Toward targeting B cell cancers with CD4⁺ CTLs: identification of a CD19-encoded minor histocompatibility antigen using a novel genome-wide analysis

To identify the mHag recognized by clone 21, we started with a genetic approach, the pair-wise two-point linkage analysis. In this method, the genomic locus of the mHag is identified by association of thousands of predefined genetic markers to mHag phenotypes (mHag⁺ or mHag⁻) in large pedigrees registered in the Centre d'Etude du Polymorphisme Humain (CEPH) (17). The CEPH families are suitable for this approach because not only have their genomes been screened for genetic markers but also EBV-LCLs are available from each individual. Upon transduction with the appropriate HLA molecules, these cell lines are used as APCs for mHag-specific T cells to determine the mHag phenotype of the CEPH individuals. Thus, we first tested the reactivity of clone 21 against (HLA-DQA1*05/B1*02 transduced) EBV-LCLs of several CEPH families (Fig. 1 A and Fig. S1) and performed the pairwise two-point linkage analysis in which the mHag phenotype data were correlated with predefined genetic markers. Analysis of the data from three families (1331, 1362, and 1408; Fig. 1 A and Fig. S1) revealed a significant linkage between the mHag phenotypes and a large cluster of markers on chromosome 16, with multiple lod scores >3 (θ = 0.001; Fig. 1 B). According to the CEPH database, two children (1189 and 2387) in a fourth family (1416) displayed a crossing over or recombinant haplotype in this region. As depicted in Fig. 1 C, the mHag in this family was inherited together with the indicated paternal allele because the mHag⁻ child 1197 inherited the other allele from the mHag⁺ father. Thus, because the two children with recombinant haplotypes were mHag⁺, the mHag gene should be located in the shared part of their paternal allele, between markers D16S3093 and D16S3034. This area was 16.8 cM on the Marshfield Map and contained 270 genes according to National Center for Biotechnology Information Map Viewer.

The identified mHag locus in chromosome 16 contained too many genes to be analyzed with previously reported strategies (8, 9, 17). Therefore, we sought an alternative strategy. Over the last decade, millions of SNPs have been genotyped in CEPH individuals by the International HapMap Project (http://www.hapmap.org/) to facilitate the identification of polymorphic genes associated with human diseases (18). We hypothesized that direct correlation of mHag phenotypes with the HapMap SNP genotypes might be useful to fine map the genetic locus of our mHag. From the previous linkage analyses, we already knew the mHag phenotype of 23 CEPH individuals included in the HapMap databases. For 15 of these individuals, we also could deduce the mHag zygosity (+/+, +/−, or −/−) from the inheritance pattern of the mHag in the pedigrees (Fig. S1 and Table S1). Thus, using a specific software that can correlate the zygosity information with the SNP genotypes, we correlated the mHag zygosity information of 15 CEPH individuals with the 4146 HapMap SNPs present in the defined region. Despite the seemingly too low numbers of CEPH individuals entered in the analysis, we found a 100% correlation between the mHag zygosities and the genotypes of two intronic SNPs (rs7184597 and rs3924376; Fig. 1 D). According to HapMap, these two SNPs were located within the same linkage disequilibrium (LD) block, posing the possibility that the SNP encoding for our mHag could also be located in this block.

