HLA-B8 and HLA-A3 Coexpressed With HLA-B8 Are Associated With a Reduced Risk of the Development of Chronic Myeloid Leukemia

CHRONIC MYELOID leukemia (CML) is a malignant disease affecting an immature hematopoietic precursor cell. Previous studies failed to identify an association between class I antigens and CML1-3 with the exception of a study in which HLA-A19 appeared potentially protective against the acquisition of CML.4 5 

CML is characterized by the chromosomal t(9;22) translocation resulting in the chimeric bcr-abl gene that encodes the P210 fusion protein.6,7 Depending on the exons involved in the breakpoint on chromosomes 9 and 22, variants of this fusion protein may occur, of which b3a2 and b2a2 are the most frequent, each having a frequency of about 45% in the patients.8,9 Recent studies showed that the breakpoint region of the b3a2 protein contains amino acid sequences with binding motifs for HLA-A3, HLA-A11, and HLA-B8.10 This indicates that if the b3a2 protein is processed in the leukemic cell allowing presentation of parts of the protein in HLA molecules on the cell membrane, a T-cell response against this fusion protein may occur in individuals expressing HLA-A3, HLA-A11, or HLA-B8. Cytotoxic T-cell responses have been induced in vitro by bcr-abl–derived synthetic peptides presented by HLA-A3 or -B8 molecules.11 12 If presentation of the b3a2-specific peptides occurs in vivo, autologous T-cell responses may be reflected in diminished frequencies of HLA-A3, -A11, or -B8 in patients with CML compared with those in normal healthy individuals.

To test this hypothesis we performed a multicenter case-control study and a meta-analysis comparing HLA frequencies in patients with CML with those in unaffected individuals.

A multicenter case-control study and a meta-analysis addressing the hypothesis whether there are diminished frequencies of HLA -A3, -A11, or -B8 in CML patients was performed. The frequencies of individuals coexpressing two of these HLA molecules were also compared between patients and controls. The same analysis was performed for control purposes on HLA-A2, an antigen not reported to have a binding motif for bcr-abl–derived peptides and for HLA-A1, also without a binding motif for the fusion protein, but in linkage disequilibrium with HLA-B8. To assess a possible protective effect of the A1B8 haplotype itself, also the frequencies of the combination HLA-A1 positive/HLA-B8 negative phenotype and of the combination HLA-A1 negative/HLA-B8 positive phenotype were compared between patients and controls.

The multicenter case-control study comprised 1,899 patients, consisting of patients who underwent bone marrow transplantation obtained from the data base of the Chronic Leukemia Working Party of the European Group for Blood and Marrow Transplantation (EBMT), restricted to 10 countries with at least 15 patients, of untransplanted patients from two centers also contributing to EBMT, and 512,363 controls, matched per country as present in the Bone Marrow Donors Worldwide Registries.13-15 All HLA data from the EBMT have been reviewed and checked by one of the authors (D.N.).

The results of studies previously reported in literature were combined in a meta-analysis. These studies were obtained by searching the MEDLINE data base and checking references in text books. Studies were included in which at least the frequencies of the HLA-A3, -A11, (-A1 and -A2), and HLA-B8 typings in CML patients were compared with those in controls. The meta-analysis consisted of 463 patients with CML and 4,912 controls.

Differences, evaluated through odds ratios (OR) with 95% confidence interval (CI), were assessed by Woolf-Haldane analysis.16,17 The ORs of the different studies were combined in a meta-analysis using as study weights the inverse variance or precision per study.18 In a similar way, the ORs of the different countries of the EBMT were combined.

In the EBMT case-control study, 23% of 1,899 patients and 26% of 512,363 controls were HLA-A3 positive, indicating a 10% reduced risk for HLA-A3 (OR, 0.90; 95% CI, 0.81 to 1.00). The OR for HLA-A11 was 1.16 (95% CI, 1.02 to 1.33). A total of 18% of 1,899 patients and 23% of 512,363 controls were HLA-B8 positive, resulting in a reduced risk of 27% for HLA-B8 (OR, 0.73; 95% CI, 0.65 to 0.82) (Table 1). The HLA-A1 frequency was 27% in patients and 31% in controls, which gives a reduced risk of 17% (OR, 0.83; 95% CI, 0.75 to 0.92), probably due to the linkage disequilibrium with HLA-B8, as the frequency of HLA-A1 positive/HLA-B8 negative individuals did not differ between patients and controls (OR, 0.95; 95% CI, 0.83 to 1.09), whereas both the frequencies of HLA-A1 negative/HLA-B8 positive and HLA-A1 positive/HLA-B8 positive individuals were diminished among patients compared with controls (OR, 0.73; 95% CI, 0.58 to 0.92, respectively, OR, 0.75; 95% CI, 0.66 to 0.86) (Table 2). Combination of the two protective major histocompatability complex (MHC) molecules HLA-A3 and -B8 gave an even stronger protective effect of 49% (OR, 0.51; 95% CI, 0.40 to 0.67) against the development of CML (Table 3). HLA-A2 frequencies did not differ between patients and controls (Table 1).

