A Cryptic Vascular Endothelial Growth Factor T-Cell Epitope: Identification and Characterization by Mass Spectrometry and T-Cell Assays

Vascular endothelial growth factor (VEGF) is involved in various physiologic processes, such as angiogenesis or wound healing, but is also crucial in pathologic events, such as tumor growth. Thus, clinical anti-VEGF treatments have been developed that could already show beneficial effects for cancer patients. In this article, we describe the first VEGF-derived CD8⁺ T-cell epitope. The natural HLA ligand SRFGGAVVR was identified by differential mass spectrometry in two primary renal cell carcinomas (RCC) and was significantly overpresented on both tumor tissues. SRFGGAVVR is derived from a cryptic translated region of VEGF presumably by initiation of translation at the nonclassic start codon CUG⁴⁹⁹. SRFGGAVVR-specific T cells were generated in vitro using peptide-loaded dendritic cells or artificial antigen-presenting cells. SRFGGAVVR-specific CD8⁺ T cells, identified by HLA tetramer analysis after in vitro stimulation, were fully functional T effector cells, which were able to secrete IFN-γ on stimulation and killed tumor cells in vitro. Additionally, we have quantitatively analyzed VEGF mRNA and protein levels in RCC tumor and normal tissue samples by gene chip analysis, quantitative reverse transcription-PCR, in situ hybridization, and bead-based immunoassay. In the future, T cells directed against VEGF as a tumor-associated antigen may represent a possible way of combining peptide-based anti-VEGF immunotherapy with already existent anti-VEGF cancer therapies. [Cancer Res 2008;68(7):2447–54]

Vascular endothelial growth factor (VEGF), also known as vascular permeability factor, is a mitogen with high specificity for vascular endothelial cells. Its biological activities range from angiogenesis (1, 2), endothelial cell growth (3), and expression of antiapoptotic proteins, such as Bcl-2, in endothelial cells (4) to induction of vascular leakage (5) by activation of matrix metalloproteinase activity (6). Thus, VEGF is crucial in various physiologic and pathologic states, such as the formation of new blood vessels, wound healing, and tumor growth (see refs. 7, 8 for reviews).

The human vegf gene is organized in eight exons, which can be spliced alternatively into four major isoforms (VEGF₁₂₁, VEGF₁₆₅, VEGF₁₈₉, and VEGF₂₀₆; refs. 9, 10) and into two minor isoforms (VEGF₁₄₅ and VEGF₁₈₃; ref. 7). With increasing length of the VEGF isoform, the ability to bind to heparin also increases. VEGF₁₈₉ and VEGF₂₀₆ are highly basic and almost completely bound to the extracellular matrix. VEGF₁₂₁ is acidic, does not bind heparin, and thus is freely diffusible (11). VEGF₁₆₅ has intermediate properties regarding its secretion. As the mitogenic potency of VEGF decreases on loss of the heparin-binding region, VEGF₁₆₅ is twice as active as VEGF₁₂₁ (12) and thus has optimal bioavailability and biological activity (13–15). It has been shown that translation of the VEGF mRNA may be initiated not only at its presumed start codon at nucleotide 1039 (AUG¹⁰³⁹) but also at an in-frame CUG codon at nucleotide 499 (CUG⁴⁹⁹). Use of CUG⁴⁹⁹ has been described for VEGF₁₂₁, VEGF₁₆₅, and VEGF₁₈₉. Although the exact extent of CUG⁴⁹⁹ usage differs between reports (16, 17), one can assume that transcription of VEGF₁₂₁ and VEGF₁₆₅ starts at CUG⁴⁹⁹ in >50% of the cases.

VEGF is in the spotlight of cancer therapy as its presence is crucial for neoangiogenesis. A tumor growing beyond a volume of 1 to 2 mm³ requires its own blood vessels for a sufficient supply of nutrition and oxygen (18). Therefore, several strategies have been developed to hamper tumor vascularization by tackling VEGF signaling or levels of secreted VEGF. Several small molecules that interfere with VEGF signaling are being incorporated at present in clinical phase II trials. VEGF binds to two receptor tyrosine kinases (VEGFR-1 and VEGFR-2) present at the surface of most blood endothelial cells. VEGF binding leads to receptor dimerization, subsequent autophosphorylation, and activation of intracellular signaling cascades (19). VEGFR-1–expressing cells can be selectively killed by a recombinant fusion construct of VEGF₁₂₁ with the toxin gelonin (20). VEGFR-2, the major mediator of the mitogenic, angiogenic, and permeability enhancing effects of VEGF (7), can be blocked by several kinase inhibitors, such as sunitinib, sorafenib (21, 22), KRN951 (23), and AMG706 (24). To prevent VEGF from binding to its receptors, the monoclonal antibody bevacizumab has been developed. Bevacizumab is a humanized neutralizing VEGF antibody (25) that binds all biologically active forms of VEGF. In 2004, it was approved by the Food and Drug Administration for treatment of metastatic colorectal cancer. To summarize, various anti–VEGF-based cancer therapies have been developed, several of which have a substantial clinical effect.

