The Signal Peptide of the Tumor-shared Antigen Midkine Hosts CD4+ T Cell Epitopes*

Background: The CD4 T cell response to the tumor antigen Midkine was unknown. Results: Most of the T cell response to Midkine relies on T cell epitopes contained in its signal peptide. Conclusion: The signal peptide of Midkine is accessible to HLA class II pathway for CD4 T cell presentation. Significance: It is a new function for signal peptides to contribute to tumor-specific CD4 T cell response. Because of the key role of CD4 T cell response in immunity to tumors, we investigated the CD4+ T cell response to the recently identified tumor antigen Midkine (MDK). By weekly stimulations of T lymphocytes harvested from seven HLA-DR-typed healthy donors, we derived CD4+ T cell lines specific for eight MDK peptides. Most of the T cell lines reacted with the peptides 9–23 and 14–28, located in and overlapping the MDK signal peptide, respectively. Accordingly, the MDK signal peptide appeared to be rich in good binders to common HLA-DR molecules. The peptide 9–23-specific T cell lines were specifically stimulated by autologous dendritic cells loaded with lysates of MDK-transfected cells or with lysates of tumor cells naturally expressing the MDK protein. One T cell line was stimulated by HLA-compatible MDK-transfected tumor cells. By contrast, the peptide 14–28-specific T cell lines were not stimulated in any of these conditions. Our data demonstrate that CD4+ T cell epitopes present in the signal peptide can be accessible to recognition by CD4+ T cells and may therefore contribute to tumor immunity, whereas a peptide overlapping the junction between the signal peptide and the mature protein is not.

Midkine (MDK) 2 is a heparin-binding growth factor with multiple functions. It plays a major role in neuronal development during embryogenesis (1) and has been implicated in ischemic renal injury (2), inflammatory responses (3), and blood pressure regulation (4). Injection of MDK diminishes myocardial infarct size in animal models (5,6). MDK also contributes to tumor progression: blockade of MDK leads to tumor regression (7). This protein is overexpressed in various human malignancies, including bladder, prostate, breast, lung, liver, and colon tumors, although its expression is typically low or undetectable in normal adult tissues (8 -10). In tumors, MDK exhibits angiogenic (11), mitogenic (12), and antiapoptotic activities (13). Its expression in tumors (8,14) and its concen-tration in plasma (15) are associated with a poor disease outcome. Its various functions and expression in many tumors led us to investigate whether MDK can induce cellular cytotoxic responses in humans (16). By in vitro stimulation of CD8 ϩ T cells collected from HLA-A2 healthy donors and immunization of HLA-A2 transgenic mice, we identified two CD8 ϩ T cell epitopes and demonstrated that MDK-specific cytotoxic T lymphocytes can lyse tumor cells (16). One of these CD8 ϩ T cell epitopes resides in the signal peptide, as described previously for other secreted tumor antigens (17)(18)(19). These findings suggest that MDK may be novel candidate for the development of a cancer vaccine.
CD4 ϩ T lymphocytes enhance and sustain the tumor-specific CD8 ϩ T cell response by providing cytokines and co-stimulation signals. CD4 ϩ T lymphocytes also contribute to tumor regression by recruiting and activating phagocytes, by producing inflammatory cytokines, or by exhibiting direct cytolytic functions (20). Many vaccine strategies exploit CD4 ϩ T cell functions to eliminate tumors and combine CD4 ϩ and CD8 ϩ T cell epitopes delivered in various forms such as DNA, recombinant viruses, proteins, or long polypeptide fragments. We investigated whether MDK can prime a CD4 ϩ T cell response in multiple HLA-typed donors. As an overexpressed tumor antigen, a residual expression of MDK persists in healthy donors and may promote a tolerance status that would diminish the induction of a MDK-specific CD4 ϩ T cell response. Also, MDK is a relatively small protein, and MDK-specific CD4 ϩ T cell responses may be limited to particular haplotypes and hence may not be effective in all individuals. As we previously identified CD8 ϩ T cell epitopes in the signal peptide, we considered the entire sequence of MDK including the mature form of the protein secreted by tumors and the signal peptide, which remains in the cell (21). Few CD4 ϩ T cell epitopes have been found in signal peptides (22,23), but here, we demonstrate that the MDK signal peptide contains both subdominant and cryptic CD4 ϩ T cell epitopes.

