Genetic vaccines against Ep-CAM break tolerance to self in a limited subset of subjects: Initial identification of predictive biomarkers



The epithelial cell adhesion molecule, Ep-CAM, has been historically considered a target of passive immunotherapy using monoclonal antibodies, and more recently, of a first Pox-vector-based cancer vaccine Phase I trial in colorectal cancer patients. To shed further light on the use of this antigen, we isolated the mouse and rhesus homologues of human Ep-CAM and explored different genetic vaccination modalities based on the use of adenoviral vectors as well as DNA electroporation (DNA-EP). Immune responses to Ep-CAM were measured by IFN-γ ELISPOT and intracellular staining assays using overlapping sets of peptides covering the entire coding regions. We found the most powerful vaccination regimen to be constituted by DNA-EP-prime/Adeno-boost mixed-modality protocols. Vaccination in rhesus macaques resulted in breakage of immunological tolerance in a minority of cases. Similarly, a low frequency of responders was observed with the mouse Ep-CAM vaccine in outbred CD1 mice. When immunized CD1 mice were analyzed for MHC haplotype and TCR expression levels, we observed that immune responders all had the same q/q MHC class I haplotype and showed higher expression levels of the TCRVβ4 and TCRVβ8 T cell receptors. Our results underscore the current limitations in our capacity to induce efficient cancer vaccines against self antigens like Ep-CAM, but also represent a first effort to identify predictive biomarkers of response.

See accompanying commentary:




colorectal cancer


DNA electroporation


epithelial cell adhesion molecule


carcinoembryonic antigen


intracellular cytokine staining


open reading frame


rhesus Ep-CAM


spot-forming colony


tyrosinase-related proteins


Cancer vaccines are being investigated as innovative approach to cancer therapy. The field is currently flourishing following the discovery of a wealth of tumor-associated antigens, but also as a consequence of our increased understanding of the immune system and of the cellular and molecular events involved in the priming and expansion of acquired immune responses. In spite of these advances, and of the demonstration of sporadic anti-cancer objective responses in several clinical trials of cancer vaccines, demonstration of a statistically significant impact of vaccines on the course of the disease and extension of survival is still lacking. This is due to several limiting factors which include (a) the absence of powerful and reliable technologies capable to induce strong cell-mediated immune responses; (b) the lack of immune correlates of clinical responses; (c) our inability to select patients that are more likely to mount immune responses to tumor antigens and, as a consequence, to benefit from vaccine administration.

The epithelial cell adhesion molecule, Ep-CAM, is a 40-kDa transmembrane glycoprotein that consists of two epidermal growth factor-like extracellular domains, a cysteine-poor region, a transmembrane domain, and a cytoplasmic tail. Expression of Ep-CAM in human adult tissues is largely restricted to epithelial cells, with a few exceptions such as squamous epithelium and some epithelium-derived cells such as hepatocytes 1. This antigen is believed to be involved in homotypic calcium-independent cell-cell adhesion, as one of several classes of intercellular adhesion molecules 2. Additional studies demonstrated Ep-CAM colocalization with the intracellular actin cytoskeleton, and also found that selective mutational analysis of the cytoplasmic domain resulted in lack of adhesion 3.

Of particular interest is Ep-CAM overexpression in carcinomas of various origins including the colon and rectum, prostate, lung and breast 4. Recently, a comprehensive overview was reported using multitumor tissue microarrays where Ep-CAM expression was observed in 98 of 131 tumor categories 5. Different hypotheses have been postulated to explain Ep-CAM expression in different tumors, but its role therein is unclear. Additionally, RNA interference experiments showed that it may also play a role in cell proliferation and migration, at least in breast tumor cell lines 6.

Autoantibodies against Ep-CAM were observed in particular in metastatic colorectal cancer (CRC), although no correlation was observed with survival 7. In line with this observation, it has been reported that a natural cell-mediated immune response can be observed in metastatic CRC patients, but not at early cancer stages 8, 9. The biological significance of these findings remains unclear and may suggest a natural defense against a tumor antigen or merely a sign of recognition by the immune system. Due to its association with tumors, Ep-CAM has been the target of immunotherapy in CRC patients 10. Recently, a CD8 response was observed in CRC patients treated by genetic vaccination with Ep-CAM in combination with GM-CSF 11.

The evaluation of human Ep-CAM (hEp-CAM) vaccines in preclinical murine models showed that vaccination with this antigen can elicit an antibody response as well as a cytotoxic T lymphocyte (CTL) response, which were associated with partial protection in tumor challenge experiments 12, 13. To date, however, only one report has described tumor protection induced by the self mouse Ep-CAM (mEp-CAM) vaccination in the presence of a high dose of IL-2 14.

