Read the Full Video Transcript

Andrea Miyahira: Hi, everyone. I'm Andrea Miyahira with the Prostate Cancer Foundation. Please welcome Dr. Jonathan Chou of UCSF, who will present his recent paper, "Synthetic Lethal Targeting of CDK12-deficient Prostate Cancer with PARP Inhibitors," that was published recently in Clinical Cancer Research. Dr. Chou, thanks for joining.

Jonathan Chou: Thank you, Andrea. So I'm going to talk today about CDK12 and our work on identifying PARP inhibitors as a strategy for specifically targeting this subclass of prostate cancer. The outline that I'll tell you about today is: first we'll talk a little bit about the clinical characteristics of prostate cancer patients with CDK12 mutations and why we really care about this particular subclass. And then I'll talk a little bit about our target discovery approach. How do we find a new target?

I oftentimes call it the Achilles heel of this particular mutation. And how does the type of mutation affect sensitivity? And then we'll take a look at some of the data out there on patient responses. How do patients do who have CDK12 mutations? How do their mutations differ in terms of their response to the therapy?

So a few years ago, Melissa Reimers, Stephen Yip, and I published this paper in European Urology, describing the clinical characteristics and outcomes of patients with this particular CDK12 subclass. These patients—it's about a 7% frequency in advanced castration-resistant prostate cancer. And if we take a look at the mutations in this particular gene, you can see that most of the mutations are truncation mutations. And these truncation mutations actually are scattered throughout the entire gene. So these are mutations that lead to early frameshifts and stop codons. And then there's also a subset of patients who have mutations in the kinase domain here in the blue box here.

And what you can see is that there's no specific hotspot mutation in the kinase domain, but there are many different mutations that are clustered around this region over here, which we'll talk about in just a few slides. Patients with this particular mutation of prostate cancer are enriched in very aggressive, high Gleason score prostate cancers. You can see that over 90% of patients have Gleason 8 or higher disease, compared to some of the other subtypes here.

And then you can also see—if we take a look at one surrogate biomarker, time to metastasis—you can see that patients with CDK12 mutations develop metastatic disease in 35 months, compared to patients with HRD mutations or even p53 mutations, which are 61 and about 55 months, respectively. And if we take a look at some recent data at one of the newest FDA-approved therapies for men with castration-resistant prostate cancer, PSMA lutetium or Pluvicto—this is actually utilizing data from Oliver Sartor's recent paper—you can see that patients with CDK12 mutations have a fairly poor response. Only about 20% of patients achieved a PSA 50 response, compared to, again, p53 mutations, BRCA mutations, and even patients with AR mutations or amplifications. So a group of patients that not only have poor clinical characteristics but also seem to be fairly resistant to our standard-of-care therapies.

So when I was a postdoc in Dr. Felix Feng's lab many years ago, we were very interested in this particular subclass and wanted to develop and identify new ways of targeting this mutation. And so to do this, we actually generated CDK12 knockouts in both the LNCaP and C42B cell line models to be able to use and manipulate for our subsequent studies. One of the first things that we did was we took our control and our isogenic knockout cell lines and subjected them to about 1,800 FDA-approved compounds, trying to identify compounds that would kill the knockout but not the control cell line.

And so this is represented over here in panel C. And what you can see is that there were a number of different compounds that came out that corresponded to these DNA damage class of molecules. And in fact, two of our top hits were the PARP inhibitors olaparib and rucaparib. And so what we did was we further validated that our knockout cells were, in fact, sensitive to these PARP inhibitors. And you can see using the three different knockout LNCaP clones and also three different knockout C42B clones that all of the clones were more sensitive compared to the control, as shown in black.

Now, one of the interesting things is that depending on the cell line, whether it was LNCaP or C42B, you can see that the degree of sensitivity was actually a little bit different. So whereas the C42B, when we knocked out CDK12, had about a log fold change, the LNCaP cells had about a two to three log fold difference in terms of sensitivity.

One of the other interesting things that we saw was that in the process of making these knockouts using CRISPR-Cas9, we also generated multiple different heterozygous deleted mutations. And if we take a look at the two HETs here in red and the green dotted lines, you can see that their sensitivities to olaparib actually remain the same as the control. And you can see here by Western blot that the HETs had lower amounts but not a complete absence of CDK12.

