The Uromigos Episode 143: Paper of the Month—Determinants of Anti-PD-1 Response and Resistance in Clear Cell Renal Cell Carcinoma

Dr. Samra Turajlic describes her paper on clear cell renal cell carcinoma.

Episode Transcript

Tom Powles:
And welcome to our paper of the month, which Brian and I are hosting from the Canary Islands, would you believe. This is our second external … I think external’s a fair word. We’re … Podcast … We’re joined by Samara. And Samara, congratulations on a paper of the month. Your cancer cell paper, we thought was a tour de force. We really enjoyed it. And maybe you’d like to introduce yourself, and then Brian’s got the first question for you.

Samra Turajlic:
Great. Okay. Well, I’m really sorry that I’m not in the Canary Islands actually.

Tom Powles:
You’d love it. You’d love it. I can tell you. You’d love it.

Samra Turajlic:
It’s a fabulous location to host the podcast. And I’ll hope you’ll continue to find exotic locations, and to in invite your speakers in the first place.

Brian Rini:
Good idea.

Tom Powles:
In the future, we should be inviting our speakers. I agree. It does seem a bit perverse that …

Samra Turajlic:
So just for everyone’s reference, I’m Samra Turajlic. I’m a medical oncologist at the Royal Marsden. I treat kidney cancer and melanoma. And I run a research group at the Francis Crick Institute in London as well. And we are interested in cancer revolution in the context of tumor progression, tumor metastases, and then of course, very topically, treatment resistance. And we mostly do this in the context of kidney cancer, because it’s the most fun and mysterious cancer model to work with.

Brian Rini:
Fun and mysterious. I love it. Samra, thanks for joining. My first question is really around the construct of this adapter study, the tracer RX study in the piece study. And I mean, it’s such a great … It’s such a great setup for, you know, getting tissue and studying heterogeneity and all the contributions. Do you want to talk first just about how you got that up and running, challenges, successes?

Samra Turajlic:
Sure.

Brian Rini:
I imagine people around the world would love to do the same thing, but it’s very difficult as you well know.

Samra Turajlic:
Yeah, of course. No, I think it’s important to highlight how challenging this is, and why we are motivated to do it. And I know some people might think us obsessed with heterogeneity. In fact, Brian Rini called me heterogeneity police. The very first time I met him.

Tom Powles:
He’s called me worse things Samra.

Samra Turajlic:
He probably won’t remember this, but we were in St. Andrews in Scotland. And you were presenting some, you know, beautiful data, biomarker data. And you said, “And for the heterogeneity police in the room, here the … [inaudible 00:02:55]”

Brian Rini:
Well, it doesn’t sound like I called you out by name. I was just generally referring to …

Samra Turajlic:
So, first of all, to say … I mean, the study’s been a labor of love, really. It’s taken 6 years to complete. The cruel itself was difficult because we had these mandated biopsies at the baseline, then 9 weeks into treatment, and then at progression. So, you know, we did struggle with that. And then the other part that was really heavy going was actually the multiomic analysis of all these different types of samples. And then we were very humbled that a number of these patients, who unfortunately didn’t benefit from the treatment or the subsequent lines of therapy, signed up for the post-mortem study. So that obviously expanded the number of metastatic sites that we could look at. So why multi-region? And just to highlight when we took the biopsies at these various time points, they were always multi regional, with respect to the primary tumor if it was in situ with respect to the metastases. And in some instances, we had both.

So, I think there is a broad appreciation now that [inaudible 00:04:19] is a reason for failure of many biomarkers. So, we were very mindful of that. But, you know, also given the data from … the beautiful data from David Braun, he’s nicely shown that T-cells in myeloid populations are distinct in advanced versus early stages of disease. So, it was very important for us to have a baseline biopsy at stage 4 disease when probing the tumor microenvironment. Never mind the cancer mutations.

