|Home | About | Journals | Submit | Contact Us | Français|
We explore herein the effect of TRANCE costimulatory pathway blockade on islet survival after allograft transplantation.
Expression of TRANCE on murine C57BL/6 (B6) CD4+ T-cells after allogeneic activation was analysed by FACS. The effect of a TRANCE receptor fusion protein (TR-Fc) and anti-CD154 antibody (MR1) on B6 spleen cell proliferation after allogeneic activation was assessed by MLR. Three groups of B6 mice were transplanted with allogeneic islets (DBA2): Control; short-term TR-Fc-treatment (days 0–4); and prolonged TR-Fc-treatment (days -1–13). Donor-specific transfusion (DST) was performed at the time of islet transplantation in one independent experiment. Transplantectomy samples were analyzed by immunohistochemistry. TRANCE expression was upregulated in stimulated CD4+ T-cells in vitro. In MLR experiments, TR-Fc and MR1 both reduced spleen cell proliferation, but less than the combination of both molecules. Short course TR-Fc treatment did not prolong islet graft survival as compared to controls (10.6±1.9 vs 10.7±1.5 days) in contrast to prolonged treatment (20.7±3.2 days; p<0.05). After DST, primary non-function (PNF) was observed in half of control mice, but never in TR-Fc-treated mice. Immunofluorescence staining for Mac-1 showed a clear decrease in macrophage recruitment in the treated groups.
TRANCE targeting may be an effective strategy for the prolongation of allogeneic islet graft survival, thanks to its inhibitory effects on costimulatory signals and macrophage recruitment.
Pancreatic islet transplantation offers a great potential for the treatment of type 1 diabetes mellitus. The protocol introduced by the group at the University of Alberta has helped improve the results of islet transplantation to a great extent, achieving 80% insulin-independence at 1-year follow-up [1, 2]. Among the issues that have surfaced are the rapid loss of islet graft function over time and the high incidence of side effects of the IS regimen. [3, 4] For this reason, alternative, less toxic IS regimens are desirable. The use of agents blocking co-stimulatory signals of T-cell activation at the time of cognate interaction between T-cell receptor and foreign antigen is a potential approach. Blockade of the CD28-B7-1 or 2 axis with the use of belatacept has recently been successfully used in a phase II trial of kidney transplantation, with a remarkably low incidence of side effects . CD154, another key molecule for T-cell activation, has been shown critical in the delivery of co-stimulation signals during allograft rejection . Anti-CD154 monoclonal antibody (MR1) used as a blocker of this co-stimulation pathway, induces operational tolerance in allogeneic [7, 8] or xenogenic  rodent models of islet transplantation. Specific blockade of this pathway also prevents acute allograft rejection in primate models of islet  and kidney  tranplantation. Unfortunately, these promising preliminary results could not reach clinical application because anti-CD154 antibody therapy was reported to cause unexpected thrombo-embolic complications in both primates and humans [10, 12, 13].
The role of TNF-related activation-induced cytokine (TRANCE) in the immune system has recently emerged [14–16]. TRANCE provides signals for the morphogenesis of lymph nodes, participates to B and T-cell development  and belongs to another T-cell co-stimulatory pathway. TRANCE mediates its effects via the TRANCE receptor (TRANCE-R), also known as receptor-activating NF-κB (RANK), expressed on antigen-presenting cells. While CD154 is only expressed on activated CD4+ T-cells, TRANCE is expressed on both activated CD4+ and CD8+ T-cells , suggesting an independent role of TRANCE on antigen-presenting cells (APCs) through CD8+ activation. In the absence of the CD40/CD154 pathway, TRANCE can independently activate CD4+T-cells, following lymphocytic choriomeningitis virus (LCMV), vesicular stomatitis virus or Leshmania infection in mice [19, 20]. An involvement of TRANCE in the allogeneic immune response has recently been suggested in a rodent model of heart allograft .
We hypothesized that blockade of the TRANCE co-stimulatory pathway might prolong the survival of allogeneic pancreatic islets. Therefore, we have first demonstrated the independent role of TRANCE in the T-cell response in an in vitro allogeneic model. Second, we have studied the effects of TRANCE blockade on graft rejection in an in vivo model of allogeneic islet transplantation.
