CTL-Promoting Effects of IL-21 Counteract Murine Lupus in the Parent→F1 Graft-versus-Host Disease Model Free

Interleukin-21 is a member of the type I cytokine family with pleiotropic effects on the immune system. IL-21R is expressed on a variety of immune cells, including B, T, NK, and dendritic cells, whereas IL-21 production is restricted to activated CD4 T cells, T follicular helper (T_FH) cells, Th17 cells, and NKT cells (1-3). IL-21 promotes the expansion and antitumor activity of NK cells, enhances CD8 T cell maturation into CTLs, and promotes the differentiation and expansion of T_FH cells while downregulating regulatory T cells (2, 4–8). Within the B cell lineage, IL-21 regulates B cell proliferation and survival, Ig production and class switching, germinal center formation, plasma cell differentiation, and memory B cell responses (9–12).

The contribution of IL-21 to a number of autoimmune diseases, including systemic lupus erythematosus, Sjögren’s syndrome, rheumatoid arthritis, experimental allergic autoimmune encephalomyelitis, psoriasis, diabetes mellitus, and inflammatory bowel disease, was reported (13–19). Elevated IL-21 levels were found in lupus-prone BXSB.Yaa mice, and autoimmune features were attenuated in BXSB.Yaa and MRL-Fas^lpr mice following genetic deletion of IL-21R or IL-21 blockade (20–23). In patients with lupus, an association between polymorphisms in the IL-21 and IL-21R gene and disease susceptibility was reported, along with increased circulating and intracellular levels of IL-21 (24-28). Thus, IL-21 is an attractive candidate for therapeutic targeting in autoimmune diseases, including SLE.

Although the CTL-promoting effects of IL-21 are critical for antitumor immunity and viral clearance, the exact contribution of IL-21 activity on cytotoxic CD8 cells in an autoimmune setting is not known. This question is of particular importance because, at least in some animal models of SLE, agents that promote CTLs might be beneficial, whereas ineffective CTLs have a permissive effect and may contribute to disease phenotype as a result of their inability to suppress or eliminate autoreactive B cells (29, 30). Recently, a protective effect of a subset of CD8 T cells with immunosuppressive abilities was demonstrated in BXSB.Yaa mice (22). Furthermore, in vivo IL-21 blockade and selective ablation of IL-21R on CD8 cells were detrimental for survival early in the disease and accelerated morbidity, respectively, by blocking the immunosuppressive effects of CD8 T cells (22, 31). Thus, although IL-21 blockade may attenuate autoimmunity by diminishing the stimulatory effects of IL-21 on B cells or by restraining CD4 T cell help to autoreactive B cells, it can also shift the balance toward disease by inhibiting the CD8-suppressive axis. It is unclear whether blocking or enhancing IL-21R signaling on CD8 T cells in an autoimmune setting can alter humoral autoimmunity by inhibiting or promoting, respectively, the cytotoxic effects of CD8 T cells.

To address this question, we took advantage of the parent-into-F1 (P→F1) model of chronic graft-versus-host disease (cGVHD) and acute GVHD (aGVHD), which allows the study of CTL-promoting or -inhibiting modalities in the setting of autoreactive B cell hyperactivity and autoantibody production. In the C57BL/6 (B6)→F1 aGVHD model, the transfer of normal parental CD4 and CD8 T cells into unirradiated normal F1 mice results in donor CD4 T cell activation and initial help for host B cells that is counteracted by the generation of donor CD8 CTLs that eliminate host splenic lymphocytes and particularly B cells, resulting in a lymphopenic state (32). In the DBA/2J (DBA)→F1 cGVHD model, the transfer of DBA splenocytes into F1 mice results in a lupus-like cGVHD as a result of the combination of donor CD4 T cell allorecognition and help for host B cells and impaired donor CD8 CTL effectors (30). The impairment in host B cell elimination allows host B cell expansion, autoantibody production, and, eventually, a lupus-like renal disease. In this study, we used IL-21R–deficient mice as donors in B6→F1 aGVHD and in vivo IL-21 administration to cGVHD mice to assess whether the CTL-promoting effect of IL-21 would counteract host B cell expansion and autoantibody production. Our results indicate that, in aGVHD, the absence of IL-21R on donor cells allows the persistence of autoreactive B cells by impairing the CTL response, whereas IL-21 administration to cGVHD mice prevents the development of lupus-like disease by rescuing the defective CTL effectors and subsequent host B cell killing that are typical of aGVHD.

Six- to eight-week-old male BDF1 (H-2^b/d), DBA (H-2^d), and B6 (H-2^b) mice were purchased from The Jackson Laboratory (Bar Harbor, ME). Breeding pairs of IL-21R^−/− mice on the B6 background were generated and provided by Dr. Michael Grusby (Harvard School of Public Health, Boston, MA) and then housed and bred at the University of Maryland Animal Care Facility (33). All procedures were approved by the University of Maryland School of Medicine Office of Animal Welfare Assurance.