And it’s important to remember that most biomarkers studies rely on archived FFP, which could be temporarily separated from the current disease and the patient by years, if not decades. And then the last aspect of this is that the sampling was longitudinal, which is very hard to do as I highlighted already. And its novel in RCC, but we’ve learned from studies in other settings that investigating these dynamic tissue-based biomarkers yields great insights. And obviously it gives us an idea of how our biomarkers vary both in space and time putting it all together.

Tom Powles:
Samra, you did whole exome sequencing, RNA sequencing. You did T-cell receptor work. You did multiplex IHC, flow cytometry, and single cell RNA sequencing. You only have 13 patients, which clearly is going to be a problem. And you took 115 samples. Why did you need to do so much biology? And were you underpowered for some of your interactions?

Samra Turajlic:
Yes. Yeah. No, that’s a very, very fair question Tom. You weren’t reviewer 3 by any chance, were you?

Tom Powles:
I wasn’t. I wasn’t actually, Samra. I wasn’t. I promise.

Samra Turajlic:
So, we had just to correct you. Not 13, but 15 patients. Which obviously, when it comes to statistics, for any single biomarker, we were underpowered, certainly. So, this was a deep exploration into, as I said, the fun and mysterious kidney cancer. So, we know that we have this background in kidney cancer of confusion, really, and contradiction when it comes to biomarkers that are perfectly good predictors of response in other settings. So be it tumor mutational burden, or indels. And then furthermore, we have the tumor microenvironment that is very stereotypical for kidney cancer, with high levels of immune infiltration, which pertained a poor prognosis outside the setting of treatment [inaudible 00:07:13].

And one thing, I remember in a meeting about 10 years, an IO … sort of early IO meeting, when I mentioned the fact that kidney cancer are the most infiltrated solid tumors there are. Someone said, “Oh, they might not be specific. These TCEs might not be specific to the tumor at all. It’s a highly vascular tumor. There’s a lot of leakiness. They could just be passing through and extravasating, and in fact, they’re not tumor specific.” And, you know, the fact that we don’t have an understanding of the antigen, or the nature of immunogenicity of this tumor … And it is highly immunogenic.

We have multiple levels of evidence of that from spontaneous regression to IL-2 and now immune checkpoint blockade. Also meant that we didn’t really know where the priming of these T cells occurs. You know, because if the antigens of the tumor cells are presenting not the ones that they’re stimulating, the T cell responds. Then you, you know, what’s behind this? So, I think it was necessary to look with as many methods, and with very, very fine level of resolution as possible. So, in that sense, this work is hypothesis generating. It’s not a biomarker validation study. But you know …

Tom Powles:
It was a tour de force. And we’re going to start in reverse order, if I may. We’re going to come to the genetics towards the end. I know there was work on TMB, and mutations. And I’m not sure how much that added specifically.

Samra Turajlic:
No, it’s really quite boring, isn’t it? So, I think we can leave it out.

Tom Powles:
I thought the last bit was really cool. And I was really interested in this issue around the heterogeneity of the immune infiltration, and the clonality of that. The preexisting component of that, as opposed to the immune infiltration, and the work you did on that. And then the final piece around that was the effect of nivolumab and nivolumab binding. And I thought that whole piece was just fascinating.

Brian Rini:
Mm-hmm (affirmative).

Samra Turajlic:
Okay. Well thank you for those very kind comments. So, I think probably we can leave the RNA seq data, bulk RNA seq data out. Because there is a lot of it out there, and I’ll just … I won’t go into the details of the transcriptional signature. I mean, Brian and yourself have done beautiful work in this. We know that there are signatures there associated with the response. We were pleased to see that the signatures that have been described by both of you perform well in this cohort.

And that was in the heterogeneity. Something that we are doing in the lab at the moment is trying to identify what components of each signature actually don’t suffer from [inaudible 00:10:27] so that we can move them forward in a more routine fashion in the clinic where you’re not taking multiple biopsies and measuring everything from multiple biopsies.

But we did see heterogeneity of immune infiltration in a small number of patients. And we also saw a change from a pre-treatment to a post-treatment state of immune infiltration, suggesting that at least in one or two instances, there were tumors immune cold within the resolution of our multiple biopsies. And we still are probably missing a lot of heterogeneity within this.

