Central Memory CD8⁺ T cells Induce Graft-versus-Host Disease and Mediate Graft-versus-Leukemia

Abstract

In allogeneic hematopoietic stem cell transplantation, mature donor αβ T cells in the allograft promote T cell reconstitution in the recipient and mediate the graft-versus-leukemia (GVL) effect. Unfortunately, donor T cells can attack nonmalignant host tissues and cause graft-versus-host disease (GVHD). It has previously been shown that effector memory T cells not primed to alloantigen do not cause GVHD, yet transfer functional T cell memory and mediate GVL. Recently, central memory T cells (T_CM) have also been reported to not cause GVHD. In contrast, here we demonstrate that purified CD8⁺ T_CM not specifically primed to alloantigens mediate GVHD in the major histocompatibility complex (MHC)-mismatched B6→BALB/c and the MHC-matched, multiple minor histocompatibility antigen-mismatched C3H.SW→B6 strain pairings. CD8⁺ T_CM and naïve T cells (T_N) caused similar histologic disease in liver, skin and bowel. B6 CD8⁺ T_CM and T_N similarly expanded in BALB/c recipients and the majority of their progeny produced IFN-γ upon restimulation. However, in both models, CD8⁺ T_CM induced milder clinical GVHD than did CD8⁺ T_N. Nonetheless, CD8⁺ T_CM and T_N were similarly potent mediators of GVL against a mouse model of chronic phase chronic myelogenous leukemia. Thus, in contrast to what was previously thought, CD8⁺ T_CM are capable of inducing GVHD, are substantially different from T_EM, but only subtly so from T_N.

Introduction

Allogeneic hematopoietic stem cell transplantation (alloSCT) is a potentially curative therapy for hematologic malignancies, including acute and chronic leukemias and lymphomas. In an alloSCT, donor T cells in the allograft are critical for reconstituting T cell immunity in the host (1) and mediate the graft-versus-leukemia (GVL) effect (2–5). Unfortunately, donor T cells also cause graft-versus-host disease (GVHD), the broad attack by donor T cells on recipient tissues, including the skin, liver and gastrointestinal tract. All clinical transplant protocols incorporate some type of GVHD prophylaxis, either via depletion of T cells from the allograft or through pharmacologic agents that impair T cell function. This GVHD prophylaxis impairs both GVL and immune reconstitution, and even with pharmacologic immunosuppression, GVHD remains a major source of morbidity and mortality. Preserving the positive effects of donor T cells—GVL and immune reconstitution—without GVHD remains the central challenge in the alloSCT field.

Peripheral T cells can broadly be divided into those that have never been activated by antigen (naïve T cells) and antigen experienced T cells, which include effector and memory T cells (6–9). Memory T cells are themselves heterogeneous and can be subdivided into effector memory (T_EM) and central memory (T_CM) fractions by surface phenotype and by function (10). T_EM (CD62L^lo/CCR7⁻CD44⁺) quickly express effector functions upon restimulation, preferentially migrate to inflamed tissues and the spleen, and are relatively excluded from peripheral lymph nodes (LN). T_CM (CD62L⁺CCR7⁺CD44⁺) have hybrid properties of both T_N and T_EM. Like T_N, their expression of CD62L and CCR7 promotes efficient homing to peripheral LNs (8, 10–16). However, while their effector functions are more vigorous than T_N, they are slower to mount these responses than are T_EM. Nonetheless, when rechallenged with antigen, T_CM have a greater proliferative capacity than do T_EM, and such a property could be advantageous in both anti-pathogen and anti-tumor responses (11, 17–21).

We and others have previously reported that CD4⁺ and CD8⁺ T_EM have a reduced capacity to induce GVHD (22–26). CD4⁺ T_EM are nonetheless functional and can transfer immunity to a model antigen and retain some GVL activity (22, 26). However, much less is known about the GVHD- and GVL-inducing potential of CD8⁺ T_CM. Chen and colleagues reported that sorted CD45RB⁺CD62L⁺CD44⁺ T_CM (a mix of CD4⁺ and CD8⁺ cells) do not cause GVHD in the B6(H-2^b) →BALB/c (H-2^d) fully MHC-mismatched allogeneic bone marrow transplant (BMT) model (27). In contrast, in the present work we demonstrate that highly purified CD8⁺CD62L⁺CD44⁺ T_CM clearly induce GVHD, albeit less severe than that induced by T_N. We did so in the same MHC-mismatched model used by Chen et al and in an MHC-matched, multiple minor histocompatibility antigen (miHA)-mismatched strain pairing. CD8⁺ T_CM were also potent mediators of GVL.

Materials and Methods

Mice

Allogeneic bone marrow transplant (alloBMT) recipients and mice used to generate mCP-CML were 7–10 weeks of age. T cell and BM donors were between 8–16 weeks of age. B6 and BALB/c mice were obtained from the NCI (Frederick, MD). C3H.SW mice were obtained from the Jackson Labs (Bar Harbor, ME) or bred at Yale from breeding pairs purchased from the Jackson Labs.