Single-cell suspensions of splenocytes were prepared in RPMI 1640, filtered through sterile nylon mesh, washed, and diluted to a concentration of 10⁸ viable (trypan blue–excluding) cells/ml. P→F1 aGVHD was induced with 50 × 10⁶ unfractionated splenocytes from B6 wild-type (WT) or B6 IL-21R^−/− donors injected i.v. into 6–8-wk old BDF1 mice, as described (34). In some experiments, the number of donor CD4 and CD8 T celquantified first by flow cytometry, and the donor cell number was adjusted so that the amount of CD8 and CD4 T cells in the IL-21R^−/− donor inoculum was approximately equal to that of WT donor cells. In some experiments, aGVHD was induced using purified donor T cells achieved through negative selection using a Dynal Mouse T Cell Negative Isolation Kit (Invitrogen, Carlsbad, CA). Further depletion of donor CD4 or CD8 T lymphocytes was carried out using Dynabeads coated with anti-mouse CD4 or CD8 Ab, respectively. Purity was confirmed by flow cytometry and was typically >95%. DBA→F1 cGVHD was induced with 80–90 × 10⁶ unfractionated splenocytes from DBA donors injected into BDF1 mice. In some experiments, cGVHD was induced with CD8-depleted DBA donor cells containing 14 × 10⁶ CD4 T cells. Normal F1 and experimental mice received 100 μg recombinant mouse IL-21 i.p. (Zymogenetics, Seattle, WA) at the time of GVHD induction and every other day thereafter. This dose was effective in vivo in models of tumor suppression and in a model of xenogeneic GVHD (35). Control mice consisted of uninjected and age- and sex-matched F1 mice, untreated cGVHD mice, and cGVHD mice receiving PBS. Recipient and donor mice were matched for age and sex within each independent experiment.

Spleen cells were incubated with anti-murine FcγRII/III mAb (2.4G2) for 10 min and then stained with saturating concentrations of Alexa Fluor 488–conjugated, allophycocyanin-conjugated, biotin-conjugated, PE-conjugated, FITC-conjugated, PE/Cy5-conjugated, or PE/Cy7-conjugated mAbs against CD4, CD8, H-2K^b, H-2K^d, I-A^b, I-A^d, CD80, CD86, CD69, CD44, Fas, Fas ligand (FasL), perforin, IFN-γ, TNF-α, IL-17, CXCR5, CD25, annexin V, Ki-67 (all from BD Biosciences, San Jose, CA), B220, killer cell lectin–like receptor G1 (KLRG-1), IL-4 (all from BioLegend, San Diego, CA), Foxp3, and granzyme B (both from eBioscience, San Diego, CA). Biotinylated primary mAbs were detected using streptavidin-allophycocyanin (BioLegend), streptavidin-FITC, streptavidin-PE, or streptavidin-PE-Cy5 (BD Biosciences). Cells were fixed in 1% paraformaldehyde before flow cytometric analysis. Multicolor flow cytometric analyses were performed using an Accuri C6 and LSR II flow cytometer (BD Biosciences). Lymphocytes were gated by forward and side scatter, and fluorescence data were collected for a minimum of 10,000 cells. Studies of donor T cells were performed on a minimum of 5,000 cells collected using a lymphocyte gate that was positive for CD4 or CD8 and negative for MHC class I of the uninjected parent (H-2K^d− for B6→F1 aGVHD and H-2K^b− for DBA→F1 cGVHD). Gated donor T cells were analyzed for CD44, Ki-67, annexin V, FasL, KLRG-1, IFN-γ, TNF-α, IL-4, IL-17, Foxp3, programmed cell death-1 (PD-1), and Fas expression and compared with uninjected strain-matched splenocytes. The percentage positivity for PD-1, Fas, and FasL for the experimental groups was determined using a setting such that 95% of the relevant control cell population (uninjected parental or host strain) was negative. For intracellular IL-4 and IL-17 staining, spleen cells were restimulated with PMA (50 ng/ml) and ionomycin (500 ng/ml) plus Golgi Plug (BD Biosciences; at 37°C) for 4 h.

In vivo cytotoxic activity was determined using CFSE-labeled target cells, as described (34, 36, 37). Briefly, target cells from B6 and DBA parental strain splenocytes were loaded with CFSE at two concentrations as follows: 0.5 μM CFSE^low for control nontarget cells (B6 for aGVHD and DBA for cGVHD) and 5 μM CFSE^high for target cells (DBA for aGVHD and B6 for cGVHD). Cell suspensions were irradiated at 2000 rad, and 5 × 10⁶ cells of each population were mixed together (1:1 ratio) and injected i.v. into control F1 and GVHD mice on day 10. Mice were sacrificed after 5 h, and CFSE-labeled cells were analyzed by flow cytometry. Percentage specific lysis was calculated as follows: for B6→F1 mice percentage of lysis = 100 − ((% DBA CFSE^high experimental/% B6 CFSE^low experimental)/(% DBA CFSE^high in control F1/% B6 CFSE^low in control F1)) × 100). For DBA→F1 mice, percentage of lysis = 100 − ((% B6 CFSE^high experimental/% DBA CFSE^low experimental)/(% B6 CFSE^high in control F1/% DBA CFSE^low in control F1)) × 100).

Mice were bled at the times indicated, and sera were tested by ELISA for the presence of IgG Ab to ssDNA, as described (38).

