS facilitates CSCs self-renewal, expansion, invasiveness [67,68,69] and potentiated tumor progression [70,71].Myant et al. highlighted the connection between Rac1, NF-kB and ROS in CSCs, in which ROS act as connecting molecules between Rac1 and NF-kB. Rac1 is required for NF-kB activation and initiation of colon tumorigenesis. Rac1 is also part of a protein complex containing NADPH oxidase, which generates superoxide, a major constituent of the pool of intracellular ROS. On the other hand, NF-kB is a redox-sensitive transcription factor that is activated by increased levels of ROS. In this study, the authors showed that increased NADPH oxidase activity, ROS intracellular concentration and NF-kB signaling in the APC-inactivated intestinal cells are all dependent on Rac1 expression [68,72].Within tumor bulk, CSCs represent the subpopulation with the highest degree of adaptability to redox stress. The GSH system, adopted by CSCs, is the major antioxidant defense mechanism that reduces the intracellular levels of ROS. Gln-derived glutamate (Glu) is a necessary substrate for the synthesis of GSH and an exchange ion for the cellular import of cystine, which cooperates with Glu for the production of GSH, through the CD44v-xCT transporter [73]. Cells showing stem-like features strongly rely on mitochondrial OXPHOS sustained by glutamine (Gln) catabolism [23]. Of note, cancer cells are metabolically heterogeneous, since they can also retain the OXPHOS capacity and replace glucose with Gln or fatty acids (FAs), as energy supply. Despite being a non-essential amino acid, Gln is considered as a ‘conditionally essential’ amino acid since its de novo synthesis does not satisfy the demands of cancer cell, which become reliant on its exogenous uptake. Gln is also required for TCA cycle intermediate replenishment, protein translation and the biosynthesis of amino acids and nucleotides [74].Interestingly, metformin resistance can be mediated by Gln compensation. In human cancer organoids and in in vivo model it was described that the administration of metformin and a glutaminase inhibitor (GLSi), which blocks Gln consumption, efficiently overcome metformin resistance [75]. Glutaminase activity also regulates head and neck cancer metabolism though the expression of the stemness marker ALDH [25].It has been shown that survival of lapatinib-resistant breast cancer cells is supported by Gln metabolism [76], whose inhibition could be exploited for sensitizing cancer cells to standard therapies. Notably, GLSi have been recently launched to Phase II in combination with paclitaxel for the treatment of triple negative breast cancer patients (NCT03057600).Recently, different clinical studies demonstrated that leukemia stem cells are greedy of amino acids and possess enhanced catabolic activity [26]. Some amino acids, such as methionine, influence the epigenetic state of cancer cells and fuel tumor initiation [77]. In addition, glycine decarboxylase, belonging to the serine–glycine pathway, is the metabolic driving force of tumor-initiating cells in non-small cell lung cancer [22]. Metabolic targeting of amino acids resulted often in neoplasia regression [77]. However, phenomena of cancer metabolic compensation, mediated by FA oxidation, have been observed [26] (Figure 1).As we have already extensively discussed in this review, the ability to adjust cell metabolism is crucial in cancer cells, because it supports the neoplastic proliferation and survival during the different steps of cancer progression [78]. Given the high plasticity of CSCs, this phenomenon is highly pronounced in this cell subset, in particular regarding FA metabolism, which is fundamental to sustain their boosted growth, division and survival during cancer progression. FAs play an important role in different aspects of cancer cell life, being crucial for cell membranes assembly, acting as structural components of the extracellular matrix, secondary messengers and energy source [79]. FAs abundance is regulated by two different processes, de novo biosynthesis mediated by nutrients as glucose and amino acids, and exogenous uptake thanks to the presence of specific transporters, both mechanisms are finely regulated at different levels during cancer progression by the dynamic TME components and dysregulation of oncogenic signaling pathways (reviewed in [80]). Importantly, in 1984, Ookhtens and colleagues demonstrated in vivo that most of the esterified FAs in tumors were derived from de novobiosynthesis [81], as confirmed few years later by Kuhajdaet al., who showed for the first time that tumor cells (breast cancer in this study) over-express fatty acid synthase (FASN) [82]. These findings were also demonstrated in glioma, breast and pancreatic CSCs, where FASN expression dictated the acquisition of stemness features [83,84,85,86].Lipid droplets (LDs) are dynamic and multifunctional cytoplasmic organelles involved in the storage of FAs, which are fundamental to avoid lipotoxicity when cells start accumulating an excess of FAs, in particular under hypoxic conditions that induce an upregulation of FAs uptake, but also to provide an important source of ATP and NADH under stress. The generation of energy from LDs is achieved by β-oxidation of stored lipids that is sufficient to provide ATP during the metastatic cascade and essential for production of NADPH useful as detox agent against ROS [87,88]. The role of LDs biogenesis and their regulated mechanisms in cancer cells has been widely described over the last decade, and observed in all the phases of cancer development, including initiation, promotion and progression [89,90] in particular in CSCs from different organs, including colon cancer [91,92]. These findings suggest that LD biogenesis could be considered a promising target for designing innovative antitumor therapeutic strategies [93,94].In addition to de novo synthesis, FAs can be collected from exogenous environment through the expression of specialized transporters that facilitate the efficient movement across the plasma membrane. The best characterized include CD36, also known as fatty acid translocase (FAT), the fatty acid transport protein family (FATPs) and the fatty acid-binding protein (FABPs), which have been shown to be upregulated in tumors [95,96]. In particular, CD36 has been demonstrated to be crucial for metastatic dissemination of cancer cell in different tumor type, with its high expression being correlated with poor prognosis [87]. Accordingly, targeting of CD36 has shown pre-clinical evidence that its inhibition is sufficient to impair metastases [87]. Interestingly, the expression of CD36 in cancer cells is also able to regulate the metabolic crosstalk with TME by increasing their dependency on exogenous lipid uptake. Lipid uptake from cancer cells becomes crucial mostly under condition of metabolic stress. Indeed, it has been shown that in hypoxia, the glucose to acetyl-CoA flux is downregulated, as well as unsaturation of FA that is driven by stearoyl-CoA desaturase-1 SCD-1, which is an oxygen-consuming enzyme. To compensate this prohibitive microenvironment, during hypoxia cancer cells upregulate FABP4 [97], which is a target of the hypoxia-induced factor 1 alpha (HIF1α) [98]. Notably, FABP4 negatively correlated with progression-free survival [99].Adipose tissue cells establish symbiotic relationship with cancer cells. Interestingly, some cancers show preferential homing to adipose tissue [100]. Adipocytes sustain cancer cell growth by activating lipolytic processes with consequent release of free FAs that are up taken by cancer cells over-expressing FABP4, which in turn release exosomes containing pro-lipolytic factors (i.e., miR-144 and miR-126) [101,102]. We can distinguish two different populations of cancer adipocytes, the “peritumoral adipocytes” that do not penetrate tumor tissue (surrounding tumor tissue, tumor-educated adipose cells can be also found distant from tumor growth [103]), and the “metastasis-associated adipocytes”, which arise from the transient or prolonged contact between metastatic cancer cells and resident adipose cells. The most important mechanism involved in cancer cell metabolic reprogramming, is driven by cancer cell-driven lipolysis in adipose cells [104]. Indeed, it has been demonstrated that following prolonged exposure of cancer cells to adipose cells, the latter start losing their lipid content (responsible for tumor cachexia), gaining fibroblast-like phenotype, which further promotes the metastatic behavior of cancer cells. Recently, Ye et al. demonstrated that leukemic stem cells use gonadal adipose tissue as niche to support their growth and to protect themselves from chemotherapy, by inducing an inflammatory microenvironment and adipose tissue lipolysis [105]. Interestingly, the most beneficial effect was gained by leukemic stem cells over-expressing the FA transporter CD36 [105].Another important source of heterogeneity in the metabolic signature of cancer cells is due to the genetic background since several metabolic processes are directly regulated by constitutive activation of oncogenes and deactivation of tumor suppressor genes [106]. In particular, oncogenes as RAS and PI3K are usually associated with glycolysis over OXPHOS, while tumor suppressor genes, as p53, have opposite effects [107].One of the most important oncogenes in cancers is RAS. Indeed, constitutive activation of KRAS is frequent in colon and non-small cell lung cancer [108]. KRAS by activating ERK1/2 reprogramscancer cell lipid metabolism, leading to increased synthesis of glycerophospholipids and FAs [109,110]. G12V mutated KRAS is also able to boost de novo lipogenesis [111].PI3K is one of the most frequently dysregulated pathways in cancers. Constitutive activation of this signaling regulates de novo lipid biosynthesis by increasing acetyl-CoA synthesis [112], and promoting NADPH production to fuel lipogenesis [113]. Several metabolic pathways are also regulated by mTOR, which is strictly linked to dysregulation of PI3K/Akt pathway. mTOR can affect cancer cell lipid metabolism by activating OXPHOS that in turn promote lipogenesis [114], as recently demonstrated by the evidence of downregulation of lipogenic enzymes, including FASN, acetyl-CoA carboxylase 1 and ATP citrate lyase, following treatment with mTOR inhibitor or by raptor genetic knockdown [111,115]. PI3K can be activated by stimulation of growth factor receptor tyrosine kinases, such as HER2. HER2 positive tumors are characterized by enhanced de novo lipogenesis, which contributes to the aggressiveness of these tumors [116]. The overexpression of HER2 is sufficient to prompt a FASN-dependent lipogenic phenotype in non-transformed epithelial cells that recapitulate the cancer cell metabolism. Conversely, HER2 inhibition—or de novo lipogenesis—inhibit oncogenic potential and induce apoptosis [117].Among the others, p53 can directly or indirectly regulate lipid metabolism-related gene expression [118]. p53 downregulates FAs synthesis by inhibiting the pentose phosphate pathway and downregulating SREBP1 [119]. On the other hand, p53 is able to increase FAs oxidation and prevent lipid accumulation, by inhibiting the pyruvate dehydrogenase kinase at transcriptional level, thus leading to increase in pyruvate dehydrogenase and to the conversion of pyruvate to acetyl-CoA [120,121].The complex intra-tumor regulation of metabolic activities is further enriched at cellular level, as best represented by breast cancer, which includes multiple cellular subtypes characterized by specific hormone/growth factor receptor and genetic profile. A recent study has highlighted how triple-negative breast cancers over-express genes involved in exogenous lipid uptake (high FABP5 and FABP7 expression), while the receptor-positive breast cancers are associated with de novo lipogenesis [122]. For these reasons, a better characterization of lipid-associated signatures in cancers could help to guide therapeutic interventions.In additionto fuel the boosted proliferation, aberrant FA metabolism protect cancer cell from surrounding stress of different nature. It has been demonstrated that an excess of saturated FAs is responsible of mitochondrial/ER stress and dysfunction [123]. To prevent accumulation of saturated FAs, cancer cells overexpress SCD1 and SCD5, which convert saturated into monounsaturated FAs [124]. To cope with hypoxic microenvironment that often characterizes tumor progression, and which inhibits the activity of SCD enzymes, cancer cells upregulate the uptake of exogenous unsaturated FAs in order to maintain lipid homeostasis [123]. Importantly, to increase FAs uptake, it has been demonstrated that cancer cells over-express lipoprotein–lipase, which are involved in lipolysis of extracellular TG-rich lipoproteins, thus facilitating the hypoxia-induced FA uptake directly from TME [125]. However, hypoxia-induced regulation of FAs is highly cell-specific [126]. The major difference among the available studies is due to the cell culture conditions, and in particular, by the abundance of nutrients in cell culture media. Indeed, it is now clear that hypoxia or serum starving alone induce different metabolic phenotype in cancer cells, compared to their combination. Importantly, hypoxic regions are characterized by both oxygen and nutrients deprivation. In such condition, cancer cells are dependent on Gln and acetate as carbon source for the production of acetyl-CoA [127,128]. In line with these findings, when both oxygen and serum are limited, cancer cells upregulate nuclear acyl-coenzyme A synthetase that plays a dual role, aiding the use of extracellular acetate as carbon source and maintaining high histone acetylation levels, to prevent induction of apoptosis and promoting cancer cell growth [129].Another important consequence of hypoxia microenvironment is represented by the excess in ROS production. This phenomenon results from FA oxidation process, activated by patatin like phospholipase domain containing 2 (PNPLA2) and mitochondrial electron leakage, which both lead to oxidative stress. To counteract this phenomenon, cancer cell downregulate PNPLA2 through the hypoxia-inducible protein 2, to promotes cancer cell survival [130,131]. The resistance to oxidative stress in hypoxic condition is also driven by a deregulated membrane lipid saturation, with increased saturated and monounsaturated FAs, which are less susceptible to peroxidation [132].Dysregulation of membrane lipids saturation, as well as cholesterol content, could also affect cell membrane fluidity and dynamics, both mechanisms associated with the acquisition of a mesenchymal phenotype, which is crucial in metastatic cancer cells. Indeed, cholesterol synthesis is finely regulated during metastatic dissemination of cancer cells. Ehmsen, et al. recently demonstrated that elevated cholesterol biosynthesis identifies the subpopulation of breast CSCs. The administration of simvastatin, by inhibiting the 3-hydroxy-3-methyl-glutaryl-coenzyme A reductase, hampered mammosphere formation [133].All these pre-clinical findings regarding the role of FA metabolism in cancer progression arouse clinical interest for development of innovative antitumor strategies targeting FA metabolism, each one directed against specific phases of lipid metabolism: de novo synthesis, modification, uptake, activation, storage, mobilization and degradation [134].Most of the enzymes involved in FA metabolism are regulated by sterol regulatory element binding protein-1SREBP-1 at transcriptional level. However, despite targeting SREBP-1 could potentially represent a promising strategy to regulate lipid metabolism in cancer cells, transcription factors are mostly considered undruggable. For this reason, several pre-clinical studies are now considering the idea to target its protein partner, SREBP-cleavage activating protein, which drive SREBP-1 to the Golgi for activation, by treating cancer cells with fatostatin and betulin [135,136]. Another important druggable player in FA biosynthesis is FASN. Although several pre-clinical studies showed efficacy in inhibiting cancer growth, the pronounced side effects, mostly on neuronal stem cells, impeded its use in clinical settings [137,138]. FA modification is an important step in lipid metabolism. Among the enzymes driving this step we find the elongases (seven members, ELOVL1-7) and desaturases (two members, SCD-1 and SCD-5). Importantly, our group has recently shown that inhibition of SCD-1, using the betulinic acid, in vitro is sufficient to efficiently kill colon cancer cells, in particular the CSC subset, leading to mitochondria-dependent cell death [139,140]. Similarly, the SCD-1 inhibitor, CAY10566, inhibits glioblastoma stem cells expansion and in vivo tumor growth [141]. As previously mentioned, one of the most important transporters of FAs is CD36. Targeting this membrane protein has been demonstrated to impair metastases in breast cancer and melanoma, by inhibiting the metastasis-initiating cell compartment [87].Recent studies shed light on the role of microRNAs (miRNAs) in regulating stemness features of cancer cells, such as self-renewal, differentiation, metabolic reprogramming, metastasis formation and anticancer therapy resistance, associated with the pathogenesis of various types of human cancer [142].miRNAs are single strand short non-coding RNA molecules (21–23 nucleotides) transcribed as precursor molecules, which are subsequently cleaved by the endoribonucleases Drosha and Dicer. Mature miRNAs are bound by a member of the Argonaute(AGO) protein family to form the RNA- induced silencing complex (RISC) in a process termed RISC loading. The miRNA guides RISC to complementary sequences located mainly in the region of its target mRNAs, silencing gene expression [143]. Following nuclear transcription, miRNAs are processed within the "multi-vesicular bodies", that consist of phospholipid membranes and lipoprotein microvesicles of endocytic origin, known as exosomes, of about 30–100 nm. After the binding of bodies to the plasma membrane, miRNAs are released into the bloodstream [144].miRNAs have the role of coordinating gene expression programs at the base of physiological and pathologic cellular processes, including cancer [145]. Interestingly, miRNAs can act as both oncosuppressors, inhibiting proliferation and oncogenes, promoting tumor initiation, growth and metastasis formation [146]. Each miRNA can bind and regulate the expression of multiple coding or non-coding mRNAs. Therefore, the aberrant expression of a single miRNA can deleteriously influence the translation of multiple genes within a cell, leading to profound phenotypic changes [146].To better understand the interaction between miRNAs and CSCs, recent studies evaluated how miRNAs can play a key role in maintaining and regulating the functioning of CSCs by targeting various oncogenic signaling pathways, such as Notch, WNT/β-catenin, JAK/STAT, PI3K/AKT and NF-kB [142].Some miRNAs, such as miR-145, miR-200c, miR-494, miR-195-5p, miR-34, miR-519d, miR-128, miR-99a inhibit CSCs expansion by inhibiting of ADAM, BMI, Notch, caspases and mTOR, respectively, while others, such as miR-19, miR-501-5p, miR-21, miR221/222, miR-483-5p, miR-196b-5p, miR-494-3p stimulate stemness through the activation of WNT/β-catenin, PTEN, cyclin D1 and MMP-2, STAT3 and Notch 1 pathways [142].The deregulation of Drosha and Dicer has been observed in different types of cancer [147,148]. For instance, the downregulation of Dicer led to decreased miR-130b and tumorigenesis [149]. In addition to Drosha and Dicer, other enzymes involved in the miRNAs biogenesis pathway, are the trans-activation responsive RNA-binding protein (TARBP2) and the Argonaute RISC catalytic component 2 (AGO2). In sporadic and hereditary carcinomas, downregulation of TARBP2 protein expression in CSCs was shown to be important for pro-metastasis signaling [150].A key role in the miRNAs biogenesis pathway is played by Exportin-5 (XPO5) protein. It mediates the nuclear export of miRNAs precursors to the cytoplasm. Genetic mutations of XPO5 leads to an entrapment and loss of precursor miRNAs in the nucleus promoting tumor initiation through an increase in the expression of stemness-related genes such as EZH2 and MYC [151].Notably, different miRNAs have been associated with the regulation of cancer cell metabolism [152].The transport of glucose inside the cells is an important event for the correct cellular functioning. This mechanism is mediated by tissue-specific glucose transporters known as GLUTs, which are directly or indirectly targeted by several miRNAs. Presumably, deregulation of GLUTs can increase glucose uptake, satisfying the high glucose requirement and accelerating metabolism in cancer cells. However, the direct links between miRNAs deregulation and glucose transport in cancer are largely unknown therefore further studies are needed. A key enzyme for aerobic glycolysis is the hexokinase 2 (HK2), which is overexpressed in tumors. miR-143 is inversely correlated to the expression of HK2 in the head and neck squamous cell carcinoma and in lung cancer [153]. Likewise, the loss of miR-143, in glioma tissues and glioblastoma stem-like cells, promoted the expression of HK2, resulting in enhanced aerobic glycolysis and inhibition of cell differentiation [154]. The connection between miR-143 and HK2 has been studied also in colorectal cancer, where it has been shown that it downregulates HK2 expression and that its reintroduction leads to a decrease in lactate secretion by impairing the rate of glycolysis [155]. Another miRNA-regulated glycolytic enzyme is phosphofructokinase 1 (PFK1). PFK1 catalyzes the phosphorylation reaction of fructose-6-phosphate to convert it into fructose-1,6-bisphosphate. Yang et al. showed that miR-135 targets PFK1 and inhibits aerobic glycolysis and suppresses tumor growth [146].Chong et al. studied the role of miRNAs in maintaining the glycolytic metabolism of CSCs [156]. In their study, they demonstrated a critical role of the LIN28B in stemness. In breast cancer cell lines, the inhibition of LIN28B suppresses MYC expression and increased miR-34a-5p levels expression, correlating with inhibition of glucose uptake/lactate production and a better patient’s prognosis. Therefore, blocking of the LIN28B/MYC/miR-34a-5p signaling pathway, by the LIN28B specific inhibitor, causes a dramatic inhibition of tumor growth and metastatic potential in orthotopic immunodeficient mouse models of human breast cancer cells [156].A class of miRNAs that regulates the choice between self-renewal and differentiation of breast CSC is represented by miR-600. In breast CSCs, the miR-600 silencing caused cancer cell expansion, while its overexpression reduced CSC self-renewal, leading to a decreased in vivo tumorigenicity. Interestingly, the main target of miR-600 is SCD1, an enzyme required to produce active lipid-modified WNT proteins, which are involved in cell fate determination during tissue development and oncogenic processes [157]. Indeed, the authors showed that low levels of miR-600 are correlated with active WNT signaling and a poor prognosis. These findings highlighted that miR-600 is involved in breast cancer cells-fate decisions and influences tumor progression [157].In addition to glucose, cancer cells need Gln for their growth. This adaptive metabolism by cancer cells appears to provide substrates for an increase in lipogenesis and biosynthesis of nucleic acids which are fundamental for the proliferative phenotype of the cancer cell. Cell transformation has been shown to stimulate glutaminolysis and many cancer cells are closely dependent on this metabolic path [158]. One of the main regulators of glutaminolysis is MYC, which promotes not only cell proliferation, but also the generation of macromolecules and antioxidants necessary for the growth of cancer cells. Interestingly, MYC’s inhibition of miR-23A/B improves mitochondrial expression of glutaminase and Gln metabolism and correlates with the onset of a neoplastic phenotype [158]. Similarly, Wang et al. have demonstrated that MYC, via miR-33b induction, supports glioblastoma CSCs via the activation of mevalonate metabolism [159]. Pyruvate dehydrogenase kinase 4 is target of miR-122, a liver-specific miRNA. mir-122 is able to hamper the glycolytic metabolism in the CD133+ CSC compartment, decreasing spheroids formation and sensitizing to standard anticancer therapy [160].An increased expression of miR-210-3p correlates with colorectal cancer progression [161]. In fact, the stable overexpression of miR-210 in colorectal cancer spheroid culturesresulted in significantly enhanced CSC self-renewal activity [162]. In colon CSCs miR-210 inhibits mitochondrial TCA cycle activity to enhance lactate production [163]. The lactate stimulation leads to an increased self-renewal capacity of different colon CSC cultures [163].The rapid proliferation of cancer cells within a tumor mass leads to the formation of hypoxic regions caused by the absence of an efficient vascular network. Cancer cell survival in hypoxia requires the activation of adaptive pathways [164] and the downregulation of miRNA through a reduction of Dicer and Drosha [165,166].Within tumor is possible that miRNAs processed in the normoxic portion of a tumor diffuse toward hypoxic zones to promote tumor growth and regulate gene expression [149]. In colon CSCs, miR-210 promotes the self-renewal and reprograms cell metabolism toward a prompted glycolytic and lactate yield. In hypoxic conditions HIF1α induces miR-210-3p expression, this determines a reduced oxidative metabolism (TCA cycle and OXPHOS) [161]. miR-210 is an oncogenic miRNA and a target of HIF-1 and HIF-2 [167]. It has been shown that during hypoxia, miR-210 targets the mRNA that encodes the mitochondrial electron transport chain component protein succinate dehydrogenase complex subunit D (SDHD). Downregulation of SDHD results in an increased stabilization of HIF1α and cancer cell survival [168,169]. In the hypoxic microenvironment, miRNAs contribute to metabolic activities, to the maintenance of stemness in cancer cells and to therapy resistance. miRNAs, hypoxia and CSC are part of a complex signaling network that promotes tumor aggressiveness [170].Several miRNAs target important cancer cell regulatory molecules and are involved in an intricate network of signaling between cancer cells and the TME. In addition to their involvement in direct cell-to-cell signaling, several miRNAs are secreted through micro vesicles or exosomes and affect cancer cell growth and metastasis [165]. Broniszeet al. studied the role of miRNAs in the transformation of normal fibroblasts into CAFs, focusing on PTEN-regulated miR-320. Downregulation of miR-320 and upregulation of one of its direct targets, ETS Proto-Oncogene 2, concomitantly with loss of PTEN, is a key event in oncogenic process. In fact, this event leads to increased angiogenesis and tumor formation [171]. miR-320 was found to regulate CAF-secreted proteins, including MMP9, MMP2, lysyl oxidase homolog 2 and elastin microfibril interfacer 2, which are known to enhance tumor metastasis by programming the TME via degradation of extracellular matrices [165]. Hua Y et al., identified an important chemokine in the TME, CCL5, as target of miR-214, the latter being downregulated in CAFs. These data support the idea that miRNAs could alter the TME by changing protein production in CAFs, such as chemokines, sustaining tumor growth [172].Invasion and metastasis of cancer cells have been associated with