Serum from mice immunized in the context of Treg inhibition identifies DEK as a neuroblastoma tumor antigen

Advanced neuroblastoma poses a grave clinical challenge and still awaits effective therapy. Early clinical observations, combined with a slight but demonstrable positive impact of bone marrow transplantation on outcome has motivated the development of immune approaches to therapy [1‐4]. In murine models of human neuroblastoma, anti-tumor immunity can be generated using cell-based vaccines where tumor cells have been genetically modified to express soluble cytokines or cell-surface immunostimulatory molecules [5‐7]. Our own work has demonstrated that cancer cell-based vaccines expressing multiple immune co-stimulatory molecules in the murine neuroblastoma cell line AGN2a can transform this tumor cell line in to a vaccine that induces strong cell-based immunity to the unmodified parental cell line [8, 9]. Based on the ability to induce an immune response with cancer cell-based vaccines, human trials with neuroblastoma patients have been carried out [10]. Although these cell-based cancer vaccines did not prove immediately effective, they were demonstrated to be safe and are ripe for further optimization [11].

In experimental systems, immunity to neuroblastoma can be amplified by the blockade of T-regulatory cell (Treg) function with anti-CD25 antibody (B.D. Johnson, et al., 2007, J. Immunother., in press). Treg are known to suppress the immune response to self-antigens, including tumor-self antigens, and thwarting this tolerogenic role by their depletion has become a major focus in the development of new immunotherapeutic strategies to treat human malignancy [12, 13]. Golgher et al. have demonstrated that CD25+ T cell depletion uncovers immune responses to the tumor cell type used as a vaccine, and importantly that this response broadens to include other syngeneic tumor cell types [14]. Given the ability to induce immune recognition of what are normally considered "self" antigens upon Treg blockade, we reasoned that treatment of experimental animals with cell-based cancer vaccines in the context of anti-CD25 antibody treatment would induce a strong anti-neuroblastoma immune response. The proposed use of serology to uncover T cell antigens is supported by the recent description of antibody as well as T cell responses to the DBY minor histocompatibility antigen in allogeneic stem cell transplantation [15, 16]. The breaking of tolerance to self-antigens with Treg depletion may be functionally analogous to the anti-tumor effect seen in allogeneic bone marrow transplantation, whose primary side-effect, graft-versus-host disease, is evidence that tolerance to normal self antigens has been modified.

The serological analysis of recombinant cDNA expression libraries (SEREX) constructed from patient tumor was established by Sahin and Tureci who demonstrated that this process identifies T-cell antigens as well as B-cell antigens [17, 18]. SEREX continues to be employed in patient studies and has even proven to identify intracellular antigens targeted by the immune system [19]. The identification of the NY-ESO-1 antigen in patients by SEREX demonstrated that both MHC class II restricted epitopes and MHC class I-restricted (HLA-A2) epitopes, targets of cytotoxic T cell responses, could be identified with this technique [19]. We present a new means to identify immunogenic tumor antigens. In this report we employ serum from experimental animals that have been vaccinated in the context of anti-CD25 antibody treatment, as opposed to using sera from tumor-bearing animals, which would be the equivalent of using serum from newly diagnosed patients. The use of immune serum-SEREX has allowed us to identify new tumor-associated antigens in our neuroblastoma model. Notably, we demonstrate that one of the antigens identified by our immune-SEREX approach, the DEK oncoprotein, induces a T cell response as well as an antibody response. Translation of this concept in to clinical studies would require the used of post-treatment or even post-vaccination serum, as opposed to the initial samples commonly harvested at the time of diagnosis.