Cell purifications

For GVHD experiments, CD8⁺ T cells were enriched from total spleen cells using negative selection with BioMag® magnetic beads (Qiagen, Hilden, Germany). Cells were incubated for 30 minutes on ice with the following antibody supernatants (all lab-grown): anti-CD4 (GK1.5), anti-B220 (RA3–6B2), anti-CD11b (M1/70), anti-FcR (24G.2). Cells were then washed in MACS buffer (PBS, 2mM EDTA, 0.5% BSA) and incubated with BioMag® goat anti-rat magnetic beads for 30 minutes on ice in a T75 or T125 flask, which was then placed next to a strong magnet. Cells not bound to magnetic particles were collected and were 60–70% CD8⁺. For further separations of T_N and T_CM, enriched CD8⁺ cells were incubated with anti-CD8-PE (53–7.6, BD Pharmingen), anti-CD62L-Alexa488 (Mel-14, lab-prepared), anti-CD44-APC (Pgp-1, BD Pharmingen) for 20 minutes on ice, washed and incubated with streptavidin-PE (BD Biosciences, San Jose, CA) for 15 minutes on ice, washed again and resuspended in PBS with 0.5% FBS. Cells were sorted into CD8⁺CD62L⁺CD44⁻ T_N and CD8⁺CD62L⁺CD44⁺ T_CM using a FACS Aria (BD Biosciences). Bone marrow (BM) was depleted of T cells as previously described (28). For experiments wherein intracellular cytokine expression and the in vivo CTL assay were performed, CD8⁺ T cells were positively selected from splenocytes using anti-CD8α beads (Miltenyi) and the AutoMACs, stained with antibodies against CD8, CD62L and CD44 and sorted on the FACS Aria as described above. For analysis of cytokine production from freshly isolated T_CM and T_N, splenocytes were sorted using anti-CD8⁻Pacific Blue (clone TIB105, lab-prepared), anti-CD44⁻APC, and anti-CD62L⁻PE (BD Biosciences). B6 and BALB/c B cells were purified by depleting T cells from LN cells using anti-Thy1.2-biotin and streptavidin (SA)-microbeads (Miltenyi) and the AutoMACs.

GVHD transplant protocol

All transplants were performed according to protocols approved by the Yale University Institutional Animal Care and Use Committee. On day 0, B6 or BALB/c mice were irradiated (1000 cGy or 800cGy, respectively) and reconstituted with 7×10⁶ T cell-depleted (TCD) C3H.SW or B6 BM, with no T cells, CD8⁺ T_N, or CD8⁺ T_CM (all cells given intravenously). Mice were weighed approximately every three days. GVHD in the B6→BALB/c model was nonlethal. Based on a large experience in the C3H.SW→B6 strain pairing (28–30), deaths prior to day 13 (which occurred equally in recipients of BM alone, T_N and T_CM) were attributed to the acute toxicity of transplant and not to GVHD and these events were censured in the survival analyses.

Histologic analysis

Tissues were fixed in 10% phosphate buffered formalin, paraffin-embedded, sectioned, and stained with hematoxylin and eosin. Slides were read by pathologists expert in skin (J.M.), liver and gastrointestinal disease (D.J.) without knowledge as to experimental group. Scoring was as described previously (28). Images of bowel and liver were obtained with an Olympus BX40 microscope (Olympus America, Lakewood, CO) using a 10× eyepiece and an 20× objective, with a QImaging QColor5 camera (Olympus) and QCapture software (Olympus). Images were processed with Adobe Photoshop 7.0 software (Adobe, San Jose, CA). Images of skin were obtained with an Olympus BX61 microscope (Olympus) using a 10× eyepiece, 20× objective, a SPOT RT Slider camera (Diagnostic Instruments, Sterling Heights, MI), SPOT software version 4.09 (Diagnostics Instruments) followed by processing with Adobe Photoshop 8.0.

Immunofluorescent staining

Tissues were fixed in 0.7% formaldehyde overnight followed by dehydration in 30% sucrose and freezing in Tissue-TeK OCT compound (Sakura Finetek; Torrance, California). 7 micron sections were incubated overnight at 4°C with DAPI and antibodies against CD8 (Alexa647; clone TIB105; lab-prepared). Sections were imaged with an Olympus BX40 microscope using a 10× eyepiece, 40× objective, a Scanalytics SPOT RT Slider camera (model 2.3.1, Diagnostic Instruments, Sterling Heights, MI) using SPOT software version 4.06 (Diagnostics Instruments). Pictures were reconstituted with Adobe Photoshop 8. To quantitate infiltration by CD8 cells, we counted CD8 cells in four 40× fields each from two mice that received only BM, 3 or 4 fields from two mice that received T_N, and 3 or 4 fields from 2 mice that received T_CM. We treated the number of CD8 cells in a field as an independent observation and compared numbers of cells using Mann-Whitney (Graphpad Prism).

Intracellular cytokine staining

Intracellular cytokine staining was performed as we have described (26). Briefly, splenocytes and LN cells (interscapular, axillary, cervical, mesenteric and inguinal) were harvested and cultured with phorbol myristic acid (PMA) and ionomycin for 5 hours; monensin was added for the final 2 hours. Prior to permeabilization, cells were incubated with ethidium monoazide (EMA) to allow exclusion of dead cells. Cells were stained with antibodies against CD45.1 (PE, BD Biosciences), CD8 (Pacific Blue, clone TIB105, lab prepared), permeabilized and then stained with antibodies against IFN-γ (APC; clone XMG1.2, BD Biosciences) and TNF-α (FITC, clone MP6-XT22, BD Biosciences) or isotype controls (for the anti-cytokine antibodies). Analysis was performed on an LSRII (BD Biosciences).

