Case: A younger man with metastatic melanoma and a high PD-L1 expression receives nivolumab monotherapy

DR LUKE: I think in the biomarker space, a real interesting case was a younger patient who had been diagnosed with a melanoma on their forearm. And unfortunately, over the course of about a year and a half, recurred with lung metastases and liver metastases. And this is a person who otherwise was in fully good health and really wanted to say, “I really want to be aggressive, but I also want to live my life and do the things I need to do.”

And so the pertinent question really became, when we considered immunotherapy, should we start with combination immunotherapy, or should we go with monotherapy PD-1? And so the gut reaction is, he just wants to be aggressive, so why don’t we do this? But when we talked through it in clinic, I informed him, “The data are pretty interesting in that in PD-L1-high patients, it really looks like you don’t need to take the combination to get the same kind of benefit.”

So while I don’t always order a PD-L1 assay, in the conversation with this patient who was really looking for maximizing that therapeutic window, as we say, with the maximum benefit/least toxicity, so we did send it. We actually delayed doing his treatment for a few weeks to get it back. And, lo and behold, it came back as PD-L1-positive. So I was pretty strong with him that “I think you should take the monotherapy and not go with combination up front.”

DR LOVE: And what was the assay, specifically, that you did? And how did his specifically read out? Is there a number on it, or high or low?

DR LUKE: The number we got back was an IHC ranking, so 1 through 4-plus ranking.

DR LOVE: And he was what?

DR LUKE: I think he was 3. It was pretty high.

DR LOVE: And how did he respond to the idea of holding back on the combination, just going with PD-1 alone?

DR LUKE: So there was a lot of back and forth. And just the delaying of just in ordering the test he wasn’t really excited about. But when you kind of really explained the amount of differential toxicity that you do see between those regimens, that was the convincing factor. And when I was able to really convince him that, like, “Obviously this is a big problem with metastatic cancer, you’re not going to immediately get sick, and we have time to make the best decision and the most informed decision.” And I think that that cut through a little bit when he was able to understand the data and make a decision.

DR LOVE: So what happened?

DR LUKE: So he had a big response and —

DR LOVE: Really?

DR LUKE: Yes, it was like 60% tumor shrinkage after the first scan.

DR LOVE: What was the agent?

DR LUKE: Nivolumab monotherapy. And so we he’s currently on treatment. I think he’s out at about 8 months now and durable, big response. It’s not a CR, but a lot of these patients can go on a very long time.

DR LOVE: And any tolerability issues at all?

DR LUKE: He had a little bit of rash, dermatitis on the upper chest, but it was really Grade 1, meaning we gave him some cream and that went away. And he’s had a little bit of fatigue, but outside of that, not really.

DR LOVE: Wow, really interesting. Just out of curiosity, what would you say is the likelihood he’s going to progress? How do you see his next few years? What’s his chance of remaining without progression, and what would you do if he did progress?

DR LUKE: So very interesting data coming out of the long-term follow-up from the Phase I trials for PD-1 antibodies really showing that the people who are in deep responses, they tend to have very durable responses even out to 2 years, and even some of them that have stopped. Most of them that have stopped have continued to maintain response. So for a patient like this who really did have a deep response, about maybe 50% to 75% tumor shrinkage, I have no expectation he’ll progress anytime in the immediate future.

Now, should he do so, I think the landscape has changed quite dramatically. So we also sent a BRAF mutational analysis on the patient. And that was actually V600 mutant, so that would be an option. We could consider adding on ipilimumab, either as monotherapy or as combination in that realm. And then, obviously, there’s a lot of research going on, new agents in that space, too.

DR LOVE: What do we know about PD-L1 assays over time? For example, if he progressed, I don't know if we have any data on what happens when you biopsy people. Do they lose their PD-L1 after treatment?

DR LUKE: So it’s not very well understood just yet. And I think it’s something we really need to learn more about. I would emphasize that it’s not at all the case that people absolutely go from PD-L1-positive to -negative. And that’s emphasized by the data for the upcoming LAG-3 antibody, called relatlimab. There’s a Phase III trial being launched now for PD-1-refractory patients, where you add on a LAG-3 antibody. And LAG-3 is another T-cell exhaustion marker, another immunotherapy agent.

So if it was the case that the PD-L1 went totally negative, you would assume that that meant that the immune cells were gone. So adding another immunotherapy like that shouldn’t really work. So we expect to find it to be that it’s still there.

Overview of the immune response to tumors

DR LOVE: Why don’t we go through some of the pathophysiology and clinical research data that we have that’s addressing these issues?

DR LUKE: So in thinking about the immune response to cancer, I think it’s really important to point out that really with these immunotherapy drugs, we’re not treating the tumor, per se, but rather that immune system’s response to the tumor. So in that context, there’s really shared biology that goes into responsiveness to immunotherapy. And I think it’s important to highlight that we have this emerging data set now with many different tumor types, with approvals. And it’s really the case. We need to learn how to maximize the utility of immunotherapy across many different tumor types. And when we think about doing that, the underpinning for this response is really the totality of the immune response.

And some people have described as the cancer immunity cycle and different things and — but this is a slide that really emphasizes all the steps in the immune process that we see, and such that we know that tumor will grow and, as they grow, they shed antigens, or tumor particles. And our natural immune system is always looking around for infections and other things in our body. And those tumor antigens can be recognized by scavenger cells, most commonly known as dendritic cells.

And after those dendritic cells take up the antigens, they then process them and show them on their surface in the context of the MHC, or major histocompatibility complex. Those dendritic cells can then go in the lymph node, where they come in contact with the effector population or naïve T cells. Naïve T cells are specific for 1 antigen. And if it’s the case that a dendritic cell with a tumor antigen comes in contact with a naïve T cell that’s specific for that antigen and its dogmatic secondary signal happens, ligation of B7 and CD28, you can activate these T cells. They can then traffic out into the periphery and find the tumor and do all the things that we want them to do, elaborate cytokines, perforins, granzymes and kill the tumor.

Now, obviously, I always say this is not what happens in patients, because if it was, they wouldn’t have cancer. But this identifies all the different steps where you could intervene with immunotherapy. For the most part, when we talk about immunotherapy to date, we’ve talked about this activation process, which is where CTLA-4 blockade, we think, predominantly has effect, as well as PD-1 blockade, which is really at the site of the tumor. But you could imagine other scenarios here.

You could increase tumor antigenicity. You could increase tumor antigen presentation. There are other immune checkpoints. So there are many other ways with immunotherapy that we could think about intervening.

DR LOVE: Could you elaborate a little bit on what you were saying about the major histocompatibility proteins or complexes?

DR LUKE: Absolutely. So in the context of our immune response, we have baseline genetics that describe the way our immune system processes particles and really speaks to each other. And so that’s the MHC, which many will be aware of in the context of blood transfusions. They have heard about our blood typing. And this is a subsetting of that, and it really goes to how does our immune system communicate with each other?

So there are different HLA alleles here. And that’s important, because that’s become an emerging issue surrounding resistance to immunotherapy, because tumors can change these such that the immune system can’t see the tumor anymore. That’s a loss of Class 1 MHC, is emerging as an important resistance mechanism. And it goes to this interaction such that if the MHC is not there, then the T cell can’t see the tumor in order to kill it.

But, big picture, we really want to try to augment this process in all the steps or try to identify in any individual patient where the problem is in their cancer immunity cycles, such that we could augment that, support it to maybe get a robust response with immunotherapy in any 1 individual patient.

DR LOVE: One thing before we go on, any gut feeling or hypothesis about what fraction of cancers, solid cancers, heme cancers, the fundamental issue is the immune cycle in some way?

DR LUKE: Absolutely. So we have performed large-scale gene expression profiling of tumors to try to get at the sense of are they T-cell inflamed or non-T-cell inflamed? And really there what we’re trying to hint at was, did this activation process take place, and do you get this trafficking of the immune cells? And so far when we look across all cancers, we see about a third of human cancer is highly T-cell inflamed. About a third looks to be maybe, kinda sorta, and about a third is not T-cell inflamed, meaning this process was never started.

And when we think about translating that into individual patients, there’s a tumor level. So different tumor types have different levels of inflammation. And individual patients obviously do as well. So a big question then is, for patients who have a preexistent T-cell response, how can we augment that response? What kind of immune checkpoints could we block to let the immune response be even more active?

Conversely, in patients who have no preexistent immune response, we need to do something to get their immune response going so that perhaps we could then give drugs like anti-PD-1 antibodies, which have demonstrated such efficacy in patients who have an immune response.

DR LOVE: One other question related to this graphic. So I kind of had this understanding that anti-CTLA-4 is somehow more involved in the lymph nodes, and PD-1 is more in the tumor. Is that correct?

DR LUKE: Broadly speaking we think that is correct, with the caveat that we don’t understand the mechanism as well as we maybe act like we do sometimes. But dogmatically, we do think that CTLA-4 plays a fundamental role in the initial activation step of T cells in the lymph node and other places, potentially, as well.

I describe this interaction between the T-cell receptor and an antigen loaded on MHC and this activation. As soon as this happens, CTLA-4 starts to translocate from within the cell to the surface, where it can outcompete the binding of CD28 with B7 and is the natural off switch in our immune system. So the big observation was, what if we block the off switch in the initial activation step? And that is the actual, we think, the mechanism behind CTLA-4. In other words, you keep the immune system on longer so that more T cells would be activated to go out. They don’t naturally shut themselves off.