So, they may have been hot in some spaces. But when we bio them again on treatment, there was an infiltration of T-cells. But the question really is, is that [crosstalk 00:11:19] anti-tumor response? So, this is where the TCR sequencing was really crucial for us, because we were able to determine the clonality of the TCR, and then just for our audience to expand it … That tells us if there is clonal expansion of a particular TCR sequence, that tells us that these T-cells are expanding in response to a particular antigen.

And then furthermore, we can look at the similarity of sequences as well, which also then infers that they’re likely to respond to the same antigen. So, what we’ve seen is that the responders have a higher clonality score than non-responders at the baseline. So, this suggests that there is a preexisting immune priming, a recognition of the tumor, and that these T cells can be reinvigorated. Which is, you know … I’ll move on to discuss the nivo binding data in a moment.

Brian Rini:
So Samra, that’s key. There’s a preexisting, prior to giving these patients nivolumab, the responders already had an expanded repertoire of T-cells that were sitting there if you will.

Samra Turajlic:
That’s right. Yes.

Brian Rini:
And this might be a dumb question. We don’t necessarily know where they came from or how they got there. Is that correct?

Samra Turajlic:
So … Yeah. So, what we don’t know is … We know that they’re engaging tumor cells, so they’re responding to an antigen of some sort.

Brian Rini:
Yep.

Samra Turajlic:
In our data so far, we don’t see … And you haven’t seen correlation with tumor mutational burden, or correlation with indels, which, you know, we had hoped would prove the key to the mystery of immunogenicity here. Now, ultimately, to show what these T cells are reacting to, we need a T-cell reactivity screen.

So, we need to predict all the possible antigens, including the endogenous retroviral antigens, and I’ll come to those in a moment, as well as any neoantigens that are informed by mutations, whether they be point mutations or indels, as well as any cancer antigens. So just over expressed cancer specific antigens. And do a T-cell activity screen a la manifest, and similar, to prove what these T-cells are reacting to. But this study provides strong evidence that the antigen is presented by the tumor cells.

Now, it all also gives hope to the TCR metric being a potential biomarker. And I think it does have implications in the adjuvant setting for, in the context of data in pembrolizumab, because I would argue that the patients who are likely to benefit from adjuvant PD-1 will be those that do have, again, preexisting clonal expansions of antigen reactive T-cells, tumor-specific T-cells, in the environment. And because these patients are getting nephrectomized, there is an opportunity to get this biomarker established in that setting.

Tom Powles:
Samra, could you give us an outline of the effect of your experiments with nivolumab binding, and the effects of nivolumab on that T-cell repertoire?

Samra Turajlic:
Yeah. So, what happens on treatment, is that new clones new, T cell clones, do appear at the tumor site, both in the responders and non-responders. But only expanded T-cells are present [inaudible 00:15:06] correlate with response. And we see that also, these are the cells that are nivolumab bound, and these are the cells that up regulate the signatures of cyto toxicity.

So, you know, I just want to highlight this question of maintenance of expanded TCR clones, because there is a broader debate in the community, which hasn’t yet settled about whether the response to PD-1 is to do with replacement of clones, or maintenance and expansion of preexisting clones. And there is a recent online study in triple negative breast cancer that also shows that it’s to do with maintenance. So, I think this is addressing sort of a broader question.

Perhaps I can come to the ERVs, because I think that’s … Because it’s been nominated as a biomarker in a number of studies, I’d really like to discuss how we’ve shed light on how those associations might have emerged. So previous induction endogenous retrovirus signatures, we’ve concluded likely just reflect immune infiltration. It is very meaningful because we need to narrow down the search space for ERVs as a potential antigen. So, there are millions of ERVs in the genome. And if you ask someone to screen all of them for immunogenicity in kidney cancer, it would …

Tom Powles:
Samra, can I pause you there for one second?

Samra Turajlic:
Yes.

Tom Powles:
The reception wasn’t fantastic for a second there. You are talking about the human endogenous retroviral expressions.

