Examining Molecular Subtyping in Prostate Cancer

Adam Weiner, MD

Years of research has shown prostate cancer to originate from either basal or luminal cells. according to Adam Weiner, MD. Though basal and luminal cells may be histologically indistinguishable once they become a prostate tumor, they are still distinct in gene expression.¹

Prostate cancer has been subclassified based on molecular drivers, but utility remains limited due to the reliance on multi-omics and cost. Additionally, while commercial products have been used widely for this, they do not subclassify prostate cancer in terms of molecular subtypes.

In the breast cancer space, sub-classifying based on cell-of-origin has already been seen with the 50-gene expression model (formerly known as PAM50). This helps to classify breast cancer based on basal or luminal subtype. However, there have been no studies creating a prostate-specific model based on cell-of-origin.

Because of this, Weiner et al aimed to create and validate an RNA expression-based prostate cancer subtyping classifier for cell-of-origin expression patterns. Their model was created with the help of the originators of the GRID cohort (NCT02609269), allowing for a large cohort 1 (training cohort) of 32,000 patients and cohort 2 (validation cohort) of 68,547.

Patients were grouped based on expression patterns of a curated panel of 8 genes and 13 expression signatures, with the most important factor relying on developing a signature to differentiate basal vs luminal cells in benign prostate cells. Using the 21 characteristics, the original discovery cohort was clustered into 4 groupings: luminal-differentiating, luminal proliferating, basal immune, and basal neuroendocrine-like.

Findings showed that when directly comparing this to the PAM50 model, there was a good amount of overlap. According to a presentation given at the 2022 American Urological Association Annual Meeting, luminal-differentiating tumors had the highest AR-activity scores, and luminal proliferating tumors had the highest expression of AR and cell proliferation genes and pathways. Basal immune tumors had the highest predicted response rate to radiation and possessed significant immune infiltration. Finally, basal neuroendocrine tumors were characterized by low AR-activity expression, low immune infiltration, and enrichment with neuronal gene expression.

Luminal proliferating and basal Immune tumors had more characteristics of homologous recombination deficiency and possessed more aggressive clinical characteristics including high proportions of grade group 4-5, node positive disease, and high decipher score. In a cohort of patients who were treated with radical prostatectomy (n=855), patients with luminal-differentiating tumors had longer metastasis-free survival compared to luminal proliferating (HR, 3.56; 1.83-6.93), basal immune (HR, 2.55;1.29-5.01) and basal neuroendocrine (HR, 3.30; 1.03-10.61).

In an interview with Targeted Oncology^TM, Weiner, chief urology resident at Northwestern University in Chicago, further discussed molecular subtyping for prostate cancer based on basal and luminal cell-of-origin and how such research may impact the future of this space.

TARGETED ONCOLOGY: What is the importance of molecular testing for prostate cancer? What advice do you have on which tests to order?

Weiner: Currently, there are many commercial products out there that can help us prognosticate prostate cancer and they're certainly adding to our ability to personalize prostate cancer management. At this point, we're still working towards a perfect subtyping mechanism for prostate cancer that can really distill what treatment makes sense for which patients. I think right now we have great tests that can help us determine the aggressiveness of prostate cancer or who needs to be monitored more closely, but we're still working on a great subtyping mechanism to figure out how we can personalize treatment.

What does the targeted therapy landscape currently look like for this disease?

There is a lot of research out there that has looked at subtypes of prostate cancer, though in large part, it's been limited in terms of its ability to be applied clinically, and that's for many reasons. They use a multi-omics approach that might not be available on a wide enough scale, or the cost might not be something that can be incurred by all the patients in the providers. We are still looking for something that can be applied more broadly.

Can you explain the purpose of your transcriptome database study?

The purpose of our study was to see, if we know that prostate cancer can develop from both basal and luminal prostate cells, is there a manner in which we could subtype prostate cancer based on the cell-of-origin, and can that help us subtype or sub-classify prostate cancer in a meaningful way based on probable treatment susceptibilities?

Can you explain the design of this analysis? What were the methods used?

This was a discovery and validation study. We used a cohort of about 30,000 prostate tumor samples, all from the commercial use of the decipher product. These were prospectively collected, their RNA expression patterns were completely done, and we were able to use previously validated signatures to create and cluster patients based on expression patterns that resembled cell-of-origin. After that at the validation cohort, which included another 70,000 patients, we were able to really validate these findings and show that if you do use this expression signature, you really can break up patients based on likely treatment susceptibilities.

What were the results presented at AUA?

We presented our preliminary approach. We've completed the creation of the signature and now, we're looking at the relevant, different markers based on the 4 subgroups created by the signature, all of which are basal versus luminal cell-of-origin. It's quite striking.

Each subgroup had a specific treatment susceptibility pattern and with that, you could use the information going forward if you have to apply additional treatments following prostatectomy or treatments if the patient has a metastatic recurrence. All of these things could be incredibly useful to practitioners.

How can these findings be applied to future research?

The next step is applying this new sub-classifier to clinical data. We've relied on gene expression signatures to preliminarily show the treatment susceptibilities based on the 4 subgroups. The next step is to take clinical data and see how the luminal cells respond to docetaxel or how the basal cells respond to androgen deprivation. Which group of tumors are the most likely to respond to radiation therapy? I think if we can show that, we can show that using a signature based on cell-of-origin is something worth our time.

What excites you the most about the future for this space?

What I find exciting about this is the idea comes from a similar signature in breast cancer, the PAM50, or what's formerly known as the PAM50 gene expression signature. That has been used widely and has been very helpful for people treating breast cancer.

There is no prostate cancer specific signature based on cell-of-origin, and we have the data to do it. I think this is the perfect project and has the means to accomplish what we're trying to do here. We know it's worked in breast cancer, and it's really helped a lot of providers and patients, and we're excited to apply that to prostate cancer.

_Reference:

1. _{Weiner A, IL C, Liu Y, et al. Molecular subtyping of >80,000 prostate cancer transcriptomes identifies four classes with distinct biological and clinical characteristics with implications for targeted therapies. Presented at: 2022 American Urological Association Annual Meeting; May 13-16, 2022; New Orleans, LA. Abstract PD07-07.}