The modification of syngeneic tumor cell lines with immune co-stimulatory molecules can transform lethal tumor cell lines into effective vaccines. The immunization of A/J mice with a cell-based vaccine, AGN2a-CD80/CD86, has been demonstrated to induce a strong cellular immune response [8]. This immune response is enhanced when the vaccine is given in the context of anti-CD25 antibody (clone PC61) as a means to block and/or deplete T-regulatory cells. We sought to explore the specificity of the immunoglobulin response to this cell based vaccine, as a means to potentially uncover new tumor-associated antigens. To do so, we constructed a cDNA library using mRNA purified from the AGN2a cell line and expressed proteins encoded by tumor- derived mRNA in E. coli using the λZAPII phage system. To identify clones expressing antigenic epitopes recognized by IgG present in immune sera, a total of 2 × 10⁵ clones were screened. Fifty-eight unique plaques were identified as expressing proteins that could be recognized by IgG antibody when serum from AGN2a-CD80/CD86+PC61 treated mice was used at a dilution of 1:250. When pooled serum (n = 5 for each group) from naïve mice, tumor-bearing mice, mice treated with PC61 alone, or mice treated with two weekly injections of the vaccine cell-line AGN2a-CD80/CD86 alone was used to screen the cDNA library 0, 0, 0, or 5 unique clones, respectively, were identified. Antibody-reactive plaques identified by serum from the AGN2a-CD80/CD86+PC61-treated mice were picked, used to re-infect E. coli, and once more serologically screened. A third round of plaque purification was not required, as it always resulted in uniform recognition of phage-expressed proteins. In vivo excision was performed with helper phage and phagemid DNA isolated. Each phagemid insert was DNA sequenced using both T7 and T3 primers, and this sequence was used to search on-line databases. The 58 clones identified by immune serum-SEREX screening encoded 25 individual proteins, as listed in Table 1. Most notable was the frequency with which DEK was detected. While all antigens listed in Table 1 were detected once from a single positive clone, the DEK antigen was detected 34 separate times. Careful inspection of the cDNA library clones demonstrates that most transcripts seem to have some role in neuronal differentiation, cell cycle control, or have previously been identified as transcripts that are over-expressed in other malignancies. The final 4 transcripts listed in Table 1 have no published or annotated functional analysis associated with them, beyond their naming and characterization using bioinformatic techniques. Primers were generated for each clone identified, and expression of each verified by PCR analysis of AGN2a-derived cDNA (not shown).

To further determine the significance of the SEREX identified transcripts, we performed cDNA microarray analysis to examine the expression level of these genes in mouse neuroblastoma cell lines Neuro-2A and TBJ or in mouse tissues. A total of 22 out of 25 genes were present in the array and were used for hierarchical clustering analysis, Figure 1. Importantly, the majority of the transcripts identified by SEREX are over-expressed in neuroblastoma cell lines, indicating the ability of immune serum-based SEREX analysis to identify over-expressed proteins using serum from vaccine-treated mice. Of note, DEK is indeed over-expressed in the neuroblastoma cell lines tested.

The cluster analysis displayed in Figure 1, identifies four different categories of neuroblastoma antigen expression profiles. In Cluster 1, transcripts are over-expressed in the mouse neuroblastoma cell lines in a tumor-restricted manner: Chgb, Dus31, 2310042G06Rik, Psma6 (proteasome subunit), Csnk2a1 (casein kinase), Mtrr (metabolic activation), and Hspa8 (heat shock protein). An interesting sub-group within this cluster includes Chgb, chromogrannin B, which is known to be associated with sympathetic neurons and which is also found in adrenal tissue. In Cluster 2 we find transcripts that are overexpressed in neurboastoma cells lines as well as spleen or lung: CD9 (tetraspanin, neuroblastoma marker), Plod3 (collagen biosynthesis), Fusip1 (neuronal differentiation), Brd2 (transcription), Tuba2 (tubulin alpha), DEK, Prpf38b, Hmgb1 (inflammatory mediator), and Serbp1 (serpine). Some of the proteins in this group, like CD9 and Hmgb1function as immune modulators. Cluster 3 (Crlf3 and Mier 1) and Cluster 4 (Atp5c1, Suv39h1, Immt, Slca6) are not over-expressed in neuroblastoma cells lines and seem unlikely to play a role in neuroblastoma. They may either be cross-reactive antigens or specific targets of auto-immune responses. We propose that each of the Cluster 1 and Cluster 2 transcripts should be explored with regard to tumorigenesis and immunogenicity.