phosphoglucose isomerase, thedownregulation of which by mir-200 in breast cancer cells impaired metastasis spread [173]. Indeed, well-known miRNAs that are downregulated in cancer belong to miR-200 family. These miRNAs are involved in many different functions, such as induction of EMT via downregulation of E-cadherin and consequent increases in zinc finger E-Box (ZEB) proteins. It was demonstrated that miR-200 influences angiogenesis indirectly via downregulation of chemokine and interleukin such as CXCL1 and IL-8, which are major players in the TME [174].Knowledge about the role of miRNAs in development and diseases, particularly in cancer, encouraged the use of miRNAs as tools or as targets for novel therapeutic approaches. For instance, several approaches are based on the use of miRNA mimics to restore the expression of tumor suppressors or antimiR to target oncogenes [169]. miRNAs mimics are synthetic double-stranded small RNA molecules that match the corresponding miRNA sequence. In human disease this therapy is used to replace the lost miRNA expression and their function. In contrast, antimiRs are single stranded and based on first-generation antisense oligonucleotides, which had been designed to target mRNAs or modified with locked nucleic acids.For example, a mimic of miR-34, whose target is lactate dehydrogenase A, is used like a tumor suppressor in Phase I clinical trials (NCT01829971).On the other hand, in Phase II clinical trials for thetreatment of hepatitis, it was used antimiRs targeting miR-122 (NCT01200420; NCT01872936; NCT02031133; NCT02508090) [169].Studies have shown that miRNAs such as let-7, miR-34, miR-451 and miR-200 can sensitize cancer cells to chemotherapy, and mimics of these miRNAs could be rationally combined with chemotherapeutic drugs [175,176,177].MRX34 was the first miRNA-based therapy undergoing clinical trial for treatment of lymphoma, melanoma, multiple myeloma, liver, lung and renal carcinoma (NCT01829971) [178]. The use of MRX34 in the aggressive KRAS TP53 mutated non-small cell lung cancer mouse model, led to significant tumor reduction. Moreover, combination treatment with the epidermal growth factor receptor (EGFR) inhibitor, erlotinib and miR-34 mimic and let-7 showed synergistic effects in inhibiting the growth of non-small cell lung cancer cell lines in vitro [177].A similar example is the MesomiR-1 drug, which reintroduces miR-16, a miRNA that regulates aldolase A in glycolysis process [179,180].Another therapy approach valuated in clinical trial is the locked oligonucleotide acid-modified inhibitor for miR-155 (MRG-106) (NCT03713320) [181]. MRG-106 replaces both miR-155 and miR-143, that negatively regulates HK2 and counteracts glycolytic phenotype [182].Given that miR-29 is frequently lost in cancer and has been reported to negatively regulates monocarboxylate transporter 1 (MCT1), a lactate transporter [183,184] it has been tempted a novel approach that involves the use of miR-29 mimic (MRG-201).In conclusion, although numerous studies have been conducted to understand the role of miRNA in the CSC metabolism, further studies are needed to define their role more precisely.The TME is characterized by a high degree of metabolic heterogeneity and dynamics that involves both cancer and microenvironmental cells. Based on the metabolic activity it is possible to classify tumors in “glycolytic” (lung, liver, colorectal, leukemia) versus “oxidative” (melanoma, glioblastoma) [185]. The complexity of this scenario is even increased by the evidence that tumors originated from the same organ—or characterized by the same genetic background—are not metabolically uniform [186,187]. This intra-tumor metabolic variability is mainly dictated by different access to oxygen and glucose, which occurs in proximity of blood vessels and by the different cell population co-existing in TME. Indeed, cancer and stromal cells can compete and/or cooperate for nutrients [188].The neoplastic tissue is constituted not only by the tumor cells, but also by the stromal cells; together they constitute the TME. CAFs, a cellular component of the TME, influence tumor growth by supplying nutrients or directly by cell-to-cell communication. In addition, CAFs provide a stromal framework to cancer cells during early growth and development, leading to malignant transformation [149].The metabolic coupling between cancer and stromal cells, which is responsible for the ‘reverse Warburg effect’, is mediated by an interchange of cytokines, metabolites and miRNAs. Stromal cells allow the creation a permissive soil that sustains cancer cell growth and fuel cancer cell energy by supplying metabolic substrates, including Gln and lactate, which enhance a specific metabolic pathway, such as OXPHOS [105,189,190] (Figure 2).In nutrient-deprived conditions, CAFs can undergo metabolic adaptation and synthesize Gln that is released in TME for cancer cell consumption, thus fostering tumor growth and metastasis formation. CAF-derived lactate, in turn, promotes OXPHOS in cancer cells, favors Gln uptake and catabolism and is responsible for the resistance to glutaminase inhibitors [191]. Monocarboxylate transporter (MCT) 4 exports lactate from CAFs, while MCT1 allows lactate uptake in CSCs [192].Another example is the alanine uptake by pancreatic cancer cells. This essential amino acid is released by pancreatic stellate cells by autophagy and fuel TCA cycle in pancreatic ductal adenocarcinoma cells for the production of essential amino acids and lipids [193].On the contrary, it seems that CSCs prefer glycolysis as source of ATP in a low oxygen microenvironment. Inhibition of glycolysis with a derivate of 3-bromopyruvate ester (pBr-PE) sensitized chemotherapy-resistant glioblastoma stem-like cells to standard therapeutic agents and counteracted tumor growth in vivo [194]. Accordingly, the activation of an EMT program repressed fructose-1,6-bisphosphatase through a Snail-G9a-Dnmt1 complex causing the production of ROS and enhancement of glycolysis under hypoxia in breast CSCs [15].Cancer metabolic reprogramming is indeed influenced by proximity to blood vessels and thus the availability of oxygen. By using a methodology capable of sorting of stroma-free glioblastoma cells and based on the distance of cells from vasculature, Kumar et al. demonstrated that glioblastoma cells positioned in close proximity to blood vessels were highly chemo- and radio-resistant and possessed an OXPHOS metabolism [195]. These observations shed light on the presence of metabolic zonation, which means that cancer cell metabolism is highly heterogeneous within the same tumor due to the presence of different metabolic niches that favor the expansion of CSCs [105,196].Besides being directly involved in nutrients exchange, CAFs secreted cytokines are involved in the activation of EMT, which is associated with metabolic reprogramming of CSCs [197]. In particular, several studies demonstrated that EMT prompts glycolytic flux in CSCs, at the expense of mitochondrial respiration [3]. Interestingly, glucose consumption by cancer cells can in turn can suppress the immune response [198] and create an acidic microenvironment that favors cancer progression [199].Beyond the cytokine-driven effects, cancer and TME cells can communicate and benefit from the exchange of vesicles-contained messengers. Exosomes are nanovesicles (40–100 nm in diameter) that are released from most cell types into the extracellular space after fusion with the plasma membrane [200,201]. Many reports have highlighted that exosomes play important roles in immune response and tumor progression [200]. It was demonstrated that cancer cells secrete a larger number of exosomes than normal cells. Tumor-derived exosomes are necessary for cell-to-cell communication. They transport a cargo of growth factors, chemokines, miRNAs and other small molecules [202,203]. Moreover, it has been demonstrated in multiple cancer models that the exposure of cancer cells to the conditioned media of CAFs promotes cancer cell growth. Besides the essential role played by CAFs in secreting factors that shape TME to allow the expansion of cancer cell population, CAFs secrete micro vesicles that act as cargo for nutrients, including intermediate metabolites, lipids and amino acids, which are ultimately absorbed by cancer cells [204]. Microvesicles released in the TME by stromal cells and loaded with several metabolites, following the uptake by cancer cells, directly promote cell proliferation and fuel tumor progression [204].Moreover, CAF-derived exosomes loaded with miRNAs impact metabolism, migration, invasion and metastasis in CSCs. Exosomal miR-21, miR-378e and miR-143, horizontally transferred from CAFs to breast CSC, encouraged the formation of mammospheres concomitantly with the acquisition EMT and stemness markers [205].Interestingly, in metastatic breast cancer, extracellular vesicles released by CAFs contain mitochondrial DNA that is internalized by CSCs and used to sustain OXPHOS and exit from quiescence induced by hormonal therapy [206].Herein we show an overview on the metabolic demand of CSCs highlighting the importance of looking at the ‘whole picture’ of tumor metabolism. CSCs possess a heterogeneous metabolism that can vary in different zones of the tumor in response to nutrients and oxygen supply.Novel frontiers of cancer treatment are focused on the use of biocompatible particles, usually conjugated with anticancer compounds, which are capable of releasing the drug directly at the tumor site. For instance, gold nanoparticles conjugated with salinomycin, caused negligible damage to healthy cells while inducing cell death of the subpopulation of CD44+/CD24− breast CSCs by inducing ferroptosis, consisting in the accumulation of iron and ROS derived from lipid oxidation [207].Thus, the identification of metabolic traits of CSCs and the adaptation of CSCs metabolism in response to a hostile or supportive TME led to the development of promising therapeutic treatments, summarized in Table 2, which interfere with CSCs metabolism and thus, disease progression.A.T., G.P., C.D. and G.S. conceived and wrote the manuscript. S.D.F., F.V., F.C., S.F., D.G., L.M. and M.T. wrote the manuscript. All authors have read and agreed to the published version of the manuscript.This work wassupported by grants from AIRC (14415) to M.T. and AIRC 5x1000 (9979), AIRC IG (16746) and PRIN (2017WNKSLR) to G.S.A.T. is a research fellow funded by European Union- FESR FSE, PON Ricerca e Innovazione 2014–2020 (AIM line 1).The authors declare no conflict of interest.Mitochondrial metabolism in cancer stem cells and therapeutic strategies.Cancer stem cells exhibit enhanced catabolism of glutamate, pyruvate and fatty acids (FAs) to generate intermediate metabolites, which converge in tricarboxylic acid (TCA) cycle. The oxidative phosphorylation (OXPHOS) (1), mediated by the electron transport chain (ETC), use the reducing equivalents (FADH2 and NADH) produced by tricarboxylic acid (TCA) cycle (2) and fatty acids oxidation (FAO) (3), with the scope of producing ATP molecules. Mitochondrial ETC is also the main endogenous source of reactive oxygen species (ROS), which possess several scavenger enzymes and molecules in CSCs such as superoxide dismutase (SOD), thioredoxin (TXN) and glutathione (GSH). These antioxidant mechanisms are targeted by Disulfiram, auranofin and L-buthionine-sulfoximine (BSO), respectively. Cancer stem cells possess a high number of mitochondria, mitochondrial ribosomes, necessary for protein synthesis are targetable with antibiotics such as tetracyclines and macrolides (4).Metabolic fingerprint of cancer stem cells and their coupling with tumor microenvironment. The tumor microenvironment and every single part of it is an important element to be taken into consideration in tumor development. Through the production and excretion by the cancer associated fibroblast (CAF) of intermediate metabolites, such as lactate and glutamine (Gln), they are able to meet the continuous and demanding energy needs of CSCs. Gln is internalized by its transporter ASCT2 and converted in glutamate (Glu), by the enzyme glutaminase (GLS), finally inducing OXPHOS and cell survival of CSCs. GLS activity is blocked by GLU inhibitors. Glu contributes to the formation of reactive oxygen species (ROS) inhibitor molecule glutathione (GSH) by acting as an exchange ion, through the CD44v-xCT transporter, for the import of cystine, which is necessary for the production of GSH together with Glu. The uptake of glucose through the GLUT foster a glycolytic phenotype in CSCs boosting cell survival. The administration of the glycolysis inhibitors 2-deoxyglucose (2-DG), 3-bromopyruvate ester (pBr-PE) or miR-143 proved to effective in dampening the expansion of the CSC subpopulation.The crosstalk of CAF and CSC is possible through CAF-derived exosomes/miRNAs, including miR-21, miR-378e and miR-143. CAF-derived exosomes/miRNAs are internalized by CSC, sustaining stemness and positively influences tumor progression. Conversely, miR-320 counteracts pre-metastatic niche formation by CAFs.Adipocyte-secreted adipokines are taken upby CSC and stimulate the production and release of exosomes and pro-lipolytic factors, including miRNA-144 and miRNA-126, which in turn active lipolysis and autophagy pathways in adipocytes. Consequently, free fatty acids (FFA) are released in the tumor microenvironment (TME) that, upon absorption by CSCs through the fatty acid transporters (fatty acid translocase, FAT/CD36; fatty acid transport protein, FATP and fatty acid-binding protein, FABP), undergo fatty acid oxidation (FAO) to sustain CSCs expansion and metabolic reprogramming.Metabolic phenotypes of cancer stem cells in various cancer models.Metabolic Therapeutic Targets of Cancer Stem Cells.