And so what this told us was that there was a difference between these biallelic mutations and these heterozygous monoallelic mutations in terms of their PARP sensitivity. And we think that this is actually probably important for multiple different types of drug sensitivities.

The other thing that we saw that was interesting was that the canonical hypothesis in the field was that CDK12 loss generated PARP sensitivity because of downregulation of genes involved in homologous recombination, or HR. And what we saw in the knockouts, when we looked at genes that were broadly involved in both homologous recombination and other forms of DNA repair, was that there was not a significant decrease in these HR genes like BRCA1, BRCA2. This was in contrast to an acute inhibition of CDK12, where we used a small molecule called THZ531 that inhibits both CDK12 and CDK13 in the kinase domain, in which we did see that BRCA1 and BRCA2 were significantly downregulated, in addition to other genes such as ATM, ATR, and DNA-PK.

And so what this told us was that these chronically depleted cells—these CDK12 cells that had been chronically depleted of CDK12—actually had a different genetic phenotype as well, and that the HR downregulation may not be that important in terms of generating the PARP sensitivity.

The other thing that we did was we took a look at the kinase domain point mutations. And again, there's not really any hotspot mutations, but we generated constructs that contained these two more common mutations—the R858W and the D918G mutations—and re-expressed them in our knockout background to ask whether or not re-expressing these mutations would change the PARP inhibitor sensitivity.

What you can see here is the dose-response curve. So again, the knockout here in blue. The control parental cell line in black. If we re-express the full-length wild-type form of CDK12 here in magenta, you can see a near-complete restoration of the PARP sensitivity. And so the knockouts, when we re-express wild type, basically look like the control cells. When we re-express these two mutants in gold and green here, you can see that there's a partial restoration.

So two things that are important here. One is that these kinase domain mutants are more sensitive than the control and the wild type. But the second important thing that we saw here was that they were not as sensitive as the complete knockouts. We also generated these different truncation mutations. So again, here, I'm showing you the full-length wild type. And then we truncated either after the kinase domain, after this proline-rich motif, or after this arginine-serine domain here and re-expressed these constructs in the knockout cells.

And again, what you can see is that all of these truncation mutations remain pretty sensitive, with the exception of a slight shift to the right with this RED construct, which is this kinase domain truncation mutant.

So finally, what we wanted to do is take a look at patients and some of the patient data that had come out of the multiple different trials that had been conducted. And so we reanalyzed the TRITON2 patients. This was a phase II single-arm study for patients who were post-abiraterone and enzalutamide and also post-docetaxel chemotherapy. And in this particular trial, there were actually 15 patients that had CDK12 mutations. When we now subgroup these CDK12 mutations into patients who had true biallelic mutations versus these gene rearrangements versus these monoallelic mutations in gray and red, respectively, what you can see is a very different response to the PARP inhibitor rucaparib. So in fact, 6 out of the 11 patients with biallelic mutations actually had PSA declines, compared to 0 out of the 4 patients with these monoallelic or these gene rearrangements.

I'm showing you the PSA trends of two of these patients—one who had a PSA 50 response that lasted for about 26 weeks before it started going up, and then another patient who only achieved a PSA 30 response but had actually a fairly stable PSA for almost 35 weeks before PSA progression. And if we take a look at a recent paper that was published by the FDA, in which the authors grouped across multiple different trials that have been conducted with various different PARP inhibitors, including rucaparib, olaparib, niraparib, and talazoparib, you can see in blue here where the CDK12 patients who got the PARP inhibitor plus an ARPI—and you can see that there was a hazard ratio under 1 for both progression-free survival and OS, suggesting that there was improvement for patients with CDK12 mutations if you added a PARP inhibitor. And here, again, just showing you, in terms of second-line treatments, the response rates were fairly low. But again, we think this is probably due to lumping all of the CDK12 mutations into one cohort. And again, with the addition of the ARPI, you can see that there was a 30% increase in responses. And these are responses—radiographic responses by BICR.

