• Open Access

Mismatched human leukocyte antigen class II-restricted CD8+ cytotoxic T cells may mediate selective graft-versus-leukemia effects following allogeneic hematopoietic cell transplantation

Authors


To whom correspondence should be addressed.
E-mail: yos-akatsuk@umin.ac.jp

Abstract

Partial human leukocyte antigen (HLA)-mismatched hematopoietic stem cell transplantation (HSCT) is often performed when an HLA-matched donor is not available. In these cases, CD8+ or CD4+ T cell responses are induced depending on the mismatched HLA class I or II allele(s). Herein, we report on an HLA-DRB1*08:03-restricted CD8+ CTL clone, named CTL-1H8, isolated from a patient following an HLA-DR-mismatched HSCT from his brother. Lysis of a patient Epstein–Barr virus-transformed B cell line (B-LCL) by CTL-1H8 was inhibited after the addition of blocking antibodies against HLA-DR and CD8, whereas antibodies against pan-HLA class I or CD4 had no effect. The 1H8-CTL clone did not lyse the recipient dermal fibroblasts whose HLA-DRB1*08:03 expression was upregulated after 1 week cytokine treatment. Engraftment of HLA-DRB1*08:03-positive primary leukemic stem cells in non-obese diabetic/severe combined immunodeficient/γc-null (NOG) mice was completely inhibited by the in vitro preincubation of cells with CTL-1H8, suggesting that HLA-DRB1*08:03 is expressed on leukemic stem cells. Finally, analysis of the precursor frequency of CD8+ CTL specific for recipient antigens in post-HSCT peripheral blood T cells revealed a significant fraction of the total donor CTL responses towards the individual mismatched HLA-DR antigen in two patients. These findings underscore unexpectedly significant CD8 T cell responses in the context of HLA class II. (Cancer Sci 2011; 102: 1281–1286)

Allogeneic hematopoietic stem cell transplantation (HSCT) has been used successfully for the treatment of hematological malignancies. Although HSCT from human leukocyte antigen (HLA)-identical siblings or unrelated donors is feasible to minimize the risk of acute graft-versus-host disease (aGVHD), HSCT from HLA-mismatched donors can be performed when a patient has advanced disease and no HLA matched donor is available.(1) It has been shown that aGVHD and survival rates are comparable between patients receiving HLA-mismatched unrelated HSCT and those receiving fully HLA-matched HSCT when the mismatch combination is not non-permissive.(2) Because the mismatched HLA molecule(s) may serve as a target for donor T cells, the immune response to these HLA in patients receiving a zero non-permissible mismatch HSCT could give rise to a favorable graft-versus-leukemia (GVL) effect with minimal risk of aGVHD. Following HLA-mismatched HSCT, it is commonly believed that CD8+ or CD4+ T cell responses are induced, depending on the mismatched HLA class I or II allele(s), based on the binding of cognate coreceptors to MHC molecules stabilizing weak interactions between T cell receptors (TCR) and MHC.(3)

In the present study, we characterized an HLA class II-restricted CTL clone isolated from a patient with acute myeloid leukemia who received HLA-DR/DP loci-mismatched HSCT. The CTL clone, named CTL-1H8, was CD8+ and its cytotoxicity was blocked by an anti-CD8 antibody as well as by an anti-HLA-DR antibody. The CTL-1H8 clone lysed primary leukemic cells possessing the mismatched HLA-DRB1*08:03, but not cytokine-treated recipient dermal fibroblasts. Engraftment of HLA-DRB1*08:03-positive primary leukemic stem cells in immunodeficient mice(4) was completely inhibited by in vitro preincubation with CTL-1H8. Furthermore, we demonstrated by CTL precursor (CTLp) frequency analysis that a significant fraction of the total donor CD8+ CTL response in this patient was directed against the HLA-DRB1*08:03 molecule. These findings underscore the in vivo immunological relevance of a CD8+ T cell response against mismatched HLA class II molecule(s).