Med-MDPI/cancers/cancers-12-06-01437.txt

@@ -0,0 +1 @@
+These authors contributed equally to this work.Cancer therapies induce differential cell responses, ranging from efficient cell death to complete stress resistance. The BCL-2 proteins BAX and BAK govern the cellular decision between survival and mitochondrial apoptosis. Therefore, the status of BAX/BAK regulation can predict the cellular apoptosis predisposition. Relative BAX/BAK localization was analyzed in tumor and corresponding non-tumor samples from 34 hepatocellular carcinoma (HCC) patients. Key transcriptome changes and gene expression profiles related to the status of BAX regulation were applied to two independent cohorts including over 500 HCC patients. The prediction of apoptotic response was tested using cell lines and polyclonal tumor isolates. Cellular protection from BAX was confirmed by challenging cells with mitochondrial BAX. We discovered a subgroup of HCC with selective protection from BAX-dependent apoptosis. BAX-protected tumors showed enrichment of signaling pathways associated with oxidative stress response and DNA repair as well as increased genetic heterogeneity. Gene expression profiles characteristic to BAX-specific protection are enriched in poorly differentiated HCCs and show significant association to the overall survival of HCC patients. Consistently, addiction to DNA repair of BAX-protected cancer cells caused selective sensitivity to PARP inhibition. Molecular characteristics of BAX-protected HCC were enriched in cells challenged with mitochondrial BAX. Our results demonstrate that predisposition to BAX activation impairs tumor biology in HCC. Selective BAX inhibition or lack thereof delineates distinct subgroups of HCC patients with molecular features and differential response pattern to apoptotic stimuli and inhibition of DNA repair mechanisms.Hepatocellular carcinoma (HCC) ranks among the most common and rapidly evolving cancers in the Western world [1]. The majority of HCCs emerge as the consequence of chronic inflammatory liver damage that promotes dysregulation of multiple cellular signaling pathways [2]. Herein, alterations of the hepatic microenvironment induce changes of the balance between proliferation and cell death, which overall predisposes malignant transformation. Therefore, failure of immunological clearance of damaged pre-neoplastic hepatocytes as well as the emergence of resistance to death signaling can be considered hallmarks of hepatocarcinogenesis [3]. The molecular profile of resulting HCCs frequently reflects the underlying type, and the magnitude of cell damage with consecutive induction of cytotoxic stress response. The resulting heterogeneity with diverse genetic alterations and various different phenotypes also induces adaptive changes that confer properties of chemoresistance in cancer cells and, therefore, impairs the development of effective therapeutic strategies [2].Cytotoxic cell stress can induce apoptosis dependent on the proapoptotic BCL-2 proteins BAX and BAK [4]. The activation of BAX and BAK induces permeabilization of the outer mitochondrial membrane (OMM) and activation of the caspase cascade [5]. Human cells protect themselves from BAX/BAK activity by constant shuttling of BAX and BAK from the mitochondria into the cytosol [6,7,8]. This BAX/BAK retrotranslocation depends on prosurvival BCL-2 proteins [8,9]. The retrotranslocation rate determines the amount of mitochondrial BAX/BAK and cellular apoptosis response [8,9]. BAX shuttling is typically faster than BAK retrotranslocation in cultured cells. Therefore, BAX resides predominantly in the cytosol and BAK on the mitochondria. In addition, prosurvival BCL-2 proteins can sequester “activator” BH3-only proteins, thereby preventing BH3-only protein mediated activation of BAX and BAK [10,11,12,13,14,15]. The BCL-2 protein interplay can be characterized by the titration of BH3 mimetic peptides in lysed cells, termed “BH3 profiling”, to identify the most suited BH3 mimetic to induce apoptosis in a tumor [16,17]. Apoptotic response to cell stress and therapeutic outcome can be predicted by the analysis of relative BAX/BAK localization [18]. Furthermore, changes to BCL-2 protein regulatory network and compensatory emergence of antiapoptotic effectors are among the most prominent changes during liver cancer development [19]. However, the molecular underpinnings as well as the subsequent molecular alterations induced by disruption of BCL-2 family members including potential clinical and therapeutic implications remain elusive in HCCs.We here demonstrate that changes to the localization equilibrium of BAX are frequent during hepatocarcinogenesis. Resulting resistance to cytotoxic stress divides tumors into BAX-protected and non-protected HCCs with differential sensitivity towards apoptotic stimuli. While non-protected HCC display disruption of mitochondrial function and high levels of mitochondrial BAX protected HCC show predominant cytosolic BAX localization and experience elevated oxidative stress. Furthermore, higher proliferative capacity renders non-protected HCC sensitive to classical chemotherapy. In contrast, continuous cellular stress and increased DNA damage with subsequent activation of DNA repair mechanisms led to increased susceptibility to PARP-inhibitors in BAX-protected HCCs. We first analyzed whether BAX/BAK regulation determines the natural course of disease in HCC. Therefore, relative BAX/BAK localization was measured in a cohort of HCC tumor isolates and corresponding non-tumor tissue (Figure 1A). Relative mitochondrial (BAX: cyan; BAK: yellow) and cytosolic protein (BAX: Red; BAK: Orange) were combined to relative protein localizations (relative BAX localization (blue); relative BAK localization (green)) to cross-compare samples as described previously [18]. High values point to increased mitochondrial BAX/BAK localizations and, thus, increased apoptosis predisposition. The results show a strong correlation between BAX and BAK localization in tumor and non-tumor samples (Figure 1B). The BAX localization shows a slight inverse correlation to total protein levels in the tumor consistent with human AML [18]. However, the opposite tendency is apparent in corresponding non-tumor cells (Figure 1C and Figure S1). BAX is shifted towards the cytosol of tumors compared to BAK (Figure 1D). This shift is expected based on the differential localization of both BCL-2 proteins in cell culture [5]. Surprisingly, no significant difference between BAX and BAK can be detected in non-tumor tissue (Figure 1E). These results suggest that BAX and BAK retrotranslocate at similar rates in non-tumor cells. Therefore, the establishment of the differential BAX/BAK localization seems to occur during malignant transformation and result from specific acceleration of BAX retrotranslocation.Mitochondrial BAX in non-tumor tissues associates with strong cytosolic shifts to corresponding tumors (Figure 1F). Uniform low mitochondrial BAX level in HCC cells point to a specific selection for cytosolic BAX (Figure 1G and Figures S2–S4). By contrast, BAK shows no tendency to shift to the cytosol in tumors but indicates stress-induced shifts to the mitochondria (Figures S5–S7). The importance of limiting the mitochondrial BAX pool is particularly apparent for a subgroup of tumors with high mitochondrial BAX levels in corresponding non-tumor cells (Figure 1H,I and Figure S8). This subgroup is selected for minimizing the mitochondrial BAX pool manifested in a predominant cytosolic BAX localization in tumor cells to protect itself specifically from BAX activation. The selection yields in a BAX localization shift typically of about 2 log scales to the cytosol, regardless of the apoptosis predisposition in non-tumor cells and BAK regulation. Therefore, these tumors will be referred to as BAX-protected HCC.The regulation of the BAX localization could not only influence overall apoptosis predisposition, but also differential sensitivity to apoptotic stimuli. Key transcriptome changes and specific signaling pathways affected in BAX-protected and non-protected tumors regarding their BAX regulation were identified. A total of 66 genes were differentially expressed between BAX-protected and non-protected tumors in non-tumor tissue and 47 genes in tumor tissue. The results suggest a common molecular underpinning of selective BAX inhibition (Figure 2A). Further, Ingenuity Pathway Analyses demonstrated that differences in key signaling pathways and gene sets affected in non-tumorous tissue between BAX-protected tumors and non-protected tumors included mitochondrial dysfunction and apoptosis, whereas changes in tumor tissues were associated with proliferation, oxidative stress response and DNA damage (Figure 2B,C). Therefore, cellular BAX localization was characterized by key changes in apoptosis resistance within the hepatic microenvironment that potentially predisposed malignant transformation. Resulting tumors showed distinct patterns of molecular changes in hallmark oncogenic signaling validating the functional differences in BAX regulation and the tumor classification. To investigate whether the differential BAX localization has a potential impact on biological traits of tumors, we integrated our results with two independent cohorts of authentic HCC patients and assessed clinical outcomes by sub-clustering the tumors based on BAX-protected gene expression signatures from non-tumor and tumor tissues. Notably, survival data was only available for the minority of patients in the corresponding cohort used to establish the signatures due to short or lost follow-up, thus preventing detailed clinical analyses in this cohort. However, a significant association to overall survival of patients could be revealed for both the non-tumor (p-value < 0.0001) as well as tumor (p-value = 0.0014) signature in both independent patient cohorts (Figure 2D) [20,21]. Moreover, gene set enrichment analyses confirmed that protection associated gene expression signatures during malignant transformation were highly enriched in patients with poor differentiation and adverse clinical outcome represented by the well-established prognostic subclass of tumors, i.e., the subtype A defined by Lee et al. (Figure 2D,E) [21]. Thus, protection against BAX-dependent apoptosis is characteristic for a subclass of patients with adverse tumor features and poor clinical outcome.Next, we individually compared the transcriptome profile between non-tumorous and tumor tissue to dissect differences during malignant transformation and identify potential molecular drivers in BAX-protected cancers. A total of 726 genes in the subgroup of BAX-protected tumors and 2029 genes in the subgroup of non-protected tumors showed differential regulation in tumor vs. non-tumorous tissue, effectively separating tumor from non-tumorous tissue in both groups (Figure 3A). Major cellular processes in non-protected tumors involved activation of metabolic processes and protein synthesis (Figure 3B, Table 1) and centered around apoptosis, proliferation, and stress response pathways (Figure 3C). Further, this group showed enrichment in gene sets involved in mTOR and MYC signaling (Figure S9). In contrast, BAX-protected tumors showed downregulation of apoptotic signaling, but activation of DNA replication, recombination, and repair pathways (Figure 3C, Table 1). In addition, enrichment of gene sets associated with DNA damage as well as BRCA signaling was observed (Figure S9). Accordingly, a trend for increased genetic alterations has been observed in BAX-protected tumors (Figure 3D,E).Next, we tested whether cells challenged with mitochondrial BAX would display a BAX-protected gene expression pattern using HCT116 BAX/BAK DKO cells devoid of intrinsic BAX and BAK [22] challenged with either BAX S184E predominantly localized in the cytosol or BAX S184V with largely mitochondrial localization (Figure S10). Transient expression of both BAX variants confirms that cells with predominant mitochondrial BAX show increased apoptosis predisposition and are therefore exposed to increased selection pressure (Figure 4A and Figure S11). The BAX-protected gene signature suggests a strong dependence on DNA damage repair. We, therefore, analyzed the response of cells to induced DNA damage by DNA-intercalating Daunorubicin and inhibition of DNA damage repair by the FDA-approved inhibitor of the poly ADP ribose polymerase (PARP) Olaparib alone and together. In parallel to BAX-protected HCCs, establishment of stable BAX S184V expression results in reduced apoptosis predisposition (Figure 4B,C and Figure S12). Consistently, gene expression profiles of BAX S184V-challenged cells showed activation of mitochondrial networks and DNA repair as well as apoptosis and proliferation (Figures S13 and S14). Furthermore, in accordance with HCC patients, HCT116 cells with cytosolic BAX expression, representing the non-protection subgroup, showed enrichment of gene sets involved in apoptosis as well as cell cycle. The analysis of transiently vs. stably expressed wild type BAX, however, does not reveal a clear difference in localization (Figure 4D,E, Figures S15 and S16). The precise analysis of the mitochondrial BAX pool by carbonate extraction separating BAX in loose association with the mitochondria in the carbonate supernatant from OMM-integral BAX in the carbonate pellet shows a prominent shift (Figure 4F,G, Figures S15 and S17). Stably expressed wild type BAX is shifted significantly to the mitochondria-associated fraction, suggesting selection towards minimizing the membrane-integral BAX pool. These results validate that challenging cells with mitochondrial BAX resembles the selection of BAX-protected HCC overall confirming the importance of BAX localization for subclassification and therapeutic implications of HCC patients.To validate a functional role of BAX/BAK regulation in primary cell lines, cultured tumor isolates from three different patients were analyzed [23]. Although reflecting largely the tumors, these polyclonal isolates underwent significant changes towards similar BAX/BAK regulation after initial cell culturing (Figure 5A and Figure S18). However, the link between intracellular BAX shuttling and sensitivity to apoptosis induction remains apparent (Figure 5B). In parallel, established hepatoma cell lines reflect only a small fraction of BAX/BAK regulation present in human cells (Figure 5C). Despite genetic differences, HUH7 cells and Hep3B cells display similar relative BAK localization. Therefore, differences in their mitochondrial apoptosis regulation are dependent on BAX. HUH7 cells with largely cytosolic BAX show low and delayed induction of caspase 3/7 activity by kinase inhibitors and topoisomerase inhibitors, whereas Hep3B cells with increased mitochondrial BAX pools show strong apoptotic response (Figure 5D, Figures S19 and S20). Targeting the BCL-2 family with the BH3 mimetics ABT-737 and UMI-77 did not induce differential response alone, suggesting apoptosis predisposition is independent of BCL-2 protein levels (Figure S21).Next, differentially regulated pathways in BAX-protected vs. non-protected tumors were targeted. To this end, HUH7, Hep3B, and HCC68 cells were tested, because they reflect the full range of BAX regulation in cultured hepatoma cells but show similar BAK localization, connecting differential apoptotic responses to BAX regulation (Figure 5C). Upregulation of mTOR signaling in non-protected tumors suggests increased susceptibility to the mTOR inhibitor rapamycin. Among the tested cell lines, HCC68 with predominant mitochondrial BAX show the highest sensitivity towards rapamycin and HUH7 cells with largely cytosolic BAX show the lowest apoptosis response with significantly less caspase activation (Figure 5E). Based on the increased activation of DNA repair in BAX-protected tumors, the PARP inhibitor Olaparib was used (Figure 5F). HUH7 cells are sensitive to 10 µM and higher concentrations of Olaparib with significant increase in apoptosis and reduced colony formation. Notably, HUH7 sensitivity to Olaparib occurs despite insensitivity to a large range of cell stresses (Figure 5D, Figures S20 and S21). Therefore, HUH7 cells reflect BAX-protected tumor cells, whereas Hep3B cells and HCC68 cells show characteristics of non-protected tumor cells. It should be noted that these differences originate not necessarily from the present relative BAX/BAK localization, but from processes during their tumorigenesis. Combination of Olaparib with DNA-intercalating Daunorubicin induces synergistic caspase activation further emphasizing the sensitivity of HUH7 cells to DNA repair inhibition (Figure 5G). Olaparib-dependent enhancement of Daunorubicin-induced apoptosis is particularly remarkable, given that apoptosis response to Olaparib alone occurs delayed. Interestingly, Olaparib administration induces a shift of BAX towards the cytosol (Figure 5H,I and Figure S22). However, BAX localization in Olaparib-sensitive HUH7 cells is restored after 24 h, while Hep3B and HCC68 cells maintain their BAX localization, suggesting lower stress response capacity in BAX-protected cells. Together, these results show that BAX-protected cells show higher susceptibility to DNA repair inhibition by Olaparib.Resistance to cell death is a recurrent hallmark of many cancers including primary liver cancer [24]. We here demonstrate regulation of BAX, and to a lesser extent of the functionally redundant BAK, within the diseased hepatic microenvironment during malignant transformation. In a subgroup of patients, BAX shifts to the cytosol accompanied by fundamental molecular changes in the tumor cells. Despite common cellular control of BAX and BAK particular low mitochondrial BAX levels are apparent in tumor cells (Figure 1). Our analysis identifies increased apoptosis predisposition through mitochondrial BAX prior hepatocarcinogenesis as a potential driver. Therefore, an increased apoptotic predisposition prior tumorigenesis could result in selection towards a BAX-protected tumor. Differences in the equilibrium between cytosolic and mitochondrial BAX delineates distinct subgroups of HCC patients and induces subsequently/as a downstream effect substantial changes in the transcriptome profile of respective patients. Challenging cells with mitochondrial BAX S184V supports these results (Figure 4). As BAX localization is dependent on BAX retrotranslocation, an increased presence of prosurvival BCL-2 proteins on the mitochondria could contribute to this process. However, it is not obvious how altered prosurvival BCL-2 protein levels on the mitochondria would inhibit BAX selectively and not both proapoptotic BCL-2 proteins. On the other hand, BAX mutants, for instance, of the C-terminal TMD, are not common in tumors. Therefore, the molecular underpinning of selective BAX inhibition in tumors will be an important subject of future research.Strikingly, a significant shift between the localizations of BAX and BAK is not apparent in non-tumor tissue, contrasting the differential BAX/BAK localizations in cultured cells [25,26,27,28,29]. BAX and BAK shuttle between cytosol and mitochondria with different retrotranslocation rates, manifesting differential steady state localization [6,7,8]. However, BAX/BAK regulation in human patients has a much wider variety of scenarios. Tumorigenesis of a subgroup of tumors could establish differential regulation of BAX and BAK, perhaps through stress and BH3-only protein-dependent signaling [16,30]. In addition to a significant resistance to apoptotic stimuli, BAX-protected tumors show increased chromosomal instability with subsequent oncogenic dependence on DNA repair (Figure 5). This is mirrored when cells are challenged with mitochondrial BAX. BAX/BAK-dependent apoptosis induction during chronic liver damage in response to oncogene activation, DNA damage, and senescence is well-established and a key mechanisms of cancer prevention [31]. In contrast, evasion of cell death and proapoptotic stimuli in non-transformed hepatocytes is frequently observed during liver cancer development and accompanied by downregulation of BAX [32]. We here demonstrate that transformation of diseased hepatocytes favors enhanced BAX retrotranslocation, identifying a previously unrecognized mechanism during hepatocarcinogenesis that is independent of direct transcriptional regulation. Differences in the equilibrium between cytosolic and mitochondrial BAX delineates distinct subgroups of HCC patients and induces substantial changes in the transcriptome profile of respective patients. BAX-protected tumors show activated oxidative stress and DNA damage suggesting that alterations within the hepatic microenvironment potentially predispose malignant transformation of HCC. These observations are in accordance with the causative role of inflammatory cell death in disease development and progression observed in the majority of HCC [33]. Consistently, GSEA analyses demonstrated that the BAX-protected profile was highly enriched in a prognostic subtype of 139 patients with aggressive tumor biology and poor differentiation [21]. Furthermore, integrative molecular analyses confirmed the prognostic impact of transcriptomic changes that conferred protection against apoptotic stimuli. Therefore, different therapeutic strategies might be required according to the apoptosis regulation of patients to induce treatment effect. While metabolic changes and classical oncogenic pro-proliferative signaling in the non-protection group suggests response to classical chemotherapy and MTOR inhibitors, the dependence on DNA repair and “BRCAness” in BAX-protected tumors might confer high sensitivity to PARP inhibitors. Consistently, HUH7 cells resembling BAX-protected tumors showed selective sensitivity to PARP inhibition by Olaparib. Blots for quantification were detected with LAS400 CCD camera or Vilber Lourmat Solo S, quantification was performed using ImageJ. Proteins were normalized using a titration of the same HeLa whole cell lysate standard curve within each individual blot; BAX (E63, Abcam, Cambridge, MA, USA) and BAK (EMD Millipore, Billerica, MA, USA) were then quantified using SigmaPlot™. The HeLa cell lysate serves as standard protein mix to provide consistent amounts of all analyzed proteins for of Western blot standardization. The amounts of mitochondrial and cytosolic protein were then determined as relative to COX IV (Invitrogen) and β-ACTIN (Millipore), respectively. Finally, ratios were built from the mitochondrial and cytosolic values to obtain the relative protein localization, to allow comparison of samples analyzed not on the same blot, relative mitochondrial and relative cytosolic protein was determined using the fractionation loading controls COXIV and Actin. Tissue from 34 patients with confirmed HCC undergoing resection at the Department of Surgery, University of Mainz, Germany were collected following patient informed consent and local ethics committee approval. This research has been approved by the ethic committee of the Landesärztekammer Rheinland-Pfalz on 05/15/13 (ethic code: 837.199.10 (7208). Detailed description of the cohort can be found [33]. Total RNA was extracted using the Qiagen RNEasy mini Kit (Qiagen GMBH, Hilden, Germany) following the manufacturer’s instructions. RNA quantity and purity were estimated using a Nanodrop ND-1000 Spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA), and integrity was assessed by Agilent 2100 Bioanalyzer (Agilent, Palo Alto, CA, USA). DNA was extracted using Qiagen Qiamp DNA Kit (Qiagen GMBH, Hilden, Germany) following the manufacturer’s instructions.Hepatocellular tissue patient samples (tumor and surrounding tissue) were extracted during surgery and samples were immediately shock frozen and stored in liquid nitrogen. Samples were thawed on ice and washed with ice-cold 1 × PBS. Tissues and cells were resuspended in SEM buffer (10 mM HEPES, 250 mM sucrose, pH 7.2) supplemented with protease inhibitors for 20 min on ice and homogenized using the MINILYS (PEQLab, Erlangen, Germany) glass mill system. Subsequently, samples were centrifuged at 1500× g for 5 min at 4 °C. The supernatant was transferred with a 30 G cannula to a 1.5 mL tube and centrifuged for 20 min at 13,000× g at 4 °C. While sedimented mitochondria were washed with SEM buffer, the supernatant was ultracentrifuged at 150,000× g for 1 h at 4 °C to obtain the cytosolic fraction. Finally, the samples were prepared in SDS sample buffer and separated by SDS-PAGE (NuPage Novex 4–12% Bis-Tris Midi) and subjected to Western blot analysis.Mitochondrial pellets were resuspended in 100 mM Na2CO3 at pH = 11.5 and incubated on ice for 20 min. Subsequently, membranes were pelleted at 15,000× g for 30 min at 4 °C. The supernatant, OMM-associated proteins, was subjected to protein precipitation by acetone. The pellet resuspended once more in 100 mM Na2CO3 at pH = 11.5, incubated on ice for 20 min and centrifuged at 15,000× g for 30 min at 4 °C to obtain OMM-integral proteins. Both fractions were assayed by Western blot.Expression values were extracted from Gene Expression Omnibus database (http://www.ncbi.nlm.nih.gov/geo, accession number: GSE84598). Gene expression values were normalized by quantile normalization method across all samples following subtraction of background noises in each spot by GenomeStudio (Illumina®). Signal intensity with a detection p > 0.05 was treated as a missing value, and only genes with sufficient representation across the samples were included in further data analysis. Differentially expressed genes were determined using Student’s t-test included in R version 3.3.3.Hierarchical cluster analyses were based on Pearson Correlation and complete-linkage clustering, including a filter of 80% presence for each gene. Results were visualized with Complexheatmap (version 1.12.0) [34]. Ingenuity Pathway Analysis (Ingenuity Systems Inc.) tools were used for functional classification and network analyses. The significance of each network, function and pathway was determined by the scoring system provided by Ingenuity Pathway Analysis tool. Gene Set Enrichment analysis (GSEA) was performed using GSEA software provided by Broad Institutes (http://www.broad.mit.edu/gsea/). Gene sets with a NOM p-value < 0.05 and FDR < 0.25 were considered significantly enriched in a priori defined set of genes.Genomic analyses were performed using GenomeStudio (version 2011.1). All SNVs with call Frequencies > 0.95 and Gen Train Score > 0.7 were included for further analyses. CNV were detected by cnvPatition (version 3.2.0) Plugin GenomeStudio. LOH were detected using Bioconductor package VegaMC (version 3.22.0) [35].Cells were treated according to the outlined protocol. Whole-cell lysates were prepared as described and incubated with Caspase 3/7 substrate (BD Pharmingen) for 60 min at 37 °C in a 96-well plate (OptiPlate, Perkin Elmer) and protein concentration was determined by a Bradford Assay (Roth). Substrate cleavage was measured for 50 cycles with 10 s delay (excitation at 380 nm, emission at 430–460 nm, Victor X4, Perkin Elmer). Kinetics were measured and calculated to the amount of protein per sample. Cells were treated according to the outlined protocol. For PARP cleavage analysis the samples were precipitated by acetone, boiled in 3× SDS-sample buffer for 10 min at 95 °C, and subjected to SDS-PAGE and Western blot analysis, using anti-Actin C4 (Sigma) and PARP polyclonal (Cell Signaling).Cells were seeded to 70% confluency and treated with indicated concentrations of Olaparib (72 h). Medium was exchanged and cells were allowed to recover for 24 h, then cells were splitted and transferred to a new 6-well plate. Cells were cultured for 10–12 days followed by fixation with 4% PFA in 1× PBS and staining with 1% Methylene Blue in 50% methanol. Colonies were identified and counted using ImageJ software.HeLa cell lines were cultured in DMEM 1 g/L glucose medium supplemented with 10 mM Hepes and 10% heat-inactivated fetal bovine serum in 5% CO2 at 37 °C. Hep3B, HepG2, and HuH7 cell lines were cultured in DMEM (Gibco, Thermo Fisher Scientific Inc.) medium supplemented with 10 mM Hepes (BioWest), 10% heat-inactivated fetal bovine serum (Biochrom, Schaffhausen, CH), and 1 g/L glucose in 5% CO2 at 37 °C, and the patient-derived cell lines (HCC9, HCC31, and HCC68) were supplemented with 5% heat-inactivated fetal bovine serum. Mammalian cell were seeded on 15 cm dishes and harvested at 80% confluency or indicated time of treatment with a cell scraper in ice cold 1× PBS, transferred to a 50 mL falcon tube and cells were pelleted by centrifugation (1200× g for 5 min at 4 °C). Cells were regularly (in 4 week intervals) tested for potential mycoplasma infection using the Venor GeM kit (Biochrome).BAX constructs were cloned into the pEGFP.C3 vectors that were kindly provided from RJ. Youle Lab, NIH, NINDS, USA, the BAX C-terminal serine 184 residue was substituted with valine residue to direct BAX to the mitochondria or with glutamic acid to direct BAX towards the cytosol. HCT 116 BAK−/− BAX−/− cells were cultured in McCoy’s 5A medium supplemented with 10% FCS and 1 mM HEPS at 37 °C in 5% CO2. Cells were transfected with Turbofect (Thermo Fisher Scientific Inc.) with the mutation bearing constructs of BAX, according to the manufacturer’s instructions. The day before the experiment HCT116 DKO cells were seeded into the 6-well plates or 15 cm dishes, which had the 70–80% of confluency at the time of transfection. G418 (1000 µg/mL) was added and refreshed every second day, to induce stable transfection. Cells were treated for 4 h or 24 h to induce apoptosis and analyzed by Caspase-3/7 activity assay. HeLa cell lysate was used to establish a standardized mix of all analyzed proteins, ensuring similar standardization of all patient samples. Cells were harvested, washed with ice-cold 1× PBS and subsequently resuspended in cell lysis buffer (20 mM Tris, 100 mM NaCl, 1 mM EDTA, 0.5% Triton X-100, pH 7.5) supplemented with protease inhibitors. Upon incubation on ice for 15 min, the samples were centrifuged at 15,000× g for 10 min at 4 °C. The supernatants were subjected to acetone precipitation, followed by resuspension in SDS sample buffer and storage at −80 °C.Statistical analysis was performed using Student’s t-test, Friedman test for multiple group comparisons, followed by Dunn’s post hoc test or one-way ANOVA using Holm–Sidak method as indicated. p-values ≤ 0.05 were considered statistically significant. Results are presented as means ± SD or means ± SEM as indicated. Survival analyses were performed using log rank (Mantel-Cox) tests.For integration of patients, publically available expression data sets were used [36]. Hierarchical cluster analyses were performed using Euclidean distance by