Within the LD block defined by rs7184597 and rs3924376, RABEP2 and CD19 were the only genes with hematopoietic-restricted expression. Therefore, we explored whether RABEP2 and CD19 genes contained nonsynonymous or transcription/translation-altering SNPs that were linked with the SNPs rs7184597 and rs3924376. Genotyping of five representative CEPH family members for nine additional SNPs in RABEP2 and CD19 (Fig. 1 E) revealed a single SNP, rs2904880, that was fully matching with the LD block defined by rs7184597 and rs3924376. The substitution of guanine for cytosine involved in this SNP causes a change of the valine (V) at position 174 into a leucine (L) within the third exon of CD19 (Fig. 2 A). The fact that the SC recipient was positive for both CD19^L- and CD19^V-encoding alleles, whereas the SC-donor lacked the CD19^L-encoding allele (Fig. 1 E), substantiated the idea that the mHag could be derived from the CD19^L allele. Confirming this possibility, clone 21 showed strong reactivity against donor EBV-LCLs transduced with a CD19^L-encoding vector but not with an empty vector (Fig. 2 B). Furthermore, clone 21 showed a dose-dependent reactivity against a 15-meric CD19^L peptide PEIWEGEPPCLPPRD, but not against the allelic counterpart peptide, demonstrating the crucial importance of leucine at position 174 for the immunogenicity of the CD19^L-derived mHag (Fig. 2 C). Finally, testing the reactivity of clone 21 against the overlapping 15-mer peptides derived from the polymorphic region revealed that only those peptides containing the core sequence WEGEPPCLP were recognized (Fig. 2 D). Analysis of eight other CD4⁺ mHag-specific T cell clones isolated from the same patient revealed the CD19^L specificity of three additional T cell clones (unpublished data). All CD19^L-specific T cell clones used the same TCR Vβ16-CDR3 sequence (unpublished data). We could trace this CD19^L-specific TCR in the patient ∼3 mo after SCT, concomitant with the occurrence of an acute graft versus host disease (GvHD) grade III and a strong anti-tumor response, resulting in significant reduction of serum M protein levels (Fig. 2 E). The TCR signal disappeared rapidly after starting immunosuppressive treatment but reappeared soon after tapering immunosuppression, indicating the long-term persistence of CD19^L-specific T cells. Discontinuation of immunosuppression led to the development of an extensive and steroid-resistant chronic GvHD, despite the disappearance of CD19^L-specific TCR signal from periphery (Fig. 2 E).

Our novel strategy was highly effective to fine map the genetic locus of the CD19^L-mHag defined on chromosome 16. Because it would be more convenient if our strategy could be directly executed on a genome-wide basis, we correlated the mHag zygosity of 15 CEPH individuals with the complete set of HapMap SNPs from all chromosomes (public release 21a). This first attempt revealed, in addition to the correct locus, several other 100% false-positive correlations throughout the genome (unpublished data), suggesting that mHag zygosity data from more individuals was required. Therefore, we extracted the mHag zygosity of eight additional CEPH individuals from a recent HapMap release (public release 22), which included the genotypes for the CD19^L-encoding SNP. Indeed, by analyzing the data from 23 CEPH individuals we identified the precise mHag locus (Fig. 3) without any other false positive 100% correlations, demonstrating the feasibility of genome-wide application of our strategy.

To determine whether our method is able to locate the genomic locus of other mHags, we first selected six known clinically relevant mHags (Table S2). After deriving the zygosity of 14 CEPH trios (father-mother-child) for these mHags from the HapMap database, we executed a series of correlation analyses by including data from an increasing number of trios. The results for the representative example mHag HMSD are depicted in Fig. 4. Although the analysis of the data from seven trios revealed several loci with 100% correlations, false-positive correlations were effectively reduced by entering data from more trios. Finally, the correct mHag locus was revealed after analyzing the data from 11 trios. Similarly, the genomic locus of the other five mHags could be identified after analysis of, at most, 14 trios (Table I). These successfully mapped mHags displayed frequencies between 35 and 70%. As there are not many known mHags beyond these ranges, in further evaluation, we analyzed 149 nonsynonymous nonlinked HapMap SNPs with frequencies between 5 and 95% (Table S3) as if they are encoding for immunogenic mHags. As depicted in Fig. 5 A, we successfully mapped 121/124 mHags with frequencies of 10–85%, 4/10 mHags with frequencies of 90%, and 1/15 mHags with frequencies of 5 or 95% using 11–30 trios. These analyses illustrated that our approach was highly powerful for mapping of mHags with frequencies between 10 and 85%. The power beyond these limits decreased significantly.

A potential technical problem of genetic analyses is phenotyping errors. Therefore, we also questioned whether our approach still has any power if the datasets would contain ∼10% false-positive phenotypes. Calculations revealed that such errors cause a certain frequency-dependent reduction in the r² values (Fig. S2). Nonetheless, we evaluated whether these calculated r² values could be used as threshold to discriminate true and false correlations by reexecuting the analyses with 149 SNPs after introducing 7–10% false-positive phenotypes in the datasets. These analyses resulted in the successful mapping of 96/124 (77%) mHags with frequencies of 10–85% and only 2/10 SNPs with frequencies of 90% (Fig. 5 B), indicating that, despite significant power loss at extreme allele frequencies, our approach may still be useful for the genomic mapping of mHags with frequencies between 10 and 85%, even if the datasets contain 10% false-positive phenotypes.

