Computational prediction of MHC anchor locations guides neoantigen identification and prioritization

Neoantigens are tumor-specific peptide sequences resulting from sources such as somatic DNA mutations. Upon loading onto major histocompatibility complex (MHC) molecules, they can trigger recognition by T cells. Accurate neoantigen identification is thus critical for both designing cancer vaccines and predicting response to immunotherapies. Neoantigen identification and prioritization relies on correctly predicting whether the presenting peptide sequence can successfully induce an immune response. Because most somatic mutations are single-nucleotide variants, changes between wild-type and mutated peptides are typically subtle and require cautious interpretation. A potentially underappreciated variable in neoantigen prediction pipelines is the mutation position within the peptide relative to its anchor positions for the patient's specific MHC molecules. Whereas a subset of peptide positions are presented to the T cell receptor for recognition, others are responsible for anchoring to the MHC, making these positional considerations critical for predicting T cell responses. We computationally predicted anchor positions for different peptide lengths for 328 common HLA alleles and identified unique anchoring patterns among them. Analysis of 923 tumor samples shows that 6 to 38% of neoantigen candidates are potentially misclassified and can be rescued using allele-specific knowledge of anchor positions. A subset of anchor results were orthogonally validated using protein crystallography structures. Representative anchor trends were experimentally validated using peptide-MHC stability assays and competition binding assays. By incorporating our anchor prediction results into neoantigen prediction pipelines, we hope to formalize, streamline, and improve the identification process for relevant clinical studies.

Personalized ctDNA micro-panels can monitor and predict clinical outcomes for patients with triple-negative breast cancer

Circulating tumor DNA (ctDNA) in peripheral blood has been used to predict prognosis and therapeutic response for triple-negative breast cancer (TNBC) patients. However, previous approaches typically use large comprehensive panels of genes commonly mutated across all breast cancers. Given the reduction in sequencing costs and decreased turnaround times associated with panel generation, the objective of this study was to assess the use of custom micro-panels for tracking disease and predicting clinical outcomes for patients with TNBC. Paired tumor-normal samples from patients with TNBC were obtained at diagnosis (T0) and whole exome sequencing (WES) was performed to identify somatic variants associated with individual tumors. Custom micro-panels of 4-6 variants were created for each individual enrolled in the study. Peripheral blood was obtained at baseline, during Cycle 1 Day 3, at time of surgery, and in 3-6 month intervals after surgery to assess variant allele fraction (VAF) at different timepoints during disease course. The VAF was compared to clinical outcomes to evaluate the ability of custom micro-panels to predict pathological response, disease-free intervals, and patient relapse. A cohort of 50 individuals were evaluated for up to 48 months post-diagnosis of TNBC. In total, there were 33 patients who did not achieve pathological complete response (pCR) and seven patients developed clinical relapse. For all patients who developed clinical relapse and had peripheral blood obtained ≤ 6 months prior to relapse (n = 4), the custom ctDNA micro-panels identified molecular relapse at an average of 4.3 months prior to clinical relapse. The custom ctDNA panel results were moderately associated with pCR such that during disease monitoring, only 11% of patients with pCR had a molecular relapse, whereas 47% of patients without pCR had a molecular relapse (Chi-Square; p-value = 0.10). In this study, we show that a custom micro-panel of 4-6 markers can be effectively used to predict outcomes and monitor remission for patients with TNBC. These custom micro-panels show high sensitivity for detecting molecular relapse in advance of clinical relapse. The use of these panels could improve patient outcomes through early detection of relapse with preemptive intervention prior to symptom onset.

Abstract 5639: Computational prediction of MHC anchor locations guide neoantigen prediction and prioritization

