Development of Cellular Immune Responses against PAX5, a Novel Target for Cancer Immunotherapy

The ultimate goal of cancer immunotherapy is the development of specific, sustained, and potent antitumor immune responses with minimal toxicity. Impressive clinical and immunologic responses have followed adoptive transfer of expanded lymphocytes derived from autologous melanoma tumors (1, 2). Indeed, the study of melanoma tumor infiltrating lymphocytes has been instrumental in the identification of immunogenic tumor antigens and the subsequent development of cancer vaccine therapies. However, in many cancers, naturally occurring tumor infiltrating lymphocytes have not been characterized, and identification of tumor antigens is required, especially for poor-prognosis tumors for which novel treatment approaches are needed. Tumor antigens can also be discovered using “reverse immunology.” Here a targetable cancer-specific protein is first predicted on the basis of its expression pattern and function, and then putative antigenic peptides are assessed for evidence of immunogenicity. Ideal putative tumor antigens have the following three characteristics: high expression in the tumor, low expression in normal tissues, and a critical role in oncogenesis. This “candidate antigen” approach has been used successfully for a number of potential immunotherapy targets lacking evidence of natural immunogenicity. For example, PSA (3), WT1 (4), CD45 (5), and CD19 (6) were not identified through analysis of naturally occurring tumor reactive lymphocytes but have all been used as targets in models of human immunotherapy. However, none of the above-mentioned antigens have all three of the hallmarks.

We and others have recently proposed the PAX family member PAX3 as a target for immunotherapy, and we have shown evidence of the potential clinical utility of generating anti-PAX3 immune responses (7). The PAX family comprises transcription factors with important developmental roles. The PAX5 protein meets the three criteria of an ideal putative tumor antigen. In hemopoiesis, PAX5 controls the identity of B lymphocytes throughout B-cell development from the pro-B to the mature B-cell stage (8). The developmental restriction of PAX5 to B cells has been deduced from the study of targeted PAX5 germ-line deletions. Pax5^−/− mice are born alive although growth retarded, and the most striking phenotypic deficit is the absence of functional B cells (9–11). All mutants fail to produce pre-B cells, B cells, and plasma cells due to a complete arrest in B-cell development at an early precursor B stage. Apart from altered morphogenesis of the midbrain, mice were phenotypically normal, suggesting a limited spectrum of functional developmental expression (9). Moreover, PAX5 expression is thought to be absent from most adult tissues (12). It is up-regulated and highly expressed in B-cell malignancies (13–15) and a number of other cancers (12, 16, 17) including neuroblastoma (18). Recently, in a genome-wide screen, PAX5 was shown to be the most commonly mutated gene in pediatric acute lymphoblastic leukemia (19). PAX5 has been shown to be oncogenic in vitro (20). Moreover, overexpression of full-length PAX5 as a consequence of chromosomal translocation t(9;14)(p13;q32), which juxtaposes PAX5 with regulatory elements of the immunoglobulin heavy chain gene, is described in B-cell malignancies and is likely to function by inhibiting terminal B-cell differentiation (21–23). Developmentally, PAX5 can inhibit apoptosis (24). Recently, a number of alternate isoforms of PAX5 resulting from differential splicing have been described, although the relative oncogenic and developmental importance of these different forms has not been well characterized (25). Taken together, these findings suggest that PAX5 is likely to play a critical role in tumor maintenance. The implication is that PAX5-targeted immunotherapy is therefore less likely to result in immune escape variants with silencing of PAX5 expression.

In this study, we have identified an immunogenic HLA-A2–restricted epitope sequence (TLPGYPPHV) in human PAX5. We show that this epitope is a potent stimulator of specific anti-PAX5 immune responses from peripheral blood of normal donors. High-avidity T cells reactive against the epitope can be generated from normal donors. This knowledge can be used for therapeutic generation of CTL lines or cloning of high-avidity T-cell receptor genes for use in adoptive immunotherapy.