In vivo CTL assay

Equal numbers of LN-derived B cells from B6 and BALB/c mice were pooled and stained with carboxyfluoresceindiacetate succinimidyl ester (CFSE; Invitrogen, Carlsbad, CA). On day 7 post BMT, 10⁷ B cells were injected into BALB/c recipients of TCD donor B6 BM, with no B6 T cells, or 10⁶ sort-purified B6 T_CM or T_N.

Cytokine measurements

Serum cytokines were measured using the Bioplex mouse Th1/Th2 panel (Bio-Rad, Hercules, CA) and a BioPlex system.

Retrovirus production and progenitor infections

MSCV2.2 expressing the human bcr-abl p210 cDNA and a non-signaling truncated form of the human low affinity nerve growth factor receptor (NGFR) driven by an internal ribosome entry site (Mp210/NGFR) was a gift of Warren Pear. Retroviral supernatants were generated by transfection of Plat-E retrovirus packing cell line as described (28, 29, 31, 32). p210-infected bone marrow progenitors were generated as previously described (28, 29, 32). Briefly, B6 mice were injected on day −6 with 5mg 5-fluorouracil (5FU; Pharmacia & Upjohn, Kalamazoo Michigan). On day −2, BM cells were harvested and cultured in prestimulation media (DMEM, 15% FBS, IL-3 (6 ng/ml), IL-6 (10 ng/ml), and SCF (10 ng/ml) (all cytokines from Peprotech; Rocky Hill, New Jersey). On days –1 and 0 cells underwent “spin infection” with Mp210/NGFR retrovirus.

GVL transplant protocol

On day 0, host mice received 900cGy (450cGy × 2) and were reconstituted with 5×10⁶ TCD C3H.SW BM with 7×10⁵ B6 BM cells that underwent spin-infection with retroviral supernatant, with or without donor CD8⁺ T_N or T_CM. Cause of death was attributed to mCP-CML if mice had a high NGFR⁺ blood count on a recent analysis of peripheral blood (PB) and by spleen weight at necropsy.

Flow cytometry analysis in GVL experiments

Whole blood was stained with antibodies against Gr-1 (FITC; clone RB6–8C5, BD Pharmingen), TER-119 (PE; clone TER-119, BD Pharmingen Biosciences) and NGFR (Alexa647; clone 20.4; lab-prepared) followed by RBC lysis with ACK (Cambrex Bio Science, Walkersville, MD). Propidium iodide was added to exclude dead cells. BM and splenocytes were stained following ACK lysis. Cells were analyzed on a FACS Calibur (Beckton Dickinson Immunocytometry Systems, San Jose CA) and results analyzed with FlowJo (Treestar, Ashland, OR).

Statistical methods

Significance for differences in weight loss was calculated by an unpaired t-test evaluating measurements from all mice in a given group against all mice in a second group as we have previously described (22, 26, 28, 30). P-values for survival curve were calculated by log rank test. P-values for comparisons of histology scores and numbers of NGFR⁺ cells in blood were calculated by Mann-Whitney. P-values comparing numbers of T_N or T_CM derived cells and serum cytokine measurements were calculated using an unpaired t test (GraphPad Prism).

Results

CD8⁺ T_CM induce GVHD in MHC-mismatched and MHC-matched alloBMT models

We first studied CD8⁺ T_CM in the MHC-mismatched B6 (H-2^b) → BALB/c (H-2^d) model. Lethally irradiated BALB/c mice were reconstituted with TCD B6 BM with no T cells or with 1×10⁶ CD8⁺CD62L⁺CD44⁻ T_N or CD8⁺CD62L⁺CD44⁺ T_CM (Figure 1A). Beginning early post-BMT, CD8⁺ T_N recipients lost weight relative to recipients of only TCD BM (Figure 1B) and this weight difference persisted until the experiment was concluded. CD8⁺ T_CM recipients also lost weight relative to recipients of only TCD BM. However, at most timepoints past day 7, they had less weight loss than did recipients of CD8⁺ T_N, and by the conclusion of the experiment, weight change was indistinguishable from recipients of only TCD BM. Thus, as measured by weight change, CD8⁺ T_CM induced GVHD, but less than did CD8⁺ T_N. 2/14 T_N recipients died prior to day 8 and no deaths occurred in T_CM recipients. Mice were sacrificed at approximately day +40 post transplant to harvest skin, small intestine, colon and liver for histologic GVHD. In both T_N- and T_CM-recipients GVHD was only present in the liver, manifest by periportal inflammatory infiltrates and bile duct damage (Figure 1C; representative pathology, Supplemental Figure 1) with a trend towards lower scores in T_CM recipients (P=0.086; two-tailed).

Figure 1. CD8⁺ T_CM induce milder but definite GVHD in the B6→BALB/c strain pairing.

A. Cell sorting. T_N and T_CM CD8 cells were purified and sorted from splenocytes as described in Methods. We gated on CD8⁺ cells (first panel), which were sorted into T_N and T_CM based on the expression of CD62L and CD44. B and C. GVHD in B6→BALB/c model (data combined from 2 experiments). BALB/c mice were lethally irradiated and reconstituted with TCD B6 BM, with no T cells or 1×10⁶ CD8⁺ T_N or T_CM. B. Weight change. P<0.005 at days 5, 8, 12 and 14 comparing recipients of BM versus CD8⁺ T_N or T_CM; P=0.013 comparing BM alone versus T_N on day 20. P<0.01 at day 20, 23, 27, 30, 34, 37 and 40 comparing recipients of CD8⁺ T_N versus CD8⁺ T_CM. C. Liver pathology. Each symbol represents a score from an individual mouse; horizontal lines are mean scores. P=0.0018 comparing BM and CD8⁺ T_N recipients; P=0.0085 comparing BM and CD8⁺ T_CM recipients. P=0.086 comparing CD8⁺ T_N versus CD8⁺ T_CM. D and E. BALB/c mice were irradiated and reconstituted with TCD B6 BM, with no T cells, 1×10⁶ T_N, 1×10⁶ T_CM or 2×10⁴ T_N (T_N control) B6 CD8 cells. D. Weight change. E. Liver pathology scores. P<0.005 for recipients of T_N or T_CM versus T_N control. P value is not significant between BM and T_N control recipients.

