Pancreatic Islets Engineered with SA-FasL Protein Establish Robust Localized Tolerance by Inducing Regulatory T Cells in Mice

Type 1 diabetes (T1D) is an autoimmune disease caused by the destruction of insulin-producing β cells by a complex set of immunological events initiated and coordinated by CD4⁺ T cells responding to a set of β cell-specific Ags (1–3). Restoration of insulin-secreting β cell mass using allogeneic islet transplantation has been viewed as a preferred treatment modality, and its efficacy in restoring physiological glycemic control has been demonstrated in clinical trials (4). However, the success of allogeneic islet transplantation is compromised by immunological rejection and secondary graft failure due to the continuous use of immunosuppressive drugs to control rejection (5). Therefore, novel approaches that specifically target and control destructive auto- and alloimmune responses without continuous immunosuppression remain to be developed for the successful application of allogeneic islet transplantation in the clinic.

Inasmuch as T cells play a critical role in the initiation of islet-destructive auto- and alloreactive immune responses (6), specific elimination of these cells or control of their function through active regulatory mechanisms may prove effective in achieving long-term islet allograft survival without the continuous use of immunosuppression (7). In this context, immunomodulation with FasL presents an attractive approach due to the critical role played by Fas/FasL-mediated apoptosis in activation-induced cell death (8), an important homeostatic molecular mechanism that controls T cell responses to self-Ags (9). The immunomodulatory function of FasL has been extensively exploited for the induction of tolerance to auto- and alloantigens using gene therapy (10–15). However, although gene therapy showed efficacy in some settings (10, 12–15), the controlled ectopic expression of FasL in transfected cells and tissues is not only technically challenging, but also poses safety concerns.

We recently generated a chimeric form of FasL protein, streptavidin (SA)-FasL, in which the extracellular domain of FasL lacking potential metalloproteinase sites was cloned C terminus to the core SA (16). This molecule exists as tetramers and oligomers with potent apoptotic activity and can be displayed on the surface of biotinylated cells in an efficient and rapid manner (16). Most importantly, systemic immunomodulation with SA-FasL–engineered donor splenocytes resulted in tolerance to cardiac allografts (17). However, the application of this novel approach to engineering tissues remains to be demonstrated. In this study, we tested if pancreatic islets, instead of isolated cells, can be engineered with SA-FasL protein and whether the engineered islets overcome rejection and establish euglycemia following transplantation into chemically diabetic allogeneic hosts. Our data demonstrate for the first time, to our knowledge, that pancreatic islets can be engineered with SA-FasL in a rapid and efficient manner, and such engineered islets under transient cover of rapamycin induce localized allotolerance that was initiated and maintained by CD4⁺CD25⁺Foxp3⁺ regulatory T (Treg) cells homing to the graft and graft-draining lymph nodes (LNs).

C57BL/6 (CD45.2; H-2^b), C57BL/10 (CD45.2; H-2^b), C57BL/6 Foxp3^EGFP, and BALB/c (H-2^d) mice were purchased from The Jackson Laboratory. Congenic C57BL/6.SJL (CD45.1; H-2^b) and TCR-transgenic OT-I (CD8⁺ T cell) mice on the rag2^−/− background were purchased from Taconic Farms (Germantown, NY) and bred in our specific pathogen-free animal housing facility at the University of Louisville using protocols approved by the Institutional Animal Care and Use Committee. Recombinant SA, human SA-CD40L, and rat SA-FasL proteins were produced in our laboratory using the Drosophila DES expression system (Invitrogen) as previously described (16, 18).

Pancreatic islets were harvested from 8–12-wk-old BALB/c mice under anesthesia using a standard protocol as previously described (16). Islets were engineered by first incubating in 5 μM EZ-Link Sulfo-NHS-LC-Biotin solution (Thermo Scientific) in PBS at room temperature for 30 min followed by extensive washing to remove free biotin. Biotinylated islets were then incubated in PBS containing SA-FasL or SA-CD40L proteins (200 ng protein/450–550 islets/200 μl PBS) or equal molar of SA as a control at room temperature for 30 min.

