Neoantigen landscape supports feasibility of personalized cancer vaccine for follicular lymphoma

Abstract

Personalized cancer vaccines designed to target neoantigens represent a promising new treatment paradigm in oncology. In contrast to classical idiotype vaccines, we hypothesized that "polyvalent" vaccines could be engineered for the personalized treatment of follicular lymphoma (FL) using neoantigen discovery by combined whole-exome sequencing (WES) and RNA sequencing (RNA-seq). Fifty-eight tumor samples from 57 patients with FL underwent WES and RNA-seq. Somatic and B-cell clonotype neoantigens were predicted and filtered to identify high-quality neoantigens. B-cell clonality was determined by the alignment of B-cell receptor (BCR) CDR3 regions from RNA-seq data, grouping at the protein level, and comparison with the BCR repertoire from healthy individuals using RNA-seq data. An average of 52 somatic mutations per patient (range, 2-172) were identified, and ≥2 (median, 15) high-quality neoantigens were predicted for 56 of 58 FL samples. The predicted neoantigen peptides were composed of missense mutations (77%), indels (9%), gene fusions (3%), and BCR sequences (11%). Building off of these preclinical analyses, we initiated a pilot clinical trial using personalized neoantigen vaccination combined with PD-1 blockade in patients with relapsed or refractory FL (#NCT03121677). Synthetic long peptide vaccines targeting predicted high-quality neoantigens were successfully synthesized for and administered to all 4 patients enrolled. Initial results demonstrate feasibility, safety, and potential immunologic and clinical responses. Our study suggests that a genomics-driven personalized cancer vaccine strategy is feasible for patients with FL, and this may overcome prior challenges in the field. This trial was registered at www.ClinicalTrials.gov as #NCT03121677.

Graphical abstract

Figure 1.

Overview of the FL personalized cancer vaccine pipeline. Patient samples are acquired and then sequenced (top left). Somatic variants of various types, including SNVs (blue), deletions (red), insertions (green), and fusions (pink), are predicted. Sequence data are analyzed to determine HLA types and B-cell clonotypes for each patient. Variant and clonal B-cell peptide sequences are inferred from variants and analyzed with respect to their predicted expression, proteasome processing, and ability to bind the patient’s MHC class I complexes. Candidates are then selected for vaccine design, and additional analyses are performed to assess manufacturability. Bioinformatic tools used for each step are indicated in italics. CDR3, complementarity-determining region 3; IEDB, Immune Epitope Database. Adapted, per CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/), from Richters et al.

Figure 2.

Recurrently mutated genes and mutation burden observed for all patients. The bar graph on the top corresponds to the number of total mutations per patient and is colored by mutation type. The bar graph on the left corresponds to the percentage of mutations for a given gene for the entire cohort. Columns represent each sample in the cohort (1 patient, LYM120, with both tumor [t] and relapse [r] samples is shown) and are ordered by the presence of mutations in the most to least frequently mutated gene. The third plot indicates the presence or absence of a mutation for each patient and gene combination, colored by mutation type. If a patient has multiple mutations for an individual gene, it is colored according to the priority order as indicated in the mutation type legend, from top to bottom. A white star indicates which mutations are predicted to result in high-quality neoantigen vaccine candidates.

Figure 3.

Clonality analysis of BCR populations within healthy normal samples and FL samples. The plots show the composition of BCR repertoire of both normal (left half) and malignant (right half) samples for each of Ig chains separately: panels A-C correspond to heavy (blue), kappa (orange), and lambda (green) chains, respectively. In each panel, the upper histogram shows the coverage of the given Ig chain in the sample (log₁₀ of read counts), whereas the lower stacked bar plot shows BCR repertoire structure. Each bar from bottom to top is composed of 10 dark sections representing the fraction of repertoire for the 10 largest clones in the sample and a single top light gray section representing all other (minor) clones. The colored bottom sections depict the single most dominant clone that exceeded the cutoff value of fraction in the sample (9%) and had sufficient overall coverage (>40 reads for malignant samples whereas normal samples were preselected having >100 reads for each chain, see “Methods”). When both light chains passed the cutoff, only the one with the larger fraction was selected and colored. Stars below each panel indicate major clonotypes predicted to result in one or more high-quality neoantigen vaccine candidates.

Figure 4.

Personalized neoantigen cancer vaccine identification and prioritization. (A) Swarm plots display the number of vaccine candidates on the y-axis for the entire cohort at each stage of filtering (x-axis), moving from left to right. The bar graphs depict the numbers of final neoantigen vaccine candidates for each patient, colored according to source (SNV, indel, BCR, and fusion) and are sorted from least to most total candidates. Bar graph (B) includes nonneoantigen mutations that did not pass neoantigen filtering, whereas bar graph (C) only contains final candidates. The red line in panel C depicts the minimum cutoff of 2 candidates required for potential vaccine design.

Figure 5.

Molecular functional profiling of APM and MHC. (A) Somatic mutations in APM genes. Only 7 mutated genes of the total 32 APM genes considered. (B) Depicts median transformed log₂(TPM + 1) RNA expression levels of MHC class I and II genes. ssGSEA, single sample gene set enrichment analysis; TPM, transcripts per million.

Figure 6.

Implementation of neoantigen vaccines in pilot clinical trial. (A) Trial overview and timeline for patient FLNA-04 (see supplemental Figures 11 and 12 for more detailed versions). (B) The table lists all variants that were predicted to result in high-quality neoantigen vaccine candidates along with corresponding HLA allele, selection criteria, long peptide sequence submitted, synthesis success status, and vaccination pool (supplemental Tables 11 and 12). Short epitope sequences that binding predictions were based on are bolded within the long peptide sequences submitted column. (C) Peptide stabilization of selected candidate neoantigens. Various concentrations of peptides were incubated overnight with ICP-47 expressing (TAP deficient) B-cell line expressing HLA A∗68:01 heavy chain, washed and then stained with W6/32 APC. Mean fluorescence intensity (MFI) of cells pulsed with decreasing peptide concentrations were compared with no peptide pulsed control cells to validate predicted peptide stabilization of MHC class I molecule. (D) PBMCs from C6 apheresis were pulsed with vaccinating peptides and cultured for 12 days in vitro and then challenged with predicted short peptides overnight on IFN-γ–coated ELISPOT plates. (E) Bar graph shows triplicate value SPUs per million PBMCs from D12 culture. (F) An example positive antigen-specific CTL enrichment detected by peptide loaded tetramers stained with tetramer PE and tetramer APC gated on live, CD3/CD8DP CTL from D12 CTL. (G) Antigen-specific CTL from D12 cultures were challenged with artificial APCs pulsed with specific peptides, incubated for 6 hours, and then IFN-γ or TNF–expressing CD3/CD8 DP cells were detected by FACS. APC, antigen-presenting cells; CTL, cytotoxic T lymphocyte; D12, day 12; EOT, end of treatment; FACS, fluorescence-activated cell sorting; MT, mutant; PE, phycoerythrin; PET/CT, positron emission tomography-computed tomography; TAP, transporter associated with antigen processing; TNF, tumor necrosis factor; Tx, treatment.