Stimulating T cell responses against patient-derived breast cancer cells with neoantigen peptide-loaded peripheral blood mononuclear cells

The PC-B-142CA (HER2⁺ subtype) and PC-B-148CA (triple-negative subtype) BCA cells were derived from the cancer tissues of two patients admitted at the Faculty of Medicine Siriraj Hospital, Mahidol University [13]. The cells were maintained in DMEM/F12 medium (Gibco, ThermoFisher Scientific, Waltham, MA) supplemented with 10% fetal bovine serum (FBS) and 1 U/ml penicillin G sodium and 1 mg/ml streptomycin (ThermoFissher Scientific, Waltham, MA) at 37 °C in humidified 5% CO₂ incubator.

The WBCs of BCA patients were extracted and used for DNA sequencing. The obtained DNA sequences were used as normal sequence references to identify the non-synonymous mutation (NSM) of PC-B-142CA and PC-B-148CA cells. PBMCs from healthy donors partially HLA-matched to these two cells were used to activate by neoantigen peptides for investigation of the neoantigen-activated T cell response. This process was approved by SIRB (COA no. Si 776/2019). All donors signed the informed consent forms.

Both genomic DNA and total RNA were extracted from PC-B-142CA and PC-B-148CA cells using an AllPrep DNA/RNA mini kit (Qiagen, Hilden, Germany), whereas germline control DNA was extracted from the patient’s WBCs (Fig. 1a). Genome libraries were prepared using TruSeq DNA Nano kit (Illumina, San Diego, CA), and sequenced on Illumina NovaSeq-6000 sequencer to generate 150 PE reads (Macrogen®, Seoul, South Korea). RNA sequence libraries were prepared using TruSeq Stranded mRNA Library Prep kit (Illumina) and sequenced as 100-bp paired-end reads using NovaSeq 6000 as a service of Macrogen®.

DNA sequence reads were aligned to the human reference genome (GRCh38/hg38) using BWA-mem (v0.7.15). Duplicated reads were marked and removed using the Picard tool. Somatic single-nucleotide variants (SNVs) were called using MuTect2 (v2.1) from the GATK package (v4.1.5). A matched normal sample was used as part of the somatic calling. The SNVs were annotated for NSM by variant effect predictor (VEP) with the Ensembl transcript set. Transcriptome sequence reads were processed using FASTQC for sequencing base quality control and aligned to the (GRCh38/hg38) using Spliced Transcripts Alignment to a Reference (STAR) tool. The number of reads that aligned to each gene was used to calculate gene expression as fragments per kilobase of transcript per million mapped reads (FPKM) (Fig. 1a).

For binding affinity prediction of neoantigen candidates, the HLA alleles of each patient were determined by PCR-sequence-specific primers at the Department of Transfusion Medicine Siriraj Hospital, Mahidol University. The 8- to 12-mer peptide sequences that contained the MT peptide and corresponding WT peptide sequence were predicted for binding affinities with HLA class I matched to the patient using MHCflurry [14], MHCnuggetsI [15], NetMHC [16], SMM [17], SMMPMBEC [18], and Pickpocket [19].

Neoantigen candidates were predicted using Personalized variant antigens by cancer sequencing (pVAC-Seq) pipeline [20] with some modifications including: (1) tumor DNA/RNA depth ≥ 10x, (2) tumor DNA/RNA variant allele fraction (VAF) ≥ 0.4, (3) FPKM > 1, (4) NetMHC of MT peptide (IC₅₀) ≤ 500 nM, (5) IC₅₀ MT ≤ 500 in 3 of 5 algorithms including MHCflurry, MHCnuggetsI, SMM, SMMPMBEC, and Pickpocket, and (6) Corresponding fold change (IC₅₀ WT/ MT) > 1 (Fig. 1a).