DNA electroporation (DNA-EP) was shown to increase the efficacy of plasmid DNA vaccination in viral vaccines 15. It was also shown in rhesus macaques that a mixed vaccination modality, such as DNA followed by adenovirus, later may elicit a more effective immune response including a persistent CD8 response 16. We have previously assessed the efficacy of a mixed-modality DNA-EP/Adeno against carcinoembryonic antigen (CEA) in hCEA-transgenic mice and showed this vaccine to be able to induce anti-CEA cell-mediated immunity as well as partial protection from subsequent tumor challenge 17. In the present study, we have therefore started to evaluate Ep-CAM as immunogen when delivered by DNA-EP and adenovirus. To this aim, we generated vectors carrying human, mouse and rhesus homologues and tested their immunogenicity in self and non-self contexts. While DNA-EP and Adeno vaccines induced potent immune and anti-tumor responses against human Ep-CAM in mice, breaking tolerance to the self antigen in rhesus and mice proved to be particularly difficult and was observed only in a subset of vaccines. A careful analysis of mouse immune responders allowed a correlation of the immune response with particular genetic and phenotypic markers. Our data, while providing a further evidence of our current limitations in the development of cancer vaccines, suggests that strategies aiming at the identification of predictive biomarkers may be necessary to select patient subsets where efficacy can be clinically demonstrated.


Adeno and DNA-EP vaccines induce immune responses against hEp-CAM in mice

We first set out to assess the capacity of plasmid DNA-EP 15 or Adeno-based vaccines to induce an immune response to Ep-CAM. For this purpose, we cloned the full-length hEp-CAM into an adenovirus (Ad) serotype 5 vector to generate Ad-hEp-CAM, as well as into a plasmid vector to give rise to pV1J-hEp-CAM. The two vaccine vectors were utilized with three different vaccination protocols: (a) single-modality DNA-EP-prime/DNA-EP-boost; (b) single-modality Adeno-prime/Adeno-boost; or (c) mixed-modality DNA-EP-prime/Adeno-boost. Cellular immunity was measured by intracellular IFN-γ staining 2 wk after the last injection. hEp-CAM-specific IFN-γ secretion from stimulated splenocytes was evaluated with a pool of 15-mer peptides overlapping by 11 amino acids and encompassing the entire protein 18.

As shown in Fig. 1A, hEp-CAM vaccination with DNA-EP followed by adenovirus resulted in a detectable CD8 and CD4 response as measured by analysis of IFN-γ intracellular cytokine staining (ICS). Vaccine splenocytes restimulated in vitro with hEp-CAM peptides showed similar levels of IFN-γ+ positive CD8 cells in the range of 0.2–0.4% of total CD3+CD8+ cells. A lower but significant CD4 response was also observed (0.05–0.2% of CD3+CD4+ cells). Although the immune responses elicited with the three protocols did not differ significantly, DNA-EP-prime/Adeno-boost showed a trend towards a stronger induction of both CD8 and CD4 IFN-γ+ antigen-specific T cells (Fig. 1B).

Figure 1.

Comparison of different hEp-CAM vaccination modalities. C57BL/6 mice were vaccinated at 2-wk intervals with: DNA-EP followed by DNA-EP; DNA-EP followed by Ad-hEp-CAM; Ad-hEp-CAM followed by Ad-hEp-CAM. (A) An example of IFN-γ ICS analysis in the DNA-EP/Ad-hEp-CAM-vaccinated mice. At 2 wk after the last injection, splenocytes were tested for IFN-γ production with a peptide pool covering the entire hEp-CAM protein. Whole-lymphocyte gating and also gating for CD8+ (R3) and CD4+ T cells (R4) are shown. (B) Percentages of IFN-γ-positive CD3 cells measured by ICS analysis. Histogram reports the average of five mice per group analyzed individually as shown in (A). CD8 or CD4 values from restimulated splenocytes with Ep-CAM peptides are in black, whereas control DMSO is the blank.

To determine the components of the peptide pool eliciting the responses, spleens from three mice of the DNA + Adeno cohort were analyzed in an IFN-γ ELISPOT assay against each of the 76 hEp-CAM peptides. Peptides that provided a positive result were further analyzed by intracellular IFN-γ assay to define the T cell specificity. While peptide 45, DP177KFITSILYENNVI191, showed a CD4 response, peptides 17 and 18 showed a CD8 response (Table 1). To define the location of the CD8+ peptide within the two overlapping peptides, a series of 8-aa peptides shifted by one residue were synthesized and tested against vaccine splenocytes. Peptide 17-2, C66LVMKAEM73, showed a result similar to peptide 17. These peptides are useful tools to quantify CD4 and CD8 responses induced against hEp-CAM in C57BL/6 vaccinated mice. In conclusion, both DNA-EP and Adeno, individually or in combination, are capable to induce an immune response against human Ep-CAM in this non-self setting. Although responses are relatively weak, they are reproducible and allow the identification of both CD8 and CD4 immunogenic peptides.

Table 1. Sequence of immunoreactive hEp-CAM peptides in C57BL/6 micea)
  1. a) C57BL/6 mice (n = 10) were immunized with two injections of 1 × 107 PFU Ad-hEp-CAM 2 wk apart. IFN-γ intracellular staining was carried out with individual peptides using a pool of splenocytes. These results are representative of two independent experiments. Data refer to percentages of IFN-γ-producing CD8+ (CD4+) CD3+ cells.

