NEW YORK – A study published Monday in Nature Biotechnology demonstrates how an estimation of tumor-specific total mRNA levels from patient tumor samples can be used to predict patient survival.
Total mRNA levels have been linked to tumor phenotype previously in single-cell RNA sequencing studies. Theoretically, total mRNA expression has potential as a measurable pan-cancer biomarker. However, the labor- and computationally intensive nature of single-cell sequencing has been a hindrance in translating that theory into a tool for guiding cancer treatment.
Now, a team from MD Anderson Cancer Center led by Wenyi Wang, a professor of bioinformatics and computational biology, has developed a way to estimate total mRNA in tumor cells by mathematically deconvoluting bulk DNA sequencing. And they're developing that metric into a biomarker test that can be used in clinical studies or, eventually, patient care, including an upcoming analysis of data from MD Anderson's Dynamo clinical trial in castration resistant prostate cancer.
Total mRNA data is captured during bulk sequencing, but standard analysis procedures mask it. Wang and her team started with single-cell RNA sequencing data from 48,913 cells collected from 10 patients in four cancer types — colorectal, liver, lung, and pancreatic — with documented survival outcomes. Using a surrogate measure for total mRNA, they found that cells with higher total mRNA counts were less differentiated, exhibiting a "stem-like" cell state, and were enriched in genes associated with stemness and epithelial-mesenchymal transition.
The MD Anderson group developed a mathematical method for deconvoluting bulk DNA and RNA sequencing data to yield a quantitative metric, TmS, representing the per-cell, per-haploid genome total RNA expression for a tumor.
With the single-cell analysis in hand, they pooled the data to create a "pseudo-bulk" sequencing analysis, and calculated a ratio of mRNA levels in tumor cells versus non-tumor cells for each sample. That bulk-level analysis revealed increased tumor mRNA content in the four patients with more advanced disease and worse survival outcomes compared to the other patients in the group.
Taking that metric to a data set of 6,590 tumor samples from four large patient cohorts linked with long-term outcomes data, they found that higher tumor-specific mRNA levels correlated with reduced progression-free survival and overall survival. "Across all 15 cancer types, we did find consistent signals that high levels of total mRNA expression in tumor cells correspond to a worse prognosis," Wang said.
In addition to making a connection between TmS and prognosis, Wang's group also probed the biological underpinnings of that relationship in different types of cancer and at the genetic level. The researchers examined data sets from The Cancer Genome Atlas, the early-onset prostate cancer cohort from the International Cancer Genome Consortium, the "Molecular taxonomy of breast cancer international consortium" (METABRIC) study, and the TRACERx study, finding that certain genes were enriched across cancer types, including housekeeping genes, essential genes, cancer hallmark genes, and transcriptional regulation pathway genes. A further analysis of biologic correlates found upregulation of the pentose phosphate pathway and glucose metabolism.
Wang and her colleagues observed an unexpected inversion of the relationship between TmS and outcome in 4 of the 12 cancer types they investigated based on the stage of the disease, including breast cancer subtypes. In those cancers, early-stage patients with high TmS levels had more favorable outcomes, possibly because the prognostic effect of TmS was modified by treatment.
Taking a closer look at breast cancer patients in METABRIC, they found that high TmS was associated with increased disease-free survival in patients with ER-positive, HER2-negative breast cancer, even after adjusting for chemotherapy and Oncotype Dx risk status. Conversely, patients with low TmS seemed not to have benefited from chemotherapy. Those results suggest a role for TmS in identifying and stratifying high-risk patients for breast cancer treatments, and the pattern may extend to other cancer types, as well.
Wang said this inversion effect has been documented in the literature with other biomarkers, where the marker can predict a worse prognosis, but at the same time better response to treatment. "This is the trickiest, but most interesting and real part of the biology that we found," said Wang.
Translating TmS to clinical practice as a test for cancer patients presents one potential challenge in that measurement of TmS would require matched whole-exome sequencing, whole-genome sequencing, or SNP array analysis plus RNA sequencing, a more extensive genomic analysis than testing that is currently routinely carried out for cancer patients.
"The other ongoing project is obviously that we want to identify signatures and molecular pathways that underlie TmS so that can be more easily measurable clinically," Wang said.
Next steps for Wang's group include working closely with clinicians at MD Anderson to deconvolute other molecular markers from bulk sequencing data, as well as studies further developing TmS as a biomarker. One such study will be funded by a Prostate Cancer Research Program Data Science award that Wang received from the US Department of Defense to investigate whether TmS can serve as a biomarker to predict how patients with aggressive variant prostate cancer (AVPC) may respond to various treatments.
"Aggressive variant prostate cancer is a subgroup of prostate cancer where these patients have much worse, much shorter survival outcomes than usual," said Wang. "And the challenge in the clinic is how to treat them and what are the molecular mechanisms that can be blocked to save these patients' lives."
Wang will be working on isolating TmS profiles collected in the Dynamo clinical trial. That study tested various drug combinations in men with castration-resistant prostate cancer with one of its primary outcome measures being identification of androgen response marker signatures. Results were presented at the 2022 American Society of Clinical Oncology annual meeting earlier this month.
Wang said prostate cancer is an example of an indolent cancer type amenable to investigations of whether TmS can help identify patients who are being overtreated or those with aggressive cancer subtypes who need alternative treatments.
She added that TmS is just one piece of the larger puzzle of tumor heterogeneity, which makes studying cancer and developing cures very challenging, and on which her lab continues to focus.