Total RNA isolation, quantitation, and reverse transcription were performed as described (34). 18S rRNA was used as an internal control. All primers (myxovirus [influenza virus] resistance 1 [Mx-1], IL-12, 18S) were purchased from SABiosciences (Frederick, MD). Real-time PCR was carried out on an Applied Biosystems StepOnePlus PCR machine (Applied Biosystems, Foster City, CA).

Proteinuria was assessed using Siemens Albustix reagent strips every 2 wk up to week 12 after GVHD induction. All measurements were performed in the morning to avoid the effects of diurnal variation. Kidneys were harvested at 12 wk after disease induction. Formalin-fixed kidney sections (4 μm) were stained with H&E. The sections were examined in a blinded fashion for glomerular, tubular, and interstitial pathology. Disease was scored on a semiquantitative scale using published criteria with modifications, as previously reported (38, 39). The severity of glomerulonephritis (GN) was graded on a 0–3 scale, where 0 = normal; 1 = mild to moderate increase in cellularity with mesangial proliferation; 2 = moderate increase in cellularity with endocapillary and mesangial proliferation, increased matrix, and/or karyorrhexis; 3 = marked increase in cellularity with endocapillary proliferation, crescent formation, and/or necrosis, and/or sclerosis. Scores from 20 glomeruli were averaged to obtain a mean score for each kidney section. Tubulointerstitial nephritis (TIN) was scored as 0, normal; 1, focal mild infiltrate of mononuclear cells (MNCs); 2, multifocal, moderate MNC infiltrate and mild and focal tubular damage (dilatation, atrophy, necrosis); and 3, extensive MNC infiltrate, extensive and severe tubular damage, and/or interstitial fibrosis.

Normally distributed data were analyzed by a two-tailed Student t test (for single comparison) or ANOVA (for multiple comparisons) using GraphPad Prism software version 6.0 (GraphPad, La Jolla, CA). Nonparametric data were analyzed by the Wilcoxon rank-sum test.

Because CTLs can limit autoreactive B cell expansion in aGVHD, we asked whether lack of IL-21R signaling on donor cells in this model impairs B cell elimination and shifts the balance toward a lupus-like cGVHD. To this end, we injected F1 mice with splenocytes from WT or IL-21R^−/− B6 donors. Donor splenocytes were adjusted so that all mice received comparable numbers of CD8 cells. Typically, at 14 d after donor cell transfer, B6→F1 aGVHD mice exhibit a cytotoxic phenotype characterized by significant engraftment of both CD4 and CD8 donor cells and profound elimination of host splenocytes mediated by donor CD8 CTLs and no autoantibody production (30). As expected, at 2 wk after disease induction, WT B6→F1 aGVHD mice exhibit a 2.9-fold reduction in total splenocyte numbers compared with normal F1 mice, near-complete elimination of host B cells, engraftment of both CD4 and CD8 donor cells, and no significant elevation of anti-ssDNA Ab (Fig. 1A–C). In contrast, IL-21R^−/−→F1 mice display splenocyte and B cell numbers that remain at levels similar to F1 controls, a trend toward decreased donor CD8 T cell engraftment, and significantly elevated anti-ssDNA autoantibodies.

The impaired host B cell elimination in IL-21R^−/−→F1 GVHD mice suggests a defective donor CD8 anti-host response. To further characterize this defect, we assessed whether the kinetics of donor CD8 T cell engraftment and expansion are altered in the absence of IL-21R signaling. In aGVHD, donor CD8 T cells undergo three phases: activation and expansion of naive cells (days 0–7), followed by expansion of mature effector CTLs (days 7–10) and downregulation and contraction of effector T cells (days 10–14). The kinetics of donor CD8 expansion and contraction progressed similarly in IL-21R–sufficient and -deficient donor CD8 cells, with the peak effector number reached on day 10, followed by their contraction by day 14 in both groups (Fig. 1D). However, significant differences were noted in the magnitude of the expansion of WT and IL-21R^−/− donor CD8 cells. Naive CD8 cells from IL-21R–deficient or -sufficient donors expanded to a similar extent from days 0 to 7. WT and IL-21R^−/− donor CD8 T cells also displayed a similarly activated CD44^high phenotype (data not shown), suggesting that the initial activation and expansion of donor CD8 cells are not altered in the absence of IL-21R signaling. However, further expansion of mature effector CTLs from days 7 to 10 was markedly decreased in IL-21R^−/−-injected F1 mice and reached a significantly lower peak number of donor CD8 cells (5.2 × 10⁶ ± 0.7 versus 2.3 × 10⁶ ± 0.8, p < 0.05). The decreased peak clonal expansion of IL-21R^−/− donor CD8 T cells was due to decreased proliferation, as demonstrated by the significantly decreased proportion of proliferating Ki-67⁺ cells at the peak of their expansion on day 10 (Fig. 1E). After reaching peak levels, both IL-21R–sufficient and -deficient donor CD8 cells underwent contraction and decreased by 42 ± 5% and 34 ± 4%, respectively. Consistent with the less robust contraction of IL-21R^−/− donor CD8 cells, the proportion of apoptotic donor CD8 cells, as assessed on day 10 by annexin V staining, was significantly lower in IL-21R^−/− donor CD8 cells (Fig. 1F). Furthermore, the number of donor CD8 cells expressing Fas and PD-1, effector molecules important for the homeostatic contraction and downregulation of CD8 cells, was significantly lower for IL-21R–deficient donor CD8 cells (Fig. 1G). In contrast to donor CD8 cells, the engraftment and expansion of donor CD4 cells did not differ between the two groups at any time point (Fig. 1H).