DMSO  0.010.01
Peptide SequenceCD8+ 
17-1 KCLVMKAE0.01 
17-2 CLVMKAEM0.4 
17-3 LVMKAEMN0.01 
17-4 VMKAEMNG0.01 
Pool A  0.3 
DMSO  0.01 

Cloning of rhesus Ep-CAM and assessment of its immunogenicity in mice

As an important step to evaluate Ep-CAM as potential tumor antigen for DNA-EP- and/or Adeno-based vaccine development, we decided to assess its capacity to break tolerance and to induce immune responses to self in a non-human primate rhesus model. This required, however, the initial isolation of the rhesus Ep-CAM (rhEp-CAM) cDNA homologue. This step was accomplished by RT-PCR of RNA extracted from biopsies of rhesus intestine (see Materials and methods). Analysis of different independent colonies revealed a conserved open reading frame (ORF), which was 96.4% homologous to the human sequence at the protein level (Fig. 2A). The rhEp-CAM ORF was found to contain 34 base changes compared to the published sequence of hEp-CAM. Additionally, a BLAST search in the data bank showed the best similarity with the hEp-CAM protein. Finally, an anti-hEp-CAM antibody could recognize epitopes present both in HeLa cells transfected with the positive control (hEp-CAM) and in HeLa cells transfected with rhEp-CAM (Fig. 2B). This evidence indicates that the rhEp-CAM ORF is able to code for an exposed plasma membrane protein and that the rhEp-CAM clone encodes a homologue of hEp-CAM.

Figure 2.

Cloning and characterization of rhEp-CAM gene. (A) Alignment of the predicted amino acid sequence of rhEp-CAM (DQ341470) with hEp-CAM (NM_002354). Amino acid differences are shaded. Boxed residues correspond to the CD8 epitope. (B) FACS analysis of rhEp-CAM expression. To show that rhEp-CAM cDNA codes for the rhesus homologue of hEp-CAM, HeLa cells were transfected with pV1J-rhEp-CAM or with the positive control pV1J-hEp-CAM. Expression was revealed with an antibody raised against hEp-CAM. (C) Tumor challenge experiment in C57BL/6 mice vaccinated with two injections of Ad-hEp-CAM, Ad-rhEp-CAM or control (mock-injected). Vaccinated mice were challenged with MC38-hEp-CAM cells or MC38 cells 30 days after the last immunization. The histogram reports tumor-free mice at 46 days after tumor challenge.

To explore DNA-EP and Adeno vaccines expressing rhEp-CAM, we generated Ad-rhEp-CAM and pV1J-rhEp-CAM and tested their immunogenicity in C57BL/6 mice. Interestingly, the IFN-γ CD3/CD8+ responsive 17-2 peptide, which was identified in Ad-hEp-CAM-vaccinated C57BL/6 mice (see previous paragraph), is conserved in both the human and rhesus sequences (Fig. 2A). Indeed, similar levels of CD8 responses were observed in mice vaccinated with Ad-hEp-CAM or Ad-rhEp-CAM vectors and analyzed by IFN-γ ICS using the 17-2 peptide (data not shown). Similar responses were also observed with DNA-EP of pV1J-rhEp-CAM.

In line with these observations, mice vaccinated with either Ad-hEp-CAM or Ad-rhEp-CAM were protected from a tumor challenge with MC38-hEp-CAM cells (Fig. 2C), which shows that the immune response induced against hEp-CAM with both vaccines is sufficient to produce an effective biological anti-tumor effect. The specificity of hEp-CAM vaccination was further supported by the absence of tumor protection in Ad-hEp-CAM-vaccinated mice challenged with MC38 cells, which do not express hEp-CAM.

Low frequency of immune responders to DNA-EP/Adeno Ep-CAM vaccine in rhesus monkeys

We initially vaccinated four rhesus monkeys with pV1J-rhEp-CAM (six monthly DNA-EP injections) followed by Ad-rhEp-CAM (two biweekly injections) separated by 8 wk. Antibody responses were assessed with a recombinant rhEp-CAM protein expressed and purified in baculovirus. The cell-mediated immune response was analyzed by IFN-γ ELISPOT assay using PBMC restimulated in vitro with three peptide pools of 15-mers overlapping by 11 amino acids that collectively encompass the entire rhEp-CAM protein. Pool A covers amino acids 1–111, pool B covers amino acids 101–211, and pool C covers amino acids 201–315. Results were completely negative (data not shown).

Following to that, a second group of four monkeys was subjected to a xenogeneic vaccination protocol with six monthly DNA-EP injections of the pV1J-hEp-CAM plasmid followed, after an interval of 8 wk, by two injections of the Ad-rhEp-CAM expressing the self antigen. As shown in Fig. 3A, an antibody response was observed at various time points in one monkey, RI497, and sporadically also in others. The serum from this monkey reacted against recombinant rhEp-CAM also in Western blot analysis (Fig. 3B). The same monkey also developed a cell-mediated immune response against Ep-CAM. As shown in Fig. 4B, the cell-mediated immune response was primarily directed towards the N-terminal region of the human Ep-CAM protein, as indicated by the spot-forming colony (SFC) value measured with peptide pools A and B. The same monkey that showed a response against human peptides also showed a response against rhesus peptides. The anti-rhEp-CAM response (Fig. 4A) is mainly directed against the N terminus of the rhesus protein, as indicated by SFC values measured with pool A. Since these responses returned to background at wk 28, monkeys were boosted with Ad-rhEp-CAM at wk 29 and 32. At wk 36, the IFN-γ response against Ep-CAM rose up again in RI497 with both human peptides (pools A and B) and rhesus peptides (pool A) and was further confirmed at wk 40 and 44. Values became undetectable at wk 51. These experiments show that, in spite of a very intense vaccination scheme, it is very difficult to break the immunological tolerance against rhEp-CAM in a non-human primate model.