MATERIALS AND METHODS
Peptides and Proteins-The human MDK sequence was retrieved from Uniprot (P21741) and comprises the signal peptide (amino acid positions 1-22) and the mature protein (23-143). Overlapping 15-amino-acid-long MDK peptides were optimized for the requirement of aliphatic or aromatic residues in the N-terminal part of the peptide for binding to HLA class II molecules and therefore covered the sequence 1-133. MDK and biotinylated peptides were purchased from Activotec (Cambridge, UK) or synthesized using standard N-(9-fluorenyl)methoxycarbonyl (Fmoc) chemistry on an Advanced ChemTech Apex synthesizer (Advanced ChemTech Europe, Brussels, Belgium). Biotin was introduced using biotinyl-6aminocaproic acid (Fluka Chimie, St. Quentin Fallavier, France) on the N terminus. As appropriate, peptides were purified by RP-HPLC on a C 18 Vydac column (Interchim, Montluçon, France) to ϳ90% purity as determined by analytical HPLC.
HLA Class II-specific Peptide Binding Assays-HLA-DR and HLA-DP4 molecules were immunopurified from homologous EBV cell lines by affinity chromatography using the monomorphic MAbs, L243 and B7/21, respectively (24 -26). Binding to HLA-DR and HLA-DP4 molecules was assessed by a competitive ELISA as reported previously (24 -26). The competition for the immunopurified HLA class II molecules is performed between a reporter peptide (biotinylated) and the peptide to be tested. The binding specificity for each HLA class II molecule was ensured by the choice of the biotinylated peptides as described previously (25,26). Unlabeled forms of the reporter peptides were used as reference peptides to assess the validity of each experiment. A binding experiment is validated if the IC 50 of the reference does not differ from more than a factor of 3 with its common value. Their sequences and IC 50 values were as follows: HA 306 -318 (PKYVKQNTLKLAT) for DRB1*01:01 (2 nM), DRB1*04:01 (6 nM), DRB1*11:01 (9 nM), and DRB5*01:01 Cell Lines and Cell Transfection-Human hepatocellular carcinoma cells (HepG2), cervical cancer cells (HeLa), and melanoma cells (A375) were purchased from the American Type Culture Collection (Manassas, VA). The pcDNA3.1-MDK (pMDK) was developed as described previously (16). Twenty g of this plasmid was introduced into HeLa cells (3 ϫ 10 6 cells) by electroporation (0.9 kV, 250 microfarads, 0.4-cm gap) with a Gene Pulser II (Bio-Rad). The treated cells were suspended in 40 ml of prewarmed RPMI 1640 medium containing 10% fetal bovine serum and incubated under 5% CO 2 /air at 37°C. They were harvested 48 h later, and MDK expression was assessed by intracellular staining with an anti-MDK-phycoerythrin antibody (BD Biosciences) as described previously (16).