Sorted CD8⁺ T_CM had between 1% and 2% contaminating CD8⁺ T_N. We therefore in the second of the two experiments from which data was combined in Figures 1B and 1C, included a group that received 2×10⁴ CD8⁺ T_N (T_{N control}), the number of T_N in the sorted CD8⁺ T_CM. 1×10⁶ CD8⁺ T_CM and CD8⁺ T_N induced both weight loss and hepatic GVHD, whereas 2×10⁴ CD8⁺ T_N induced neither (Figure 1, D and E). Thus the GVHD induced by CD8⁺ T_CM was not due to contaminating CD8⁺ T_N.

Next we determined whether sorted CD8⁺ T_CM mediate GVHD in the C3H.SW(H-2^b)→ B6 (H-2^b) MHC-matched, multiple miHA-mismatched strain pairing. Lethally irradiated B6 mice were reconstituted with C3H.SW TCD BM with no T cells or with 1.5×10⁶ CD8⁺ T_N or T_CM. Recipients of only TCD C3H.SW BM did not develop GVHD whereas most CD8⁺ T_N recipients developed clinical GVHD manifest by death (Figure 2A; 40% by day 43; P=0.0029 comparing BM and T_N), weight loss (Figure 2B), diarrhea, ruffled fur and skin lesions (not shown). In contrast, only 1/13 CD8⁺ T_CM recipients died by day 43 (P=0.2138 comparing BM and T_CM; P=0.054 comparing T_N and T_CM). Weight loss in T_CM-recipients was less than in recipients of T_N and not statistically different than recipients of only donor BM, though there was a trend towards lower weights (as compared to BM controls) at late time-points. Tissues were harvested for histologic analysis on day +43 post BMT. T_CM caused statistically significant pathologic GVHD in the skin, liver and small intestine (Figure 2C; representative histology in Supplemental Figure 1), whereas T_N induced significant GVHD in these tissues and in the colon (Figure 2C). Although T_CM did not induce GVHD in the colon (relative to BM alone controls), scores in comparison to T_N recipients did not reach significance (P=0.159). In the liver, aside from 3 mice with high scores, the majority of mice had relatively mild liver GVHD. Nonetheless, even if the scores from these mice are excluded or assigned a value of 5, the P value comparing liver GVHD in BM alone controls and T_CM recipients is still significant (P<=0.021), though this would render disease in T_N recipients more severe than in T_CM recipients (P<=0.043). Thus, CD8⁺ T_CM induce GVHD in the C3H.SW→B6 strain pairing. However, GVHD was clinically less severe than that induced by T_N and this difference may at least in-part be explained by less colon and liver GVHD in T_CM recipients.

Figure 2. CD8⁺ T_CM induce milder but definite GVHD in C3H.SW→B6 strain pairing.

Data combined from 3 experiments. B6 mice were lethally irradiated and reconstituted with TCD C3H.SW BM, with no T cells, 1.5×10⁶ CD8⁺ T_N or T_CM. A. Survival. P=0.0029 comparing BM and T_N; P=0.2138 comparing BM and T_CM; P=0.054 comparing T_N and T_CM. B. Weight change. P<0.001 at days 26, 30, 34, 37 and 43 comparing recipients of BM only versus CD8⁺ T_N; P<0.05 at days 26 and 43 comparing recipients of T_N versus T_CM. P values not significant at any time-point comparing BM alone versus T_CM. C. Pathology scores. Each symbol represents a score from an individual mouse; horizontal lines are mean scores. P values are noted below each panel.

To further investigate the nature of GVHD in B6 recipients of C3H.SW T_CM we performed immunofluorescence staining on small bowel specimens to identify infiltrating CD8⁺ T cells. CD8⁺ T cells invaded the mucosa with penetration into villi in both T_CM and T_N recipients while such cells were rare in recipients of only TCD C3H.SW BM (Figure 3A, representative images; Figure 3B, number of CD8 cells/40× field). CD8 cells were more frequent in T_CM and T_N recipients than in recipients of only BM (P=0.003) whereas counts in T_CM and T_N recipients were similar (P=0.8). Thus T_N- and T_CM-derived effectors shared the ability to cause infiltrative GVHD of the small intestine.

Figure 3. CD8⁺ T_N and T_CM infiltrate small bowel in the C3H.SW→B6 strain pairing.

Frozen sections of small bowel from recipients of only TCD BM and T_N or T_CM recipients with confirmed histologic GVHD (by blinded scoring) were stained with anti-CD8 (Alexa647; rendered in green). DAPI-stained nuclei are rendered in blue. Shown in A are representative sections from two recipients of T_N, T_CM or only TCD BM. In T_N and T_CM recipients, CD8 cells percolate through the bowel and invade crypts, characteristic of GVHD. B, quantitation of the number of CD8 cells. CD8 cells were counted in 3–4 40× fields from two mice from each group. Each circle represents the number of CD8 cells in an individual field; horizontal lines are means. P=0.003 comparing BM alone versus T_N or T_CM; P=0.8 comparing T_N and T_CM.