So in summary, the take-home messages that I hope to have shown you today are that CDK12 mutations occur in about 5% to 8% of advanced prostate cancer patients, the majority of them biallelic mutations, and that the different types of mutations—whether they're frameshift, loss-of-function mutations versus kinase domain point mutations—seem to have differential sensitivity, and that downregulation of these homologous recombination genes may not be a critical factor in determining sensitivity, at least to PARP inhibitors. We have to look closely at the type and also the allelic frequency of the mutation. In other words, not all mutations are created equally in terms of sensitivity to the drug that we're testing, and so we may not be able to lump them together in clinical trials. And that additional therapeutic targets are really needed for CDK12 mutant prostate cancer patients, and that we should be exploring combinations of PARP inhibitors plus other targeted therapies.

And so with that, I would like to acknowledge my lab as well as my mentors, Dr. Felix Feng and Dr. Alan Ashworth, as well as external collaborators Lixing Yang at the University of Chicago, Jean Tien and Arul Chinnaiyan at the University of Michigan, my funding sources including the Prostate Cancer Foundation, and UroToday for inviting me to give this short seminar. Thank you very much.

Andrea Miyahira: OK. Thank you for sharing that. So what degree of CDK12 loss is typically observed in patients with CDK12 mutations? And how does this relate to PARP inhibitor sensitivity?

Jonathan Chou: Yeah. We think, depending on the cohort that we're taking a look at, somewhere between 5% to 8% of patients. And oftentimes, I get the question, "Well, Jon, you're studying a relatively rare subtype of prostate cancer." And I remind people that these relatively rare subtypes can actually have really important and profound implications for how we should be treating them.

And if we take a look at, for example, lung cancer, which has really subtyped their patients with EGFR, ALK, ROS, BRAF, et cetera mutations, many of these subclasses are 5% or even less. And so I do think that we have to be thinking about our patients and how we utilize the wealth of genomic information that we have now on trying to figure out the best therapy and potentially combinations of therapy, particularly up front.

Andrea Miyahira: Thanks. And so can other DNA repair genes compensate for CDK12 loss for cell survival, for PARP inhibition sensitivity, et cetera?

Jonathan Chou: Yeah. That's a great question. One of the things that we had hypothesized and we've been working on is that CDK13, which is the most closely related CDK to CDK12, may be compensating for its function when it's lost. And in fact, there may be another synthetic lethality and vulnerability there.

And so with Dr. Tien and Dr. Chinnaiyan, who have developed CDK13-specific inhibitors and degraders, we've been looking and testing whether or not these CDK12 knockouts would be more sensitive. And in fact, we've shown that they are, at least in the cell line models and PDX models that we've used.

Andrea Miyahira: Thanks. And based on this data, are there specific CDK12 alterations that would recommend patients for treatment with PARP inhibitors?

Jonathan Chou: Yeah. That's a great question. I've had a couple of patients who have these true biallelic frameshift mutations that I've recommended PARP inhibitors for. And we've seen, again, I would say like a moderate PSA response. Typically, these are patients who are pre-chemotherapy. I see a better response than patients who are post-chemotherapy.

I really do think that we need to be exploring combinations with PARP inhibitors and trying to figure out how to augment the PARP inhibitor response for these CDK12 mutations. Because I think the majority of patients, unfortunately, either have a very short response or have no response.

Andrea Miyahira: OK. Thank you. And how do you think tandem duplications caused by the CDK12 alterations impact PARP inhibitor sensitivity? And can this generate novel resistance mechanisms?

Jonathan Chou: Yeah. That's a really fantastic question. It's something that we are very, very interested in exploring. And so one of the really interesting characteristics of the CDK12 mutant prostate cancers is that the genomes are littered with these large tandem duplications. And this is a genomic signature that is very different than the BRCA2 signature that we typically see.

And in forming these tandem duplications, we certainly see that many of these are located in AR-driven genes like AR itself, FOXA1, Cyclin D. And so we definitely think that these amplifications in these oncogenes or these pro-tumor genes are really, really important for generating the PARP inhibitor resistance. One of the other interesting findings is that these tandem duplications seem to be forming over time. And so maybe there's an opportunity to be using PARP inhibitors up front, where maybe there's fewer of these tandem duplications. And so maybe the patients would be more sensitive. But yeah, it's a really wonderful question, and we really need to do a lot more to understand the relationship between the tandem duplications found in these CDK12 mutant patients and the response to therapy.

Andrea Miyahira: OK. Well, thank you so much for sharing this study with us today.

Jonathan Chou: Thank you so much for inviting me.