Testing of T cell–based cancer therapeutics often involves measuring cancer antigen–specific T-cell populations with the assumption that they arise from in vivo clonal expansion. This analysis, using peptide/MHC tetramers, is often ambiguous. From a leukemia cell line, we identified a CDK4-derived peptide epitope, UNC-CDK4-1 (ALTPVVVTL), that bound HLA-A*02:01 with high affinity and could induce CD8+ T-cell responses in vitro. We identified UNC-CDK4-1/HLA-A*02:01 tetramer+ populations in 3 of 6 patients with acute myeloid leukemia who had undergone allogeneic stem cell transplantation. Using tetramer-based, single-cell sorting and T-cell receptor β (TCRβ) sequencing, we identified recurrent UNC-CDK4-1 tetramer–associated TCRβ clonotypes in a patient with a UNC-CDK4-1 tetramer+ population, suggesting in vivo T-cell expansion to UNC-CDK4-1. In parallel, we measured the patient's TCRβ repertoire and found it to be highly restricted/oligoclonal. The UNC-CDK4-1 tetramer–associated TCRβ clonotypes represented >17% of the entire TCRβ repertoire—far in excess of the UNC-CDK4-1 tetramer+ frequency—indicating that the recurrent TCRβ clonotypes identified from UNC-CDK-4-1 tetramer+ cells were likely a consequence of the extremely constrained T-cell repertoire in the patient and not in vivo UNC-CDK4-1–driven clonal T-cell expansion. Mapping recurrent TCRβ clonotype sequences onto TCRβ repertoires can help confirm or refute antigen-specific T-cell expansion in vivo. Cancer Immunol Res; 3(3); 228–35. ©2015 AACR.
Part of measuring the efficacy of T cell–based therapies such as cancer vaccines, adoptive T-cell therapies, and allogeneic stem cell transplantation (SCT) involves quantifying antigen-specific T-cell responses, which is usually accomplished using peptide/MHC tetramers (1–4). Although the enumeration of tetramer+ cells is considered a measurement of antigen-specific T-cell expansion, this assumption may be invalid in the setting of low-frequency, or low-affinity, responses—as is often the case in the measurement of cancer antigen–specific T-cells.
Antigen-specific T-cell expansion should yield clonal populations with recurrent TCRβ sequences in tetramer+ populations (5). In this report, we describe the use of tetramer-based fluorescence-activated cell sorting (FACS) to identify recurrent, potentially “antigen-specific” TCRβ CDR3 sequences, i.e., TCRβ clonotypes, which were then mapped onto a patient's TCRβ CDR3 repertoire. By comparing the frequency of tetramer+ cells with the frequency of tetramer-associated TCRβ CDR3 sequences in a patient's CD8+ T-cell repertoire, we were able to support in vivo clonal T-cell expansion to a viral antigen (pp65NLV) and refute clonal expansion to a potentially novel leukemia-associated antigen (UNC-CDK4-1, ALTPVVVTL) in an SCT patient despite detection of recurrent UNC-CDK4-1 tetramer–associated TCRβ clonotypes.
Materials and Methods
Detailed descriptions are included in Supplementary Materials and Methods.
Identification of HLA-A*02:01–restricted peptides by high-performance liquid chromatography–mass spectrometry (HPLC-MS)
A lysate of 6 × 109 HLA-A*02:01–transfected U937 cells (U937.A2) was cleared by ultracentrifugation, and the supernatant passed over a BB7.2-loaded HiTRAP recombinant protein A column. The BB7.2/HLA/peptide complexes were eluted with acetic acid, and the eluate passed through Microcon 3 K filters to yield peptide epitopes (6).
A Hitachi NanoFrontier Nano LC/linear ion trap time-of-flight mass spectrometer was used for online LC/MS-MS experiments. The peptide mixture was injected and subjected to data-dependent acquisition using collision-induced dissociation for peptide ion activation. MS/MS ion searching was performed using the Mascot search engine, with the no-enzyme option and non-identical protein database (NCBInr).