Finally, we explored the possible immunotherapeutic advantages of targeting the CD19^L-encoded HLA class II–restricted antigen by testing the helper and effector functions of clone 21. In a coculture system, antigen-specific activation of clone 21 with the synthetic peptide or with mHag⁺ APCs facilitated antigen-dependent proliferation of a CD8⁺ T cell clone specific for the mHag LRH-1 (9) (Fig. 6 A). In another assay, clone 21 induced CD19^L peptide-dependent maturation of HLA-DQ–matched DCs and polarized them to produce significant levels of IL-12 (Fig. 6 B), which is considered essential for licensing DCs to induce CTL responses (23). Clone 21 also maturated the DCs that were loaded with apoptotic CD19^L-positive EBV-LCLs (Fig. 6 C). In evaluation of its effector functions, clone 21 appeared to secrete high levels of granzyme-B and displayed antigen-specific cytotoxic activity against recipient-derived but not against donor-derived EBV-LCLs (Fig. 7, A and B). Furthermore, when tested against malignant cells derived from various B cell chronic lymphoid leukemia (B-CLL) patients, clone 21 produced significant levels of IFN-γ (Fig. 7 C) and displayed significant antigen-specific cytotoxic activity only toward HLA-matched and CD19^L-positive malignant cells (Fig. 7 D). Collectively, these results indicated that upon adoptive transfer, CD19^L-mHag–specific CD4⁺ T cells may not only provide effective help to other effector cells but may also directly contribute to an antitumor effect by lysing CD19^L-positive tumor cells.

In this paper, we describe the first genuine hematopoietic mHag presented by HLA class II (HLA-DQA1*05/B1*02) molecules to CD4⁺ T cells. This antigen is encoded by a SNP in the B cell lineage-specific CD19 gene. To identify the CD19^L-encoded mHag, we developed a powerful genetic strategy, which is nonlaborious and genome-wide applicable to identify a wide range of HLA class I– as well as HLA class II–restricted mHags.

What may be the clinical importance of this novel HLA class II–restricted mHag? Like CD20, CD19 is a B cell lineage-specific molecule, with constitutive expression in acute and chronic B cell lymphoid leukemias, B cell lymphomas, and in a subset of acute myeloid leukemias. The expression of CD19 is rarely down-regulated in B cell malignancies. Furthermore, CD19 is not expressed in pluripotent stem cells (24). For these reasons, CD19 is widely considered to be an ideal target for immunotherapy of several B cell malignancies. Immunotherapy strategies based on CD19 antibodies (immunotoxins and chimeric TCRs) have been developed and have used both CD4⁺ and CD8⁺ T cells as effector cells (25, 26). Numerous investigators have searched for HLA class I CTL epitopes on the CD19 molecule (27). Although these approaches may be very beneficial, we think that in an immunotherapy setting, targeting of CD19 with CD4⁺ T cells also may provide several additional benefits; in our assays, the CD19^L-specific CD4⁺ T cells not only mediated potent helper functions to activate DCs and stimulate CD8⁺ CTLs but also directly functioned as effector cells to lyse HLA-matched CD19^L-positive malignant cells. Thus, immunotherapy with CD19^L-specific CD4⁺ CTLs alone or in combination with CD8⁺ T cells may permit the exploitation of both helper and effector functions whereby more effective and durable antitumor responses may be established. Because CD19 is a B cell–specific molecule, CD19^L-specific therapy may be feasible not only within but also beyond the allo-SCT settings. The CD19^L allele is expressed in 53% of the Caucasian population, and its antigen-presenting molecule HLA-DQA1*05/B1*02 has a frequency of 15.3% in the Caucasian population. Thus, after an HLA-matched SCT, ∼2.3% of the donor-recipient pairs will be CD19^L mismatched and eligible for a treatment with CD19^L-specific CD4⁺ T cells or vaccination with peptide-loaded or gene-transfected DCs. This percentage can increase up to 8.1% (15.3 × 53%) in an HLA-matched unrelated donor SCT setting or when patients are not transplanted. In all settings, adoptive immunotherapy with ex vivo–generated CD19^L-specific T cells may be feasible, in particular after generation of CD19^L-specific T cells using the so-called TCR transfer approach (15).