Figure 3.

Induction of rhEp-CAM-specific antibody response in rhesus macaques. Four rhesus macaques were immunized intramuscularly with pV1J-hEp-CAM DNA in the presence of electroporation at weeks 0, 4, 8, 12, 16 and 20 followed by Ad-rhEp-CAM injection at weeks 28 and 32. (A) Sera were collected from the monkeys and assayed for the induction of rhEp-CAM-specific antibodies at a dilution of 1 : 50. (B) Induction of rhEp-CAM-specific Western blot reactivity in immunized macaques. The serum collected from monkey RI497 at 8 wk post immunization was analyzed by Western blot analysis against recombinant rhEP-CAM protein (lane 2). The rhEp-CAM band at the molecular weight of 29 kDa is indicated by an arrow. The serum from pre-immune RI497 was negative (lane 1).

Figure 4.

Frequency of Ep-CAM-specific IFN-γ-secreting cells (as measured by ELISPOT assay) in rhesus macaques vaccinated as depicted in Fig. 3. Numbers of SFC per 106 PBMC against pool A covering the N-terminal region (25 peptides, amino acids 1–111, ♦), pool B covering the central protein region (25 peptides, amino acids 101-211, ▪), and pool C covering the C-terminal region (26 peptides, amino acids 201–315, ▴) are reported at different time points; mock responses to DMSO (no peptides) ranged from 0 to 96 per 106 PBMC and were subtracted from these reported responses; asterisks indicate responses threefold over the corresponding DMSO value. (A) ELISPOT analysis of rhesus RI497 against rhesus peptide pools. (B) ELISPOT analysis of rhesus RI497 against human peptide pools.

mEp-CAM vaccination breaks inmmunological tolerance only in a subgroup of CD1 mice

The low frequency of immune responders to Ep-CAM in rhesus monkeys prompted us to go back and re-examine the vaccine capability to break tolerance to self in smaller animal models in mice, where a variety of strains with well-defined MHC haplotypes are available and where it is possible to work with larger numbers of individuals 14.

Ad-mEp-CAM was, therefore, generated and used with the same vaccination protocol reported above for Ad-hEp-CAM in a series of mouse inbred strains (Table 2). The cell-mediated immune response was evaluated using IFN-γ ICS with a peptide pool specific for mEp-CAM. Splenocytes restimulated in vitro with mEp-CAM peptides did not show any detectable IFN-γ response in vaccinated inbred strains. Since we could break the immunological tolerance in rhesus using the hEp-CAM sequence, we asked whether the same xenogeneic protocol could be effective in inbred mouse strains. C57BL/6 mice were vaccinated with Ad-hEp-CAM as described above, and splenocytes were analyzed with human and mouse Ep-CAM peptides by IFN-γ ICS analysis. A positive signal was observed with human peptides, but not with mouse peptides.

Table 2. Immune response induced by Ad-mEp-CAM vaccination in different mouse strainsa)
 ICSTumor protection
  1. a) Groups of mice were immunized with two injections of 1 × 107 PFU Ad-mEp-CAM 2 wk apart. IFN-γ intracellular staining was carried out with splenocytes from each mouse restimulated in vitro with the peptide pool covering the entire mEp-CAM protein. These results are representative of two independent experiments. Data refer to percentages of IFN-γ-producing CD8+CD3+ cells.

StrainRespondersAverange CD8+/IFN-γ+

We then examined this vaccination protocol in outbred CD1 mice. Ad-mEp-CAM vaccination resulted in a subgroup of responders (23%) as shown by the presence of Ep-CAM-specific IFN-γ+ CD8+ cells, which ranged between 0.3 and 1.2% of CD3+CD8+ cells (Fig. 5A). This response was not observed in a control group of 30 mice that were not vaccinated with Ad-mEp-CAM (data not shown). To further characterize responsive CD1 mice, the CD8 response was analyzed over time by ICS analysis using PBMC. IFN-γ-secreting CD3+CD8+ positive cells were detected also in the periphery of vaccinated mice and confirmed at different time points. The signal became undetectable at 2 months post injection (Fig. 5B). Finally, the CD8 epitopes were mapped at the level of single peptides by ELISPOT assay and shown to be contained within peptide 1, M1AGPQALAFGGLLLA15. To exclude that the immune response was induced by an amino acid difference between the vaccine sequence and the resident gene, mEp-CAM cDNA was isolated and sequenced in all responsive mice. No mutations were observed in the host mEp-CAM cDNA sequence, indicating that the immune response was elicited against the resident antigen; thus, it is likely that other genetic differences were associated with a “responsive phenotype”.