CD8⁺ T_CM mediate GVL

To study the efficacy of CD8⁺ T_CM in mediating GVL, we used a mouse model of chronic phase CML (mCP-CML) created via retroviral transduction of BM cells with the human bcr-abl fusion cDNA (p210) (28, 29). The construct also expresses a nonsignaling form of NGFR, which allows detection of mCP-CML cells by flow cytometry. mCP-CML is characterized by a high white blood cell count and splenomegaly, with hematopoiesis dominated by maturing myeloid cells (32, 33). In the C3H.SW→B6 strain pairing GVL is directed towards miHAs and not against p210 and NGFR (29, 34). B6 recipients were lethally irradiated and reconstituted with TCD C3H.SW BM, p210-infected B6 BM and no C3H.SW T cells or 3× 10⁵ CD8⁺ T_N or T_CM. We chose a low dose of donor T cells that mostly prolong survival rather than curing 100% of mice, thereby allowing the detection of minor differences in the abilities of T_N and T_CM to induce GVL that could be masked by higher T cell doses. All mice that did not receive donor CD8⁺ T cells died from mCP-CML by day 22. In contrast, recipients of either CD8⁺ T_N or T_CM had prolonged survival with 1/10 in each group surviving to the end of the experiment at day 75 (Figure 4A; all deaths were from mCP-CML). T_CM-mediated GVL was not due to contaminating T_N as recipients of 7.5×10³ CD8⁺ T_N (T_{N control}), the number of T_N in the sorted CD8⁺ T_CM cells, died from mCP-CML with similar kinetics as did mice that received no donor T cells (Figure 4A). We serially analyzed peripheral blood using NGFR expression to identify mCP-CML cells (Figure 4B, representative flow cytometry). Overall T_N and T_CM recipients had comparable numbers of NGFR⁺ cells (Figure 4C) with T_CM-recipients having fewer NGFR⁺ cells on day +11 (P=0.0039), more on day +18 (P=0.0185), without statistically significant differences at the remaining time-points. Similar results were observed when 1×10⁵ T_N and T_CM were used to induce GVL (not shown). Thus, CD8⁺ T_N and T_CM have comparable capacities to induce GVL.

Figure 4. CD8⁺ T_CM mediate GVL against mCP-CML.

B6 mice were irradiated and reconstituted with TCD C3H.SW BM, B6 mCP-CML with no T cells, 3×10⁵ CD8⁺ T_N, 3×10⁵ CD8⁺ T_CM, or 7.5×10³ CD8⁺ T_N (T_N control) C3H.SW CD8 cells. A. Survival data. P<0.001 for recipients of CD8⁺ T_N or CD8⁺ T_CM versus only BM. B. Representative serial flow cytometry of peripheral blood. Each column is from an individual mouse; two representative mice are shown for T_N and T_CM recipients. C. Numbers of NGFR⁺ cells in the peripheral blood of transplanted mice at different time points. Each symbol represents an individual animal; solid lines are mean values.

T_CM and T_N expansion, cytokine production and cytolytic activity post alloBMT

To better understand how donor T_CM and T_N expand and mature into effectors after transfer, we analyzed progeny of T_CM and T_N 7 days after BMT in the B6→BALB/c model wherein GVHD was induced by 10⁶ CD8⁺ T_N or T_CM. We used CD45.1⁺ and CD45.2⁺ B6 mice as T cell and BM donors, respectively, so as to be able to clearly identify infused T_N and T_CM (BALB/c mice are CD45.2). As controls we performed syngeneic B6→B6 transplants. Both BALB/c T_CM- and T_N-recipients had approximately 4×10⁶ and 4×10⁵ CD45.1⁺CD8⁺ cells in spleen and lymph nodes, respectively (Figure 5, C and D; P>0.6 comparing T_N and T_CM recipients). This expansion was dramatically greater than was seen in B6 recipients of T_CM or T_N wherein spleens and LNs contained 20–30-fold and 10-fold fewer CD8⁺CD45.1⁺ cells, respectively (P<0.0015 comparing BALB/c versus B6 recipients of T_N or T_CM). Although it is unlikely that in BALB/c recipients all progeny of T_CM and T_N at day +7 are alloreactive, the difference in expansion of infused T_CM and T_N in BALB/c relative to B6 hosts is attributable to the presence of and effects from alloreactive T cells, and by this measure T_N and T_CM behaved similarly. Interestingly, relative to T_N-recipients, a greater number of T_CM-derived cells were found in B6 spleens (P=0.026) but not in LN. Decreased expression of CD8 is a marker for T cell activation. Although sorted CD8⁺ T_CM and T_N had a similar fluorescent intensity prior to transfer, in syngeneic transplants T_CM progeny had a lower mean geometric mean fluorescent intensity than did T_N (T_CM: 11870 +/− 520; T_N: 26110 +/−2051; P=0.0025). CD8 expression on both T_N and T_CM was dramatically downregulated in allogeneic recipients (T_CM: 4388 +/−211; T_N: 4642 +/−186; P=0.42; comparing T_N to T_CM in BALB/c recipients; P<= 0.0005 comparing T_N or T_CM in BALB/c versus B6 recipients).

