A repertoire-independent and cell-intrinsic defect in murine GVHD induction by effector memory T cells

In allogeneic stem cell transplantation (alloSCT), a potentially curative therapy for malignant and nonmalignant disorders of hematopoiesis, donor T cells can attack recipient tissues, resulting in GVHD.¹  Removal of allograft T cells prevents GVHD, but abrogates T cell–mediated immune reconstitution and graft-vs-leukemia (GVL). The goal of alloSCT research is to limit GVHD without sacrificing the positive effects of donor T cells. In a CD4-mediated mouse model of GVHD, we showed that naive phenotype T cells (T_N) cause GVHD, while effector memory phenotype cells (T_EM) do not.²  Importantly, despite their inability to cause GVHD, T_EM engrafted, responded to antigen and mediated GVL.³  Several labs have reported similar findings for both CD4 and CD8 T cells using either major- or minor-histocompatibility antigen (miHAg) disparate systems.^4-9  Thus, the failure of T_EM to cause GVHD is a generalizable phenomenon. However, the mechanisms that limit the ability of T_EM to cause GVHD are unknown. A better understanding of these mechanisms could lead to new targets for preventing and treating GVHD induced by T_N. Further, understanding how and why T_EM, T_N, and T_CM differ is important for developing such cells as therapeutics for promoting GVL and immune reconstitution with a lower risk for GVHD.

We considered 3 main hypotheses to explain why T_EM fail to cause GVHD. First, there could be differences in migration and priming, because T_EM do not express CD62L and CCR7, and they are therefore relatively excluded from lymph nodes,^10,11  which have been proposed as key sites for alloreactive T-cell activation. Second, relative to T_N, T_EM may have a TCR repertoire with fewer alloreactive specificities.¹²  Third, T_EM may be inherently limited in their ability to induce GVHD because of reduced clonal expansion or less pathogenic effector functions. Studies from Beilhack and our group indicated that trafficking alone is not responsible for the inability of T_EM to cause GVHD.^13,14  However, the roles of TCR repertoire and inherent limitations in T_EM are currently unresolved.⁹ 

It has been difficult to distinguish the effects of repertoire and intrinsic functional properties of T_EM in GVHD models. In polyclonal models of GVHD, the specificities of the disease-causing T cells are unknown and/or cannot be measured, precluding a precise estimation of the frequency of such cells. Tracking alloreactive and pathogenic T cells by expression of activation markers is also not feasible, as essentially all donor T cells after transplant express them.¹⁵ 

A complementary approach to address the issue of repertoire, which is not limited by these technical barriers, would be to compare the GVHD-inducing ability of T_N and T_M with identical specificities, using TCR Tg model systems. Indeed, there have been several reports of GVHD models in which disease is mediated by TCR Tg T cells of a single specificity. These include the CD8-mediated, major MHC-mismatched, 2C Tg model (L^d-specific)^16-18 ; the CD4-mediated, MHC-mismatched 3BBM74 (I-A^bm12-specific)¹⁹  and D10 (I-A^b-specific) models²⁰ ; and the CD4-mediated miHAg-mismatched TEα model.^20-22  These models have improved our understanding of GVHD but they each have limitations. Disease was mild in the MHC-mismatched models and required very large numbers of T cells to induce. In the TEα model, the alloantigen is highly expressed and restricted to the host MHCII⁺ hematopoietic compartment. This model causes only moderate disease in a sublethal irradiation model even with high doses of T cells (4 × 10⁶), and the recipients develop only some of the features of GVHD.^20-22  Thus, despite the great utility of TCR Tg models in other scenarios of T cell–mediated immunopathology, such as diabetes or EAE, use of TCR Tg models of GVHD has been relatively limited, to date.

Recognizing the potential utility of this approach to address the question of repertoire versus cell-intrinsic differences in the GVHD-inducing potential of T_N vs T_M, we sought to generate a TCR Tg model: (1) that reproduces the clinical and histopathologic GVHD induced by polyclonal cells; (2) that is induced by the transfer of a smaller numbers of cells, better modeling the number of alloreactive T cells transferred in polyclonal systems; and (3) in which hematopoietic and nonhematopoietic cells express the target antigen, also paralleling the typical clinical GVHD situation. We chose to transfer CD4⁺ TS1 TCR Tg T cells,²³  which recognize the S1 epitope of influenza hemagglutinin (HA), into irradiated recipients in which HA is expressed at low levels in all cells by virtue of the HA104 Tg (HA104 mice²⁴ ). We found that the transfer of small numbers of TS1 T_N into irradiated HA104 Tg recipients precipitated the major clinical and histopathologic features of GVHD.