Western blot analysis
Twenty micrograms each of three human acute myelogenous leukemia (AML) peripheral blood mononuclear cell (PBMC) lysates, a healthy donor PBMC lysate, and a Jurkat cell lysate were electrophoresed on a 4% to 12% NuPAGE gradient gel and transferred to a polyvinylidene difluoride membrane. CDK4 was detected with a primary antibody (Abcam; ab75511) followed by a horseradish peroxidase–conjugated anti-mouse antibody. Bands were visualized using Amersham ECL Western blotting reagents.
iTopia affinity and off-rate assays
Epitope binding was measured using the iTopia Epitope Discovery System. For binding affinity, peptides were incubated in HLA-A*02:01–coated wells overnight, in the presence of the anti-HLA antibody, and fluorescence was read on a Synergy 2 microplate reader with results compared with the binding of the positive control peptide (FLPSDFFPSV, from Hepatitis B core protein) at 10−4 mol/L. The EC50 was determined using GraphPad Prism's nonlinear regression “log (agonist) versus response − variable slope (four parameter)” curve. For the off-rate assay, peptides were incubated in HLA-A*02:01–coated wells at 11 μmol/L overnight, then washed. Fluorescence was read at the times indicated on the graph. The t1/2 was calculated using GraphPad Prism's nonlinear regression, “dissociation − one phase exponential decay” curve.
UNC-CDK4-1–specific cytotoxic T-cell generation
Antigen-specific T cells were generated based on the method of Wölfl and Greenberg with some modifications (7). HLA-A*02:01–expressing monocyte-derived dendritic cells (DC) were generated following adherence to plastic and incubation with IL4 10 ng/mL and GM-CSF 800 IU with the addition of 10 ng/mL LPS, 100 IU/mL IFNγ on the fifth day. The DCs were pulsed with 20 μg/mL UNC-CDK4-1 peptide and irradiated at 30 Gy. Naïve CD8+ cells were isolated from the nonadherent fraction by negative selection using Miltenyi MACS beads with subsequent negative selection using anti-CD57 and anti-CD45RO beads. The naïve CD8+ cells and peptide-pulsed DCs were co-incubated at a ratio of 4:1 with IL21 at 30 ng/mL. On day 3 of coculture, IL15 at 5 ng/mL and IL7 at 5 ng/mL were added. Cultures were analyzed on day 11.
CD107/IFNγ T-cell activation assay
Antigen-specific activity was measured by flow cytometry quantifying CD107 and IFNγ expression described by Betts and colleagues (8). Autologous DCs were pulsed with 20 μg/mL of PR1 (VLQELNVTV) peptide, 20 μg/mL of UNC-CDK4-1, or left unpulsed. As a positive control, nonspecific stimulation with phytohemagglutinin (PHA) was also performed. T cells were mixed with DCs at a 1:1 ratio and incubated with phycoerythrin (PE)-labeled anti-CD107a, PE-labeled anti-CD107b, and anti-CD28/49d. After 1 hour, the cells were treated with Brefeldin A and monensin. After incubation for an additional 5 hours, cells were washed, permeabilized, and fixed. The cells were blocked with IgG and incubated with PerCP-labeled anti-CD8 and FITC-labeled anti-IFNγ. After 30 minutes, the cells were analyzed by flow cytometry.
Tetramer flow cytometry
PBMCs (5 × 105 to 1 × 106) from patients with cryopreserved post-SCT AML were incubated in DPBS with Pacific Blue–conjugated CD4, CD14, CD16, and CD19 (lineage) antibodies, FITC-conjugated CD8 antibody, and PE-UNC-CDK4-1/HLA-A*02:01 tetramer at 4°C for 25 minutes. Live/Dead Fixable Far Red Stain was added, and cells were incubated for 5 minutes at 4°C. Samples were washed and analyzed on a MACSQuant flow cytometer. Tetramer-positive cells were enumerated in the live (far red-negative), lymphocyte (forward scatter/side scatter), “lineage” (Pacific Blue) negative, CD8 (FITC)-positive gate (9–11). A negative tetramer and/or a fluorescence minus one (FMO) sample was used to set the tetramer gate.
FACS tetramer analysis and single-cell sorting
PBMCs from patient 5 were stained with CD8-Pacific Orange along with PE-HLA-A*02:01 tetramers generated with cytomegalovirus (CMV)-pp65NLV (NLVPMVATV), UNC-CDK4-1, and negative control peptides. For single-cell sorting, sort gates were determined by setting the CD8+ tetramer+ gate so that it included no cells in the negative tetramer sample. Tetramer+CD8+ T cells were sorted by an iCyt Reflection high-speed sorter at 1 cell/well into a 96-well PCR plate, each well containing 4 μL buffer (0.5× PBS, 10 mmol/L DTT, and 8 U RNaseOUT; ref. 12). The same method was used for single-cell CD8+ sorting, except no tetramers were used.
