Identification of Rare High-Avidity, Tumor-Reactive CD8⁺ T Cells by Monomeric TCR–Ligand Off-Rates Measurements on Living Cells

Binding of T-cell receptor (TCR) to peptide–MHC (pMHC) complex is the key step for T-cell activation and cellular immune responses (1). Efficient triggering of T-cell responses critically depends on the strength of TCR-mediated antigen recognition, namely it has been shown that strong binding to pMHC confers superior effector function than weak interactions (2–7). This is of particular interest for immunotherapy based on adoptive T-cell transfer, aiming to convey immune reactivity against tumor-associated antigens, for which endogenous T-cell responses are usually weak. Many tumor antigens are in fact self-antigens that are expressed in the thymus, and accordingly, most tumor-reactive T cells of high avidity become negatively selected, in contrast with pathogen-specific T cells (8). Therefore, there is a need for a robust technology that allows rapid identification and isolation of CD8⁺ T cells expressing TCRs capable to efficiently activate and enhance T-cell function against malignant cells.

TCR–pMHC binding parameters are typically assessed by surface plasmon resonance (SPR), which requires laborious and expensive production of soluble TCRs and ignores the TCR–pMHC avidity effects related to CD8 coreceptor binding. On the other hand, kinetic measurements using pMHC tetramers or multimers have been extensively studied on the surface of antigen-specific CD8⁺ T cells, but due to the multivalent and heterogeneous composition of these multimers, this approach does not allow to accurately determine TCR–pMHC dissociation rates (9). Binding and dissociation measurements should ideally be performed using monomeric pMHC complexes. Because TCR–pMHC interactions typically exhibit weak affinities and fast dissociation rates, this has long precluded conclusive measurements with monomeric pMHCs (10).

Nauerth and colleagues (11) recently measured monomeric TCR–pMHC dissociation kinetics using reversible Streptamers and reported that virus-specific CD8⁺ T cells with longer half-lives (low k_off) conferred increased functional avidity and better in vivo protection than T cells exhibiting shorter t_1/2 (high k_off). However, the Streptamer assay (11) needs a significant lag time until monomeric TCR–pMHC dissociation starts to become detectable, limiting thereby off-rate analyses to antigen-specific T cells of relative long half-lives, typically found in immune responses against pathogens. Moreover, accurate measurements of TCR–pMHC binding parameters on living T cells (11–14) require specialized equipment, which is currently not available for the high-throughput screen of antigen-specific T-cell populations.

To overcome these limitations, we here applied pMHC multimers built on reversible chelate complexes of Ni²⁺-nitriloacetic acid (NTA) with oligohistidines (15) allowing the efficient and direct cytometry-based assessment of monomeric TCR–pMHC dissociation kinetics on the surface of (self) tumor-specific CD8⁺ T cells. We found that the NTAmer technology accurately predicted T-cell biologic responses within a large panel of tumor-specific T-cell clones, providing novel means for the direct isolation of rare functionally relevant CD8⁺ T cells for adoptive cell transfer therapy.

The three HLA-A*0201–positive patients had stage III/IV metastatic melanoma and were included in immunotherapy studies (patient LAU 50, NCT00112242; patient LAU 155, NCT00002669; and patient LAU 618, NCT00112229; www.clinicaltrials.gov). The studies were designed and conducted according to the relevant regulatory standards, upon approval by the ethical commissions and regulatory agency of Switzerland. Patient recruitment, study procedures, and blood withdrawal were done upon written informed consent.

Human primary HLA-A*0201^pos CD8⁺ T lymphocytes were obtained following positive enrichment using anti–CD8-coated magnetic microbeads (Miltenyi Biotec), and cultured in RPMI supplemented with 8% human serum (HS) and 150 U/mL recombinant human IL2. HLA-A*0201-NTAmer^pos (NY-ESO-1_157–165–specific or Melan-A^MART-1-_26–35–specific) T cells from melanoma patient LAU 50 and LAU 618, respectively, were sorted by flow cytometry, cloned by limiting dilution and expanded in RPMI-1640 medium supplemented with 8% HS, 150 U/mL recombinant human IL2, 1 μg/mL phytohemagglutinin (PHA; Oxoid), and irradiated (30 Gy) allogeneic PBMCs as feeder cells. For patient LAU 155, twenty HLA-A*0201/NY-ESO-1_157–165–specific T-cell clones expressing dominant TCR BV1, BV8, and BV13 clonotypes and a nondominant TCR BV2 clonotype were selected from our previously generated database of T-cell clones (16), thawed and in vitro expanded before further use. Primary bulk CD8⁺ T cells and tumor-specific T-cell clones were expanded by periodic restimulation with 30-Gy irradiated allogeneic PBMCs and 1 μg/mL PHA.