An additional advantage of this model is that the TS1 Tg mouse has been used extensively to study T cell memory.^25-34  Protocols for the generation of the required numbers of T_M are available based on in vitro priming followed by in vivo resting.^{25,26,30,33,34}  Similar systems have been used to generate T_M by others in other TCR Tg systems.^35-37  This method is advantageous, because T_M generated by infection or immunization with killed virus in vivo are very few and their frequencies decay with time (Hataye et al³⁸  and K.W.J., W.D.S., and M.J.S., unpublished observations, June 2009). This limitation of cell number precluded the use of this approach in our studies,which required purification and transfer to many simultaneous recipients. Importantly, T_M produced via the method we used have authentic memory cell properties, gene expression and behavior.^25-37,39  Such cells function similarly to memory T cells generated by influenza infection in vivo, including providing protective immunity.^29,30,33,34 

We compared the GVHD-inducing ability of TS1 T_N and T_EM and found an inherent defect in the ability of T_EM to cause GVHD. We were able to trace this defect to differential accumulation of T_N and T_EM in the colon, but, surprisingly, not in spleen, as well as to a decrease in cytokine production at both sites. We also used the system to isolate T_CM that were definitively Ag-experienced. This allowed us to address whether T_CM can cause GVHD, a point that is currently controversial,^5,40  showing clearly that they do cause GVHD. Taken together, these experiments help explain the intriguing finding that T_EM are incompetent at causing disease. They also have important implications for the use of T_M in the alloSCT setting in patients.

Mice were from the following sources: BALB/c (National Cancer Institute or The Jackson Laboratory), BALB/c RAG2^−/− (Taconic), BALB/c TS1 and B10.D2 TS1 (Adam Adler, University of Connecticut), and BALB/c Thy1.1 (The Jackson Laboratory). BALB/c TS1 RAG2^−/−, B6.C TS1 RAG1^−/− and BALB/c TS1 Thy1.1 RAG2^−/− were generated at Yale (see supplemental Methods, available on the Blood Web site; see the Supplemental Materials link at the top of the online article). Hosts were HA104 heterozygotes²⁴  except in Figure 1A and B in which homozygotes were used.

Seven to 12-week-old hosts received 750 cGy and BALB/c HA-negative RAG^−/− BM alone or in combination with sorted TS1 T_N, T_CM, or T_EM. Recipients were weighed 2-3 times per week. Mice that lost 30% of their body weight were killed and scored as dead; their last recorded weight remained in the dataset.

Slides were prepared and scored as described previously.^2,41 

Antibodies were from commercial sources or lab-prepared as described in supplemental Methods. Surface and intracellular cytokine staining and dead cell exclusion were as described³  and as in supplemental Methods. For BrdU analysis, mice were injected with 3 mg of BrdU (Sigma-Aldrich) 30 minutes before sacrifice. BrdU staining and apoptosis assays were performed as in supplemental Methods. Colon and hepatic cells were isolated as described in supplemental Methods.

TS1 CD4 T cells were enriched to > 75% purity from splenocytes using a CD4⁺ T cell–enrichment kit (StemCell Technologies). TS1 cells (10⁶ cells/mL) were cultured with HA S1 peptide 110-119 (SFERFEIFPK; 10ug/mL; Keck Facility, Yale University) and T cell–depleted, irradiated splenocytes (3 × 10⁶ cells/mL) for 3 days.^25,26  BALB/c or B6.C TS1 cells (5 × 10⁶) were transferred intravenously into each RAG^−/− BALB/c or B6.C recipient, respectively.

TS1 were sort-purified by excluding cells expressing F4/80, CD11b, CD8α, Gr-1, CD25, and Ter119. T_N (CD62L⁺ CD44^dim) were sorted from unmanipulated TS1 mice and T_CM (CD62L⁺ CD44⁺) and T_EM (CD62L⁻ CD44⁺) from “memory” mice. Sorted cells were 40%-50% CD4⁺TS1⁺. Purities of T_N, T_EM, and T_CM among TS1 were > 99%.

The significance of comparisons of weight change and the percentage of BrdU⁺ TS1 in colon were calculated by Student t test. The significance of comparisons of histopathology scores, numbers of TS1 cells, and percentages of TS1 that were BrdU⁺ (in spleen), cytokine⁺, or apoptotic were calculated by the Mann-Whitney U test.