Adult male C57BL/6 (B6) and DBA2 mice were purchased from Janvier Laboratories (Le Genest Saint Isle, France). Eight to ten week-old mice were used for in vitro experiments and 6–9 week-old mice were used for in vivo experiments. Animals were maintained in our own housing facilities with free access to food and water. All experimental protocols were reviewed and approved by the Institutional Animal Care and Use Committee and by the State of Geneva Veterinary Authorities.
After skin disinfection, spleens from B6 or DBA2 mice were removed though a large midline laparotomy. Cells were obtained by gentle mechanical disruption of the spleen followed by filtration through a 100 µm mesh. Erythrocytes were lysed using a commercial kit (Mouse Erythrocyte Lysing kit, R&D Systems, Minneapolis, MN), and mononuclear cells were purified by Ficoll-hypaque density gradient centrifugation (Histopaque-1077, Sigma, St Louis, MO).
Supra-physiological activation of murine CD4+ T-cells was induced by incubation with anti-CD3 and anti-CD28 mAb as described previously. Briefly, 400 ×105 mononuclear cells isolated from the spleens of B6 mice were transferred into 96-well round bottom plates (Nunclon Surface, DK 4000, Roskilde, Denmark) previously coated with saturating amounts of rat anti-mouse CD3 IgG (Serotec, Oxford, United Kingdom) in 200 µl Iscove modified Dulbecco’s culture medium (Gibco, Basel, Switzerland) supplemented with 10% fetal calf serum (Gibco), 1mM sodium pyruvate (Gibco), MEM Non-Essential Amino Acids (Gibco) diluted 100×, 100 units/ml penicillin, 100 µg/ml streptomycin (Sigma, St Louis, MO), 0.292 mg/ml L-Glutamin (Sigma), and 50 µM 2-Mercapto-ethanol (Sigma). 2.5 µg/ml of a hamster anti-mouse CD28 IgG (Becton Dickinson, Basel, Switzerland) antibody was added to each well. As control, equal numbers of cells were transferred into 96-wells plates untreated with anti-CD3 and anti-CD28 antibodies. After 2 days incubation cells were harvested, stained for CD4 and TRANCE, and analyzed by flow cytometry as described below. Experiments were run in triplicate and repeated 5 times.
Cells were stained using a monoclonal Phycoerythrin (PE)-conjugated rat anti-mouse CD4 IgG (1/400; Becton Dickinson, Basel, Switzerland, réf. 553048) and a monoclonal goat anti-mouse TRANCE IgG (dilution 1/1, R&D Systems, ref. AF 462) followed by Fluorescein (FITC)-conjugated donkey anti-goat F(ab’)2 IgG (1/400, AffiniPure, Jackson Laboratories, Basel, Switzerland). As negative isotype control, cells were incubated with an irrelevant goat IgG (1/1, AB108c, R&D Systems).
All antibodies were incubated with cells for 1 hour. Fluorescence of stained cells was measured on a FACSTrack fluorocytometer (Becton Dickinson) and analyzed with the WinMDI software (Scripps Research Institute, La Jolla, CA).
To assess TRANCE expression in response to allogeneic stimulation, 3 ×105 DBA2 spleen cells and 3 ×105 B6 spleen cells were co-cultured in 96-well round-bottom plates (Nunc, Wiesbaden, Germany). As controls (non activated cells), either 6 ×105 DBA2 spleen cells or 6 ×105 C57BL/6 spleen cells were cultured independently. Spleen cells were cultured in 200 µl Iscove modified Dulbecco’s culture medium supplemented with 10% fetal calf serum, 1mM sodium pyruvate, MEM Non-Essential Amino Acids diluted 100×, 100 units/ml penicillin, 100 µg/ml streptomycin, 0.292 mg/ml L-Glutamin, 50 µM 2-Mercapto-ethanol and 2.5 µg/ml of anti-CD28 IgG. For the experimental and control groups, cell harvesting and culture were performed in parallel, in quadruplicates. Plates were incubated at 37°C, in a 5% CO2 atmosphere. After 5 days, cells were harvested and washed three times with PBS (Sigma) supplemented with BSA (Sigma) 0.1% and Azide 0.1%. Experiment was repeated 5 times. To study the effect of co-stimulatory blockade on T-cell proliferation, spleen cells obtained from C57BL/6 mice were used as responder cells, and spleen cells from DBA2 as stimulator cells. Stimulator cells were irradiated with a total dose of 35 Gy using a Cesium source. Cells were cultured in a 2:1 stimulator/responder ratio, i.e. 4 × 105 DBA2 cells were cultured with 2 × 105 C57BL/6 cells. Culture was performed in quadruplicate for each condition, in 96-well round bottom plates containing 200 µl fresh Iscove modified Dulbecco’s medium supplemented as described above. Each experiment was repeated three times. Cells were incubated for 5 days at 37°C in 5% CO2. Inhibition of TRANCE was obtained by adding a recombinant protein of the extracellular domain of TRANCE-R fused to the constant region of human IgG1 (TR-Fc; Dr. Y. Choi, University of Pennsylvania, Philadelphia). Production of TR-Fc has been described previously . Inhibition of CD154 was performed using a hamster anti-mouse CD40L antibody, MR-1 (Bio Express, Inc, West Lebanon, NH). After 4 days incubation, 1mCi 3H-thymidine (Amersham, Otelfingen, Switzerland) was added and cells were further incubated for 6 hours. Plates were harvested and cell proliferation was assessed by measuring 3H incorporation using a beta counter (Wallac, Perkin-Elmer; Geneva, Switzerland).
