Colin Pritchard: Well, it’s really an honor to be here. Thanks for the organizers. This was one of my favorite meetings a couple of years ago and I am anticipating that it will be even better this year in 2019.
I’m going to talk very quickly about the state-of-the-art in molecular characterization of advanced prostate cancer. Talking with my colleagues in the session, we tried to split out some of the elements. You’re not going to hear all of it here, I’m going to focus mostly on DNA damage repair and specifically DNA mutations and how those are important in advanced prostate cancer.
The emerging model for advanced prostate cancer, I think, it’s fair to say that, increasingly, germline testing, tumor testing, and liquid biopsy, that is plasma cell-free DNA testing, is being evaluated in an increasing number of our patients. Therapy is being guided by both germline and somatic findings, so we have to do both tests to be able to decide which treatments to give. Then, increasingly, as an oncologist, which I’m not I’m a pathologist, but our oncologists or maybe even our pathologists are having to think about genetic counseling based on these findings, as well as … What most people are probably interested in in the room, which is guiding therapy. This is this interesting mishmash of doing germline, tumor, plasma-based testing, indications all over the place. Genetic counseling, treatment, and also risk stratification which we’ll hear a little bit in one of the subsequent talks.
What are we talking about with DNA repair and prostate cancer? There are at least four DNA repair pathways that matter, up to 12, depending on if you talk to Alan D’andrea or someone like that, maybe even more. The DNA repair pathways that we know something about in terms of being actual right now are homologous or combination DNA repair and those are genes like BRCA-1/BRCA-2 but also genes like PALB2, RAD51C, RAD51D. Then, mismatched repair which is … I’m putting MSH2 as the key gene here because it is for prostate cancer but there are four genes we think about there, MSH2, MSH6, MLH1, and PMS2. There are germline-inherited syndromes associated with defects in both of these pathways, then, importantly, there are treatment implications. With homologous of combination DNA repair, we now know very well that that is predictive of PARP-inhibitor response in prostate cancer, also predictive of Platinum chemotherapy response and mismatch repair is predictive of checkpoint blockade immunotherapy. These are now standard of care kinds of indications, at least on the immunotherapy side. It hopefully, will be standard of care in the PARP-inhibitor setting soon.
If we’re thinking about DNA damage repair which I’m abbreviating DDR here, and mutation prevalence estimates in advanced prostate cancer … I’ve boiled this down, this isn’t exactly right but what I did is I went through a subset of the papers that have been published to try to really nail down, now that we have literally thousands of men who’ve been sequenced for these genes both on the germline and the somatic side, what is the overall prevalence if we just mush together all these populations? I try to make the numbers easy to remember too, they’re not exactly right but that’s the point is to have something to take home.
BRCA-2, about 5% of men with advanced prostate cancer, metastatic prostate cancer are going to have germline pathogenic variance, about 5% are going to somatic, together 10% are going to have BRCA-2 approximately. BRCA-1 about 1% and 1%, it’s probably a little bit higher than that, but again, just to keep things simple about 2% overall. ATM, maybe 2-3% somatic, maybe more depending on what study you’re looking at. About 2% germline, about 4-5% overall. Then, a mismatched repair which is largely MSH2, although some MSH6, some MLH1, about 4-5% somatic, about 1-1.5% germline. Then, together, about 5-6%, maybe higher in advanced prostate cancer depending on what study you look at.
It’s interesting. The one disconnect here, if you’re going to take something away from this slide is that for the homologous combination DNA repair genes, it seems about half the time when you find it, it’s germline, half the time you find it, it’s somatic. For mismatched repair, it’s not that way, it’s about 4-1 maybe 3-1. Where most of the time when you find it, it’s somatic and less often when you find it, it’s germline, but germline mismatch repair definitely matters.