After several washes, islets were cultured in vitro for various days, stained first with anti–SA-FITC Ab (Vector Laboratories) to detect SA-FasL and SA-CD40L, washed several times in PBS, and then stained with SA-allophycocyanin (BD Biosciences) to visualize biotin. Z-stack analysis was performed to measure fluorescence intensity of SA-FasL on engineered islets using LAS AF software on Leica TCS SP5 confocal microscopy (Leica Microsystems). Unmodified or SA-engineered islets were used as controls for background staining.

Diabetes was induced in C57BL/6 mice by i.v. injection of streptozotocin (200 mg/kg) and confirmed by two consecutive blood glucose readings >300 mg/dl. Pancreatic islets were harvested, cultured overnight, and then engineered with SA, SA-CD40L, or SA-FasL proteins. These islets were then immediately transplanted under the kidney capsule of diabetic mice (450–550 islets/mouse). Unless otherwise indicated, graft recipients were injected i.p. with 0.2 mg/kg rapamycin (LC Company) starting on the day of transplantation daily for 15 d. Animals were monitored for diabetes, and those with two consecutive daily measurements of ≥250 mg/dl blood glucose level were considered diabetic and confirmation of graft failure.

To investigate the nature of observed tolerance, we performed two sets of experiments on long-term (>100 d) graft acceptors. In the first set, unilateral nephrectomy was performed to remove the SA-FasL–engineered islet graft. After confirmation of hyperglycemia, a second set of unmodified islet graft was transplanted under the contralateral kidney capsule. To eliminate the effect of surgery associated inflammation on graft rejection, in a second model, unmodified donor islets were transplanted under the contralateral kidney capsule 40 d prior to surgical removal of the kidney harboring SA-FasL–engineered islet graft.

Lymphocytes from peripheral LNs, spleen, and kidney-draining LNs (KDLNs) of various treatment groups were harvested, stained with Abs to mouse CD4-allophycocyanin and CD25-PE (BD Pharmingen) molecules, washed with PBS, and permeabilized/fixed overnight at 4°C using a Permeabilization/Fixation kit from eBioscience. The FcγII/III receptors were blocked using 2.4G2 Ab, followed by staining with anti–Foxp3-FITC Ab according to the manufacturer’s protocol (eBioscience). The cells were run on the FACSCalibur (BD Biosciences), and the data were analyzed by FlowJo software (Tree Star).

Snap-frozen sections of islet grafts were incubated in a blocking solution (0.5% Triton X-100, 0.1% BSA, 5% goat serum, and 1:400 FcγII/III receptor block). Guinea pig anti-insulin Ab (DakoCytomation) and a rat anti-mouse CD4 mAb (BD Pharmingen) were used to detect islets and CD4⁺ T cells, respectively. Staining of these Abs was visualized using secondary Abs conjugated with Alexa Fluor-647 (CD4) and Alexa Fluor-555 (insulin) followed by direct staining with FITC-conjugated rat anti-Foxp3 Ab (eBioscience) to visualize Treg cells. Hoechst (Molecular Probes) was used to stain the nucleus of the cell. Fluorescent images were obtained using Leica TCS SP5 confocal microscopy (Leica Microsystems) under ×20 original magnification. H&E staining was performed on formalin-fixed and paraffin-embedded kidney tissue blocks as previously described (17).

Selected groups of mice were injected i.p. with 300 μg purified anti-CD25 Ab/animal (PC61; Bio X Cell) to deplete CD25⁺ T cells on days 14 or 100 posttransplantation of SA-FasL–engineered allogeneic islets. Depletion was confirmed using the 7D4 Ab recognizing a different epitope of CD25 than PC61 Ab (19) to stain PBLs harvested at various times post-PC61 treatment. Groups of mice were also injected i.p. with 300 μg/animal isotype Ab or NK1.1 Ab (clone PK136; Bio X Cell) on day 14 posttransplantation as controls.