DR LOVE: And is that also the reason you see — I mean, you have the global feeling there’s more toxicity with anti-CTLA-4?

DR LUKE: Again, yes, we think it does, because — maybe 1 way to say it is, like, the natural immune rheostat is brought down slightly, whereas individual patients normally wouldn’t have autoimmunity. If you bring them down slightly in their natural immune state, all of a sudden you do trigger it, because it’s been lurking under the surface, but not enough. But now since you brought that level down, now it is an issue.

The immunoediting process: Shaping tumor development

DR LUKE: Another important concept in immuno-oncology to understand, really dovetails with the way we’ve thought about targeted therapy over decades, is this immuno-editing hypothesis that was advanced by Bob Schreiber and others. And really what is suggested here is that when tumors initially develop — and it’s probably happening all the time — the immune system is initially able to eliminate those tumors. And over a period of time, however, different genetic events or immune selection can happen in tumors such that the immune system doesn’t completely eliminate it.

And you can think about this as a tumor that pops up, but it’s just naturally eliminated or perhaps is resected, and then in situ or Stage I state, but its equilibrium phase could be after you had maybe a larger lesion that’s removed, maybe a Stage II or III lesion where the tumor is gone but we know, in reality, it’s not actually gone in some patients, because they have recurrence. And that’s this concept of escape, where further genetic instability and immune selection comes forward, but the tumor has figured out how to get outside of the immune system and then to grow at length.

And this is really the same concept that we’ve thought about with targeted therapy, which is, mutations accumulate over time, which can become drivers. And maybe we can go after them.

It’s really important, when thinking about the efficacy of immunotherapy, to think through this conceptual model that can help guide which treatments would be best in combination.

And some colleagues in the field going back, like Tom Gajewski and others, proposed this idea that in a subset of patients, they have a spontaneous immune response. And, in fact, those are the patients that predominantly benefit from immunotherapy.

And for model purposes, you can bucket them into 2 groups, one being those who are non-T-cell inflamed or noninflamed tumors. And these are the people we think probably don’t benefit. And we think that the primary problem here is that the activated CD8 cells are really unable to get into the tumor. So while you can generate an immune response, say, to a vaccine, for reasons that are still being elucidated, those T cells, they don’t come into the tumor. And when we biopsy the tumor, we see the tumor is there and lots of stromal elements like fibroblasts and endothelial cells, but you don’t see things like PD-L1 and IDO and so on and so forth.

And that’s in contrast with the inflamed tumor type, which is where we’ve seen the predominance of the benefit to checkpoint blockade. So here, now the T cells can get in. And through a just mildly complicated biology, they make those cytokines, which actually the tumor will react to, so predominantly the cytokine called interferon gamma. That will upregulate a number of pathways in the tumor for the tumor to try to evade these T cells. And the most famous of those is PD-L1, but others include indoleamine dioxygenase, which is another immune target that’s coming forward, and also a secretion of cytokines which recruit in other kinds of immune cells like T regulatory cells, all of which can dampen the immune response.

But now, going back to our model then, thinking about combination immunotherapies, we’re going to try to stratify what we’re going to do. For the people who have an inflamed phenotype, we want to give multiple inhibitors of tumor inflamed phenotype milieu, so PD-1/IDO or PD-1 and a T-reg depleter.

In converse, the patients who don’t have inflammation, we need to do something first. And maybe those are patients that need chemotherapy with immunotherapy or radiation first, followed by immunotherapy, and so on and so forth.

DR LOVE: For practical purposes, how can — is there a way to clinically — how do you clinically separate inflamed and noninflamed?

DR LUKE: So right now we don’t have a great clinical assay, but broadly speaking, this is what PD-L1 immunohistochemistry attempts to get at. And it’s part of the reason that PD-L1 IHC isn’t really the greatest test, because it only tests for the PD-L1. But in reality, all this other stuff is happening as well. There are assays that are being developed that may or may not come into clinical practice around gene expression profiling to look more broadly in the tumor, but those are still in development.

DR LOVE: What do we know about heterogeneity in terms of when you assess PD-L1 levels or assess inflamed, when you use more sophisticated testing in different locations within the same patient and over time, as the tumor evolves?

DR LUKE: So that’s a very good point. And just like in the context of targeted therapy, we’re learning that if you biopsy different tumors in different parts, you get different answers. The one caveat I would say there, though, is that the immune system is antigen specific. So even if you had an inflamed tumor in 1 part, it’s still possible you could amplify that response to get those other sites inflamed. And that’s really the hope around some of the combination immunotherapies.

Prevalence and importance of PD-L1 as a predictor of response to PD-1 blockade

DR LUKE: So I was mentioning how T cells get into the tumor microenvironment, but they’re then regulated by a lot of things that are happening in that tumor microenvironment. So you can see that there are many other different kinds of immune cells that have PD-1 and PD-L1 interactions, as well as several other kinds of interactions. All of these things are impacting on the T cell.

Really importantly, however, we really think it’s this loop of T cells making interferon gamma in the tumor, noticing that interferon gamma via the interferon gamma receptor and upregulating these immune-evasion molecules, that really regulates whether or not T cells are effective in the tumor microenvironment.

We do know that PD-L1 is an inducible marker and it’s part of the reason that we sometimes have problems. But on the right-hand side, what you can see is immunohistochemical staining for PD-L1. And what you can see is, it’s really a cell-surface marker on tumor cells, trying to keep the immune cells at bay. And predominantly, this is an effect that — what we call the leading edge of the tumor. The tumor, as it grows out, wants to express the PD-L1 that, when the T — comes in contact with the T cells that turn it off. It doesn’t really gain much advantage by doing PD-L1 on the middle of the tumor, because that’s not where the T cells are.

And so that’s one of the issues around the biopsies that can be used. If you actually biopsy different parts of a tumor, you can get different PD-L1 results, and it’s part of this reason, because the physiologic purpose of PD-L1 is to block those activated T cells that are trying to kill the tumor. Those are, generally speaking, on the edge of the tumor, not so much in the very middle. So if you did a perfect biopsy at the center of the tumor, you might actually miss the PD-L1, because that’s not where the tumor needs to express PD-L1.

DR LOVE: What do we know about how conventional anticancer therapy — you were referring to radiation, I mean, chemotherapy, et cetera — affects PD-L1 expression?

DR LUKE: So there’s a lot of literature around this, but it’s still debated. There is this concept of immunological cell death that’s been described for certain chemotherapies and radiation, where, by giving that treatment, you do prime the pump, so to say, meaning that will drive some Type 1 interferon and eventually PD-L1 expression. So a lot of clinical trials are trying to look at this question of giving chemotherapy before immunotherapy or even just doing a biopsy pre and post. I think it’s an area that needs more investigation, but I think it’s hopeful, maybe, that even for patients who don’t have inflamed tumors at baseline, we might be able to do something to manipulate them, to put them in a position, maybe, to benefit long term from immunotherapy.

DR LOVE: Did you, by any chance, go to the presentation, or are you aware of the presentation, looking at amplification of PD-L1, looking at the signature that was seen in Hodgkin, looking at other cancers? Are you familiar with that data?

DR LUKE: Yes. So that’s a really interesting story surrounding amplification of the PD-L1 locus — PD-L1 and PD-L2 locus in solid tumors.

So in the context of Hodgkin lymphoma, one of the first investigations to really look at the reason that PD-1 works in Hodgkin’s emphasized that almost all Hodgkin lymphomas have amplification of the PD-L1/PD-L2 axis. And it was thought that maybe that is the predominant reason why then if you give PD-L1 antibody, it’s really effective.

There’s an emerging story there that says that’s not quite as clean as we used to think and that many of the other immune cells in the area around the tumor are also contributing to the immunosuppression. But it raises this question in solid tumors: Do we have the same phenomenon, where you get amplification?

So there have been large-scale efforts now to try to look in different data sets, in Foundation Medicine’s data set as well as some institutional data sets. And what we’re finding is that in solid tumors, it’s a fairly rare event to get amplification of PD-L1, on the order of about 1% of tumors. And so then the question becomes, what is the functional relevance of that finding?

And interesting data was presented to suggest that in patients they can profile and find this, 6 out of 9 that were presented did have a response. It’s a little hard to say, because there wasn’t more research done in those same patients to know whether or not is that the only thing that they had, or did they have high tumor mutational burden? Did they have high tumor infiltrating lymphocytes in those same patients? But it definitely is worth further investigation.

I would say that I don't know if that’s a biomarker yet for prime time, like for CLIA, used in clinical practice, but I definitely think it’s something that we should be looking for.

DR LOVE: We looked at the poster. And there’s scans of a patient with glioblastoma who had this and who had, clearly, an objective response. I think that, as you mentioned, they reported 6 responses. What do you think the pathophysiology is? How would you explain to a fellow or to me what goes on when you have PD-L1 amplification?