Bioconductor package multiClust (version 1.4.0). Missing values were computed by k-Nearest Neighbour Imputation with CRAN package VIM (version 4.7.0) [37]. Survival analyses were performed by CRAN package survival und survminer (version 0.4.3) using log rank (Mantel-Cox) tests.Together, we discover a previously unrecognized link between cellular BAX localization and apoptosis resistance that characterizes a differential mechanism of malignant transformation in affected patients and links adverse tumor biology to high genetic instability and poor outcome. The following are available online at https://www.mdpi.com/2072-6694/12/6/1437/s1, Figure S1: Relative BAX localization in non-tumor samples vs. relative BAX level in non-tumor tissue (double log10 scale) is displayed for 34 patients. Pearson’s correlation is shown by r- and p-values, Figure S2: Relative cytosolic BAX (S2) and BAK (S3) levels are displayed for tumor (T) and non-tumor (N) samples from 34 HCC patients. p-values according to t-test is displayed, Figure S3: Relative cytosolic BAX (S2) and BAK (S3) levels are displayed for tumor (T) and non-tumor (N) samples from 34 HCC patients. p-values according to t-test is displayed, Figure S4: Relative mitochondrial BAX levels in tumor tissues from HCC patients with predominantly cytosolic (C, N = 11), neutral (N, N = 12) and mostly mitochondrial (M, N = 11) BAX localization in non-tumor tissue. Mean is displayed, Figure S6: Changes from non-tumor to tumor mitochondrial BAX levels vs. changes from non-tumor to tumor cytosolic BAX levels in 34 HCC patients (double log10 scale). r- and p-valuess according Pearson’s correlation are displayed, Figure S7: Changes between non-tumor and tumor mitochondrial BAK levels vs. changes between non-tumor and tumor cytosolic BAK levels (double log10 scale). r- and p-valuess according Pearson’s correlation are displayed, Figure S8: Relative mitochondrial BAK levels are displayed for tumor (T) and non-tumor (N) samples from HCC patients of the protected group (N = 11), Figure S9: Gene set enrichment analysis (GSEA) for protection and non-protection associated gene expression signatures during malignant transformation were performed. Normalized enrichment score (NES) reflects degree of overrepresentation for each group at the peak of the entire set. Statistical significance calculated by nominal p-values of the ES by using an empirical phenotype-based permutation test, Figure S10: Subcellular localization of BAX S148E and BAX S184V expressed for 4 h in HCT116 BAX/BAK DKO cells after fractionation analyzed by Western blot. Separation of cytosol and heavy membrane fraction (mitochondria) was controlled using AKT1 and COX IV, respectively. N = 3, Figure S11: Caspase 3/7 activity induced by 1 µM Rapamycin in HCT116 BAX/BAK DKO cells expressing either the largely cytosolic BAX S184E (green) or the predominantly mitochondrial BAX S184V (red) for 4 h. Data ± SEM. N = 3 and p-values according to One Way ANOVA using the Holm-Sidak method are displayed. Data was adjusted to BAX variant expression levels based on GFP fluorescence, Figure S12: (A) Caspase 3/7 activity induced by 1 µM Rapamycin in HCT116 BAX/BAK DKO cells with either transient (red, trnst) or stable (yellow, stable) expression of predominantly mitochondrial BAX S184V. Data ± SEM. N = 3 and p-valuess according to One Way ANOVA using the Holm-Sidak method are displayed. Data was adjusted to BAX S184V expression based on GFP fluorescence. (B)Uncropped Western blot depicted in Figure 3H with density quantifications below (UB—upper band, LB—lower band), Figure S13: set enrichment analysis (GSEA) for protection and non-protection associated gene expression signatures during malignant transformation were performed. Normalized enrichment score (NES) reflects degree of overrepresentation for each group at the peak of the entire set. Statistical significance calculated by nominal p-values of the ES by using an empirical phenotype-based permutation test, Figure S14: set enrichment analysis (GSEA) for protection and non-protection associated gene expression signatures during malignant transformation were performed. Normalized enrichment score (NES) reflects degree of overrepresentation for each group at the peak of the entire set. Statistical significance calculated by nominal p-values of the ES by using an empirical phenotype-based permutation test, Figure S15: Uncropped Western blot depicted in Figure 4D,F, Figure S16: Density quantification of Western blot depicted in Figure 4D, Figure S17: Density quantification of Western blot depicted in Figure 4F, Figure S18: Relative BAX localizations (log10 scale) of polyclonal cell lines (HCC9, HCC31, HCC68) in early (E) and late (L) passages are displayed, Figure S19: Caspase 3/7 activity of HUH7 cells and Hep3B cells after treatment with 1 µM ActD or 50 µM Etoposide. Data ± SEM. N = 3, Figure S21: Caspase 3/7 activity resulting from treatment of HUH7 cells and Hep3B cells with 1 µM ABT-737 or 1 µM UMI-77. Data ± SEM. N = 3, Figure S22: Relative BAX localization of HCC68 cells in the presence of either Olaparib (black) or Rapamycin (gray) monitored at 1, 6 and 24 h. Data represent averages ± SEM. N ≥ 3.Conceptualization, F.E. and J.U.M.; methodology, K.F., C.C., R.K., D.B., J.H., A.O., F.R., F.F.-T. and J.L.; software, C.C.; validation, C.C., D.B. and J.U.M.; formal analysis, C.C.; investigation, K.F. and C.C.; resources, P.R.G.; data curation, C.C.; writing—original draft preparation, K.F.; writing—review and editing, F.E., J.U.M. and C.C.; visualization, F.E., J.U.M. and C.C.; supervision, F.E. and J.U.M.; project administration, F.E.; funding acquisition, F.E., J.U.M., C.C. and A.O. All authors have read and agreed to the published version of the manuscript.F.E. was supported by the DFG Heisenberg program, the Collaborative Research Cluster (CRC) 746, the Else Kröner-Fresenius-Stiftung, and Germany’s Excellence Strategy (CIBSS–EXC-2189–Project ID 390939984). J.U.M. is supported by grants from the German Research Foundation (MA 4443/2-2; SFB1292) and the Volkswagen Foundation (Lichtenberg program). J.U.M. and F.E. are supported by a grant from the Wilhelm-Sander Foundation (2017.007.1). C.C. is supported by a TransMed Fellowship of the University of Mainz. A. O. was supported in part by the Excellence Initiative of the German Research Foundation (GSC - 4, Spemann Graduate School) and in part by the Ministry for Science, Research and Art s of the State of Baden - Wuerttemberg.The authors declare no conflicts of interest.(A) BAX and BAK constantly translocate to the mitochondria and retrotranslocate back into the cytosol. Localization equilibriums are characterized by determining relative mitochondrial protein (rel. mitochondrial BAX (cyan), mitochondrial BAX (blue circle) per COX IV (light gray rectangle), rel. mitochondrial BAK (yellow), mitochondrial BAK (green circle) per COX IV (light gray rectangle)), and relative cytosolic protein (rel. cytosolic BAX (red), combining cytosolic BAX (blue cycle) with β-ACTIN (dark gray rectangle); rel. cytosolic BAK (orange), combining cytosolic BAK (green cycle) with β-ACTIN (dark gray rectangle)). Relative protein localization (BAX: blue; BAK: green) is the quotient of relative mitochondrial protein and relative cytosolic protein. High relative protein localization values indicate shift towards the mitochondria independent of the cellular protein concentration. (B) Relative BAX localization vs. relative BAK localization (double log10 scale) for 34 HCC patient samples (full circle) and corresponding non-tumor samples (open circle). R- and p-values according Pearson’s correlations. (C) Relative BAX localization in tumor tissue vs. relative BAX level in tumor samples (double log10 scale) for 34 patient samples. R- and p-values according Pearson’s correlations. (D) Relative BAX/BAK localization in 34 HCC patient tumor tissues (log10 scale). p-value according to t-test. (E) Relative protein localization (log10 scale) of BAX and BAK in 34 HCC patient non-tumor tissue samples. p-value according to t-test. (F) Relative BAX localization in non-tumor tissue vs. changes in relative BAX localization between non-tumor and tumor samples from 34 HCC patients (double log10 scale). R- and p-values according Pearson’s correlation. (G) Relative mitochondrial BAX (left) and BAK (right) levels for tumor (T) and non-tumor (N) samples from 34 HCC patients. p-value according to t-test. (H) Changes from non-tumor to tumor relative BAX localization of 34 HCC categorized according to the mitochondrial BAX levels in non-tumor tissue (N = 8 for extreme levels each and N = 9 for moderate levels each). p-value according to One Way ANOVA using Holm-Sidak method. (I) Relative mitochondrial BAX levels for tumor (T) and non-tumor (N) samples from 11 HCC patients with pronounced cytosolic BAX shift from non-tumor to tumor samples compared to samples without this shift (N = 23).(A) Unsupervised clustering analyses based on genes differently regulated between BAX-protection and non-protection subgroups in non-tumorous and tumor tissue. (B) Activated signaling pathways between BAX-protected and non-protected subgroups in non-tumorous and tumor tissue identified by Ingenuity pathway analyses. (C) Gene set enrichment analysis (GSEA) for non-tumorous and tumor tissue. Normalized enrichment score (NES) reflects degree of overrepresentation for each group at the peak of the entire set. Statistical significance calculated by nominal p-value of the ES by an empirical phenotype-based permutation test. (D) Kaplan–Meyer analyses based on the specific transcriptome profiles in non-tumorous and tumor tissue using public available data from authentic human HCC of 139 patients from Lee et al. and of 395 patients from the TCGA database. (E) Gene set enrichment analysis (GSEA) for BAX-protected and non-protected tumor associated gene expression signatures during malignant transformation on prognostic HCC subgroups (panels A and B). Normalized enrichment score (NES) reflects degree of overrepresentation for each group at the peak of the entire set. Statistical significance calculated by nominal p-value of the ES by an empirical phenotype-based permutation test.(A) Unsupervised clustering analyses based on corresponding genes differently regulated between non-tumorous and tumor tissue in BAX-protected and non-protected subgroups. (B) Venn diagram demonstrating the overlap of differentially regulated genes in BAX-protected and non-protected tumor subgroups during malignant transformation. (C) Activated signaling pathways during malignant transformation examined by comparative ingenuity pathway analyses based on BAX-protected and non-protected tumor subgroup transcriptome profiles. (D) Presence of loss of heterozygosity (LOH) in percentage based on SNP array analyses for BAX-protection and non-protection in non-tumorous and tumor tissue. (E) Copy number alterations based on SNP array analyses for BAX-protected and non-protected tumor subgroup in non-tumorous and tumor tissue.(A) Cells were challenged either with mainly cytosolic BAX (S184E, top) or predominantly mitochondrial BAX (S184V, bottom). The experiments were designed to increase the apoptosis predisposition through a large mitochondrial BAX pool, testing the hypothesis that cells challenged with predominantly mitochondrial BAX select for genetic alterations associated with BAX-protected tumors. (B) Caspase 3/7 activity induced by 1 µM Daunorubicin (D), 30 µM Olaparib (O) or the combination (OD) for 24 h in HCT116 BAX/BAK DKO cells with either transient (red, trnst, T) or stable (yellow, stable, S) expression of predominantly mitochondrial BAX S184V. Data ± SEM. N = 3 and p-values according to one-way ANOVA using the Holm–Sidak method. Data adjusted to BAX S184V expression. (C) Western blot analysis of caspase substrate PARP cleavage in HCT116 BAX/BAK DKO cells with either transient or stable expression of predominantly mitochondrial BAX S184V as in panel (G) following treatment with 1 µM Daunorubicin (D), 30 µM Olaparib (O), or the combination (OD). N = 3. (D) Subcellular localization of wild type BAX transiently (T, 4 h) or stably (S) expressed in HCT116 BAX/BAK DKO cells after fractionation analyzed by Western blot. Separation of cytosol (C) and heavy membrane fraction (HM, mitochondria) was controlled using AKT1 and VDAC or cyt c, respectively. N = 3. (E) Relative wild type BAX localization (log10 scale) after transient (T) and stable (S) expression in HCT116 BAX/BAK DKO cells analyzed by fractionation depicted in D. (F) Carbonate extraction (pH 11.5) of wild type BAX transiently (T, 4 h) or stably (S) expressed in HCT116 BAX/BAK DKO cells analyzed by Western blot. Separation of supernatant (S) containing OMM-associated proteins and pellet (P) with OMM-integral proteins was controlled using cyt c (released during the procedure) and VDAC, respectively. N = 3. p-value according to t-test. (G) Relative wild type BAX distribution on the OMM between OMM-integral and OMM-associated protein pool determined by carbonate extraction (log10 scale) after transient (T) and stable (S) expression in HCT116 BAX/BAK DKO shown in F.(A) Relative BAX localizations of polyclonal cell lines (PCL, HCC9, HCC31, and HCC68) and the corresponding tumor samples (T) are displayed in log10 scale. (B) Comparison of apoptotic response of PCL (HCC9, HCC31, and HCC68) after 24 h treatment with ABT-737 (1 µM), ActinomycinD (ActinoD, 1 µM), Daunorubicin (Dauno, 1 µM), Doxorubicin (Doxorub, 5 µM), Etoposide (100 µM), Sorafenib (5 µM), Staurosporine (Stauro, 1 µM), or Umi-77 (1 µM) according to caspase 3/7 activation. (C) Relative BAX localization vs. relative BAK localization (double log10 scale) for HepG2 cells, HUH7 cells and Hep3B cells in black. For comparison area of relative localizations in 34 HCC patient samples (orange broken line) and 3 polyclonal cell lines (red) are displayed. HUH7 cells, Hep3B cells and HCC68 cells (underlined in blue) share a similar BAK localization but show the full range of different BAX localizations in the tested cell lines. (D) Caspase 3/7 activity in HUH7 cells and Hep3B cells in response to 1 µM Daunorubicin or 5 µM Doxorubicin treatment for 24 h. Data ± SEM. N = 3. Apoptotic response of both cell lines after treatment with ABT-737 (1 µM), ActinomycinD (ActinoD, 1 µM), Etoposide (100 µM), Sorafenib (5 µM), Staurosporine (Stauro, 1 µM), or Umi-77 (1 µM) in Supporting Figures S19–S21. (E) Caspase 3/7 activity induced in HUH7 (green) Hep3B (red) and HCC68 cells (black) in response to 5 µM Rapamycin. Data ± SEM. N = 3. (F) Clonogenic survival of HCC68 (red), Hep3B (orange) or HUH7 (yellow) following incubation with 0, 10, 20, 30, 40, or 50 µM Olaparib for 72 h. Data represent averages ± SEM. N ≥ 3 and p-values according to one-way ANOVA using the Holm–Sidak method. (G) Caspase 3/7 activity induced by 30 µM Olaparib (O), 1 µM Daunorubicin (D) or the combination (OD) in HUH7 cells. Data ± SEM. N = 3. (H) Changes of relative BAX localization of HUH7 cells with either Olaparib (black) or Rapamycin (gray) within 24 h. Data represent averages ± SEM. N ≥ 3 (I) Effect of Olaparib (black) or Rapamycin (gray) on relative BAX localization of Hep3B cells within 24 h. Data represent averages ± SEM. N ≥ 3.Top molecular and cellular functions identified by Ingenuity Pathway Analysis in the protection and non-protection subgroup during malignant transformation.