Single-cell PCR and sequencing
RT-PCR amplification and sequencing of TCRβ clonotypes was performed using multiplex primers (Supplementary Table S1) and nested PCR covering all human TCRβ variable region genes and the β-chain constant region (12). PCR products were treated with Exonuclease I and shrimp alkaline phosphatase and sequenced by the UNC Genome Analysis Facility. TCRβ sequence identifications were made using the SoDA software tool (13). Tetramer–PE fluorescence versus CD8-Pacific Orange fluorescence values were plotted for each T-cell's TCRβ clonotype using an R script.
Bulk TCRβ repertoire sequencing
PBMCs were isolated from a second sample from patient 5 obtained 2 months after the initial sample. Bulk CD8+ T cells were sorted, and RNA was extracted from 3 × 106 CD8+ T cells using the RNEasy Kit. The TCRβ repertoire was amplified using an RT-PCR protocol and the iRepertoire multiplex primer set (14). The sequencing library was run using 2 × 100 paired-end chemistry on an Illumina HiSeq 2000 sequencer in the UNC High Throughput Sequencing Facility.
High-throughput sequencing data analysis
Sequence data were processed using Python and R scripts developed in the laboratory. Paired-end reads were analyzed for the presence of the conserved invariant cysteine and FGXG motifs that define the CDR3 region in one reading frame. Sequences that did not exhibit this motif, had stop codons in the motif reading frame, or were present at less than three copies in the dataset were excluded. Vβ and Jβ gene identifications were made by either exact alignment or the highest-scoring Smith–Waterman alignment to the germline reference gene sequences annotated in IMGT (http://www.imgt.org/IMGTrepertoire; refs. 15, 16).
In vitro analysis of the UNC-CDK4-1 antigen
By HPLC-tandem MS, we identified 34 candidate leukemia-associated antigens (LAA) from U937.A2 cells (Supplementary Table S2) and focused on the CDK4-derived peptide ALTPVVVTL (Fig. 1A) because CDK4 is a therapeutic target in leukemia (17–19). Western blotting confirmed CDK4 overexpression in human AML but not in normal PBMCs (Fig. 1B). Using the iTopia assay, we measured an equilibrium EC50 of 2.4 μmol/L for UNC-CDK4-1 and a dissociation t1/2 > 8 hours, the longest time point in the assay (Fig. 1C and D).
Using UNC-CDK4-1–pulsed autologous DCs as stimulators, we generated a T-cell product that showed antigen-specific killing against UNC-CDK4-1 (Fig. 1E–H), and in experiments using UNC-CDK4-1–pulsed T2 cells as stimulators, we generated a T-cell population with 0.2% of CD8+ cells specific for UNC-CDK4-1 (Supplementary Fig. S1).
Measurements of UNC-CDK4-1 T-cell responses in post-SCT patients
PE-UNC-CDK4-1/HLA-A*02:01 tetramers were used to probe for UNC-CDK4-1–specific T-cell responses in 6 patients with HLA-A*02:01–expressing, post-SCT, AML (Table 1). Based on in vitro expansion experiments, we set a threshold for a tetramer+ population of >0.01% as possibly representing a UNC-CDK4-1–specific T-cell response. We observed UNC-CDK4-1/HLA-A*02:01 tetramer frequencies of >0.01% of CD8+ cells in 3 of the 6 patients (Fig. 2A–F; Supplementary Fig. S2), but all frequencies were low, raising the possibility that the identified cells did not represent a true antigen-specific response.
Antigen-specific T-cell responses generate clonally expanded T-cell populations with the same TCRβ clonotypes (5). We performed tetramer-based single-cell sorting with TCRβ CDR3 sequencing on a PBMC sample from patient 5 (Fig. 2E) using tetramers specific to UNC-CDK4-1 or the CMV pp65NLV epitope. The patient and donor were CMV seropositive before SCT. The patient had developed CMV viremia 2 months after SCT, but he was not viremic at the time of this sample collection. There were 5 recurrent TCRβ clonotypes in the pp65NLV-tetramer+–sorted cells and 6 recurrent TCRβ clonotypes in the UNC-CDK4-1 tetramer+ cells (Fig. 2G–J; Supplementary Table S3). The cells in each recurrent clonotype had the same nucleic acid and amino acid CDR3 sequence. For the pp65NLV tetramer+ cells, the TCRβ clonotype containing epitope ASRPFVNTEA had significantly higher tetramer affinity compared with the next two most prevalent pp65NLV-specific TCRβ clonotypes (Fig. 2I); however, no UNC-CDK4-1–specific TCRβ clonotype showed a statistically significant higher affinity (Fig. 2J).