CD4⁺CD25⁺ T cells were sorted from KDLNs or spleens of long-term SA-FasL–engineered islet graft acceptors, SA-engineered islet graft rejectors, and naive C57BL/6 mice using Abs against CD4-FITC and CD25-PE in flow cytometry. CD4⁺CD25⁻ effector T (Teff) cells from naive C57BL/6 mice spleen were sorted using flow cytometry. The purity of cells was >95%. One million Teff cells were adoptively transferred i.v. into chemically diabetic OT-I mice 1 d prior to the transplantation of unmodified BALB/c islets either alone or with 2,500–10,000 CD4⁺CD25⁺ Treg cells. Mice were monitored for graft survival by assessing blood glucose levels on a regular basis.

BALB/c islets engineered with SA-FasL or SA proteins were transplanted under the kidney capsule of streptozotocin diabetic C57BL/6 Foxp3^EGFP mice subjected to daily rapamycin (0.2 mg/kg) treatment starting on the day of transplantation. Three days posttransplantation, mice were euthanized, and grafted islets were dissected with fine forceps from the kidney capsule and mechanically dispersed into single cells. Cells were stained with fluorochrome-conjugated Abs against CD3, CD4, CD8, CD25, Gr-1, and CD11b molecules and Annexin V, run on a BD LSR II flow cytometer (BD Biosciences), and analyzed using Diva software (BD Biosciences).

C57BL/6.SJL (CD45.1⁺) splenocytes were engineered with SA-FasL (40 ng SA-FasL protein/10⁶ cells) or equal molar of SA protein (20 ng SA protein/10⁶ cells), and 5–10 × 10⁶ of the engineered cells were injected i.p. into C57BL/6 (CD45.2⁺) mice. LPS (Sigma-Aldrich) injected i.p. at 10 μg/mouse served as positive control. Animals were terminated at various times (17–44 h) postinjection, and peritoneal lavage was aspirated under aseptic conditions. Lavage cells were stained with Abs against CD3, CD19, Gr-1, CD11b, CD45.1, CD45.2, and F4/80 molecules. Percentages of lavage neutrophils (CD11b⁺Gr-1^high) were assessed by gating on recipient (CD45.2⁺) cells using an LSR II and Diva software (BD Biosciences).

The t test and ANOVA were used to determine significance between two and multiple groups, respectively. Graft survival was assessed using the Kaplan-Meier method and the log-rank test. Data are expressed as mean ± SD. The p values <0.05 were considered significant. Statistical analysis was performed using SPSS 13.0 software (SPSS).

Various conditions were used to optimize the rapid and efficient display of SA-FasL on the surface of mouse pancreatic islets ex vivo without compromising their survival and function. After testing various doses of biotin and SA-FasL, concentrations of 5 μM EZ-Link Sulfo-NHS-LC-Biotin and 200 ng SA-FasL/450–550 islets were found to be optimum for islet engineering (Fig. 1A). Z-stack analysis demonstrated intense staining for both biotin and SA-FasL on the external surface of islets, with the inner core showing only moderate staining (Fig. 1B). Flow cytometric analysis of mechanically dispersed SA-FasL–engineered pancreatic islets further confirmed the presence of both biotin and SA-FasL on the majority of islet cells (Fig. 1C). Under these conditions, SA-FasL did not have any toxic effect on islets, and the protein persisted on the surface of islets for >1 wk ex vivo (Fig. 1D). Importantly, transplantation of SA-FasL–engineered islet grafts into chemically diabetic syngeneic C57BL/10 mice resulted in normalized blood glucose levels and survival in all mice over a 100-d observation period (Fig. 1E). Collectively, these results demonstrate that pancreatic islets can be engineered with SA-FasL protein in a rapid and efficient manner without compromising their function for establishing glucose homeostasis in diabetic hosts in the absence of detectable acute toxicity.