Patient samples. No patient included in the experiments had received any anti–VEGF-based therapy. The following amounts of tissues were used for HLA peptide analysis: RCC099 (HLA-A*02, HLA-A*03, HLA-B*2705, and HLA-B*57): tumor tissue, 7.7 g; normal tissue, 12.4 g; RCC110 (HLA-A*02, HLA-A*68, HLA-B*18, and HLA-B*2705): tumor tissue, 7.0 g; normal tissue, 5.0 g.

Elution of HLA-presented peptides. HLA-presented peptides were obtained by immunoprecipitation of HLA molecules from solid tissues using an adapted protocol (26). One volume of 2× lysis buffer containing PBS, 0.6% CHAPS, and complete protease inhibitor (Roche Diagnostics) was added to shock frozen tissue samples. Subsequently, the samples were homogenized using a blender and afterwards by a potter. One volume of 1× lysis buffer was added and the lysate was stirred for 1 h at 4°C. After four times of 30 s of sonication, the sample was centrifuged at 3,000 × g, 4°C for 20 min, and afterwards at 150,000 × g, 4°C for 1 h, to remove cell debris. Finally, the sample was passed through a 0.2-μm filter (Sartorius). For immunoprecipitation, this lysate was applied for at least 12 h to a CNBr-activated Sepharose 4B column (40 mg Sepharose per mg tissue; GE Healthcare) to which the HLA-A–, HLA-B–, and HLA-C–specific antibody W6/32 had been coupled (1 mg antibody per mg tissue) as described by the manufacturer. After binding, the column was rinsed with 250 mL PBS and subsequently with 500 mL double-distilled water. For elution and dissociation of bound HLA-peptide complexes, the column was shaken at least four times in one bed volume of 0.1% trifluoroacetic acid (TFA) for 20 min at room temperature. The four TFA eluates were subsequently collected and combined. Finally, the HLA-presented peptides were isolated by ultrafiltration through a Centricon 10-kDa cutoff membrane (Millipore). For liquid chromatography-mass spectrometry (LC-MS) analysis, samples were freeze dried and resuspended in 0.1% formic acid or 0.1% TFA for modification.

Peptide modification and analysis. Modification of peptides (27) and peptide analysis (28) were carried out as described. For quantification, tumor and normal samples were mixed in a total peptide ratio of 1:1 and recorded in a single LC-MS experiment without fragmentation using a Q-TOF (Micromass). For sequence analysis, tumor and normal samples were analyzed separately in individual LC-MS/MS experiments. Interpretation of MS/MS fragmentation spectra was done as described elsewhere in detail (27).

Database searches. BLAST searches for SRFGGAVVR peptide sequence were performed using the PAM30 matrix (expect value < 1,000).⁶

Protein versus protein searches (pBlast) were done in 245,584 sequences deposited in the SwissProt database; all species were included in the search. Translated protein versus nucleotide searches (tblastn) were done for all species using the National Center for Biotechnology Information (NCBI) nonredundant database (4,934,654 sequences), and expressed sequence tag sequences were excluded.

Histology and in situ hybridization. Cells expressing VEGF mRNA were detected by radioactive in situ hybridization (ISH) using single-stranded ³⁵S-labeled RNA probes, which were synthesized from human VEGF₁₂₁ cloned into pBluescript II SK (Stratagene). This pBSK-hVEGF121 plasmid was generously provided by Georg Breier (Institute of Pathology, Dresden, Germany). Paraffin tissue sections (5 μm) were treated as described previously (29).