Tetramer-associated TCRβ clonotypes map to the TCRβ repertoire at different frequencies
Using a 2-month subsequent sample from patient 5, we measured the frequency of the tetramer-associated TCRβ clonotypes (Fig. 2G–J) in the patient's CD8+ TCRβ repertoire. After removing sequences supported by ≤2 reads as possible sources of PCR or sequencing errors, our TCRβ repertoire contained 1.02 × 108 reads with 9.5 × 107 yielding functional TCRβ sequences. The frequency of different Vβ and Jβ combinations varied greatly (Fig. 3A). Furthermore, the diversity of CDR3 region amino acid lengths (analogous to TCRβ spectratyping) showed a nonparametric distribution, suggesting a highly constrained/oligoclonal TCRβ repertoire (Fig. 3B, Supplementary Table S4; refs. 20–22).
We graphed the 5,043 unique clonotypes according to their frequency in the dataset. The clonotype frequencies varied from one clonotype (V6-5_J1-1_ASWTGEGNTEA) representing 0.28 (28%) of the repertoire to 1,346 unique clonotypes, each having the minimum clonotype frequency (supported by three reads in the repertoire) of 3.16 × 10−8. All six UNC-CDK4-1 tetramer+–associated TCRβ clonotypes were present at unexpectedly high frequencies from 0.038% to 12%, whereas the three pp65NLV tetramer+–associated TCRβ clonotypes identified in the repertoire were present at between 0.00011% and 0.0016% (Fig. 3C). To test for amplification bias in the repertoire analysis, we obtained another sample from patient 5, 6 months after the sample used for repertoire analysis, and performed single-cell sorting of CD8+ cells with TCRβ CDR3 sequencing. From 95 sorted CD8+ cells, we obtained 86 TCRβ CDR3 sequences. All six recurrent UNC-CDK4-1 tetramer+–associated clonotypes were again present at extremely high frequencies (Fig. 3D). By comparing the frequency of the UNC-CDK4-1 and pp65NLV-associated TCRβ clonotypes in their respective tetramer+ gates (Fig. 2G and H) to their frequencies in the repertoire (Fig. 3C), we measured an enrichment ratio for each clonotype, i.e., clonotype frequency in tetramer+ gate/clonotype frequency in bulk repertoire (Fig. 3E). The mean enrichment ratio for the pp65NLV-associated clonotypes was 7.63 × 105, whereas the mean enrichment for the UNC-CDK4-1 was only 1.04 × 101.
Peptide/MHC tetramers are often used to measure antigen-specific T-cell responses in vivo; however, the data they provide can be ambiguous. As part of an antigen discovery project, we identified UNC-CDK4-1 as a potential LAA. CDK4 is overexpressed in AML, the UNC-CDK4-1 epitope has high affinity to HLA-A*02:01 (23), and tolerance to UNC-CDK4-1 can be broken. Based on these in vitro data, we tested for UNC-CDK4-1 responses in vivo.
Using UNC-CDK4-1/HLA-A*02:01 tetramers, we identified tetramer+ T cells at a frequency of >0.01%, higher than our sensitivity cutoff, in 3 of 6 patients with post-SCT AML. The tetramer results were ambiguous because of the low number of tetramer+ cells, and many of the gated CD8+ T cells had low affinity for the tetramer. To address this ambiguity, we performed tetramer-based single-cell sorting and TCRβ CDR3 sequencing to test for recurrent TCRβ clonotypes (5).
We isolated UNC-CDK4-1 and pp65NLV tetramer+ T cells from patient 5 and identified tetramer-associated recurrent TCRβ clonotypes. Although recurrent TCRβ clonotypes identified in a tetramer-defined population typically confirm the discovery of antigen-specific T-cell clonal populations (5), our repertoire mapping studies show that this is not the case in all situations.