What are some of the tumor features or other clinical features that can help predict whether you might have one of these? For homologous or combination DNA repair, a family history of breast and ovarian cancer is important. Also, histology. We have Dr. Bristow sitting here who’s done some of the pioneering work to show the role of intraductal histology and how that’s associated particularly with BRCA-1 and BRCA-2 defects. Also, ductal histology which is different but also looks like it’s associated with that. For mismatched repair, family history of colon or endometrial cancer, the Lynch syndrome tumors, is predictive. Also, ductal histology, probably intraductal histology as well. Then, primary Gleason pattern 5. A study of a Hopkins’ group showed that maybe even up to 8-10% of men who have primary Gleason pattern 5, that’s like five plus five, five plus four, cancers will have mismatch repair deficiency.
Not all DNA damage repair genes are the same. I’ve already highlighted some of those differences but, again, I think it’s worth just hammering this home. Particularly because our guidelines as they’re written right now, tend to lump these together without any granularity of this gene versus that gene. BRCA-2 I think is probably the most important DNA damage repair gene in prostate cancer. It’s got a high germline prostate cancer risk, pretty well established now of PARP-inhibitor and Platinum chemotherapy response if you have, in fact, bi-allelic inactivation, that is both copies of the gene is taken out. Some emerging data with checkpoint blockade, PD/PD-L1-inhibitors, but that’s still emerging.
This next category in orange would be what I’d consider moderate. Both moderate in terms of germline risk and more moderate in terms of the data that they predict response to therapy. Of course, BRCA-1 and BRCA-2 are often lumped together in this way, but I think if you look at the data carefully, the data is stronger even for the therapy side in BRCA-2 than BRCA-1, although we could argue about that. PALB2. It’s an emerging evidence but I do believe it may end up being as important as BRCA-2, although much more rare but there’s still emerging evidence there.
If we skip down to mismatch repair, MSH2 is really the heavy hitter here and that’s different than if you look at colon cancer or endometrial cancer and that’s interesting. There’s some interesting biology here. Why is it that MSH2 of the four key mismatch repair genes is so important in advanced prostate cancer compared to the other three? I don’t know but it does look like it is more important. MSH6 and MLH1 are also important genes that are definitely associated with prostate cancer risk and associated with a response to checkpoint blockade immunotherapy when there’s evidence of microsatellite instability, so that’s a key thing. Just because you have an inherited mutation that is definitely a damaging mutation in one of these genes, like say you do germline mutation testing on your patients and you find this, that doesn’t mean that it actually caused the prostate cancer. Unfortunately, we have a phenotypic readout for that in all cancers that’s microsatellite instability. One thing to emphasize here is that PMS2, it’s very limited data about the role here, so I think it does have a role but if you get a germline PMS2 result and you have a patient with Gleason 6 prostate cancer, I wouldn’t necessarily say that they’re related. It could be true-true unrelated, it could be related.
This is a paper that just came out from the Lynch syndrome screening study and I thought it was nice because these are thousands and thousands of patients that were ascertained on cancers other than prostate cancer. Then, what they looked at was accumulative cancer incidence by gene for each of the four Lynch genes across a number of cancers including prostate cancer. So, what I did is pulled out the data for prostate cancer for this because I thought it’d be interesting for this group. If you look at the cumulative prostate cancer incidence in this very large cohort that was ascertained in a nice way, you can see that MSH2 really stands out as the key gene in terms of cancer predisposition. MLH1 and MSH6 are both above baseline, so they’re definitely associated with prostate cancer predisposition, but lesser. PMS2, I think there were only one or two patients so that’s why it’s just 4.6% down the board. When you’re looking at age 50, 60, 70, 75. Nonetheless, once you get to age 75, it’s not really significantly above baseline which is interesting.
Getting to the guidelines. I’m talking with Dr. Eeles and Dr. de Bono, we were kind of trying to talk through how to divvy this up. I’m the guy from the United States, so we’ll talk about United States guidelines. The NCCN guidelines that are relevant in terms of DNA damage repair: Consider metastatic biopsy; Consider tumor testing for microsatellite instability-high or for deficient mismatch repair; Then, the guidelines actually specifically state to use a MSI test that’s validated for prostate cancer. I’ll talk about that very briefly in a couple of slides; Consider germline or tumor testing for homologous or combination DNA repair. These are the genes listed: BRCA-1; BRCA-2; ATM; PALB2; FANCA; RAD51D; and CHEK2. Hopefully, I’ve impressed upon to you that these genes are not the same. To me, that list is just wildly different in terms of the level of evidence. On the lower end would be FANCA, maybe RAD51D. On the higher end, of course, BRCA-2 and probably PALB2.