We next tested if the SA-FasL protein on islet grafts is effective in preventing rejection in allogeneic hosts in the absence of immunosuppression. SA-FasL–engineered BALB/c islets, although they showed significantly (p = 0.001) prolonged survival in the absence of any immunosuppression in chemically diabetic allogeneic C57BL/6 mice as compared with unmodified or SA protein-engineered islets, only a moderate percentage (∼18%) of grafts survived over the 100-d observation period (Fig. 2A). To improve long-term survival, SA-FasL–engineered islets were transplanted in conjunction with a short course of rapamycin treatment. The rationale for using rapamycin was 2-fold. First, rapamycin has been shown to cause apoptosis of alloreactive T cells (20, 21) and as such may serve to physically eliminate circulating alloreactive T cells evading apoptosis induced by SA-FasL displayed on islet grafts. Second, rapamycin has also been shown in various settings to induce CD4⁺CD25⁺Foxp3⁺ Treg cells (22, 23). All SA-FasL–engineered islet grafts (n = 45) survived indefinitely and normalized blood glucose levels in chemically diabetic allogeneic C57BL/6 mice treated with 0.2 mg/kg rapamycin daily for 15 doses starting on the day of transplant (Fig. 2B). In marked contrast, all control islets engineered with SA (n = 10) or human SA-CD40L (n = 5) protein that does not interact with mouse CD40 receptor (18) underwent acute rejection within 30 d. Taken together, these data demonstrate that SA-FasL alone is effective in prolonging allogeneic islet graft survival with a modest effect on long-term survival and that, in combination with a short course of rapamycin treatment, SA-FasL is effective in inducing tolerance in all islet graft recipients. Therefore, we used rapamycin as a component of our immunomodulation protocol for the rest of the studies.

To test if euglycemia is maintained by the transplanted SA-FasL–engineered allogeneic islet grafts, the kidney harboring the grafted islets was surgically removed 100 d posttransplantation. All hosts developed hyperglycemia within 3 d (n = 5). These mice were then transplanted with a second set of unmanipulated donor allogeneic islet grafts under the remaining kidney capsule. All islet grafts were rejected in acute fashion (n = 5; mean survival time [MST] = 18 ± 7.3 d; Fig. 3A). To test that surgery-associated trauma/inflammation or temporary lack of donor Ags in the mice with unilateral nephrectomy was not the cause of secondary islet graft rejection, another group of long-term survivors (n = 4) were first transplanted with a second set of unmanipulated allogeneic islets under the contralateral kidney capsule and then nephrectomized 40 d later to remove the kidney harboring the primary graft. All mice maintained euglycemia until the removal of the kidney harboring the primary islet grafts, which then resulted in the development of hyperglycemia within 3 d (Fig. 3A). These data demonstrate that the secondary unmanipulated grafts are rejected without an effect on the survival of the primary grafts, thereby demonstrating the localized nature of tolerance.

Lack of systemic tolerance was further tested by performing BALB/c skin allografts into C57BL/6 mice with long-term (>100 d) surviving islet allografts. All skin grafts rejected at a similar tempo (n = 4; MST = 12.75 ± 1.26 d) to skin grafts (n = 6; MST = 11.3 ± 1.21 d) transplanted onto naive C57BL/6 mice (Fig. 3B). Importantly, unlike islet grafts (Fig. 3A), skin graft rejection resulted in acute rejection of long-term islet grafts within 21 d (Fig. 3B). Taken together, these data demonstrate that the transient display of SA-FasL on pancreatic islet grafts is effective in inducing localized tolerance, which can be overcome by alloreactive responses against donor skin, but not islet, grafts.

Although SA-FasL–engineered islet grafts may induce localized tolerance by clonal deletion of alloreactive T cells (17), this mechanism is expected not to operate when the levels of SA-FasL on pancreatic islet grafts decline with time. Therefore, newly arising alloreactive T cells late posttransplantation need to be controlled by ongoing immunoregulatory mechanisms, such as Treg cells. To investigate the role of Treg cells in localized tolerance, lymphoid organs from long-term syngeneic and allogeneic SA-FasL–engineered allogeneic islet graft acceptors and SA-engineered islet graft rejectors were analyzed at various times posttransplantation. There were no detectable differences in the percentages of Treg cells in the spleen, mesenteric LNs, or KDLNs of all three groups (Fig. 4A, 4B, and data not shown).