To resolve the contribution of TCR repertoire to the different abilities of T_EM and T_N to induce GVHD, we transplanted TS1 CD4 Tg into BALB/c HA104 Tg mice, which express HA as a model miHAg. TS1 mice were RAG2-deficient, ensuring that all TS1 cells express a single TCR.

HA104 mice were irradiated and reconstituted with RAG^−/− BALB/c BM and 1000, 3000, or 9000 sort-purified BALB/c TS1 T_N (CD44^dimCD62L⁺). As negative controls, we transferred BM without TS1 cells into irradiated HA104 recipients and 3000 TS1 T_N into irradiated BALB/c recipients lacking the HA104 Tg. Transplant of TS1 cells into HA104 recipients induced death and weight loss. Death was dose-dependent, with 100% mortality in mice that received 9000 TS1 cells, 40% in mice that received 3000 TS1 cells, and 10% in mice that received 1000 TS1 cells (Figure 1A). There was no significant difference in weight loss between recipients of 3000 and 1000 TS1 cells (Figure 1B). Importantly, there was no death or long-term weight loss in irradiated HA104 mice that received BM cells alone or in HA-negative BALB/c mice that received 3000 or 9000 TS1 cells (Figure 1 and not shown), verifying that disease is dependent on both donor TS1 cells and expression of HA by recipients.

To investigate whether TS1 cells induce pathology typical of GVHD, HA104 mice were transplanted with 3000 TS1 T_N and histopathology was evaluated on day 21 (Figure 1C-D). TS1 recipients developed skin and colon disease and mild, but statistically significant liver disease. Skin pathology included dermal fibrosis, fat loss, inflammation, epidermal interface changes and mild follicular drop-out. In the colon there were inflammatory infiltrates, apoptotic bodies, neutrophilic cryptitis and crypt loss. Affected liver had central perivenulitis and bile duct injury. These pathologic findings are typical of GVHD induced by polyclonal cells. In samples taken at later time points, pathologic changes typical of GVHD were also observed below. GVHD-type pathologic changes in the colon were always observed, while there was some variability in the intensity of skin disease (K.W.J., W.D.S, and M.J.S., unpublished observations, May 2010), as is sometimes the case in miHAg mismatched systems.⁴¹  Notably, TS1 Tg T cells on the B6.C RAG^−/− background (H-2^d) also caused GVHD in BALB/c HA104 mice, with a similar clinical and pathologic syndrome (data not shown and supplemental Figure 1).

Having established that TS1 cells induce GVHD in HA104 mice, we next generated TS1 T_M so we could compare TS1 T_N, T_EM, and T_CM in the HA104 GVHD model (Figure 2A). TS1 cells were stimulated in vitro for 3 days with HA peptide and APCs and then rested in syngeneic RAG^−/− recipients for 2-3 months to allow memory differentiation, as demonstrated in prior work.^25-34  By day 3 of culture, TS1 cells increased in FSC and uniformly up-regulated CD25. CD44 was also up-regulated and a fraction of cells lost expression of CD62L (Figure 2B,C). After residence in RAG^−/− mice for 2-3 months, ∼ 80% of TS1 cells were CD62L⁺CD44⁺ and therefore of a T_CM phenotype, while ∼ 20% of TS1 cells were CD62L⁻CD44⁺ and therefore of a T_EM phenotype (Figure 2D). All memory cells were CD25⁻ and FoxP3⁻, as were TS1 T_N (not shown).

To confirm that our TS1 T_M behaved as previously reported, sort-purified TS1 T_CM and T_EM recovered from RAG^−/− recipients or fresh T_N were stimulated with PMA and ionomycin for 4 hours and assayed for production of IFN-γ by intracellular cytokine staining (Figure 2E). As predicted, and consistent with their memory phenotypes, TS1 T_EM and T_CM, but not T_N, produced IFN-γ.