Quantitative reverse transcription-PCR and transient transfection. Total RNA from cell lines was generated using Trizol reagent (Invitrogen), and RNA from normal human tissues was purchased from Clontech. RNA (100 ng/μL) was reverse transcribed into cDNA using the SuperScript II Reverse Transcriptase Kit (Invitrogen) according to the manufacturer's protocol. Gene-specific primers and probes for the NH₂ terminus region of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and exon 8 of PAX5 were purchased from Applied Biosystems. PAX5 expression was normalized to GAPDH, and PAX5 quantification in cell lines and normal tissues was relative to its expression in PAX5-positive RH18 cells. Relative quantitation was determined by the 2^−ΔΔCT method (Applied Biosystems Users Bulletin 2). Lipofectamine 2000 (Invitrogen) was used transiently to transfect HLA-A2–positive 293-T cells with a mammalian expression vector, pcDNA3.1/Hygro(+) (Invitrogen), containing the full-length PAX5 coding sequence (1,176 bp) together with 77 bp of the 5′-untranslated region (UTR) and 2,026 bp of the 3′-UTR sequence (kindly provided by Andreas Himmelmann, Internal Medicine, University of Zurich, Zurich, Switzerland).

Peptides. Two independent computer-based programs, BioInformatics and Molecular Analysis (BIMAS)⁵

Cell lines. T2 (TAP deficient), 293-T (embryonic kidney), SW480 (colon carcinoma), SH-SY5Y, LAN-1 (neuroblastoma), MDA-MB231 (breast cancer), and TC32 (Ewing's sarcoma) were all from American Type Culture Collection (ATCC); NALM1, NALM6, and NALM27 (Pre-B leukemia) were kindly provided by Dr. Akira Harashima (Fujisaki Cell Center, Okayama, Japan). RH18 (rhabdomyosarcoma) were originally from Peter Houghton (St. Jude Children's Research Hospital, Memphis, TN). Autologous B lymphoblastoid cell lines were established by incubation of peripheral blood mononuclear cells (PBMC) with supernatant from the EBV-producing marmoset cell line B95.8 (ATCC) as described (26). Cell lines were maintained in RPMI 1640 (T2, RH18, SW480, MDA-MB-231, and pre-B leukemia cell lines) or DMEM, supplemented with 10% fetal bovine serum (v/v), l-glutamine, nonessential amino acids, sodium pyruvate, and gentamicin (complete medium). All culture materials were purchased from Life Technologies.

Antibodies and pentamer staining. The phycoerythrin (PE)-labeled pentamer Pro5 (Proimmune) was titered and used at optimal concentrations (5–10 μg/mL) to assay the PAX5-specific CD8⁺ T cells. For staining, 1 × 10⁶ PBMCs or 1 × 10⁵ T-cell clones were washed and resuspended in pentamer staining buffer consisting of Dulbecco's PBS, 0.1% NaN₃, and 0.1% bovine serum albumin (BSA). The cells were stained with 10 μL of Pro5 for 20 min at room temperature, washed, and then stained in buffer with 5 μL of antihuman CD8-FITC (BD Biosciences) for 20 min at room temperature. The cells were washed in buffer and fixed in Dulbecco's PBS, 1% paraformaldehyde and run through a Cyan flow cytometer (DAKO). 7-Amino-actinomycin D (BD PharMingen) was used to exclude dead cells. The pentamer-positive cells were analyzed on a two-color plot with CD8-FITC. Results are shown as percent pentamer-positive cells for each culture. Data were analyzed with Summit (DAKO).

Peptide-binding assay. T2 cells were used to determine binding of peptides to HLA-A2. T2 cells (2 × 10⁵ per well in 96-well plates) were incubated for 18 h with 0 to 100 μmol/L of each synthetic peptide, washed twice in PBS, and stained in the dark at 4°C for 30 min with 2 μg/mL PE-conjugated anti–HLA-A2 (BB7.2) or PE-conjugated mouse IgG2 isotype control (both BD PharMingen). Fluorescence index was calculated as the mean fluorescence intensity.