Reverse transcription-PCR and quantitative reverse transcription-PCR. RNA from cells was isolated using Trizol reagent (Invitrogen) according to the manufacturer's recommendations. cDNA was synthesized from 1 μg total RNA. Real-time quantitative reverse transcription-PCR (qRT-PCR) using SYBR Green PCR Master Mix (Applied Biosystems) and expression analysis was performed as described (30). Primers were purchased from biomers.net: VEGF₁₂₁, 5′-AACATCACCATGCAGATTATGC-3′ (forward) and 5′-GGCTTGTCACATTTTTCTTGTC-3′ (reverse); VEGF₁₆₅, 5′-GTGAATGCAGACCAAAGAAAG-3′ (forward) and 5′-TTTTTGCAGGAACATTTACACG-3′ (reverse); and 18S, 5′-CGGCTACCACATCCAAGGAA-3′ (forward) and 5′-GCTGGAATTACCGCGGCT-3′ (reverse). For absolute quantification, standard curves of VEGF₁₂₁ and VEGF₁₆₅ were prepared using plasmids coding for the respective VEGF. Linear ranges of standard curves ranged from 10 fg to 10 ng of VEGF coding DNA. To calculate absolute VEGF copy numbers per cell, C_T values of the VEGF standard curve were allocated to C_T values of 18S RNA isolated from 2,000 JY-BLCL cells.

Gene expression analysis by high-density oligonucleotide microarrays. RNA isolation from tumor and autologous normal kidney specimens as well as gene expression analysis by Affymetrix Human Genome U133 Plus 2.0 oligonucleotide microarrays (Affymetrix) were performed as described (31). Microarray data are available from the Gene Expression Omnibus repository⁷

with the accession number GSE8050.

VEGF protein quantification. Before the VEGF protein quantification, anti-VEGF beads were generated by covalent coupling of the anti-VEGF capture antibody to xMAP COOH Microspheres (Luminex). Tissue (40–80 mg) or 1 × 10⁷ X63-B*2705 cells were pulverized and proteins were extracted with a 10-fold excess of solubilization buffer: 50 mmol/L Tris (pH 7.5), 400 mmol/L NaCl, 1 mmol/L CaCl₂, 1 mmol/L MgCl₂, 1% Triton X-100, and 1× complete protease inhibitor (Roche Diagnostics). After 1-h incubation on ice, cellular debris was removed. The total amount of solubilized protein was determined by Bradford assay. Subsequently, VEGF protein quantification was done as described (32) using a Luminex 100 IS (Luminex) and counting at least 100 beads.

Transfection of eukaryotic cells. The mouse myeloma cell line X63-B2705 [X63-Ag8.653 transfected with human HLA-B*2705 and human β2-microglobulin; generously provided by Elisabeth Weiss, Anthropologie und Humangenetik, Ludwig-Maximillians-Universität München, Munich, Germany (33)] was electroporated with various VEGF constructs. Transfected cells were maintained in RPMI 1640 (C.C.Pro) medium containing 10% FCS (Pan), 1 mg/mL G418S (Biochrom AG), and 0.4 mg/mL hygromycin (Roche Diagnostics).

In vitro stimulation of human CD8⁺ T cells using dendritic cells. T-cell culture medium consisted of RPMI 1640, 25 mmol/L HEPES, 2 mmol/L l-glutamine (all from Invitrogen), 10% heat-inactivated human AB serum (C.C.Pro), 50 units/mL penicillin, 50 μg/mL streptomycin, and 20 μg/mL gentamicin (all from Lonza). Isolation of peripheral blood mononuclear cells (PBMC) and dendritic cells (DC) was performed as described (34). DCs were pulsed for 2 h with 25 μg/mL of SRFGGAVVR peptide in T-cell culture medium and washed thrice with PBS before subsequent usage. In vitro stimulations were initiated in 24-well plates with 25 × 10⁵ responder cells (autologous PBMCs, in vitro cultured for 1 wk) plus 5 × 10⁵ DCs per well in 1.5 mL of T-cell culture medium. Cryopreserved, peptide-loaded, irradiated autologous PBMCs were used for restimulation after priming with DCs.