Full-length codon-optimized TCR AV23.1 and TCR BV13.1 chain sequences of a dominant HLA-A*0201/NY-ESO-1_157–165 specific T-cell clone of patient LAU 155 (16) were cloned in the pRRL third-generation lentiviral vectors as an hPGK-AV23.1-IRES-BV13.1 construct and structure-based amino acid substitutions were introduced into the WT TCR sequence as described previously (5). Lentiviral production was performed using the calcium-phosphate method and concentrated supernatant of lentiviral-transfected 293T cells was used to infect TCRα knockout CD8⁺ SUP-T1 cells (ATCC number CRL-1942, mycoplasma free), CD8⁻null Jurkat T cells (ATCC number TIB-152; mycoplasma free), or primary CD8⁺ T cells overnight. Levels of transduced TCR expression on SUP-T1, Jurkat cells, and primary bulk CD8⁺ T cells were monitored with PE-labeled HLA-A*0201/NY-ESO-1_157–165 specific multimers (TCMetrix Sàrl) and FITC-conjugated BV13.1 antibody (Beckman Coulter).

NTAmers were synthetized by TCMetrix Sàrl (www.tcmetrix.ch) as described in ref. 15. NTAmers are composed of streptavidin-phycoerythrin (SA-PE; Invitrogen) complexed with biotinylated peptides carrying four Ni²⁺-nitrilotriacetic acid (NTA₄) moieties and noncovalently bound to His-tagged HLA-A*0201 monomers containing β2m or a Cy5-labeled β2m. Monomers were obtained by refolding of the HLA-A*0201 heavy chain in the presence of β2m or Cy5-labeled β2m containing the S88C mutation with Cy-5-maleimide (GE Healthcare) and the analog NY-ESO-1_157–165 [SLLMWITQA] or Melan-A^MART-1_26–35 [ELAGIGILTV] tumor antigenic peptide, optimized for enhanced HLA-A*0201 binding. After purification on a Superdex S75 column, pMHC monomers were mixed at a 10-fold ratio with SA-PE-NTA₄ in the presence of Ni²⁺ as described in ref. 15. The same procedure was used to prepare CD8-binding deficient HLA-A*0201 monomers bearing the D227K/T228A mutations in the HLAα3 domain (17). Dually labeled NTAmers containing SA-PE and Cy5-labeled monomers were used for dissociation kinetic measurements as described below. Single labeled NTAmers containing SA-PE and unlabeled monomers were used for flow-cytometric-based sorting of tumor-specific T cells. Once sorted, tumor-specific T cells were treated with 100 mmol/L imidazole at 4°C, allowing the rapid dissociation of the SA-PE-NTA₄ scaffold and pMHC monomers before in vitro T-cell cloning by limiting dilution.

TCR-transduced CD8⁺ SUP-T1 or CD8⁻ Jurkat cells (5 × 10⁵ cells) and tumor-specific CD8⁺ T-cell clones (2 × 10⁵ cells) derived from patients LAU 155, LAU 50, and LAU 618 were incubated for 40 minutes at 4°C with HLA-A*0201/NY-ESO-1_157–165 or HLA-A*0201/Melan-A^MART-1_26–35 NTAmers containing streptavidin-phycoerythrin and Cy5-labeled monomers in 50 μL FACS buffer (PBS supplemented with 0.5% BSA, 15 mmol/L HEPES, and 0.02% NaN₃). After a washing step, cells were suspended in 500 μL FACS buffer at 15°C (for SUP-T1 and Jurkat cells) or 200 μL FACS buffer at 4°C (for primary T-cell clones from melanoma patients) and cell surface-associated mean fluorescence was measured under constant temperature using a thermostat device (15°C for SUP-T1 and Jurkat cells and 4°C for primary T-cell clones) on a SORP-LSR II flow cytometer (BD Biosciences) following gating on living cells. PE-NTA₄ and Cy5-pMHC monomer fluorescence was measured before (between 30 seconds to 1 minute; baseline) and during 5 to 10 minutes after the addition of imidazole (100 mmol/L). High-resolution microscopy flow analysis was performed on an Amnis ImageStream^x Mark II instrument (Merck Millipore) using a ×40 objective. Staining was performed as described above for SUP-T1 cells. PE-NTA₄ and Cy5-pMHC monomer fluorescence was measured before (t = 0) and during 5 minutes upon the addition of imidazole (100 mmol/L) at 20°C. A 60 seconds lag time due to the automated handling of cell suspension by the Amnis instrument precluded earlier time-point measurements.

