Female donors contribute to a selective graft-versus-leukemia effect in male recipients of HLA-matched, related hematopoietic stem cell transplants

The efficacy of allogeneic hematopoietic stem cell transplantation (HSCT) for leukemia is in part due to a graft-versus-leukemia (GVL) effect, which is mediated by T cells contained in the stem cell inoculum and is often associated with the development of graft-versus-host disease (GVHD).^1-3  The lower risk for leukemic relapse for recipients in whom GVHD develops does not always translate into a survival advantage because these patients are at risk for complications related to GVHD-mediated tissue damage and to immunosuppressive drugs administered for the treatment of GVHD.^1,4  Nevertheless, identifying the GVL effect has fostered hope that approaches to allogeneic HSCT based on augmenting immune-mediated antitumor activity can improve outcomes. Examples of efforts to exploit the GVL effect for clinical benefit include the infusion of donor lymphocytes to treat recurrent leukemia after transplantation and the use of low-intensity conditioning regimens that lack significant antitumor activity but sufficiently suppress recipient immunity to allow the engraftment of donor T cells capable of mediating GVL activity.^5-10  The results of both interventions have validated the potency of the GVL effect, but, unfortunately, current approaches have not evolved sufficiently to allow the separation of GVL activity from GVHD.^7,11 

Studies in murine models of allogeneic HSCT and in humans have implicated recipient minor histocompatibility antigens (minor H antigens) as the targets of donor T cells that mediate antileukemic activity and GVHD.^2,12-15  Minor H antigens are peptides derived from cellular proteins encoded by polymorphic genes that differ between the transplant donor and recipient and that are presented to CD8⁺ and CD4⁺ T cells by major histocompatibility complex (MHC) class 1 and 2 molecules, respectively. In humans, only a small number of minor H antigens have been molecularly characterized, and the contribution of T cells specific for individual determinants to GVL activity or GVHD remains largely speculative. It has been suggested that minor H antigens, which are limited in their expression to hematopoietic cells including leukemic cells, may be targets for a selective GVL effect, whereas those that are broadly expressed are targets for GVHD.^16,17  However, the very large number of minor H antigens that are likely encoded by autosomal genes and are involved in allogeneic responses has complicated analysis of the contribution of individual determinants in the clinical setting.

Transplantation of stem cells from a female donor to a male recipient (F→M HSCT) is a special circumstance in which donor T cells specific for minor H antigens, encoded by genes on the recipient Y chromosome that are polymorphic to their X-chromosome homologues, may make a contribution to GVHD and GVL activity. Four genes on the Y chromosome have already been identified to encode minor H antigens, and the role of these antigens in GVHD and GVL responses is being investigated.^18-22  Hematopoietic and nonhematopoietic cells express epitopes that are encoded by SMCY and are presented by HLA-A2 and -B7 to CD8⁺ T cells, and this development of T-cell responses to SMCY has been implicated in GVHD in small studies of male recipients of HSC transplants from female donors.^23,24  In contrast, an HLA–B8-restricted epitope encoded by UTY is not presented to CD8⁺ T cells by nonhematopoietic cells in vitro but is expressed on hematopoietic cells, including leukemic progenitor cells.^20,25  These observations suggest it is possible that some H-Y antigens may be differentially associated with GVHD and GVL responses after transplanting stem cells from female donors into male recipients. The Y-chromosome genes that encode minor H antigens exhibit substantial polymorphism with their X-homologues, suggesting additional epitopes may be presented by other HLA alleles that will also be targets for T cells mediating GVHD or GVL activity after F→M HSCT.

Earlier studies of transplantation outcomes have shown that male recipients of female HLA-identical HSC transplants experience increased GVHD,^4,26-28  but these studies have not specifically examined whether such patients exhibit a GVL effect that is independent of GVHD. We analyzed the effect of donor/recipient sex on GVHD and on relapse of the underlying malignancy in a cohort of more than 3000 patients with hematologic malignancies who underwent myeloablative HSCT from an HLA-identical sibling at a single institution.