Tetramer analysis of T cells of patient 5 identified a rare T-cell population (0.04%) recognizing UNC-CDK4-1; however, the recurrent UNC-CDK4-1–associated TCRβ clonotypes represented approximately 15% of the patient's CD8+ repertoire. In contrast, the patient had a T-cell population (0.07%; Supplementary Fig. S2) recognizing the pp65NLV antigen, and the recurrent pp65NLV-associated TCRβ clonotypes only represented 0.002% of the patient's TCRβ repertoire. The low frequency of pp65NLV tetramer+ cells is consistent with antigen-specific memory T cells in a CMV-exposed but not viremic subject, like this patient. Although the differences in clonotype frequency measured by tetramer compared with TCRβ repertoire analysis for pp65NLV-associated clonotypes could be explained by the 2-month difference in sample collection dates, tetramer gating and TCR repertoire sampling, it is unlikely that these factors explain the vast differences observed in the UNC-CDK4-1–associated clonotypes. By measuring the frequency of TCRβ clonotypes in a tetramer+ sample and in the bulk T-cell repertoire, we demonstrated that the pp65NLV-associated clonotypes were highly enriched in their tetramer+ gate, whereas the UNC-CDK4-1–associated clonotypes were not—suggesting that the recurrent UNC-CDK4-1–associated clonotypes are not antigen-specific, whereas the pp65NLV-associated clonotypes are antigen specific (Fig. 3E).
Overexpression of TCRs in some cells or preferential amplification of specific TCRβ CDR3 regions could affect TCRβ repertoire analysis. However, the primers used in the TCRβ CDR3 repertoire analysis minimize amplification bias (14), and our single-cell CD8+ TCRβ CDR3 sequencing reports the TCRβ clonotype for individual cells. Because the frequencies between the repertoire and CD8+ single-cell analysis are so similar (Fig. 3C and D), the highly restricted repertoire observed is likely an accurate representation of the T-cell repertoire of patient 5.
Our study did not analyze TCRα sequences, and although it is conceivable that the highly prevalent TCRβ sequences found in patient 5 could be paired with multiple different TCRα sequences, this phenomenon is unlikely, given the mechanism behind TCR assembly and has not, to our knowledge, been described (5, 24, 25).
In this patient's highly restricted (i.e., oligoclonal) T-cell repertoire, which is common in patients following SCT (21, 22), the majority of T cells identified in the UNC-CDK4-1 tetramer+ gate are derived from the expanded pools of T cells that are not antigen specific. Tetramer-based TCRβ clonotype analysis in a highly restricted T-cell repertoire can lead to clonotypically identical T cells stochastically being captured in a tetramer+ gate and misinterpreted as antigen specific. This potential error can be addressed by TCRβ repertoire mapping.
Disclosure of Potential Conflicts of Interest
S. Sarantopoulos is a consultant/advisory board member for Pharmacyclics. No potential conflicts of interest were disclosed by the other authors.
Conception and design: S.A. Hunsucker, B.G. Vincent, J.A. Frelinger, P.M. Armistead
Development of methodology: B.G. Vincent, A.A. Enyenihi, K.P. McKinnon, A.S. Buntzman, J.A. Frelinger, G.L. Glish, P.M. Armistead
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.A. Hunsucker, C.S. McGary, B.G. Vincent, A.A. Enyenihi, J.P. Waugh, L.M. Bixby, P.A. Ropp, J.M. Coghill, W.A. Wood, D.A. Gabriel, S. Sarantopoulos, T.C. Shea, G. Lizée, G.L. Glish, P.M. Armistead
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.A. Hunsucker, C.S. McGary, B.G. Vincent, A.A. Enyenihi, J.P. Waugh, J.S. Serody, G. Alatrash, G.L. Glish, P.M. Armistead
Writing, review, and/or revision of the manuscript: S.A. Hunsucker, B.G. Vincent, A.A. Enyenihi, T.C. Shea, J.S. Serody, G. Alatrash, G. Lizée, A.S. Buntzman, J.A. Frelinger, G.L. Glish, P.M. Armistead
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): B.G. Vincent, K.P. McKinnon, T. Rodriguez-Cruz, P.M. Armistead
Study supervision: P.M. Armistead
B.G. Vincent received salary support through T32 HL007149-37. P.M. Armistead received salary support through KL2 TR000084 and K08 HL113594-01.
Note: Supplementary data for this article are available at Cancer Immunology Research Online (http://cancerimmunolres.aacrjournals.org/).
- Received January 7, 2014.
- Revision received December 15, 2014.
- Accepted December 29, 2014.
- ©2015 American Association for Cancer Research.