Materials and Methods

Cells, HLA transfectants, and antibodies.  Peripheral blood mononuclear cells (PBMC) were collected and cryopreserved before and after HSCT from a male patient who had received his brother’s bone marrow (BM) for AML (M6; French American British subtype, M6). The HLA genotype of the recipient was A*24:02/*33:03, B*52:01/*44:03, C*12:02/*14:03, DRB1*08:03/*13:02, DQB1*06:01/*06:04, DPB1*02:02/*04:01, whereas that of the donor was mismatched by DRB1*15:02 instead of DRB1*08:03 and by DPB1*05:01 instead of DPB1*02:02. The patient developed grade II aGVHD limited to the skin and extensive chronic GVHD, but has been free from disease recurrence for over 2 years. B-Lymphoblastoid cell lines (B-LCL) were established from the donor and recipient, as well as from normal volunteers. All blood, BM, and tissue samples were collected after the subjects had provided written informed consent, and the study was approved by the Institutional Review Board of Aichi Cancer Center. The B-LCL, including the HLA class I negative B-LCL line 721.221, were maintained in RPMI 1640 medium supplemented with 10% FCS (Immuno-Biological Laboratory, Gunma, Japan), 2 mM l-glutamine, and penicillin/streptomycin (referred to as “culture medium”). The B-LCL were transduced with retroviral vectors carrying individual HLA class I or class II cDNAs, as described previously.(5) The mAbs used in the present study were against the following antigens: pan HLA class I, HLA-DR, HLA-DQ, HLA-DP, CD4, CD8, CD45, and CD34 (all from BD Biosciences, Franklin Lakes, NJ, USA); HLA-DR8 (One Lambda, Canoga Park, CA, USA); and FITC-conjugated rabbit anti-mouse IgM (BD Biosciences). The mAb blocking experiments were performed using final concentrations of 20 μg/mL mAb. Stained cells were analyzed with a FACSCalibur flow cytometer and CellQuest software (BD Biosciences).

Generation of CTL lines and clones.  The CTL lines were generated from the CD8+ fraction of post-HSCT PBMC after three stimulations with irradiated (33 Gy) pre-HSCT recipient PBMC. Interleukin (IL)-2 (20 U/mL; Chiron, Emeryville, CA, USA) was added on Days 1 and 5 after the second and third stimulation. The CTL clones were generated by limiting dilution and expanded as described previously(6,7) and were frozen until use. All cultures were performed in RPMI 1640 medium supplemented with 4% pooled human serum, 2 mM l-glutamine, and penicillin/streptomycin (referred to as “CTL medium”).

Purification of CD34+ leukemia cells using magnetic beads.  Primary leukemic cells carrying HLA-DRB1*08:03 that had been collected and cryopreserved at the time of diagnosis were thawed and positively selected for CD34+ subsets using phycoerythrine (PE)-conjugated anti-CD34 mAb (BD Biosciences) and anti-PE immunomagnetic beads through MACS MS columns (Miltenyi Biotec, Bergisch Gladbach, Germany).

Cytotoxicity assays.  Target cells were radiolabelled with 3.7 MBq 51Cr for 2 h and 1 × 103 target cells/well were mixed with CTL at various effector/target (E/T) ratios in a standard 4-h cytotoxicity assay using 96-well round-bottomed plates. All assays were performed at least in duplicate. Primary dermal fibroblasts from the skin were treated with interferon (IFN)-γ (100 U/mL; Endogen, Woburn, MA, USA) and tumor necrosis factor (TNF)-α (10 ng/mL; Endogen) for 48 h or 7 days, as indicated. Percentage specific lysis was calculated as follows:

(Experimental c.p.m. − Spontaneous c.p.m.)/(Maximum c.p.m. − Spontaneous c.p.m.) × 100

Leukemic stem cell engraftment assay in immunodeficient mice.  Non-obese diabetic/severe combined immunodeficient/γc-null (NOG) mice(4) were purchased from the Central Institute for Experimental Animals (Kanagawa, Japan). All mice were maintained under specific pathogen-free conditions in the Aichi Cancer Center Research Institute. The Ethics Review Committee of the Institute approved the experimental protocol. The CD34+ fraction (3.0 × 106) of Philadelphia chromosome (Ph)-positive primary acute lymphoblastic leukemia (ALL) cells was preincubated for 16 h in CTL medium supplemented with 20 units/mL IL-2 at 37°C with 5% CO2 either alone or in the presence of CTL-1H8 or a control CTL-1B9 (HLA-A*24:02-restricted, minor histocompatibility antigen-specific CTL(8)) at a T cell:ALL cell ratio of 1:1. Thereafter, the cultures were harvested, resuspended in a total volume of 300 μL CTL medium, and inoculated via the tail vein into 8–10-week-old NOG mice. Six to 7 weeks after inoculation, mice were killed, peripheral blood was aspirated from the heart, and BM cells were obtained by flushing the femora with complete medium. Nucleated cells were analyzed for the expression of human CD45, human CD34, or HLA-DR.