Figure 5.

Analysis of CD8 immune response elicited against mEp-CAM by Ad-mEp-CAM vaccination in CD1 mice. CD1 mice received two intramuscular injections of Ad-mEp-CAM, at 2-wk intervals. (A) IFN-γ ICS analysis performed as described in Fig. 1. (B) Persistence of the anti-mEp-CAM response over time evaluated in three responsive CD1 mice. PBMC were restimulated in vitro with mEp-CAM peptides and the CD3/CD8 response was measured by IFN-γ ICS analysis.

Identification of biomarkers associated with immune responsiveness to Ep-CAM vaccine in CD1 mice

To verify whether the induction of a CD8 response against mEp-CAM was associated with a specific genetic background, we determined the MHC class I haplotype in vaccinated CD1 mice (Fig. 6A). A statistically significant association was observed between responding animals (7 out of 30) and the q/q MHC class I haplotype (p = 0.0003). Non-responding animals were identified as q/j, q/b, b/k or j/j (Fig. 6B). Thus, the immune response is likely to be associated with a genetic factor that segregates with the q/q haploytype. Responsive MHC class I q/q mice were also characterized for MHC class II. All of them showed expression of the q/q allele. However, we also observed that in other strains with the same q/q haploytype, such as FVB and DBA/1j (Table 2), there were no responders, thus suggesting that additional factors influence the immune reactivity against Ep-CAM.

Figure 6.

Correlation between anti-mEp-CAM response and MHC class I haplotype in CD1 mice. Splenocytes from Ad-mEp-CAM-vaccinated mice were analyzed at 2 wk post injection for the induction of the anti-mEp-CAM CD8 response by IFN-γ ICS analysis and were characterized for MHC class I haplotype expression as described in Materials and methods. (A) A representative experiment is depicted in the histogram, where letters on the black bars indicate the MHC class I haplotype. (B) Summary Table of the anti-mEp-CAM CD8 response and the MHC class I haplotype; a statistically significant association was observed between responders and the q/q haplotype by the Fisher exact test (p <0.005).

Many different genes are present in the vicinity of the MHC class I locus, including the T cell receptor (TCR) which is a potential modulator of the immune response. It has been shown that autoimmune diseases are linked to expression not only of a specific MHC haplotype but also of specific TCR sequences such as TCRVβ8 19. To verify this hypothesis, the expression of TCRVβ8 and TCRVβ4 was compared in mouse strains having the same MHC class I (q/q). FVB mice do not express TCRVβ8 whereas a statistically significant increase of both TCRVβ4 and TCRVβ8 expression levels was observed in responsive CD1 q/q mice with respect to non-responder DBA/1j (q/q) mice (p <0.05) (Table 3). These data suggest an association between TCRVβ expression and the anti-Ep-CAM response.

Table 3. Expression of TCRVβ4 and TCRVβ8 in MHC q/q micea)
  1. a) TCRVβ4 and TCRVβ8 expression in different mouse strains. Splenocytes obtained from different mouse strains (FVB, DBA/1j, CD1 q/q) were examined for TCR expression by FACS analysis. Values are the average of expression in at least six animals per group ± standard deviation. A statistically significant increase (p <0.05) in both TCRVβ4 and TCRVβ8 expression was observed in CD1 (q/q) mice with respect to DBA/1j (t-test).

CD1 q/q+2.2 ± 0.912.0 ± 3.2
FVB q/q7.0 ± 1.1
DBA/1j q/q1.3 ± 0.38.5 ± 1.6

We then asked whether the responsive phenotype could be transmitted to the progeny by crossing q/q mice. CD1 q/q mice were backcrossed together and the F1 progeny was again vaccinated with Ad-mEp-CAM. In this case, all 15/15 (100%) mice developed an anti-mEp-CAM cell-mediated immune response (not shown), as expected from a Mendelian transmission of this character. The isolated MHC class I q/q colony will be further characterized to identify additional genetic markers responsible for this phenotype.


Ep-CAM has been the objective of many clinical trials and it is still under evaluation for both passive and active immunotherapy 4. Although an immune response has been reported in Ep-CAM-vaccinated CRC patients, the host factors that correlate with the response are not well characterized. Here, we observed that an intensive immunization protocol is effective only in a small subset of outbred mice and rhesus macaques.

Previous reports have shown that DNA injection in the presence of electroporation or Ad vectors are potent immunization vehicles 15, 16. Here, we used these vectors to set up vaccination conditions for hEp-CAM. Indeed, C57BL/6 mice injected with Ad-hEp-CAM alone or in combination with DNA-EP showed induction of Ep-CAM-specific CD8+ and CD4+ IFN-γ-secreting T cells. Previous reports have shown that viral vectors expressing hEp-CAM could elicit antibody as well as CTL responses in BALB/c mice 12, 13. Here we have extended these observations showing that also DNA-EP in addition to Ad-hEp-CAM vaccination elicits a cell-mediated immune response. The response was characterized using peptides covering the entire Ep-CAM amino acid sequence. This approach allowed the responsive peptides to be identified experimentally 20. Using these peptides, a CD4 epitope (P177KFITSILYENNVI191) and a CD8 epitope (C66LVMKAEM73) were mapped. This information may help in evaluating the efficacy of hEp-CAM vaccines in a hEp-CAM-transgenic C57BL/6 mouse model 21.