Flow-cytometric–based data were processed using the FlowJo software (v.9.6, Tree Star, Inc.). After gating on living cells, PE or Cy5 mean fluorescence intensity was derived using the kinetic module of the FlowJo software. Gates of 6 seconds period were created following addition of imidazole at the following time points: 15 seconds, 30 seconds, 45 seconds, 60 seconds, 90 seconds, 120 seconds, and then every minute for 10 minutes. Geometric MFI was measured at each time point after gating on HLA-A*0201/NY-ESO-1_157–165– or HLA-A*0201/Melan-A^MART-1_26–35–specific staining. Irrelevant Flu-specific gMFI values were systematically subtracted at the various time points. Corrected gMFI values were plotted and analyzed using the GraphPad Prism software (v.6, GraphPad).

A total of 5 × 10⁴ TCR-transduced SUP-T1 or primary bulk CD8⁺ T cells were loaded with 2 μmol/L Indo 1-AM (Sigma-Aldrich) for 45 minutes at 37°C. Cells were washed and resuspended in 250 μL prewarmed RPMI containing 2% FCS. Baseline was recorded for 30 seconds before 1 μg/mL of undissociated HLA-A*0201/NY-ESO-1_157–165 NTAmers (for SUP-T1) or 1 μg/mL HLA-A*0201/NY-ESO-1_157–165–specific multimers (for primary CD8⁺ T cells) were added to the cells allowing specific stimulation. Intracellular Ca²⁺ flux was assessed over 5 minutes under UV excitation and constant temperature of 37°C using a thermostat device on a LSR II SORP (BD Biosciences) flow cytometer. Indo-1 (violet)/Indo-1 (blue) 405/525 nm emission ratio was analyzed by FlowJo kinetics module software (TreeStar).

The functional avidity of antigen recognition was analyzed in a 4-hour ⁵¹Cr-release assay using TAP^−/−-deficient T2 (HLA-A*0201^pos) target cells pulsed with serial dilutions of the natural NY–ESO-1_157–165 peptide [SLLMWITQC] for tumor-specific T-cell clones derived from patient LAU 155 and LAU 50 or the natural Melan-A^MART-1_26–35 peptide [EAAGIGILTV] for tumor-specific T-cell clones derived from patient LAU 618. The NY-ESO-1 peptide was preincubated for 1 hour at room temperature with the disulfide-reducing agent Tris [2-carboxyethyl] phosphine (TCEP; 2 mmol/L, Pierce Biotechnology) before functional assays. The percentage of specific lysis was calculated as follows: 100 × (experimental − spontaneous release)/(total − spontaneous release). For T-cell clones of defined TCRα,β clonotypes previously derived from patient LAU 155 (16), antigen-specific recognition and lytic activity were further assessed against the melanoma cell lines Me 275 (HLA-A2^pos/NY-ESO-1^pos) and NA8 (HLA-A2^pos/NY-ESO-1^neg).

Statistical analyses were done with the GraphPad Prism software. A minimum of three independent experiments were performed to ensure a statistical power of 100% at α = 0.05 (Figs. 2 and 3), and a sample size ≥ 23 to ensure a statistical power of 80% at α = 0.05 (Fig. 5). Correlation analyses were performed using Pearson (Figs. 2 to 4) or Spearman (Fig. 5) coefficient r. The associated P value (two-tailed, α = 0.05) quantifies the likelihood that the correlation is due to random sampling. Two different investigators performed each TCR–pMHC dissociation kinetic experiment independently, and experiments described in Fig. 5 were performed blinded for both sample allocation and outcome assessment.