Limiting dilution-based CTLp frequency assay.  The proportion of CTLp specific for the HLA-DRB1*08:03 of the total CTLp against potential recipient alloantigens was quantitated using a standard limiting dilution assay. Purified CD8+ T cells from the PBMC obtained on specific days after HSCT, as indicated, were cultured at twofold serial dilutions with 33 Gy-irradiated 3 × 104 CD40-activated B (CD40-B) cells generated from pre-HSCT recipient PBMC in 96-well round-bottomed plates in CTL medium.(5) On Days 2 and 5, 50 U/mL IL-2 was added after each restimulation. There were at least 12 replicates for each dilution. After three rounds of stimulation, a split-well analysis was performed for HLA-DRB1*08:03-specific cytotoxicity against 51Cr-radiolabeled donor B-LCL with or without HLA-DRB1*08:03 cDNA transduction or recipient B-LCL. The wells were considered to be positive if the total c.p.m. released by the effector cells was >3 SD above that in control wells (mean c.p.m. released by the target cells incubated with irradiated stimulator cells alone). In addition, CD8+ cells from another recipient receiving HLA class II-mismatched HSCT were tested in a similar way. Finally, the CTLp frequency was calculated using L-Calc software (StemCell Technologies, Vancouver, BC, Canada).(9)

Results

Cytotoxicity of the CD8+ CTL clone against allogeneic HLA-DRB1*08:03-positive hematopoietic cells.  In all, 27 clones cytotoxic to recipient but not donor B-LCL were isolated by limiting dilution from CD8+ T cells obtained on Day 207 after HSCT. Based on HLA restriction analysis using partially HLA-matched panel B-LCL, three groups of clones were identified: the first two groups (five in group 1 and 14 in group 2) were potentially restricted by HLA-A*24:02, A*33:03, B*44:03, and C*14:03 and showed lytic activity against cytokine-treated fibroblasts; the remaining eight clones in group 3 showed no lytic activity against cytokine-treated fibroblasts and were potentially restricted by HLA-A*24:02 or C*14:02. Because our primary goal was to generate CTL clones that recognized hematopoietic cells, including leukemic cells for selective GVL effect induction,(10) we omitted the group 1 and 2 clones. Of the eight group 3 clones, we chose CTL-1H8 as a representative CTL clone for further analysis owing to its superior lytic and expansion performance.

The CTL-1H8 clone was CD8+ (Fig. 1a) and efficiently lysed recipient B-LCL and phytohemagglutinin-stimulated T cell lines (PHA-blasts) but not donor LCL (Fig. 1b), indicating that CTL-1H8 recognized recipient-specific alloantigen. Surprisingly, antibody-blocking experiments revealed that lytic activity against recipient B-LCL was significantly inhibited by the addition of anti-HLA-DR mAb and anti-CD8 mAb (Fig. 1c). This led us to re-examine CTL-1H8 HLA restriction using B-LCL with or without cDNA transduction of HLA-DRB1*08:03, which was mismatched between the recipient and donor. As shown in Figure 1(d), CTL-1H8 lytic activity was observed only when donor B-LCL, irrelevant B-LCL, or 721.221 B-LCL (all HLA-DRB1*08:03 deficient) were transduced with HLA-DRB1*08:03 cDNA, indicating unexpectedly that CTL-1H8 was restricted by HLA class II molecules, which are generally thought to be recognized by CD4+ T cells. Because 21.221 B-LCL was deficient for HLA class I molecules, the possibility of presentation of the HLA-DBB1*0803-derived peptides to CTL-1H8 is unlikely.

Figure 1.