In vitro stimulation of human CD8⁺ T cells using artificial antigen-presenting cells. Untouched CD8⁺ T cells were MACS enriched by negative depletion of PBMCs (Miltenyi Biotec). Stimulations were initiated in 96-well plates with 1 × 10⁶ responder cells in 250 μL of T-cell culture medium complemented with 5 ng/mL human interleukin (IL)-12 (PromoKine) plus 2 × 10⁵ microbeads coated with HLA-B*2705-SRFGGAVVR complexes, HLA-A*0201-NLVPMVATV complexes, plus anti-CD28 and anti-4-1BB antibodies as costimulatory signals (35). Subsequent stimulations with 80 units/mL human IL-2 (Chiron) were performed for 3 wk as described (34).

Tetramer staining. Tetramer analyses were performed with SRFGGAVVR-tetramer-APC, NLVPMVATV-tetramer-PE, and antibody CD8-PerCP clone SK1 (BD Biosciences). Cells were incubated with the antibody at 4°C for 20 min in the dark followed by 30-min incubation with fluorescent HLA tetramers in the dark at room temperature. After resuspending with PBS containing 1% paraformaldehyde, 0.5% bovine serum albumin, 2 mmol/L EDTA, and 0.02% NaN₃, cells were analyzed by flow cytometry on a four-color FACSCalibur cytometer (BD Biosciences).

Intracellular IFN-γ staining. Autologous PBMCs were pulsed with 10 μg/mL peptide and incubated with effector cells in a ratio of 1:1 for 6 h. GolgiStop (BD Biosciences) was added for the final 5 h of incubation. Cells were analyzed using a Cytofix/Cytoperm Plus kit (BD Biosciences), CD8-PerCP clone SK1 (BD Biosciences), and IFN-γ-PE (BD Biosciences). For negative controls, cells were incubated either with irrelevant peptide or without peptide, respectively. Stimulation with phorbol 12-myristate 13-acetate (PMA)/ionomycin was used as positive control. Cells were analyzed on a four-color FACSCalibur.

⁵¹Cr release assay. SRFGGAVVR-PE tetramer-positive CD8⁺ T cells were isolated using anti-PE MACS beads (Miltenyi Biotec). Cells were subsequently cultured in the presence of 5 × 10⁵ cells/mL irradiated fresh allogenic PBMCs, 5 × 10⁴ cells/mL irradiated LG2-EBV cells, 150 units/mL IL-2, and 0.5 μg/mL PHA-L (Roche Diagnostics). Target cells were labeled with 100 μCi Na₂ ⁵¹CrO₄ (Hartmann Analytic). Per well, 10,000 target cells were used at different E:T cell ratios in a standard 6-h ⁵¹Cr release assay. Varying between different cell types, spontaneous releases ranged from 800 to 1,400 cpm, and maximal releases were in the range between 8,000 and 15,000 cpm.

A cryptic VEGF-derived HLA ligand identified by MS. Renal cell carcinoma of the clear cell type (RCC) is a frequent genitourinary tumor that is very resistant to standard therapy. To discover potential targets for T-cell–based tumor immunotherapy, we have analyzed the HLA ligandome of several RCCs and their autologous normal kidney tissue in detail (27, 31). Therefore, we isolated HLA-presented peptides from tumor and normal tissue, labeled the peptides with a stable isotope tag (28), and subjected them to LC coupled to an electrospray ionization mass spectrometer (ESI-LC-MS). All normal tissue–derived peptides were labeled NH₂ terminally with nicotinic acid (H₄-NIC); all peptides derived from tumor tissue were labeled in exactly the same manner but with a nicotinic acid bearing four deuterium atoms (D₄-NIC). The introduced isotope labels allowed, due to the isotopic mass shift, an accurate allocation of peptide signals to tumor or normal tissue after mixing of the two samples in a total peptide ratio of 1:1. Comparing the signal intensity of a heavy D₄-NIC peptide with its light H₄-NIC counterpart allowed peptide quantification between tumor and autologous normal tissue in one ESI-LC-MS experiment (27).

In such an analysis of RCC099, a very prominent tumor-derived peptide with m/z 529.3 attracted attention (Fig. 1A). After normalization to the average peptide presentation ratio (27), the quantitative analysis of this peptide showed a >6-fold overpresentation of the peptide in tumor (D₄-NIC) compared with normal tissue (H₄-NIC; Fig. 1B^,, left). In RCC110, the same peptide seemed >5-fold overpresented on tumor tissue (Fig. 1B^,, right). The corresponding peptide sequence (SRFGGAVVR) was identified by de novo sequencing using MS/MS (Fig. 1C). Furthermore, SRFGGAVVR fragmentation spectra from RCC samples were identical with fragmentation spectra of synthetic SRFGGAVVR (Fig. 1C , bottom). The peptide matched the HLA motif of HLA-B*2705 (score = 30)⁸

for which both RCC099 and RCC110 were positive.