Overall, these results indicate that IL-21R^−/− donor CD8 cells exhibit decreased proliferation and expansion to peak levels and impaired elimination of host B cells, resulting in an intermediate, low-intensity cGVHD phenotype.

To address whether the defective elimination of host B cells in IL-21R^−/−→F1 mice is due only to the observed quantitative defect in peak donor CD8 T cell number or is also indicative of a qualitative defect, we first assessed the effect of increased IL-21R^−/− donor cell inoculum on aGVHD parameters. The injection of greater numbers of IL-21R^−/− donor cells resulted in increased engraftment of donor CD8 T cells, from 1.3 × 10⁶ ± 0.38 in mice receiving 50 × 10⁶ IL-21R^−/− donor cells to 2.1 × 10⁶ ± 0.7 and 3.99 × 10⁶ ± 1.2 in mice receiving 75 × 10⁶ and 100 × 10⁶ donor cells, respectively. However, despite the greater number of IL-21R^−/− donor CD8 T cells engrafted, elimination of host B cells remained significantly impaired; even the injection of 100 × 10⁶ IL-21R^−/− donor cells failed to decrease host splenic B cells to the levels observed in WT aGVHD (host B cell number = 41 × 10⁶ ± 3.6 in IL-21R^−/−→F1 versus 1.2 × 10⁶ ± 0.4 in WT→F1, p = 0.001) (Fig. 1I). Thus, transferring donor cell numbers sufficient to engraft IL-21R^−/− donor CD8 cells in numbers equal to or even higher than those of WT donor cells does not correct the defect in host splenocyte elimination. These results support the idea that, in addition to quantitative defects, the absence of IL-21R signaling in donor CD8 T cell results in qualitative defects in effector CTL function.

To confirm a qualitative defect in IL-21R–deficient donor CD8 cells, we assessed effector CTL activity in vivo using CFSE-loaded target cells of B6 donor (CFSE^low) and host DBA (CFSE^high) origin injected i.v. in equal amounts into aGVHD mice on day 10; the percentage of donor anti-host lysis was determined 5 h later. IL-21R^−/−→F1 aGVHD mice exhibit a significant reduction in the ability to lyse DBA target cells compared with WT→F1 mice (51.2 ± 3.4% versus 71 ± 4.7%, p < 0.05), as demonstrated by the higher percentage of remaining CFSE^high target cells (Fig. 2). These results indicate that the cytolytic effector function is impaired in the absence of IL-21R signaling and that the difference in the donor CD8 expansion cannot account for the difference in CTL activity.

To determine whether the impaired expansion and cytolytic function of IL-21R^−/− donor CD8 cells reflect differential upregulation of the signal 3 cytokines IFN-α and IL-12, we measured the mRNA expression of IL-12 and the IFN-α inducible gene, Mx-1. Both WT→F1 and IL-21R^−/−→F1 aGVHD mice exhibit a significant increase in Mx-1 and IL-12 gene expression at 7 d after GVHD induction. Upregulation of IL-12, but not Mx-1, gene expression is sustained at days 10 and 14 of disease (Supplemental Fig. 1). We did not detect a significant difference in Mx-1 or IL-12 gene expression between IL-21R^−/−→F1 and WT→F1 mice at any time point, demonstrating that the impaired expansion and cytolytic function of IL-21R^−/− donor CD8 cells are not due to a defect in signal 3 cytokines.

Host B cell elimination in aGVHD is mediated primarily by the granzyme B/perforin and Fas/FasL pathways (40, 41). The impaired CTL activity of IL-21R^−/− donor CD8 T cells could be due to a defect in one or both killing pathways. Using flow cytometry analysis at 10 d after parental cell transfer, IL-21R^−/− donor CD8 T cells exhibit a significant decrease in the percentage and absolute numbers of granzyme B⁺ cells versus WT donor CD8 T cells (Fig. 3A, 3B). Similarly, the absolute number of perforin-expressing donor CD8 T cells was lower among IL-21R–deficient donor CD8 T cells, although the percentage did not differ from WT donor CD8 T cells (Fig. 3C, 3D). The absence of IL-21R also impaired FasL upregulation on day 10 after disease induction, as demonstrated by the lower percentage and absolute number of FasL⁺ donor CD8 T cells (Fig. 3E, 3F). The number of host B cells that upregulated Fas on day 7 or 10 did not differ significantly in WT→F1 versus IL-21R^−/−→F1 mice (data not shown). These data indicate that the donor anti-host CTL effectors that arise in IL-21R^−/−→F1 mice are impaired in the granzyme B/perforin and Fas/FasL pathways.

