IDDM is characterized by leukocyte invasion to the pancreatic tissues followed by immune destruction of the islets. Despite the important function of Th17 cells in other autoimmune disease models, their function in IDDM is relatively unclear. In this study, we found association of elevated Th17 cytokine expression with diabetes in NOD mice. To understand the function of Th17 cells in IDDM, we differentiated islet-reactive BDC2.5 TcR transgenic CD4+ cells in vitro into Th17 cells and transferred them into NOD.scid and neonate NOD mice. NOD.scid recipient mice developed rapid onset of diabetes with extensive insulitic lesions, whereas in newborn NOD mice, despite extensive insulitis, most recipient mice did not develop diabetes. Surprisingly, BDC2.5+ cells recovered from diabetic NOD.scid mice, in comparison with those from neonate NOD mice, showed predominant IFN-γ over IL-17 expression, indicating conversion of donor cells into Th1 cells. Moreover, diabetes progression in NOD.scid recipients was dependent on IFN-γ while anti-IL-17 treatment reduced insulitic inflammation. These results indicate that islet-reactive Th17 cells promote pancreatic inflammation, but only induce IDDM upon conversion into IFN-γ producers.
CD4+ Th cells play essential pathogenic function in autoimmune diseases. After activation, antigen-specific Th cells differentiate into cytokine-secreting effector cells that have been historically classified into Th1 or Th2 cells 1. Th1 cells make IFN-γ whereas Th2 cells produce IL-4, -5 and -13. Although Th1 cells were associated with diabetes in NOD mice, NOD mice lacking IFN-γ 2, IFN-γ receptor 3 or IL-12 4 developed T1D similarly to wild-type NOD mice. However, islet-reactive Th1 cells generated from BDC2.5 TcR transgenic T cells were reported to drive aggressive diabetes 5, possibly via IFN-γ induction of apoptosis of insulin-producing β cells 6. Moreover, transfer of Th1 but not Th2 cells into neonatal NOD mice caused T1D 5 and BDC2.5 mice lacking IFN-γ receptor are resistant to cyclophosphamide-induced diabetes 6.
Recent studies have identified a new subset of Th cells, called Th17, which produce IL-17, IL-17F, IL-22 and IL-21 and mediate tissue inflammation 7, 8. There is growing evidence that Th17 cells are pathogenic in several autoimmune disease mouse models such as EAE and collage-induced arthritis 9–12. However, there is little information on Th17 or IL-17 in type 1 diabetes mellitus. Recently, Jain et al.13 reported that treatment with a fusion protein consisting of IgG and GAD peptide 206–220 confers diabetes protection to hyperglycemic NOD mice, correlating with a reduced number of IL-17-producing cells present in the spleen and induction of IFN-γ-producing cells.
To address the involvement of Th17 cells in T1D, we assessed IL-17 and IL-17F expression in NOD mice pancreas at different stages of T1D development and found that both cytokines were increased in expression in diabetic mice. In order to evaluate the function of Th17 cells, we differentiated islet-reactive BDC2.5 transgenic CD4+ cells into Th17 cells and transferred them into NOD.scid and newborn NOD mice. To our surprise, NOD.scid recipient mice developed full-blown diseases but newborn NOD recipients were mostly resistant. Although BDC2.5 Th17 cells consistently caused massive islet infiltration in both types of recipients, donor cells in NOD.scid mice predominantly expressed IFN-γ but not IL-17. A blocking antibody against IFN-γ inhibited diabetes in NOD.scid mice, while anti-IL-17 only reduced insulitic inflammation. Therefore, islet-reactive Th17 cells primarily function by promoting inflammation but their conversion to Th1 cells in lymphopenic hosts results in diabetes.