 Characteristics of the human leukocyte antigen (HLA)-DR8-restricted CTL clone, 1H8. (a) The cytolytic activity of CTL-1H8 was evaluated in a standard 4-h 51Cr release assay. The recognition by CTL-1H8 of target cells was examined against B-lymphoblastoid cell lines (B-LCL) derived from the recipient (Re) and donor (Do), and phytohemagglutinin-stimulated T cell lines (PHA blasts). (b) Flow cytometric analysis of CTL-1H8 for CD4 and CD8. (•), Re B-LCL; (○), Do B-LCL; (bsl00066), Re PHA blasts. (c) Antibody blocking of cytolysis was performed at an effector/target (E/T) ratio of 3:1, with the mAbs indicated at a final concentration of 10 μg/mL. (d) The HLA-DR8-negative B-LCL from the donor and an unrelated individual were transduced with mock or HLA-DRB1*08:03-encoding retroviral vector and tested with CTL-1H8 at an E/T ratio of 10:1.

Because HLA class II expression is restricted to hematopoietic cells and a fraction of activated non-hematopoietic cells, CTL recognizing HLA class II molecules could selectively mediate the GVL effect without GVHD.(11,12) Thus, we examined whether HLA-DRB1*08:03 expression on dermal fibroblasts and their susceptibility to CTL may change before and after cytokine treatment. To this end, the recipient dermal fibroblasts were incubated with IFN-γ and TNF-α for 2 or 7 days and analyzed for HLA-DR8 expression with a DR8-specific mAb. As shown in Figure 2(a), cytokine treatment for 2 days did not induce HLA-DR8 expression at all, whereas 7 days of treatment resulted in an approximate fourfold upregulation. However, the expression level was 1 log lower than that observed for recipient B-LCL (Fig. 2a, right panel). Despite HLA-DR8 upregulation, fibroblasts treated for 7 days were not lysed by CTL-1H8 at all (Fig. 2b), suggesting that the recognition of the HLA-DRB1*08:03 complex by CTL-1H8 may require HLA-bound antigenic peptides that are not produced in fibroblasts or that such weak upregulation may not be sufficient for recognition by CTL-1B8. The latter possibility may be less likely because primary ALL cells with similar HLA-DR8 expression were moderately lysed by CTL-1H8 (see below).

Figure 2.

 Induction of human leukocyte antigen (HLA)-DR8 by cytokines and susceptibility of cytokine-treated dermal fibroblasts to CTL-1H8. (a) The recipient (Re) dermal fibroblasts (Fibro.) were incubated with 100 U/mL interferon (IFN)-γ and 10 ng/mL tumor necrosis factor (TNF)-α for 2 or 7 days and analyzed for HLA-DR8 expression with DR8-specific mAb. A recipient B-lymphoblastoid cell lines (B-LCL) was used as a positive control. (b) Recipient fibroblasts with or without 7 day cytokine treatment were tested for lysis by CTL-1H8 at indicated effector/target (E/T) ratios, in conjunction with recipient and donor (Do) B-LCL as positive and negative controls, respectively. (•), Re B-LCL; (○), Do B-LCL; (bsl00083), Re Fibro.; (bsl00066), Re Fibro. (IFN + TNF, 7 days.)

HLA-DR8 expression in primary leukemia cells and their susceptibility to CTL-1H8.  Expression of DR8 on primary leukemia cells was first examined in conjunction with CD34, which has been shown to be a stem cell marker in humans.(13) Of 51 PBMC or BM specimens from leukemia patients, five had the HLA-DRB1*08:03 or DRB1*08:02 genotype, of which three samples contained a substantial fraction of CD34+ cells, all of which were from patients with ALL (Ph-ALL: HLA-DRB1*08:02; B-ALL#1 and B-ALL-#2: HLA-DRB1*08:03), and had a significant fraction of double-positive cells (Fig. 3a). We next tested whether positively selected CD34+ fractions from the three ALL samples (Fig. 3b) were susceptible to CTL-1H8. As shown in Figure 3(c), the CD34+ fraction from all three ALL samples was lysed by CTL-1H8 and no natural killer activity against HLA-deficient K562 cells was observed. Although the Ph-ALL sample carried the HLA-DRB1*08:02 genotype, the cells were lysed by CTL-1H8, suggesting that the single amino acid difference in the HLA-DRB1 α1 domain between *08:03 and *08:02 did not affect recognition by CTL-1H8.

Figure 3.