We previously developed a method for the isolation and analysis of antigen-specific cytotoxic T cells by reversible NTAmers (10, 15). Upon addition of imidazole at low, nontoxic concentration, the SA-PE-NTA₄ moieties rapidly decay (average dissociation half-life of 2.5 seconds), thereby releasing the pMHC monomers bound at the cell surface. Consequently, NTAmers allow FACS-sorting of CD8⁺ T cells without inducing adverse effects on the cell integrity (e.g., activation-induced cell death; refs. 10, 15). Taking advantage of the fact that NTAmers can be switched from stable binding to rapid dissociation, a two-color NTAmer was engineered to assess monomeric TCR–pMHC dissociation kinetics directly on living CD8⁺ T cells (Fig. 1A). We first evaluated this novel approach by direct visualization of the dissociation process on individual human CD8⁺ SUP-T1 cells expressing TCRs of increasing affinities for the HLA-A*0201–restricted tumor epitope NY-ESO-1_157–165 (18), using a flow cytometer generating simultaneous high-resolution microscopy imaging (ImageStream^X MarkII; Fig. 1B). As predicted, we observed strongest and most sustained fluorescence levels for T cells of very high TCR affinities (e.g., QMα and wtc51m). Because of a 60-second lag time due to the automated handling of cells following the addition of imidazole, the imaging approach precluded the visualization of labeled monomeric pMHC dissociation for T cells expressing TCRs ranging within the physiologic range (e.g., V49I, wild-type and DMβ). To solve this issue, we used a conventional flow cytometer (LSRII-SORP) equipped with a thermostat, which drastically reduced the lagging time to less than 5 seconds (Fig. 1C), thereby allowing the accurate assessment of monomeric pMHC dissociation kinetics on the surface of CD8⁺ T cells across a wide TCR affinity spectrum (Fig. 1D and Supplementary Fig. S1A). To validate the NTAmer technology, we next compared these dissociation values with SPR kinetics data obtained with the corresponding soluble TCRs (Table 1). Robust correlations were found between NTAmer-based cell surface monomeric pMHC dissociation kinetics (half-live t_1/2 and k_off), with both k_off rates and affinities measured by SPR, contrasting with the weaker correlations obtained with pMHC tetramers (Fig. 2A and Supplementary Fig. S2A).

To precisely quantify the contribution of CD8 coreceptor to monomeric TCR–pMHC dissociation, we generated NTAmer^227–228 variants containing the HLA-A*0201 D227K/T228A mutations in the MHC α3 domain that abolish CD8 binding to pMHC (17). Using the same panel of TCR-transduced CD8⁺ SUP-T1 cells, we show that abrogating CD8–MHC binding drastically diminished dissociation half-lives (t_1/2) over the whole spectrum of TCR affinities (Fig. 2B and Supplementary Fig. S1B and S1C). Monomeric off-rate (k_off and t_1/2) correlations between NTAmers^227–228 and SPR were highly significant (Fig. 2A and Supplementary Fig. S2A). Comparison between wild-type and CD8-binding deficient NTAmer-based half-lives revealed that CD8 stabilized by a factor of 3- to 4-fold the TCR-pMHC complex, and this was irrespective of the TCR affinity (Fig. 2B). A similar effect of CD8 binding to pMHC was observed when using TCR transduced Jurkat T cells that lack CD8 expression (Fig. 2C and D and Supplementary Fig. S2B) with k_off and t_1/2 values that were highly comparable with those found by NTAmers^227–228 or by SPR on CD8⁺ T cells (Table 1). Altogether, the NTAmer technology permits (i) the precise measurements of monomeric TCR–pMHC dissociation kinetics directly on living T cells expressing a wide TCR affinity range, including those typically found within the self/tumor-specific repertoires, and (ii) the quantification of the contribution of CD8 binding on the TCR–pMHC complex stability, which is not possible in SPR measurements.

Next we compared k_off rates obtained using NTAmers and the intracellular calcium mobilization on CD8⁺ SUP-T1 cells (Fig. 3A and B) and primary CD8⁺ T cells (Fig. 3D and E) expressing the panel of affinity-optimized TCRs against NY-ESO-1 tumor antigen. In line with our previous reports (19, 20), the Ca²⁺ flux was enhanced with increasing TCR affinity, up to reaching a plateau, while further affinity increase (K_D < 1 μmol/L) resulted in reduced calcium signaling. Within the physiologic TCR–pMHC affinity range, however, both SUP-T1 cells and primary T cells exhibited strong correlations between dissociation rates and Ca²⁺ mobilization (Fig. 3C and F); there was an enhanced Ca²⁺ flux in T cells of longer half-lives (low k_off), contrasting with the reduced signaling observed in T cells of shorter t_1/2 (high k_off). An overall reduction in Ca²⁺ flux was observed when using mutant multimer/HLA-A*0201^227–228 for specific stimulation, but this effect was less stringent for T cells expressing very high TCR affinities (e.g., TMα, QMα, and wtc51m; Fig. 3A and B). As previously reported (20–22), these data indicate that the degree of CD8 dependence for T-cell activation inversely depends on the affinity and dissociation half-life of the TCR–pMHC interaction.