Expression of IL-17 and IL-17F during diabetes progression in NOD mice
To determine whether Th17 cells play a role in diabetes development, we first assessed the expression of two characteristic cytokines produced by Th17 cells, IL-17 and IL-17F, in the pancreas of NOD mice. RNA was extracted from pancreas of 2-wk-old, 11-wk-old (non-diabetic mice), and recently detected diabetic NOD mice (16–25-wk-old), followed by real-time RT-PCR for IL-17 and IL-17F. We found that both genes were increased in mRNA expression in older mice and there was a significant expression of IL-17 and IL-17F in the pancreas of diabetic NOD mice (Fig. 1A). Thus, IL-17 and IL-17F expression correlates with established insulitis and diabetes. These results prompted us to search for IL-17-producing cells in NOD mice. To detect IL-17 and IFN-γ by intracellular staining (ICS), LN and spleen cells were activated with PMA and ionomycin for 5 h. Since IL-17 can be produced by other cell types, such as γδ T cells and macrophages 14, we analyzed the IL-17+ cells in the TCR-β+ and CD4+ population. IL-17+ cells were found in small quantities in young adults, pre-diabetic (16–18-wk-old) and diabetic NOD (16–23-wk-old) mice but further increased by 2–3 times in the later especially in MLN, pancreatic LN (PLN) and spleen (Fig. 1B and C). In contrast, IFN-γ-expressing cells were found predominantly in PLN of pre-diabetic and diabetic NOD mice (Fig. 1B and C). The increase of IL-17-producing cells was not observed in control C57BL/6 mice (Fig. 1C). Our results indicate that increased Th17 cells were associated with pathogenesis of T1D in NOD mice and suggest their functional involvement in this autoimmune disorder.
Polarized BDC2.5 Th17 cells cause diabetes in NOD.scid mice but only insulitis in newborn NOD mice
To analyze the capacity of Th17 cells to cause type 1 diabetes mellitus, we differentiated CD4+ T cells from BDC2.5 TcR transgenic mice into Th17 cells. CD4+ cells were cultured with the mimotope 63 (RTRPLWVRME) 15 under Th17 differentiation conditions (IL-6, TGF-β, IL-23, anti-IFN-γ and anti-IL-4) and irradiated antigen-presenting cells 16. On day 3, a sample of the cells was restimulated with PMA and ionomycin or mimotope 63 and IL-17 and IFN-γ production was evaluated by ICS. Under these conditions, we would typically obtain 10–30% of IL-17-producing cells, 50% of IL-17F-producing cells and between 0.5 to 5% of IFN-γ-producing cells (Fig. 2A and data not shown). These cultured cells had 90–95% viability and did not require the addition of new media or more IL-23. On day 4, 1 million cells of this preparation were transferred into NOD.scid mice as reported previously 5, 17 and diabetes was monitored daily. One hundred percent of incidence of diabetes was observed between day 6 and day 20 after transfer of BDC2.5 Th17 cells (Fig. 2B and C). Disease incidence was dependent on the amount of BDC2.5 cells transferred and 0.5 million cells were the minimum amount of cells necessary to cause 100% disease (Fig. 2C).
When we examined the islet lesions generated by Th17 cells by histology, we found that the leukocyte infiltrate was present in the islets but also covering extensive areas of the exocrine tissue (Fig. 2D and Supporting Information Fig. 1), which are different from the lesion infiltrates reported for BDC2.5 or BDC2.5 Th1 cells 5, 18.
We then asked whether Th17-polarized BDC2.5 cells could cause disease in neonatal NOD mice with a full T-cell repertoire with Treg cells, in which there is no influence of homeostatic expansion. One million of BDC2.5 Th17 cells were transferred into 7-day-old NOD mice; similar Th1 cell dose has been reported to cause diabetes (31). To our surprise, only 20% of the recipient mice developed diabetes within 25 days (Fig. 2E). However, when examined histologically, all the mice were found to have developed insulitis (ranging from 20 to 70%) and perinsulitis (45% average) of the islets by day 25 (Fig. 2F). Interestingly, the islet infiltrates generated by transferred cells also cover the exocrine tissue (Fig. 2G). These results suggest that the Th17 cells can cause insulitis but are arrested in the islet environment in neonatal NOD mice.
To confirm that the transferred BDC2.5 Th17 cells were active and the cause of insulitis in the neonatal NOD recipient mice, we repeated the transfer experiment but this time, NOD recipient mice were sacrificed on day 6 after cell transfer. IL-17 was quantified in the serum of the NOD recipients and pancreata were evaluated for insulitis. We found that those animals with insulitis and perinsulitis at 6 days after transfer had high levels of IL-17 in the serum, when compared with control littermates that received PBS (Fig. 2H).