DR LUKE: So I think there’s a couple of different possibilities. It’s possible that PD-L1 amplification, in and of itself, just massively overdrives and suppresses in and of itself. It’s also possible that there’s still an interplay between tumor infiltrating lymphocytes and PD-L1. Dogmatically, we think that the tumor infiltrating lymphocytes have to drive the PD-L1. In other words, if they don’t go there first, then you don’t get PD-L1. But it is possible that genetically, by having amplification, you somehow attract immune cells to suppress them. That’s not really worked out well yet. But I think it’s definitely worth adding this to our profiling of tumors, given the limited biomarkers that we do have. This is definitely one that looks interesting.

Characteristics of T-cell-inflamed tumor microenvironment

DR LOVE: Are you going to go through TILs, or do you want to talk about that now?

DR LUKE: Surrounding this idea of T-cell-inflamed tumor microenvironment, I think it’s important to talk about the functional characteristics of the tumor infiltrating lymphocytes, because we really do think that they sit as the nidus that really generates this whole phenomenon. And while this data is really interesting relevant to checkpoint-blocking antibodies like PD-1, it’s actually quite relevant to other therapies as well, like chemotherapy and others.

And the underlying biology is that in the subset of patients, they have a spontaneous CD8 T-cell response in the tumor, and these are these T-cell-inflamed tumors. But really what’s important is that after the CD8 cells go there, they make the rest of the T-cell inflammation happen. So those interferon gamma-driven pathways, like PD-L1 and indoleamine dioxygenase, get upregulated in these same tumors. But we think it’s really that T cells have to go there first and, in response to the T cells, the tumor upregulates all these phenomenon, because if you look in tumors that do not have CD8 cells, you don’t see these other things.

And that’s really important for immunotherapy, because, obviously, we’re trying to go after the PD-1 — or PD-L1 with the antibody. But this is actually important in other tumor types as well. And so there’s an emerging story in breast cancer and in ovarian cancer. It’s very clear that patients who have this kind of response have better, longer-term survival in general, stage for stage, as well as response to chemotherapy.

DR LOVE: Now, the TILs, are they actually seen histologically?

DR LUKE: Yes. So there are different ways to do it. So here, we show an immunohistochemistry that’s really not at a very high power, but there’s emerging work in the breast cancer space in the pathology groups to really look at higher power, to actually count TILs, because that’s such a powerful prognostic indicator now that’s been incorporated into standard practice.

DR LOVE: And again, do you need to stain these, or you just see them on typical H&E histology?

DR LUKE: I’m not a pathologist. I think you can pick them out somewhat, but obviously you’re augmented if you do stain them for a CD8 stain. That makes them obviously colorful, and then you can see them more easily.

Importantly, in addition to being able to see them under H&E, we can start to do more powerful genomic techniques to try to characterize these phenomenon, because one caveat is that we do know that sometimes you don’t see very many T cells. And so if you really went on absolute number, sometimes you might miss this phenomenon. And so multiple groups have started to use gene expression profiling. And, really, the take-home point here is that you can characterize, based on several different gene signatures that have been published, which patients have this T-cell-inflamed phenotype.

And if you use that data, that’s now been used in retrospective data sets and clinical trials to stratify patient outcomes. And it’s very, very powerful. But it completely makes sense, because the patients who have this T-cell-inflamed phenotype, those patients do very well with PD-1 immunotherapy here, progression-free survival and overall survival, compared to those patients that do not have this inflamed phenotype. But that, of course, is obvious now, because these are the patients that have the immune cells in their tumor. They have PD-L1. They do very well. If you don’t have that, then the drug really isn’t going to work.

DR LOVE: That paradigm, obviously, it’s out there in terms of melanoma. But do we have clinical research data in other cancers in the same way?

DR LUKE: Yes. So the data that’s shown here is really melanoma, but in this same presentation at ASCO, now going back to 2015, there were actually data sets for 4 cancers that were presented. And this is broadly the case in all cancer types, because this is really an immunological phenomenon.

Now, the extent to which this is true does vary slightly. And how it overplays with mutation burden and some other biomarkers that are in development isn’t entirely clear yet. So this isn’t the only biomarker, but this is the most biologically informed biomarker that we have so far, meaning if you don’t have an immune response, you’re not very likely to respond to immunotherapy.

This plot really identifies what we’ve been talking about on a clinical level, of this inflamed or noninflamed tumor groups. And what you can see here are the interferon gamma gene expression score on the bottom, as I discussed, versus the progression-free survival, shown in here on the Y axis. And this is head and neck cancer in the context of treatment with pembrolizumab immunotherapy.

And what you can see is that those patients in the blue box who have a low interferon gamma score have very short progression-free survival. And, of course, that’s the case, whereas patients who have a very high score predominantly have long progression-free survival. And so this really goes to that biology I talked about. Those patients that are inflamed do very well, because they have this interferon gamma score. They benefit from PD-1, whereas those patients who are really noninflamed are really not going to benefit from monotherapy PD-1.

And I would suggest that this could be clinically useful as it comes into practice, because if patients really are very, very low in their interferon gamma score, that really could be a useful biomarker, because it would tell you, you really should not treat them with monotherapy PD-1 immunotherapy. These would be patients in this blue box with a low score that really need either combination or need to consider a different kind of treatment.

DR LOVE: I’m trying to think about the whole paradigm of predictors of response. When you mentioned head and neck cancer, it kind of triggered something I’ve always been curious about. It actually came up last week. We were out here doing a symposium on HCC, which is, from what I can see, I think, if I understand this correctly, in terms of, for example, head and neck, you see, globally, about the same response rate whether it’s the smoking/alcohol phenotype or the nonsmoker/HPV phenotype.

Similarly, when we were talking about various etiologies of hepatocellular cancer, it doesn’t seem like it matters whether it’s even viral or nonviral. Any way to explain that?

DR LUKE: I think the best ideas that we have right now is that those are different kinds of carcinogenesis models, but both of the models turn out to be immunotherapy responsive. And what I mean by that is, we’ve known for — PD-L1 was really discovered in the context of viral infection. And so the HPV version of head and neck cancer has very high PD-L1 expression, because when viruses infect, one of the first mechanisms that goes up is PD-L1.

Flip side of that is, if it’s not HPV-infected lung — head and neck cancer, those others are predominantly then smoking/tobacco/alcohol, high carcinogen associated. So there’s a lot of DNA damage in those tumors that are not HPV. And that can translate into very high mutational burden. And so then there we know that there’s a responsiveness to PD-1 immunotherapy for high mutational burden. So you have these 2 different cohorts of patients, but it turns out both of the underlying etiologies are immunotherapy responsive.

DR LOVE: Same paradigm for HCC. You have NASH. You have viruses. And you have alcoholic.

DR LUKE: Right.

DR LOVE: So again, all roads end up, though, the same place.

DR LUKE: We think that — yes. Paradigm wise, we think that’s generally the case. You can have the hepatitis-associated viruses that can drive PD-L1 expression or, really, these carcinogens that really are causing DNA damage and lead to the disease, again, but that puts you at a higher responsiveness level to immunotherapy.

And on a higher level, we have performed gene expression profiling across the Cancer Genome Atlas to really try to get a sense of this inflamed and noninflamed phenotype across all tumors. And as I already mentioned, what we can see is that in the inflamed category across any different tumor type, we see about a third of tumor types are T-cell inflamed. And that is shown here, about a third. Whereas about a third are intermediate and about a third are noninflamed.

And this was done in the TCGA database, which is not associated with outcomes, but it’s very interesting to see this, because if you broadly ask the question, “What percentage of all patients with cancer respond to a PD-1 antibody, independent of histology,” you get a number around 20% to 30%. And what percent really get no benefit at all? Again, you get a number probably in this sort of range. So we really think this is really true underlying biology of the immune system’s response to cancer independent of histology.

But that being said, we were able to really drill down then by different tumor types and try to look at their inflamed frequency in that context. And what we can see is, when we rank individual samples by what tumor type they came from, the group with those that are more inflamed, being on the right-hand side here in red. And you can see that those are predominantly tumor types that have a high level of responsiveness to PD-1. And that’s the converse of the ones over here on the left, who are very non-T-cell inflamed. And they tend not to respond.

So over here, we see kidney cancer and lung cancer and melanoma and head and neck that are responsive, and they’re T-cell-inflamed, whereas on the left-hand side, the noninflamed tumors include prostate cancer, glioblastoma, uveal melanoma. These are tumor types that we very rarely see responses to anti-PD-1.

DR LOVE: It’s interesting that you see mesothelioma there with a high mutational burden, which is kind of understandable, but yet clinically I don’t get the feeling that — at least at this point — that checkpoint inhibitors have been that effective in mesothelioma. Is that your take?

DR LUKE: I think that’s an evolving story.

So there are registrational trials ongoing for mesothelioma. And the reports to date do describe a RECIST response rate in the 20% to 30% range, which looks a lot like lung cancer. Just the story hasn’t quite come forward as big so far.

DR LOVE: Yes. I’m anxious to learn more about that, because I wonder whether part of it is just the anatomy of mesothelioma and some of the secondary things that start happening by the time you actually diagnose the patient. A lot of tumor bulk — I don't know — a lot of fibrosis. Maybe that is more of an issue than in other cancers.

DR LUKE: So it’s a very good question, because similarly here, we see that pancreas cancer looks to be highly inflamed. And that’s another tumor where we actually honestly don’t usually see responses. So one of these questions then becomes, what are the other tumor-specific factors that could be influencing whether or not there’s response to immunotherapy? So fibrosis is one that potentially could be an issue.