What may be the impact of our mHag identification strategy? Our novel genetic approach involves a correlation analysis between mHag zygosities and HapMap SNP genotypes to locate the genetic locus of the mHag. Our results demonstrate that this approach is genome-wide applicable and able to precisely map the genetic locus of a wide range of mHags with phenotype frequencies of 10–85%. The basic idea behind our strategy is similar to a recently introduced method, which utilizes genome association scans after performing SNP array analysis on DNA samples pooled from mHag⁺ and mHag⁻ individuals (28). A major advantage of our strategy is that it covers more SNPs. Although the other method can only analyze 65% of the HapMap SNPs after performing 500,000 SNP arrays, we analyze millions of HapMap SNPs covering the whole genome without even performing an SNP array. Furthermore, our strategy is less laborious. In our analyses, the genome-wide identification of the precise LD block of six known mHags required the mHag phenotyping of 27–42 CEPH individuals, whereas the other approach had to phenotype at least 100 individuals to identify the locus of these mHags (28). Thus, our method can dramatically reduce the work load and the time required to identify mHags, which are the two most important drawbacks of all current mHag identification strategies. The only apparent limitation of our strategy may be its difficulty in identifying mHags with allele frequencies beyond 10–85%. Nonetheless, it should be emphasized that mHags with very low or high phenotype frequencies are of limited immunotherapeutic value. As illustrated in Fig. S3 , the chance of an mHag mismatch between the recipient and donor is <10%, if the mHag phenotype frequency is beyond 26–78% in the HLA-matched sibling setting or beyond 11–89% in the HLA-matched unrelated donor setting. Because the number of eligible patients will also be limited by HLA restriction, mHag frequencies beyond these limits are of little practical value for clinical application. Thus, our strategy, with sufficient power to identify mHags with frequencies of 10–85%, is actually very suitable for the identification of a vast majority of mHags that are of high value for immunotherapy.

A frequent problem in genetic linkage analyses is phenotyping errors. In our approach, we devoted an extreme attention to avoid such errors because our analyses are executed with a very limited set of CEPH individuals. As phenotype errors will result in the reduction of correlation coefficient, our standard criterion of r² = 1 to discriminate true-positive hits from false-positive ones cannot be applied for datasets containing phenotype errors. Yet our analyses with datasets containing 10% false-positive phenotypes revealed that the reduction in correlation coefficient can be calculated (Fig. S2) and that these calculated values could be used as threshold values to eliminate false hits without dramatically reducing the power of the analyses. Thus, our approach, with theoretically adjusted threshold values, may still be useful for identification of clinically relevant mHags even when phenotyping errors cannot be excluded. Nonetheless, as the power decreases significantly we still think that the best strategy for success is to use all means to avoid phenotype errors.

Finally, we think that our strategy is universally applicable because there are five ethnic panels with trios genotyped in the Phase III HapMap. Using one of these panels may be sufficient for analysis of a wide range of mHag-specific T cell clones, even if they are obtained from a different population. In fact, using the Caucasoid CEPH panel, we have been able to map the genomic locus of HMSD, ACC-1^Y, and ACC-2, mHags which were originally described in the Japanese population. Nonetheless, not all HapMap panels consist of trios, which are required for deducing the zygosity information. Because this may still be a potential limitation for identification of mHags in some ethnic populations, we are currently evaluating the possibility of using unrelated HapMap individuals in our strategy. In conclusion, the first HLA class II–restricted hematopoietic mHag as well as the powerful mHag identification strategy described here can significantly facilitate the application of mHag-based immunotherapy in a broader clinical setting.