To identify the source protein from which the peptide SRFGGAVVR was derived, BLAST searches were performed against protein and translated nucleotide databases. Using the SwissProt protein database, no protein in any species was found containing the peptide SRFGGAVVR. However, searching in a translated nucleotide database, the SRFGGAVVR peptide was found in the “untranslated” region of VEGF in humans, chimpanzee, and rhesus monkey. SRFGGAVVR is in frame with the actual VEGF protein sequence and can be transcribed using the nonclassic CUG⁴⁹⁹ (16, 17) as start codon, thus being cryptic (Fig. 2). Furthermore, the COOH terminus of the peptide is predicted to be generated very efficiently by the proteasome (NetChop 3.0 software; score = 0.97).⁹

Integrated bioinformatic approaches, assessing the whole antigen processing pathway, rank SRFGGAVVR at least in the top 2% of all HLA-B*27–presented peptides of VEGF (WAPP score MHC = 1.05, TAP = −48.58, proteasome = 3.31;¹⁰ NetCTL, score = 1.68;¹¹ EpiJen, score = 1.498¹²).

Analysis of VEGF expression by ISH. To further characterize the cellular origin of VEGF leading to presentation of the SRFGGAVVR peptide in RCC tissue, ISH experiments with radioactively labeled VEGF RNA probes were performed in RCC tumors and autologous normal tissues. Lacking appropriate paraffin-embedded sample material for RCC099 and RCC110, this analysis was performed using three other clear cell RCCs (RCC135, RCC180, and RCC190) and their autologous normal tissues. In all analyzed samples, VEGF transcripts were only detected in tumor cells (Fig. 3A–C), thus being responsible for VEGF protein expression and subsequent presentation of the VEGF-derived peptide SRFGGAVVR. All autologous normal tissues (Fig. 3D–F), inflammatory cells (Fig. 3D,, if), as well as endothelial cells of blood vessels within the tumor tissues (Fig. 3A –C, bv) were negative for VEGF mRNA.

Expression profile of VEGF variants in RCCs. Gene chip analysis of RCC099 and RCC110 showed a 2- to 4-fold overexpression of VEGF in tumor tissue (Table 1). However, the gene chip probe set detecting VEGF did not allow determination of the variant of VEGF expressed in tumor and normal tissue, respectively. Similarly, the ISH probe allowed no specific detection of any VEGF variant because it had a minimum of 90% sequence overlap with each VEGF variant.

A RT-PCR was therefore set up using primers covering the VEGF sequence from the classic ATG¹⁰³⁹ start codon to the TGA¹⁶⁸⁶ stop codon. VEGF₁₂₁ (459 nucleotides) and VEGF₁₆₅ (591 nucleotides) could subsequently be identified as the major transcript variants of VEGF in all analyzed samples (Fig. 4A). Moreover, minor amounts of VEGF₁₄₅ (531 nucleotides) and VEGF₁₈₉ (663 nucleotides) could also be detected.

In the same samples, VEGF amounts were analyzed by qRT-PCR in an absolute manner using defined amounts of reference plasmids coding for VEGF₁₂₁ and VEGF₁₆₅. Thus, VEGF₁₂₁-specific qRT-PCR primers were designed. In the case of VEGF₁₆₅, no specific qRT-PCR primer pair could be defined, which did not also recognize VEGF₁₈₉. Due to the very low amount of VEGF₁₈₉ expression (Fig. 4A), its contribution to VEGF₁₆₅-derived signals in qRT-PCR was put aside. VEGF₁₆₅ was expressed in RCC099 3-fold higher amounts than in VEGF₁₂₁ and in RCC110 7-fold (Fig. 4B), respectively. Comparing tumor and normal tissue, RCC099 showed only minor differences of mRNA expression levels for both VEGF variants. In contrast, RCC110 showed a minimum of 12-fold overexpression of VEGF in tumor tissue.