That being said, I do want to point out a great paper that was recently published in Nature looking at desmoplastic melanomas. And these are melanomas that tend to be much more fibrous than standard superficial spreading melanomas. And they tend to be much more locally recurrent, but they also tend to have a high mutation burden. If you look in the histology, they’re highly fibrous. And so one of the questions was this same argument: Will they respond?

And it turns out response rate in those tumors, 70%.

DR LOVE: Wow!

DR LUKE: So higher than normal in melanoma. And we think the rationale there is very high mutational density and a lot of T cells that have actually been attracted to the tumor. So even though you would postulate that maybe they can’t get in, with PD-1 immunotherapy, you actually do get a very high response rate.

DR LOVE: That’s really interesting, though, your pointing out this about pancreatic cancer. Is it your thought that maybe, theoretically, checkpoint inhibitors could be effective in pancreatic cancer?

DR LUKE: I think, theoretically, but we’ve got to figure out what are the tumor-specific issues that are limiting our ability to do that so far. And whether it’s T-cell trafficking or entry due to physical barriers like fibrosis or whether or not it’s that there’s a slightly different immune response in pancreas cancer — pancreas cancer tends to be associated with a high infiltration of myeloid cells. And those can be immunosuppressive, so that maybe the other combinations of checkpoint inhibitors specific to a tumor type may also be necessary.

Association of mutational load and T-cell receptor diversity
 
DR LUKE: So this mutational load is also an important biomarker that’s really evolving out and is associated with the inflamed score, but it’s actually slightly different. Here what we’re plotting is the mutational density, as published by many groups from the TCGA on the upper panel here. And this canonical increasing mutational density has been previously described, so melanoma and lung cancers and bladder cancer, I mean, a very high mutational load.

So one of the questions then is, does just that, in and of itself, drive a neoantigen burden or new immune system molecules that could be recognized? And on the bottom, what’s plotted is the diversity of the T-cell receptor, which is commonly associated with whether or not you’re having an immune-recognition pattern. And what you can see is that, broadly speaking, they actually tracked fairly well. And that really does make sense. It suggests that as you get more mutations, you get more immune recognition.

And those tumor types tend to overlap as you go to higher degrees. And in this unification plot you can see, again, the tumors that are most likely to be recognized by the immune system are out here. So mutational load has obviously come forward with MSI and other sorts of mutations. T-cell diversity is another biomarker that’s really being evaluated in clinical trials right now, although how it exactly overlaps and would be useful clinically is still an evolving process.

DR LOVE: What exactly is that, TCR diversity?

DR LUKE: So the number of T-cell receptors that are available for different kinds of antigens can be associated with a more or less productive immune response. So if you induce a very broad immune response, you’ll get greater T-cell diversity. And so that’s been a biomarker that correlates well with whether or not you get responses to checkpoint blockade.

One point to make, however, is, we’ve talked about the inflamed tumor signature or inflamed phenotype. Again, I mentioned these tumor types. That’s here, and they’re in red, the hot versus cold. A question that comes up is, how does that overlay with the mutational burden? Are they the same phenomenon?

And we were one of the first groups to point out that really, they’re not the same biomarker. And so what we can see is that even though we have increasing frequency of the inflamed phenotype, when you look at the corresponding number of mutations in the same samples, we don’t see that same trend going this way. And, in fact, when we look between what we call hot and cold tumors, we actually see that the number of mutations is actually often fairly similar.

So these are getting at 2 different biological processes that can be along the same lines but are not necessarily the same thing, which is why I think it is useful to have both a PD-L1 test to look at whether or not the tumor’s inflamed but to also have the mutational burden be assessed in those patients.

When we tried to ask the question of, like, why is there no difference between these, we asked, looking again in TCGA, about the different — in the nonsynonymous mutations, so different mutational events. And when we looked across and used our T-cell signature, we saw there statistically was no difference in the number of mutations between inflamed and noninflamed tumors.

We even went so far as to try to look at which epitopes or candidate binding peptides were available in different kinds of tumors; so from inflamed and noninflamed tumors, which actual antigens were there. We actually synthesized those antigens and exposed them to T cells. And what we observed was that the responsiveness was actually the same, suggesting it’s not necessarily just the number but the quality of the interaction of the immune cells with the mutational-derived epitopes probably matters as well.

DR LOVE: In non-small cell lung cancer what we’ve been hearing from investigators for about the past year is separating out patients with tumor mutations, EGFR, for example, from wild type in terms of immune responsiveness. They talk about the fact these are younger patients. They’re nonsmokers, quote, lower mutational burden. What about this paradigm in non-small cell lung cancer in terms of targetable-type cancers versus nontargetable, or wild type, in terms of this inflamed phenotype?

DR LUKE: Yes. So this is a really important clinical point to make, where the genetics hits the clinic. So what we observe is the underlying etiology for why these different kinds of lung cancers — and, actually, this applies to other cancers, as well — why they develop is very different. And that correlates with clinical. So we know that who gets EGFR and ALK mutations are nonsmokers. So the underlying etiology is not carcinogenesis. When we think about how does that dovetail with immunotherapy, we know that the carcinogen-associated tumors are more likely to be inflamed, and they’re more likely to have mutational load.

So when it comes to considering your options, we see that EGFR-mutant and ALK-translocated tumors in lung cancer tend not to be T-cell inflamed and tend to have low mutational burden, which we think is the reason why they don’t respond the same way. And given the effectiveness of targeted therapies for those diseases, we prioritize giving them targeted therapy before considering whether or not to give checkpoint blockade.

DR LOVE: That’s a practical clinical point, but I’m thinking more in terms of the implications in terms of future research, et cetera. And if you think about it, I mean, why are these people developing EGFR-mutated lung cancer? Is it just a mutation that occurs spontaneously, or could it be that these people actually have primary immune deficiencies or problems?

DR LUKE: Yes. So I think that’s a great insight. And it’s probably a little bit of both. But we do think that we’re learning more, that certain driver oncogenes are actually associated with immunosuppression, meaning we’ve always thought of them as growth drivers, but, in reality, they have downstream immunosuppressive effects as well, or, just as you suggest, maybe one of the primary things that EGFR activation does is actually lead to suppression of antigen presentation and, therefore, you don’t get a T-cell-inflamed phenotype.

And that would suggest that maybe we should consider paradigms of clinical trials where you combine a targeted inhibitor to allow for the immune system to be more effective, then come back in with immunotherapy.

DR LOVE: Or some other kind of new strategy. But it also gets into — sometimes when I look across oncology and cancer research, it kind of feels like there’s kind of an antimicrobial model there, maybe coming out of the history with chemotherapy. And I wonder if that’s even fundamentally correct.

DR LUKE: And when you say “microbial,” do you mean a —

DR LOVE: That there’s a mutation that — it has a unique ability to cause a harmful effect. And you have to kill it, so to speak, as opposed to looking at the host.

DR LUKE: I think there is something to be said for that, because some of these pathways to get activated as driver oncogenes, I think, have what we call pleiotropic effects. They can have many different things that they do, make it grow, but they can also suppress the immune system. So I think, to some degree, that’s true. I would not discount, however, how important thinking about the host is as well, meaning that our body’s natural immunity is also an important element to this, because we know that there are definitely associations of people who have suppressed immune systems in the context of bone marrow transplants, HIV status and so on and so forth who are more likely to get cancer in the first place.

T-cell-inflamed versus noninflamed tumor microenvironments

DR LOVE: I have to ask you, because I ask everybody whether they want to say anything about the whole idea of the microbiome in immunotherapy and cancer.

DR LUKE: We definitely think that that’s important. And in our group at the University of Chicago, we’ve really started to think about the response to the immune system to cancer in really 3 different buckets.

One is what we’ve talked about already with the somatic differences in terms of mutational load and oncogenic pathways that would be driving resistance. But we think that two more very important but less investigated areas so far are the germline genetic differences of the host, meaning polymorphisms and immune regulatory genes.

And we know from our colleagues in rheumatology, this is a very important issue. In other words, your baseline polymorphisms in your immune regulatory subject can really put you at risk for lupus or rheumatoid arthritis or Type 1 diabetes. We think it’s very likely the case that some element of that will be true in cancer as well.

And the third area then is environmental differences, meaning, what have we been exposed to in our lives that would be different, say, from you versus me? And there, to date, we’ve been the most interested in the commensal microbiome of the GI tract, meaning which bugs are in our gut may help drive whether or not we develop a productive immune system more globally in our bodies. And again, that space, there’s emerging data in many different areas of medicine to suggest that what bacteria are there can influence on whether or not we get Alzheimer’s disease, whether or not we get rheumatological problems.

And, in fact, our group and others were one of the first now to publish on whether or not you respond to a PD-1 antibody can be mitigated by which bacteria are actually in your gut.

DR LOVE: And — because I was talking with Keith Flaherty about this. And he brought up something I hadn’t heard before, which is the idea of what is going on normally in the gut? And also, he brought up the liver in terms of the fact that there’s bacteria there that the body needs to protect itself against. And his thinking was that maybe that’s part of the reason you see toxicity in the liver and colon. Any thoughts about this interface with bacteria?