The initial atomic coordinates were built from the HLA-A*11:01 X-ray crystal structure complex with SSCSSCPLSK positive peptide (10 amino acids) (Protein Data Bank (PDB) ID: 5GRD) [21], HLA-A*24:02 complex with TYQWIIRNW positive peptide (9 amino acids) (PDB ID: 7JYW) and TYQWIIRNWET positive peptide (11 amino acids) (PDB ID: 7JYX) [22], whereas the HLA-C*07:02 complex with RYRPGTVAL positive peptide (9 amino acids) (PDB ID: 5VGE) [23]. The amino acid side chains of the positive peptides within HLA were modified to the neoantigen peptide by the Discovery Studio program (BIOVIA, Dassault Systèmes, Discovery Studio Visualizer Software, V21.1.0.20298, 2020). The missing hydrogen atoms of the HLA protein were added with the help of the LEaP module in the Assisted Model Building with Energy Refinement (AMBER) version 2020 [24]. The AMBER ff14SB [25] was taken for both the HLA and peptides. The complex geometries with the added hydrogen atoms were minimized with 1000 steps of the steepest descent method and subsequently by 3000 steps of conjugated gradient [26]. The solvation of each system was performed by the transferable intermolecular potential 3P water molecules in a periodic box. The systems were neutralized using Cl⁻ or Na⁺ counter ions. The water molecules were minimized only with 1000 steps of steepest descents and continued by 3000 steps of the conjugate gradient. Lastly, all systems were fully minimized by the same minimization process. The MD simulations were performed for 100 ns to reach equilibrium. The final 20 ns (from 80 to 100 ns) was then used to calculate the binding free energy of peptides using the molecular mechanics with generalized Born and surface area solvation (MM/GBSA) method. The steps of this method are summarized in Supplementary Figure S1.

Total RNA was extracted using a Total RNA extraction kit (Genemark, Taichung, Taiwan) and converted into cDNA by reverse transcription-polymerase chain reaction using SuperScript III First-strand Synthesis system (Invitrogen, ThermoFisher Scientific, CA). The cDNA template was amplified using specific primers (Supplementary Table S1) for each neoantigen with the Platinum™ Taq DNA Polymerase kit (Invitrogen). PCR products were run in 2% agarose gel, and the DNA was extracted from the gel with QIAquick gel extraction kit (Qiagen), measured the amount by NanodropTM Spectrophotometer (ThermoFisher Scientific, Waltham, MA), and sent for Sanger DNA sequencing (Celemics, Inc., Seoul, South Korea).

Neoantigen peptides and the corresponding WT peptides were purchased from GeneScript (Piscataway, NJ) with more than 98% purity. Reverse-phase HPLC was used to produce lyophilized peptides which then were reconstituted in DMSO (Sigma-Aldrich, Merck, Burlington, MA) and used for neoantigen-specific T cell stimulation.

Donors 1–3 (Supplementary Table S2) had HLA partially matched to PC-B-142CA cells, while donors 4–6 were partially matched to the PC-B-148CA cell. The PBMCs were used to generate neoantigen-specific T cells [27], with minor modifications. Briefly, PBMCs were isolated from 50 ml of peripheral blood using density gradient centrifugation in Lymphocyte Separation Medium (Corning, Corning, NY). RBCs were lysed using red blood cell lysis buffer. PBMCs were cultured at a density of 2 × 10⁶ cells/well with 10 μg/mL MT or WT peptides in AIM-V medium (Invitrogen) supplemented with 10% human AB serum (Sigma-Aldrich) and 20 U/ml recombinant human IL-2 (ImmunoTools, Friesoythe, Germany) and 10 ng/ml IL-7 (ImmunoTools) (only day 0) for 14 days at 37 °C in a 5% CO₂ atmosphere. PBMCs were re-stimulated with peptides in a medium containing 20 U/ml IL-2 at days 4, 8, and 12. PBMCs without peptide stimulation and cytomegalovirus, Epstein–Barr virus, and influenza virus (CEF) pooled peptides (Mabtech, Cincinnati, OH) were used as the negative and positive control.

IFN-γ production of neoantigen-specific T cells was determined using an IFN-γ ELISpot assay kit (Mabtech, Inc, OH). The effector lymphocytes (E) were added together with target cancer cells (T) in a ratio of E:T as 10:1. The 1 µg/ml biotinylated Mab 7-B6-1 (Mabtech, Inc, Cincinnati, OH) in 0.5% human AB serum was incubated at room temperature for 2 h and then 1:1000 alkaline phosphatase (ALP)-conjugated streptavidin (Mabtech, Inc) for 1 h. Then, 100 µl/well of BCIP/NBT plus (Mabtech) was added. The spots were photographed and automatically calculated by the CTL ImmunoSpot® Software (ImmunoSpot, Cleveland, OH).