Human CTL line and clone generation. Unless indicated, all cytokines and granulocyte macrophage colony-stimulating factor (GM-CSF) were purchased from PeproTech. Fresh peripheral blood was centrifuged over Ficoll-Hypaque to obtain PBMC. Adherent cells (1.5 × 10⁶ per well in a six-well plate) were cultured in 10% AB serum, interleukin (IL)-4 (30 ng/mL), and GM-CSF (100 ng /mL) for 7 d, with replenishment on days 3 and 5. On day 6, dendritic cells were matured with keyhole limpet hemocyanin (10 μg/mL), CD40L (500 ng/mL; Biosource), and prostaglandin E2 (500 ng/mL; Cambridge Laboratories). CD8 T cells were stimulated twice at weekly intervals with autologous dendritic cells pulsed with peptide (10 μmol/L) and a further two stimulations with autologous CD40-activated B cells. Briefly, day 7 autologous matured dendritic cells were pulsed with 10 μmol/L peptide for 4 h, γ-irradiated, and cocultured with autologous CD8⁺ T cells in the presence of IL-12 (20 IU/mL) and IL-7 (10 ng/mL), with a ratio of T cells/dendritic cells of 10:1. T cells had been positively selected from nonadherent PBMC using MACS beads (Miltenyi). T cells were restimulated under the same conditions for a second week. The next two further weekly stimulations used autologous peptide-loaded (10 μmol/L) CD40-activated B cells. Autologous CD40-activated B cells were generated as previously described (27). Briefly, CD40L stably transfected mouse fibroblast cells (t-CD40L⁻ cells, from Dr. John Gordon, MRC Centre for Immune Regulation, University of Birmingham, Birmingham, United Kingdom) were γ-irradiated at 100 Gy, plated at 3.5 × 10⁵ per well in six-well plates in basal Iscove's medium supplemented with 10% FCS, and incubated overnight. CD8 T-cell–depleted PBMC were added at 1 × 10⁶/mL to 2 × 10⁶/mL in the presence of 5.5 × 10⁻⁷ mol/L cyclosporin A (Sandoz Pharmaceutical) for 4 d. Fresh irradiated t-CD40L⁻ cells were added every 4 d. The B cells generated were 75% CD19 positive. T cells were harvested after four rounds of stimulation and were further cloned by the limiting dilution method (at 0.4 and 1 cell per well) using γ-irradiated allogeneic peripheral blood leukocytes at 1 × 10⁵ per well (25 Gy) and B-lymphoblastoid cell lines at 1 × 10³ per well (75 Gy) as feeder cells in RPMI medium, containing 250 IU IL-2 and 1 μg/mL phytohemagglutinin. After 12 d, growing T-cell clones were tested by ⁵¹Cr cytotoxic assay and subsequently expanded in RPMI medium.

T-cell expansion from leukemia patients. Peripheral blood samples were obtained from patients with chronic lymphocytic leukemia following local ethics committee approval. CTL from patient 1 and the HLA-A2–negative patient were generated from CD19-depleted PBMC using dendritic cells from an allogeneic HLA-A2⁺ donor pulsed with PAX5 peptide (10 μmol/L) and stimulated for 1 wk with 10 ng/mL IL-7, following which 20 IU/mL IL-2 was added after 3 d of stimulation. CTL were then stimulated a further week with T2 cells pulsed with peptide and the same concentrations of IL-7 and IL2. The subsequent CTL from patients 2 to 4 were generated by stimulation of CD19-depleted T cells with PBMC pulsed with peptide (10 μmol/L), IL-7, and IL-2 as above for 1 wk. ELISPOT assay is conducted with 1 × 10⁴ cells per well and using T2 pulsed with PAX5 or nonrelated murine peptide as targets (10 μmol/L).

Cytotoxicity assays. Target cells were labeled with 100 μCi Na₂⁵¹CrO₄ in cell culture medium containing 10% FCS for 60 min at 37°C. The cells were washed twice in culture medium. When required, the targets were pulsed with specific or nonspecific control peptides (0.0001–10 μmol/L) for 1 h, washed once in culture medium, and resuspended at 50,000 cells/mL. After 4 to 18 h of incubation, 25 μL of the assay supernatant were placed into a 96-well Lumaplate (Perkin-Elmer). Scintillant (100 μL) was added and radioactivity was counted using a Microbeta scintillation counter (Perkin-Elmer). The results are expressed as percent specific lysis, calculated as (experimental release − spontaneous release / total release − spontaneous release) × 100. For blocking assays, ⁵¹Cr-labeled tumor cells were treated with anti–HLA-A2 or isotype control monoclonal antibody (mAb) for 1 h at 37°C before addition to effector cells. For CD8 blocking, effector cells were incubated with anti-CD8 for 30 min at 37°C before addition to ⁵¹Cr-labeled target cells.