The HLA-DQA1*05/B1*02-restricted CD4⁺ T cell clone 21 and the HLA-DP4–restricted CD4⁺ control clone 3AB11 were previously isolated from a multiple myeloma patient during the development of acute GvHD grade III and a strong GvT response after allo-SCT from his HLA-identical brother (29). In brief, after SCT, PBMCs of the recipient were stimulated with irradiated pre-SCT PBMCs for 3 wk and thereafter expanded with EBV-LCLs derived from pre-SCT PBMCs. T cell clones were obtained by limiting dilution. The HLA-B7–restricted CD8⁺ T cell clone specific for the mHag LRH-1 was described elsewhere (9). T cell clones were expanded using a feeder cell–cytokine mixture as previously described (15). EBV-LCLs were cultured in RPMI-1640, 10% FBS (Integro), and antibiotics. The Phoenix packaging line (φ-NX-A) and the CD40L-expressing fibroblasts were cultured in DMEM (Invitrogen), 10% FBS, and antibiotics. PBMCs containing 84–99% CD5⁺ malignant B cells (partly provided by R. van Gent, University Medical Center Utrecht, Utrecht, Netherlands) were obtained from 18 B-CLL patients after written informed consent. Experiments with these human materials were approved by local medical ethical committee of the University Medical Center Utrecht.

Cells labeled with fluorochrome-conjugated antibodies (all obtained from BD, except the anti–ΔNGF-R [culture supernatant of clone 20.4]) were analyzed with a FACSCalibur flow cytometer (BD). FACS data were analyzed with CellQuest software (BD).

Commercially synthesized and purified (Pepscan) 15-mer peptides were dissolved in DMSO to 100 mM and diluted in PBS to 6 mM to use in functional assays.

The LZRS-based retroviral vectors containing HLA-DQA1*0501-IRES-GFP or HLA-DQB1*0201-IRES-ΔNGF-R were provided by F. Koning (Leiden University Medical Center, Leiden, Netherlands). The pMX-CD19^174L-IRES-GFP vector was generated by cloning the CD19^L gene derived from an expression vector (OriGene Technologies). Generation of retroviral supernatants and retroviral transduction of EBV-LCLs were described elsewhere (15).

To determine the mHag phenotype (mHag⁺ or mHag⁻) of the CEPH individuals, their HLA-DQA1*05/B1*02–positive (naturally positive or positive after retroviral transduction) EBV-LCLs were used as APCs for the mHag-specific clone 21 at a 1:1 ratio, in a final volume of 200 μl in U-bottomed 96-well plates. After 18 h of incubation, supernatants were collected and tested for the presence of IFN-γ using standard ELISA kits. EBV-LCLs were judged mHag⁺ if the mean OD value at 450 nm of triplicate cultures was >0.250, about four times the background OD value. For <10% of the CEPH individuals, the OD value was between 0.1 and 0.25. In these cases, assays were repeated at least four times to exactly determine the mHag phenotype of the CEPH individual. Individuals remaining doubtful after five attempts were excluded from the analyses.

Cytokine content of cell-free supernatants was determined using commercial ELISA kits according to the manufacturer's instructions (IFN-γ, Invitrogen; granzyme-B, Sanquin [provided by P de Koning, University Medical Center Utrecht, Utrecht, Netherlands]; IL-12-p70, eBioscience). The SEM of triplicate measurements never exceeded 10%.

The mHag phenotypes of the CEPH individuals and 533 markers genotyped on CEPH families 1331, 1362, and 1408 were used as input to perform a genome-wide two-point linkage analysis using the subroutine MLINK of the FASTLINK program as previously described (8, 9, 17, 30, 31).

Genotype data for chromosome 16 were extracted from the CEPH database (version 10; http://www.cephb.fr/cephdb/) to determine recombinant haplotypes in the CEPH families (32, 33). Candidate genes and SNPs were derived from Map Viewer (http://www.ncbi.nih.gov/mapview/) and Ensembl (release 45; http://www.ensembl.org/), respectively.

