Tumour neoantigen repertoire prediction in malignant peripheral nerve sheath tumours define private and public targets for immunotherapy

This study characterizes the neoantigen landscape of malignant peripheral nerve sheath tumors (MPNSTs), revealing that while PRC2-loss and PRC2-wildtype subtypes share private neoantigens, PRC2-loss tumors exhibit immune resistance and reduced antigen presentation, thereby necessitating a dual therapeutic strategy of personalized neoantigen vaccines for all patients and cell-surface TAA-directed therapies specifically for PRC2-loss cases.

The Gist

The Big Picture: A Tough Enemy with a Hidden Weakness

Imagine MPNST (Malignant Peripheral Nerve Sheath Tumour) as a very tough, aggressive fortress. It's a type of cancer that is hard to treat with standard drugs, and surgery is often the only way to beat it. Scientists have been trying to teach the body's immune system (the "police force") to recognize and destroy this fortress, but so far, it hasn't worked very well.

This paper is like a digital detective story. The researchers didn't go into a lab to test blood samples; instead, they used powerful computer programs to scan the genetic blueprints (DNA and RNA) of these tumors. Their goal? To find the "wanted posters" (neoantigens) that the immune system can use to identify and attack the cancer.

They discovered that these tumors come in two main "flavors" (types), and each flavor needs a different strategy to defeat.

---

The Two Types of Fortresses: The "Open" vs. The "Bunker"

The researchers found that MPNSTs fall into two categories based on a specific genetic switch called PRC2.

1. The "Open" Fortress (PRC2-WT): These tumors still have the PRC2 switch working. They are a bit more open to the immune system. The immune police can see them, and there are some immune cells hanging around outside the walls.
2. The "Bunker" Fortress (PRC2-Loss): These tumors have broken the PRC2 switch. This is the dangerous type. They are like a bunker that has locked all its doors and windows. They actively hide from the immune system, and the area around them is a "desert" with almost no immune police officers.

The Search for "Wanted Posters" (Neoantigens)

To get the immune system to attack, you need to show it a picture of the criminal. In cancer terms, these pictures are called neoantigens. These are unique "badges" or "scars" on the cancer cells caused by genetic mutations.

What the computer found:

• Every Criminal is Unique: The researchers found that almost every single patient has their own unique set of "wanted posters." There is no "one size fits all" poster for MPNST.
  • Analogy: Imagine trying to catch a group of thieves. You can't just make one "Wanted" poster for "Thief #1" and expect it to catch everyone. You have to make a custom poster for each specific thief.
• The Numbers are Low: The tumors don't have many mutations (scars) compared to other cancers like melanoma. This makes the job harder because there are fewer targets to aim at.
• The "Bunker" Problem: In the "Bunker" type (PRC2-Loss), the tumor doesn't just hide; it also breaks the machinery inside the cell that usually displays these wanted posters. Even if the tumor has a mutation, it can't show it to the immune system. It's like having a wanted poster but keeping it locked in a safe deep inside the bunker.

The Twist: Finding a Back Door (Tumor-Associated Antigens)

Since the "Bunker" tumors are so good at hiding their unique mutations, the researchers asked: "Is there another way to spot them?"

They noticed that in the "Bunker" tumors, a specific part of the genetic blueprint (Chromosome 8) gets copied over and over again. Think of this like a factory that accidentally prints 100 copies of the same instruction manual instead of just one.

Because of this "copy-paste" error, these tumors are forced to produce huge amounts of specific surface proteins (proteins on the outside of the cell).

• The Strategy: Since the immune system can't see the hidden "wanted posters" inside, the researchers suggest using a different weapon: CAR-T cell therapy.
• Analogy: Instead of trying to find the criminal's face (which is hidden), you just look for the giant, bright red hat they are all wearing because the factory made 100 of them. You can train the immune police to ignore the hidden face and just attack anyone wearing that specific red hat.

The Takeaway: Two Different Strategies for Two Different Enemies

The main conclusion of this paper is that we can't use the same treatment for all MPNST patients. We need a personalized approach:

1. For the "Open" Fortresses (PRC2-WT):

   • Strategy: Personalized Vaccines.
   • How it works: Since these tumors have unique mutations, we can make a custom vaccine for each patient that teaches their immune system to recognize their specific "wanted posters." This is like giving the police a custom sketch of the specific criminal they are hunting.