We compared the ability of TS1 T_N, T_EM, and T_CM to induce GVHD by transferring graded doses of sort-purified TS1 T_N, T_CM, or T_EM, with BALB/c BM cells, into irradiated HA104 recipients. Although all TS1 recipients initially lost more weight than did BM-only controls, weight loss in the T_N and T_CM recipients was far more severe and prolonged as compared with that in T_EM recipients (Figure 3A). Weight change in the 3000 T_EM group did not differ at most time points from that in the BM alone control. However, T_EM were not inert, as 9000 T_EM induced statistically significant weight loss, albeit mild, relative to BM alone controls. Weight changes in the T_CM and T_N recipient groups were similar at all time points. Although mortality was limited in this experiment, all deaths occurred in the T_N and T_CM groups. In other experiments, T_N and T_CM were more lethal, with death in as many as 50% of mice receiving 3000 T_N and 20% of mice receiving 3000 T_CM. In an analysis of survival in 5 independent experiments in which TS1 T_N, T_EM and T_CM were infused, no T_EM recipients died (combined survival data, supplemental Figure 2). Overall, TS1 T_CM were less lethal than were TS1 T_N though this difference was not reflected in weight loss and histopathologic changes (Figures 3 and 4).

Multiple surviving mice from one experiment were killed at day 60 and pathology was scored (Figure 3B). Although potentially biased by death in the T_N and T_CM recipients, which would presumably eliminate mice with the most severe pathology, there were significant differences between the recipients of T_EM and T_CM or T_N. Recipients of TS1 T_N and T_CM developed typical histopathologic GVHD of the skin, liver and colon whereas T_EM induced no skin or liver GVHD. 9000, but not 3000, T_EM induced only mild pathology in the colon. Disease in T_N and T_CM recipients did not significantly differ. In 2 other experiments, colon disease was always significantly milder in T_EM recipients compared with T_N recipients; however, T_N-induced skin disease was variable (mild in one experiment and absent in another, not shown). Liver disease was also mild and not significantly different in these 2 other experiments (not shown). In both of these experiments, TS1 T_N induced death in 35%-50% of mice, which most likely eliminated the mice with the most severe pathology. In parallel experiments in the B6.C TS1→HA104 system, T_N and T_CM induced severe clinical GVHD whereas again TS1 T_EM induced only mild weight loss, significantly less than induced by either T_N or T_CM (supplemental Figure 1).

In sum these experiments demonstrate that when T_EM and T_N have an identical TCR, T_EM are inherently limited in their ability to cause GVHD, whereas T_CM behave similarly to T_N. Thus, repertoire alone cannot completely account for the different abilities of T_EM and T_N to induce GVHD.

To investigate why TS1 T_EM induce less GVHD than do TS1 T_N or T_CM, we tested the hypothesis that the lack of GVHD relates to differences in expansion of T_EM vs T_N or T_CM progeny posttransfer. HA104 mice were irradiated and reconstituted with BALB/c RAG^−/− BM with no T cells or with 3000 TS1 T_N, T_CM, or T_EM. At weekly intervals posttransplant, TS1 cells in spleen, liver, BM and colon were enumerated by flow cytometry (Figure 4A-B). Strikingly, in spleen and BM similar numbers of TS1 cells were recovered in T_EM, T_N, and T_CM recipients at all time points. In liver, aside from a small lag in T_EM recipients at day 7, TS1 numbers were also similar in all recipient groups. Thus, alloreactive TS1 T_EM engraft, survive, and robustly expand in vivo. However, there were substantially fewer TS1 cells in the colons of T_EM recipients than in T_N or T_CM recipients at all time points, and by day 28 the difference was approximately 20-fold (Figure 4B). T_EM thus have a selective defect in accumulation in the colon, which likely contributes to their reduced ability to cause clinical GVHD.

Given that T_EM expanded at early time points, we compared histopathology in T_EM, T_CM and T_N recipients at days 7, 14, and 21 after transplant (Figure 4C; representative pathology, supplemental Figure 3). There was no skin GVHD in any group on day 7. Skin scores were similar in all TS1 recipient groups on day 14, but by day 21 T_EM recipients had lower scores than did T_N or T_CM recipients. This was not driven by fat loss (a measure of weight loss) as P values were significant even when fat scores were not included (not shown). Liver GVHD was only detectable on day 14, and T_EM induced mild but significant disease that was similar to that induced by T_N or T_CM. In the colon, T_N and T_CM recipients had disease at all time points while in T_EM recipients significant pathology was only seen on days 7 and 21. However, scores were lower in T_EM recipients than in recipients of T_N or T_CM at day 14 and as compared with scores in T_N and T_CM recipients combined at day 21. Therefore, although T_EM recipients had little GVHD at late time points, there was histologic disease at early times, albeit of a relatively milder nature, suggesting that T_EM can initiate, but are limited in their ability to sustain, a pathologic GVHD response.