Interestingly, in contrast to granzyme B and perforin, other markers of effector CD8 T cells were not altered in the absence of IL-21R. The proportion of IFNγ⁺ donor CD8 T cells was comparable in WT and IL-21R^−/− injected mice, and it peaked on day 7 in both groups (Supplemental Fig. 2A). Similarly, the expression of TNF-α and KLRG-1 did not differ significantly between the two groups (Supplemental Fig. 2B, 2C).

Lack of IL-21R signaling on donor T cells may contribute to the impaired CTL response through defective IL-21R signaling in CD8 T cells alone, CD4 T cells alone, or both. To assess whether IL-21 responsiveness in CD8 cells is sufficient for the generation of effective CTLs or whether IL-21R signaling in donor CD4 T cells is also required, we transferred purified CD4 (10 × 10⁶) and CD8 (5 × 10⁶) T cells from WT and IL-21R^−/− donors into F1 mice in all four possible matched or mixed combinations and assessed disease phenotype on day 14 using host B cell elimination as a surrogate marker of donor anti-host CTL function. Uninjected and F1 mice injected with CD8-depleted WT or IL-21R^−/− donor cells containing 10 × 10⁶ CD4 T cells were used as normal and cGVHD controls, respectively. Mice that received WT CD8 T cells either matched with WT CD4 or mixed with IL-21R^−/− CD4 T cells display marked elimination of host B cells, whereas mice that received IL-21R^−/− CD8 T cells in matched or mixed combination with CD4 T cells exhibit decreased B cell elimination, as demonstrated by B cell numbers similar to the uninjected host (Fig. 4A). Consistent with the number of host B cells, anti-ssDNA Ab production increased during the first week in mice injected with WT CD8 T cells, irrespective of the presence or absence of IL-21R on donor CD4 T cells, but it decreased to baseline levels by day 14 as a result of B cell elimination (Fig. 4B). In contrast, in mice injected with combinations of IL-21R^−/− CD8 T cells with WT or IL-21R^−/− CD4 T cells, the levels of anti-ssDNA Ab remained elevated through day 14, in keeping with the persistence of host B cells (Fig. 5C). These results demonstrate that CD8, but not CD4, T cell responsiveness to IL-21 is critical for optimal CTL generation and the aGVHD phenotype. However, even when paired with high numbers of WT CD4 cells, IL-21R^−/− CD8 T cells induced an attenuated phenotype compared with WT CD4 →F1 cGVHD mice, as demonstrated by the decreased host B cell expansion (Fig. 4A, columns 5 and 6) and anti-ssDNA Ab production (Fig. 4C), suggesting a partial, but not complete, failure to eliminate host B cells. Thus, IL-21/IL-21R interaction on CD8 T cells is critical, but not exclusive, for full CD8 CTL effector maturation and suppression of cGVHD.

In agreement with our previous study, in the absence of CD8 T cells in the donor inoculum, IL-21R^−/− donor CD4 cells induced an attenuated cGVHD phenotype (Fig. 4A, columns 6 and 7, 4C) compared with WT CD4 donor cells (38). These data suggest that IL-21R^−/− donor CD4 cells are impaired with regard to providing help to autoreactive B cells but not with regard to WT CD8 differentiation and maturation into CTLs.

As the result of defective in vivo CTL effector development, the DBA→F1 model of cGVHD is characterized by a phenotype of lupus-like cGVHD: predominant engraftment of donor CD4 T cells, impaired expansion of donor CD8 T cells, host B cell proliferation, and autoantibody production. The observation that IL-21R deficiency decreases donor anti-host CTL function and converts aGVHD into a low-intensity chronic phenotype raises the possibility that, in the DBA→F1 model, IL-21 administration could convert the phenotype from cGVHD to aGVHD. To evaluate this possibility, DBA→F1 mice received 100 μg of recombinant mouse IL-21 or PBS every 2 d beginning at the time of donor cell transfer. Untreated and PBS-treated DBA→F1 control mice exhibit typical features of cGVHD: >2-fold increase in total spleen cells and host B cells compared with uninjected F1 mice (Fig. 5A), engraftment of predominantly donor CD4 T cells (Fig. 5B), 7-fold elevation in serum anti-ssDNA Ab compared with normal F1 mice (Fig. 5C), and no IFN-γ production (Fig. 5D–F). IL-21 administration to DBA→F1 cGVHD mice significantly altered the cGVHD phenotype, resulting in a significant reduction in total splenocytes and elimination of host B cells (Fig. 5A), engraftment of CD8 donor T cells (Fig. 5B), a significant reduction in serum anti-ssDNA Ab (Fig. 5C), and a significant increase in the proportion of IFNγ⁺ donor CD8 T cells, amount of IFN-γ/cell, and amount of IFN-γ transcripts (Fig. 5D–F). Mimicking parameters of aGVHD induced with WT donor cells, IL-21 administration increased the proportion of proliferating Ki-67⁺ donor CD8 T cells, their sensitivity to apoptosis, and expression of Fas and PD-1 (Fig. 6I). Interestingly, the reduction in anti-ssDNA Ab levels in IL-21–treated mice occurred despite evidence of enhanced IL-21–mediated B cell help, as demonstrated by increased MHC class II expression on host B cells and an increased number of CD4⁺ PD1⁺ CXCR5⁺ donor T_FH cells (Fig. 5J, 5K). F1 mice receiving IL-21 without donor splenocytes did not exhibit alterations in total splenocyte, B cell, or CD4 or CD8 T cell numbers (Fig. 5A, data not shown).