The software ssSNPer (provided by D. Nyholt, Queensland Institute of Medical Research, Brisbane, Australia), a program for identifying genetically indistinguishable SNPs (34), was used to analyze the correlation between the mHag zygosities and individual HapMap SNPs. For genome-wide analyses, ssSNPer was slightly modified. We have shared this modified ssSNPer at http://www.umcutrecht.nl/subsite/dcch/Research/Hemato-Oncology/Identification-of-novel-GvT-associated-minor-H-antigens.htm. mHag zygosities (+/+, +/−, and −/−) of CEPH individuals were deduced from the Mendelian segregation pattern of the mHag phenotypes in their pedigrees. The HapMap SNP genotypes were downloaded from the HapMap website (http://www.hapmap.org/). The input file for ssSNPer was assembled by inserting the deduced mHag zygosities of CEPH individuals as a fictive SNP in the downloaded HapMap file.

60 ng of genomic DNA isolated from PBMCs or EBV-LCLs was amplified at 94°C for 5 min for 35 cycles (94°C for 30 s, 55°C for 30 s, and 72°C for 30 s) and 72°C for 10 min using primers (Table S4) and directly sequenced using BigDye Terminator kit (v3.1; Applied Biosystems) in an automated ABI PRISM 3100 sequencer (Applied Biosystems).

The TCR-β chain of clone 21 was amplified from genomic DNA isolated from PBMCs collected periodically after allo-SCT using a TCR-specific primer set (Table S4). The specificity of the PCR was controlled using genomic DNA of clone 21 (positive control) and of several irrelevant PBMCs (negative control).

LRH-1–specific CD8⁺ T cell clone (9) was stimulated with HLA-matched CD19^L-mHag–negative or –positive EBV-LCLs at a ratio of 1:2.5 in the presence of 5 μM LRH-1 peptide. 30 Gray–irradiated clone 21 was added with or without 0.6 μM of the 15-mer CD19^L-peptide. ³H-Thymidine incorporation was determined after 48 h.

Immature DCs, generated from HLA-matched monocytes culturing with GM-CSF and IL-4, were maturated by coculture with T cell clones at a DC/T cell ratio of 1:2.5 in the presence of 6 μM of either the CD19^L peptide PEIWEGEPPCLPPRD or irrelevant peptide LPPRDSLNQSLSQDL. In some assays, apoptotic CD19^L-positive EBV-LCLs (apoptosis was induced by incubation with 500 ng/ml FasL for 24 h) were added as the source of CD19^L antigen (DC/EBV-LCL ratio, 1:20). As positive control, DCs were cocultured with CD40L-expressing fibroblasts in a DC/fibroblast ratio of 1:1. After 72 h, DC maturation was assessed by determining the percentage of CD80⁺ and CD83⁺ cells, the mean fluorescence intensity of CD86, HLA-DR, and/or CD40 on the cells, and the content of IL-12 in the culture supernatants.

The capacity of clone 21 to lyse various mHag⁺ or mHag⁻ targets (EBV-LCLs or B-CLL cells) was measured in FACS-based 48-h cytotoxicity assays as previously described (15). The SEM of duplicate cultures never exceeded 15%.

Where indicated, differences between groups were analyzed in two tailed Student's t tests with Prism software (GraphPad Software, Inc.). P-values < 0.05 were considered significant.

Fig. S1 shows the mHag phenotypes for the individuals of CEPH families 1408, 1416, and 1331. Fig. S2 shows the reduction in correlation coefficient in datasets containing 10% false-positive phenotypes. Fig. S3 shows the chance of mHag mismatch between recipient/donor pairs (sibling or unrelated). Table S1 shows detailed phenotypes and zygosities of 15 CEPH individuals and Table S2 shows the input phenotype/zygosity data for retrospective genome-wide analysis. Table S3 shows the 149 SNPs used in the power analyses. Table S4 displays the used PCR primer sets.

We thank Drs. M. Theobald, J. Kuball, and B. Koeleman for stimulating discussions and critical reading of the manuscript, Dr. F. Koning for providing the HLA-DQ retroviral constructs, Dr. D. Nyholt for providing the standalone version of ssSNPer, P. de Koning for providing reagents for granzyme-B assays, and R. van Gent for providing B-CLL samples.

This work was supported by the University Medical Center Utrecht (Utrecht, Netherlands).

The authors have no conflicting financial interests.