The ability of T_EM progeny to accumulate to a significant level in spleen, BM and liver was unexpected, given the lack of clinical GVHD. An informative way to compare the fitness of 2 populations is in a competition assay wherein latent defects not apparent when single populations are tested can be revealed. This is especially relevant for testing the impact of T_EM in GVHD, as in the clinic mixed populations of T cells are infused. We therefore cotransferred 3000 Thy1.1 TS1 T_N and 3000 Thy1.2 TS1 T_EM into irradiated HA104 recipients. The Thy1 difference allowed the fates of the 2 types of cells to be separately tracked. As controls, 3000 TS1 Thy1.1 T_N or Thy1.2 T_EM were transferred individually. Mice were killed on days 14, 21, and 28 and spleen and colon cells were analyzed.

As in prior experiments, when transferred individually, progeny of TS1 T_EM and T_N expanded similarly in spleen while T_EM expanded less in colon. However, the presence of TS1 T_N markedly suppressed the expansion of TS1 T_EM in both spleen and colon at all time points (Figure 5A-B). Thus, the presence of T_N inhibits the expansion of T_EM and suggests that in vivo, even if there are alloreactive T_EM, their function and presumably pathogenicity will be suppressed by alloreactive T_N. Clinical GVHD, as measured by weight loss, was similar in T_N and T_N+T_EM recipients (supplemental Figure 4), indicating that T_EM do not suppress T_N-induced GVHD, consistent with the failure of T_EM to suppress the accumulation of T_N progeny in spleen and colon.

To further investigate why there were fewer TS1 cells in colons of recipients of T_EM than T_N, we used BrdU pulse-labeling to assess the frequency of dividing TS1 cells in both the individual and competitive situations. On days 14, 21, and 28 cohorts were given BrdU 30 minutes before sacrifice, which labels only cells currently in S phase. This short pulse excludes the possibility of cells being labeled outside the colon and subsequently migrating into tissues. At all time points a smaller fraction of T_EM than T_N TS1 progeny recovered from colon were dividing, and this percentage was further suppressed by competing T_N at day 14, but not on days 21 or 28 (Figure 6A-B). In spleen, BrdU incorporation by TS1 cells was similar in all groups at day 14 (Figure 6C). On days 21 and 28, however, a smaller fraction of T_EM than T_N progeny incorporated BrdU, though the difference was less marked than that in the colon. In contrast, there were no significant differences among any of the groups in TS1 cell apoptosis, nor did the competitive presence of T_N increase cell death among T_EM (supplemental Figure 5). Thus, the reduced accumulation of T_EM in the colon is explained at least in part by a reduced ability to divide in situ. The further suppression of T_EM by T_N (Figure 5A-B) primarily occurs by mechanisms other than diminishing proliferation of T_EM.

Because the integrin α4β7 promotes recruitment of T cells into bowel, we investigated whether TS1 cells in T_N, T_EM, and T_CM recipients differed in its expression. In the spleen at day +28 the MFI of α4β7 staining was highest in TS1 T_N recipients as compared with TS1 T_EM recipients (supplemental Figure 6). α4β7 expression of T_CM was between that of TS1 T_N and T_EM. In the colon, however, α4β7 was low in all TS1 recipients, consistent with prior reports that α4β7 is down-regulated in bowel-infiltrating T cells.⁴²  Overall, it is likely that differences in numbers of colonic TS1 cells can be attributed to both differences in proliferation as well as the small differences seen in α4β7 expression.

In addition to quantitative differences, we investigated whether progeny of TS1 T_N, T_CM and T_EM differ qualitatively in ways that may contribute to their reduced ability to induce GVHD. In the spleen, the majority of TS1 cells in all recipients were CD62L⁻CD44⁺ and therefore of an effector/T_EM phenotype (Figure 7A). However, at day 7 a substantial fraction of TS1 cells in T_N and T_CM recipients maintained CD62L expression, though by day 14 all TS1 cells had down-regulated CD62L. As expected, there was little if any reexpression of CD62L by T_EM progeny. In spleen and colon, CD69 expression was elevated on all TS1 T cells regardless of origin (supplemental Figure 7). In contrast, CD25 was uniformly elevated only in colon-resident TS1 progeny of both T_N and T_EM recipients (supplemental Figure 7), suggesting a difference in environment affecting the quality of T cell activation. Consistent with the activated phenotype,⁴³  splenic and colon TS1 cells in both T_N and T_EM recipients also up-regulated GITR (supplemental Figure 7). In sum these data, together with the BrdU data (Figure 6), reflect sustained in situ activation of TS1 T cells locally in the colon as well as spleen.