In view of several recent reports providing evidence for in situ intrapancreatic immune reactions playing a critical role in the development of T1D in NOD mice (24, 25), we analyzed islet grafts using immunohistochemistry and confocal microscopy for the presence of Treg cells. On day 5 posttransplant, SA-FasL–engineered islet grafts had significantly higher numbers of Treg cells that were localized to the grafts as compared with SA-engineered control grafts (Fig. 4C, 4D). Importantly, long-term (>100 d) SA-FasL–engineered allogeneic islet grafts also had higher numbers of Treg cells residing in the periphery of islet grafts as compared with syngeneic grafts (Fig. 5A, 5B) and otherwise looked similar to syngeneic grafts with respect to insulin expression and lack of significant levels of inflammatory infiltrates (Fig. 5C, 5D).

The critical role of Treg cells in the localized tolerance was further confirmed by in vivo elimination of these cells using an Ab (PC61) against the CD25 molecule. Consistent with a previous report (26), i.p. treatment of mice with PC61 Ab (300 μg/mouse) partially depleted Treg cells because a significant percentage (∼6%) of CD4⁺ T cells in the blood remained Foxp3 positive with downregulated expression of CD25, as assessed by a second Ab (7D4) to a different CD25 epitope (Fig. 6A). Partial depletion of Treg cells either early (day 14 posttransplant) during the establishment of tolerance or late (day 100 posttransplant), when tolerance has already been established, resulted in prompt islet graft rejection (Fig. 6B). There was a direct correlation between the depletion of CD4⁺Foxp3⁺ Treg cells and the tempo of graft rejection. Mice that rejected islet grafts at a rapid tempo had fewer CD4⁺Foxp3⁺ Treg cells remaining in the periphery as compared with mice that rejected at slower tempo (data not shown). Indeed, one out of five mice that did not reject the allograft had inefficient depletion of CD4⁺Foxp3⁺ Treg cells (29 versus 39–66% depletion in rejecting mice). Treatment with an isotype Ab or NK1.1-depleting Ab did not prompt graft rejection in either setting (Fig. 6B), demonstrating that it is not mere depletion of a lymphocyte population that sets off the rejection reaction. Collectively, these data demonstrate that Treg cells play a critical role in the maintenance of localized tolerance.

Although the partial depletion of Treg cells using anti-CD25 Ab demonstrates the importance of these cells in the induced tolerance, it does not assess if they are the primary mechanism of tolerance. We, therefore, tested the function of these cells in an adoptive transfer allogeneic islet model. Given that the established tolerance was not systemic and localized to the graft, we also tested if there was a functional difference between Treg cells residing in the graft draining LNs and spleens of tolerant mice. Treg cells were sorted from the KDLNs or spleens of long-term (>100 d) SA-FasL–engineered islet graft acceptors, SA-engineered islet graft rejectors, and unmanipulated naive C57BL/6 mice. The sorted cells were then adoptively transferred into chemically diabetic rag2^−/− OT-I mice on C57BL/6 background 1 d prior to BALB/c allogeneic islet transplantation. Animals transplanted with unmodified BALB/c allogeneic islets did not reject their grafts (n = 5; >100 d) due to the specificity of the TCR in OT-I mice for the OVA Ag (Fig. 7). However, adoptive transfer of 1 × 10⁶ CD4⁺CD25⁻ T cells alone from naive C57BL/6 mice resulted in acute rejection of allogeneic islets (n = 5; MST = 29.0 ± 9.2 d). Cotransfer of 2,500–10,000 Treg cells sorted from graft-draining LNs of SA-FasL–islet graft acceptors prevented islet graft rejection induced by 1 × 10⁶ CD4⁺CD25⁻ T cells (n = 6; MST >100 d). In marked contrast, cotransfer of the same number of Treg cells sorted from either SA-islet graft rejectors or naive C57BL/6 mice did not prevent rejection mediated by CD4⁺CD25⁻ T cells (n = 4; MST = 36 ± 2.3 d and n = 5; MST = 34.4 ± 3.5 d, respectively; Fig. 7). Importantly, splenic Treg cells sorted from SA-FasL–islet graft acceptors had significantly reduced efficacy in preventing the rejection of unmanipulated donor islet allografts in the adoptive transfer model as compared with Treg cells sorted from graft-draining LNs. Indeed, only one out of six mice, which received the highest number of Treg cells, in this group accepted the donor graft, whereas all of the rest had rejection, but in a delayed fashion (MST = 48 ± 3.8 d) as compared with controls (Fig. 7). These studies provide direct evidence for the role of Treg cells in the induced tolerance and further demonstrate that graft-protective Treg cells preferentially home to the graft-draining LNs, but not distant lymphoid organs, such as spleen.