C57BL/6 mice were rendered diabetic by a single intraperitoneal injection of 200 mg/kg streptozotocin (Sigma, Buchs, Switzerland). Blood glucose levels were monitored on whole-blood samples collected from the tail vein using a strip glucometer (Precision Q.I.D; Abbott MediSense, Wiesbaden, Germany). Mice with blood glucose > 20 mmol/L for more than 3 consecutive days were used as transplant recipients.
Murine islets were isolated using a technique routinely used in our laboratory. Briefly, a collagenase (type XI, Sigma) solution was prepared at a final concentration of 2 mg/ml in Hank’s balanced saline solution (HBSS). After anesthesia by intraperitoneal ketamine injection (12 mg / 100g of animal body weight), mice were euthanized, the abdomen was opened, and the pancreas was exposed. After clamping at the porta hepatis, the main bile duct was cannulated with a 10×0.4 mm needle (Venisystem, butterfly-27G) and injected in retrograde fashion with 2 ml of collagenase solution in order to distend the pancreas. The pancreas was then removed and placed in 7.5 ml ice-cold HBSS in 50-ml conical tubes (2 pancreata per tube). Digestion was started by raising the temperature to 37°C. After 17 minutes, tubes containing the pancreata were put on ice and cold HBSS was added up to 50 ml in order to stop the digestion. Mechanical disruption of the digested pancreatic tissue was achieved by vigorous tube shaking. Islets were then purified on a discontinuous Ficoll gradient, washed twice with HBSS solution and kept on ice until transplantation.
Allogeneic murine islet transplantation was performed in the fully mismatched DBA2 (H2d) to B6 (H2b) combination. Prior to transplantation, islets were divided in aliquots of 750 islets. Islet viability, assessed by fluoresceine diacetate/propidium iodide staining was always > 90%. Under general anaesthesia by isoflurane (Forene®, Abbott AG, 6340 Baar) inhalation using a custom-made breathing apparatus, the left kidney was exposed. Islets were slowly injected under the kidney capsule, as described previously . Islet transplantation was considered successful when blood glucose levels < 11 mmol/L were measured on at least 2 consecutive days, and rejection was defined as sustained hyperglycaemia > 15 mmol/L on at least 3 consecutive days. Primary non function was defined as sustained hyperglycaemia after transplantation in the absence of technical problem.
Animals were kept up to 100 days if no islet rejection occurred, or 3 to 10 days after recurrence of diabetes. Animals were sacrificed and graft-bearing kidneys were retrieved. Kidneys were embedded in Tissue-Tek (Miles, Elkhart, IN), snap-frozen in liquid methylbutane pre-equilibrated with liquid nitrogen and stored at −80°C.
In a first set of experiments, the treated group (n=8) received 0.2 mg TR-Fc intra-peritoneally on days 0, 2 and 4, day 0 being the day of islet transplantation (Short-term treatment). Controls (n=7) were injected intraperitoneally with vehicle. In a second set of experiments, the treated group (n=4) received 0.2 mg of TR-Fc intra-peritoneally on days -1, 1, 3, 5, 7, 9, 11, 13 (Long-term treatment). In a third set of experiments, 5×106 donor-specific spleen cells were injected intravenously into the tail vein 1 day before transplantation and TR-Fc injections (treated group, n=5) or vehicle (control group, n=9) were administered at the same dose and on the same days as in the short-term treatment group.