Before we explored the specific immune response to DEK we sought to demonstrate that the immune response to a cell-based tumor vaccine is fundamentally altered by treatment with Treg-depleting antibody. The was first demonstrated qualitatively, whereupon western blotting of AGN2a cell lysates with serum from naïve, vaccinated (AGN2a-CD80/CD86), or vaccinated and Treg-depleted mice, different protein bands were recognized by the different sera, figure 2. Although prominent bands were seen with both the vaccine only group (figure 2A, lane 4), these differed from the prominent bands in the mice vaccinated in the context of Treg depletion (lane 5). To examine the differences in the response to a single protein we compared the antibody titer in an ELISA assay using recombinant DEK as the target and the same set of sera. Recombinant DEK protein was produced in the pET15b bacterial vector encoding an amino-terminal 6-histidine sequence to facilitate purification on a nickel affinity column. Enhanced green fluorescent protein (EGFP) was produced as a negative control. To obtain sufficient protein for immune function assays pET15b vectors were used to transform the BL21 (DE3) pLysS E. coli strain, protein expression induced with IPTG, and proteins containing the 6x-histidine tag purified from bacterial lysates using a Ni-NTA column. When columns were eluted with imidazole, distinct bands of 55 kD for DEK and 30 kD for EGFP were readily resolved by SDS-PAGE and Coomassie blue staining of the resolved proteins, Figure 3A.

We then tested if purified DEK was recognized by anti-human DEK antibody and anti 6x-histidine antibody. Western blot analysis demonstrated that bacterially-expressed mouse DEK was recognized by anti-human DEK antibody and that anti-6x-histidine antibody recognized the histidine tag encoded by both recombinant proteins, Figure 3B. Having confirmed that the anti-human antibody recognized murine DEK, we looked at two cellular targets for the expression of DEK by immunofluorescence (IF). We analyzed both the wild-type/parental AGN2a cell line and AGN2a transfected with pcDNA3.1-Hygro/DEK (AGN2a/DEK), in order to induce over-expression of the DEK oncogene. Both AGN2a and AGN2a/DEK showed strong nuclear immunoreactivity by IF, with AGN2a/DEK showing more intense staining, Figure 4. While the expression of DEK was variable in AGN2a, strong expression was seen in every AGN2a/DEK transduced cell, as demonstrated by exact overlap of the DAPI and DEK signals. To further confirm that the phage-expressed DEK cDNA was a genuine immune target we tested serum from tumor-vaccinated mice for the presence of DEK-reactive immunoglobulin by ELISA. Using polystyrene plates coated with recombinant DEK, DEK-reactive IgG could be readily detected in serum from mice immunized with AGN2a-CD80/86 in the context of PC61 treatment, Figure 5. Anti-human DEK antibody was used as a positive control antibody in the ELISA. This assay also supports our hypothesis that vaccination in the context of Treg depletion alters the nature of the immune response to cell-based tumor vaccines. High titer DEK antibody production required the depletion of Treg, while vaccination with the cell based vaccine, AGN2a-CD80/CD86, did not produce anti-DEK IgG.

To investigate the CD8⁺ T cell response against DEK in mice that had received a cell-based cancer vaccine, A/J mice were given two weekly subcutaneous (s.c.) injections of 2 × 10⁶ irradiated AGN2a-CD80/CD86 cells cultured in DMEM supplemented with 2% mouse serum or 10% FBS, and Treg activity inhibited by the i.p. injection of clone PC61 anti-CD25 antibody three days prior to the first vaccination. Mouse serum-cultured vaccine cell lines were used in order to control for potential FBS reactivity, which has been reported in the analysis of CD4 T cell responses. AGN2a-CD80/86 was cultured in normal mouse serum at least two weeks prior to immunization studies. Splenocytes were collected 5 days after the second vaccination, depleted of RBCs, and CD8⁺ T cells purified by immunomagnetic selection. T cell responses to DEK were detected in IFN-γ ELISPOT assays using two different sources of antigen-presenting cells. In the first assay, peritoneal exudate cells (PEC, which are primarily composed of macrophages) loaded with recombinant DEK protein were used. Harvested PEC were plated in ELISPOT plates, incubated with recombinant DEK or EGFP for 4 hours at 37°C, and then immune CD8⁺ T cells added to the plates for an additional 18 hours. Control cultures containing T cells plus non-protein-loaded PEC, or T cells plus EGFP-loaded PEC, had similar low numbers of IFN-γ-producing cells, Figure 6A. In contrast, significant anti-DEK reactivity was seen in T cells from mice immunized with AGN2a-CD80/86 and treated with PC61, whether the vaccine cell line had been cultured in FBS or normal mouse serum (ms), Figure 6A.