Med-MDPI/cancers/cancers-12-06-01438.txt

@@ -0,0 +1 @@
+The data on the prognostic significance of low volume metastases in lymph nodes (LN) are inconsistent. The aim of this study was to retrospectively analyze the outcome of a large group of patients treated with sentinel lymph node (SLN) biopsy at a single referral center. Patients with cervical cancer, stage T1a-T2b, common tumor types, negative LN on preoperative staging, treated by primary surgery between 01/2007 and 12/2016, with at least unilateral SLN detection were included. Patients with abandoned radical surgery due to intraoperative SLN positivity detected by frozen section were excluded. All SLNs were postoperatively processed by an intensive protocol for pathological ultrastaging. Altogether, 226 patients were analyzed. Positive LN were detected in 38 (17%) cases; macrometastases (MAC), micrometastases (MIC), isolated tumor cells (ITC) in 14, 16, and 8 patients. With the median follow-up of 65 months, 22 recurrences occurred. Disease-free survival (DFS) reached 90% in the whole group, 93% in LN-negative cases, 89% in cases with MAC, 69% with MIC, and 87% with ITC. The presence of MIC in SLN was associated with significantly decreased DFS and OS. Patients with MIC and MAC should be managed similarly, and SLN ultrastaging should become an integral part of the management of patients with early-stage cervical cancer.One of the main controversies in the management of cervical cancer is currently the uncertain prognostic importance of micrometastases (MIC) and isolated tumor cells (ITC). These small metastatic lesions, by definition ≤2 mm, were in the past reported extremely rarely by a standard pathological assessment of pelvic lymph nodes (LN). As the acceptance and popularity of SLN has increased, and pathological processing is much more intensive in SLN than in other pelvic LN, and about 10% to 15% of patients with early-stage tumors are detected with MIC or ITC in their SLN [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16]. Available data on the impact of MIC or ITC for prognosis are not consistent [1,14,15,16,17,18,19,20]. Since the risk of recurrence is very low in early-stage cervical cancer, any assessment of prognostic importance of MIC requires large cohorts.The controversy about the prognostic significance of MIC and ITC has major consequences for clinicians, who must decide if they should manage these cases as LN-positive, and this is even more important for pathologists. If small-size metastases showed no impact on the outcome of patients, the intention of SLN examination would be limited to the detection of macrometastases above 2 mm in size. In such cases, a protocol for SLN assessment would be much less time-consuming, less expensive, and would not include immunohistochemistry. An intraoperative one-step nucleic acid amplification method (OSNA) has been recently proposed as an alternative method to ultrastaging [21].The aim of the study was to retrospectively analyze data from a large cohort of patients treated in a tertiary gynecologic oncology center, where SLN biopsy has been used in the management of cervical cancer since 2004, and where a standardized intensive protocol for SLN pathologic ultrastaging has been applied since 2009.In total, 226 patients were included in the analysis (Table 1).The majority had squamous cell cancer (76%), non-fertility-sparing surgical procedure (88%), and stage pT1b1 (69%). Parametrial invasion was reported by pathology in nine cases; in six cases, an initial invasion was known before surgery, and three were upstaged based on the pathology report (cT1b → pT2b). Out of those six cases treated by primary surgery, two were symptomatic cases with severe bleeding from the tumor, and four patients with initial parametrial invasion diagnosed by imaging preferred surgery over primary chemoradiation. Radical parametrectomy (radical hysterectomy or radical trachelectomy) type C1 or C2 was performed in 47% and 36% of patients, respectively.Overall SLN detection rate reached 93% at least on one side of the pelvis; bilateral detection rate was achieved in 80% of all cases and was comparable in subgroups with tumors <2 cm, 2–3.9 cm, and ≥4 cm (79%, 83%, 76%) [22]. There were two patients with negative SLN and macrometastases in non-SLN from systematic pelvic lymphadenectomy; the false negativity of sentinel lymph node ultrastaging reached 1% only (Table 2). Lymph node involvement was diagnosed in 38 cases (17%), including MAC in 14, MIC in 16, and ITC in 8 cases. Adjuvant radiotherapy or chemoradiation was given to 37 cases (16%), due to LN involvement (27 cases), positive vaginal margin (2), and parametrial involvement (8). Amongst patients with positive LNs, adjuvant treatment was not received by 4/14 cases with MAC (three T1b1 cases who rejected radiotherapy; one case with early cervical progression after fertility-sparing treatment), 2/16 cases with MIC (one patient rejected radiotherapy; one case was treated in an early period when MIC was not considered an indication for adjuvant treatment in the absence of other prognostic risk factors), and 5/8 cases with ITC (all five cases treated in an earlier period when ITC had not been considered an indication for adjuvant treatment in the absence of other prognostic risk factors).With the median follow-up of 65 months, 22 recurrences occurred: eight in the pelvis only, four in distant sites only, and ten combined. Six recurrences developed in patients after fertility-sparing treatment. Three cervical (2) or pelvic (1) recurrences after abdominal radical trachelectomy were salvaged by further treatment. Three pelvic (2) or combined (1) recurrences after conization were fatal (Table 3).Patient 17, who had a 25 mm tumor and deep stromal invasion (tumor free distance TFD = 0), refused both radical trachelectomy and adjuvant radiotherapy. Amongst patients with positive LNs who did not receive adjuvant treatment (11), only one with ITC developed recurrence. Disease-free survival (DFS) with the median follow-up of 65 months reached 90% in the whole group, 93% in LN-negative patients, 89% in patients with MAC, 69% in patients with MIC, and 87% with ITC. DFS was significantly worse in cases with MAC (p = 0.037) and MIC (p = 0.001) in comparison to LN-negative cases (Figure 1). Similarly, OS was significantly worse in groups with MAC (p < 0.001) and MIC (p < 0.001) in comparison to LN-negative patients (Figure 2). Both DFS (p = 0.717) and OS (p = 0.839) were similar in patients with MAC and MIC. Parameters significant for the risk of recurrence by the univariate analysis included adenosquamous tumor type (HR = 5.08; p = 0.032), presence of LVSI (HR = 2.95; p = 0.018), number of positive LNs (HR = 1.5; p = 0.015), LN positivity (MAC or MIC) (HR = 4.03; p = 0.002), MAC in LN (HR = 3.61; p = 0.046), MIC in LN (HR = 4.62; p = 0.004), TFD binarized (cut-off value ≤3.5 mm) (HR = 9.0; p = 0.033), tumor size binarized (cut-off value >33.5 mm) (HR = 2.56; p = 0.029), and adjuvant treatment (HR = 3.46; p = 0.005) (Table 4). None of the parameters significant in univariate analysis remained significant in the multivariate model (Table 5).In a large retrospective cohort of patients from a single institution, the presence of MIC in SLN was a significant independent negative prognostic factor. Patients with MAC and MIC had significantly and similarly decreased DFS and OS in comparison to LN-negative patients.Numerous papers on SLN in cervical cancer presented data on the prevalence of SLN [11,12,13,23,24], and some suggested an association of MIC with other traditional tumor-related risk factors such as tumor size or LVSI [15,16,25]. Very few papers, however, evaluated the impact of MIC on the prognosis and the data are varying (Table 6).For the first time, the potential significance of MIC was suggested by the French group (Marchiole 2005) [15]. In a case-control study, they compared a group of 26 recurred patients with the same number of matched controls without recurrence. A hysterectomy specimen was reassessed for LVSI; LN samples were serial-sectioned and stained using cytokeratin. The relative risk of recurrence was 2.44 (p < 0.01) for MIC and 2.64 (p < 0.01) for LVSI. In a Brazilian study, all pelvic LNs from 289 patients in stages IB–IIA were reassessed, finding 11 cases with MIC (3.8%) and 37 cases with MAC (12.8%) (Fregnani 2006) [17]. The low prevalence of MIC corresponded to a very low intensity of LN pathological processing. With the median follow-up of 8.5 years, 43 recurrences (15%) occurred. The presence of MIC was a significant independent prognostic factor (HR = 3.2; 95% CI: 1.1–9.6) with five-year DFS at 89%, 80%, and 50% in patients with N0, MIC, and MAC, respectively. In 2008, a German group presented the outcome of a large group of 894 patients with IB–IIB cervical cancer. They re-examined samples from positive LN, measuring the size of metastases, using original slides without any further processing (Horn 2008) [14]. Five-year DFS was significantly lower in both groups with MAC (62%) and MIC (69%) in comparison to those with negative LN (87%). In the largest retrospective study published so far, data from 645 cases were collected from seven institutions (Cibula 2012) [1]. All patients had SLN biopsy followed by pelvic lymph node dissection, and SLNs were processed by pathological ultrastaging. Both MAC and MIC were associated with similar and significantly decreased overall survival (MAC: HR = 6.85; 95% CI: 2.59–18.05; MIC: HR = 6.86; 95% CI: 2.09–22.61). In another multi-institutional retrospective study, tissue blocks were recut and evaluated for the presence of MIC in a group of 129 patients who were LN-negative at the time of primary treatment (Stany 2015) [18]. Any immunoreactive tumor cells were classified as MIC, not distinguishing MIC and ITC. This can explain the high proportion of 26 (20%) patients with MIC detected by re-evaluation. The presence of MIC was not associated with a negative outcome. There were, however, only 11 recurrences in this group (8.5%), and patients with MIC were more likely to receive adjuvant radiotherapy than those with negative LN (39% vs. 18%). In a similar study, LN tissue was reviewed and stained by immunohistochemistry in a group of 83 LN-negative patients. The presence of MIC was the strongest independent predictor of the recurrence by multivariate analysis (OR = 11.73; 95%CI: 1.57–87.8; p = 0.017), outweighing all traditional tumor-related variables such as LVSI, stromal invasion, or tumor size (Colturado, 2016) [19]. Recently, data from the prospective French study SENTICOL were analyzed for the presence and impact of MIC and ITC (Guani 2019) [20]. All LNs from 139 patients were reprocessed, although the protocol for ultrastaging of that many hundreds of LNs is not fully described. Positive LNs were found in 25 patients (18%), including eight cases with only MIC and eight cases with only ITC. Since 14 cases with MIC or ITC were reported in the original report, it seems that two more cases were identified by an additional pathological review of non-SLNs (Bats 2012) [26]. With the median follow-up of 36 months, only 13 (9%) recurrences occurred. Surprisingly, all types of LN metastases, MAC as well as MIC or ITC, were associated with a decreased survival. There are multiple reasons which can explain discrepant results in the literature. Firstly, the risk of recurrence is low in the early stages of cervical cancer, usually around 10%, so it requires a large cohort to demonstrate any significant impact on the prognosis. Even in our study, which is, to our knowledge, the largest single institutional retrospective cohort, we reported only 22 (10%) recurrences with the medium survival of more than five years. Secondly, and more importantly, in the absence of any universal protocol for SLN ultrastaging, pathological processing is so different that such discrepancies inevitably impact the accuracy of detection of not only MIC but also small MAC [27]. Other than this paper, only two out of nine previously published studies evaluating the impact of MIC on prognosis included pathological ultrastaging of SLN (Cibula 2012, Guani 2019) [1,20]. Thirdly, the designs of the studies differed considerably. In some of them, SLN was prospectively