We extended these observations to a clinically relevant setting, and characterized k_off rates and functional avidity of T cells obtained from a large set of HLA-A*0201/NY-ESO-1_157–165–specific CD8⁺ T-cell clones expressing well-defined TCRαβ clonotypes derived from LAU 155, a melanoma patient with long lasting antitumor T-cell responses (16). The dissociation rates of those cytotoxic T cell clones clustered within a narrow range (Fig. 4A and Supplementary Fig. S3A), and strongly correlated with EC₅₀ target cell killing (50% maximal lysis; Fig. 4C) and tumor cell recognition (Fig. 4E), with TCR clonotypes having long half-lives (low k_off) exhibiting increased T-cell cytotoxicity (Fig. 4B and D). Conversely, NY-ESO-1–specific TCR clonotypes with short half-lives (high k_off) showed reduced T-cell killing capacities.

To find out whether NTAmers allow direct identification of tumor-specific CD8⁺ T cells with high tumor killing capacity, we derived 147 tumor-specific T-cell clones from two other melanoma patients with detectable ex vivo T-cell responses against the HLA-A*0201–restricted tumor antigens NY-ESO-1_157–165 and Melan-A^MART-1_26–35, respectively, and screened them for monomeric dissociation rates (Fig. 5A and D and Supplementary Fig. S3B and S3C). Tumor-specific T-cell clones were distributed according to their TCR Vβ family expression using monoclonal antibodies and k_off rate analyses revealed large differences in half-lives between specific TCRs (Supplementary Fig. S4). Next, we performed killing assays with 57 clones selected for relatively low or high k_off values (Fig. 5B and E). We observed, for both antigenic specificities, a robust correlation between off-rates and the functional avidity (EC₅₀) determined with the cytotoxicity assay (Fig. 5C and F), demonstrating that NTAmers allow reliable assessment of surface-based TCR–pMHC dissociation kinetics and rapid selection of highly potent tumor-specific T-cell clones derived from different patients with cancer.

Identifying antigen-specific T cells that confer efficient effector function is critical for successful adoptive cell therapy. The stability of TCR binding for the peptide–MHC complex is a key determinant for T-cell activation. Several lines of evidence argue for a close relationship between cell surface dissociation kinetics (k_off) and functional T-cell responses (5, 23–26). Yet, rapid and accurate screening methods to measure TCR–pMHC binding kinetics are still needed to isolate antigen-specific T cells expressing TCRs of high binding avidity. Recently, we developed a novel approach combining reversible peptide–MHC multimers (NTAmers; ref. 15) and real-time flow cytometry (Fig. 1). Using SPR, we had previously determined the TCR–pMHC binding strength of sequence-optimized HLA-A*0201/NY-ESO-1_157–165–specific TCRs with increasing affinity of up to 150-fold from the wild-type receptor, including two outliers, a very low- and a very high-affinity TCR (5, 20). This TCR panel provides a unique model for validating monomeric TCR–pMHC dissociation kinetics at the surface of T cells by NTAmers. Here, we demonstrate that NTAmer-based off-rates (k_off, t_1/2) followed the same TCR–pMHC binding hierarchy than previously established (20), in excellent agreement to both binding parameters, k_off and the dissociation constant K_D, obtained by SPR (Fig. 2; Table 1).

Because of their switch ability, that is, high stability and rapid reversibility (<5 sec), NTAmers allowed accurate determination of dissociation rates, even for weak TCR–pMHC interactions, that is, fast off-rates, such as those found for (self) tumor-specific CD8⁺ T-cell repertoires (Fig. 4). Notably, the NTAmer approach differs from the Streptamer one, which is mostly limited to the detection of CD8⁺ T cells of high avidity such as virus-specific cells, as it requires a significant lag time (60 seconds) before monomeric TCR–pMHC dissociation becomes detectable (11). Moreover, the NTAmer-based assay represents a rapid and straightforward approach for the quantitative assessment of monomeric k_off rates on a large set of cloned antigen-specific CD8⁺ T cells derived from different patients and tumor epitopes (HLA-A*0201/NY-ESO-1_157–165 and HLA-A*0201/Melan-A^MART-1_26–35; Fig. 5). Importantly, we demonstrate robust correlations between dissociation off-rates and the biologic responses (e.g., calcium flux and target cell killing) on a large panel of CD8⁺ T-cell clones specific for two distinct tumor antigens, indicating that TCR–ligand k_off rate is a reliable predictor of T-cell function (Figs. 3–5).