Figure 5. Both T_N and T_CM expand in syngeneic and allogeneic recipients and produce IFN-γ.

BALB/c (CD45.2) and B6 (CD45.2) mice were irradiated and reconstituted with TCD B6 CD45.2 BM and 10⁶ CD8⁺CD45.1+ T_N or T_CM. Seven days post BMT, spleen and LN cells were stimulated with PMA and ionomycin and then stained for CD45.1, CD8, IFN-γ and TNF-α. Positive gates for IFN-γ were based on isotype staining (not shown). A and B, representative flow cytometry of CD8⁺CD45.1⁺ splenocytes and LN cells, respectively. The percentage of IFN-γ+ and IFN-γ very bright cells are shown in the upper left and upper right corners, respectively. Total numbers of CD45.1⁺CD8⁺, CD45.1⁺CD8⁺IFN-γ⁺ or IFN-γ-bright and the percent of CD8⁺CD45.1⁺ cells that are IFN-γ⁺ is shown for spleen (C) and LN (D). Each circle is the value from an individual mouse; horizontal lines are mean values. E, IFN-γ expression after PMA/ionomycin treatment of freshly isolated T_CM (right panel) and T_N (left panel).

To analyze differentiation into effectors, spleen and LN cells from transplant recipients (isolated on Day+7) were stimulated with PMA and ionomycin and analyzed by intracellular cytokine staining for the presence of IFN-γ and TNF-α. We observed little, if any, TNF-α production (Figure 5A, representative staining). However, approximately 40–50% of CD8⁺CD45.1⁺ spleen cells in BALB/c recipients of T_N or T_CM produced IFN-γ, with comparable overall numbers of both IFN-γ⁺ and IFN-γ-very bright cells in T_N and T_CM recipients (Figure 5A, representative staining; Figure 5C, quantitation; P>0.29). The percentage of T_N or T_CM progeny that were IFN-γ+ in LN was similar to that in spleen (Figure 5B, representative staining; Figure 5D, quantitation); however, a smaller percentage were IFN-γ very bright (P<0.02 comparing T_N or T_CM in LN versus those in spleen). Syngeneic recipients of T_CM had more splenic CD8⁺CD45.1⁺IFN-γ⁺ cells than did T_N recipients, which paralleled the difference in overall numbers of donor-derived CD8 cells (P=0.028). We were also interested in how well freshly isolated B6 CD8⁺ T_CM and T_N produced IFN-γ after PMA/ionomycin stimulation (Figure 5E). Only 9% of T_N produced IFN-γ, and only weakly. In contrast, 70% of T_CM produced IFN-γ, with a geometric mean expression nearly 6-fold higher than in T_N. Nearly 25% of T_CM brightly expressed IFN-γ as compared to no IFN-γ-bright T_N. In sum, after transfer to BALB/c recipients, T_CM and T_N similarly differentiated into IFN-γ-producing cells, although T_CM may have already been polarized to do so.

We also analyzed serum samples taken on day +7 from these transplant recipients for cytokines using the Bio-Plex system. We analyzed levels of IFN-γ, IL-2, TNF-α (Figure 6), IL-12 and GMCSF (not shown). Of these, only IFN-γ was elevated in allogeneic recipients relative to syngeneic recipients, (P<0.02 comparing T_N or T_CM in B6→BALB/c versus B6→B6 recipients). There was no difference in IFN-γ levels in allogeneic T_CM versus T_N recipients (P=0.41).

Figure 6. Allogeneic T_CM and T_N induce serum IFN-γ.

Serum was harvested from B6→B6 and B6→BALB/c recipients of B6 T_N or T_CM on day +7. Cytokine concentrations were determined by Bio Plex beads. IFN-γ but not IL-2 or TNF-α were elevated in allogeneic T_CM or T_N recipients. Each symbol is a measurement from an individual mouse; horizontal lines are means.

To compare cytolytic function of transferred T_N and T_CM, we injected 5×10⁶ CFSE-labeled B6 and 5×10⁶ CFSE-labeled BALB/c B cells into BALB/c mice that had been transplanted 7 days prior with TCD B6 BM, with no T cells or with 10⁶ CD8⁺ T_N or T_CM. Twenty hours later, mice were sacrificed and the survival of CFSE⁺B220⁺ B6 and BALB/c cells in spleen and LN was assessed by flow cytometry. In recipients of only TCD BM, the ratio of infused B6 and BALB/c B cells was nearly 1:1 (Figure 7). In contrast, in both T_N and T_CM recipients, greater than 94% of CFSE⁺B220⁺ cells were H-2K^b+ and therefore B6 in origin. Thus, both T_N and T_CM mediated potent killing of allogeneic BALB/c B cells, consistent with their GVL activity.

Figure 7. Allogeneic BALB/c B cells are preferentially killed in vivo in recipients of B6 T_CM or T_N 7 days post transplant.

Representative flow cytometry of splenocytes (SPL) and lymph node cells (LN). Plots are gated on B220⁺ cells. Note that greater than 94% of CFSE⁺B220⁺ cells are H-2K^b+ and therefore B6 in recipients of T_N or T_CM.