To evaluate effector functions of TS1 progeny in T_N and T_EM recipients, we stimulated splenocytes and cells extracted from colon with PMA and ionomycin and stained for intracellular IFN-γ, IL-2, and IL-17. At day 21 post transplantation, in colon, ∼ 16% of TS1 T_N progeny produced IFN-γ as compared with ∼ 7.5% of TS1 T_EM progeny (Figure 7B-C). The IFN-γ MFI was also higher among IFN-γ⁺ TS1 T_N progeny. The frequencies of IL-2⁺ (Figure 7B,C) and IL-17⁺ cells (supplemental Figure 8) in the colon were < 1.5% in all groups, suggesting that they were less important for pathogenesis relative to IFN-γ⁺ cells. Nonetheless, there were differences consistent with less effector function deriving from T_EM in that the frequency of IL-2⁺ cells as well as the MFI of IL-17 staining were both lower in TS1 T_EM progeny. In the spleen, we saw a parallel picture, although the frequencies of cytokine-producing cells in both T_N and T_EM recipients were higher than in the colon (Figure 7D-E). Seventy percent and 25% of T_N progeny produced IFN-γ and IL-2 respectively, in contrast to 38% and 11.5% of T_EM progeny, respectively. Again, similar to colon, the MFI of IFN-γ but not IL-2 was higher among IFN-γ producing TS1 T_N progeny. Overall, analysis of cytokine secretion clearly showed enhanced effector function generated by T_N progeny compared with T_EM progeny, with the most striking effects seen with IFN-γ but subtle effects observed with both IL-2 and IL-17.

We also considered whether there were differences in expression of the inhibitory receptor PD-1. PD-1 was uniformly up-regulated on TS1 in both spleen and colon in TS1 T_N, T_EM, and T_CM recipients as early as day 7 after transplantation (not shown). Strikingly, at day 28, the MFI of PD-1 was significantly higher in TS1 T_EM progeny than in T_N progeny (supplemental Figure 7), suggesting that T_EM progeny may be more susceptible to inhibition by PD ligands, which in turn could contribute to both their reduced numbers and diminished IFN-γ and IL-2 production.

The main goal of the present studies was to determine whether differences in TCR repertoire account for the relative inability of CD4⁺ T_EM to induce GVHD. We approached this by devising a system in which highly purified T_M and T_N with identical repertoires could be compared. We developed a TCR Tg model of GVHD and used it to compare the abilities of T_EM, T_CM, and T_N to induce GVHD. We found that T_EM differ from T_N in their ability to induce GVHD, even when expressing the same TCR, and thus there are repertoire-independent explanations for the reduced ability of T_EM to cause GVHD. In contrast, T_CM behaved similarly to T_N.

The role of repertoire differences between T_N and T_EM in GVHD has been controversial. Dutt et al found that prior vaccination of donors with host splenocytes increased the potency of CD4⁺ T_EM to induce GVHD.⁹  However, disease induced by T cells from primed donors was different in character from that induced by T_N. Such differences could have been because of shifts in the repertoire and recognition of different allogeneic epitopes, intrinsic differences between T_N and T_EM, or both. These possibilities could not be distinguished in that polyclonal system because alloreactive and GVHD-inducing T cells could not be tracked, as was possible in the TCR Tg model. Zhang et al reported that donor CD4⁺CD62L⁻ cells isolated from mice with GVHD induced GVHD in secondary recipients whereas fresh T_EM did not.⁴⁴  However, these studies do not bear on the role of repertoire among T_M, because CD4 cells extracted from hosts undergoing GVHD would have been activated effectors rather than resting T_M, as they would have been continuously exposed to alloantigen.

Our studies used T_N and T_M with a fixed TCR, allowing a focus on the repertoire-independent differences. By comparing equal numbers of cells with an identical specificity that only differ in their histories of antigen experience, we could directly test the inherent differences. In this well-controlled setting, we could_, reach the clear conclusion that T_EM have substantial inherent and repertoire-independent defects compared with T_N. Clinically and histopathologically, TS1 T_EM caused only mild and transient early disease, which was much less severe than that caused by TS1 T_N, even when almost 10-fold more T_EM were transferred. In this context, it is important to point out that our studies do not contradict a role for repertoire—it could also play a role if different between T_N and T_EM. Rather, our data complement studies investigating repertoire, such as those by Dutt et al⁹  and our own recent studies,⁴⁵  by demonstrating that T_EM and T_N do differ fundamentally independent of repertoire. Thus, both polyclonal and TCR-restricted models of GVHD can make unique contributions to understanding the roles of repertoire and intrinsic defects in the reduced potential of T_EM to cause GVHD.