Blood Samples and HLA-DR Genotyping-Blood cells were collected from anonymous healthy donors at the Etablissement Français du Sang (EFS, Rungis, France) as buffy coat preparations after informed consent and following EFS guidelines. Peripheral blood mononuclear cells (PBMCs) were isolated by density centrifugation on Ficoll-Hyperpaque gradients (Sigma-Aldrich Generation of Antigen-specific T Cell Lines from Healthy Donors-Monocyte-derived dendritic cells (DCs) were generated from plastic-adherent cells among PBMCs after 5-7 days of culture in AIM V medium (Invitrogen) supplemented with 1000 units/ml of rh-IL-4 (R&D Systems) and rh-GM-CSF (Miltenyi Biotech, Paris, France). Immature DCs were collected on day 5 or 6. Mature DCs were generated after the addition of 1 g/ml LPS (Sigma). CD4 ϩ T lymphocytes were isolated from the nonadherent phase by positive selection using an anti-CD4 monoclonal antibody coupled to magnetic microbeads (Miltenyi Biotech) and magnetic cell sorting, as recommended by the manufacturer (Miltenyi Biotech). Mature DCs (5 ϫ 10 5 ) were incubated at 37°C, under 5% CO 2 , for 4 h in 1 ml of complete IMDM containing a pool of MDK peptides, each peptide being at a concentration of 10 g/ml. Complete IMDM was IMDM (Invitrogen) supplemented with 0.24 mM glutamine, 0.55 mM asparagine, 1.5 mM arginine (all amino acids from Sigma), 50 units/ml penicillin and 50 g/ml streptomycin (Invitrogen), and 10% human serum. Pulsed cells were washed, and 10 4 was added to each round-bottom microwell already containing 10 5 autologous CD4 ϩ lymphocytes in 200 l of complete IMDM supplemented with 1000 units/ml of IL-6 (R&D Systems) and 10 ng/ml IL-12 (R&D Systems). The CD4 ϩ T lymphocytes were restimulated on days 7, 14, and 21 with autologous DCs freshly loaded with the MDK peptide pools and were grown in complete IMDM supplemented with 10 units/ml IL-2 (R&D Systems) and 5 ng/ml IL-7 (R&D Systems). The specificity of the T cell lines was investigated by IFN-␥ enzyme-linked immunospot (EliSpot) assays on days 28 and 29 as described previously (27). Two to three consecutive EliSpot assays were performed in the same week. Antigen-presenting cells were autologous PBMCs or autologous immature DCs or murine L cells transfected with one HLA class II molecule (28). T cell lines were considered peptide-specific when their production of spots in the presence of antigens was at least two times higher than in their absence with a minimum of 20 spots. Positivity in a statistical unpaired Student t test was also evaluated (significance determined at p Ͻ 0.05). Peptide specificity of each T cell line was evaluated in at least two independent experiments.

CD4 ϩ T Cell Response Specific for MDK Peptides in Healthy
Donors-We first investigated the capacity of 18 peptides overlapping the MDK sequence to prime specifically CD4 ϩ T cells from seven HLA-typed healthy donors. These donors were selected to represent all the most frequent HLA-DR molecules in the Caucasian population ( Fig. 1 legend). CD4 ϩ T cells were seeded in 96-well plates and stimulated weekly by mature DCs loaded separately with one of the two peptide pools. Peptide specificity of the T cell lines was evaluated by IFN-␥ EliSpot using autologous PBMC as antigen-presenting cells (Fig. 1A). Four peptides were common to multiple T cell lines isolated from the seven donors: peptide 9 -23 induced four T cell lines; peptide 14 -28, ten; peptide 99 -113, three; and peptide 105-119, two (Fig. 1B). Other peptides showed stimulating activity but induced only one single T cell line (Fig. 1). The mean number of MDK-specific T cell lines per donor was three, and only peptide 14 -28 elicited a CD4 ϩ T cell response in at least half of the donors. Two donors (332 and 342) were nonresponders. Overall, the CD4 ϩ T cell response to MDK was of moderate intensity and involved mainly peptides 9 -23 and 14 -28. These two peptides are both located in the N-terminal part of the protein and overlap the signal peptide substantially (amino acid positions . Restriction to HLA-DR and -DP4 of MDK Peptides-Because the MDK peptide-specific CD4 ϩ T cell responses were observed in donors of different HLA-DR allotypes, we evaluated the capacity of the overlapping MDK peptides to bind to ten HLA-DR and two HLA-DP molecules that are common allotypes in the Caucasian population (Table 1). Consistent with the findings reported above, peptides 9 -23 and 14 -28 bound to nine and four HLA class II molecules, respectively. Peptides 1-15 and 4 -18, entirely within the signal peptide, also exhibited a broad specificity for HLA class II molecules. In contrast, few peptides of the mature protein bound to HLA class II molecules, the most promiscuous peptide being peptide 110 -124, which bound efficiently to four molecules. To identify the HLA class II molecules involved in the CD4 ϩ T cell response to MDK peptides, we introduced L cells, transfected with one specific HLA class II molecule, as antigen-presenting cells in the EliSpot assay (Fig. 2). CD4 ϩ T cell recognition of the peptide 9 -23 was restricted to T cell recognition by HLA-DR11, -DR7, -DR1, and -DP4 (top panels), whereas the CD4 ϩ T cell response to peptide 14 -28 was restricted to HLA-DR4, -DR7, -DRB4, and -DP4 (middle panels). Peptide 14 -28 also seemed to bind to HLA-DR13 as the peptide was presented to the T cell line 403.13 by the 06AD EBV cell line (DRB1*13:01/DRB3*01:01) but not by STEILIN (DRB1*03:01/DRB3*01:01). The peptides 99 -113 and 64 -78 seemed also to be presented by HLA-DR13 to the T cell line 403.5 and 403.7, respectively. The presentation of the peptide 74 -88 seemed to be restricted to HLA-DR3 (bottom panels). Multiple HLA-DR and DP4 molecules therefore appeared to be involved in the presentation of MDK peptides to T cells. This was particularly true for the 9 -23 and 14 -28 peptides that bound to several HLA-DR molecules.