The rhEp-CAM gene was cloned and evaluated as genetic vaccine both in a mouse tumor model and in rhesus macaques. The amino acid sequence of rhEp-CAM showed a high identity with hEp-CAM, as indicated by sequence alignment and by the observation that an anti-human Ep-CAM antibody could recognize both proteins (Fig. 2A, B). Due to the lack of tumor models in rhesus, the possibility of using rhesus and human sequences as cross-reactive vaccines was explored in a mouse tumor model. IFN-γ ELISPOT analysis using single rhesus peptides showed a positive response with the same CD8 peptide identified with the human sequence (peptide 17-2). Indeed, this peptide is conserved in human and rhesus sequences (Fig. 2A). This observation may explain, at least in part, the comparable tumor protection observed in mice vaccinated with either the human or the rhesus sequence and challenged with a tumor cell line expressing hEp-CAM (Fig. 2C). These results appear to be superior to previous reports where only a partial tumor protection was observed upon Ad-hEp-CAM vaccination in a BALB/c tumor model 12. Differently from the previous report, the higher level of tumor protection observed in the C57BL/6 model was specific for hEp-CAM, as indicated by tumor growth in Ad-hEp-CAM-vaccinated mice challenged with MC38 cells which do not express the antigen.

Vaccination in rhesus with the self rhEp-CAM antigen alone did not lead to any detectable IFN-γ ELISPOT response, whereas priming with human sequence resulted in one responsive monkey, RI497 (Fig. 3, 4). Xenogeneic DNA vaccination has been previously reported to break immunological tolerance in mice with different antigens including HER2, metalloproteainase-2, tyrosinase-related proteins (TRP)-1 and -2, and human gp100 2225. Different mechanisms can explain the efficacy of this approach, such as the cross-reactivity of conserved peptides which may be recognized differently when presented in a different sequence context. A CTL response against an H-2Kb-restricted peptide of mTRP-2 was observed only in mice vaccinated with human sequence and not in mice vaccinated with the self antigen 26. It is remarkable to note that this peptide is conserved in the TRP-2 sequence of both human and mouse. Similarly, the IFN-γ ELISPOT response observed in RI497 with both human and rhesus peptides may be explained by a conserved epitope present in peptide pool A. Deconvolution of peptide pool A was not conducted due to the limited amount of PBMC available, which prevents the possibility of using single peptides in the ELISPOT assay. Future experiments may be designed to answer this question. These experiments indicate that a low frequency of responders is present in an outbred rhesus macaque population.

To answer the question of whether genetic variability may impact on the anti-Ep-CAM response, the same vaccination approach was utilized with mEp-CAM in different mouse strains. An immune response was not observed in inbred mouse strains, whereas a CD8 response was observed in a subgroup of outbred CD1 mice (Fig. 5). To our knowledge, this is the first evidence that Ad-mEp-CAM vaccination can induce a CD8 response in a tolerant model. Interestingly, a statistically significant association between the mEp-CAM-specific response and the q/q MHC class I haplotype was observed (Fig. 6). Previous reports showed that this haplotype is associated with autoimmune diseases, and only mice of the q/q haplotype are susceptible to arthritis induction upon immunization with type II collagen in CFA 27. In addition to the MHC class I haplotype, other factors seem to play a role in inducing the anti-mEp-CAM response. Apparently, the presence of the MHC class I q/q haplotype is not sufficient to induce an Ep-CAM-specific response since inbred mouse strains with the same MHC class I haplotype such as DBA/1j and FVB were negative in the same vaccination protocol (Table 2). A potential explanation is the statistically significant increase of expression of TCRVβ4 and of TCRVβ8 observed in responsive CD1 mice versus unresponsive DBA/1j mice (Table 3). Expression of these TCR has been correlated with the occurrence of autoimmune diseases 19. In addition, recent clinical trials with Ep-CAM vaccines showed that the CD8+IFN-γ+ response, measured by ELISPOT assay, was associated with a specific TCR 28. It is reasonable to speculate that cell-mediated immune responses can be induced in the presence of a combination of a given MHC class I allele and a specific TCR. It would be interesting to verify whether the Ep-CAM-specific CD8 response induced in CD1 mice is skewed toward a particular TCR such as TCRVβ4 and TRCVβ8.

