Erin Newburn, MS, PhD
Sr. Director, Field Application Scientist

Leveraging RNA Sequencing Data in Predicting Immunotherapy Response

Can transcriptomics help us fine tune tumor specific responses and transform personalization of therapy?

The complexity of understanding immune checkpoint blockade response has been abundantly documented in recent studies. Also increasingly studied is the newest biomarker in clinical practice: Tumor Mutational Burden (TMB).  The basic genomic rationale for TMB is that if a tumor possesses a greater number of mutations or “non-self” elements, it would appear to be “more foreign” to our immune system during immunosurveillance, thus eliciting a response. Despite recent FDA approval of pembrolizumab for metastatic tumors with high TMB, and the valiant efforts of the scientific community to harmonize TMB reporting (Personalis January 2021 blog, Tumor Mutational Burden: A continually evolving biomarker for immunotherapy), there are still uncertainties and unexplored elements. The relationship between mutational load and patient immunotherapy response is proving to be an imperfect one (McGrail et al. 2021).  Various intrinsic factors may be at play, such as the need for cancer indication- (or subtype-) specific thresholds for the metric (Hilke et al. 2020). Further, other influences such as the patient’s prior treatment history may also be important as reports have shown that TMB associates with better survival in patients naïve to immunotherapy (Riviere et al. 2020).

Beyond TMB, it seems that there are further vital facets in how T cells recognize the foreignness of tumor cells.  Among these would be the influence of a patient’s human leukocyte antigen (HLA) genetic variation. The diversity of the major histocompatibility complex (MHC) (Chowell et al. 2018) as well as catastrophic genomic events such as HLA loss of heterozygosity (McGranahan et al. 2017; Montesion et al. 2021) may be driving forces in dictating treatment efficacy. Considering all these different factors, it seems impossible that a pan-cancer IO biomarker can be achieved. In particular, meta-analyses have pointed to IO predictive biomarkers as being highly tumor type and context dependent (Litchfield et al. 2021). Due to the inherent complexities of the interplay between the tumor and the immune system, a unified model of patient response seems daunting.  Yet, there are many emerging biomarkers under investigation, including gene expression signatures, that may hold promise in helping to better stratifying responses and give a deeper understanding of the multiple suppressive pathways in the tumor.

In recent years, gene expression signatures have been invaluably leveraged in immunotherapy translational research. Analyzing gene expression profile data has provided a better understanding of the differences between inflamed and non-inflamed Tumor Microenvironments (TME). Efforts to determe genes that identify activated T cells, chemokine expression and other components, such interferon-responsive genes, have improved predictions for immunotherapy response in various indications such as melanoma, gastric, and head and neck squamous cell carcinomas (Ott et al. 2018).  Further, large publicly available datasets like The Cancer Genome Atlas (TCGA) have been invaluably leveraged in immunotherapy translational research (Danaher et al. 2018).  Likewise, large cohorts from clinical trial studies are providing fresh insights into the dynamics of response to immunotherapies beyond the established biomarkers like TMB and PD-L1 expression. As an example, in a recent publication from the JAVELIN Renal 101 trial, one of the largest renal cell carcinoma cohort of its kind, the authors found significantly prolonged progression-free survival (PFS) with first-line avelumab + axitinib versus a multi-target tyrosine kinase inhibitor (TKI) (Motzer et al. 2020). In this cohort, they found that neither (PD-L1 expression) nor tumor mutational burden was associated with PFS. However, they developed a unique expression signature using the Personalis’ NeXT TranscriptomeTM that was significantly associated with PFS.  In this signature, a set of 26 genes was associated with immune-related functionality and was validated in an independent cohort. In another study, Thompson et al. 2020 developed an antigen presentation machinery (APM) score consisting of key APM genes including B2M, CALR, PSME1 and others. Data demonstrated that a higher APM score in both lung cancer and melanoma was able to predict outcomes to immune checkpoint blockade (ICB) therapy. Supporting the strength of this biomarker and pathway, the APM score was actually more predictive of patient response than looking at an inflammatory gene signature.

Personalis ImmunoID NeXTTM RNASeq (NeXT TranscriptomeTM):

ACE TechnologyTM Improves RNA Performance Critical for Biomarker Discovery

In most oncology translational studies, the majority of samples available for analysis are formalin fixed paraffin embedded (FFPE) materials. Historically, these specimens have been quite challenging for RNA-based next generation sequencing.  Due to the chemical modification process, FFPE materials often result in degraded materials and shortened nucleic acid fragment sizes.  Enrichment methods such as poly-A selection require intact, high quality RNA molecules.  Likewise, approaches such as ribodepletion may result in inconsistent rRNA removal, thereby requiring greater sequencing depths.

Figure 1. NeXT Transcriptome: Optimized approach for FFPE materials

Comprehensive characterization of tumor gene expression is an important overlay for interpreting somatic mutations, identifying neoantigens, assessing expression of checkpoint genes and immune signatures, and identifying prognostic expression signatures. Applying ACE TechnologyTM to an exome-capture transcriptome protocol produces high quality transcriptome sequencing results from FFPE samples. This enrichment protocol directly selects for transcripts using complementary capture probes, eliminating background and focusing the sequencing on regions of interest. Since multiple probes target each transcript, this assay captures transcripts even when the poly-A tail is lost due to RNA degradation making it ideal for cancer FFPE samples. Additionally, the NeXT Transcriptome can work with a lower input of material.

Another common challenge in oncology biomarker discovery is analyzing the large volume of data generated from high content approaches such as transcriptome solutions. At Personalis, we have also created various analytical modules to enable the generation of informative immuno-oncology insights from the comprehensive RNA data generated by ImmunoID NeXTTM. In particular, we have developed RNA bioinformatic pipelines that can be mined for various outputs (Table 1).

Table 1. NeXT Transcriptome Bioinformatic Output
 • Raw data files – FASTQ, BAM files
 • Variant calls – VCF
 • Variant annotations report
 • Somatic variant (SNV, indel) analysis
 • Fusion gene analysis
 • Expression analysis (gene, transcript)
 • QC Summary

As part of the ImmunoID NeXT platform, the NeXT Transcriptome approach is a robust method for analyzing gene expression in FFPE tissues for advanced biomarker discovery. Questions on how this approach can be applied to your translational research efforts? Please contact us at [email protected]