Neoantigens are novel peptide sequences resulting from somatic mutations in tumors that upon loading onto major histocompatibility complex (MHC) molecules allow recognition by T cells. Accurate neoantigen identification is thus critical for designing cancer vaccines and predicting response to immunotherapies. Neoantigen identification and prioritization relies on correctly inferring whether the presenting peptide sequence can successfully induce an immune response. As the majority of somatic mutations are SNVs, changes between wildtype and mutant peptide are subtle and require cautious interpretation. An important, yet underappreciated, variable in neoantigen-prediction pipelines is the mutation position within the peptide relative to its anchor positions for the patient’s specific HLA alleles. While a subset of peptide positions are presented to the T-cell receptor for recognition, others are responsible for anchoring to the MHC, making these positional considerations critical for predicting T-cell responses. However, a systematic method for determining anchor locations for the wide range of HLA alleles present in the population and application of these to evaluate MT/WT peptide pairs arising in tumors has not been reported. As a result, many neoantigen studies have either failed to adequately consider this crucial factor or have used conventional assumptions to guide their neoantigen identification process. Here, we provide a computational workflow for predicting anchor locations for a wide range of HLA alleles, using a reference dataset generated from clinical and The Cancer Genome Atlas (TCGA) patient samples. We calculated high probability anchor positions for different peptide lengths for over 300 common HLA alleles. Analysis of these results showed clusters of different anchor trends among the HLA alleles analyzed. A subset of these HLA anchor results were orthogonally validated using protein crystallography structures. Analysis of 923 tumor samples showed that 7-41% of neoantigen candidates were potentially misclassified in the neoantigen selection process and can be rescued using allele-specific knowledge of anchor positions. These anchor predictions are currently undergoing experimental validation using both peptide-MHC stability assays as well as fluorescence-based competition binding assays. By incorporating our anchor prediction results into neoantigen prediction pipelines, such as pVACtools, we hope to formalize and streamline the identification process for relevant clinical studies.

Citation Format: Huiming Xia, Joshua McMichael, Michelle Becker-Hapak, Onyinyechi C. Onyeador, Rico Buchli, Ethan McClain, Patrick Pence, Suangson Supabphol, Megan M. Richters, Anamika Basu, Cody A. Ramirez, Cristina Puig-Saus, Kelsy C. Cotto, Jasreet Hundal, Susanna Kiwala, S. Peter Goedegebuure, Tanner M. Johanns, Gavin P. Dunn, Antoni Ribas, Christopher A. Miller, William Gillanders, Todd A. Fehniger, Obi L. Griffith, Malachi Griffith. Computational prediction of MHC anchor locations guide neoantigen prediction and prioritization [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 5639.

Abstract 1215: pVACsplice: Predicting neoantigens from tumor-specific alternative splicing events derived from cis-acting regulatory mutations using whole exome and RNA sequencing data

Neoantigens are tumor-specific peptides on the cell surface that can be recognized by the adaptive immune system. Personalized immunotherapies, such as cancer vaccines, rely on neoantigen prediction to identify sequences that can activate T cells to recognize and destroy the tumor. The majority of cancer vaccine trials have utilized neoantigens derived from missense mutations and small insertions and deletions. However, other mutation types could contribute to the overall neoantigen landscape, such as aberrantly spliced transcripts arising from cis-acting regulatory mutations. In this study, we explore the potential immunogenicity of alternative splicing events by creating pVACsplice, a tool to expand the capability of pVACtools, a suite of tools for neoantigen prediction (http://www.pvactools.org). pVACsplice assembles alternative transcripts from tumor-specific splicing patterns, identifies sequence changes by comparison to a reference, and predicts neoantigens from the novel regions. To verify the accuracy of alternative transcript assembly, we ran pVACsplice with HCC1395 cell line samples and performed long-read sequencing to detect the transcripts in vitro. Matched whole exome sequencing and RNA sequencing datasets from glioblastoma, melanoma, and colorectal cancer cohorts will also be analyzed with pVACsplice to obtain binding affinity estimates. We will compare these results to neoantigen predictions from other mutation sources and across cancer types to discover the prevalence of immunogenic splicing events. Finally, we will perform immunogenicity testing with a set of high quality candidates to validate our predictions. We hope to increase the number of candidates for patients’ vaccines by adding this functionality to our standard neoantigen prediction workflow. This tool could help generate a more accurate portrait of the neoantigen landscape in tumors, and in turn, enhance responses to personalized immunotherapies.

Citation Format: Megan M. Richters, Kelsy C. Cotto, Susanna Kiwala, Huiming Xia, Beatriz M. Carreno, Gavin P. Dunn, Antoni Ribas, Obi L. Griffith, Malachi Griffith. pVACsplice: Predicting neoantigens from tumor-specific alternative splicing events derived from cis-acting regulatory mutations using whole exome and RNA sequencing data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 1215.