The Peptides 9 -23 and 14 -28 Are Presented Differentially from Tumor Lysates to T Cells-We then investigated whether the identified MDK CD4 ϩ T cell epitopes were appropriately presented by autologous DCs loaded with lysates of tumor cell lines expressing MDK (Fig. 3). T cell lines specific for the peptides 64 -78 (403.7), 99 -113 (403.5), and 78 -92 (403.37) were indeed stimulated by autologous DCs loaded with a lysate of the hepatocellular carcinoma cells HepG2 expressing the MDK protein. They however did not react to DCs loaded with a lysate of the cervical cancer cells HeLa, which did not express MDK. In contrast, the three T cell lines specific for the peptides 14 -28 (Fig. 3A) were not stimulated by either HepG2 or HeLa lysates. To confirm these results, we used HeLa cells transiently transfected with MDK in the T cell presentation assays (Fig. 3B). Two T cell lines specific for the peptide 9 -23 were activated by the lysates of MDK-transfected HeLa cells and HepG2 cells but not by a lysate of nontransfected HeLa cells. By contrast, the two T cell lines specific for the peptide 14 -28 reacted with none of the lysates (HeLa, HeLa transfected with MDK, and HepG2). We concluded from these experiments that epitopes 64 -78, 78 -92, and 99 -113 present in the mature form of MDK were processed appropriately by immature DCs loaded with lysates of MDK-expressing tumor cells. This was also the case for    epitope 9 -23, mostly in the signal peptide (position 1-22). By contrast, epitope 14 -28, which straddles the junction between the signal peptide and the mature protein, did not appear to be presented by immature DCs loaded with the lysates. The Peptides 9 -23 and 14 -28 Are Presented Differentially to T Cells from a MDK-expressing Tumor Cell Line-To investigate the direct presentation of CD4 ϩ T cell epitopes from MDK-expressing tumor cell lines, we looked for tumor cell lines that concomitantly expressed MDK and HLA-DR molecules. Unfortunately, none of the HLA-DR expressing tumor cells of our cell collection also expressed MDK. We therefore produced a new cell line by transfecting A375 melanoma cells that express HLA-DR (HLA-DRB1*04:01/07:01) with an MDK-encoding plasmid (Fig. 4). The resulting stable clone (A375.MDK) exhibited a homogeneous level of MDK expression as assessed by intracellular staining and FACS analysis (data not shown). We then produced T cell lines specific for the 9 -23 and 14 -28 peptides using healthy donors matched for HLA-DR with the A375 cells: T cell lines 821.14 and 820.31 were isolated from donors 821 (HLA-DRB1*01:01/04:01) and 820 (HLA-DRB1*01:01/07:01), respectively. The 9 -23-specific T cell line 821.14 was stimulated by the peptide presented by PBMCs and by A375.MDK cells but not by A375 cells, demonstrating the specificity of the response. This response was inhibited by a anti-HLA-DR antibody (L243) but not by a anti-HLA-DQ antibody (SPVL3), strongly suggesting that the T cell line was restricted to HLA-DR molecules. By contrast, the 14 -28-specific T cell line responded to its peptide presented to PBMCs but not to A375.MDK or A375. The peptide MDK 9 -23, which is mostly within the signal peptide, was therefore displayed at the surface of the tumor cell A375.MDK bound to HLA-DR molecules, whereas peptide 14 -28, which overlaps the junction between the peptide signal and the mature protein was not.