DR LUKE: So I think that’s absolutely the case, that we are, as human beings, the collection of all of the cells in our body, some of which are not actually ourselves. They’re the commensal microbiome or micro-organisms that naturally live within us. And our interaction of our immune system can be both positive and negative. So you could make the argument that if bad bacteria are there and you give a drug, you might get a bad outcome.

Conversely, though, you could make the opposite argument to say, if the good bacteria aren’t there either, then you won’t have a good outcome. So it isn’t necessarily just the negative. It’s that we need to understand better what’s our natural physiology and homeostasis? What bacteria do we want to be there? It’s not the case that we don’t want bacteria, because we naturally have them. We have to have them. We want to know which ones give us the best chance to have a primed immune response as opposed to a suppressed immune response.

DR LOVE: What are your thoughts in terms of what the mechanism is, why various different bacteria in the gut, et cetera, might influence response to checkpoint inhibitors?

DR LUKE: So I think there are 2 categories. And I would say we don’t know the answer yet. But one would be, do the bacteria or the endothelium make a molecule that then is in the systemic circulation that could impact on immune cells somewhere else? Or are immune cells at the gut lumen activated that then traffic somewhere else and do something? And to date, we don’t know the answer, but our better hypotheses are on the latter, meaning that we think that there are dendritic cells, those scavenger cells that are noticing immune signals, who are resident at the gut lumen, who can be more or less activated.

And they can then go to the lymph node, they can go to other organs, the spleen, and they can tell everybody else, “Hey, we’re in a good situation for immune activation. Be productive,” or, “Hey, we got bad bacteria here. We’re immunosuppressed. We’re not doing what we need to do.” And there we think the future could be we can figure out which bacteria actually make that difference and maybe start to think about an intervention. Could it be fecal transplant? Could it even be a probiotic? Now, it’s not to say that we’re ready to do that already, but it’s not impossible to think of a world in the short-term future where we could do that.

DR LOVE: Just a practical question: If a patient who you want to put on a checkpoint inhibitor or is on a checkpoint inhibitor, “Can I take a probiotic,” what do you say?

DR LUKE: I tell them, “Yes, you can take one, but I can’t give you a recommendation on which one to take.” And we’ve published some high-profile papers on this to date in Science and Nature, but even such, I can’t tell my patients that, “You should do this.” We are going to launch a clinical trial of supplying a specific bacteria to see if that augments checkpoint blockade, but that is a clinical trial and that’s not ready for patients to do yet.

DR LOVE: I’ve heard people say that you really can’t change, like, say, the gut microbiome. You give antibiotics, you give probiotics, whatever, it changes temporarily, but then it goes back to where it was. Is that the case?

DR LUKE: It varies by model. So in some tumor models, you can change it and it will stay. In others, you need to persistently keep it changed. And this is one of the questions about how would you intervene? If you did a fecal transplant and then a year later, would it just go back the way it was, or do you need to chronically supply, in the form of a probiotic, a bacteria that you want to be there? And I think we don’t know the answer to that yet, but it’s very much the case in the models. If you keep giving the probiotic, you can keep the bacteria there. In other words, it doesn’t happen so fast that you lose it. You can repopulate on a prospective level.

Interaction and independent predictive values of mutational load and gene-expression profile in immunotherapy response

DR LUKE: We’ve talked throughout our discussion around these inflamed signatures and mutations. And one of the big questions that still exists in the field is, how do these things overlap? In other words, are the same patients the ones that are inflamed as the people who have high mutational burden? And this is great work that was done in the context of the pembrolizumab development program, which really highlights that, in fact, while there’s a correlation, it’s not perfect, meaning some patients can either be highly inflamed and low mutation, or vice versa, low mutation and high.

And that’s emphasized in this panel here on the left, where this gene expression profiling of these 18 genes is compared with the whole exome sequencing on the Y axis. And what you see is there’s a general trend going up, but these dots don’t all overlap, meaning there are patients, again, who it goes along this continuum, but you can be highly inflamed with a low mutation or not inflamed and high mutation.

And I think, again, going to clinical practice, that’s a really important point, because when you get your patient in front of you, you kind of want to know either one of them. Because I think either one of them could be a good trigger to use immunotherapy.

If your patient in front of you both had a, like, no PD-L1 staining and a low mutational burden, depending on what the other options for treatment would be, I don’t know that immunotherapy would be my go-to thing in that situation. And that is an area where biomarkers are already coming into our practice.

DR LOVE: Now, these type of readouts, do you get that when you do some of the multiplex assays that are out there, FoundationOne, Caris, et cetera?

DR LUKE: So you do now. Foundation and, I think, Caris as well. So initially they just gave you the different oncogenes that we were looking for, but now they have incorporated that, a PD-L1 assay using an antibody that’s a clone of some of the other commercially available assays. And it will give you a PD-L1 score.

In addition, they give you a general call across the genome to say, “Do you have a lot of mutations or not very many?” And that can be useful, because even if you don’t find the drivers you’re looking for, even if you don’t find PD-L1, it is sometimes still the case that they’ll say, “High mutational density,” which would be a suggestion that considering checkpoint blockade would be reasonable.

Traditional paradigms for FDA tumor-specific approvals in oncology

DR LE: Today I’m going to be talking about integrating immunotherapy into the treatment of patients with MSI-high solid tumors from any histology. And so the objective is really to review the studies that support PD-1 inhibitor in mismatch repair-deficient or MSI-high cancers.

And so when we think about traditional paradigms for approval in oncology in the past — and still, currently, oftentimes we base it on a specific tumor type, a stage of disease and a line of therapy. And I’m a GI medical oncologist, so a perfect example of that would be first-line metastatic pancreas cancer using gemcitabine and nab paclitaxel.

We’ve also had approvals of a biomarker within a specific tumor type. And again, another GI example would be first-line metastatic HER2-positive gastric cancer using trastuzumab with a platinum-based regimen. And, oftentimes, these approvals are based on randomized controlled trials, again, the example being gem/nab paclitaxel versus gemcitabine in metastatic pancreatic cancer.

So what is tissue-agnostic development? It’s actually taking a biomarker across histologies and trying to develop that as a predictive biomarker regardless of the origin of the cancer. And so to use this approach, you really have to think about whether it makes sense scientifically. Originally, there was the idea that we could target BRAF across multiple tumor types. However, that can be heterogeneous, as we’ve seen with BRAF-mutated melanoma versus BRAF-mutated colon cancer — have to be targeted quite differently.

We have to think about whether we can study all these diseases within 1 cohort, if they are prognostically different. Can we really study patients with prostate cancer and pancreas cancer in the same study? And how would we measure treatment effects across different tumor types? Likely, we’d have to look at actual objective response across multiple tumor types.

We would have to think about whether it would actually slow development. If it’s a common tumor type, perhaps it would make more sense to target those common tumor types. But it might make more sense if these are rare patient populations. And the incidence has to be kind of, at least, evident across multiple tumor types to develop this approach. We would also want to know if there was an unmet medical need. And the response would really have to be clinically meaningful. It would have to affect a good number of patients and the responses would have to be durable, so that the signal wouldn’t be lost.

We also have to consider who’s going to enroll these patients. Many of us at academic centers are very disease specific. And should we include pediatric patients? If it makes sense biologically in adult patients, should we also include pediatric patients? And then there’s the issue of the companion diagnostic. Is it really validated across multiple tumor types?

FDA-approved immune checkpoint inhibitors for the treatment of solid tumors

DR LE: So I’m going to review some of the studies that went behind the approval for pembrolizumab for MSI-high or mismatch repair-deficient cancers, which is the first tissue-agnostic approval. This approval came in May of 2017.

You can see the first checkpoint inhibitor, ipilimumab, was approved in March 2011, followed by pembrolizumab 3 years later. And there have been multiple approvals, all disease specific, with our approval in May of 2017. Subsequently in July 2017, nivolumab was also approved for MSI-high colorectal cancer.

The most commonly thought of tumor when we’re thinking about mismatch repair deficiency is colorectal cancer. This recurs from genetic and epigenetic defects in mismatch repair proteins. This results in microsatellite instability. This can occur from both germline or sporadic mutations in the mismatch repair proteins. It can also occur from epigenetic silencing from MLH1 hypermethylation in the sporadic cases.

It occurs in approximately 15% of colorectal cancer across all stages, but as you can see, as you get to more advanced-stage disease, it’s only 4% in metastatic patients.

Evaluating microsatellite instability (MSI), mutational burden and baseline PD-L1 expression in colorectal cancer and other gastrointestinal cancers

DR LE:  In the studies where we selected patients, the tumors were selected based on either IHC or PCR. This is an example of an MLH1- and PMS2-deficient tumor. It’s proficient in MSH6 and MSH2. And oftentimes, MLH1 and PMS2 are deficient together, because they function as heterodimers.

This is an example of a PCR-based test looking for microsatellite instability. This is looking at normal compared to tumor. And you can see that there are shifts in the microsatellites where errors occur in repetitive DNA sequences.

So the idea behind the study is, as we were getting evidence of PD-1 inhibition effects in multiple tumor types, it was noted that many of the tumor types were high mutation burden tumors, such as lung cancer and melanoma.

If you look at colorectal cancer, there’s a patient population that has hypermutated colorectal cancer with thousands of mutations, compared to the majority of microsatellite-stable colorectal cancer.

And you can see here, looking at the genome atlas, looking from right to left, that there are definitely a range of mutation frequencies in a human colorectal cancer, with most of the high mutation burden tumors being microsatellite unstable.