Western blot analysis. Cell lines were lysed in 100 μL of radioimmunoprecipitation assay buffer (Pierce) containing protease inhibitor cocktail (Pierce). Cell lysates corresponding to 30 μg of total protein were separated on 10% SDS polyacrylamide gels and transferred onto nitrocellulose membranes (Hybond enhanced chemiluminescence; Amersham Biosciences Europe). After blockage for 1 h in 5% milk powder in PBS, the membrane was washed and probed with rabbit anti-human PAX5-specific mAb (1 μg/mL; Invitrogen) diluted in 5% dry milk in PBS. Filters were developed with horseradish peroxidase–conjugated antirabbit IgG and enhanced by chemiluminescence (Roche).

Adoptive transfer. Groups of 6- to 8-wk-old “triple knockout” mice (RAG^−/− common γ chain^−/− complement C5^−/−, completely deficient in T cells; ref. 28) were derived from the Institute of Child Health. In the first experiment, 5 × 10⁶ tumor cells were injected s.c. After 3 to 4 wk, with the tumors 2 to 3 mm in diameter, 10 × 10⁶ PAX5-specific T-cell clones were injected i.v. Mice were sacrificed 24 h and 7 d following adoptive transfer and organs were collected. In the subsequent tumor growth rate assays, 10 × 10⁶ SW480 tumor cells were injected s.c. in the back of mice. After 7 d, with the tumors reaching a size of ∼2 mm in diameter, 10 × 10⁶ T cells or PBS were injected i.v. Thereafter every 3 to 4 d, all mice received 25 μg of IL-2 IP until the tumor volume size reached ∼500 mm³. For the analysis of tumor infiltrating cells, tissues were harvested 48 h after adoptive transfer, crushed and passed through a cell strainer, washed with PBS, resuspended in 0.1% BSA/PBS buffer, and incubated with antihuman CD8-FITC or isotype control antibodies before analysis by flow cytometry. Alternatively, tissues were fixed in formalin and analyzed by immunohistochemistry for the presence of infiltrating lymphocytes. Mice were treated in accordance with U.K. Home Office–approved protocols.

High PAX5 expression is limited to hemopoietic cells and cancer. Previous workers have shown that PAX5 is expressed in early B-cell lineages during hemopoiesis (15). Expression in other adult tissues has not been well characterized. We measured PAX5 expression in a panel of adult tissues and cancer cell lines to confirm its limited tissue expression in normal tissues and its relative up-regulation in malignancy. We used quantitative reverse transcription-PCR (RT-PCR) using primers amplifying exon 8, which is thought to be present in most, but not all, splice variants (25). We also chose exon 8 because it encodes the epitope TLPGYPPHV, which was subsequently chosen for targeted immune response generation (see below). As expected, significant expression was seen in B cells and in spleen, presumably reflecting the presence of B cells here. In contrast, in 7 of 13 cancer cell lines tested, the expression at the RNA level was equal with or significantly greater than that seen in spleen, reflecting deregulation of PAX5 in these cells (Table 1). In the remaining nonhemopoietic normal human tissues, low-level expression was observed (range of 0.02–6.9% of the level seen in the relatively low expressing RH18 rhabdomyosarcoma cell line). We performed Western blot analysis of the cell lines to confirm that high-level expression, as shown by quantitative RT-PCR, corresponded with high protein expression (Supplementary Fig. S1). We concluded that there is a range of deregulated expression in human malignancy, including non–B-cell cancers, and that specific immune responses directed against PAX5 would not be predicted to cause autoimmune toxicity apart from B lineage hemopoietic cells.