Serial frozen sections were cut at 5 µm intervals using a cryostat (Leica, Glattbrugg, Switzerland). Sections were stained with hematoxilin and eosin (HE). For immunofluorescence staining, tissue sections were fixed with absolute ethanol for 1 min and then incubated with phosphate-buffered saline supplemented with 0.5% bovine serum albumin (Sigma, St. Louis, MO) for 15 min to block non-specific binding. Insulin staining was performed using a guinea pig anti insulin primary antibody (Ab) (1/100, DAKO, Glostrup, Denmark) and a rhodamine-conjugated goat anti guinea pig secondary Ab (1/400, Jackson Laboratories, Basel, Switzerland). T lymphocyte staining was performed using a rat anti mouse CD4 Ab (1/20, Becton Dickinson, Basel, Switzerland) and rat anti mouse CD8 Ab (1/100, Serotec, kidingtown, UK). Macrophage staining was performed with a rat anti mouse Mac-1 Ab (1/10, Imunokontact, Switzerland). For all T cell and macrophage stainings, FITC-conjugated donkey anti rat was used as secondary Ab. Slides were examined under a fluorescence microscope (Zeiss Axiophot, Göttingen, Germany). Sections were analyzed by 2 independent observers (AW, MA), and intensity of staining was classified semi-quantitatively as absent (0), borderline (+/−), mild (+), moderate (++), or strong (+++).
The Statistica 6.0 software package (Statsoft, Tulsa, OK) was used for statistical analysis. Data were expressed as median and range or mean ± SD, wherever appropriate. Comparison of continuous variables was done using bilateral Student’s t-test. Kaplan Meier analysis was performed for diabetes free survival determination, and differences were assessed with the Mantel Cox log-rank test. Values of p<0.05 were considered as significant.
To test whether TRANCE/TRANCE-R interaction plays a role in T-cell activation, surface expression of TRANCE was analyzed on activated CD4+ T-cells after stimulation with both anti-CD3 and anti-CD28 monoclonal antibodies (Figure 1, A, B, C). After CD4/TRANCE double staining of murine B6 spleen cells, CD4+T-cells showed a clear increase in TRANCE expression as compared to control conditions in which spleen cells were incubated in non-treated wells with culture medium only. No increase in TRANCE expression was detected after stimulation by anti-CD3 or anti-CD28 alone.
Five-day mixed lymphocyte culture of spleen cells isolated from B6 and DBA2 mice led to an increased expression of TRANCE by B6 CD4+ T-cells, as detected by a slight but sizeable shift to the right of the fluorescence peak on FACS analysis (Figure 1, A, F). TRANCE expression by CD4+ T-cells was not increased in control B6 or DBA/2 spleen cells cultured independently (Figure 1, panels D and E).
In mixed lymphocyte culture experiments of spleen cells isolated from B6 and DBA/2 mice, cell proliferation was largely increased in control conditions as a result of allogeneic stimulation by DBA2 irradiated cells on proliferation of B6 stimulator cells. This response was inhibited individually by blockade of the TRANCE , with a trend toward significance (p=0.19) and CD154 pathways (p<0.02), using TR-Fc and anti-CD154 mAb (MR-1) respectively. A cumulative effect on the inhibition of cell proliferation was observed with simultaneous blockade of both pathways (p<0.01; Figure 2).
Based on these results, the effects of systemic TR-Fc treatment in an in vivo allogeneic model of islet allograft transplantation were tested. A full mass islet graft (750 IEQ) restored normoglycaemia in all recipient mice within 2 days. In the short-term treatment group, mice did not exhibit a significant prolongation of graft survival (n=8; 10.6±1.9 days) compared to controls (n=6; 10.7±1.5 days, p>0.1; Figure 3). A significant prolongation of graft survival was observed in the long term-treatment group (n=4; 20.7±3.2 days, p<0.05; Figure 3). Primary non function did not occur in the control or treated groups.
One mouse in each group did not reject its allograft (graft survival > 100 days). In these mice, blood sugar levels increased after nephrectomy showing that the observed normoglycaemia was only due to a functional islet graft.