The second source of antigen-presenting cells used to monitor anti-DEK responses was the AGN2a cell line we produced that over-expresses DEK (Figure 2). CD8⁺ T cells were purified from naïve mice, mice treated only with PC61 (anti-CD25 antibody), or from mice that received both the AGN2a-CD80/86 vaccine and PC61. While naïve or PC61-treated mice did not show any IFN-γ ELISPOT activity in response to AGN2a, vaccinated mice showed relatively strong ELISPOT reactivity against both unmodified AGN2a tumor cells and heightened responses against the DEK-over-expressing cell line, Figure 6B. Taken together these assays demonstrate that anti-DEK CD8 T cell responses are induced by our vaccine+PC61 immunization protocol. This also validates that our serologic cDNA screening assay, based on the use of immune serum rather than tumor-onset serum, can identify important T cell epitopes.

SEREX analysis has been used to identify tumor-associated antigens in a number of malignancies. Primarily, patient serum has been used to screen tumor-derived cDNA libraries. The immunological rationale for the SEREX approach is that unique tumor antigens should induce an antibody response. However, the induction of tolerance to tumor antigens is now recognized to be a formidable obstacle to inducing anti-tumor immunity. It may well be that the antibody specificities present in tumor-bearing patients or animals represent antigens to which a tolerized, or non-tumoricidal immune response has been generated.

We have used serum from mice immunized twice weekly in the context of Treg-depletion to carry out a SEREX analysis, reasoning that unique neuroblastoma antigens may be uncovered in vaccinated, as opposed to tumor-bearing, animals. When an AGN2a cDNA library we prepared was screened for the ability to express IgG-reactive antigens, a number of transformation-associated proteins were identified, Table 1. These antigens were over-expressed in two other murine neuroblastoma cell lines as well, when compared to normal tissues, Figure 1, demonstrating that immunization in the context of Treg inhibition can identify unique and potentially important transcripts. What was most striking about our data was the abundant number of "hits" that were generated against the DEK oncogene. DEK was indeed over-expressed in neuroblastoma cell lines, Figure 1, and was recognized by immune serum in ELISAs, Figure 5. Our SEREX analysis also identified some antigens that were not over-expressed in neuroblastoma, yet these antigens still induced an antibody response. This response may either be due to the generation of cross-reactive antibody, or the antigens may be targets of an induced autoimmune response. It is also possible that point mutations in these antigens may have generated an immune response to them. Direct sequence analysis of these proteins will be required to confirm or refute this possibility and will be explored in future studies.

The combination of Treg depletion and vaccination may also induce CD8 effector cells. Recent descriptions of auto-reactive CD4 T cells that have expanded in vitro upon Treg depletion from healthy individuals including those specific for NY-ESO-1, tyrosinase, and GAD65 (a type 1 diabetes-associated autoantigen) support this hypothesis and suggest that Treg modulation may be essential for inducing anti-tumor immunity [20, 21].