Samra Turajlic:
Endogenous retroviruses. So, these are retroviral elements, which were incorporated into human genome long time ago. They’re normally transcriptionally repressed. So, they should not, you know … Although they have open reading frames, there are intragenic regions, and in normal cells, they’re not expressed because they’re repressed. Now, why they’re exciting as potential immunogenic antigens? Well, because they’re transcriptionally repressed, they’re not necessarily subject to central tolerance because the immune has not seen them. And therefore, when they’re re expressed, which is what happens in cancers in general …

And it happens a lot more in kidney cancer than in other tumor types, potentially because we have so many mutations in chromatin modifiers. And as these heterochromatin structures are altered, these retroviral elements become expressed, and therefore can be presented to the immune system.

So, what previous studies have shown is that there are certain signatures of these retroviral elements. And as I said, there are millions of them. So, us studying them one by one is never going to happen. And what we wanted to do with these signatures is firstly, align the … Do [inaudible 00:18:24] alignment of these elements. And we collaborated with a very talented scientist here at Crick called George Kassiotis who’s an ERV expert.

So, he helped us with the alignment. But the second thing that we did was to look at the expression of ERVs in immune cells. And immune cells have a very specific signature of ERV expression. And in fact, what we found is that those previously reported signatures that were thought to be originating from the tumor cells, in fact, originate from immune cells in the tumor.

So, the correlation makes sense. So basically, the more immune infiltration you have, the more of these ERVs you express. But it shows us that the correlation is not direct. And these ERVs that have been identified are not tumor specific at all. And therefore, the correlation is not about the antigenicity of the ERVs. It’s simply reflecting immune infiltration, which we know, associates with the response.

Brian Rini:
And did … Samra, in that previous work, I know there was, in your paper, there was a lot of about the technical nature of how they were determined. I didn’t quite follow that. There were different methods used across different papers, and …

Samra Turajlic:
Yeah. So, I think that’s probably a less relevant element of the validation of the two. But basically, what we had was a more complete custom repeat annotated library, which was built by this group working on ERVs. And what we found is that some regions that were previously called as ERVs were either fragmented transcripts, so actually they shouldn’t contribute to ERVs, or they were transcripts actually overlapped with other genes. And some of them included immune genes. So that’s where you could really end up mixing up the signals. So, if I see there is a transcript that’s up regulated, but actually also overlaps a coding region of immune gene, then they will correlate in terms of expression.

Brian Rini:
Yeah.

Samra Turajlic:
But the most important part is that these … most of the ERVs found in the signatures were found to be highly expressed in immune cells. When we looked at a purified immune population. Using independent RNA sequencing datasets. So, there is an opportunity now to subtract these ERVs that are expressed in immune cells so that, what we are left with, are kidney cancer specific ERVs, which we know have previously been reported, and including ERV E, for which there is a TCR transduced T-cell trial open I believe at NIH. And I don’t know if either of you are involved in that. So, it’s not to deny the fact that ERVs are potential antigens in this setting, but the number of ERVs reported in these signatures, most of them are not specific to the kidney cancer cells.

Brian Rini:
Got it.

Samra Turajlic:
They’re coming from the immune environment.

Tom Powles:
Samra, my last question. I was intrigued by one of the individuals on the trial who unfortunately developed brain metastasis. And I think in the postmortem series identified a tumor that was very … what looked very foreign from a kidney cancer. It had very high TMB, it had DDR signatures. What does this tell us about targeting immune refractory kidney cancer for the future?

Samra Turajlic:
Yeah. Thanks Tom. No, we were quite taken aback to see that. And obviously we went back several time to check. So, there were a couple of sites of disease in this patient that had really excess tumor mutational burden. And they had a signature of mismatch repair defect. And as you know, mismatch repair defect is not something that we see commonly in kidney cancer. And most of the mutations that accumulated in kidney cancer simply have this age-related signature, which supports the fact that there isn’t really an exogenous mutagen, and there isn’t a common DNA repair defect somatically acquired in kidney cancer.