To confirm that conversion of cGVHD to an acute phenotype following IL-21 administration is CD8 dependent, cGVHD was induced with CD8-depleted DBA donor cells. In the absence of donor CD8 T cells, untreated and IL-21–treated GVHD mice displayed total splenocyte numbers, host B cell expansion, and anti-ssDNA Ab levels typical of cGVHD (Fig. 5A, 5C).

Among CD4 T cell subsets, the frequency of IL-4⁺ and IL-17⁺, but not IFN-γ⁺, donor CD4 T cells was increased by IL-21 administration, although only the increase in IL-4⁺ CD4 cells reached statistical significance (Fig. 5L). The proportion of donor Foxp3⁺ CD25⁺ regulatory T cells was not altered by IL-21 administration.

Next, we assessed whether, similar to aGVHD, elimination of host B cells in cGVHD mice treated with IL-21 results in quantitative changes in the engraftment and expansion of donor CD8 T cells, as well as qualitative effects on CTL effector maturation. In vivo killing of B6 target cells was significantly higher in IL-21–treated DBA→F1 mice (55 ± 4%) than in untreated or PBS-treated control DBA→F1 mice (22 ± 9% and 28 ± 10%, respectively), as demonstrated by the higher percentage of remaining CFSE^high B6 target cells (Fig. 6A, 6B). Furthermore, the expression of CTL effector molecules, granzyme B, and perforin in donor CD8 T cells was also upregulated in response to IL-21, as shown by the increased proportion and absolute number of granzyme B–expressing (Fig. 6C–E) and perforin-expressing (Fig. 6F–H) donor CD8 T cells. Similarly, FasL expression on donor CD8 T cells (Fig. 6I, 6J) and on host B cells (Fig. 6K, 6L) was upregulated significantly after IL-21 administration. Among other markers of CTL effector maturation, expression of TNF-α, but not KLRG-1, was increased in IL-21–treated mice (Fig. 7). Taken together, these data suggest that IL-21 administration to DBA→F1 cGVHD mice rescues defective donor CD8 cells and converts the phenotype from chronic to acute by promoting the proliferation, expansion, and maturation of donor CD8 T cells into effective CTLs.

Next, we determined whether the rescue of the DBA donor anti-host CTL activity requires IL-21 administration only during the first week after disease induction when the initial activation and expansion of naive CD8 cells occur or IL-21 is also required during the maturation phase of CTLs that occurs after day 7. For this purpose, mice were treated with IL-21 from days 0 to 7 and compared with cGVHD mice that received IL-21 from days 0 to 14. cGVHD mice receiving IL-21 from days 0 to 7 exhibit a statistically significant increase in donor CD8 cell engraftment compared with DBA→F1 control mice that received PBS. The increase in donor CD8 engraftment was even more pronounced in mice that received IL-21 from days 0 to 14, although the increment was not statistically significant (Fig. 8A). Furthermore, mice receiving IL-21 during the first week displayed a significant reduction in host total spleen and B cell numbers that was further reduced in mice receiving IL-21 from days 0 to 14 (Fig. 8B). Compared with PBS-treated GVHD mice, anti-ssDNA Ab levels were significantly lower in GVHD mice that received IL-21 from days 0 to 14 but not in those treated only from days 0 to 7 (Fig. 8C). These results indicate that, although the presence of IL-21 signaling during the initial activation, expansion, and differentiation of naive CD8 cells was sufficient for generating effective CTLs, an optimal CTL response requires IL-21 administration during the expansion and maintenance of mature effector CTLs. Delayed administration of IL-21 from days 8 to 14 failed to alter cGVHD phenotype, because donor CD8 engraftment, host B cell number, and autoantibody production were comparable to those in PBS-injected cGVHD mice (data not shown).

Long-term features of cGVHD in the P→F1 model are lupus-like autoantibody and immune complex GN (30). To determine whether the elimination of host B cells seen at early time points in IL-21–treated DBA→F1 mice results in long-term improvement of autoantibody levels and lupus-like GN and, conversely, whether the low-intensity cGVHD features noted in aGVHD induced with IL-21R^−/− donor cells are associated with lupus-like GN, we assessed the kinetics of anti-ssDNA Ab levels and parameters of renal disease by proteinuria and kidney histology.

As shown in Fig. 9, untreated and PBS-treated control DBA→F1 mice exhibit a significant increase in serum anti-ssDNA Ab levels that peak at 6 wk after disease induction and remain elevated, although they gradually decline through week 12 (Fig. 9A). Proteinuria is detected starting at week 4 (data not shown) and peaks at >3⁺ at week 10 (Fig. 8B). As expected, at week 12, untreated (data not shown) or PBS-treated cGVHD mice exhibit typical features of GN, as demonstrated by glomerular enlargement and mesangial and endocapillary hypercellularity with occasional crescent formation (Fig. 9J). In contrast, cGVHD mice that received IL-21 display significantly lower, but sustained, levels of anti-ssDNA Abs (Fig. 9A), significantly lower proteinuria and glomerular score (Fig. 9B, 9C), much milder features of GN (Fig. 9K), and a trend toward a lower TIN score (Fig. 9D).