The identification of DEK as a tumor-associated antigen in murine neuroblastoma cell lines is fascinating due to the expression of DEK in an expanding number of human cancers. Both DEK and E2F3 have been identified as over-expressed transcripts due to chromosome 6p gains in retinoblastoma, a common pediatric malignancy [22]. Genomic gains of oncogenes like DEK are likely to be selected for because they confer a growth advantage to the malignancy, and 6p gains have been described in a number of malignancies including bladder and gastroesophageal cancer, osteosarcoma, and melanoma [23‐26]. DEK was first identified as part of a fusion protein with the CAN/NUP 214 nucleoprotein in an acute myeloid leukemia, AML, sub-type through an "in-frame" translocation of chromosome 6 and 9 [27]. Biochemically, DEK is known to regulate transcription and contains a DNA binding domain, several phosphorylation sites, and a SAF-Box (scaffold attachment factor). DNA binding is dependent on phosphorylation by casein kinase 2 (CK2) which appears to regulate its transcriptional regulatory function [28‐30]. Of note, we also identified CK2 by immune serum-SEREX, Table 1. Most recently DEK was demonstrated to reside in the spliceosome, and that upon phosphorylation, DEK associates with U2AF, enforcing 3' splice site discrimination, preventing U2AF from binding to pyridine tracts not followed by AG sequence [31]. The association with DEK over-expression or inactivation with human disease may be related to alterations in splice-site recognition and intron removal [32]. Evidence that DEK regulates key genomic responses to DNA damage was demonstrated by the ability of a partial fragment of DEK, isolated by a cDNA library screen, to complement an ataxia-telangiectasia phenotype in vitro [33]. In our immune-SEREX screen we identified partial transcripts as well, generated from an internal ATG sequence, that were recognized by immune sera (not shown). The association of DEK with cellular transformation has left little doubt of its oncogenic potential. DEK has been shown to inhibit senescence in cells infected with high-risk papillomavirus and to associate with the latency protein LANA-1 expressed by KSHV (Kaposi's sarcoma-associated herpesvirus) [34, 35]. In the AML subset containing the t(6;9)(p23;q34) chromosomal abnormality, monitoring of the DEK-CAN transcript by RT-PCR shows an exact correlation to therapeutic outcome [36].

DEK is a nuclear protein, and antibody responses to DEK have been proposed to be indicators of autoimmune disease. However, in a report by Dong et al., it was proposed that anti-DEK antibodies are not a marker for any specific disease, but a marker for a subset of autoimmunity associated with IFN-γ production [37]. This is fascinating, as the breaking of tolerance to a self/tumor-associated antigen, as demonstrated by an IFN-γ mediated immune response, is a key characteristic of generating Th1 immunity. Further evidence for the ability to generate an immune response specific for DEK was seen when a human CD4 T cell clone specific to the DEK-CAN fusion protein produced IFN-γ upon co-culture with dendritic cells loaded with either apoptotic or necrotic t(6;9) leukemia cells [38]. Another potential mechanism for the induction of an immune response to self-proteins is the production of a truncated transcript by the tumor itself. Immune responses to a truncated HER-2/neu protein are far greater than to native protein and have become a new focus for vaccine development [39]. Although full-length recombinant DEK was used throughout our studies, the isolation of a partial transcript from the AGN2a cDNA library leaves this a possibility in our in vivo vaccine studies.

SEREX-based analysis of human cancers is entering a new phase of development. Studies from the laboratory of Dr. L. Old are beginning to describe how in different clinical situations, it is the SEREX-identified antibody targets that define discrete sets of over-expressed protein antigens that predict tumor pathophysiology. Notably, their current hypothesis is that SEREX specifically identifies tumor/self antigens that are recognized by CD4⁺CD25⁺ regulatory T cells, and that without induction of an inflammatory immune response that includes CTL activity, immunization with SEREX-identified antigens alone may actually enhance tumor progression [40, 41]. Our approach avoids this concern, in that tumor antigens we identify in the context of cell-based vaccination and Treg inhibition occur in a physiological environment where protective cytolytic T cell responses are being induced by the tumor cell-based vaccine [8]. We anticipate testing our hypothesis in human subjects immediately after the initial course of chemotherapy, where tumor antigen has been newly loaded in to the antigen-presenting cells, or following specific immunotherapeutic trials.

We have presented a new rationale for tumor antigen identification using serum from mice vaccinated in the context of altered tolerance, by blocking the function of Treg cells. This procedure, which was developed for inducing strong T cell responses, also generated a strong serological response. Few if any antigens were identified in tumor-bearing, or Treg-blocked only animals. However, vaccination in the context of Treg depletion produced serum from which we were able to generate a candidate tumor antigen list, and we demonstrate that for one of these antigens, the DEK oncogene, a strong T cell response is also induced. A number of these antigens are associated with tumorigenicity and should be explored in their own right. We conclude that tolerance mechanisms are operative in tumor-bearing animals, and that these may block effective tumor antigen identification by serological screening of cDNA libraries (SEREX). To extrapolate these findings to human disease, serology-based antigen discovery should be carried out not with onset or diagnostic serum (which is most commonly banked), but with serum derived from treated patients in which antigen loading in to the immune system is optimal and in which Treg effects may be minimized.