Retrospective analysis was performed on data from 3238 patients who underwent a myeloablative preparative regimen for a hematologic malignancy and who received an unmanipulated HSC transplant from an HLA-identical sibling between 1969 and 2001 at the Fred Hutchinson Cancer Research Center (FHCRC). Patients who expressed at least one of these class 1 HLA alleles—HLA-A1, -A2, -A3, -B7, -B8, -B40, or -B60—were included in the analysis because these alleles are the most frequent in the North American population and have been shown to present at least one of the defined H-Y antigens or are predicted by computer algorithm to bind polymorphic peptides encoded by one or more Y chromosome genes. More than 85% of the patients who underwent HLA-identical HSCT at the FHCRC for hematologic malignancy in the study period expressed one or more of these HLA alleles and were included in the analysis. Patient and donor characteristics are shown in Table 1. Our interest in this study was to evaluate the GVL effect, and we elected to categorize patients based on tumor burden at the time of transplantation because tumor burden has been identified as a factor in the efficacy of immune-mediated tumor regression. Thus, less advanced disease was defined as chronic myelogenous leukemia in chronic phase (CML-CP), acute leukemia in remission, or myelodysplastic syndrome classified as refractory anemia or refractory anemia with ringed sideroblasts. All others were considered more advanced disease. Assessing and grading acute and chronic GVHD were performed as previously described.^29,30 

Multivariable logistic regression models were fit to examine the association of donor/patient sex with the probability of acute grades II to IV GVHD. Proportional hazards regression models were fit to examine the association of sex with the risk for failure for time-to-event endpoints (overall mortality, relapse, chronic GVHD). The impact of acute and chronic GVHD on mortality and relapse was assessed by treating GVHD as a time-dependent covariate. Unadjusted survival estimates were obtained using the Kaplan-Meier method. All 2-sided P values associated with regression models were derived from the Wald test, and no adjustments were made for multiple comparisons.

After controlling for patient and donor age, GVHD prophylaxis regimen, conditioning regimen, patient/donor CMV serostatus, and disease status at the time of HSCT, male patients who underwent HSCT with female donors (F→M) had the greatest odds of acquiring grades II to IV acute GVHD among the 4 donor/recipient sex categories. This difference was statistically significant when the F→M subset was compared with F→F or M→M HSCT patients (Table 2). After adjusting for the same variables, the odds ratio of clinical extensive chronic GVHD was also significantly higher in male recipients with female donors compared with recipients in each of the other donor/recipient sex categories (Table 3). In other studies, donor parity has been identified as a risk factor for the development of acute GVHD after HSCT from HLA-identical female donors,^4,26,27  though it has not been a risk factor for overall survival after adjusting for other factors.²⁶  The parity data for our population of HSCT patients was incomplete, and we were unable to evaluate their potential contribution to GVHD. However, the F→F HSCT recipients had the lowest odds ratio of acquiring acute and chronic GVHD, and the median age ranges of the donors in the F→F and F→M groups were identical, suggesting it was unlikely that differences in donor parity were solely responsible for the observed differences in GVHD.

We anticipated that the increased risk for acute and chronic GVHD observed in male recipients of HSC transplants from female donors may be associated with a reduction in the risk for leukemic relapse based on the findings of prior studies that have linked acute and chronic GVHD with the GVL effect.^1-3,31  As shown in Table 4, this was in fact the case because male HSCT patients with female donors had a lower risk for relapse compared with all other donor/recipient sex categories. The difference was statistically significant when compared with the M→M and F→F cohorts and reflected a similar trend when compared with female recipients and male donors. The statistical model was then expanded to include acute and chronic GVHD as time-dependent covariates. If acute or chronic GVHD is largely responsible for the reduction in the risk for relapse, then adjusting for GVHD should decrease the hazard ratios summarized in Table 4. Remarkably, the hazard ratios changed very little after controlling for GVHD (Table 5), suggesting that the GVHD effect is not the major driver in the reduction of the risk for relapse among male recipients of transplants from female HSC donors.