But it was interesting … I think what happened in this patient is that they acquired a mismatch repair defect through a mutation. And I think that was purely stochastic. But what that went on to cause is an accumulation of a huge tumor mutational burden.

And it showed us that actually, in the presence of a high level of mutations, and therefore neoantigens, which is what was happening in these sites, you actually see mechanisms of resistance, or immune evasion that are exactly the same as those that described in melanoma, or lung cancer. So, in this patient, there was subsequent loss of beta 2M, which is obviously part of the antigen presenting machinery, through a mutation and a loss of hetero zygosity. So, both copies of the gene were lost.

We saw that a protein level, there was no beta 2M at all. And the thing with beta 2M is that if … it doesn’t matter how many mutations you have, or how many good quality neoantigen that are different to self and so on, if you don’t have beta 2M, you can’t present them to the immune system.

So, it was very interesting to see that when you push the mutational burden up, you kind of see the convergence onto the same mechanisms of immune resistance. Now, I know that there is work, via [inaudible 00:24:08] and others, looking at the role of mismatch repair in kidney cancer. I, I don’t think that this … what we’ve learned from this particular patient is going to apply in very many settings. I mean, it allowed this to explain why there was a divergence in response to treatment, and a hydrogenous response, which is obviously what we often see in patients. And in this case, you know, there was a very elegant path that had led to that.

Brian Rini:
Samra, I have …

Samra Turajlic:
I have a question for you.

Brian Rini:
Well, I have two.

Samra Turajlic:
Well, I have one to follow.

Brian Rini:
One is, and I’ll just ask them both. You know, we commonly use nivo now with ipilimumab with CTLA 4 blockade. So, if you want to just comment on, you know, these data in light of throwing CTLA 4 blockade in there.

Samra Turajlic:
Yeah, of course. Yeah. Yeah, so …

Brian Rini:
And then I have one more quick one. Yeah.

Samra Turajlic:
So yes, of course the standard of care is ipi-nivo. I think it is very important to dissect the effect of these drugs individually. I think we as a field have experience of what it’s like to try and do the biology backwards where you have combination treatments. And I know Tom is a big proponent of triple, quadruple …

Brian Rini:
I think that was a shot at you and I Tom.

Tom Powles:
I’m afraid so. I’m afraid so, Brian. It won’t be the last. It won’t be the last. Just don’t take it personally.

Samra Turajlic:
I think we need to understand them in isolation, and we need to understand them in combination, whether in synergy, or just as additive. So, one thing we did note is that, in the non-responders, there was a higher level of T regs. Now this is a widely debated question in [inaudible 00:25:58] but CTLA 4 is thought to be T reg depleting. And therefore, I think this offers up evidence that in some of these patients that were refractory to nivo, if that was driven by dominance T regs in the tumor microenvironment, then addition of ipilimumab could have helped. I think we do need to do a study like this with ipi-nivo, to be frank. I’m not. I have in me to do another 5 years.

Brian Rini:
You definitely do. Yeah.

Samra Turajlic:
Lewis Au, who’s been a real champion of this study. He’s a clinical research fellow here at the Crick at the Marsden, is at the moment sort of writing a blueprint for how a translational study like this can be conducted. And he’ll be sharing that experience through a protocol manuscript. So, I really do hope that that people will get into doing this more to allows us to build the knowledge.

Brian Rini:
Yeah. Well, I have many more questions, but we’ll save it for another time offline. I just … Congratulations to you and the entire group on this massive effort. I mean, these are great insights that you’re repeatedly giving us in kidney cancer that I think will be the foundation of better care for patients, frankly. So, congratulations.

Samra Turajlic:
Thanks very much. And I really just want to take the opportunity to acknowledge the huge number of people that have contributed to this. And multiple translational studies that have allowed us to go as far as this.

Tom Powles:
Samara, I love the paper. Thanks for joining us as always.

Samra Turajlic:
Thank you. Enjoy the sunshine.

Tom Powles:
We are enjoying the sun. We’re going for a swim right now. Body surfing, I think. Body surfing is the key. I’ll see you.