Consistent with the low-intensity cGVHD observed at 2 wk in mice with IL-21R^−/− →F1 aGVHD, serum anti-ssDNA Ab levels in IL-21R^−/− injected aGVHD mice peak at 2 wk and then slowly decline to reach levels comparable to WT aGVHD mice by week 10 (Fig. 9E). In accordance with the lower autoantibody titers, mice displayed slightly elevated glomerular score and >2⁺ proteinuria at 8–12 wk and features of mild GN and TIN at 12 wk (Fig. 9F–H, 9L). No evidence of lupus-like glomerular disease or proteinuria was seen in WT-injected aGVHD mice at any time point (data not shown). Taken together, these data demonstrate that administration of IL-21 to DBA→F1 mice results in long-term attenuation of autoantibody production and renal disease, whereas aGVHD induced with IL-21R^−/− donor cells exhibits a disease phenotype of low-intensity cGVHD with transient autoantibody production and mild GN.

The stimulatory effect on T_FH cells, B cells, and CTLs supports a critical role for IL-21 in the development of autoimmune diseases on one hand and tumor rejection and viral clearance on the other hand (13). Using the P→F1 model of acute and chronic GVHD, we assessed whether the CTL-promoting effects of IL-21 in the setting of ongoing B cell activation and humoral autoimmunity could limit autoreactive B cell expansion, and, conversely, whether the lack of IL-21R signaling impairs CTL-mediated elimination of autoreactive B cells. Our data reveal that lack of IL-21R signaling on donor T cells in aGVHD results in a low-intensity cGVHD due to impaired expansion and functional maturation of donor anti-host CD8 CTL effectors, as shown by decreased peak donor CD8 T cell numbers, decreased upregulation of FasL and impaired granzyme B and perforin expression in donor CD8 T cells, and reduced in vivo killing of host target cells resulting in the persistence of activated host B cells and autoantibody production. Complementing these findings, IL-21 administration to DBA→F1 cGVHD mice rescued the defective donor CD8 T cells, leading to the expansion and maturation of effective CTLs and conversion to an acute phenotype, as demonstrated by the elimination of autoreactive B cells and a decrease in anti-ssDNA Ab production; engraftment of donor CD8 T cells; increased production of IFN-γ; upregulation of donor CD8 FasL, granzyme B, and perforin expression; and increased in vivo killing of host target cells.

Depending on the context, IL-21 can influence multiple aspects of CD8 development and function, such as the primary activation and expansion of naive CD8 T cells, their survival, and/or the functional differentiation into effector or memory cells (42–46). In both GVHD models, IL-21R signaling was required to regulate the magnitude of donor CD8 T cell expansion and the effector function of donor anti-host CTLs. The enhanced expansion of IL-21R–sufficient or IL-21–treated donor CD8 T cells in aGVHD and cGVHD, respectively, was due primarily to enhanced proliferation that leads to significantly higher peak donor CD8 T numbers, despite a simultaneous increase in apoptosis. The observation that IL-21R signaling exerts opposite effects on donor CD8 T cells, as demonstrated by the increase in proliferation and apoptosis, was reported previously in vitro (47). In that study, the proapoptotic effect prevailed over the proliferative one and prevented the expansion of CD8 T cells; in our in vivo model, the apoptotic effect of IL-21 was offset by the increased proliferation, ultimately resulting in the expansion of donor CD8 T cells. The weaker proliferative response of IL-21R^−/− donor CD8 T cells also elicited a diminished downregulatory response, as demonstrated by the decreased expression of PD-1 and Fas.

Nearly all of the CTL activity exerted in vivo in aGVHD can be accounted for by the granzyme B/perforin and Fas/FasL pathways (40). Our data indicate that both pathways are impaired in IL-21R^−/−→F1 aGVHD mice and enhanced in IL-21–treated cGVHD mice. The accumulation of granzyme B and perforin in CD8 T cells is a widely recognized effect of IL-21 (48, 49). The low-intensity cGVHD phenotype observed in IL-21R^−/−→F1 aGVHD mice is similar to that observed following the transfer of donor T cells defective in perforin in the same model (29). Interestingly, our data also show that the lack of IL-21R was associated with a decrease in the proportion and absolute number of FasL⁺ donor CD8 cells, suggesting that lack of IL-21R may also affect FasL regulation and FasL-mediated CTL activity. Although previous reports suggested a role for IL-21–dependent CTL effector pathways in addition to granzyme B/perforin, the specific involvement of the Fas/FasL pathway has not been described (50). Previous studies in the aGVHD model showed that the complete loss of either pathway can tip the balance toward expansion of autoreactive B cells and autoimmunity (29, 51). Therefore, a partial inhibition of both granzyme B/perforin and Fas/FasL-mediated CTL activities in the absence of IL-21R is likely to result in a significant loss of CTL function.