A recent clinical report showed that ALVAC-Ep-CAM vaccination in CRC patients led to the induction of an Ep-CAM-specific IFN-γ response in two out of six patients untreated with GM-CSF: A cell-mediated immune response was observed 4 months after the last vaccination in two metastatic patients 11. Spontaneous weak cell-mediated responses may seldom develop against Ep-CAM in metastatic CRC patients 9. It cannot be excluded, therefore, that the anti-Ep-CAM immune response observed after ALVAC-Ep-CAM vaccination is due to a boost of a pre-existing natural immune response. The efficacy of Ep-CAM vaccines may be further evaluated in breast cancer patients where the antigen is clearly overexpressed 5 but where natural immune responses have so far not been reported 8. Our results showed that in the absence of adjuvant therapy or ongoing cancer, breakage of the immunological tolerance against mEp-CAM was achieved in a subgroup of mice and in one out of eight monkeys. The identification of a “responsive” mouse model (CD1 q/q) where 100% of mice respond to mEp-CAM vaccination may prove to be a useful tool in understanding the key factors that enable an immune response against a self antigen. Preliminary vaccination experiments in this animal model with an additional tumor-associated antigen such as mouse Survivin indicate that the responsive phenotype is not restricted to mEp-CAM. In conclusion, the low frequency of immune responders implies that an Ep-CAM-based vaccine, in order to become clinically successful, will necessarily require a combination with adjuvants/immunomodulatory agents as well as the development of adequate strategies to identify predictive markers of the response to the vaccine.

Materials and methods


Six-week-old C57BL/6 (H-2b), Balb/c (H-2d), DBA/1j (H-2q), FVB (H-2q) and CD1 mice were obtained from Charles River (Calco, Italy). All animals were kept under specific pathogen-free conditions.

Antigens and vaccine vectors

hEp-CAM and mEp-CAM cDNA clones were generated by RT-PCR on RNA extracted from CaCo2 cells and mice intestine (C57BL/6), respectively. A cDNA clone specific to rhEp-CAM was generated by RT-PCR of total RNA isolated from rhesus intestine. To obtain intestinal RNA, frozen tissues were obtained from The Biomedical Primate Research Center (BPRC, Rijswijk, The Netherlands). Tissues were mechanically pulverized and combined with the Ultraspec RNA reagent (Biotecx Laboratories, Houston, TX) according to the manufacturer's instructions. To perform the PCR step, the following oligonucleotide primers were designed based on the alignment of the 5′ and 3′ conserved non-translated regions of the human (Accession No. NM_002354), mouse (Accession No. NM_008532) and rat (Accession No. AJ001044) Ep-CAM genes: EpcamA, 5′-GGAAGATCTCGCGCGCAGCATGGCG-3′ (SEQ ID NO:11); and EpcamB, 5-ACGCGTCGACGAGTACAGGTTTCACTATTACAAAT-3′. The primer sequences were modified to introduce the cloning sites Bgl II and Sal I. The amplification product was analyzed by agarose gel electrophoresis and the expected band isolated with JetSorb (Genomed, Bad Oeynhausen, Germany). Purified DNA bands obtained by RT-PCR were cloned into plasmid vectors for further analysis. In the first cloning experiment, five colonies were picked and the inserts were sequenced and found to be >99% identical to each other. To confirm the DNA sequence, total RNA was isolated from the intestine of a different macaque. Plasmid DNA isolated from seven independent colonies was sequenced to verify the accuracy of the sequence data. Results indicate that two out of the seven colonies contained the same rhEp-CAM sequence (colonies 2 and 10) with an overall nucleotide identity of 99.1% (seven colonies). Colonies 2 and 10 contained the same sequence as three of the colonies from the previous experiment, with the exception of two nucleotides at positions 184 and 942. This novel rhEp-CAM sequence, which contains a “C” at position 184 instead of a “T,” and a “G” at position 942 instead of an “A”, nevertheless shows 100% amino acid identity with the proteins encoded by the first and second rhEP-CAM sequences, suggesting that this mutation may be a polymorphism present in two different macaques.

To generate a plasmid vaccine, hEp-CAM, mEp-CAM and rhEp-CAM were subcloned into the pV1J expression vector 29. Ad vectors were constructed as described 30. A tumor cell line expressing hEp-CAM was obtained using pV1J-Ep-CAM-Zeo, which contains the zeocyn-selectable marker. The plasmid was transfected into MC38 cells, whereupon zeocyn-resistant clones expressing hEp-CAM were evaluated by FACS analysis. The clone used in subsequent tumor challenge experiments expressed hEp-CAM in 94% of cells.

Sequencing mEp-CAM gene and MHC genotyping in CD1 mice

mEp-CAM cDNA was generated by RT-PCR of total RNA isolated from CD1 mouse intestine. The DNA sequence was confirmed for the 5′ cDNA region using the following primers: CGCGCCACACCTCAGTCCGT and ATTCAGAGAGCAACTTGTTG. MHC class I and class II genotyping was performed according to a previous publication 31.