Since it had been shown that TRANCE blockade could prolong murine heart graft survival only when combined with donor-specific spleen cell transfusion , we then used the same strategy in our model of islet transplantation. A marked increase in the rate of PNF occurred in the control group (n=5, 56% of transplanted mice) compared to the treated group (n=0). Median islet graft survival was 13.8 ± 5.0 days in the treated group (n= 5) vs 10.3 ± 2.6 days in the control group, when excluding grafts that underwent PNF from analysis (n=4; p= 0.31). When islet grafts that underwent PNF were considered in the control group, survival was 5.9 ± 4.5 days (n=9, p= 0.026; Figure 4). No long term survival was observed.
Histological examination of islet allografts was performed 3 to10 days after rejection in all groups of recipients. Islet grafts were identified by HE and insulin staining. Inflammatory cells were stained for Mac-1, CD4 and CD8..
In controls (Figure 5 A, E, I, M), classic signs of rejection were evident, with infiltration by T-cells, identified by strong (+++) CD4 or CD8 staining, and macrophages, identified by strong (+++) Mac-1 staining. Disruption of islet architecture was observed, as well as a decrease of insulin staining (mild, +).
In the group of mice that received DST but no TR-Fc treatment (Fig 5: B, F, J, N), intense inflammation and near-complete islet destruction were seen. Only rare islets stained positive for insulin (+/−) and a marked macrophage infiltration (Mac-1; +++) with mild (+) CD4+ and CD8+ T-cell infiltration was observed. Only small remnants of insulin-containing tissue permitted islet identification.
In TR-Fc-treated mice with DST (Fig 5 C, G, K, O), inflammatory cell infiltration was observed on HE staining and, despite rejection, macrophage infiltration was borderline (+/−). CD4+ T-cell infiltration was borderline (+/−) and CD8+ T-cell infiltration was moderate (++). Insulin staining was moderate (++), with preservation of islet structure.
In long-term TR-Fc-treated mice (no DST; Fig 5 D, H, L, P), macrophage infiltration was borderline (+/−). CD4+ T-cell infiltration was absent (0) and CD8+ T-cell infiltration was moderate (++). Insulin staining was strong (+++) with preserved islet architecture.
In this study, we have shown in an allogeneic model of islet transplantation that blockade of the TRANCE pathway can prevent islet graft primary nonfunction and delay graft rejection. We first showed that TRANCE expression is upregulated during T-cell activation, and that TRANCE blockade impairs T-cell proliferation. Second, we demonstrated in vivo that TRANCE blockade delays acute islet allograft rejection in a murine model, while decreasing macrophage infiltration in the graft.
The role of TRANCE/TRANCE-R interaction as a co-stimulatory pathway of T-cell activation was first suggested by Josien et al . We showed here that the expression of TRANCE at the membrane surface of T-cells was highly upregulated upon antibody stimulation of the CD3 and CD28 T-cell surface molecules. Moreover, we have pointed out the role of allogeneic stimulation in the increase of TRANCE expression on CD4+T-cells in a mixed lymphocyte reaction (MLR) experiment using DBA2 and B6 spleen-cells. In the same MLR combination, TRANCE blockade by a recombinant protein of the extracellular domain of TRANCE-receptor fused to the constant region of human IgG1 (TR-Fc) led to a decrease in T-cell proliferation. Blockade of the CD154 pathway led to a decrease of T-cell proliferation to a similar extent, and blockade of both pathways further decreased T-cell proliferation to levels close to background. This suggests that TRANCE and CD154 engagement have at least an independent additive effect on T-cell activation secondary to an allogeneic stimulus. Taken together, these results suggest that TRANCE and CD154 may share similar in vivo co-stimulatory functions during T-cell activation in response to an allogeneic stimulus. TRANCE may also be responsible for CD154-independent CD4+ T-cell alloimmune responses, as observed in murine models of viral infections [19, 20].
To determine whether the role of TRANCE as a co-stimulatory role is significant enough to prevent or delay graft rejection, we blocked the TRANCE pathway by in vivo treatment with TR-Fc in a murine model of allogeneic islet transplantation. Short-term TR-Fc treatment had no effect on the timing and occurrence of rejection. By contrast, long-term TR-Fc treatment delayed graft rejection and doubled graft survival time. Long-term graft survival was not achieved. Interestingly, graft rejection occurred in both treatment groups despite low graft infiltration by T-cells, as shown by weak CD4 staining. These experiments show that, in contrast to CD154 blockade, TRANCE blockade alone is not sufficient for long-term prevention of rejection. We did not attempt to combine TRANCE and CD154 blockade in vivo, because in our hands, as in others’, CD154 blockade alone is able to yield indefinite graft survival in the murine islet allogeneic graft model .