PAX5-reactive T-cell lines can be generated from the autologous repertoire. Using two computational algorithms (SYTHPETHI and BIMAS), we identified three sequences from human PAX5 predicted to bind HLA-A2 strongly (Table 2). However, on testing the affinity of these peptides in T2 binding assays, only the TLPGYPPHV peptide (hereafter referred to as TLP) showed significant stabilization of HLA-A2 on T2 cells (Fig. 1A). The TLP epitope is derived entirely from exon 8 of PAX5. Immune responses directed against TLP would therefore be predicted not to target normal body tissues (Table 1), with the exception of B cells, but to be cytotoxic against a range of PAX5 exon 8–expressing cancer types.

We next investigated whether the T-cell repertoire of normal blood donors contains CTL with avidity for the HLA-A2/TLP complex, by generating T-cell lines from buffy coat preparations from five normal blood donors. We generated CTL lines using CD8 positively selected cells, CD3 positively selected cells, or unselected PBMC as the starting cell population and used a standardized T-cell expansion protocol. This involved four stimulations, at weekly intervals, with peptide-pulsed antigen-presenting cells as described previously (7). In the first two stimulations TLP peptide is pulsed onto autologous dendritic cells, whereas in the third and fourth stimulations TLP-pulsed autologous B cells were used as antigen-presenting cells. The killing activity of all five CTL lines against TLP-pulsed T2 cells ranged from 10% to 80% at E/T ratio of 30:1 (Table 3). We assessed the frequency of TLP-reactive T cells in these CTL lines and in PBMC of HLA-A2–positive blood donors by staining with a PE-labeled pentamer comprising TLP and the HLA-A2 complex. The TLP-reactive T cells were undetectable in PBMC from normal donors, whereas 2% to 5% of cells in the best T-cell line (donor BC-24) were stained with the pentamer (Fig. 1B). To confirm that the donor BC24–derived CTL line recognizes endogenously processed PAX5, we transiently transfected 293-T cells (HLA-A2 positive, PAX5 negative) with a cDNA construct expressing full-length PAX5 (all exons included) and used the paired transfected and untransfected cell populations as targets in a standard chromium release assay. Specific killing was seen only against TLP-pulsed T2 cells or the PAX5-transfected cells (Supplementary Fig. S2), confirming that the TLP epitope can be intracellularly processed from the PAX5 polypeptide and presented on the cell surface in HLA-A2–positive human cells.

CTL clones directed against PAX5 can specifically lyse PAX5-expressing cancer cells. To assess specific cytotoxicity against human PAX5-expressing cancer cells, we generated CTL clones from one of the TLP CTL lines (from donor BC24) by limiting dilution. We initially screened for killing following addition of T2 cells pulsed with TLP peptide at concentrations of 0, 0.01, and 10 μmol/L by chromium release assay. Five high-avidity clones from ∼300 screened were found to exhibit specific killing at 0.01 μmol/L and no killing at 0 μmol/L, and three were further characterized to determine the avidity of the interaction with the TLP:A2 complex. The clone with the highest avidity (clone 33) specifically lysed T2 cells pulsed with as little as 100 pmol/L peptide (Fig. 2A). We confirmed that these cells had the surface phenotype of CTL by staining with CD8 and an MHC pentamer specific for HLA-A2 and TLP (Fig. 2B) and by showing positive staining with pan α/β T-cell receptor antibody (data not shown). Clone 33 also caused lysis of 293 cells transfected with PAX5, but not vector-transfected controls, confirming the specificity of its killing activity (Fig. 2C).

Therefore, anti-PAX5 clone-33 was selected to determine the spectrum of the cytolytic activity of anti-PAX5 CTL. The highest specific killing in standard chromium release assays was seen against the HLA-A2–high and PAX5-positive cancer cell lines NALM6, RH18, and SW480 (Fig. 2D). Importantly, Raji (PAX5 positive, HLA-A2 negative), MEL5 (PAX5 negative, HLA-A2 low), and MDA-MB-231 or 293-T (PAX5 negative, HLA-A2 positive) were not killed by the clone, confirming its specificity (Fig. 2D; Table 4). LAN-1 and SH-SY5Y cells were only killed significantly after the addition of IFNγ, which up-regulated class 1 MHC (Supplementary Fig. S3A). Cytotoxic specificity was further confirmed by showing that clone 33 killing of SW480 cells was inhibited by anti-CD8 and anti–MHC class 1 antibodies (Supplementary Fig. S3B).