The GVHD developing in the absence of IL-21R on donor cells has an intermediate phenotype between aGVHD and cGVHD, with host B cells at near-normal levels and more transient autoantibody production. The ensuing low-intensity cGVHD phenotype is in keeping with the lower number of donor CD4 T cells (∼7.5 × 10⁶) present in the unfractionated donor inoculum in aGVHD versus 10–14 × 10⁶ donor CD4 T cells in the inoculum for cGVHD and the lack of IL-21R signaling in donor CD4 T cells leading to suboptimal B cell help, decreased expansion in host B cells, and decreased autoantibody production (38, 41). However, even when coinjected with a WT donor CD4 inoculum equal to the threshold for GVHD induction in the mix-and-match experiments, IL-21R^−/− donor CD8 cells induced an intermediate phenotype, suggesting that other IL-21–independent pathways may contribute to CD8 CTL effector maturation and suppression of cGVHD. However, despite a less robust phenotype, the long-term assessment showed development of a mild lupus nephritis, confirming the cGVHD phenotype.

Experiments mixing CD4 and CD8 T cells from WT and IL-21R^−/− donor mice in aGVHD indicate that IL-21R signaling by CD8 T cells alone was required for the differentiation and function of mature CTLs. Specifically, only injection of IL-21R^−/− donor CD8 cells resulted in defective elimination of host cells and a low-intensity cGVHD phenotype, regardless of whether they were paired with WT or IL-21R^−/− CD4 T cells. The observation that IL-21R signaling in CD4 T cells is not required for optimal CTL maturation is similar to reports demonstrating that IL-21 responsiveness in CD8 T cells alone was sufficient for allograft rejection and antitumor or antiviral responses (43, 50, 52).

There are conflicting reports regarding the effect of IL-21 on IFN-γ production in CD8 T cells and other cells (4, 42, 44, 53). Acquisition of IFN-γ–producing capacity by donor CD8 cells in aGVHD is a distinguishing feature from the cGVHD model in which IFN-γ production is impaired (54). Consistent with this requirement, the acute phenotype following the treatment of DBA→F1 cGVHD mice with IL-21 was accompanied by a significant increase in the proportion of IFNγ⁺ donor CD8 cells and the amount of IFN-γ produced per cell. TNF-α followed a similar trend, in keeping with previous reports that IFN-γ modulates TNF-α production in aGVHD (55). In contrast, the donor CD8 IFN-γ response was preserved in aGVHD, regardless of whether IL-21R was intact, likely as the result of the strain-related robust response of B6 donor CD8 cells in the initial 2–4 d, which determines the subsequent IFN-γ production (56, 57). Nevertheless, in the absence of IL-21R on donor CD8 cells, the CTL maturation phase that followed the initial IFN-γ production was impaired. In contrast to CD8 T cells, IL-21 administration to cGVHD mice did not alter the proportion of IFNγ⁺ donor CD4 cells. However, it increased the frequency of T_FH and Th2 cells, likely contributing to the low, but sustained production, of anti-ssDNA Abs from residual host B cells.

A number of studies indicated defective in vitro CTL function in murine and human lupus (58–61). Effective CTLs are particularly important in lupus patients in the setting of a concomitant viral infection. A large body of evidence suggests that certain infections, particularly viral in nature, might participate in disease initiation, disease flare, or worsening of active lupus. Neutralization of IL-21 in this setting could result in the reduction of pathogen-specific CD8 T cells and a delay in viral clearance. In addition, because CD8 CTLs appearing during a flare could also eliminate autoreactive B cells, thereby preventing them from responding to a potentially increased autoantigen burden, neutralization of IL-21 may further delay disease remission. Therefore, the safety of IL-21 neutralization in lupus, especially in the setting of concomitant viral conditions, should be evaluated carefully.

Our findings demonstrating that IL-21–mediated enhancement of CD8 cytotoxic function attenuates autoimmune parameters in an alloantigen-induced GVHD model by eliminating autoreactive B cells through granzyme B/perforin and Fas/FasL pathways complement other reports in the BXSB.Yaa spontaneous model of lupus demonstrating the role of IL-21 in mediating the immunosuppressive effects of CD8 T cells (31). Although the exact nature and mechanism of action of these CD8 suppressor T cells remain to be characterized further, these data emphasize the significant role of the CD8 axis in a model of CD4 T cell–driven B cell disease and raise the question whether IL-21 blockade weakens this suppressive axis and shifts the balance toward disease.

Taken together, our results in the P→F1 GVHD model indicate that IL-21R signaling on CD8 T cells opposes the stimulatory effect of IL-21 on B cells and attenuates the lupus-like phenotype by promoting CTL activity and elimination of activated, autoreactive B cells. The effect of IL-21 blockade in autoimmunity may depend on the relative contribution of these effects to disease phenotype.

We thank Dr. Michael Grusby for initially providing the IL-21R–deficient mice, Stacey Dillon (Zymogenetics) for providing the recombinant mouse IL-21, and Dr. Charles S. Via (Uniformed Services University of the Health Sciences, Bethesda, MD) for helpful discussions.

The authors have no financial conflicts of interest.