To further ascertain that the transferred BDC2.5 Th17 cells migrate into pancreas, immunohistochemistry for IL-17F, CD4 and macrophages (F4/80 staining) were performed on frozen sections of pancreas of recipient NOD mice. Interestingly, IL-17F was detected only in the islets of NOD mice that developed diabetes (Supporting Information Fig. 2). This population of IL-17F-producing cells were localized adjacent to macrophages and were in the perimeter of the islet structures (Supporting Information Fig. 2, IL-17F). Thus, Th17 cells participate in the insulitic lesion and may be involved in the attraction and activation of macrophages in the islet. Another interesting observation of Th17 lesion of neonatal NOD mice was the presence of macrophages localized in the perimeter of the islets but never in the central area of the islet (Supporting Information Fig. 2). Our results strongly suggest that Th17 cells might recruit and activate macrophages to the islets to promote tissue inflammation and active insulitis.
IFN-γ is necessary for Th17-induced diabetes in NOD.scid mice
Since BDC2.5 Th17 cells caused significant diabetes in NOD.scid but not neonatal NOD mice, we further characterized the donor cells in different hosts. When BDC2.5 T cells recovered from the diabetic NOD.scid mice were analyzed for cytokine production, we found 8% of IL-17-producing cells in almost every lymphoid organ; however, higher numbers of IFN-γ-producing BDC2.5 T cells were recovered, especially in PLN (Fig. 3A). Moreover, the predominant cytokine in all the mice and organs analyzed was IFN-γ (Fig. 3A, graph). Thus, although BDC2.5 T cells differentiated into Th17 were transferred, some cells continued to produce IL-17 but a majority of cells produced IFN-γ (Fig. 3A).
In contrasts, when neonatal NOD recipient mice were analyzed, the number of IL-17- and IFN-γ-producing cells in all the secondary lymphoid organs were less than 3% and there was no dominance of IFN-γ production in the PLN (Fig. 3B).
Since we found IFN-γ-producing cells in LN of NOD.scid mice that received Th17 cells, we then asked whether IL-17 or IFN-γ was more important to provoke diabetes. Blocking antibodies against IFN-γ or IL-17 or control rat IgG were administered to NOD.scid mice together with BDC2.5-Th17-differentiated cells, and continued the antibody administration every 4 days. We found that diabetes was significantly reduced in NOD.scid mice that received anti-IFN-γ when compared with the group-receiving rat IgG (Fig. 4A). However, mice treated with anti-IL-17 blocking antibody developed similar diabetes as the group of mice treated with rat IgG (Fig. 4A). This result indicates that transferred cells required IFN-γ to cause diabetes.
Evaluation of the islet lesions from each group revealed that blockade of IL-17 reduced the extensive leukocyte infiltration observed in mice treated with rat IgG (Fig. 4B, D and E) but not the total insulitis. The massive infiltration was only observed in 30% of the islets of mice treated with anti-IL-17 blocking antibody (Fig. 4E), in comparison to 70% observed in the group treated with rat IgG (Fig. 4E). As expected, insulitis in the mice treated with anti-IFN-γ blocking antibodies was limited to the islets and in the majority of these mice about 10–20% of islets were free of insulitis (Fig. 4C and E). This let us to conclude that IL-17 produced by donor BDC2.5 cells does not directly participate in the destruction of islet- cells but is required to promote extensive inflammation in the islets. More importantly, IFN-γ was indispensable to cause islet destruction.
To compare Th17 cells with Th1 cells in their capacity to provoke disease, we analyzed NOD.scid mice that received BDC2.5 cells that were differentiated into Th1 cells. These mice developed diabetes as has been reported 5 with a similar onset as the mice that received polarized Th17 cells. Furthermore, the disease was prevented by blocking antibodies against IFN-γ, confirming that IFN-γ produced by Th1 cells participates in the destruction of the islet-β cells (Fig. 4C, F and G). Some mice treated with anti-IFN-γ did not show full insulitis and in 20% of the islets of these mice there were no leukocyte infiltrates. Although both Th17 and Th1 cells can cause disease, the leukocyte infiltrates and the inflammatory process surrounding the islet appears different. Th1 cells promoted an islet-restricted lesion whereas the Th17 cells recruited a massive infiltrate that extended to the exocrine pancreas (Fig. 4B and G and Supporting Information Fig. 1).