Quantitative analysis of VEGF protein levels in RCCs. Correlation of mRNA with protein levels (36) and HLA ligand levels (27) is weak. Thus, protein levels of VEGF were assessed by a bead-based sandwich immunoassay (32) in RCC099 and RCC110 (Fig. 4C , left). Additionally, two other RCCs (RCC147 and RCC160), for which protein samples were available, were tested for VEGF protein expression. In all samples analyzed, VEGF protein expression was >4-fold higher in tumor compared with normal tissue. Protein ratios in both RCC099 and RCC110 were about four times higher as their corresponding mRNA ratios. This agrees with reports showing enhanced VEGF translation under hypoxic conditions (37).

Assessment of VEGF on various cellular levels. We analyzed VEGF on mRNA, protein, and HLA ligand levels (Table 1). Expression of VEGF mRNA in RCC tumor tissue has been described in several reports (38, 39). It is induced under hypoxic conditions mediated by the hypoxia-inducible factor (HIF)-1. In RCCs, in particular, it has been shown that inactivation of the von Hippel-Lindau tumor suppressor gene leads to excessive VEGF production (40) due to constitutive HIF-1 (41) and HIF-2 (42) activation. In 16 gene chip experiments (data not shown), we compared the expression of VEGF in primary RCC tissue and autologous normal kidney tissue and found VEGF overexpression (>3-fold) in ∼70% of all cases. In 25% of all cases, overexpression was >9-fold. For RCC099 and RCC110, overexpression in tumor tissue was 2.0- and 4.7-fold, respectively. These gene chip data were confirmed by qRT-PCR. VEGF₁₂₁ and VEGF₁₆₅ were the predominant VEGF variants expressed in RCC, both being significantly overexpressed in tumor tissue. On the protein level, VEGF overexpression in tumor tissue was even more prominent. Here, differences ranged from 4- to 40-fold. Furthermore, the VEGF-derived peptide SRFGGAVVR was overpresented on HLA molecules in both tumor tissues. The extent of HLA peptide overpresentation did not match its corresponding mRNA levels; this observation has been described previously for other gene products (27).

Generation of VEGF-expressing cell lines. For subsequent T-cell experiments, the mouse myeloma cell line X63-B*2705 (X63-Ag8 cells stably transfected with human HLA-B*2705 and β2-microglobulin) was used to circumvent contamination due to endogenous expression of human VEGF, which we could observe in human cell lines (see below). X63-B*2705 was stably transfected with two different VEGF constructs: VEGF-UTR165 was transfected with VEGF₁₆₅ containing the untranslated region and thus also containing the SRFGGAVVR peptide. VEGF-165 was transfected with VEGF₁₆₅ lacking the untranslated region and thus the SRFGGAVVR peptide. Surface expression of HLA-B*2705 was confirmed by fluorescence-activated cell sorting (FACS) analysis (data not shown), and VEGF protein expression was tested as described above (Fig. 4D).

Generation of VEGF-specific T cells. The VEGF-derived peptide SRFGGAVVR is presented by HLA-B*2705. To investigate whether the HLA-bound peptide can be recognized by functional cytotoxic T cells, CD8⁺ T cells from healthy blood donors were stimulated in vitro for 4 weeks with DCs or artificial antigen-presenting cells (aAPC). All blood donors were HLA-A*0201 and HLA-B*2705 positive as well as seropositive for human cytomegalovirus (HCMV). For stimulation experiments, DCs were loaded with SRFGGAVVR peptide; aAPCs were loaded with costimulatory antibodies, HLA-B*2705-SRFGGAVVR complexes, and HLA-A*0201-NLVPMVATV complexes. The immunodominant HCMV-derived peptide NLVPMVATV served as positive control. After 4 weeks of stimulation, cells were stained with fluorescence-labeled B*2705-SRFGGAVVR or A*0201-NLVPMVATV tetramers and assayed in a FACS experiment. Both the HCMV-derived peptide and the VEGF-derived SRFGGAVVR were able to induce a specific T-cell population (Fig. 5A). In total, 16 HLA-B*27–positive healthy blood donors were tested for reactivity against the SRFGGAVVR peptide. In five donors, SRFGGAVVR tetramer-positive T cells could be expanded. Regarding these donors, ∼5% of all stimulated wells (each containing 1 × 10⁶ CD8⁺ T cells) were positive in aAPC stimulations. The frequency of SRFGGAVVR tetramer-positive cells ranged between 0.5% and 83.4%. Thus, the precursor frequency of SRFGGAVVR-specific T cells in healthy blood donors was estimated being in the magnitude of 0.5 × 10⁻⁷ to 1 × 10⁻⁷ according to the calculations by Carroll and Ashcroft (43). Due to this low frequency of SRFGGAVVR-specific T cells even after stimulation, positive reactions in the nine nonreactive blood donors may have been missed.