Our TCR Tg model of GVHD differs from prior TCR Tg models in important ways. We chose to target a peptide antigen ubiquitously expressed at a low level in the host, as this recapitulates a typical miHAg targeted in a polyclonal system. Hence, there is presentation of the peptide by both recipient and donor antigen presenting cells (APCs). This is clinically realistic–given that donor APCs promote CD4-mediated GVHD by presenting hostderived peptides^46-48 —and allows the testing of Tg T_N, T_EM, and T_CM under conditions in which donor T cells can continue to encounter Ag-bearing APCs as GVHD evolves. In contrast, prior TCR Tg systems have mostly used TCR Tgs that are reactive against the host MHC and as a consequence T cell stimulation diminishes as host APCs are cleared. Perhaps for this reason, other systems required the transfer of very large numbers of donor T cells, at least 10 to 1000 fold greater than in our system.^16-21,49-51  Further, our Tg T cells were on a RAG^−/− background, assuring that T_N and T_M repertoires were not skewed by endogenous TCR rearrangements that could be selected for as the immune process progresses. Importantly, our model mirrors key features of GVHD in polyclonal systems. TS1 recipients lost weight and developed hunched posture. At T cell doses of only 3000 to 9000 per recipient, GVHD was lethal in a substantial fraction of mice. Skin, liver and bowel had histopathologic changes characteristic of typical GVHD caused by polyclonal T cells.

Whether T_CM can cause GVHD has also been an area of controversy. It was previously reported that a mix of polyclonal CD4 and CD8 T_CM did not induce GVHD,⁵  whereas we found that purified polyclonal CD8⁺ T_CM did induce GVHD,⁴⁰  though milder than that induced by T_N. Here we found that T_CM and T_N both induced similar weight loss and histopathologic disease, though T_CM were modestly but significantly less lethal. That TS1 T_CM were less lethal than TS1 T_N indicates that T_CM also have repertoire-independent differences from T_N. Separation of T_CM and T_N is based on the degree of CD44 expression. In a polyclonal system we could achieve this convincingly with CD8 cells. However, we were unable to cleanly separate CD4 T_CM from T_N based on CD44 expression, which in our hands overlapped because of the broad and continuous distribution of expression levels (B.E.A, W.D.S., and M.J.S., unpublished observations, October 2010). This difficulty was overcome in the TS1 system because we know that CD44⁺CD62L⁺ T_CM were stimulated during the in vivo culture period, and became homogeneously CD25⁺, indicating that they were definitively Ag-experienced. Despite the clear induction of GVHD by Ag-induced T_CM, it is worth noting that our results do not exclude that a subset of “natural” T_CM, perhaps generated via Ag-independent means, may yet be impaired in GVHD-inducing capacity.

An advantage of our model is the ability to quantitatively track the TS1 progeny posttransplant. One of the most surprising findings from such studies was that T_EM expand in the spleen and elsewhere, despite their relative inability to cause GVHD. This seems to contradict findings in polyclonal systems; however, we believe our finding is not necessarily inconsistent with observations in polyclonal GVHD models. Negrin and colleagues demonstrated reduced expansion of T_EM relative to T_N in the FVB→BALB/c model.⁶  Our group similarly recovered fewer progeny of CD4⁺ T_EM relative to T_N in the B6^bm12→B6 system.³  A likely explanation to reconcile these results with those in the TS1 system is that in the polyclonal models alloreactive cells were more frequent among T_N than T_EM. If so, this would argue for a contribution of repertoire in addition to repertoire-independent effects in the polyclonal situation. Also of relevance, we found that T_EM expansion was markedly inhibited by the presence of Ag-specific T_N, potentially indicating that T_EM can be affected by surrounding T cells, which could have impacted results in polyclonal settings.

Another difference between the TS1 and polyclonal models is that polyclonal T_EM likely included cells that arose via homeostatic expansion rather than Ag-specific activation.⁵²  In the TCR Tg system, all T_EM are Ag-experienced; such T cells might represent a particular subset among the polyclonal T_EM phenotype cells with more potent ability to expand on encountering alloantigen in a GVHD setting. Nonetheless, though TCR Tg T cells do indeed expand, albeit to a limited extent in colon, they do not cause sustained clinical disease, in agreement with polyclonal models.

