Regulatory T Cell–Derived TRAIL Is Not Required for Peripheral Tolerance

TRAIL (Tnfsf10/TRAIL/CD253/Apo2L) is an important immune molecule that mediates apoptosis. TRAIL can play key roles in regulating cell death in the tumor and autoimmune microenvironments. However, dissecting TRAIL function remains difficult because of the lack of optimal models. We have now generated a conditional knockout (Tnfsf10L/L) for cell type–specific analysis of TRAIL function on C57BL/6, BALB/c, and NOD backgrounds. Previous studies have suggested a role for TRAIL in regulatory T cell (Treg)–mediated suppression. We generated mice with a Treg-restricted Tnfsf10 deletion and surprisingly found no impact on tumor growth in C57BL/6 and BALB/c tumor models. Furthermore, we found no difference in the suppressive capacity of Tnfsf10-deficient Tregs and no change in function or proliferation of T cells in tumors. We also assessed the role of TRAIL on Tregs in two autoimmune mouse models: the NOD mouse model of autoimmune diabetes and the myelin oligodendrocyte glycoprotein (MOG) C57BL/6 model of experimental autoimmune encephalomyelitis. We found that deletion of Tnfsf10 on Tregs had no effect on disease progression in either model. We conclude that Tregs do not appear to be dependent on TRAIL exclusively as a mechanism of suppression in both the tumor and autoimmune microenvironments, although it remains possible that TRAIL may contribute in combination with other mechanisms and/or in different disease settings. Our Tnfsf10 conditional knockout mouse should prove to be a useful tool for the dissection of TRAIL function on different cell populations in multiple mouse models of human disease.

TRAIL was initially discovered as a molecule that specifically targets malignant cells and spares nonmalignant cells. TRAIL-or DR5-deficient mice are more susceptible to tumor growth and metastasis, implicating an important role for TRAIL in controlling tumor growth (23)(24)(25)(26)(27)(28)(29)(30)(31). This tumor-specific killing is primarily mediated by NK cells and CD8 + T cells in the tumor microenvironment (TME), although other cells express TRAIL in the TME (17,31,32). Moreover, although TRAIL is a molecule that targets cell death, it can also regulate immune cell function and proliferation (33).
Regulatory T cells (T regs ) are an immunosuppressive subset of CD4 + T cells that can suppress activated immune cells and limit autoimmunity. For example, T regs are critical for limiting multiple models of autoimmunity such as the NOD mouse, a spontaneous model of autoimmune diabetes, and the myelin oligodendrocyte glycoprotein (MOG) C57BL/6 model of experimental autoimmune encephalomyelitis (EAE). T reg depletion in these models rapidly results in overt diabetes and exacerbated EAE disease severity, respectively (34)(35)(36). Despite this important role, T regs can also suppress the antitumor response and therefore are an effective barrier to limiting tumor growth (37,38). T regs have multiple mechanisms of suppression and can use these mechanisms in the TME and autoimmune environment. T regs can suppress through production of inhibitory cytokines, targeting of dendritic cell function, metabolic disruption, and direct cytolysis (39)(40)(41). Our laboratory has shown that T regs from IL-10-and IL-35-deficient C57BL/6 mice upregulated TRAIL to suppress responding T cells and that T regs from BALB/c mice express higher levels of TRAIL than T regs from C57BL/6 mice (42). In addition, it has been reported that T regs produce TRAIL in an allogenic skin graft model to suppress activated T cells (43). Taken together, these observations suggest that T regs can use TRAIL to suppress immune cells in various disease environments.
In this study, we had two specific goals: 1) investigate TRAIL function in an inducible, cell type-specific manner by generating Tnfsf10 L/L mice on C57BL/6, BALB/c, and NOD backgrounds, as studies thus far have only used blocking Abs or constitutive Tnfsf10 knockout mice, and 2) assess if T regs require and/or are dependent on TRAIL as a mechanism of suppression within the tumor or autoimmune microenvironment by use of Tnfsf10 L/L Foxp3 Cre mice.