It has been shown that FasL can cause tissue destruction by serving as a chemotactic factor for neutrophils (27–29). We, therefore, conducted two sets of studies to directly test if SA-FasL has such a function. In the first set of studies, SA-FasL–engineered syngeneic splenocytes were injected i.p. into mice, and peritoneal exudate cells were harvested at various time points to assess the recruitment of neutrophils using flow cytometry. Splenocytes engineered with SA served as negative control, whereas LPS was used as positive control. We observed about the same percentages (∼12%) of neutrophils among peritoneal exudate cells harvested from mice injected with SA-FasL– and SA-engineered splenocytes (Fig. 8A). In marked contrast, ∼60% of peritoneal exudate cells were neutrophils in the LPS-positive group. In a second set of studies, we assessed the presence of neutrophils among recipient cells infiltrating into SA- and SA-FasL–engineered BALB/c islet grafts transplanted into C57BL/6 mice under the cover of rapamycin. SA-FasL–islet grafts had low levels of neutrophils as compared with SA-engineered islets (Fig. 8B). Taken together, these data demonstrate that SA-FasL displayed on splenocytes or pancreatic islets lacks chemotactic activity for neutrophils.

We demonstrate for the first time in this study, to our knowledge, that pancreatic islets can be engineered ex vivo in a rapid and efficient manner to display on their surface an apoptotic form of FasL protein, SA-FasL, and that such islets induce robust localized, rather than systemic, tolerance when transplanted into fully allogeneic hosts under transient cover of rapamycin without detectable toxicity to the graft or host. Tolerance by this protocol is initiated and sustained by Treg cells, as physical depletion of these cells in vivo early (14 d) or late (>100 d) posttransplantation resulted in acute graft rejection. Furthermore, sorted Treg cells from long-term, but not rejecting, graft recipients prevented the rejection of unmanipulated donor allogeneic islets in an adoptive transfer model, further confirming their role in the induced tolerance.

Tolerance in this model is completely dependent on SA-FasL protein and accentuated by a short course of rapamycin treatment. This observation is consistent with the demonstrated role of Fas-meditated apoptosis in lymphocytes for the regulation of chronic immune responses and control of autoimmunity (9). Upon activation, T cells upregulate both Fas and FasL and become sensitive to autocrine and paracrine apoptosis following repeated engagement with the challenge Ag (9, 30). Therefore, the presentation of alloantigens in the context of FasL may specifically eliminate alloreactive T cells that upregulate the Fas receptor and thereby become sensitive to Fas/FasL-mediated apoptosis. Indeed, analysis of T cells infiltrating into the grafts 3 d posttransplantation revealed higher percentages of CD4⁺ Teff and CD8⁺ Teff cells undergoing apoptosis in SA-FasL–islet grafts as compared with SA-islet grafts (Supplemental Fig. 1). These findings are also consistent with our recent studies demonstrating that systemic immunomodulation with SA-FasL–engineered donor splenocytes induces apoptosis in alloreactive T cells, resulting in the inhibition of primary and secondary alloreactive immune responses (16) and induction of tolerance to cardiac allografts (17). Our findings are also consistent with several studies by others using cells (13), tissues (15), or organs (31) genetically manipulated to express FasL for immunomodulation.