An important question is why T_EM are less capable of causing GVHD. Among the most interesting observations was that despite robust expansion in spleen, far fewer progeny of T_EM accumulated in colon than did progeny of T_N or T_CM, commensurate with reduced pathology induced there by T_EM. One possible explanation for this would be that fewer T_EM progeny entered the colon. This is possible, as expression of α4β7, required for optimal trafficking into bowel,⁵³  was greater in a subset of progeny of TS1 T_N versus T_EM (supplemental Figure 6), although it did not differ in cells already resident in colon. However, migration differences are unlikely to fully explain different numbers of T_N and T_EM progeny in the colon, as the frequency of TS1 cells in S phase was substantially greater in T_N recipients (Figure 6), indicating that T_EM are relatively less proliferative in the colon. That so many TS1 are dividing in situ is in itself a significant finding and suggests initial priming in secondary lymphoid tissues is but the first step in alloreactive T-cell activation, and that further stimulation and division occurs in target tissues, even weeks posttransplant. Continued uniform expression by colonic TS1 cells of activation markers such as CD69, CD25 and GITR also supports this conclusion.

In addition to cell number, at time points when cell numbers and disease scores were diverging between recipients of T_N and T_EM, T_EM were less able to generate cytokines, chiefly IFN-γ. Both cell frequency and the intensity of intracellular staining were lower in progeny of T_EM than T_N. Though why T_EM progeny are less proliferative and functional remains to be determined, an important observation is that they up-regulate PD-1 expression to a greater extent. PD-1 is well-documented, particularly in CD8 T cells⁵⁴  but also in CD4 T cells,⁵⁵  to mediate exhaustion and reduced functional capabilities. It also limits the capacity of naive polyclonal T cells to cause GVHD in MHC-mismatched GVHD models.⁵⁶  Thus, while more work is required to understand qualitative differences between T_N and T_EM progeny and how this relates to pathogenesis, our findings are important mechanistic clues, particularly in conjunction with the observed differences in cell numbers.

TS1 T_EM showed an even more pronounced impairment in cell accumulation when in competition with TS1 T_N. In the spleen, competition with T_N reduced the total number of TS1 T_EM whereas in isolation T_N and T_EM expanded to almost equal extents. This observation could indicate additional and different impairments in T_EM that only operate in the context of more “fit” competitors. It is possible that T_N actively inhibit division or promote the death of T_EM. In any case, it is interesting that T_N were more successful than T_EM in competing for limiting factors or niches that promote division and survival.

Overall, our studies have provided a key insight into why T_EM fail to cause GVHD. While T_EM may have a different repertoire than T_N, by controlling for this we found that T_EM have cell-intrinsic defects that include reduced ability to proliferate in target tissues as well as an inability to cause sustained tissue damage. Critically, we found that T_EM can indeed engraft, be activated and significantly expand in secondary lymphoid tissue; this in turn could explain why T_EM can carry out some beneficial effector functions, such as GVL and recall responses to immunogens.^2,3  The surprising capacity of T_EM to expand bodes well for their use to provide GVL and immunity without clinically significant GVHD. Future work will focus on the nature of the cell intrinsic defects that restrict both proliferation as well as presumably effector function. The system we describe in this report should be useful for such studies, which should shed light on the basic biology of T_EM as well as provide concepts that could guide the prevention and therapy of GVHD.

The authors thank Yale Animal Resources Center technicians for excellent animal care.

This work was supported by National Institutes of Health grants R01-HL066279 (M.J.S. and W.D.S.), P01-AI064343 (W.D.S. and A.J.D.), R01-AI24541 (A.J.C.), and T32-HL007974 (B.E.A.). W.D.S. is a recipient of a Clinical Scholar award from the Leukemia & Lymphoma Society.

National Institutes of Health

Contribution: K.W.J. designed and performed experiments, analyzed data, and wrote the paper; B.E.A. designed and performed experiments and analyzed data; C.Z. performed experiments; histopathology slides were scored by J.M.M. (skin) and A.J.D. (colon and liver); A.J.C. created the HA104 mice; D.L.F. developed the methods to create TS1 memory cells, provided technical advice and edited the manuscript; and W.D.S. and M.J.S. designed experiments, analyzed data, and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Mark J. Shlomchik, Yale University School of Medicine, 333 Cedar St, PO Box 208035, New Haven, CT 06520-8035; e-mail: mark.shlomchik@yale.edu.