Discussion

Prior studies have shown that T_EM not specifically primed to alloantigens are greatly impaired in their ability to induce GVHD (22–27, 35, 36). The capacity of T_CM to do so has been less well studied, and one group recently reported in a single MHC-mismatched model that T_CM do not cause GVHD (27). In contrast to that study, in the present work we clearly show in MHC-matched and MHC-mismatched strain pairs that CD8⁺ T_CM not specifically primed to alloantigens do cause clinical and pathologic GVHD. However, T_CM were less potent as measured by weight loss, death, and by less pathologic colonic GVHD in the C3H.SW→B6 strain pairing. There was also a suggestion that overall, T_CM-induced less severe liver GVHD in both models. Thus our in-depth studies show that T_CM are capable of inducing GVHD and that they differ substantially from T_EM, but more subtly from T_N.

In contrast to their abilities to induce GVHD, CD8⁺ T_CM and T_N were comparable mediators of GVL as measured both by survival and by the numbers of NGFR⁺ mCP-CML cells in peripheral blood at several time points. For GVL experiments we employed a T cell dose well below that which cures 100% of mice so as to be able to detect small differences in T cell potency, making it unlikely that we missed a major difference. These data are consistent with prior work with TCR-transgenic T_CM in an antitumor model (17, 19). Thus the selective infusion of T_CM may be able to mediate GVL with less GVHD, though the reduction in GVHD may only be modest.

In our studies, T_CM were not generated specifically against alloantigens. Rather they likely differentiated into T_CM after reacting against allergens or commensual and pathogenic organisms that can infect laboratory mice (37–42). Alternatively, these T_CM phenotype cells could be the product of lymphopenia-induced proliferation which occurs early in the development of the immune system (43–46). Consistent with having been previously activated, 70% of freshly isolated CD8⁺ T_CM produced IFN-γ after PMA/ionomycin stimulation. Regardless of their precise activation history, the T_CM used in our studies clearly responded against both allogeneic MHC-peptide and syngeneic-MHC-miHA peptide complexes. Such cross-reactivity, or heterologous immunity, has been well-demonstrated in T cells reactive against both allogeneic MHC (47–62) and against viral peptides on a shared MHC (61, 63–73), analogous to our results in the B6→BALB/c and C3H.SW→B6 models, respectively. Alloreactive memory cells are thought to be barriers to successful solid organ transplantation (48, 51–53, 55, 74), and the present data that T_CM mediate both GVHD and GVL are consistent with these studies.

B6 CD8⁺ T_CM and T_N similarly expanded by day +7 in allogeneic recipients, and expansion of both was 20–30-fold greater than in syngeneic recipients. This difference highlights the alloreactivity in both populations. It is surprising that even though T_CM and T_N differ in TCR repertoire and in their prior activation history, that similar numbers accumulated in spleen and LN. In contrast, in syngeneic transplants, T_CM expanded to larger numbers. These data suggest that there may be T cell intrinsic and/or extrinsic factors that restrict T cell expansion in allogeneic recipients, such as limitation on the amount of antigen (in this case in the form of recipient antigen presenting cells) or cytokines. This picture parallels observations in other models wherein the peak of antigen-driven T cell expansion is not directly proportional to the number of precursor cells that initially are activated (75, 76).

Both T_CM and T_N progeny expressed effector functions in vitro and in vivo. Similar fractions of T_N or T_CM-derived cells produced IFN-γ in both syngeneic and allogeneic recipients, though a greater percentage of both T_N and T_CM were IFN-γ-bright in allogeneic recipients. Although we cannot determine whether IFN-γ-producing progeny of T_CM were derived from cells that had previously been polarized to produce IFN-γ, given that 70% of freshly isolated T_CM produce IFN-γ, it is reasonable to suggest that the precursors of many of the T_CM progeny were already polarized. In contrast, most T_N were likely polarized after transfer in both syngeneic and allogeneic recipients. Allogeneic T_N or T_CM recipients also had elevated serum IFN-γ levels which roughly paralleled the numbers of IFN-γ producing cells.

We also compared CTL activity of T_N and T_CM progeny against allogeneic targets in vivo, and both specifically killed BALB/c B cells relative to syngeneic B6 B cells. These data do not address the frequency of T_N or T_CM-derived CTL that recognize allogeneic targets or their specific potencies on a per cell basis; however they clearly demonstrate that both T_CM and T_N differentiate into CTLs as indicated in the GVL experiments.

Our results differ from those of Chen and colleagues who reported that a mix of CD4⁺ and CD8⁺ T_CM induced neither clinical nor histologic GVHD in the B6→BALB/c strain pairing, which we also used. In that study, T_CM were identified as being CD62L⁺CD44⁺CD45RB⁺ whereas we did not stain for CD45RB. In our hands and in their published data, CD8⁺CD62L⁺CD44⁺ cells are all CD45RB⁺ and therefore the inclusion of CD45RB should not have resulted in their sorting a population distinct from that in our work. In their experiments, 10⁶ total T_CM were transferred, of which approximately half were CD8⁺. Thus it is possible that GVHD was not induced because fewer CD8⁺ T_CM were transferred. If so then 5×10⁵ CD4⁺ T_CM must neither induce GVHD nor promote the activation of alloreactive CD8⁺ T_CM. The failure of a mix of CD4⁺ and CD8⁺ T_CM to induce GVHD was not due to CD62L⁺CD44⁺CD4⁺CD25⁺ regulatory T cells as in their studies, when these cells were depleted, T_CM still did not induce GVHD. CD4⁺ T_CM may in some other way reduce the capacity of CD8⁺ T_CM to cause GVHD, though this is not generally the case with unfractionated CD4 and CD8 cells in GVHD models (77).