Characterization of the Genomic and Immunological Diversity of Malignant Brain Tumors Through Multi-Sector Analysis

Despite some success in secondary brain metastases, targeted or immune-based therapies have shown limited efficacy against primary brain malignancies such as glioblastoma (GBM). While the intratumoral heterogeneity of GBM is implicated in treatment resistance, it remains unclear whether this diversity is observed within brain metastases and to what extent cancer-cell intrinsic heterogeneity sculpts the local immune microenvironment. Here, we profiled the immunogenomic state of 93 spatially distinct regions from 30 malignant brain tumors through whole exome, RNA, and TCR-sequencing. Our analyses identified differences between primary and secondary malignancies with gliomas displaying more spatial heterogeneity at the genomic and neoantigen level. Additionally, this spatial diversity was recapitulated in the distribution of T cell clones where some gliomas harbored highly expanded but spatially restricted clonotypes. This study defines the immunogenomic landscape across a cohort of malignant brain tumors and contains implications for the design of targeted and immune-based therapies against intracranial malignancies.

Splicing factor SF3B1 promotes endometrial cancer progression via regulating KSR2 RNA maturation

Although endometrial cancer is the most common cancer of the female reproductive tract, we have little understanding of what controls endometrial cancer beyond the transcriptional effects of steroid hormones such as estrogen. As a result, we have limited therapeutic options for the ~62,000 women diagnosed with endometrial cancer each year in the United States. Here, in an attempt to identify new prognostic and therapeutic targets, we focused on a new area for this cancer—alternative mRNA splicing—and investigated whether splicing factor, SF3B1, plays an important role in endometrial cancer pathogenesis. Using a tissue microarray, we found that human endometrial tumors expressed more SF3B1 protein than non-cancerous tissues. Furthermore, SF3B1 knockdown reduced in vitro proliferation, migration, and invasion of the endometrial cancer cell lines Ishikawa and AN3CA. Similarly, the SF3B1 inhibitor, Pladienolide-B (PLAD-B), reduced the Ishikawa and AN3CA cell proliferation and invasion in vitro. Moreover, PLAD-B reduced tumor growth in an orthotopic endometrial cancer mouse model. Using RNA-Seq approach, we identified ~2000 differentially expressed genes (DEGs) with SF3B1 knockdown in endometrial cancer cells. Additionally, alternative splicing (AS) events analysis revealed that SF3B1 depletion led to alteration in multiple categories of AS events including alternative exon skipping (ES), transcript start site usage (TSS), and transcript termination site (TTS) usage. Subsequently, bioinformatics analysis showed KSR2 as a potential candidate for SF3B1-mediated functions in endometrial cancer. Specifically, loss of SF3B1 led to decrease in KSR2 expression, owing to reduced maturation of KSR2 pre-mRNA to a mature RNA. Importantly, we found rescuing the KSR2 expression with SF3B1 knockdown partially restored the cell growth of endometrial cancer cells. Taken together, our data suggest that SF3B1 plays a crucial oncogenic role in the tumorigenesis of endometrial cancer and hence may support the development of SF3B1 inhibitors to treat this disease.

Abstract 4413: Accurate neoantigen prediction depends on mutation position relative to patient-specific MHC anchor location