So the idea is that mismatch repair deficiency results in the presentations of neoantigens due to these repair deficiencies. These proteins are then expressed on the cell surface. There are T cells in the tumor microenvironment that are then shut off by PD-L1 expression either on tumor cells or immune-infiltrating cells and that the PD-1 inhibitors could take the brakes off of these cells.

So work by Nico Llosa at Hopkins has shown that you can see with the MSI-high tumors at top compared to the microsatellite-stable tumors at the bottom that in these tumors, the PD-L1 expression is mostly actually on tumor-infiltrating myeloid cells and not on the tumor themselves.

So the idea when we first did the original study, which we refer to as KEYNOTE-016, was that these mutations would encode proteins that could be recognized and targeted by the immune system and that the average number of mutations in an average tumor would be dozens, whereas mismatch repair-deficient tumors would harbor thousands. We knew that these tumors across multiple tumor types were actually infiltrated with T cells and that immune augmentation with PD-1 blockade would be highly effective in these tumors.

If this hypothesis were true, then histology should not actually matter. And so when we designed our original study, we thought it was important, actually, to have noncolorectal patients.

This is an example — actually, these patients are from our study cohort with baseline PD-L1 and CD8 T-cell infiltration. The top row is a patient with mismatch repair-deficient colorectal cancer. The middle row is a mismatch repair-proficient colorectal cancer. And the bottom panel is actually endometrial cancer with mismatch repair deficiency.

And you can see the nice PD-L1 and CD8 T-cell expression at the tumor, a normal invasive front, as well as CD8 T-cell infiltration in the parenchyma of the mismatch repair-deficient tumors.

So again, the idea is, these tumors have high mutational load, PD-L1 expression and CD8 T-cell infiltration.

And the idea that we could target multiple tumor types, we really had to look whether there was enough tumors of other types that had MSI to be able actually to enroll to our third cohort. And you can see it depends on the assay, stage of disease, but you can see it’s fairly common in the sense that you’ll see it in multiple tumor types, although at low frequency.

One of the obvious next diseases to consider would be endometrial cancer. Endometrial cancer, this is looking at, again, the sequencing. And you can see to the far left there’s the ultramutated POLE-mutated tumors, and then a higher percentage of microsatellite-unstable tumors in endometrial cancer.

Investigators have also done neoantigen prediction in T-cell infiltration looking at the hypermutated cancers. And if you look at the POLE and MSI, they have much higher neoantigen load, CD3 and CD8 T-cell infiltration into the hypermutated tumors compared to the microsatellite-stable tumors.

This is looking at microsatellite-stable on the top row, MSI-high in the middle row and POLE-mutated tumors looking at CD8 T-cell infiltration in these tumors. And you can see the difference between the tumors, again, the idea that this could happen across multiple tumor types.

And we’ve also looked at the data looking at gastric cancer, where there are subtypes of cancers, perhaps the EBV-associated tumors or the MSI-high tumors that could potentially be some of the good responders in some of these studies as well.

This is looking at PD-L1 and PD-L2 expression in molecular subtypes of gastric cancer. In the EBV-associated tumors, you see both PD-L1 and PD-L2 expression. In the microsatellite-unstable tumors, again, you see the PD-L1 expression mostly on the immune infiltrating cells in gastric cancer.

This is a data set where they looked at over 12,000 tumors and saw that in a majority of tumor types, there is at least some level of microsatellite instability. And again, the more common ones being the endometrial and the GI cancers. This, the bar coding, the blue and the red, refer to whether these are earlier- or late-stage patients, because it’s a mix of tumor types and stages. And again, it’s more frequent in early-stage disease.

So this is another slide, again looking at NGS to look for microsatellite instability. Now, there are algorithms where they actually look at thousands of microsatellites when they’re doing sequencing and are able to detect microsatellite instability using these new panels. And this is where they used a particular algorithm and saw that there were 3.8% of all cancers with microsatellite instability in this particular cohort.

DR LOVE: In the adjuvant situation, what’s the percent MSI compared to metastatic in colon?

DR LE: Yes. So in the adjuvant situation, it’s more like 15%. And metastatic disease in colon, it’s more like 4%. It’s variable across the actual disease types. If you look at endometrial cancer, it’s more prevalent in both advanced-stage and early-stage disease compared to colon cancer. But if you look at the gastric data, I mean, sometimes they say it’s up to 30%. But if you actually look at the metastatic patients, it’s probably more in the 8% range. And so I think it is common across some tumor types but not all tumor types.

DR LOVE: And what’s the reason that you see it more commonly in earlier-stage disease?

DR LE: Yes. I mean, I don’t think we know for sure why we see it more in early-stage disease. But again, I think the idea perhaps is that because the immune system is able to kind of keep these diseases in check that you’re going to see it less commonly in the advanced-disease setting. And then once it hits the advanced-disease setting, the immune system has been turned off. And so perhaps then the patients might do actually a little bit worse.

Pathophysiology of MSI-high tumors

DR LOVE: Can you talk more about what’s the normal function and then from a functional point of view, pathophysiologically, what’s going on when you have an MSI-high tumor?

DR LE: Yes. So there’s a lot of confusion about the words “mismatch repair deficiency” and “MSI.” So mismatch repair deficiency is exactly that, a deficiency in the mismatch repair proteins. So that’s MSH2, MSH6, MLH1 and PMS2. And these are DNA repair proteins. So if you’re deficient in one of them, then you are mismatch repair deficient. And that’s why a simple test like IHC, just looking for missing proteins. So when we’re looking at those test results, we’re actually looking for a loss of the protein. So the people are a little bit confused when they see “positive.” We’re looking for the loss of the protein.

DR LOVE: How does that tie into things like BRCA? Is that, like, completely different, or is it —

DR LE: It is. There are different types of repair enzymes. And they work in different parts of the repair. So for mismatch repair, these deficiencies occur in these areas called microsatellites. And we usually see frame-shift mutations, whereas the BRCA is more homologous recombination. So it’s a little bit different in terms of the areas where it is actually repairing. Mismatch repair-deficient tumors in general have a lot more mutations because of the type of repair that’s occurring in the deficiencies, whereas the BRCA patients, they’re — again, people are looking at whether the BRCA patients could also be sensitive to immunotherapy. But I think the mechanism there is not going to be really due to mutation burden.

DR LOVE: Interesting. And how many mismatch repair proteins are there?

DR LE: There’s 4 classic ones. Those are the ones they test for by IHC.

DR LOVE: So 4 proteins. And when you have mutations, is it just that there’s less protein, or is the protein abnormal?

DR LE: Yes. So it’s variable. And that’s why sometimes the IHC actually can miss patients, because they can have intact protein. So when you’re doing the assay, the patients look proficient, but the protein’s not working. And so that’s where microsatellite instability testing, PCR testing, might pick up some of those patients, because they’re still functionally mismatch repair deficient even though they have the presence of the protein. And so MSI testing is almost more of a functional test, because you’re looking at the instability in the microsatellites, the shifts. The product of the protein is not working properly. So sometimes we do test for both, not always. But you can miss some with IHC.

DR LOVE: And again, drawing it out a little more pathophysiologically, okay, so you have less effective or ineffective DNA repair. You get more mutations. And then how does that tie into the PD-1 axis?

DR LE: Yes. So the idea behind having a lot of mutations that are actually neoantigens is that the tumor looks more foreign. And so the more foreign something looks, the more likely you’re going to have a T cell that could recognize it, because it hasn’t been deleted from the system, which is a normal kind of process that you have so that you don’t have autoimmunity. And so these T cells can exist, because these are neoantigens, antigens that weren’t present at baseline.

And so we know that they are there and are circulating or are actually sitting in the T cells that you can see, just by doing simple IHC staining, but they’re turned off.

We don’t know exactly what turns off and why we have a PD-1 and PD-L1 upregulation in the tumor microenvironment. We know they’re expressed on these immune infiltrating cells. We don’t know what drives those cells into the tumor microenvironment, but we know that they’re there.

DR LOVE: So — and then when you bring in a PD-1 inhibitor, what happens? How does that change the balance of things?

DR LE: Yes. So what happens is, the PD-1 blocks the PD-L1, so it actually — PD-L1 binds to PD-1 on the T cell and turns it off. So you’re blocking the negative interaction. And so now you’re able to accelerate a response for the T cell.

DR LOVE: They’re presenting these data on the amplification that’s seen with Hodgkin lymphoma that they think is related, I guess, to even the histology and the benefits that you see for checkpoint inhibitors and the idea — they saw it in all kinds of cancers, lung, breast, colon, head and neck. There was a basal cell that had an objective response, metastatic basal cell, incidentally, to a PD-1 inhibitor. What do you think this means, not just in terms of clinical decision-making — okay. Fine. We’re going to go out and find these rare patients who have this, but more globally as to how we can take the lessons that we’re learning about this pathology and get greater efficacy and better, more effective drug development. How can we take these as clues to deal with more than just the 3% of patients, or whatever it might be? What are your thoughts about that?