Vaccination protocols

DNA vaccination in mice was performed via both electroporation according to Rizzuto et al.32 and adenovirus vaccination according to Mennuni et al.17. Monkey immunization studies were performed at The Biomedical Primate Research Center (BPRC, Rijswijk, The Netherlands) using a group of four rhesus macaques (Macaca mulatta). For priming, animals were vaccinated intramuscularly at wk 0, 4, 8, 12, and 20 by injection of DNA, followed by electroporation. The DNA injection consisted of a 1-mL solution (split over two sites with 0.5 mL/site) containing 5 mg pV1J-rhEp-CAM DNA. For electroporation, two trains of 100 square bipolar pulses (1 s each) were delivered every other second for a total treatment time of 3 s. The pulse length was 2 ms/phase with a pulse frequency and amplitude of 100 Hz and 100 mA, respectively (constant current mode). Animals were injected with either the pV1J-rhEp-CAM or a dose of 1011 viral particles of the adenovirus construct Ad-rhEp-CAM. Blood was collected every 4 wk, and PBMC were isolated by Ficoll density gradient centrifugation. Fresh or frozen PBMC samples were analyzed for antigen-specific IFN-γ secretion by an IFN-γ ELISPOT assay.

FACS analysis

FACS analysis was performed using the anti-hEp-CAM Ab-3 monoclonal antibody (clone 323/A, Lab Vision Corp., Fremont, CA).

Mouse PBMC (1 × 106 or 2 × 106) were used for TCR analysis. Cells were stained for 30 min with (i) allophycocyanin-conjugated anti-mouse CD3ϵ, (ii) isothiocyanate-conjugated anti-mouse Vβ4 TCR, and (iii) r-phycoerythrin-conjugated anti-mouse Vβ8 TCR (BD PharMingen). Cells were analyzed in a FACSCalibur (Becton Dickinson, San Jose, CA) using CellQuest software (Becton Dickinson).

Cell-mediated immune response

Antigen-specific IFN-γ-secreting cells were detected using a standard ELISPOT assay. Spleen cells were prepared from immunized mice and resuspended in R10 medium (RPMI 1640 supplemented with 10% fetal calf serum, 2 mM L-glutamine, 50 U/mL penicillin, 50 mg/mL streptomycin, 10 mM HEPES, 50 mM 2-mercaptoethanol). Multiscreen 96-well filtration plates (Millipore Corp., Bedford, MA) were coated with purified rat anti-mouse IFN-γ antibody (PharMingen, San Diego, CA). After overnight incubation, the plates were washed with 1 × PBS/0.005% Tween-20 and blocked with 250 µL R10 medium per well. Splenocytes at a density of 2.5 × 105 or 5 × 105 per well were incubated for 24 h in the presence of 2 µg/mL peptide pools or single peptide. After extensive washing (1 × PBS/0.005% Tween-20), biotinylated rat anti-mouse IFN-γ antibody (PharMingen) was added and incubated overnight at 4°C. For development, streptavidin-alkaline phosphatase (PharMingen) and 1-Step NBT-BCIP Development Solution (Pierce, Rockford, IL) were added. A pool of 15-mer peptides encompassing the entire Ep-CAM protein was used to reveal the Ep-CAM-specific IFN-γ-secreting T cells. Spots were counted using an automated detection system.

Intracellular IFN-γ was measured according to the BD PharMingen standard protocol. Briefly, 2 × 106 splenocytes from immunized mice were incubated overnight in R10 medium with or without 10 µg/mL peptide pool and brefeldin A as a protein transport inhibitor (Cytofix/Cytomer Plus kit; BD PharMingen). DMSO was tested with the splenocytes as a background control. Cells were stained with allophycocyanin-conjugated anti-mouse CD3ϵ, phycoerythrin-conjugated anti-mouse CD4, and peridinin chlorophyll a protein-conjugated anti-mouse CD8α antibodies (PharMingen). The splenocytes were then washed, fixed, permeabilized, and stained for intracellular IFN-γ by fluorescein isothiocyanate-conjugated anti-IFN-γ monoclonal antibody. All samples were acquired within 24 h with a FACSCalibur flow cytometer and CellQuest software (BD Immunocytometry Systems). T-lymphocyte IFN-γ was calculated as 100 × [(IFN-γ+, CD3+ and CD4+ or CD8+)/(CD3+ and CD4+ or CD8+)].

Antibody response

The amino acid region 23–265 of rhEp-CAM was expressed in baculovirus with a 6-His tag at the C terminus. Recombinant protein was expressed and purified using a standard protocol at Protein'expert (Grenoble, France). Induction of anti-rhEp-CAM antibodies in rhesus was monitored by enzyme-linked immunosorbent assay (ELISA) as described 33. Nunc ELISA plates (96-well) were coated overnight at 4°C with recombinant rhEp-CAM protein (250 ng), washed with washing buffer (PBS/0.05% Tween-20), and incubated with blocking buffer (5% bovine serum albumin, 1 × PBS/0.05% Tween-20) to remove nonspecific binding. Bound antibodies were detected with anti-monkey IgG Fc-specific alkaline phosphatase-conjugated antibody (catalog number A-1929; Sigma) diluted 1 : 2000 in blocking buffer.

Tumor challenge

MC38 and MC38-hEp-CAM cells were injected subcutaneously into C57BL/6 mice at a dose of 3 × 105 cells per mouse. Tumor growth was monitored every week.

Statistical analysis

Data obtained from ELISA experiments were analyzed using a t-test to evaluate the significance of antibody titers. For the mouse challenge experiments, protection significance was evaluated using the log rank test.


  1. 1


  2. 2


  3. 3


  4. 4


  5. 5


  6. 6