Anti-PAX5 CTL activity against primary cells and primary tumors. To determine whether the PAX5-reactive CTL were also capable of specific recognition and lysis of primary human cancer cells, we made use of two primary pre-B leukemia samples (G7 and PL1), which were available to us and known to be HLA-A2 positive (data not shown) and PAX5 expressing (Table 1). PL1 cells had been engrafted in a nonobese diabetic-severe combined immunodeficient mouse model from a primary pre-B-cell leukemia sample (29), whereas G7 cells were a short-term in vitro culture from a patient with pre-B-cell acute lymphoblastic leukemia. Both of the primary patient leukemia samples expressed PAX5 exon 8 RNA (Table 1), and both were lysed in vitro by anti-PAX5 clone 33, with killing blocked by an anti–HLA-A2 mAb (Fig. 3A). To test the cytotoxicity against PAX5-expressing primary B cells, we positively selected CD19 cells from an HLA-A2–positive donor and tested them before and after 25 days of in vitro stimulation with CD40 ligand. Freshly isolated CD19-positive cells from the peripheral blood of HLA-A2–positive blood donors had higher PAX5 expression than CD40 ligand cultured B cells (Table 1). In accordance with this differential PAX5 expression, whereas killing of freshly isolated B cells was observed at a similar level to the primary leukemia samples, only a relatively low level killing of activated B cells was observed (Fig. 3B).

We next determined whether there was evidence of anti-PAX5 CTL activity in patients with PAX5-expressing leukemia. To this end, PBMC from HLA-A2–positive patients with chronic lymphocytic leukemia were stimulated with TLP peptide for 2 weeks and tested for specific IFNγ secretion following coculture with TLP-pulsed T2 cells. In four of four patients, there was an increase of IFNγ secretion following addition of TLP peptide to T2 targets, and this was reversed following addition of an HLA-A2 blocking antibody. In an HLA-A2–negative chronic lymphocytic leukemia (CLL) patient, however, there was no evidence of increased IFNγ secretion (Fig. 3C). Therefore, there is evidence of PAX5-specific T cells in the blood of HLA-A2–positive patients with PAX5-expressing leukemia.

Anti-PAX5 CTL infiltrate PAX5-expressing tumor in vivo to inhibit tumor growth. To investigate whether the anti-PAX5 T cells were capable of infiltrating into PAX5-expressing tumors, we adoptively transferred 10 × 10⁶ clone 33 anti-PAX5 T cells into the tail vein of immunodeficient triple knockout (RAG^−/− common γ chain^−/− complement C5^−/−) mice bearing small palpable s.c. tumors of the PAX5-positive, HLA-A2–positive SW480 cell line. Mice were sacrificed at 48 hours or 7 days following T-cell infusion, and internal organs and tumor were analyzed for the presence of T cells by staining for CD8. At both time points, a detectable T-cell population was seen only in the tumor and, to a lesser extent, the draining lymph nodes, but no significant numbers of T cells were seen in spleen, lung, liver, or kidneys (Fig. 4A). This is consistent with the T cells accumulating at the site of PAX5 overexpression.

To investigate whether T-cell infiltration could inhibit tumor growth, further adoptive transfer experiments were performed in which SW480 tumor–bearing mice were treated with clone 33 T cells and given IL-2 i.p. every 3rd day to promote T-cell survival. Control mice were given nonspecific activated T cells and IL-2, or IL-2 alone. Remarkably, whereas IL-2 alone and nonspecific T cells had no significant effect on tumor growth, clone 33–treated mice developed only small tumors up to 25 days after tumor challenge (Fig. 4B). At 48 hours following adoptive transfer, an infiltrate of lymphocytes surrounded the outside of the tumor in both clone 33– and control CD8 cell–treated mice, but this was not apparent in the untreated tumors; fluorescence-activated cell sorting staining indicated a larger number of infiltrating T cells in clone 33– versus control CD8–treated mice (Fig. 4C). Therefore, T cells were attracted toward SW480 tumors, but only PAX5 targeted T cells were capable of preventing tumor outgrowth. After 25 days, inhibition of tumor growth was not associated with any histologic evidence of organ damage in clone 33–treated mice as shown by normal morphologic appearances of liver, spleen, and kidneys and normal peripheral blood appearance compared with untreated controls (data not shown). Importantly, in parallel adoptive transfer experiments, clone 33 had no effect on the in vivo growth of the PAX5-negative HLA-A2–positive tumor cell line MDA-MB-231 compared with IL-2 alone–treated mice (data not shown).