Neoantigens (neoAgs) are short peptide sequences resulting from somatic mutations specifically found in tumor populations. They can be loaded onto major histocompatibility complex (MHC) class I or II molecules to allow recognition by cytotoxic T cells. Accurate neoAg prediction is critical for the design of personalized vaccines and may improve prediction of response to immune checkpoint blockade therapy. The effectiveness of a neoAg-based vaccine relies in part on whether the sequence presented to T cells has previously been exposed to the immune system. Incorrectly selecting for wildtype (WT) peptides will result in susceptibility to central tolerance and potentially induce auto-immunity. As the vast majority of somatic mutations found are single nucleotide variants, changes between the WT and mutant (MT) peptide are subtle and must be interpreted cautiously. An important yet currently overlooked factor in neoAg predicting pipelines is the position of the mutation within the peptide relative to its anchor positions for the patient's human leukocyte antigen (HLA) alleles. Current pipelines consider simple filtering strategies (e.g. MT peptide IC50 < 500 nM and WT/MT binding affinity fold change (agretopicity) > 1); however, only a subset of positions on the loaded peptide sequence are presented to the T cell receptor for recognition, while other positions are responsible for anchoring to the MHC, making these positional considerations critical for predicting T cell responses. We have collected peptide data, through clinical collaborations and The Cancer Genome Atlas (TCGA), for over 200 commonly observed HLA alleles and computationally predicted high probability anchor positions with respect to different peptide lengths (8-11mers). We observed unique anchoring patterns among different HLA alleles, varying in sequence positions and number of anchoring locations within the binding groove. To demonstrate the importance of positional information on prioritization of neoAgs, we are additionally analyzing 1000 TCGA patient samples where potential neoAgs are filtered according to different criteria including: A) MT IC50 < 500 nM, B) MT IC50 < 500 nM and agretopicity > 1, C) MT IC50 < 500 nM with data-driven filtering based on peptide mutation position, MHC anchor position and WT IC50 value. The number of potentially misclassified neoAg candidates will be accessed using experimentally validated neoAg datasets. Currently, prediction pipelines have relatively low accuracy when prioritizing neoAgs for T cell response in patients. By accounting for additional positional information, we hope to significantly reduce the number of false positive neoAgs and increase prediction accuracy. Our anchor results will be implemented as a visualization guide for tumor immunogenomics boards for clinical trials and incorporated into our existing neoAg prioritization pipeline, pVACtools.

Citation Format: Huiming Xia, Megan M. Richters, Cody A. Ramirez, Cristina Puig-Saus, Kelsy C. Cotto, Gavin P. Dunn, Todd Fehniger, Antoni Ribas, William E. Gillanders, Obi L. Griffith, Malachi Griffith. Accurate neoantigen prediction depends on mutation position relative to patient-specific MHC anchor location [abstract]. In: Proceedings of the Annual Meeting of the American Association for Cancer Research 2020; 2020 Apr 27-28 and Jun 22-24. Philadelphia (PA): AACR; Cancer Res 2020;80(16 Suppl):Abstract nr 4413.

Best practices for bioinformatic characterization of neoantigens for clinical utility

Neoantigens are newly formed peptides created from somatic mutations that are capable of inducing tumor-specific T cell recognition. Recently, researchers and clinicians have leveraged next generation sequencing technologies to identify neoantigens and to create personalized immunotherapies for cancer treatment. To create a personalized cancer vaccine, neoantigens must be computationally predicted from matched tumor-normal sequencing data, and then ranked according to their predicted capability in stimulating a T cell response. This candidate neoantigen prediction process involves multiple steps, including somatic mutation identification, HLA typing, peptide processing, and peptide-MHC binding prediction. The general workflow has been utilized for many preclinical and clinical trials, but there is no current consensus approach and few established best practices. In this article, we review recent discoveries, summarize the available computational tools, and provide analysis considerations for each step, including neoantigen prediction, prioritization, delivery, and validation methods. In addition to reviewing the current state of neoantigen analysis, we provide practical guidance, specific recommendations, and extensive discussion of critical concepts and points of confusion in the practice of neoantigen characterization for clinical use. Finally, we outline necessary areas of development, including the need to improve HLA class II typing accuracy, to expand software support for diverse neoantigen sources, and to incorporate clinical response data to improve neoantigen prediction algorithms. The ultimate goal of neoantigen characterization workflows is to create personalized vaccines that improve patient outcomes in diverse cancer types.

Reactive attachment disorder of infancy or early childhood

Since its introduction into DSM-III, reactive attachment disorder has stood curiously apart from other diagnoses for two reasons; it remains the only diagnosis designed for infants, and it requires the presence of a specific etiology. This paper describes the pattern of disturbances demonstrated by some children who meet DSM-III-R criteria for reactive attachment disorder. Three suggestions are made: (1) the sensitivity and specificity of the diagnostic concept may be enhanced by including criteria detailing the developmental problems exhibited by these children; (2) the etiological requirement should be discarded given the difficulties inherent in obtaining complete histories for these children, as well as its inconsistency with ICD-10; and (3) the diagnosis arguably is not a disorder of attachment but rather a syndrome of atypical development.