1 × 10⁴ PC-B-142CA and PC-B-148CA cells were cultured in 96-well plates and peptide-pulsed lymphocytes were added at E:T of 20:1 and 40:1 for 24 h. The CellTiter 96 Aqueous One Solution Reagent (3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, inner salt; MTS) (Promega Corporation, Madison, WI) was used to measure the viable cancer cells by a microplate reader at OD490 nm. The cancer cell lysis was calculated according to the following formula:

Statistical analysis was performed using GraphPad Prism V (GraphPad Software, San Diego, CA). The results were presented as mean ± standard error of the mean (SEM) of three independent experiments. Statistical comparisons of two groups were performed using Student’s t test, and comparisons of more than two groups were performed using one-way analysis of variance (one-way ANOVA) with Tukey’s post hoc test. Data were considered significantly different when the P value < 0.05.

A total of 2513 and 3490 somatic mutations in PC-B-142CA and PC-B-148CA cells were detected (Fig. 1b, Supplementary Table S3a and b). After variant annotation, 646 NSMs were identified for PC-B-142CA and 652 for PC-B-148CA (Fig. 1b, Supplementary Tables S4a and b). The results showed 52 candidate neoantigens for PC-B-142CA cells and 4 neoantigens for PC-B-148CA cells (Fig. 1b, Supplementary Table S5a and b). The top 3 candidate neoantigens ranked by the highest IC₅₀ WT/IC₅₀ MT fold change and the characters of each peptide including amino acid substitution, HLA specificity, and binding affinity are summarized (Table 1). For PC-B-142CA cells, neoantigens identified from both cells and cancer tissue (Supplementary Table S6) were selected. No cancer tissue of PC-B-148CA was available; hence, the neoantigens were identified from only PC-B-148CA cells. The final 3 neoantigens of PC-B-142CA cells were ADGRL1^E274K, adhesion G protein-coupled receptor L1 with a mutation at position 274 where E (glutamate) was replaced by K (lysine); PARP1^E619K, poly (ADP-ribose) polymerases-1 with a mutation at position 619 where E was replaced by K; and SEC14L2^R43Q, SEC14-like lipid binding 2 with a mutation at position 43 where R (arginine) was replaced by Q (glutamine). Those of PC-B-148CA cells were LSR^I158P, lipolysis-stimulated lipoprotein receptor with a mutation at position 158 where I (isoleucine) was replaced by P (Proline); ALKBH6^V83M, alkB homolog 6 with a mutation at position 83 where V (valine) was replaced by M (methionine); and GAA^I823T, alpha-glucosidase with a mutation at position 823 where I was replaced by T (Threonine).

The Sanger sequencing results confirmed that PC-B-142CA cells contained the heterozygous point mutation of ADGRL1 (G–A transition) and homozygous point mutations of PARP1 (G–A transition) and SEC14L2 (G–A transition) (Fig. 2a–c), while PC-B-148CA had heterozygous point mutations of LSR (A–T transition), ALKBH6 (G–A transition), and GAA (T–C transition) (Fig. 2d–f).

The anchor residues of candidate neoantigen peptides that well bound in pocket groove of HLA class I molecules were measured as the binding distance between the P2 and the P9 of the candidate neoantigen peptides (Supplementary Figure. S2a). The results showed 16 Å or longer binding distance between the P2 and the P9 of candidate neoantigen peptides or corresponding WT peptide (Fig. 3a and 3c) demonstrating the suitable distance for the peptide binding to HLA [28].