2. For the "Bunker" Fortresses (PRC2-Loss):

   • Strategy: Targeted Cell Therapy (CAR-T).
   • How it works: Since these tumors hide their mutations and have a "desert" around them, vaccines might not work. Instead, we should target the "Red Hats" (the surface proteins caused by the Chromosome 8 copy errors). We can engineer immune cells to hunt down anyone wearing that specific protein, regardless of whether the tumor is trying to hide.

Why This Matters

This study is a roadmap. It tells doctors and scientists:

• Don't just try one drug for everyone.
• Check the patient's tumor type first.
• If it's the "Open" type, try a custom vaccine.
• If it's the "Bunker" type, try targeting the surface proteins.

It's a step toward turning a "one-size-fits-all" approach into a "tailor-made" solution for a very difficult cancer.

Technical Summary

1. Problem Statement

Malignant Peripheral Nerve Sheath Tumours (MPNSTs) are aggressive soft-tissue sarcomas with poor prognosis and limited effective systemic therapies. While immunotherapies like Immune Checkpoint Blockade (ICB) have succeeded in other solid tumors, they have shown limited efficacy in MPNSTs.

• The Gap: There is a lack of comprehensive characterization of the neoantigen landscape in MPNSTs. It is unclear whether the failure of immunotherapy is due to a lack of neoantigens, defects in antigen presentation, or an immune-desert tumor microenvironment (TME).
• The Stratification: MPNSTs are increasingly stratified into two molecular subgroups based on the status of the Polycomb Repressive Complex 2 (PRC2): PRC2-Wild Type (WT) and PRC2-Loss (characterized by mutations in EED or SUZ12 leading to loss of H3K27me3). PRC2-Loss tumors are known to have reduced immune infiltration, but the specific antigenic targets for these subgroups remain undefined.

2. Methodology

The study employed a comprehensive in silico pipeline integrating public genomic and transcriptomic data to predict neoantigens and tumor-associated antigens (TAAs).

• Data Sources: Two independent cohorts were analyzed:
  1. dbGaP Cohort: 16 patients (12 fresh-frozen tumors) with Whole Exome Sequencing (WES) and RNA-seq.
  2. GeM Consortium Cohort: 88 patients with Whole Genome Sequencing (WGS) and RNA-seq.
  • Total: ~104 patients after strict quality control and sample matching (using SMaSH for identity verification).
• Stratification: Tumors were classified as PRC2-WT or PRC2-Loss based on genomic alterations in SUZ12, EED, EZH2, or EZH1.
• Neoantigen Prediction Pipeline (pVACtools):
  • Somatic Variants: Identified using Strelka2, VarScan2, and Mutect2.
  • Fusion Genes: Identified using STAR-Fusion and validated with FusionInspector.
  • HLA Typing: Performed using OptiType (Class I) and HLA-HD (Class II) from RNA-seq data.
  • Binding Prediction: Used pVACseq and pVACfuse to predict peptide-MHC (pMHC) binding affinity using multiple algorithms (NetMHCpan, MHCflurry, etc.).
  • Filtering Criteria: High-confidence neoantigens required:
    • Clonal origin (DNA VAF ≥ 0.20).
    • Expression (RNA depth ≥ 10, RNA VAF ≥ 0.25).
    • Strong binding (IC50 < 500 nM).
    • No match to the reference proteome (to exclude self-antigens).
• TAA Discovery: Focused on Chromosome 8 (recurrently amplified in MPNST). Identified cell-surface proteins with copy number gains (CN ≥ 3) and correlated them with mRNA expression levels.
• Immune Profiling: Used ImSig and CIBERSORTx to deconvolute immune cell abundance and assess antigen-processing machinery (APM) gene expression.

3. Key Contributions

• First Comprehensive In Silico Landscape: Provided the first systematic characterization of both somatic mutation-derived and fusion-derived neoantigens in MPNSTs, stratified by PRC2 status.
• Definition of "Private" vs. "Public" Targets: Demonstrated that MPNST neoantigens are almost exclusively private (patient-specific), with no shared epitopes across cases, even among those with common driver mutations (e.g., TP53, NF1).
• Mechanistic Insight into Immune Resistance: Linked PRC2-Loss status directly to a "cold" tumor microenvironment characterized by the downregulation of antigen processing machinery (specifically MHC Class II and TAP/NLRC5) and reduced immune infiltration.
• Identification of Alternative Targets: Identified a distinct set of MHC-independent targets (Cell Surface TAAs) driven by Chromosome 8 amplification, specifically in PRC2-Loss tumors, offering a rationale for CAR-T or antibody-based therapies.