Earlier studies have demonstrated that the GVL effect of allogeneic HSCT is greater for patients with chronic myelogenous leukemia (CML) than for patients with acute myelogenous leukemia (AML) or acute lymphoblastic leukemia (ALL). Comparing HSCT male recipients/female donors with the other donor/recipient sex categories showed a lower risk for relapse for patients with CML, AML, and ALL (Table 6). Consistent with findings from earlier studies, our analysis showed the hazard ratio for relapse was lower among patients who underwent transplantation for CML than it was among patients who underwent transplantation forAML orALL, though the confidence interval for each of these groups overlapped, indicating that the differences between the disease categories were consistent with what one might expect to have been caused by chance alone. After adjusting for acute and chronic GVHD, the reduction in the risk for relapse among male patients after HSCT with female donors was diminished among the CML patients but was equivalent in patients with AML or ALL (Table 7).

Increased GVHD and GVL activity would be expected to have opposing effects on survival, with GVHD having a detrimental effect and the reduction in relapse having a beneficial effect. When overall survival was analyzed, female recipients had higher unadjusted survival rates compared with male recipients (Figure 1). After adjusting for the same variables as before, the reduced survival in male recipients was largely confined to the F→M group (Table 8). Male recipients with female donors had a risk for death that was statistically significantly greater than that for recipients in other donor/recipient sex categories, and the odds ratio for M→M HSCT overlapped with that of F→F and M→M (Table 8). The increased risk for death with F→M HSCT was also observed when patients were stratified according to type of leukemia, but the difference is considered to be only a trend among ALL patients (Table 9). Confidence intervals for these disease-specific comparisons overlap, however, indicating that the observed differences among the 3 disease groups are consistent with what would be expected to have been caused by chance alone.

Donor T cells play a crucial role in the antileukemic activity of allogeneic HSCT, but they are also responsible for the development of GVHD.^1-3,16,32,33  There is substantial interest in identifying the molecular nature of minor H antigens that are targets of GVL responses and GVHD because this may assist in developing strategies for improving the efficacy of transplantation. Genes on the Y chromosome have been identified to encode minor H antigens that may contribute to GVHD and GVL responses in male recipients of stem cell grafts obtained from female donors. Prior studies that have analyzed transplantation outcomes in relation to the sex of the recipients and donors found that male recipients of transplants from female HSC donors had an increased risk for acute and chronic GVHD.^1,3  The data reported here in a large number of HLA-matched HSCT patients at a single institution also found that male recipients of grafts from female donors had a significantly higher probability for acute and chronic GVHD than male recipients with male donors or female recipients with female donors. Of interest, female recipients with male donors also had a higher likelihood of GVHD, albeit lower than male recipients with female donors. This finding has been observed previously after HLA-matched, related HSCT,^4,26  but it remains unexplained. It has been suggested that proteins encoded by the X chromosome may be selectively expressed in female but not in male cells and that they provide minor H antigens that could be recognized after transplantation in females with cells from male donors.⁴  An alternative is that autosomal genes regulated by sex hormones could be differentially expressed in males compared with females and could encode minor H antigens. The differential expression of an autosomal gene by donor and recipient cells as a consequence of a gene deletion in the donor has recently been shown to result in the generation of a minor H antigen in humans, providing a precedent for such a mechanism.³⁴ 

In addition to the increase in GVHD, we found that male recipients with female donors had a decreased risk for relapse compared with all other sex categories. Gratwohl et al²⁸  have shown that male recipients with CML in first chronic phase who receive grafts from female donors have lower relapse rates in the interval between 2 and 5 years after transplantation than female recipients with female donors. When the patients in our study were stratified based on type of leukemia, a statistically significant reduction in relapse was also observed for male patients with CML who received transplants from female donors compared with other recipient/donor sex categories. In addition, male recipients who underwent transplantation for AML and ALL with female donors exhibited a reduced risk for relapse, and this approached statistical significance for ALL. In contrast to the findings of Gratwohl et al,^28,35  we found, when comparing relapse rates of CML-CP among male recipients of transplants from female HSC donors with other recipient/donor sex combinations, there was no departure from the assumption of proportional hazards (data not shown), suggesting that in our cohort the beneficial GVL effect observed was not limited to a particular window of time.