The binding affinity of candidate neoantigen peptide and HLA class I molecule was calculated by MM/GBSA method at the 20-ns MD simulation where the system’s conformational stability is showing in the number of atom contacts, number of hydrogen bonds (Supplementary Figure. S2b–c), and video files (Supplementary Video. S1–S16). The binding affinity of the control peptides to the specific HLA class I molecules is shown in Supplementary Figure. S2d). The result showed that the ΔG_MM/GBSA of complexes between neoantigen peptides and HLA molecule: ADGRL1^E274K, PARP1^E619K or SEC14L2^R43Q with HLA-A*11:01 were − 14.54 kcal/mol, − 0.60 kcal/mol, and − 7.07 kcal/mol (Fig. 3b), while the ΔG_MM/GBSA of WT peptides with HLA-A*11:01 were 8.54 kcal/mol for ADGRL1^WT, − 8.23 kcal/mol for PARP1^WT, and − 4.28 kcal/mol for SEC14L2^WT (Fig. 3b). For PC-B-148CA-derived neoantigens, the ΔG_MM/GBSA of MT peptides complexed with HLA-A*24:02 were − 30.27 kcal/mol for LSR^I158F and − 24.88 kcal/mol for ALKBH6^V83M, while GAA^I823T peptide complexed with HLA-C*07:02 revealed − 16.73 kcal/mol (Fig. 3d). In comparison, the ΔG_MM/GBSA of LSR^WT, ALKBH6^WT and GAA^WT were − 29.02, − 18.01, and − 12.33 kcal/mol (Fig. 3d).

For PC-B-142CA-derived neoantigens, the SEC14L2^R43Q peptides induced neoantigen-specific T cell response exhibited significantly increased IFN-γ secretion compared to SEC14L2^WT peptide in all 3 donors (donor1: 32.33 ± 7.31 vs. 9.67 ± 7.31%, donor2: 52.67 ± 1.45 vs. 10.00 ± 2.51%, donor3: 598.30 ± 71.61 vs. 217.30 ± 13.86%) (Fig. 4a and b). The PARP1^E619K peptide significantly induced IFN-γ compared to PARP1^WT peptide in 2 donors (donor1: 14.67 ± 2.33 vs. 6.67 ± 1.33%, donor3: 442.00 ± 151.00 vs. 77.33 ± 35.24%) (Fig. 4a and b). The ADGRL1^E274K peptide, however, showed no induction of T cells compared to its WT peptide in all donors. Interestingly, the pooled MT peptides could activate T cells to produce IFN-γ with statistical significance to peptide-unpulsed T cells (UP) in donor 1. In addition, induction of IFN-γ secretion from SEC14L2^R43Q peptide-activated T cells was observed in all donors (Fig. 4a and b).

ALKBH6^V83M, GAA^I823T, and pooled MT peptides significantly induced IFN-γ from neoantigen-specific T cells obtained from only donor4 compared with the WT peptide counterpart (ALKBH6: 65.33 ± 23.18 vs. 14.50 ± 3.57%, GAA: 61.00 ± 16.09 vs. 12.67 ± 6.74%, pooled MT peptides: 57.00 ± 5.77 vs. 17.67 ± 6.69%) (Fig. 5a and b). The ALKBH6^V83M and GAA^I823T peptides could induce IFN-γ from T cells significantly more than those of unpulsed T cells in 2 donors (donors 4 and 5). Interestingly, the pooled MT peptides exhibited a significant increment of IFN-γ from peptide-activated T cells compared to the unpulsed T cells in all 3 donors. All 3 MT peptides could not activate T cells in donor6, while the pooled MT peptides induced T cells to produce IFN-γ more than that of unpulsed T cells.

To further characterize the responses induced by each candidate neoantigen, intracellular cytokine staining was performed to measure the levels of IFN-γ, CD107a, and CD69 in CD8⁺ T cells. The representative gating strategy of PARP1^E619K -induced CD8⁺ T cells activation from donor 1 and the percentages of IFN-γ, CD107a, and CD69 in CD8⁺ T cells were detected higher in PARP1^E619K MT peptide-pulsed T cells than those of WT peptide (Fig. 6a).