Interactions between TCRs and pMHC are usually measured by SPR or pMHC tetramer or multimer staining, which requires one binding partner in soluble form. Both approaches have caveats. Because of their incomplete dissociation and multivalent nature, accurate off-rates data from pMHC tetramer/multimer staining measurements are imprecise. Conversely, monomeric dissociation k_off rates measured by the NTAmer technology were within the range of seconds to minutes, spanning a broad range (2-logs), as compared with a narrow range of minutes observed when using pMHC tetramers. Moreover, SPR fails to take into account rapid rebinding of the TCR to the same pMHC, because one of the two binding partners is constantly moving in the fluid phase, which impacts on the binding kinetics. Increased k_on rates have been shown to allow rapid rebinding after TCR–ligand dissociation, resulting in enhanced effective dissociation half-life of the TCR–pMHC interaction (25, 27). This may explain our observation that TCR variants with faster k_on (e.g., TMα and QMα) showed prolonged NTAmer-based dissociation half-lives compared with soluble monomeric off-rates measured by SPR (Table 1 and Supplementary Fig. S5).

The NTAmer-based approach further deviates from SPR measurements, as it provides data on living cells and includes contributions of CD8 to TCR–pMHC interactions. The CD8 coreceptor enhances antigen recognition and T-cell activation by stabilizing TCR–pMHC interaction at the cell surface (28–31) and recruiting p56^lck to TCR/CD3 complex promoting cell signaling (32, 33). Here, we demonstrate that CD8 strengthened the TCR–pMHC binding mainly by decreasing the TCR–pMHC dissociation by a factor of 3- to 4-fold (Fig. 2), as anticipated by previous tetramer dissociation assays (22, 30). Interestingly, the CD8 stabilization factor was independent of the TCR–pMHC affinity, in contrast with the CD8 dependence for T-cell activation, which can be directly linked to the affinity (3, 20, 22), and allows tuning the sensitivity and specificity of T-cell responses (34).

We recently provided new evidence that T-cell signaling and activation are optimal within a given TCR–pMHC affinity window (20), controlled through TCR affinity-mediated regulatory molecules, involving the inhibitory receptor PD-1 and SHP-1 phosphatase (19). Furthermore, while high-avidity T cells have been shown to control tumor growth, they become preferentially tolerized in the tumor microenvironment (35) or can target normal tissues expressing the cognate antigen (36, 37). Therefore, tumor-specific T cells of high avidity may not always be functionally better, and it remains to be fully determined to which degree intermediate or high-avidity T cells contribute to protective immunity. In this regard, NTAmers constitute a highly valuable tool for assessing the TCR–pMHC avidity and its relation to cell activation, signaling, and function in naturally or therapeutically induced tumor-specific T-cell responses, with TCR–pMHC affinities spanning within the physiologic range.

In summary, NTAmer technology enables efficient and direct interrogation of monomeric TCR–pMHC dissociation kinetics on a large set of living antigen-specific T cells by flow cytometry, and provides novel perspectives for rapid identification of rare functionally relevant tumor-reactive CD8⁺ T cells. Our approach may also be applicable to the analysis of other weak protein–protein interactions. Precise and widespread characterization of TCR–pMHC avidities will likely improve the development of T-cell–based immunotherapies in patients with cancer.

No potential conflicts of interest were disclosed.

Conception and design: M. Hebeisen, J. Schmidt, D.E. Speiser, N. Rufer

Development of methodology: M. Hebeisen, J. Schmidt, P. Guillaume, D.E. Speiser

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Hebeisen, J. Schmidt, P. Baumgaertner, N. Rufer

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Hebeisen, N. Rufer

Writing, review, and/or revision of the manuscript: M. Hebeisen, J. Schmidt, D.E. Speiser, N. Rufer

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.E. Speiser, I. Luescher

Study supervision: N. Rufer

The authors thank P. Gannon, M. Irving, O. Michielin, P. Romero, S. Siegert, and L. Zhang for essential collaboration and advice, and N. Montandon and M. van Overloop for outstanding technical assistance.

This work was supported by the Department of Oncology of the University of Lausanne and the Ludwig Center for Cancer Research of the University of Lausanne.

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.