 Expression of human leukocyte antigen (HLA)-DR8 in primary leukemia cells and their susceptibility to CTL-1H8. (a) Expression of DR8 on five primary leukemia samples possessing the HLA-DRB1*08:02 or DRB1*08:03 genotype was examined in conjunction with CD34. (b) Purity and HLA-DR8 expression of leukemia cells after positive selection using anti-CD34-PE antibody followed by capture with anti-phycoerythrine (PE)-immunomagnetic beads. (c) Lysis of the CD34+ fraction from all three acute lymphoblastic leukemia (ALL) samples by CTL-1H8, in conjunction with recipient (Re) and donor (Do) B-lymphoblastoid cell lines (B-LCL) as positive and negative controls, respectively. (•), Re B-LCL; (○), Do B-LCL; (bsl00066), Philadelphia chromosome (Ph)-ALL; (bsl00001), B-ALL#1; (⋄), B-ALL-#2; (×), K562. M2, French–American–British subtype M2; CMML, chronic myelomonocytic leukemia; E/T, effector/target. The HLA-DRB1 types of the three ALL samples were as follows: Ph-ALL, *08:02; B-ALL-#1, *08:03; and B-ALL-#2, 08:02.

Inhibition of human Ph-positive ALL cell engraftment in NOG mice by CTL-1H8.  In order to determine whether HLA-DR8 recognized by CTL-1H8 is indeed expressed on leukemic stem cells and thus may have been involved in a GVL effect, we performed the leukemic stem cell (LSC) engraftment assay, as reported previously,(14) using NOG mice.(4) Because we were unable to obtain CD34+ fractions of primary leukemic cells from the present patient, we selected Ph-positive primary ALL (Ph-ALL) leukemic cells (positive for HLA-A*24:02 and DRB1*08:02) for this assay because they were found to be negative for the HLA-A*24:02-restricted minor histocompatibility antigen ACC-1C and were not lysed by the ACC-1C-specific clone CTL-1B9(8) (data not shown), which was used as an irrelevant control (see Materials and Methods). Flow cytometric analysis of the harvested cells was conducted to investigate the expression of human CD45 and CD34. The BM cells of three control mice receiving Ph-ALL CD34+ cells that were cultured in medium alone (n = 1) or with control CTL-1B9 (n = 2) prior to inoculation were found to contain 96.5%, 32.9%, and 10.9% human CD45+ CD34+ cells (Fig. 4a–c), whereas the PBMC of the same three mice contained 65.2%, 5.7%, and 9.6% human CD45+ CD34+ cells (data not shown). In contrast, human cells were undetectable in both BM and PBMC of mice inoculated with Ph-ALL cells precultured with CTL-1H8 (n = 3; 0.07%, 0.01%, and 0.07% human CD45+ CD34+ cells in BM cells; Fig. 4d–f).

Figure 4.

 Inhibition of human Philadelphia chromosome acute lymphoblastic leukemia (Ph-ALL) stem cell engraftment in non-obese diabetic/severe combined immunodeficient/γc-null (NOG) mice by CTL-1H8. The CD34+ fraction (3.0 × 106) of a Ph-positive primary ALL sample was preincubated for 16 h either alone or in the presence of CTL-1H8 or a control CTL-1B9 (see Materials and Methods) and inoculated via the tail vein into 8–10-week-old NOG mice. Six to 7 weeks after inoculation, mice were killed and peripheral blood mononuclear cells and bone marrow (BM) cells were obtained. Flow cytometric profiles are shown for the expression of human CD45 and CD34 of BM cells from (a) control mice receiving Ph-ALL CD34+ cells cultured in medium alone or (b,c) with control CTL-1B9 or (d–f) with CTL-1H8 prior to inoculation.

After HLA-DR mismatched HSCT, HLA-DR-specific CD8+ T cells are detectable in recipient post-transplant PBMC.  A split-well assay was used to estimate the relative frequencies of CD8+ CTLp specific for HLA-DRB1*08:03 and those specific for all alloantigens expressed on the recipient’s hematopoietic cells in the post-HSCT PBMC, as reported previously.(9) As shown in Figure 5 (left panel), the frequency of CTLp reactive with recipient B-LCL and HLA-DRB1*08:03-transfected donor B-LCL in peripheral blood CD8+ cells obtained on Day 207 after HSCT, from which the CTL-1H8 was derived, was 1/1317 (95% confidence interval [CI] 1/906–1/1913) and 1/2689 (95% CI 1/1825–1/3961), respectively, indicating that nearly half the CTL responses to recipient alloantigens in this donor/recipient pair were directed at the mismatched HLA-DR8. On Day 355, the frequency of CTLp recognizing HLA-DRB1*08:03-transfected donor B-LCL was 1/22 580 (95% CI 1/14 241–1/35 801) and that for CTLp recognizing recipient B-LCL was 1/16 508 (95% CI 1/10 823–1/25 178), demonstrating that even at the later time point the CD8+ CTL responses against HLA-DR8 continued to account for a significant fraction (73%) of the total donor CTL response in this donor/recipient pair (Fig. 5, right panel).