4. Key Results

A. Neoantigen Landscape (Somatic & Fusion)

• Low Burden: MPNSTs have a low tumor mutational burden (TMB). The median neoantigen load was ~10 per sample (3–4 from somatic mutations, 5–6 from fusions).
• Privacy: All predicted neoantigens were private to individual cases. Even common driver genes like TP53 produced unique peptides in each patient due to different mutation sites.
• Subtype Comparison:
  • PRC2-WT vs. PRC2-Loss: The total number of predicted neoantigens was comparable between subtypes.
  • Variant Types: PRC2-WT tumors had a higher proportion of missense variants, while PRC2-Loss had more frameshifts.
  • Fusion Genes: Fusion-derived neoantigens were also highly private. Only 8 fusion genes were shared between subtypes, and most were unique to specific cases.
• Persistence: A subset of neoantigens arose from "persistent" mutations (clonal, single-copy or multi-copy regions), suggesting they are stable targets for vaccines.

B. Immune Microenvironment & Antigen Presentation

• PRC2-Loss is Immune-Desert: PRC2-Loss tumors showed significantly reduced infiltration of T cells, macrophages, and B cells compared to PRC2-WT.
• Antigen Processing Defects: PRC2-Loss tumors exhibited significant downregulation of:
  • MHC Class II genes (HLA-DPA1, HLA-DPB1, HLA-DMB).
  • Peptide Transporters: TAP1, TAP2.
  • Regulators: NLRC5 (a master regulator of MHC Class I).
• Implication: Even if neoantigens are present, the PRC2-Loss machinery is intrinsically impaired in presenting them to T cells, explaining the failure of ICB monotherapy.

C. Chromosome 8 Amplification & TAAs

• Genomic Driver: Chromosome 8 gain (particularly 8q) is a hallmark of PRC2-Loss MPNSTs.
• Surface Antigen Overexpression: This amplification drives the overexpression of cell-surface proteins.
• Candidate TAAs: The study identified a recurrent set of surface genes overexpressed in PRC2-Loss tumors due to copy number gain, including:
  • FGFR1, SDC2, SLCO5A1, GPR20, LY6E, MMP16, FZD6, OPRK1.
• Therapeutic Relevance: These are MHC-independent targets, making them ideal candidates for CAR-T cell therapies or antibody-drug conjugates (ADCs) in the "immune-cold" PRC2-Loss subgroup.

D. Survival Analysis

• Neoantigen burden (low, medium, high) was not significantly associated with overall survival (OS) in this cohort, likely due to the small sample size and the dominance of the immune-suppressive microenvironment over antigen load.

5. Significance and Clinical Implications

This study provides a rational framework for developing immunotherapies for MPNSTs, moving away from a "one-size-fits-all" approach to a PRC2-status-dependent strategy:

1. For PRC2-WT Tumors:

   • These tumors have better immune infiltration and antigen presentation.
   • Strategy: Personalized Neoantigen Vaccines (mRNA or peptide) targeting private somatic and fusion neoantigens could prime the existing immune system.

2. For PRC2-Loss Tumors:

   • These tumors are "immune-cold" with defective antigen presentation.
   • Strategy: ICB monotherapy is unlikely to work. Instead, therapies should target MHC-independent surface TAAs (e.g., FGFR1, SDC2) identified via Chromosome 8 amplification.
   • Approach: CAR-T cell therapy or bispecific antibodies targeting these overexpressed surface proteins.
   • Combination: Strategies to "heat up" the TME (e.g., STING agonists or viral vectors) may be required to sensitize these tumors to neoantigen-based approaches.

Conclusion: The paper establishes that while MPNSTs harbor private neoantigens suitable for vaccination, the PRC2-Loss subgroup is intrinsically resistant to standard neoantigen recognition due to antigen presentation defects. However, the recurrent Chromosome 8 amplification in this subgroup offers a distinct, actionable target for MHC-independent immunotherapies.