PAX5 is herein identified as a potential target for cancer immunotherapy by virtue of its limited normal developmental expression but high-level expression in a range of adult and pediatric cancers. Central to the argument for identification of a clinically useful new tumor antigen is the demonstration of a marked differential expression between neoplastic and normal tissues. The low expression levels in normal tissues we have shown are generally supportive of potential clinical use. No tissues have the high degree of expression seen in B-cell malignancies apart from hemopoietic tissues (normal B cells, spleen, and bone marrow) and testis (an immune privileged site). We hypothesize, but have not formally proved, that the high expression in bone marrow and spleen reflects B hemopoiesis.

It would therefore be important to investigate potential hematologic toxicity before clinical use. Previous workers have shown that PAX5 is critical for B-cell specification and for maintenance of B-cell phenotype, such that loss of PAX5 is tumorigenic for B-cell leukemia (15, 30). We have not been able to evaluate the toxicity of PAX5 immunotherapy in a murine model with targeting of the human TLP epitope because its antigenicity is restricted to human HLA-A2. The clinical consequence of PAX5-targeted therapy resulting in destruction of normal B cells might be the requirement for long-term immunoglobulin replacement. Another concern about PAX5 immunotherapy would be the emergence of PAX5 antigen loss variants among B cells predisposing to B-cell malignancy. These issues will require formal testing in appropriate animal models (e.g., humanized HLA-A2 transgenic mice). Given the expression observed in normal bone marrow, it is important to confirm that PAX5 is not seen in normal hematopoietic stem cells. However, Pax5^−/− mice have normal hemopoiesis apart from B cells, strongly suggesting that PAX5 has little significant expression outside of B-cell hemopoiesis.

Our in vivo data show that anti-PAX5 CTL accumulate in the tumor site of PAX5/HLA-A2–positive tumor cells. Moreover, accumulation is associated with a significant slowing of tumor growth despite the lack of T-cell help. It is interesting to speculate that in the immunocompetent setting, it would be possible to boost anti-PAX5 immune responses using peptide vaccinations to achieve long-term disease control. Further work is also required to show the therapeutic potential in a larger number of cancer types.

The ability to generate effective anti-PAX5 CTL for all of the normal blood donors tested was slightly surprising and maybe suggests a lack of tolerance to PAX5, confirmed by the presence of anti-PAX5 T precursors in four of four patients with CLL. However, the precursor frequency of anti-TLP T cells in the peripheral blood of normal HLA-A2–positive donors was below the level of detection by pentamer staining. Future studies to prospectively to collect samples and test for precursor frequency at diagnosis will be of interest.

Anti-PAX5–directed T-cell therapy has potential clinical application in a range of adult and pediatric malignancies. Especially attractive is the prospect of generation of vectors for gene therapy encoding high-affinity T-cell receptors directed against PAX5, with a view to transducing the patient's autologous repertoire and adoptively transferring the redirected cells back into the patient. The clone with the highest affinity described in this study (specific killing observed at as little as 100 pmol/L pulsing of T2 cells) might be of sufficiently high affinity for clinical benefit in this setting, but further in vivo evaluation is required.

No potential conflicts of interest were disclosed.

Grant support: Sport Aiding Medical Research for Kids, Cancer Research UK, and Research into Childhood Cancer.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Akira Harashima for provision of NALM cell lines, Dr. Amit Nathwani for help with obtaining blood samples from CLL patients, and Joanna Buddle for assistance with flow cytometry.