In PC-B-142-derived candidate neoantigens, the frequency of IFN-γ⁺CD8⁺ T cells in response to all 3 MT peptides (ADGRL1^E274K, PARP1^E619K, and SEC14L2^R43Q) or pooled MT peptide stimulation was significantly higher than that of unpulsed peptide condition in different donors (Fig. 6b). When compared to the corresponding WT peptide, SEC14L2^R43Q stimulation significantly increased IFN-γ⁺CD8⁺ T cells in all 3 donors (donor1: 18.65 ± 1.22 vs. 10.83 ± 1.00%, donor2: 3.40 ± 0.05 vs. 1.16 ± 0.25%, and donor3: 19.87 ± 1.42 vs. 8.43 ± 1.50%). PARP1^E619K significantly induced IFN-γ⁺CD8⁺ T cells in donor1 (26.07 ± 0.55 vs. 13.67 ± 0.50%). ADGRL1^E274K and pooled MT peptides increased IFN-γ⁺CD8⁺ T cells in donor2 (ADGRL1: 3.53 ± 0.50 vs. 1.38 ± 0.22%, pooled MT peptides: 4.93 ± 0.36 vs. 1.74 ± 0.28%) (Fig. 6b). The frequency of CD107a⁺CD8⁺ T cells that responded to PARP1^E619K was significantly increased in two donors (donor1: 4.23 ± 0.37 vs. 1.80 ± 0.06% and donor 3: 7.89 ± 0.79 vs. 4.16 ± 0.85%), and to SEC14L2^R43Q in all 3 donors (donor1: 6.16 ± 0.68 vs. 2.99 ± 0.75%, donor2: 3.76 ± 0.09 vs. 1.72 ± 0.43% and donor3: 11.78 ± 1.95 vs. 6.51 ± 1.08%) compared to those of WT peptides (Fig. 6c). CD69⁺CD8⁺ T cells were not significantly increased by the MT peptides (Fig. 6d).

ALKBH6^V83M, GAA^I823T, and pooled MT peptides significantly induced IFN-γ⁺CD8⁺ T cells compared to that of unpulsed peptide (Fig. 7a). The IFN-γ⁺CD8⁺ T cells responded to ALKBH6^V83M peptide stimulation was significantly higher than ALKBH6^WT peptide stimulation in two donors (donor4: 23.60 ± 1.68 vs. 10.70 ± 1.93% and donor6: 19.27 ± 2.07 vs. 6.26 ± 2.53%); and to GAA^I823T was significantly higher than GAA^WT stimulation in all 3 donors (donor4: 30.80 ± 7.72 vs. 10.20 ± 1.30%, donor5: 13.15 ± 1.70 vs. 5.030 ± 1.810% and donor6: 22.83 ± 1.93 vs. 14.27 ± 0.98%), and to pooled MT peptides was significantly higher than pooled WT peptides in donor4 (20.53 ± 3.92 vs. 5.92 ± 2.15%) and donor5 (11.55 ± 1.07 vs. 4.990 ± 1.79%) (Fig. 7a). The numbers of CD107a⁺CD8⁺ T cells were significantly increased by ALKBH6^V83M and pooled MT peptides stimulation compared to corresponding WT peptides stimulation in donor4 (ALKBH6: 23.10 ± 5.26 vs. 10.99 ± 1.27% and pooled MT peptides: 17.98 ± 1.05 vs. 8.150 ± 1.52%) and by GAA^I823T peptide stimulation in donor6 (16.67 ± 2.05 vs. 7.93 ± 0.80%) (Fig. 7b). GAA^I823T peptide significantly induced CD69⁺CD8⁺ T cells in donor6 compared to GAA^WT (12.93 ± 0.240 vs. 10.06 ± 0.443%) and in donors 5 and 6 compared to unpulsed condition (donor5: 12.93 ± 0.24 vs. 8.44 ± 0.25% and donor6: 12.93 ± 0.24 vs. 8.443 ± 0.25%) (Fig. 7c). The pooled MT peptides activated CD69⁺CD8⁺ T cells more than that of unpulsed in only donor6 (12.57 ± 0.29 vs. 8.443 ± 0.25%) (Fig. 7c).

The ADGRL1^E274K peptide-pulsed T cells from donor 2 (Fig. 8a) and PARP1^E619K peptide-pulsed T cells from donors 1 and 3 (Fig. 8b) significantly increased cytotoxicity against PC-B-142CA cells compared to those of WT peptide in a dose-dependent manner. SEC14L2^R43Q and pooled MT peptides-pulsed T cells from all donors exclusively killed PC-B-142CA target cells compared to those of WT peptides (Fig. 8c and d). No difference in PC-B-148CA cell lysis after co-cultured with LSR^I158F or LSR^WT peptide-activated T cells (Fig. 8e). However, the ALKBH6^V83M -, GAA^I823T peptide- and pooled MT peptides-pulsed T cells significantly induced PC-B-148CA cell killing at 40:1 ratio (Fig. 8f–h).