Mice
Foxp3 Cre-YFP mice on a C57BL/6 background were obtained from A. Y. Rudensky (Memorial Sloan-Kettering) (44). Foxp3 Cre mice on a BALB/c background were obtained from S. Sakaguchi (Osaka University) (45). Foxp3 Cre-GFP .NOD mice were obtained from J. A. Bluestone (University of California, San Francisco) (46). All animal experiments were performed in the American Association for the Accreditation of Laboratory Animal Careaccredited, specific pathogen-free facilities in Division of Laboratory Animal Resources, University of Pittsburgh School of Medicine. Female and male mice of 4-6 wk of age were used for B6 and BALB/c experiments. All tumor phenotype and functional experiments were performed at 12 d after tumor inoculation unless otherwise specified. Female and male NOD mice were followed for diabetes incidence up to 30 wk of age. All NOD phenotype and functional experiments were performed with female mice at 10 wk unless otherwise specified. Animal protocols were approved by the Institutional Animal Care and Use of Committees of University of Pittsburgh.

Generation of a Tnfsf10 L/L mouse
The Tnfsf10 L/L targeting construct was generated using standard recombineering methods (47). Initially, 26.7 kb of the Tnfsf10 locus were retrieved from a bacterial artificial chromosome plasmid and an Loxp-Neo-Loxp cassette inserted 313 bp upstream of exon 2. The Neo was removed via Cre-mediated recombination, leaving a single Loxp and an StuI restriction site (inserted into the intron of the retrieved Tnfsf10 locus). An Frt-Neo-Frt-Loxp cassette was then inserted 573 bp downstream of exon 5 to establish an alter-nate exon 2 containing the following: a SpeI restriction site, the splice acceptor from exon 2, "self-cleaving" T2A peptide sequence, a truncated version (nonfunctioning) of the human nerve growth factor receptor (hNGFR), and the SV40 polyadenylation sequence. The linearized targeting construct was electroporated into JM8A3.N1 embryonic stem cells (C57BL/6N background) and neomycin-resistant clones were screened by Southern blot analysis using StuI and SpeI restriction digests for the 5′ and 3′ ends, respectively. Correctly targeted clones were 100% normal diploid by karyotype analysis and were injected into C57BL/6 blastocysts. Chimeric mice were mated to C57BL/6 mice and transmission of the targeted allele verified by PCR. The mice were crossed with actin flippase mice to remove the Neo cassette. The mice were backcrossed >10 generations onto the BALB/c or NOD background and verified by microsatellite analysis. Genotyping primers are 5′-GCCCACGGGTGTAAAGAGCAGTTC-3′, 5′-GGTGGAACAGCTGACAGACATGATAAGATAC-3′, and 5′-GTCTCCCCAGTCCAATCACTGCTAC-3′. Primers for detection of exon 1 of Tnfsf10 are forward 5′-GCACTCCGCCTTCTAACTGT-3′ and reverse 5′-GTGCTGACTGAAGCTGAGGT-3′, exon 2 forward 5′-GACGGATGAGGATTTCTGGGAC-3′ and reverse 5′-TTCAATGAGCTGATACAGTTGCC-3′, and exon 5 forward 5′-ATGGAAAGACCTTAGGCCAGA-3′ and reverse 5′-TAGATGTAATACAGGCCCTCCTGC-3′.

Measurement of diabetes and insulitis
Measurement of diabetes and insulitis were performed as previously described (48)(49)(50). Briefly, diabetes incidence was monitored weekly through presence of glucose in the urine with Diastix (Bayer). Mice positive for glucose on Diastix were then measured for blood glucose with a Breeze2 glucometer (Bayer). Mice were considered diabetic and were marked for sacrifice when blood glucose was ≥400 mg/dl. Pancreata for histology were prepared as previously described at the University of Pittsburgh Biospecimen Core (48). Briefly, pancreata were embedded in a paraffin block and cut into 4-μm-thick sections with 150-μm steps between sections and stained with H&E. An average of 60-80 islets per mouse were scored in a blinded manner. Two methods of insulitis measurement were used as previously described (51).