There was also this group, the International Society of Urologic Pathologists that I was peripherally involved with who got together back in March at the United States and Canada Pathology meeting to come up with some guidelines from a pathologists’ perspective. So, again, because this meeting happened in the United States, I’ll talk about this as well. Basically, their recommendation was very similar to NCCN. Germline panel testing for DNA repair genes in metastatic and also in high-risk localized disease. Somatic tumor DNA testing in all metastatic patients and that should include testing for defective mismatch repair either by immunohistochemistry and/or by MSI or gene sequencing, and somatic testing for defective homologous or combination DNA repair by sequencing of BRCA-1/2 at a minimum.
Then, I think the most important thing to take away from this is they concluded that testing on metastatic tissue is a good thing but, if unavailable, a primary tissue is an acceptable surrogate. This is an important question as we do more and more testing and try to decide, “Do we do a biopsy when we already have a primary that we could use and subject that man through that?” I think the evidence is still emerging here but the evidence that does exist suggests pretty high concordance particularly for mismatch repair DNA repair deficiency status between primary and metastasis but also, probably for homologous or combination DNA repair. Of course, it’s the best thing to test the metastatic biopsy but primary tissue probably is a decent surrogate but there are going to be times where it’s discordant, so that’s important to emphasize.
What are some of the issues with homologous or combination DNA repair deficiency testing? So, confirming that it’s bi-allelic. These are tumor suppressors. You don’t really get the functional effect unless you’ve knocked out both copies of the gene, so having a mutation doesn’t mean that you’ve knocked out the pathway, doesn’t mean your patient’s going to be responsive to PARP inhibitors necessarily.
There are a few ways that these could be confirmed. A number of so-called signature assays are out there now. There’s a homologous or combination DNA repair deficiency signature analysis that’s analogous to microsatellite instability, but for that, so that’s a functional readout. If that’s positive, that’s useful. Also, for specific genes like ATM, Dr. de Bono and I had a long discussion about the role of immunohistochemistry for ATM to confirm that there is, in fact, bi-allelic ATM loss at the protein level. Unfortunately, we don’t have great IHC assays for BRCA-1 and BRCA-2, I wish we did, but we do for some of these. Those are just a couple of examples about thinking about it. When it is important, when it is in doubt, definitely think about confirming that the lesion you’re looking at is bi-allelic. Also, just call your lab. The director … If they don’t know how to look at this, then you’re in trouble. But, the director should be able to say, “Yeah, I’m looking at it and I see there’s loss of heterozygosity and I see the variant fraction. Yeah. I’m pretty sure it’s bi-allelic.” So, even if it’s not on the report, you can have that one-on-one discussion without even any further testing, is another approach.
Variant interpretation is still an issue. We talked about this two years ago. I think in the talk I gave, about half of it was about variants of uncertain significance and all of the workaround that, which is a really interesting area. Unfortunately, it’s still really an issue. Commercial labs, particularly commercial labs that focus primarily on somatic tumor testing, there’s some work to do in terms of the variant interpretation. The bottom line here, is if you get, particularly a somatic report that reports a pathogenic mutation in a gene that you’re really interested in treating your patient with, the temptation may be, “Gosh, I just want to treat right away.”, but the first thought may be, “Well, is this really a pathogenic mutation? Is this really telling me what I think it is?”, it’s worth always asking that question.