DR LE: Yes. I mean, I think it’s difficult when we see some activity. We don’t have good biomarkers a lot of times to figure out who it is is responding. I think an example would be gastric cancer. There is some activity of the agents even in the microsatellite-stable tumors. We really don’t know if they’re the EBV associated, because no one’s really looked at that. And I think what is important is in the MSI story, when we find an MSI gastric patient and we treat them, the response is quite durable. And so I think when we find these patients that are going to really respond, then it kind of is a way to kind of use our resources towards patients that are really going to benefit.

So I think that the story with the PD-L1 amplification — is it that they looked at — I think it’s going to be the same story that even if it’s infrequent and now we have these assays that are much easier to obtain and the cost of those assays are also going down, that we should develop and direct our resources toward those patients, even though they’re small in number. But, actually, if you take them across multiple tumor types, it actually ends up being a lot of patients — the same with the MSI story. I think it will be something similar for that.

DR LOVE: And. I mean, any thoughts about what fraction of cancer in general is a function of some kind of immune deficit or immune problem? Do you think that is the essential problem, or is it kind of a minor part of a cancer that we’re just tapping into here?

DR LE: I think that for a subset of patients that it is one of the major issues, obviously for, like, the MSI story. And then if you look across multiple tumor types, usually the response rate is 10%, 15%, 20%. So I think in those patients, it certainly plays a large role.

I think outside of those patients, I think it still plays a role in many — not always, because some of these are driven by different mechanisms, obviously — that there probably is a component of the immune system kind of changing the pace of the disease in terms of the equilibrium of the disease. But single-agent therapy is just not going to work. And there’s obviously tons of studies trying to figure out how we can break that equilibrium towards antitumor immunity.

DR LOVE: I guess — and I don’t want to get too far off on a tangent, but sometimes when I look at oncology, targeted therapy, chemotherapy, it almost seems like an antimicrobial paradigm that kind of to me hasn’t worked too well. And I wonder if it’s just we haven’t the right drugs or it’s just the whole completely wrong strategy. Obviously immunotherapy is a completely different strategy. And again, what hints do we have that maybe a much higher fraction of patients or cancers, this might be a critical issue?

DR LE: Yes. I mean, again, I think that when we are treating patients, I think that we want the immunotherapy to be applicable to everyone. But I treat a lot of pancreas cancer, and we’re doing a lot of work to try to change the tumor microenvironment, get T cells in the tumor microenvironment, inhibit regulatory cells, inhibit myeloid cells in the tumor microenvironment. But we have a long way to go. And I don't know, ultimately, if it will apply to those patients.

And so that — again, I think, right now we’re kind of aiming at the patients where we have clues that it’s going to work, like these patients with T-cell infiltration or RNA expression showing inflammatory signature and such. But for the vast majority of patients, I’m not 100% sure if we will see responses. But, as you know, in the microsatellite-stable colon cancer, there are a couple of studies out there now showing 18%, 20% response rates. So I think it’s just going to take us time to try to figure out how to make it more applicable.

Integrating immunotherapy into the treatment of MSI-high solid tumors

DR LE: This is kind of the story behind the PD-1 blockade and mismatch repair deficiency. And we originally presented updated data in 2016 at ASCO looking at our 3 cohorts of patients.

The first two were colorectal cancer patients. We had 28 patients with mismatch repair-deficient colon cancer. We had 25 patients with mismatch repair-proficient colorectal cancer. And we had 30 patients with mismatch repair-deficient noncolorectal cancer.

The colorectal cancer patients did have to have 2 or more prior therapies, and the noncolorectal cancer patients had to have 1 or more prior therapy.

And the reason I like this waterfall plot, again, is just the color coding that it is pretty clear that PD-1 inhibitor will only work in the mismatch repair-deficient colorectal cancer patients.

And this is the spider plot just looking at the kinetics of response. You can see that some patients respond immediately on their first scan, although the first scan in this study was actually at 12 weeks, and others just kind of have a slow regression over time.

This is looking at the Cohort C patients.

And you can see multiple color coding, showing that we see responses across multiple tumor types and many of them having complete responses, and again, those responses being durable. But I’ll show you updated data from 86 patients from this study.

So we updated our data on 40 patients with colorectal cancer and 46 patients with noncolorectal cancer, so a total of 86 patients.

And you can see there, out of multiple tumor types representative, we treated patients across 12 tumor types. The majority of the patients had 2 or more prior therapies. Approximately 45%, 46% of the patients had Lynch syndrome.

You can see that our response rate in the colorectal cohort was approximately 50%, as well as in the noncolorectal cohort. And the complete response rate was 12% in the colorectal cohort and almost 30% in the noncolorectal cohort as well.

So this is looking at treatment response by tumor type. Most clinicians really want to know kind of the granular data as they’re looking at a patient in front of them. And we really did see responses across multiple tumor types. However, the N for many of them are quite small. So you might see 0% or 100%, and that usually represents an N of 1.

This is the pie chart and the waterfall, again, looking at the 12 different tumor types. And you can see, over to the far right of that waterfall, many of the responses are quite deep, with 100% response in their target lesions. The purpose of the figure on the right was really just to show that you can have a response at 20 weeks, but some of these really take a long time to evolve. But in general, we didn’t see pseudoprogression. Patients just had a just slow response over time, so it took a while for some patients to develop their PR or CR.

The one thing to note about these swimmer plots for our patients is that in many of the pembrolizumab studies, patients actually stopped getting treated at 2 years, and then we follow them, which is different than in the nivolumab studies. And so you can see those triangles start to all end at the 2-year mark, but we’re still continuing to follow patients. And that top bar is a patient who reached his 4-year anniversary last December with 2 years of it being off of therapy. So many of the responses are still ongoing off therapy. And on the right you see the PFS and OS curves, seeing nice plateaus.

What we don’t know yet and will take longer follow-up is what happens out 4 years, 5 years, 6 years, but at least up to the 3-year point, it’s nice to see the plateau.

DR LOVE: When you look at these waterfall plots and, even though it’s really impressive you see these responses, some patients progress. And I don't know whether you’ve had people who just go right through therapy. I’m sure there must be — maybe some, based on those. What do you think is going on pathophysiologically, why they don’t respond?

DR LE: Yes. So, actually, we kind of address those with the next 2 slides.

We did try to look at whether there was something about the mutation burden or the PD-L1 expression or the CD8 T-cell infiltration. I don’t have an obvious answer. These are an example of what you see. In the blue is a primary tumor. And in the salmon color is a mutation — actually burden in the metastases in a progression, metastases.

So these 2 individuals actually had brain metastases fairly early into therapy, maybe at 4 and 7 months. And so technically they were called progression at that point. But we continued to treat them, because the protocol did allow, because their systemic disease control was actually quite good. But we were able to resect the brain metastases. If you look at the actual number of mutations between the primary and metastases, it’s not a numbers issue. It’s probably the quality of the actual antigens. And there might be other mechanisms.

And so the other thing we did was, we looked at sequencing when we could of the primary versus the recurrent tumors. And one of the things that people have been interested in is at antigen presentation. Are these antigens actually being seen? We looked at beta-2 microglobulin. And it’s interesting, though, that baseline beta-2 microglobulin alterations actually didn’t predict response. So patients could have response whether they had these alterations or not. But we did see that in the metastases that progress, they did sometimes acquire a second beta-2 microglobulin alteration, so it’s not clear.

Some of it might be other mechanisms. There might be other checkpoints that actually if you look at the MSI-high tumors, they upregulate LAG-3/IDO. So it might be some other pathways might be more important in those patients, and it might be antigen presentation or it might be other immune cells that need to be targeted in those patient populations.

Case: A 38-year-old man with heavily pretreated advanced rectal cancer receives pembrolizumab

DR LOVE: So before we go on, I want to just take a breath and talk about some of the clinical aspects of this and particularly in your own experience. And I was wondering if you could talk a little bit about the patient you were telling me about, the 38-year-old man and just briefly what happened to him.

DR LE: Yes. So this is a 38-year-old Vietnamese gentleman who presented with a very locally aggressive rectal cancer. And actually, in rectal cancer, the MSI rate is actually even lower than the standard colorectal cancer patient. And so he was treated with standard chemoradiation and actually progressed right through that, with invasion into his prostate and seminal vesicles. And he was seen at a tertiary care center and they recommended FOLFOXIRI, again, trying to be extremely aggressive in a young patient with still localized disease.

And, actually, he progressed right through that as well. Then he was recommended regorafenib, because he had a KRAS-mutated tumor, so he wasn’t a candidate for the EGFR therapy. But before he got there, he developed a bowel obstruction, and then that set him back and never started the regorafenib.

At that time — it was the summer of 2013 — he came to Hopkins and saw one of my colleagues, Nilo Azad, and his performance status, I went back and read her original note. His intake note said, “He came here to look for options,” and the quote said, “I don’t want to die.” So, actually, that was really heartwarming just now 3 years later, because he’s doing so well.

So at that time, because of his performance status, neuropathy, depression and pain, the focus was really just to try to make him more comfortable, to see if we could get his performance status improved. So during that time then, he was admitted for renal failure, with a creatinine of 11. He had obstructed ureters. He was in and out of the hospital, was on TPN.

He was transferred to Hopkins. Again, young guy, so we were just being super aggressive. And the thought was, he could have a pelvic exenteration. So he had that in October 2013, and they also did intraoperative radiation, because it was invading into the pelvic wall. He went to rehab, came back December 2013 for his first postop scan. Multiple pelvic masses despite all that aggressive therapy. The good part of that was that his functional status actually did improve.