Islet isolation and lymphocyte preparation
Islets were prepared as previously described (48,52). Briefly, 3 ml of collagenase (600 U/ml in complete HBSS with 10% FBS) was perfused through the pancreatic duct. Pancreata were then incubated for 30 min at 37°C. Pancreata were then washed two times and resuspended in clear complete HBSS with 10% FBS, and islets were isolated by hand under a dissecting microscope. Isolated islets were dissociated with 1 ml dissociation buffer (Life Technologies) for 15 min at 37°C with vortexing every 5 min. Cells were washed, resuspended, counted, and used.

EAE induction
Induction of EAE was performed as described previously (53,54). Briefly, IFA (Difco) at was supplemented with 5 mg/ml Mycobacterium tuberculosis (Difco) to make CFA. MOG peptide (AAPPTec) was diluted to 1 mg/ml in PBS, and the CFA and MOG peptide were mixed at a 1:1 ratio. Mice were injected with 100 μl of the emulsion on both flanks s.c. Pertussis toxin (200 ng/200 μl PBS; Sigma-Aldrich) was injected i.p. on day 0 and day 2 of injection. Animals were scored blinded for clinical symptoms as follows: 0, no change; 1, limp tail; 2, partial hind limb paralysis; 3, full hind limb paralysis; 4, full hind limb paralysis and partial front limb paralysis; and 5, moribund or death.

In vitro assays
Microsuppression assays were performed as previously described (59,60). Briefly, T reg cells were isolated from the spleen of naive mice or nondraining lymph node (NDLN) and tumor-infiltrating lymphocytes (TIL) of mice 12 or 18 d after injection with B16 or CT26. Isolated T regs were cocultured with CellTrace Violet (Life Technologies)-labeled CD4 + Foxp3 − responder T cells in the presence of mitomycin C-treated, TCRβ-depleted splenocytes and anti-CD3ε (1 μg/ml) for 72 h at 37°C.

Quantification and statistical analysis
Statistical analysis was performed with Prism version 8.0.0. Student t tests were used when only two experimental groups were involved. Tumor growth and EAE curves were analyzed using two-way ANOVA with multiple comparisons correction with sequential time point measurements. The log-rank (Mantel-Cox) test was used for diabetes incidence statistical analysis. Number of mice used in the experiment is represented by "n," with number of individual experiments listed in legend. All p values were two sided, and statistical significance assessed at ≤0.05.

TRAIL is expressed on T regs in the TME
We hypothesized that T regs use TRAIL to suppress the antitumor response. Therefore, we initially assessed TRAIL expression in multiple cell populations isolated from the TME of B16 tumor-bearing mice, and we found substantial upregulation of Tnfsf10 transcript in the TIL compared with the NDLN (Fig. 1A). Interestingly, T regs and CD4 + Foxp3 − were trending to have higher Tnfsf10 levels in the TME compared with other cells in the TME.
It is important to note that TRAIL protein expression was difficult to discern, as previously reported, which may be due to its low level of expression (61).