Bi-allelic inactivation matters. There was this paper that came out recently from the [inaudible 00:13:57] Group where they looked at a number of cancers and they looked at BRCA-1 and BRCA-2 mutation status both monoallelic or bi-allelic. Then, they looked at cancer types that were traditionally driven by BRCA-1 and BRCA-2 like prostate cancer, and cancer types that are not driven by it. What’s shown in this figure here is when you have a germline pathogenic mutation in BRCA-1 and BRCA-2, that’s the blue, what percentage of the time do you see evidence of loss of heterozygosity in the tumor? That is what percentage of the time does the tumor look like this gene caused the cancer? That’s the way I think about it. You can see for breast and ovarian on the far left, there’s the highest level up above 75%. For pancreatic and prostate, well above 50%. Great news, but not 100%. Just because you have a mutation, you can’t assume bi-allelic. Of course, this is just looking at LOH.
You can see for the other tumor types, there’s basically no evidence. Bi-allelic is associated with homologous or combination DNA repair deficiency as measured by score.
Another issue to think about is with mismatched DNA repair. For those, the specific to think about are the accuracy of the MSI test, the accuracy of the IUC if you’re using that, and the fact that it’s technically challenging to detect underlying mismatch repair mutations, particularly because they tend to be large rearrangements, structural rearrangements, which are technically challenging.
Microsatellite instability patterns are not the same between cancers. Prostate has it’s own specific microsatellite instability pattern. So, if you’re using a generic MSI test that’s not validated for prostate, you might run into trouble. We and others have looked at this. When you’re using a prostate-validated MSI test, it works much better but there can still be issues.
In the last two slides, and I know that the cowbell is coming, we’re going to talk briefly about liquid biopsy. Liquid biopsy is very exciting, I definitely encourage using it. The big take-home message there, I think, is think about which patients you’re testing when you do liquid biopsy. It’s probably only the patients with high-burden disease in which this testing is really going to be informative. It’s sort of like garbage in, garbage out. I like to talk about if you just put a needle in someone and you miss the tumor and then you test that, obviously, that test has no meaning. It’s the same idea with the liquid biopsy. If it’s a non-diagnostic biopsy, if you don’t have enough cancer there, the test is meaningless. Yet, right now there’s not a lot of standardization around that, so you can just send in your liquid biopsy to a commercial lab and say there’s prostate cancer. They’ll test it and report it back out as if there’s prostate cancer there when, in fact, sometimes there’s not. That’s even with these ultra-sensitive assays. They’ll say 1 in 100,000 sensitivity, doesn’t matter, still a problem. Basically, a nice cutoff that we published just recently that seems to hold up fairly well in our experience is, metastatic prostate cancer, PSA over 10, or variant histology, you’re probably in pretty good shape. PSA less than 10, with standard histology is going to be more dicey.
This is just my last slide to show the pitfalls of liquid biopsy. Ken Pienta and his colleague did a split sample send out of 40 metastatic prostate cancer patients to two commercial labs. Basically, there was very poor congruence but when you split that data out by PSA, if you look at the four patients they had of those 40 that had PSA greater than 10, they actually look like they have decent congruence and they actually have prostate cancer mutations in their plasma. If you take all the other congruent patients, which there were only five of those 40 that were congruent at all, with PSA less than 10, unfortunately, the congruence might be double false-positives in both places. What I mean by that, is probably clonal hematopoiesis that’s being misinterpreted as prostate cancer. So, that’s even worse than being not congruent. We’re speculating there but it makes you wonder because the congruence was all in mutations P53 and ATM which are well known to be contaminating due to clonal hematopoiesis when you do plasma-only testing. So, think about a lab that does plasma and whole blood control when you’re thinking about liquid biopsy, is the bottom line there. This can be used for treatment resistance, for example, for reversion mutations.
In summary, germline and tumor NGS testing is increasingly being recommended to guide therapy in advanced prostate cancer. Liquid biopsy testing is increasingly useful in place of tissue testing but beware of what I just talked about. Interferences with clonal hematopoiesis and selecting your patients carefully so they have adequate burden disease to get an adequate result. Think about a molecular pathologist because sample methods and interpretation matter and there’s an increasing role there.
So, thanks very much. A ton of people have done this work but I’m over time and I really appreciate everyone listening. Thanks so much.