These results suggest a possible conversion of BDC2.5 T cells from Th17 to Th1 phenotype in the NOD.scid mice. To explore this possibility, we differentiated BDC2.5 T cells under Th17 conditions as in Fig.1 and on day 3 the cells were stimulated with peptide 63 for 3 h followed by staining for IFN-γ producers using a bi-specific capture system (Miltenyi Biotec), which was revealed with a fluorescently labeled antibody against IFN-γ (Fig. 5A). CD4+ cells negative for IFN-γ were FACS-sorted and transferred into NOD.scid mice. A sample of the sorted cells was cultured for 5 h with peptide 63 and irradiated APC and for analysis of IFN-γ, IL-17 and IL-17F expression. We observed that the purified cells did not produce IFN-γ but continued to produce IL-17 and IL-17F (Fig. 5A). NOD.scid recipient mice developed diabetes on day 8 after the transfer (Fig. 5B) and on day 10 LN and spleen cells from these mice were stimulated with peptide 63 and evaluated for IL-17, IL-17F and IFN-γ production. PLN and spleen cells predominantly produced IFN-γ and few cells expressed IL-17 (Fig. 5C and D). Interestingly, there was a 12.5% of IL-17F+ IL-17− cells in the spleen and only 2.3% in the PLN (Fig. 5C and D). Nonetheless, the predominant population was IFN-γ-producing cells in all the animals analyzed. Therefore, these results indicate that fully differentiated Th17 cells can be converted to IFN-γ-producing cells in NOD.scid recipients to produce diabetes.
Despite their prominent function of CD4–Th17 cells in mouse models for rheumatoid arthritis and multiple sclerosis, the role of Th17 cells in IDDM has been unclear. In this study, we found that IL-17 and IL-17F gene expression in pancreas of NOD mice correlated with diabetes. We show that BDC2.5 T cells that were polarized to Th17 cells in vitro induced T1D in lymphopenic NOD.scid mice and insulitis in neonatal NOD mice.
The onset and incidence Th17-induced disease in NOD.scid mice was similar to Th1-mediated diabetes reported before 5. However, Th17-caused islet lesions were characterized by severe insulitis with destruction of exocrine tissues in contrast to the Th1 lesions that were localized in the islets. In newborn NOD recipient mice that developed diabetes, we found IL-17F+ CD4+ cells in the periphery of the islet where macrophages were also present. Our results strongly suggest that Th17 cells might recruit or activate macrophages to the islets to promote an uncontrolled inflammation that leads to profuse leukocyte infiltration inside the islet and very distinctively over the islet borders to the exocrine pancreas, thus facilitating diabetes development. It is interesting that we found detectable levels of IL-17 in the sera of NOD neonates that received Th17 cells and those mice were the ones establishing fastest insulitis. Our results also suggest that BDC2.5 Th17 cells in NOD neonates are a component of the disease onset and may complement Th1 cells, since the transfer of 1 million or less Th1 cells to NOD neonates 5, 17 or ConA activated BDC2.5 cells can induce full diabetes 19. Since IL-17 and IL-17F seem to be elevated in mice with diseases, it would be of interest to explore levels of these cytokines in serum as an additional marker for T1D onset in humans.
Blockade of IL-17 does not prevent diabetes onset; however, we cannot discard the role of IL-17F and IL-22 in the disease process, both of which are produced by Th17 cells. An unexpected finding was that IFN-γ was indispensable for the Th17-induced diabetes process, and as described before, IFN-γ mediates cell destruction 6. After diabetes onset, we found an increased number of BDC2.5 T cells producing IFN-γ in the PLN and MLN, which suggest a possible conversion from Th17 to Th1 phenotype of the diabetogenic T cells in the NOD.scid mice. Depletion of IFN-γ producers before the transfer did not attenuate IFN-γ expression in donor cells, thus supporting the idea of Th17 to Th1 conversion. At this stage, we cannot discard the possibility that Th17 cells activate macrophages and dendritic cells in the islets that can further sustain Th1 differentiation or IFN-γ production by macrophages and NK T cells that will promote islet-β-cell death in situ2, 6, 20. Our data differ from a recent report by Jain that finds that blockade of IL-17 can prevent BDC2.5-Th17-mediated disease in NOD.scid mice when transferred 10 million Th17-polarized cells. However, Jain's report does not analyze the blockade of IFN-γ in BDC2.5-Th17-mediated disease, or the BDC2.5 Th17 present in diabetic NOD.scid mice or when blockade of IL-17. Moreover, the focus of Jain's report is the vaccination with a fusion protein GAD 206–220 Ig where protection is mediated by IFN-γ-producing cells and reduction of IL-17 cells in the spleen. The protective role of IFN-γ-producing cells upon vaccination of NOD mice with BCG or CFA has previously been reported by Serreze et al.21 and it is not opposing to the pathogenic effect of IFN-γ in the initiation of disease as reported in an acute induction of diabetes by cyclophosphamide treatment 22. Our report, however, does agree with Jain's report on the pathogenic role of Th17 in NOD.scid mice and the fact that only the diabetic NOD mice that received Th17 have detectable levels of IL-17F in the pancreas and IL-17 in serum further confirms the role of Th17 during the destructive phase of diabetes.