Generation of a Tnfsf10 L/L mouse
To directly access the importance of TRAIL expression in distinct cell types in the TME, in particular in T regs , we generated a novel Tnfsf10 L/L mouse. LoxP sites were inserted in the intron between exons 1 and 2 and following exon 5 along with an artificial exon containing a truncated nonfunctional version of the hNGFR (Fig. 1B, 1C). The hNGFR was intended to serve as a reporter for Cre-mediated deletion of Tnfsf10. However, upon validation of the strain, it was found that expression of hNGFR was minimal following Cre-mediated deletion, likely because of the weak transcription strength of the Tnfsf10 promoter consistent with challenges experienced in detected TRAIL expression (data not shown). This may also have been due to inefficient splicing into the artificial exon. To assess the role of TRAIL in T regs , we crossed the Tnfsf10 L/L mice with Foxp3 Cre-YFP .B6 mice, and fidelity of T reg -specific deletion was verified by cell specific genotyping (Fig. 1D, 1E). Taken together, we have successfully generated a Tnfsf10 L/L murine model, thus enabling us to specifically examine the role of TRAIL in T regs .
T reg -restricted deletion of Tnfsf10 does not affect tumor growth or suppression in C57BL/6 mice Our laboratory and others have suggested that T regs from C57BL/6 mice can use TRAIL to suppress the immune response (42,43). To assess this, we first examined the suppressive capacity of T regs from naive Tnfsf10 L/L Foxp3 Cre-YFP .B6 mice. Surprisingly, the suppressive capacity of Tnfsf10-deficient T regs was equivalent to wild-type (WT) T regs (Fig. 2A). Next, to assess if T regs primarily depend on TRAIL to suppress the antitumor response, we injected Foxp3 Cre-YFP .B6 and Tnfsf10 L/L Foxp3 Cre-YFP .B6 mice with B16 melanoma. We chose this model because of studies describing the important role of T reg suppression in B16 tumor growth (59,62). However, we found no difference in B16 tumor growth in Tnfsf10 L/L Foxp3 Cre-YFP .B6 mice (Fig. 2B).
Furthermore, T regs from the NDLN or TIL of Tnfsf10 L/L Foxp3 Cre-YFP .B6 mice with B16-bearing tumors were fully capable of suppressing in vitro (Fig. 2C). Moreover, the suppressive activity of T regs from Tnfsf10 L/L Foxp3 Cre-YFP .B6 mice did not change if T regs were isolated at a later time point (Supplemental Fig. 1A). We also examined an additional tumor model, MC38 colon adenocarcinoma, which has been shown to be sensitive to TRAIL-induced cytotoxicity, but found no differences in tumor growth between Foxp3 Cre-YFP .B6 and Tnfsf10 L/L Foxp3 Cre-YFP .B6 mice (Fig. 2D) (63). In an effort to understand if T reg -restricted deletion of Tnfsf10 would impact tumor growth in a model of an active immune response that justifies a strong involvement of T reg -mediated negative feedback, we treated Tnfsf10 L/L Foxp3 Cre-YFP .B6 mice with anti-PD-1 therapy and found no change in response to the immunotherapy (Fig. 2D).
T regs use TRAIL to suppress through induction of cell death in CD4 + Foxp3 − T cells (42,43). However, in Tnfsf10 L/L Foxp3 Cre-YFP .B6 mice, we did not find a difference in activation/cleavage of the main downstream executioner caspase 3 in CD4 + Foxp3 − or CD8 + T cells when compared with Foxp3 Cre-YFP .B6 mice (Fig. 2E, 2F). We also assessed other immune and nonimmune populations, including tumor cells, but did not find differences in cell death (Supplemental Fig. 1B-E). This indicated that loss of TRAIL in T regs did not affect cell death in immune and nonimmune populations in the TME. Interestingly, the low expression of the murine TRAIL agonistic cell DR5 may explain the lack of effect of T reg -mediated deletion of TRAIL (Supplemental Fig. 1F).
TRAIL can also suppress responding cells by inhibiting proliferation and T cell activation/ function rather than cytotoxicity (64)(65)(66)(67). However, the proliferation of CD4 + Foxp3 − and CD8 + T cells, measured by Ki67 expression, was not affected (Fig. 2G, 2H). We also analyzed the functional status of CD4 + Foxp3 − and CD8 + T cells and found no changes in production of proinflammatory cytokines TNF-α and IFN-γ (Fig. 2I-L). We conclude that T reg -restricted deletion of Tnfsf10 does not affect T reg suppression, tumor growth, cell death, or proliferation and function of T cells.
Next, we hypothesized that T reg -restricted deletion of TRAIL may not lead to a change in tumor growth because Tnfsf10 L/L Foxp3 Cre-YFP .B6 T regs still retain other mechanisms of suppression. Thus, we examined the expression of suppressive molecules IL-10, LAP-TGF-β, CTLA4, CD39, and CD73, and indeed, expression was equivalent between WT T regs and TRAIL-deficient T regs (Supplemental Fig. 1G-K). Moreover, expression of the proliferation marker, Ki67, and markers of activation/exhaustion, PD-1 and LAG3, remained unchanged in the T regs in tumors of Tnfsf10 L/L Foxp3 Cre-YFP mice (Supplemental Fig. 1L-P). These results further indicate that the suppressive phenotype of Tnfsf10-deficient T regs is unaffected.
We also found no change in the proportion of T regs or proportion of total immune cells in the tumor at day 12 (Supplemental Fig. 1Q and 1R) or day 18 (Supplemental Fig. 1S). Finally, although others have argued that TRAIL plays a role in T reg apoptosis, we found no change in T reg cell death in the TME (Supplemental Fig. 1T) (68). Taken together, these data suggest that T regs are not primarily dependent upon TRAIL to suppress in the TME via cell death, inhibition of cell proliferation, or function. This may be due to minimal expression of DR5 and/or the use of other suppressive molecules.