Given the link between recipient/donor sex and GVHD, it was reasonable to hypothesize that the association of recipient/donor sex with risk for relapse was mediated through GVHD. Previous studies have shown that recipients of allogeneic HLA-matched HSCT who do not acquire GVHD are less likely to experience relapses than recipients undergoing autologous HSCT or HSCT from a syngeneic donor,^1,36  suggesting there is a GVL effect associated with allografting that is independent of clinical GVHD. We determined the risk for leukemic relapse for each of the recipient/donor sex categories, controlling for the presence of acute and chronic GVHD as time-dependent covariates. Male recipients of HSC transplants from female donors still exhibited a lower risk for relapse compared with other sex combinations, demonstrating that the reduction in relapse in this group could not solely be explained by the presence of GVHD. When the patients were stratified according to type of leukemia, a lower risk for relapse in male recipients with female donors was only significant (P = .04) for patients with CML. Although a similar trend was observed for patients with AML and ALL, these results lacked statistical significance, perhaps because of smaller numbers of patients within each group.

Despite the beneficial GVL effect observed in male recipients with female donors, these patients had significantly reduced survival compared with all patient/donor sex combinations, suggesting that other variables, particularly GVHD, contribute to transplantation-related mortality. Indeed, after controlling for the presence of acute and chronic GVHD, the increased risk for death associated with F→M transplantation was statistically significantly diminished (data not shown). The reduction in survival in male recipients with female donors was greatest for patients with CML, presumably reflecting the lower risk for relapse in these patients, the development of effective therapy for posttransplantation relapse in the past decade, and fewer causes, other than GVHD, of non–relapse-related death.^6,7,37  Our data support the choice of a male rather than a female sibling donor for a male transplant recipient, particularly for patients at low risk for relapse. However, there may be situations in which the additional GVL effect associated with using a female donor would be beneficial, and our data suggest that strategies could be developed to augment the GVL effect without GVHD, such as by the adoptive transfer of donor T cells specific for minor H antigens encoded by genes on the Y chromosome.

A potential explanation for the reduction in relapse in male recipients of transplants from female HSC donors that is not attributable to increased GVHD is the presence of minor H antigens encoded or regulated by the Y chromosome that are targets for a selective GVL effect. CD8⁺ T cells specific for an epitope encoded by the UTY gene and presented by HLA B8 were isolated from a male patient who did not acquire GVHD after receiving a female HLA-identical graft.²⁰  The CD8⁺ UTY-specific cytotoxic T-lymphocytes (CTLs) lysed hematopoietic target cells in vitro and eliminated AML progenitors in a NOD/SCID mouse model of human leukemia but failed to lyse nonhematopoietic cells in vitro, despite the detection of lowlevel transcription of this gene in nonhematopoietic cells.²⁰  UTY is highly polymorphic with its X homologue. Additional polymorphic UTY peptides that are predicted to bind to other common HLA molecules, including HLA A2, and are recognized by T cells from male recipients of transplants from female HSC donors have recently been described.³⁸  Thus, a careful search for Y-chromosome–encoded minor H antigens may uncover targets that could be exploited for selective GVL activity. An alternative is that minor H antigens may be encoded by autosomal genes that are differentially expressed in male cells because of the presence of male hormones. Androgens have an effect on normal and abnormal hematopoiesis, and a variety of androgen-responsive genes have been described. Thus, the possibility that androgen-responsive genes may be overexpressed in male leukemic cells and may provide antigenic targets for donor T cells cannot be excluded. The selective GVL effect observed among recipients of F→M HSCT in this study provides the impetus for further study of potential mechanisms, including the characterization of HY antigens at the molecular level, to identify potential targets of GVL for use in the immunotherapy of leukemia and to identify those targets responsible for GVHD.