DISCUSSION
In this paper, we have evaluated the capacity of MDK to prime CD4 ϩ T lymphocytes in humans and localized several CD4 ϩ T cell epitopes of MDK-restricted to different HLA-DR molecules. Two CD4 ϩ T cell epitopes, overlapping the MDK signal peptide but differing in their processing outcome in tumor cells, were responsible for a large proportion of the T cell response.
By in vitro stimulation of CD4 ϩ T lymphocytes collected from healthy donors, we derived peptide-specific T cell lines and identified several CD4 ϩ T cell epitopes in the MDK sequence including in its signal peptide. Antigen-specific T cell line derivation is a very sensitive method to detect antigenspecific T cells in multiple HLA-typed subjects (27,29). It usually generates a sufficient number of specific T cells to characterize the restriction elements and the antigen processing of the T cell epitopes, as intended in this study. We found that T cell lines induced by the peptides 9 -23, 64 -78, 78 -92, and 99 -113 were specifically stimulated by DCs loaded with lysates of tumor cells naturally expressing MDK or transfected with MDK. We also derived a T cell line specific for the peptide 9 -23, which was stimulated by a HLA-DR-compatible tumor cell line transfected with MDK. As proposed recently (30), these epitopes can be qualified as subdominant, at least, as T cells primed with the corresponding peptides were stimulated by the whole antigen presented in native conditions. They may therefore contribute to a tumor-specific response. The T cell epitopes we identified include the peptide 9 -23, which is more than half of the MDK signal peptide. The peptide 9 -23 was found T cell stimulating in vitro for approximately one third of the 25 naïve donors tested (data not shown), and the magnitude of T cell response varied from 1 to 6 independent 9 -23-specific T cell lines per donor. To our experience, this peptide induces a substantial T cell response. To our knowledge, few CD4 ϩ T cell epitopes contained in signal peptides have been reported (22,23). They have been delineated in signal peptides of the tumor antigen Muc-1 (23) and the Caf1 antigen of Yersinia pestis (22). As stated by the authors, BALB/c mice inoculated with metastatic MUC1-transfected murine DA3 mammary tumor cells exhibited significantly prolonged survival following vaccination with MUC1 signal peptide (23). Some HLA-DR1-transgenic mice infected with Y. pestis responded to peptides contained in the signal peptide of Caf1, suggesting that after cleavage from Caf1, the signal sequence is available for antigen presentation during infection (22). Signal peptides contain numerous hydrophobic residues that serve to anchor the nascent polypeptide to the membrane (31). These residues may also serve as anchor residues for HLA-DR molecules. Indeed, this is consistent with our binding experiments that revealed a large specificity of the MDK signal peptide for HLA-DR molecules. Hydrophobic and aromatic residues constitute primary anchor residues for the pocket P1 of all HLA-DR (32) and HLA-DP4 (26) molecules. These binding data also revealed that some of the MDK peptides were promiscuous, for example peptides 1-15 and 4 -28, and did not elicit any T cell response, as described previously for other peptides (27). We reported previously (16) that the MDK signal peptide also hosted HLA-A2restricted CD8 ϩ T cell epitopes, as has been found for numerous tumor and viral antigens (17)(18)(19). Again, the presence of several hydrophobic residues in signal peptides favors their binding to HLA-A2 (33). Thus, signal peptides appear to harbor both CD4 ϩ and CD8 ϩ T cell epitopes.