T reg -restricted deletion of Tnfsf10 does not affect tumor growth or suppression in BALB/c mice
Although we did not observe a primary role for TRAIL in T regs in C57BL/6 mice, we hypothesized we may see differences in BALB/c mice given our previous studies in which TRAIL had a more predominant role in BALB/c T regs compared with T regs from C57BL/6 mice (42). Moreover, other studies have revealed TRAIL can play a part in regulating the Th1/Th2 balance (69)(70)(71)(72). Therefore, we backcrossed the Tnfsf10 L/L mice to the Th2-prone BALB/c background and then crossed it to the BALB/c Foxp3 Cre mouse (45).
Initially, we assessed the function of naive TRAIL-deficient T regs in a standard in vitro suppression assay, and interestingly, the level of suppression was equivalent to WT T regs (Fig. 3A). Next, we assessed tumor growth in Foxp3 Cre-YFP .BALB/c, Tnfsf10 L/L .BALB/c, and Tnfsf10 L/L Foxp3 Cre-YFP .BALB/c mice using the BALB/c CT26 colon carcinoma model in which T regs suppress the antitumor response (73,74). Although we did not observe a difference in tumor growth (Fig. 3B), we did see a small decrease in suppression in TRAIL-deficient T regs isolated from CT26 tumors compared with WT T regs (Fig. 3C). However, this was not the case at a later time point (Supplemental Fig. 2A). Next, we determined that cleaved caspase levels in CD4 + Foxp3 − , CD8 + T cells, tumor cells, and other cell populations were equivalent (Fig. 3D, 3E) (Supplemental Fig. 2B-E), suggesting that T regs were not dependent upon TRAIL-mediated cytotoxicity in the TME of BALB/c mice, possibly because of low DR5 expression in the TME (Supplemental Fig. 2F).
Furthermore, we did not see any changes in Ki67, TNF-α, and IFN-γ in T cells, suggesting that T regs do not suppress by limiting proliferation nor function of responding T cells (Fig. 3F-K). We also observed that TRAIL-deficient T regs in the TME still retained other suppressive molecules, indicating that other molecules may aid in suppression in the TME despite loss of TRAIL (Supplemental Fig. 2G-K). Furthermore, we did not see any differences in expression of Ki67, PD-1, LAG3, and cleaved caspase 3 on T regs (Supplemental Fig. 2L-Q). The proportion of immune cells and T regs remained unchanged on both days 12 and 18 (Supplemental Fig. 2R-T). Taken together, these data suggest that despite the reported higher levels of TRAIL expression in BALB/c T regs , they are not primarily dependent upon TRAIL as a means of suppression in the TME (42).