Contradictory data demonstrating that immunomodulation with FasL does not prevent graft rejection have also been reported. For example, islets and heart grafts genetically modified to express FasL lacked protection in allogeneic hosts (11, 32). These negative results have been attributed to the use of a cleavable isoform of FasL and diverse functions associated with the membranous and soluble forms of this molecule. FasL is a type II membranous protein that is converted into a soluble form via cleavage by matrix metalloproteinases in response to various physiologic stimuli (33). The membranous form is noted for its ability to induce apoptosis in autoreactive and alloreactive T cells, thus promoting tolerance. In contrast, the soluble form may inhibit apoptosis, initiate inflammatory responses, and promote the active recruitment of neutrophils, thereby accelerating disease or allograft rejection (29, 34, 35). Although the separation of these distinct functions of the soluble versus membranous forms of FasL has been the source of great controversy, a recent study using transgenic mice expressing either membranous or secreted forms of FasL demonstrated that the membranous form only has apoptotic activity and is responsible for the elimination of self-reactive T cells and prevention of systemic autoimmunity (9). In marked contrast, the soluble form had no apoptotic activity and appeared to promote autoimmunity and tumorogenesis via nonapoptotic functions.

The SA-FasL molecule used in this study lacks potential metalloproteinase sites and as such is not cleaved into a trimeric soluble form, but rather forms stable tetramers or oligomers (16) without any detectable chemotactic activity for neutrophils as demonstrated in this study (Fig. 8). Given that oligomerization of the Fas receptor on the surface of T cells is critical for the delivery of apoptotic signals (36), SA-FasL has potent apoptotic activity in soluble form (16) or when displayed on the cell surface (17). Therefore, SA-FasL–engineered allogeneic cells or tissues have the potential to effectively eliminate alloreactive T cells via apoptosis, thereby initiating a cascade of regulatory events resulting in either systemic tolerance (17) or localized tolerance as shown in this study, with CD4⁺CD25⁺Foxp3⁺ Treg cells serving as the common denominator. Treg cells were shown to be relatively resistant to Fas-induced apoptosis under selected experimental conditions (37) and are spared by rapamycin (22). Consequently, we observed high numbers of Treg cells homing to the islet graft. Physical depletion of these cells either early or late posttransplantation resulted in the abrogation of tolerance. The critical role of these cells in the observed tolerance was further demonstrated in an adoptive transfer model in which Treg cells sorted from graft-draining LNs of tolerant, but not those from naive or SA-islet, graft rejectors prevented the rejection of unmanipulated donor islet grafts. Importantly, splenic Treg cells isolated from tolerant SA-FasL–islet graft recipients were inferior to Treg cells from graft-draining LNs in preventing rejection in the adoptive transfer model. Taken together, these data demonstrate that graft-protective Treg cells preferentially home to graft-draining LNs and potentially to the graft and play a critical role in keeping in check the pathogenic Teff cells. Although we cannot totally exclude other regulatory mechanisms in addition to Treg cells involved in the observed tolerance, our data are consistent with a recent study by Kendal et al. (38), demonstrating that Treg cells play a critical role in tolerance to skin allografts induced by nondepleting anti-T cell Abs by preferentially residing in the graft and keeping in check the pathogenic function of Teff cells.