One question that arises from our study is why Th17 cells are unable to cause disease in the NOD mice but can work in the lymphopenic model? It is possible that the Th17 cells are more susceptible to Treg suppression than Th1 cells in the islets and may not survive or stay long enough in the islet to promote diabetes. Another possibility is that in the newborn NOD mice Th17 cells cannot convert into Th1 cells to cause disease. In contrast, in the lymphopenic environment Th17 are not influenced by Treg and can benefit from the rich cytokine environment that promotes homeostatic expansion. Katz and collaborators reported similar results using Th2-differentiated BDC2.5 cells to promote diabetes 17. Therefore, in a lymphopenic condition, Th2 or Th17 cells could worsen islet-β-cell destruction. On the other hand, since transferred Th17 BDC2.5 cells were more readily converted into a Th1-dominant phenotype in lymphopenic hosts, it is likely that these expanded or reprogrammed Th1 cells may mediate the diabetes development, a hypothesis supported by our results that the disease was ameliorated by anti-IFN-γ. It would be interesting to investigate whether Th2 cells could further convert to Th1 in this same model in a similar fashion, which could indicate a dominant conversion to Th1 in lymphopenic conditions. Nonetheless, to our knowledge, this is the first time that conversion of Th17 cells into Th1 in lymphopenic mice is being reported. A recent report, on the use of polarized Th17 cells for therapy against solid melanoma showed that rejection of established melanomas was dependent on IFN-γproduction and irradiation of tumor bearing mice 23. So far we know the role of TGF-β, IL-6, IL-21 and IL-1 for Th17 differentiation and of IL-23 to promote Th17 proliferation 7, 8. IL-6 and TGF-β synergize to promote Th17, whereas TGF-β without inflammatory cytokines promotes inducible Treg (iTreg) differentiation 24. Moreover, we recently reported that IL-6 treatment of iTreg and natural Treg promoted the conversion of these lineages into Th17 cells via down regulating Foxp3. Conversely, TGF-β induction of Foxp3 antagonizes RORγt and ROR function, which is required for Th17 differentiation 24. These data indicate a greater plasticity of Th cells than previously expected. Future studies will shed more light on the exact combination of signals that promote the maintenance of Th17 phenotype, or facilitate the conversion to Th1 in vivo in a full versus empty immune compartments and help understand their pathological role in autoimmune disorders.
Then what is the function of IL-17 or Th17 cells in IDDM? Our transfer study indicates that Th17 cells induced different types of inflammation in the islets. Anti-IL-17 inhibited insulitis in NOD.scid and neonatal NOD mice. We think that Th17 cells may play an accessory role in IDDM by promoting inflammation, which may enhance the islet destruction by Th1 or CD8+ T cells. Nonetheless, our data indicate an involvement of Th17 cells in IDDM, which may be further investigated in human IDDM patients.
Materials and methods
NOD/LTJ (NOD/ShiLtJ) and NOD SCID (NOD.CB17-Prkdcscid/J) mice were purchased from The Jackson Laboratory (Bar Harbor, ME, USA). BDC2.5/NOD mice were kindly donated by Diane Mathis and Christophe Benoist. Mice were housed and bred in the SPF animal facility at the MD Anderson Cancer Center and the experiments were performed using protocols approved by the Institutional Animal Care and Use Committee.
Quantitative real-time PCR
Total RNA was prepared from pancreas using TriZol reagent (Invitrogen). cDNA were synthesized with superscript reverse transcriptase and oligo (dT) primers (Invitrogen and gene expression was examined with a Bio-Rad iCycler Optical System using an iQ SYBR green real-time PCR kit (Bio-Rad). The data were normalized to a β-actin reference. The primers for IL-17 were: forward 5′-CTCCAGAAGGCCCTCAGACTAC-3′, reverse 5′-GGGTCTTCATTGCGGTGG-3′; the primers for IL-17F were: forward 5′-CCCATGGGATTACAACATCACTC-3′, reverse 5′-CACTGGGCCTCAGCGTC-3′.