Figure 5.

 Human leukocyte antigen (HLA)-DR8-specific CD8+ T cells were detectable in recipient (Re) post-transplant peripheral blood mononuclear cells (PBMC) following HLA-DR-mismatched hematopoietic stem cell transplantation (HSCT). The proportion of CD8+ CTL precursors specific for HLA-DRB1*08:03 among total CTL precursors (CTLp) against recipient alloantigens was quantitated using a standard limiting dilution assay with L-Calc software (StemCell Technologies, Vancouver, BC, Canada). The CD8+ T cells from PBMC on Days 207 or 335 after HSCT were cultured at limiting dilution with irradiated CD40-B cells generated from pre-HSCT recipient PBMC in 96-well round-bottomed plates. After three rounds of stimulation, a split-well analysis was performed for HLA-DRB1*08:03-specific cytotoxicity against 51Cr-radiolabeled recipient B-lymphoblastoid cell lines (B-LCL, -⋄-) or donor B-LCL transduced with HLA-DRB1*08:03 cDNA (····bsl00023····) or mock-transduced donor B-LCL (bsl00066). Wells were considered positive if the total c.p.m. released by effector cells was >3 × SD above that in control wells.

To explore whether our finding is a phenomenon limited to the present patient, we performed similar assays in another patient receiving cord blood HSCT mismatched by three loci (HLA-C, DR, and DQ). As indicated in Table 1, a small fraction (2.2–12.1% at three time points after HSCT) of the total donor CD8+ CTL response in this donor/recipient pair was directed against the mismatched HLA-DRB1*12:01, whereas a slightly higher fraction (7.1–17.6%) was directed against the mismatched HLA-C*04:01. We were unable to examine the CTLp against the mismatched HLA-DQ molecule owing to an insufficient number of cells.

Table 1.   Frequency of CD8+ cells against mismatched HLA-C and DR antigens in the population of CTL precursors
 Anti-ReAnti-HLA-C*04:01 (%)Anti-HLA-DRB1*12:01 (%)
  1. The patient received cord blood transplantation for her T cell acute lymphocytic leukemia. Human leukocyte antigen (HLA) typing for the recipient (Re) and donor (Do) was as follows, with mismatched alleles underlined: Re: A*24:02/11:01, B*15:01/54:01, C*01:02/04:01, DRB1*04:06/12:01, DQB1*03:01/03:02, DPB1*02:01/03:01
    Do: A*24:02/11:01, B*15:01/54*01, C*01:02/–, DRB1*04:06/05:05, DQB1*03:02/04:02, DPB1*02:01/*03:01. HLA, human leukocyte antigen.

Day 521/18821/26 651 (7.1)1/26 101 (7.2)
Day 1021/75771/84 290 (9.0)1/346 830 (2.2)
1.3 years1/16 8071/95 577 (17.6)1/139 331 (12.1)

Discussion

To our knowledge, the present study is the first to demonstrate that CD8+ CTL restricted by a mismatched HLA-DR molecule are induced physiologically and can be cytotoxic against hematopoietic cells carrying the mismatched HLA-DR allele. The HLA-DRB1*08:03-restricted CD8+ CTL-1H8 clone was isolated from a patient who received an HLA-DR-mismatched HSCT. At 207 days after HSCT, CTLp frequency analysis demonstrated that nearly half the CD8+ T cell responses specific for any recipient-specific alloantigen were directed against the mismatched HLA-DRB1*08:03 molecule. Although we were unable to determine the magnitude of the CD4+ T cell responses against the mismatched HLA-DRB1*08:03 molecule because of a paucity of PBMC, the CD8+ CTLp frequency of 1/2689 on Day 207 is high enough to conclude that the isolation of CTL-1H8 was not an artifact. (The composition of the CD8+ CTLp against mismatched HLA-DPB1*02:02 could not be determined in the present study, but it is possible that the remaining CTLp would be partly restricted by the HLA-DP or minor histocompatibility antigens restricted by shared HLA alleles. In this setting, the involvement of tumor antigens could not be assessed because the stimulators used in the present analysis were recipient CD40L-activated normal B cells and not leukemia cells.) This unexpected finding is supported by data from another HLA class I- and II-mismatched HSCT recipient. Because the number of patients receiving HLA-mismatched HSCT from various donors is increasing, it would be of interest to determine the kinetics of T cell reactions to individually mismatched HLA molecules depending on the type of hematopoietic stem cell donor.