T reg -restricted deletion of Tnfsf10 does not affect autoimmune diabetes
Because T regs are also critical in limiting autoimmunity, we hypothesized that T regs may use TRAIL to suppress in the autoimmune microenvironment. Also, it has been reported that TRAIL can regulate cell death of diabetogenic T cells in the pancreatic islet of NOD mice (75). Although it was proposed that this was mediated by TRAIL-expressing pancreatic β cells, we hypothesized that T regs may also use TRAIL to suppress T cells in this environment (75). Indeed, T cells express the highest levels of Tnfsf10 in the islet (Fig.  4A). We hypothesized that T reg -restricted deletion of Tnfsf10 would limit suppression of diabetogenic T cells and lead to exacerbated autoimmune diabetes.
Interestingly, we found that deletion of Tnfsf10 in T regs did not significantly alter diabetes incidence or insulitis in female (Fig. 4B-D) or male (Supplemental Fig. 3A) mice, although there was a slight trend toward reduced diabetes incidence. Moreover, we did not find any changes in cell death in CD4 + Foxp3 − and CD8 + T cells in the islet (Fig. 4E, 4F). As seen with our tumor data, we found that the levels of proliferation and cytokine production in the diabetogenic T cells of the islet were similar in both WT and Tnfsf10 L/L Foxp3 Cre-GFP .NOD mice (Fig. 4G-L). This would indicate that T regs do not require TRAIL to suppress diabetogenic T cells in the pancreatic islet of NOD mice.
We also examined DR5 expression on immune and nonimmune cells in the islet and found minimal expression of DR5 on immune cells but higher expression on insulin-positive β cells (Supplemental Fig. 3B). Reports of direct TRAIL-mediated β cell killing have been inconsistent (76)(77)(78)(79)(80). However, upon examination of insulin-positive cells, we found no change in cell death (Supplemental Fig. 3C). Interestingly, we did see a reduction in cell death in the CD11c + population (Supplemental Fig. 3D). TRAIL can have an effect on dendritic cells (81); however, it is unclear what impact this may play in our system, as we did not see a consequence of altered disease. Future studies may elucidate what other impact this has in autoimmune diabetes.
We found that Tnfsf10-deficient T regs isolated from the TME retained their suppressive phenotype. We questioned if this remained true for Tnfsf10-deficient T regs isolated from the islet. We found T regs still expressed functional markers such as LAP-TGF-β, IL-10, and CD39 (Supplemental Fig. 3E-G) and even had an increase in CD73 expression (Supplemental Fig. 3H). This further indicates that Tnfsf10-deficient T regs retain their suppressive phenotype in the islet. As seen in the tumor, we found no change in T reg proliferation (Supplemental Fig. 3I), as measured by Ki67, and no change in activation/ exhaustion markers PD-1 and LAG3 (Supplemental Fig. 3J-M).
We had demonstrated above that TRAIL had no effect on T reg cell death or the proportion of immune cells and T regs in the TME. Interestingly, although we did not observe a difference in immune cell proportions within the islet (Supplemental Fig. 3N), we did see an increased proportion of intra-islet T regs in Tnfsf10 L/L Foxp3 Cre-GFP .NOD mice (Supplemental Fig.   3O). Interestingly, reduced T reg death was only observed in 10-wk-old mice (Supplemental Fig. 3P), as there was no difference in 12-wk-old mice (Supplemental Fig. 3Q). Therefore, we conclude that T regs are not dependent on TRAIL to suppress in the islet.
Finally, we examined if T reg -derived TRAIL had a role in the MOG model of EAE using the Tnfsf10 L/L Foxp3 Cre-YFP .B6 mice. As seen with the tumor and NOD models, we did not observe a difference in EAE score and initiation of the disease between WT and Tnfsf10 L/L Foxp3 Cre-YFP .B6 mice (Supplemental Fig. 3R). Therefore, we conclude that T regs do not require nor are dependent on TRAIL as a means of suppression in autoimmune microenvironments.

DISCUSSION
We report four key developments from our studies. First, we created the first conditional Tnfsf10 L/L knockout mouse, that we are aware of, which allows for cell type-specific deletion of TRAIL. Although we focused our efforts on understanding TRAIL biology in T regs , this novel resource could be used to examine the role of TRAIL in other cell populations. Second, we used the Tnfsf10 L/L mice and determined that T regs are not primarily dependent upon TRAIL as a means of suppression within the TME. Third, we found that T regs from autoimmune diabetes and EAE are not primarily dependent upon TRAIL as a means of suppression. Finally, these data, along with our previous work in which multiple mechanisms of T reg suppression were deleted, suggest that T regs are capable of using multiple mechanisms of suppression and are able to overcome or compensate when a mechanism is compromised or blocked.
Finally, although we did not determine a primary role of TRAIL in T regs within the tumor and autoimmune environments, we cannot rule out the possibility that TRAIL does play a role in T reg function, either in concert with other mechanisms or in disease models we did not examine. It may be important in future studies to assess different models in which DR5 is more highly expressed. It will also be important to examine the role of TRAIL in the absence of other mechanisms of T reg suppression, such as IL-10 or IL-35, in other cell types, and in other disease models such as infectious or inflammatory diseases.

Supplementary Material
Refer to Web version on PubMed Central for supplementary material. Vignali