Skip to main content

Blood-Based Genetic Risk Score Shows Promise for Prostate Cancer Drug Response, Survival Prediction

Premium

NEW YORK – A group of researchers has developed a multi-gene risk score based on a group of copy number alteration (CNA) mutations in circulating tumor DNA (ctDNA) that it believes can predict metastatic castration-resistant prostate cancer (mCRPC) treatment response and survival outcomes.

The team envisions developing a non-invasive liquid biopsy assay to guide targeted treatment for mCRPC patients.  

Clinicians often treat mCRPC using a combination of abiraterone acetate and prednisone (AA/P) therapy or other treatments. However, there is a lack of molecular biomarkers that predict primary and secondary resistance to prostate cancer therapies like AA/P to help clinicians decide whether to continue treatment or try something different.

The researchers therefore sought to prospectively evaluate whether CNAs could serve as predictive and prognostic biomarkers for AA/P treatment response in mCRPC patients.

"We were interested in finding out [candidate genes] that could serve as part of a potential multi-gene score," Manish Kohli, co-senior author and professor at the Huntsman Cancer Institute, explained. "We also wondered how we could capture the heterogeneity of the drug-tumor landscape interactions to predict treatment outcomes."

In the study, published last month in Prostate Cancer and Prostatic Diseases, Kohli and his colleagues collected plasma samples from 88 chemotherapy-naïve mCRPC patients at an initial visit (prior to initiating AA/P treatment) and at a follow-up visit 12 weeks later. The group followed patients until progression or until death to identify ctDNA-based somatic CNAs predictive of acquired AA/P resistance and prognostic of mCRPC status.   

After preparing DNA libraries from extracted ctDNA, the researchers performed whole-genome sequencing on the specimens using Illumina's HiSeq2500 system.

Kohli and his colleagues then performed CNA analysis on all samples before treatment and after 12 weeks, and evaluated 12-week responses for primary resistance, time to treatment change for secondary resistance, and overall survival for prognosis. In addition, the researchers performed CellSearch circulating tumor cell counts, and radiological and prostate-specific antigen measurements on the enrolled patients who were still alive after 12 weeks of treatment.

To identify CNAs associated with primary AA/P resistance,Kohli and his team labeled the 48 patients that lacked PSA progression, had an absence of new bone lesions in a bone scan, and showed no evidence of radiological progression of metastases after 12 weeks as responders. However, the 36 patients who did not meet the criteria were labeled as non-responders with primary resistance.

While ctDNA amounts in pretreatment and posttreatment samples dropped by about 4 percent in the responder cohort, they increased about 1 percent in the non-responder population. In 82 paired plasma samples, 21 showed a significant drop (3 percent) in ctDNA content, while 10 showed an increased ctDNA count (greater than 3 percent).

The study authors also noted that nine of the 10 (90 percent) of patients with increased ctDNA content were nonresponders, whereas only 38 percent of patients with reduced ctDNA contents were nonresponders.

In total, Kohli and his team tested 31 prostate cancer-related genes to evaluate candidate gene-specific CNAs and found that multiple genes, including AR, OPHN1, PIK3CA, and ZFHX3, are associated with primary resistance.

To identify gene-specific CNAs predictive of acquired resistance, the group then applied regression analysis and found significant associations with time to treatment change (TTTC) in selected gene loci.  Median TTTC was about 9 months with AR amplification, compared to a median TTTC of more than 42 months without AR amplification.

To tackle genetic heterogeneity, Kohi's team designed multivariate risk models using multiple somatic CNAs and clinical factors linked to acquired resistance. Applying stepwise regression to 11 genes that showed associations with TTTC, the analysis generated a CNA-based risk score — containing AR, NKX3.1 and PIK3CA — significantly associated with TTTC and that was independent of adjustments for CTC counts, ctDNA content, and clinical factors such as age and baseline PSA level.

The group also evaluated the CNAs' prognostic value using the 31 selected genes. Applying cox regression analysis, the team identified a significant association with overall survival in 11 of 31 selected gene loci. The team then chose four genes — ZFHX3, RB1, PIK3CA, and OPHN1 — to create a prognostic risk score, which remained significant after adjusting for CTC, ctDNA content, and clinical factors.