The meta-analysis consisted of five different studies and included 463 patients with CML and 4,912 healthy controls (Table 4). In the meta-analysis, the pooled estimated OR for HLA-A3 was 1.19 (95% CI, 0.90 to 1.56) and for HLA-A11 1.09 (95% CI, 0.80 to 1.50). The OR for HLA-B8 was 0.71 (95% CI, 0.52 to 0.97), which gave a 29% risk reduction on the development of CML. The OR for HLA-A1 was 0.80 (95% CI, 0.61 to 1.01) and 1.03 for HLA-A2 (95% CI, 0.82 to 1.30).

Recent in vitro studies have shown that synthetic bcr-abl–derived peptides can be presented in the context of HLA-A3, -A11, and -B8 molecules and can elicit a T-cell response against these peptides.10-12,19 However, no endogenous processing of bcr/abl protein resulting in the presentation of the fusion region in the HLA molecules has been demonstrated. We hypothesized that if the bcr/abl-specific peptides could be processed in vivo in these HLA molecules, the presence of any of these class I antigens may protect an individual against the development of this disease. The results of the multicenter case-control study showed a decreased incidence of individuals expressing HLA-B8 in the CML patient group, corroborating the findings of the in vitro studies. The negative asociation with HLA-B8 was confirmed in the meta-analysis. The HLA-A3 frequency in patients with CML was lowered only in the very large EBMT case-control study, but not in the meta-analysis. A recently published epidemiologic study showed an association between homozygosity for HLA-A3 and early onset CML.20 However, the strongly diminished incidence of the combination of HLA-A3/HLA-B8 in patients with CML, as found in the presented case-control study, indicates a negative association with both HLA-A3 and HLA-B8.

The discrepancy between the protective effect of HLA-B8 and HLA-A3 and the absence of effect of HLA-A11 may be explained by the inability of tumor-specific peptides to bind to HLA-A11. Alternatively, peptides with binding affinities for HLA-A11 may be processed, but are unable to elicit an immune response. In the in vitro studies, there were also differences in peptide binding and sensitivity to specific T-cell lines between these HLA molecules.10 12 

Although immune responses have been reported for both b2a2 and b3a2-derived peptides in HLA-B8,11 the protective effect of HLA-B8 and HLA-A3 may be underestimated. In vitro T-cell responses have been mainly described against b3a2-derived peptides.12 19If no or only minor effects are present in the subgroup of patients with the b2a2 translocation, an even stronger protective effect of HLA-A3 and -B8 may be expected for the b3a2 subgroup. Peptides other than the bcr-abl breakpoint region related may also be immunogenic in combination with HLA-A3 and HLA-B8.

To analyze whether the negative association of HLA-A1 with CML was due to the linkage disequilibrium of the HLA-A1B8 haplotype, the frequencies of HLA-A1 positive/HLA-B8 negative, HLA-A1 negative/HLA-B8 positive, and HLA-A1 positive/HLA-B8 positive individuals were compared between patients and controls in the EBMT case-control study. In the EBMT study, the HLA-A1 positive/-B8 negative frequency resulted in an OR of 0.95 (95% CI, 0.83 to 1.09) for patients with CML, the HLA-A1 negative/-B8 positive frequency gave an OR of 0.73 (95% CI, 0.58 to 0.92) for patients with CML compared with controls, whereas the HLA-A1/HLA-B8 both positive combination yielded an OR of 0.75 (95% CI, 0.66 to 0.86). These results illustrate the negative association of HLA-A1 is due to the linkage disequilibrium with HLA-B8.

In summary, the studies presented show a negative association of HLA-B8 and HLA-A3 with CML. The possible mechanism may be that binding of bcr/abl–derived peptides in these HLA molecules lead to an immune response resulting in a protective effect against the development of CML. These findings may have implications for a bcr-abl fusion protein-based immunotherapeutic approach in HLA-A3 and HLA-B8 positive patients with CML. It may be possible to generate either an autologous T-cell immune response against these peptides in vitro using patient lymphocytes or an allogeneic response using donor lymphocytes. In vitro-generated T-cell lines with a specificity against breakpoint specific peptides may be used as part of the treatment of CML.