SA-FasL may contribute to the generation/expansion of Treg cells by two distinct mechanisms: 1) preferential elimination of activated alloreactive Teff cells due to their enhanced sensitivity to FasL-mediated apoptosis and as such tilting the balance toward Treg cells (37); or 2) generation of induced Treg cells through mechanisms involving apoptosis (39). Unlike the human Treg cells that are not only resistant to Fas/FasL-mediated apoptosis, but also use FasL as an effector molecule to induce apoptosis in Teff cells as a means of immune suppression (40, 41), the preferential sensitivity of mouse Treg cells over the Teff cells to Fas/FasL-mediated apoptosis has been the subject of significant controversy. A series of studies in the autoimmunity and cancer settings has demonstrated that Treg cells are more sensitive than Teff cells to Fas/FasL-mediated apoptosis (42–44). For example, immunomodulation with FasL protein was recently reported to selectively deplete Treg cells from tumor (43). Similarly, the efficacy of immunomodulation with IL-12 was shown to be dependent on CD8⁺ T cells expressing FasL and eliminating Treg cells within the tumor via Fas/FasL-mediated apoptosis (42). In marked contrast, freshly isolated CD4⁺CD25⁺ Treg cells were shown to be less sensitive to Fas-mediated apoptosis as compared with CD4⁺CD25⁻ T cells in response to CD3 stimulation in vitro (45). However, this differential sensitivity to Fas/FasL-mediated apoptosis could be reversed in coculture experiments depending on the Treg/Teff cell ratios that were regulated by IL-2. Consistent with this study, we recently demonstrated relative resistance of NOD Treg cells to SA-FasL–mediated apoptosis under inflammatory conditions as compared with Teff cells (37, 46, 47). Indeed, the direct display of SA-FasL on the surface of Treg cells endowed them with better regulatory activity by inducing apoptosis in Teff cells via SA-FasL/Fas interaction (46). In marked contrast, the direct display of SA-FasL on NOD Teff cells resulted in their accelerated apoptosis and reduced onset and incidence of diabetes in an adoptive transfer model (47). Consistent with these studies, we observed higher percentages of CD4⁺ Teff and CD8⁺ Teff cells undergoing apoptosis in SA-FasL–islet grafts as compared with SA-islet group. In marked contrast, CD4⁺ Treg cells appeared to be resistant to SA-FasL–induced apoptosis (Supplemental Fig. 1). Although the molecular nature of these opposing observations is not known and remains to be elucidated, they indicate the complex nature of Fas/FasL-mediated homeostasis of Treg and Teff cells under normal physiological conditions and disease settings.

Alternative and/or additive to the sparing of Treg cells from Fas-mediated death, SA-FasL–engineered islet grafts may facilitate de novo generation of induced Treg cells via a cascade of immunoregulatory mechanism orchestrated by apoptosis of Teff cells. Consistent with notion is a recent study demonstrating that immunomodulation with anti-CD3 Ab results in apoptosis of T cells, digestion of apoptotic bodies by phagocytes, and their secretion of TGF-β that converts Ag-specific Teff cells into Treg cells (39). Importantly, Treg cells harvested from SA-FasL–engineered, but not SA-control, islet graft recipients prevented rejection of unmodified second set donor islet grafts in an adoptive transfer model, suggesting that these cells may represent the induced Treg cells that show Ag specificity.

The tolerogenic effect of SA-FasL was greatly enhanced by a short course of rapamycin treatment. Rapamycin may work in synergy with SA-FasL to enhance tolerance by eliminating Teff cells (20) and contributing to the generation of Treg cells either by sparing natural Treg cells (22) or actively contributing to the conversion of Teff cells into induced Treg cells through the regulation of the mammalian target of rapamycin (23). Important in this context is a recent study demonstrating that in the presence of rapamycin TCR/IL-2 signaling resulted in the upregulation of antiapoptotic Bcl-2 family members in Treg cells and their relative resistance to apoptosis (48). In marked contrast, under the same conditions, rapamycin downregulated the expression of Bcl-2 family members while increasing the expression of proapoptotic Bax in T conventional cells, leading to their increased sensitivity to apoptosis.

Genetic manipulation of pancreatic islets ex vivo to express immunoregulatory molecules for long-term survival in allogeneic hosts represents an attractive immunomodulatory approach that may spare the recipient from harmful immunosuppressive treatments. However, introduction of foreign DNA into the graft is not only technically challenging but also has safety concerns, and continuous expression of potent immunomodulatory molecules, such as FasL, on the graft may have long-term complications. Therefore, our approach of engineering donor grafts ex vivo to display on their surface immunomodulatory molecules, such as SA-FasL, in a rapid, effective, and transient manner represents a novel means of immunomodulation with therapeutic implications in cancer, autoimmune disease, and transplantation.

We thank O. Grimany for technical help with SA-FasL, SA-CD40L, and SA protein expression and purification and Drs. Suzanne T. Ildstad, Michele Kosiewicz, and Huang-Ge Zhang for critical reading of the manuscript.

The SA-FasL protein and ProtEx technology described in this article are licensed from the University of Louisville by ApoVax, Inc., Louisville, KY, for which H.S. serves as a Member of the Board and Chief Scientific Officer, and H.S. and E.S.Y. have significant equity interest in the company. The other authors have no financial conflicts of interest.