We point out that Chen et al reported a negative result, whereas we report a positive one—that is T_CM induced both GVHD and GVL. Therefore one must consider other possible reasons that Chen et al failed to make this observation. It is conceivable that the cell sorting procedure these investigators employed sufficiently reduced the viability, survival after transfer, or functionality of T_CM so as to result in an underestimation of their GVHD-inducing potential. In their experiments, sorted T_CM contained 3% and therefore 30,000 T_N whereas 10,000 unfractionated T cells (of which only ~50% were naïve) that did not undergo cell-sorting, induced death and weight loss in BALB/c recipients. Therefore if the majority of sorted cells were viable, one would have anticipated that some GVHD would have been seen in recipients of 30,000 contaminating T_N, unless this was suppressed by the majority T_CM. It is also possible that antibody bound to CD45RB altered the survival or function of the sorted T_CM (78). Regardless of why our results differ, our data are a critical contribution in showing with multiple models, that CD8⁺ T_CM indeed do cause GVHD.

As to why CD8⁺ T_CM have a greater capacity to induce GVHD than do CD8⁺ T_EM, we see several hypotheses. One possibility is that T_CM may have a broader T cell receptor (TCR) repertoire that includes more receptors that recognize alloantigens. In humans, T_EM and T_CM have distinct but similarly complex repertoires (79). However, it is difficult to extrapolate this data to laboratory mice in which the repertoire of these subsets is less likely to have been shaped by constant exposure to pathogens. T_CM may be more efficiently activated after transfer due to better access to LN and Peyer’s patches, owing to their expression of CD62L and CCR7. Such a mechanism has been postulated to at least in-part explain why in vitro generated CD8⁺ TCR-transgenic T_CM were more potent in an anti-tumor model than were T_EM (17). However, it was recently shown that T_N induce robust GVHD in mice that nearly completely lack LNs and Peyer’s patches (PP) (80, 81). Nonetheless it remains to be tested whether T_CM are more reliant on priming in those sites than are T_N.

Another possibility is that T_CM are more capable than T_EM of undergoing clonal expansion. T_CM have a greater in vitro proliferative capacity (reviewed in (10)). In vivo, adoptively transferred TCR-transgenic T_CM provide greater protection from LCMV infection than do adoptively transferred T_EM, and this was shown to be due to a greater proliferative capacity of T_CM (11). The relative abilities of CD8⁺ T_EM and T_CM to protect from pathogens is model-dependent and features other than proliferative capacity play a role; nonetheless, overall T_CM seem to have a greater ability to undergo clonal expansion (reviewed in (20)). These types of quantititative analyses are not possible in our studies, which used naturally occurring polyclonal T_CM, due to the inability to identify miHA-stimulated T cells and distinguish them from those that underwent lymphopenia-induced proliferation. In alloBMT models, CD4⁺ T_EM do not accumulate as well as T_N (23, 26), but again T cells activated by alloantigen could not be distinguished from those that underwent lymphopenia-induced proliferation.

A related question is why T_CM induced milder clinical and subtly less pathologic GVHD than did T_N. Both T_N and T_CM have ready access to LN and PPs due to their expression of CD62L and CCR7. Also, our data demonstrated that both T_N- and T_CM-derived effectors were able to infiltrate target tissues (Figure 3). However, T_CM may more readily migrate directly into tissues after infusion, thereby effectively reducing the number of cells available for activation in secondary lymphoid tissues. For example, T_CM more efficiently enter BM spaces (82). Nonetheless, by day 7 there were comparable numbers of T_CM and T_N-derived cells in both spleen and LN. T_CM may have a less alloreactive TCR repertoire than do T_N, and again this could be more applicable in humans wherein the TCR repertoire of memory phenotype cells is more likely to have been shaped and narrowed by pathogens (83–86). In MHC-matched, multiple miHA-disparate transplants, not only could the frequency of alloreactive T cells be reduced overall in T_CM, but the specific miHAs targeted could be different as the TCR repertoire can impact on the choice of immunodominant antigens (87–91). Thus T_CM could have chosen a different set or hierarchy of immunodominant antigens, which is known to be a property of T cell responses to miHAs (92–94).

There could also be differences in other intrinsic properties of the progeny of activated T_N and T_CM, such as the types of cytokines elaborated, expression of cytotoxic granules, TNF family ligands and resistance to activation-induced cell death. At a minimum we know that in the B6→BALB/c model, T_N and T_CM-derived cells both produced IFN-γ and had cytolytic function in vivo. However, given that such a high frequency of freshly isolated T_CM produced IFN-γ, it is likely that a substantial fraction of alloreactive T_CM were already programmed. It is possible that they differentiated into effectors less capable of causing disease than did T_N only activated in BMT recipients. Investigating potential differences between T_CM and T_N will require comparing in GVHD models T_N genetically deficient in candidate molecules by which they may differ from T_CM (reviewed in (95–97)) to intact T_CM; given the likelihood that multiple factors contribute and the relatively mild quantitative differences in GVHD induction, such studies are not likely to be informative.

A complete understanding of the relative potencies and mechanisms of action of T_CM, T_EM and T_N in GVHD and GVL models will require experimental systems wherein donor T_N, T_CM and T_EM have equivalent and defined TCR repertoires, which would help to control at least some of these variables. Nonetheless, with available systems, it is important conceptually, mechanistically and clinically to know that CD8⁺ T_CM can cause both GVHD and GVL, in contrast to what was previously thought (27).