The team also found that AR and AR enhancer amplification CNAs were linked to both primary resistance and shorter TTTC. Specifically, ZFHX3 deletion and PIK3CA amplification were linked to primary resistance, shorter TTTC, and poor overall survival.

Kohli's team therefore developed a CNA-based risk score integrating associations with TTTC, which were predictive of secondary resistance and established prognoses for survival based on CNAs in ZFHX3, RB1, PIK3CA, and OPHN1. The group also found that the multigenic risk scores were more predictive than individual genes or clinical risk factors.

"We identified multiple novel loci … as candidate biomarkers for response to AA/P, [as well as] multigene risk scores for predicting resistance to AA/P treatment, disease progression, and survival," the study authors concluded. "We believe that using plasma ctDNA CNAs and risk scores can potentially help clinicians predict mCRPC treatment and survival outcomes in patients." 

Liang Wang, study coauthor and a researcher in the tumor biology department at the Moffit Cancer Research Center, highlighted that the team was surprised to find that OPHN1 was indicative of shorter disease progression and survival.

Kohil noted that his team might have performed further analysis on the OPHN1 gene if they had access to a larger mCRPC cohort, highlighting that the gene had been amplified in about 56 percent of the mCRPC patients.

"In prostate cancer, OPHN1 is upregulated and is associated with cell adhesion and migration," the study authors noted. "However, it is unclear whether amplification of either the AR enhancer or OPHN1 are functional consequences or coincidental bystander events related to AR amplification."

Wang emphasized that the team will therefore need to collect additional evidence to demonstrate that the biomarker can act as an improved predictor of treatment and overall survival. At the same time, he believes that "it might be as equivalent as the AR gene to help predict clinical outcome."

Clinical potential

Wang and Kohli have now launched a larger trial to validate the identified biomarkers to potentially guide clinical practice in advanced prostate cancer. Kohli estimates that the multi-year prospective study will collect around 400 and 500 patients between the Huntsman Cancer Institute and the Moffitt Cancer Research Institute.

Wang added that the team will also need to validate the prostate cancer risk score and narrow it down to a smaller panel of gene loci. The team aims to capture the tumor's intratumoral heterogeneity as a classifier and develop multiple risk scores for individual drugs that can help clinicians treat patients with the most appropriate therapy. 

"While [none of these alterations] may be genomic drivers of tumor biology and in fact some may be bystander events, taken together they contribute to a genetic heterogeneity," Kohli said. "We attemted to use the increasing number of observed alterations ... to derive our multi-gene scores recognizing that a signle gene alone may not be responsible for clinical outcomes." 

If the multiple-gene scores are validated in the future for a specific drug, Kohli believes that clinicians could eventually offer a non-invasive blood-based assay that helps direct treatment response in an outpatient clinical setting. However, he noted that commercializing the risk score further will depend on the results of the larger study.

"Basically, when patients come to see [their doctor] in the clinic, we could offer a [specific] drug among a plethora of drugs," Kohli said. "[A clinician] could draw a blood sample, send it to our lab, which would look at the number of candidate genes to develop that score, find a particular score with certain genes, and provide a readout that a particular drug is effective for that patient."

However, Kohli emphasized that the study is only one of the first steps to such a workflow and that his team will require several validation steps to evaluate the genes for different cancer drugs.

Wang noted that for clinical purposes, the team may likely switch from WGS to a PCR-based approach — such as digital PCR — to analyze targeted candidate hotspot locations, rather than wasting time and costs on sequencing the entire genome.

While the researchers analyzed ctDNA in the study, Kohli said they will still examine additional molecules of interest in mCRPC patients' blood that could translate to patient care. Working with Wang's lab at the Moffit Cancer Center, Kohli has begun to analyze exosomal DNA for potential treatment and prognostic indicators.

In an upcoming European Urology study, the team demonstrated that several CTC-based ARV7-positive may also benefit from use of the mCRPC risk score. In addition, the team found that other mCRPC ARV7-negative patients, who were AR-amplified on plasma ctDNA, also had an equally poor prognosis and may benefit from a change in treatment.