CD4 ϩ T lymphocytes are known to contribute in vivo to tumor immunity by sustaining CD8 ϩ T cell responses, by providing cytokines such as IFN-␥ and by recruiting phagocytes (20,34). It has been also proposed that tumor cells expressing HLA class II can be the target of CD4 ϩ T cells by presenting their specific epitopes (35). Experiments performed with DCs loaded with lysates and with the A375 tumor cells transfected with MDK demonstrated the tumor specificity of T cells specific for the MDK T cell epitopes: 64 -78, 99 -113, 78 -92, and the MDK T cell epitope 9 -23 contained in the signal peptide. These T cells are therefore expected to be stimulated in vivo by presentation of MDK present in the tumors but captured by the DCs and possibly by the tumors themselves. They could therefore exert helper functions in vivo and contribute to tumor immunity. Because of the 9 -23 also encompasses a HLA-A2restricted T cell epitope, this peptide should be able to elicit in vivo both CD4 ϩ and CD8 ϩ tumor-specific T cells. An interesting issue would be to investigate whether the endogenous expression of MDK in the tumors initiates specific immune responses in cancer patients, including CD4, CD8, and B cell responses. Because many tumor antigens do not generate such a spontaneous response (36), we do not exclude the lack of spontaneous responses to MDK. However, the preexisting T cells we have detected could be either recruited in vivo by multiple injections or amplified by ex vivo stimulations.
Most of the CD4 ϩ T cell lines raised against MDK peptides were specific for peptide 14 -28, which straddles the junction between the peptide signal and the mature protein. However, 14 -28-specific T cell lines did not react to DCs loaded with lysates of tumor cells expressing MDK or to HLA-compatible tumor cell lines transfected with MDK. Thus, peptide 14 -28 behaved as a cryptic epitope: 14 -28-specific T cell lines did not recognize the native antigen expressed by a tumor (30). T cells primed by the MDK peptide 14 -28 are therefore not tumorspecific. The signal peptide of secreted proteins, such as MDK, is cleaved in the lumen of the endoplasmic reticulum by signal peptidase and signal peptide peptidase (21). The cleavage site is at the 22 position of the MDK sequence, the 14 -28 epitope is presumably cleaved by signal peptidase and therefore not displayed in MDK produced by the tumor to T cell presentation by HLA class II molecules. This cleavage may also contribute to protect the epitope-specific T cell repertoire from negative selection or peripheral tolerance. It is consistent with the superiority of the 14 -28 peptide over all of the other peptides for priming specific T cells. Peptide peptidase has been implicated in the processing of the preprocalcitonin tumor antigen (19). Presentation of the preprocalcitonin 16 -25 CD8 ϩ T cell epitope is inhibited by the preincubation of tumor cells with the serine protease inhibitor dichloroisocoumarin. However, in our experiments dichloroisocoumarin was toxic to the A375 mela-noma cell line, even a low dose, such that they were uninformative (data not shown). Proteases involved in the MHC class II pathway are difficult to identify unambiguously using inhibitors (37,38). This may be because the peptide protrudes from the HLA class II molecules and does not require a very precise length to be presented to T cells or because other redundant proteases might cleave the peptide (39). Formally, we therefore can only conclude that location of this 14 -28 cryptic epitope is consistent with peptide peptidase being involved in its cleavage during the MDK elongation.
We delineated several CD4 ϩ T cell epitopes with a broad specificity for HLA class II molecules, but the capacity of the mature form of MDK to prime a CD4 ϩ T cell response in humans is limited. Most of the T cell lines are specific for the peptides 9 -23 and 14 -28, and very few are specific for the mature form of MDK. The number of T cell lines specific for the mature form of MDK we obtained was small compared with the results with other antigens such as survivin (40) or viral peptides (27,29). In our assays, the yield of T cell lines depends on the number of preexisting T cells in the blood of the donors at the initiation of the T cell culture. It was not the goal of this study to evaluate the size of the MDK-specific T cell repertoire, as we have already described for other proteins (41)(42)(43), but the low yield of T cell lines suggests that there were few preexisting CD4 ϩ T cells specific for the mature form of MDK. A low level of MDK is detected in normal subjects (44), and this residual expression may contribute to reduce the MDK T cell repertoire. According to others (45,46), injection of recombinant MDK is expected to induce a CD4 ϩ T cell response in humans beings with a limited amplitude and consequently be poorly immunogenic (42,43,47). This could be an advantage for therapeutic applications of the MDK protein to treat cardiac ischemia (48) and neuronal disorders (49).
In conclusion, by investigating the CD4 ϩ T cell response specific for MDK, we mainly demonstrated the relevance of a CD4 ϩ T cell epitope contained in its signal peptide to contribute to the tumor-specific T cell response. By contrast, another epitope, overlapping the junction between the signal peptide and the matured protein, was cryptic.