Functional analysis of VEGF-specific T cells. To assess the functionality of SRFGGAVVR-specific T cells, their ability to secrete cytokines and their cytolytic potential were analyzed. For these functional experiments, only DC stimulations were used because it has been reported that strong and rapid T-cell expansion using aAPCs bears the risk of inducing dysfunctional T cells (44). On coculture with autologous PBMCs loaded with SRFGGAVVR peptide, CD8⁺ T cells were able to secrete IFN-γ. This confirmed that the effector T cells were functional and specific for VEGF (Fig. 5B). Furthermore, the cytolytic potential of these cells was analyzed in a chromium release assay. Therefore, in vitro–stimulated SRFGGAVVR-specific tetramer-positive T cells were MACS sorted and expanded. Autologous PBMCs loaded with an irrelevant peptide remained unlysed. Human C1R-B*2705 transfected with VEGF-UTR165 was lysed efficiently; however, the parental cell line C1R-B*2705 was lysed due to endogenous VEGF expression as well (data not shown). Mouse X63-B*2705 cells were recognized only after transfection with VEGF-UTR165 containing the cryptic SRFGGAVVR peptide sequence, whereas VEGF-165–transfected X63-B*2705 cells were not lysed, although VEGF-165 protein levels were 10-fold higher than in VEGF-UTR165 cells (Fig. 4D). Finally, a cell line generated from RCC110 tumor tissue was specifically recognized by these T cells. Due to the fact that further maintenance of this SRFGGAVVR-specific T-cell line was not possible, determination of T-cell receptor avidity by peptide titration as well as testing of an extended RCC panel was not possible. Furthermore, we were not able to expand all SRFGGAVVR tetramer-positive T-cell clones to an extent that performance of ⁵¹Cr release assays was possible.

Absence of VEGF-specific T cells in blood of RCC patients. Using PBMCs from healthy blood donors, SRFGGAVVR-specific T cells were inducible in vitro. To asses whether such SRFGGAVVR-specific T cells were induced in vivo in RCC patients, we analyzed PBMC samples of 12 HLA-B*27–positive RCC patients by tetramer analysis and by intracellular IFN-γ staining. In none of the samples, including RCC110, were tetramer-positive or IFN-γ–producing cells detectable, neither in an ex vivo analysis nor on 7 days of in vitro peptide stimulation (Fig. 5D). This suggests that SRFGGAVVR may not be spontaneously recognized by T cells from RCC patients, which may indicate a possibility for a therapeutic use of VEGF-specific T cells. However, it may also reflect a lower precursor frequency than in healthy blood donors or a peripheral tolerance against VEGF induced by the tumor, which would have to be broken for an immunotherapeutic approach using potent adjuvants.

Outlook. To predict possible organs in which a T-cell–based anti-VEGF therapy could have adverse effects, we analyzed VEGF mRNA expression in 32 different normal tissue specimens by gene chip experiments. Of all tissues analyzed, VEGF mRNA was only detectable in 13 tissues (see Supplementary Table S1). Only prostate tissue showed a >1.5-fold increase of VEGF expression compared with a commercially available standard kidney sample. In contrast, the average increase of VEGF in RCCs also compared with the standard kidney tissue was >5.5-fold.

However, the low amount of VEGF mRNA in the genitourethral tract, the intestine, and skeletal muscle suggests that a T-cell–based VEGF immunotherapy might lead to similar side effects as anti-VEGF antibodies, namely, hypertension, proteinuria, and bleeding or abnormalities in wound healing.

Grant support: Deutsche Forschungsgemeinschaft grants SFB-TR19 (A.O. Weinzierl, K. Klingel, and S. Stevanović) and SFB 685 (D. Maurer and S. Stevanović); Bundesministerium für Bildung und Forschung grant NGFN2 (O. Schoor); European Union, Cancer Immunotherapy, LSHC-CT-2006 518234 (H-G. Rammensee); and Jürgen Manchot Stiftung (A.O. Weinzierl).

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 Lynne Yakes for critical reading of the manuscript, Tobias Krüger for establishing a cell line from RCC110 tumor tissue, and Jörg Hennenlotter and Arnulf Stenzl for providing RCC sample material.