assessed by ultrastaging, while other pelvic LNs were processed by standard H&E evaluation. In others, tissue blocks from all pelvic LNs were re-evaluated retrospectively from patients who were LN-negative at the time of the treatment. There are substantial differences in cohort sizes (49–894), in disease stages, and in the proportion of cases who received adjuvant therapy (0–33%). Notably, only two papers reported an adjuvant radiotherapy rate in cases with MIC.The prevalence of cases with LN involvement varies widely in cohorts of patients treated by primary surgery; from 7% to 20% [1,8,28,29,30,31,32,33,34,35]. This is mostly due to selection criteria for primary surgical treatment. The occurrence of 6% of MAC cases in our study is rather low, taking into account that it entailed the whole spectrum of early stages, including 24% of tumors larger than 4 cm. The low rate of LN positivity could be explained by the intraoperative triage of patients based on SLN status. Radical surgery was abandoned if LN involvement was detected intraoperatively by frozen section, and these cases were excluded from our analysis.The strengths of our study include the uniform management of patients throughout the study period in a single center, the large cohort size, and the intensive protocol of SLN ultrastaging.Only a small proportion of patients received adjuvant treatment (16%). It should be emphasized that adjuvant (chemo)radiation was administered to a similar proportion of cases with MAC and MIC, so the outcome was not impacted by a different postoperative management. Although this study is the largest single institutional cohort evaluating the impact of MIC on prognosis, the main limitation remains the small number of patients with MIC and the limited number of recurrences. Due to the excellent outcome of current early-stage cervical cancer management and the relatively low prevalence of MIC, thousands of patients would be required for a prospective study powered to address the impact on survival. Even the two ongoing prospective trials on SLN in cervical cancer patients (SENTIX, NCT02494063; SENTICOL III, NCT03386734) are not designed to bring the evidence. This applies even more to ITC. Other than the prevalence of ITC, which usually equals to half of MIC, an additional reason lies in unreliable pathological detection. SLN ultrastaging cannot be intensive enough to detect all ITCs, so their prevalence will always be only relative, reflecting the intensity of the protocol for SLN assessment [1,31,36,37].This study was approved by the Ethics Committee of the General University Hospital in Prague (project No. 1587/17 S-IV). Patients with cervical cancer stage pT1a–pT2b (squamous cell cancers, adenocarcinomas or adenosquamous cancer), without enlarged or suspicious pelvic LN on preoperative imaging, treated by primary surgery with curative intent, with at least unilateral SLN in the pelvis detected, and treated in one tertiary center between January 2007 and December 2016, were included in the study. Excluded were patients in whom radical hysterectomy or fertility-sparing treatment was abandoned due to intraoperative detection of positive SLN, patients with rare tumor types, patients in whom SLN was not detected at least on one pelvic side, and patients who received neoadjuvant chemotherapy.A combined technique with both radioactive tracer (99Tc, long protocol, application 12 h before surgery, 4 × 20 MBq) and blue dye (application at the beginning of the surgery, 2 mL not diluted or diluted in 2 mL of saline) was used for SLN detection either by laparoscopy or by laparotomy. In small tumors, both tracers were applied superficially into the cervical stroma. The syringe was kept in place for a few seconds after application to avoid retrograde leakage of the tracer. The application technique was modified in cases with large tumors, as previously described (application into the residual stroma by a spinal needle, continuous control of vaginal leak when injected into the necrotic tissue) [22]. All well-defined pelvic regions were carefully explored and searched for all blue and/or radioactive lymph nodes with hand-held gamma probe. All identified SLNs were submitted for intraoperative pathologic evaluation. Further radical surgery was abandoned if any type of metastases, gross parametrial invasion, or any distant spread was identified intraoperatively, and such patients were referred for primary chemoradiation instead.If SLNs were intraoperatively confirmed negative, full pelvic lymph node dissection was completed, except in stage T1a/LVSI-negative patients. Systematic lymphadenectomy included removal of all fatty lymphatic tissue from seven pelvic regions which are well defined by exact anatomical landmarks, i.e., regions with the most frequent occurrence of positive lymph nodes. These involve bilateral obturator fossa, external and common iliac, plus presacral regions. Parametrial lymph nodes were removed together with parametria as part of radical hysterectomy specimen [38].The type of radical parametrectomy was classified according to the Querleu–Morrow classification system [39,40], and the extent of parametrectomy was tailored to risk factors known preoperatively [41]. Adjuvant radiotherapy or chemoradiation was administered if patients had positive LN, parametrial involvement, or positive vaginal margins.Patients were followed in the center for at least five years after treatment. Survival data were controlled by matching data with the Czech National Registry of Death.Pathological tumor stage (pT), tumor type, grading, LVSI status, parametrial involvement, LN involvement (MAC, MIC, ITC), number of positive LN, largest tumor size assessed by ultrasound (US), largest tumor size assessed by pathology (P), minimal tumor-free distance (TFD) assessed by ultrasound (US), depth of stromal invasion (DSI), and tumor volume calculated by the formula for ellipsoid from pathological measurement (P) were evaluated by a univariate analysis. Tumor-free distance (TFD) was defined as the minimal uninvolved stroma between the tumor and pericervical ring (dense hyperechogenic layer on ultrasound and hypointense layer on MRI) on either side of the cervix.At the time of surgery, all submitted SLNs were cut along their longest axis, and both halves of each node were examined with frozen sectioning techniques. SLNs with a diameter of less than 3 mm were not examined by frozen section. After that, SLNs as well as all other pelvic LNs were fixed in 10% formalin, sliced at 2 mm intervals, and embedded in paraffin. SLN ultrastaging protocol consisted of two consecutive sections (4 μm thick) obtained in regular 150 μm intervals at four levels. The first section was stained with H&E, and the second section was examined immunohistochemically with an antibody against cytokeratins (AE1/AE3, 1:50 dilution; Dako, Glostrup, Denmark). The presence of MAC, MIC, and ITC was classified according to the TNM system. Macrometastasis was defined as a metastasis >2 mm in the largest diameter, MIC as a metastasis between 0.2 and 2 mm, and ITC as individual tumor cells or small clusters of cells <0.2 mm in diameter.Standard descriptive statistics were applied in the analysis; absolute and relative frequencies for categorical variables and median supplemented with the 5th–95th percentile range for continuous variables. The influence of patient characteristics on survival was analyzed using univariate and multivariate Cox proportional hazard models and described using hazard ratios (HRs) and their 95% confidence intervals. Cut-off values for continuous variables were determined by ROC analysis; the criterion was the highest value of the sum of sensitivity and specificity. Kaplan–Meier methodology was adopted for the visualization of survival data; the statistical significance of differences in survival curves among groups of patients was tested using the log rank test. Analysis was computed using SPSS 25.0.0.1 (IBM Corporation 2018).In conclusion, based on the results of our study and a critical review of the literature, there is growing evidence that the presence of MIC in SLN is associated with significant negative impact on the survival, which is similar to patients with MAC. Despite caveats in current evidence and discrepancies in available data, patients with MIC should be managed with the same criteria as patients with MAC, and SLN biopsy and its ultrastaging should be implemented into routine management. SLN ultrastaging is undoubtedly a substantially more time and cost consuming practice if compared to routine LN assessment; however, it enables the identification of an additional subgroup of around 10% of patients with MIC who would be otherwise missed. If cases with all types of metastases (MAC, MIC, ITC) are excluded, a remaining subgroup of LN negative patients has an excellent prognosis. Conceptualization and supervision, D.C.; methodology, D.C. and R.K.; investigation and data acquisition R.K., J.S., D.F., A.G., F.F., L.D., T.B., A.B., P.D., K.N.; data curation and statistical analysis; J.D., J.J. and S.S.; writing—original draft preparation, R.K.; writing—review and editing, D.C.; funding acquisition, D.C. All authors have read and agreed to the published version of the manuscript.This work was supported by grants from the Czech Research Council (No 16-31643A; NV19-03-00023), Charles University in Prague (UNCE 204065 and PROGRES Q28/LF1) and by the project “International Mobility of Researchers at Charles University” (reg. n. CZ.02.2.69/0.0/0.0/16_027/0008495), which is supported by the Operational Programme Research, Development and Education. The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.Disease-free survival according to the type of LN involvement. ITC = isolated tumor cells, MIC = micrometastasis, MAC = macrometastasis.Overall survival according to the type of LN involvement. ITC = isolated tumor cells, MIC = micrometastasis, MAC = macrometastasis.Characteristics of the group (n = 226).1 absolute and relative frequencies for categorical variables; median supplemented with 5th–95th percentile range for continuous variables; 2 assessed by ultrasound; 3 assessed by pathology; LVSI = lymphovascular space invasion; ST = simple trachelectomy; RT = radical trachelectomy; SLNB = sentinel lymph node biopsy; LN = lymph nodes; MAC = macrometastasis; MIC = micrometastasis; ITC = isolated tumor cells; RT = radiotherapy; DOC = died of other cause; DOD = died of disease.Lymph node status. Combined results of SLN ultrastaging and pelvic non-SLN examination (n = 226).Characteristics of patients with recurrence (n = 22).A = adenocarcinoma; ART = abdominal radical trachelectomy; Comb = combined recurrence (pelvic plus distant); CHRT= concomitant chemoradiotherapy; CombRT = combined radiotherapy; DOD = died of disease; DSI = depth of stromal invasion; FST = fertility sparing treatment; ITC = isolated tumor cells; LVSI = lymphovascular space invasion; MAC = macrometastasis; MIC = micrometastasis; NA = not applicable; NED = no evidence of disease; SCC = squamous cell carcinoma.Significant parameters for the risk of recurrence from univariate analysis.1 hazard ratios are computed using Cox proportional hazards model; 2 cut-off determined by ROC analysis, the criterion was the highest value of the sum of sensitivity and specificity; LVSI = lymphovascular space invasion; TFD = tumor free distance.Multivariate model for the risk of recurrence.1 hazard ratios are computed using the Cox proportional hazards model; 2 cut-off determined by ROC analysis, the criterion was the highest value of the sum of sensitivity and specificity; LVSI = lymphovascular space invasion; TFD = tumor free distance.Overview of articles reported the impact of MIC on the prognosis.→ no impact, ↑ increased, ↓ decreased, A—adenocarcinoma, Adj. Tx—adjuvant treatment, AS—adenosquamous cancer, CI—confidence interval, DFS—disease-free survival, mF/U—median follow-up, HR—hazard ratio, LN—lymph node, N0—lymph node negative, N1—lymph node positive, ITC—isolated tumor cells, m—months, MAC—macrometastases, MIC—micrometastases, No.—number, OR—odds ratio, OS—overall survival, RecR—recurrence rate, RR—relative risk, SCC—squamous cell cancer, SLNB—sentinel lymph node biopsy, UD—undifferentiated; * MAC/MIC/ITC, # MIC + ITC.