A previous report by Guilloneau et al (21) showed that short-term TRANCE blockade was only effective in prolonging heart transplant survival when combined with donor specific transfusion  administered on the day of transplantation. Although no clear explanation was offered for this observation, we used the same approach in our model. Surprisingly, DST led to a high incidence of primary non function (PNF) in the control group. At variance with Guillonneau’s study (21), TRANCE blockade did not delay graft rejection in DST-treated animals. However, it completely prevented the occurrence of PNF. Differences in graft survival can partly be explained by the different nature of heart and islet transplantation, i.e. between a vascularized and a non-vascularized graft. TRANCE has been described as a potent angiogenetic factor (24) and its blockade may have affected neovessel formation around transplanted islets which is critical for their engraftment and survival. It should also be mentioned that in the Guilloneau study (21) DST was given on the same day as heart transplantation rather than 1 day before, and in a different mouse strain combination, both elements that could account for the differences observed.
The role of DST per se in islet transplantation must also be discussed. In our hands, DST administered on the day of transplant caused a high rate of islet PNF. A similar phenomenon was reported in the Balb/c to NOD combination, even when DST administered 5–8 days before islet transplantation was combined with CD154 costimulatory blockade (25). In contrast, the same group reported a beneficial effect of DST, which was able to indefinitely prolong islet graft survival in the DBA/2 to B6 combination, when combined with CD154 costimulatory blockade (26). Of note, DST was administered 28 days prior to islet transplantation in the latter study. These somewhat conflicting sets of data make it difficult to interpret the role of DST in allogeneic murine islet transplantation. However, it seems likely that there exists in mice a delicate balance between the tolerizing and sensitizing effects of DST, that is dependent on timing of administration and genetic backgrounds of donor and recipient.
It is well known that macrophage activation is one of the key determinants of islet of Langerhans PNF (27). In our study, DST caused a massive macrophage infiltration in the islet grafts of control mice, whereas a weak infiltration was observed in TR-Fc-treated mice. TRANCE-R (RANK) is expressed on the membrane surface of antigen-presenting cells, notably dendritic cells (DC), which belong to the same cell lineage as macrophages. TRANCE-RANK engagement not only results in costimulatory signals to T-cells, but also in survival, maturation and secretory signals to DCs (28), notably leading to the secretion of proinflammatory cytokines such as IL-1 and IL-6 (18). It is therefore likely that macrophage maturation and recruitment are inhibited by TRANCE blockade, as is the case of dendritic cells. This lack of recruitment could well explain the weak macrophage infiltration observed in islet grafts in TR-Fc-treated mice, which is even more obvious in DST-treated animals, and thus the protection against PNF conferred by TRANCE-RANK engagement blockade.
In accordance with the increasing evidence about the role of innate immunity in allograft rejection, the effect of TRANCE blockade on macrophage function at the site of engraftment could also be involved in the delay of rejection observed in long-term TR-Fc treated mice. Taken together, our data indicate that TRANCE expression is upregulated on T-cells after allogeneic stimulation and that TRANCE blockade by TR-Fc in vitro can depress T-cell proliferation in a CD154-independent pathway. This effect is not sufficient to confer operational tolerance to islet grafts in vivo, but can prolong islet graft survival if TR-Fc is administered over a prolonged period of time. This important difference with the effects of CD154 blockade can be explained by TRANCE being less critical for T-cell priming than CD154, but also by the blockade of the angiogeneic properties of TRANCE in the context of islet grafts that must revascularize in order to ensure engraftment. On the other hand, the tampering of macrophage recruitment and activity by TRANCE blockade may be of high interest in the promotion of islet engraftment.
Funding: This work was supported in part by grant # 3200BO-116562 from the Swiss National Research Foundation (To PM, DB and TB)
AW and AA contributed equally to the manuscript.
TB, PM, and DB: designed study
AW, AA, SC, MA, and CT: performed research
AW, AA, SC and MA: collected data
AW, AA and DB: analyzed data
YC: contributed major reagents
AW, AA, and TB: wrote the paper
CT, YC, DB and PM: reviewed the paper
Conflict of interest: None of the authors have conflicts of interest to disclose.