Primary Tumors Can Tell Us More If We Ask Them To

As a medical oncologist, I may rely a little less than my radiation or urology colleagues on the Gleason score for prognosis and treatment decision making. Most of our decisions are based on the pace of disease and extent, and of course whether it is castration-resistant or castration sensitive. However,  I do look at it and in particular, it factors into the data ‘stew’ that one creates within an individual case and how we approach it.

I have written previous blogs about cases that diverged from what we thought might happen. On one occasion I asked, "Why is my patient dying if he only had a Gleason 6?”, and in another, I expressed soaring optimism and was “Flying High on MSI”, in which I described a case of a patient with Gleason 9 disease who experienced a complete response to pembrolizumab after developing symptomatic metastatic castration-resistant prostate cancer (mCRPC). It is now over two years since starting that therapy and the patient still has a prostate-specific antigen (PSA) of zero, and is off all therapy (fingers crossed). So, in a couple of instances here, at least, the Gleason score failed to reveal something underneath our typical expectation. As we all know, a deeper dive into tumor biology is the key.

We have a few biological tools now that Professor Gleason didn’t have (it is remarkable that he actually charted the course of over 50 years of decision making with just two tools – a microscope and a Sharpie®!), most notably next-generation sequencing where we can look at genes as single entities (e.g. TP53 altered or not, BRCA2 mutated or not), as well as multi-gene panels such as those from Prolaris® or Decipher®.

One recent analysis has allowed me to gain some traction in my thinking. Dan Spratt and colleagues did the largest ever full transcriptomic analysis of prostate cancer focusing on the primary tumors of over 20,000 patients from a number of datasets1. The beauty of this analysis, in addition to its sample size, is the manner in which they proposed "branched thinking" (my term) and divided the tumor initially by their intrinsic androgen receptor (AR) signaling capacity using “AR-A low” for low AR activity and “AR-high” for its binary counterpart. From there they explored the difference in genomic biology comparing these two groups.

It is an important observation that androgen receptor activity is a broad spectrum even in primary tumors. This paper is an excellent example of a topic that has been of interest to me since my fellowship where a wrote a paper about a small number study of the role of AR signaling in primary tumors after neoadjuvant androgen deprivation therapy (ADT)2. That is to say, even in a primary tumor there are tumor components that are already ‘androgen-independent’ from the perspective that they are able to grow and proliferate despite low signaling of the AR. This is different biology from castration-resistant prostate cancer (CRPC), where the AR is amplified and active.

The real learning here comes when we focus on the low AR-A group, and understanding what biological pathways are active in such tumors. It gives us insight not only into the pathways that may underlie eventual castration resistance, but when viewed opportunistically this analysis can open up some potential treatment opportunities.

Spratt and colleagues identified five pathways that diverge based on AR-A low versus AR-A high: Neuroendocrine Prostate Cancer, Double Strand DNA repair, Mismatch Repair, Suppressor Immune Cells, and Effector Immune Cells. For the sake of brevity, I will focus on only a couple of these to demonstrate the point and the clinical potential. I strongly suggest reading the full paper to get a full appreciation.

First, take neuroendocrine prostate cancer (NEPC). When AR signaling was divided by decile it is revealed that NCAM1, a gene associated with NEPC, was very highly expressed in the lowest AR-A decile, while its expression was lowest in the highest AR-A decile, with a proportional (and statistically very significant) stepwise reduction in expression as the deciles go from low to high. Pretty straightforward. If we had an ideal treatment for NEPC, we could easily test its implementation across this spectrum.

Next, consider double-strand break repair (DSBR). The opposite pattern is shown here – when AR-A is low, DSBR is low, increasing stepwise by decile to a very high level in the high AR-A samples. This is a very telling observation and has been made before. We have not yet fully exploited this observation yet in the clinic. This observation is telling us that when the AR is less active that the machinery for DNA repair is reduced substantially and conversely supports the observation that AR signaling supports DNA repair. 

The observation may support why we have seen study after study that combines ADT with radiation as showing additive effects to these modalities, and why, for example, a few months of neoadjuvant ADT prior to radiation is superior to starting ADT immediately with radiation: On ADT we impair the tumor’s ability to repair the DNA damage produced by radiation.

Next, it would suggest that maybe we can exploit other therapies that induced DNA damage in the context of low AR signaling. One approach would be to identify AR-A low tumors and preferentially treat them with platinum chemotherapy – either in the CRPC or castration sensitive prostate cancer (CSPC) setting. We are indeed treating patients with newly diagnosed high volume castration sensitive prostate cancer on a clinical trial of cabazitaxel plus carboplatin and will, down the road, report outcome as a function of baseline tumor AR-A signaling, but unfortunately, those results are a couple of years away. The other setting in which this observation may be helpful is going to be in the setting of treatment with poly (ADP-ribose) polymerase (PARP) inhibitors. We have seen a signal from studies of olaparib plus abiraterone in the United Kingdom that the combination of these two drugs may be superior to abiraterone alone even in the absence of DNA repair mutations such as BRCA2. The Spratt data may help explain that – if you suppress AR and suppress DNA repair, suppress PARP and induce a ‘second hit’ on DNA repair, which is lethal to the tumor, but in a different way than when you induce DNA damage with platinum or radiation.

The missing variable in the setting of wildtype BRCA2 etc. is that we don’t really know what is happening to PARP. In the thought experiment that these observations allow, therapeutic suppression of the AR through abiraterone, enzalutamide or initial ADT, DNA repair is suppressed (not unlike it would be in a mutant BRCA2 carrier) and PARP is induced. Subsequent PARP suppression induces synthetic lethality. A clinical trial that demonstrates how and when we induce PARP would be interesting, and potentially very valuable for patients.

Bit by bit, the story is coming together. These tumors are telling us what to do, we just need to figure out how to listen.

Written by: Charles Ryan, MD, B.J. Kennedy Chair in Clinical Medical Oncology, Director and Professor of Medicine, Division of Hematology, Oncology and Transplantation, University of Minnesota, Minneapolis, Minnesota

Published Date: January 6th, 2020


1. Spratt, Daniel E., Mohammed Alshalalfa, Nick Fishbane, Adam B. Weiner, Rohit Mehra, Brandon A. Mahal, Jonathan Lehrer et al. "Transcriptomic Heterogeneity of Androgen Receptor Activity Defines a de novo low AR-Active Subclass in Treatment Naïve Primary Prostate Cancer." Clinical Cancer Research 25, no. 22 (2019): 6721-6730.

2. Ryan, Charles J., Alex Smith, Priti Lal, Jaya Satagopan, Victor Reuter, Peter Scardino, William Gerald, and Howard I. Scher. "Persistent prostate-specific antigen expression after neoadjuvant androgen depletion: an early predictor of relapse or incomplete androgen suppression." Urology 68, no. 4 (2006): 834-839.