Th differentiation and T-cell transfer for diabetes development
For Th17 differentiation, spleen and LN cells from BDC2.5 NOD mice were cultured with 0.5 μg/mL of peptide mimotope 63 RTRPLWVRME 15, 5 ng/mL of TGF-β, 10 ng/mL of IL-6 (Peprotec), 50 μg/mL of anti-IFN-γ (XMG 1.2), 10 μg/mL of anti-IL-4 (11B11) and 100 ng/mL of IL-23 (R&D Systems). For Th1 differentiation, the cells were cultured with 0.5 μg/mL of peptide 63, 10 μg/mL of anti-IL-4 and 350 pg/mL of IL-12. On day 3, a sample of the culture was reactivated with 10 ng/mL PMA and 250 ng/mL of ionomycin or 0.5 μg/mL of peptide 63 and Golgi-Plug for 5 h (BD Biosciences) to further process for ICS for IL-17 and IFN-γ. On day 4, 1 million cells of BDC2.5 Th17 cells were transferred i.v. into NOD.scid or i.p. into 10-day-old NOD mice. When diabetes was induced with purified CD4+ cells from diabetic females, CD4+ cells were purified using anti-CD4+ magnetic beads from Miltenyi Biotec following manufacturer instructions. All mice were monitored daily for urine glucose levels (Diastix; Bayer Pharmaceuticals) and high reads were confirmed by blood glucose measurements (Ascencia Elite; Bayer Pharmaceuticals). Diabetes was scored after three consecutive reads higher than 13.5 mM/L. For labeling of IFN-γ-secreting cells, we use IFN-γ secretion assay detection kit (Miltenyi Biotec). On day 3 of the Th17 differentiation, T cells were harvested, washed and cultured with 1 μg/mL of peptide 63 and irradiated APC following manufacturers instructions for activation. Labeling with anti-IFN-γ catch reagent and anti-IFN-γ detection antibody was done as recommended by Miltenyi Biotec. CD4+ IFN-γ-cells were then sorted in an FACS Aria sorter (BD Biosciences) and further transferred into NOD.scid mice. A sample of the sorted cells was activated with 0.5 μg/mL of peptide 63, irradiated APC and Golgi stop for evaluation of IL-17, IFN-γ and IL-17F production by ICS.
Pancreata from NOD.scid, NOD/LTJ mice were fixed in 10% formalin, and embedded in paraffin. Six-micrometer sections were cut every 100 μm through the total tissue and H&E staining was performed; 50–100 islets per mice were scored for presence and type of insulitis. For IL-17F staining pancreata were frozen in tissue freezing medium (Triangle Biomedical Sciences). Immuno-staining was carried out on 10-μm sections. Sections were fixed with acetone at −20°C for 10 min and then air-dried. Sections were incubated with a 1:300 dilution of rabbit polyclonal anti-IL-17F antibody (tested in house), CD4, F4/80 and GR1. Normal rabbit IgG was used as isotype control.
Antibodies and flow cytometry
CD4 PercPCy5.5, TCR-β FITC, Vβ4 PE, IFN-γ APC and IL-17 PE antibodies were from BD Biosciences. Polyclonal antibody against IL-17F was conjugated with Alexa 647 (Invitrogen) following manufacturer's instructions. All cell suspensions were incubated with anti-CD32/CD16 for 30 min at 4°C before staining with the antibody mixture. ICS for IFN-γ and IL-17 was performed using BD Cytofix/Cytoperm kit following manufacturer's instructions. Cells were analyzed on FACScalibur (BD Biosciences) and files were analyzed using FlowJo 8.4.3 (TreeStar).
We thank the entire Dong lab for their help and discussion and Pamela Grant for help with histology. This work is supported in part by grants from National Institute of Health and grants from the Center for Target Therapy of MD Anderson Cancer Center (to C.D.). S.H.C. receives a postdoctoral fellowship from the Arthritis Foundation and C.D. receives a Cancer Research Institute Investigator award, an American Lung Association Career Investigator award, a Leukemia and Lymphoma Society Scholar award and a Trust Fellowship of the MD Anderson Cancer Center.
Conflict of interest: The authors declare no financial or commercial conflict of interest.