It is generally believed that class II MHC-specific TCR transgenic mice predominantly give rise to CD4+ T cells, whereas class I-specific TCR transgenic mice predominantly give rise to CD8+ T cells. Furthermore, CD4 and CD8 are believed to activate T cells effectively when the intrinsic affinity of the TCR or antigen expression is low,(15) and these accessory molecules can work even if the interacting MHC is not directly bound to the self TCR.(16) In line with this, it has been shown that mature CD8+ T cells can develop in class II MHC-specific TCR transgenic mice when CD4 is absent(17) and that polyclonal CD4+ T cells transduced with the TCR molecules cloned from a CD8+ WT1-specific T cell clone can lyse and/or react with their target cells.(18) In addition, the allorecognition of MHC class II molecules by CD8+ T cells prepared from class II-deficient mice,(19) by those stimulated with antigen-specific B cells,(20) and by heteroclitic CD8+ T cells that also recognize a class I(21) have been described. These findings imply the flexibility of coreceptor choice under unusual conditions. Thus, HLA-mismatched HSCT could be one such unusual situation where T cells may fail to follow the lineage instruction in the thymus because of highly inflammatory and immunogenic conditions after HLA-mismatched allo-HSCT.

Leukemic stem cells have a particularly strong capacity for proliferation, differentiation, and self-renewal(22) and likely play an important role in disease relapse after HSCT. Our mouse model clearly demonstrated that at least HLA-DRB1*08:03 is expressed on such stem cells and may serve as a GVL target for CTL-1H8 in vivo. Unfortunately, however, we could not confirm the GVL potential of CTL-1H8 against recipient leukemia cells because of a limited number of leukemia cells cryopreserved at the time of diagnosis. Because it has been shown that AML (M6) cells do not always express either HLA-DR and CD34,(23) it would need to be determined whether a small fraction of patient stem cells coexpress both HLA-DR and CD34. Nevertheless, it is of note that targeting an HLA-DR molecule alone using a specific CTL clone was sufficient to inhibit Ph-ALL LSC engraftment, suggesting that most LSC were present in the HLA-DR strongly positive, and not weakly positive or negative, population (Ph-ALL in Fig. 3b).

Finally, GVHD is still the major cause of mortality and morbidity following allo-HSCT. Therefore, selective induction of GVL is crucial. Under less inflammatory conditions, MHC class II molecules are mainly expressed only on hematopoietic cells, including leukemia cells. Thus, targeting HLA-DR molecules could be an ideal approach for this purpose. The patient in the present study has been free of disease recurrence for more than 2 years, but developed grade II aGVHD and extensive chronic GVHD. At least in the Japanese population, it has been shown that disparity in HLA-DR is much less hazardous than that in HLA-A and -B in terms of the development of severe aGVHD and mortality.(2) It remains to be determined in future studies whether targeting a mismatched HLA-DR molecule, especially late after HSCT when inflammatory conditions have subsided, would induce detrimental GVHD. In addition, the potential targeting of an HLA-DP molecule whose disparity is almost permissive following HSCT(2) should be examined.

Acknowledgments

The authors thank Dr William Ho (Exploratory Clinical Development, Genentech, South San Francisco, USA) for critically reading the manuscript. The authors also thank Miwako Nishizawa and Hiromi Tamaki for their expert technical assistance and Drs Hiroo Saji and Etsuko Maruya (HLA Laboratory, Kyoto, Japan) for valuable suggestions regarding HLA typing. This study was supported, in part, by the Third Term Comprehensive Control Research for Cancer (no. 30) and Research on Allergic Disease and Immunology from the Ministry of Health, Labour, and Welfare, Japan, as well as Grants-in-Aid for Scientific Research (C) (no. 21591256) from the Ministry of Education, Culture, Science, Sports, and Technology, Japan.

Disclosure Statement

The authors declare no competing financial interests.

Ancillary