At that time, we actually had 2 studies targeting patients with MSI-high colorectal cancer, the pembrolizumab study and a second combination chemotherapy study. And back then, we didn’t have a lot of experience with the immunotherapy, and so we kind of thought, “Oh, it may be immunotherapy. If it’s going to work, it’s going to be slow. And we need to try the combination chemotherapy,” even though chemotherapy had failed previously. But it was still kind of in our minds that that might be the way to go in this gentleman.

Within his first on-study scan on that chemotherapy, again showed progression. So in March — I think now we’re in 2014 — he started on pembrolizumab. And within 3 doses, he was admitted for a partial small bowel obstruction, which was probably really due to adhesions. But his first scan, actually for the first time, showed some regression in his pelvic lesion. So despite him being hospitalized for his SBO, we were all very excited, because we saw the scan and knew that this was going to be an upward trend for him.

It took another 6 months, actually, to develop a full partial response, again because sometimes it can take a while. At that time, he then developed a lot of issues with his stoma and peristomal issues, with rash that eventually evolved into blistering, with lots of, kind of, steroid courses, which we eventually diagnosed bullous pemphigoid, based on biopsy and staining.

DR LOVE: So this was a complication of the pembrolizumab, do you think?

DR LE: Yes. Yes.

DR LOVE: Has that been seen with other — I don’t remember hearing about bullous pemphigoid.

DR LE: I think — I’ve certainly seen some rare cases of bullous pemphigoid. Sometimes I think it’s just buried in the rash category.

DR LOVE: And what exactly is it? How does it present clinically?

DR LE: For him, first it was just macular-papular rash, but then he started to develop blistering.

DR LOVE: Right.

Where was it? All over the body?

DR LE: It started around his stoma and then kind of spread out a little bit further. And then he also had a macular-papular rash throughout his body as well.

DR LOVE: So you stopped the pembro?

DR LE: We did, because sometimes for reactions, we try to do 4 to 6 weeks of steroids and then try to get the patient started. For this particular type of toxicity, it just wouldn’t — we tried to do that. It didn’t work. And usually for bullous pemphigoid, when you speak to the dermatologists, they actually treat patients for months, if not — I think he ended up actually on steroids for over a year.

DR LOVE: Wow!

DR LE: And we couldn’t stop the steroids quickly. So it took a while to taper him. But, interestingly, we kept getting the scans. And his scans continued to show regression on chronic steroids. And so he eventually developed a complete response despite — I think he got a total of about a year of pembrolizumab. And his last scan, November 2013 — 2017 still showed no evidence of disease. So now we modified the protocol, because the patients don’t want to come see us that frequently, so that now we see them every 6 months.

FDA indications for pembrolizumab in solid tumors

DR LE: The actual approval of pembrolizumab for these tumors, the approval was actually for adult and pediatric patients for solid tumors that had progressed following prior treatment and they have no satisfactory alternative treatment options. For colorectal cancer, they had to have had 5-FU/oxaliplatin and irinotecan.

The approval was actually based on 5 different studies, but most of the patients came from KEYNOTE-016, which was our study, KEYNOTE-164, which is a colorectal cancer study, a global study, and then KEYNOTE-158 has an MSI noncolorectal cohort, also a global study.

And in those 149 patients, the objective response rate was approximately 40%. But I think what was important to the FDA was really the response duration was not reached.

And here, before, we talked about 12 tumor types. I think here we might be up to 15 different tumor types being represented. And now there’s updated data from KEYNOTE-164 and KEYNOTE-158.

This was presented at ESMO 2017, although now we’re actually getting all the updated data together.

But this was the global studies looking at pembrolizumab across colorectal and noncolorectal in 2 different studies, and the data were combined.

This is looking at baseline characteristics for the colorectal study. The patients had to have had prior 5-FU/oxaliplatin and irinotecan. And for the noncolorectal study, they had to have had at least 1 standard therapy. And the more studies getting performed, now we’re looking at over 20 different tumor types being represented in the pie charts. I think the more pie charts we get and then we see, kind of, more rare tumors, such as adrenocortical, also included here.

The response rate in the MSI-high colorectal study was 28%, with the disease control at 51%. And in the noncolorectal, it was 37%, with a disease control rate of 59%.

These are the waterfall plots, again looking at the idea that most patients either had stable disease or shrinkage of their tumors.

And this is for the noncolorectal cohort as well. Again, I think important to the approval is the idea that the duration of responses are still not reached.

And I think that was very important to kind of confirm that with subsequent data for the FDA.

This is looking at the survival of the MSI-high colorectal patients, 72% at 1 year and close to 60% in the noncolorectal cohort at 1 year.

So there were some postmarketing requirements. These were single-arm studies with the idea that those were an unmet need. These were highly refractory patients. So the benefit was evident and meaningful, so those were definitely requirements. But there are postmarketing requirements. They want to look at additional colorectal and noncolorectal patient data, and they want to include children.

Now, you recall the label has children in it, but the studies did not. So the idea again is, if we can treat adults, why wouldn’t we be able to expand this to children in terms of the hypothesis? And so they’re going to have to generate that data in the children as well.

Use of checkpoint inhibitors as first-line therapy for metastatic disease

DR LOVE: You mentioned that the approvals are after prior therapy, prior chemotherapy.

DR LE: Yes.

DR LOVE: Do we have data in the chemotherapy-naïve patients, the first-line setting?

DR LE: So we don’t have any data. The KEYNOTE-177, which is pembrolizumab versus chemotherapy in first-line colorectal patients, has closed enrollment.

DR LOVE: How many patients went into that?

DR LE: I think in the end it was around 300 patients.

DR LOVE: Any data from the people who were treated, any anecdotal data? Have you treated patients first line? Is there any reason to think it would be more or less effective if you used it earlier?

DR LE: So I can give you a couple of anecdotes. So they’re actually interesting ones, because it brings up the issue of autoimmune toxicity as well. So I think that first-line patients are going to respond, just based on my anecdotal experience.

I have a patient who actually has a history of scleroderma. And the patients are allowed on study, if they don’t have any requirement for any active immunosuppression for several years. And she is a BRAF-mutated tumor, so it’s not germline. And she started on first-line pembrolizumab. And the tumors are shrinking, doing beautifully.

However, she’s developed some diarrhea, kind of low grade, but enough to really annoy her, enough to make me nervous, because she has scleroderma and I’m a little bit afraid to give her more pembro, because she’s having such a great response, antitumor wise, that I’ve been hesitant to redose her. She actually got scoped this week. The scope on visual inspection looked okay but on the biopsies are showing colitis.

So it’s interesting. I think she’s going to be one of those patients, once I get the CMV back, if that’s negative, we’ll put her on some steroids. But I probably will not rechallenge her, just based on her history. But I think she’s still going to have a great response. Antitumor wise, she’s due for scans next week.

DR LOVE: So how about if I say that in a patient, whether it’s colon or anybody else with MSI-high with metastatic disease who’s asymptomatic, low tumor burden, that when you look at risk benefit, it makes more sense to use a checkpoint inhibitor first line, than chemo?

DR LE: Yes. So I actually have another story of a patient — this is a different patient with a different autoimmune disease. She has ulcerative colitis. And she actually had prior breast cancer and so had experience with chemotherapy before. When she had her colon cancer, she had a complete colectomy due to the fact that it was UC associated, had more chemotherapy, had a horrible experience again.

She came to me with kind of low-burden metastases and has read about immunotherapy, said, “Dr Le, I know about your work. I want immunotherapy.” I was very hesitant to give it to her because of her history, even though she didn’t have a colon. I was worried that she was predisposed to other autoimmunity. And so I actually did drag my feet on her for as long as I could. And then in September, she presented with a large neck mass, growing. It was causing discomfort.

So we did actually give her — she didn’t want to go on a study. She refused chemotherapy. She said there’s no way she would take more chemotherapy. Again, very nice response. However, she has recently developed something similar to a Stevens-Johnson syndrome, which she says is related to the amoxicillin that she’s allergic to, and she would like more pembrolizumab.

However, I’m very concerned about giving her more pembro. And again, we’ll just have to wait and see. We don’t know how much pembrolizumab patients need, especially these patients who might be kind of predisposed to having immune effects.

DR LOVE: She still has tumor, though?

DR LE: She does. But it’s regressing, so I’m going to continue to follow. I think both these patients, I will just continue to follow them off of therapy.

MSI testing in solid cancers and hematologic malignancies

DR LOVE: Another question is who should get MSI testing. I’ve heard people say everybody with GI cancers should get it, everybody with any solid cancer. I don't know. What about heme cancers? From a practical point of view, obviously, there’s the economic issue. But who do you think should be tested?

And I should say with all these cancers, where you hear it is in metastatic disease, obviously, not primary. Metastatic, maybe with not great options.

DR LE: Yes. I mean, I think, in part, it does come down to an economic issue. The testing itself, if you do protein testing or PCR, run in the hundreds, not the thousands, like sequencing. I think some people are moving towards the idea that we’re going to catch more patients as we do more sequencing. But we do need those panels to get cheaper before we can apply that to everyone. But I agree that especially being a GI oncologist, we should test all of our patients, because many of those diseases are GI diseases.

DR LOVE: What about heme cancers?

DR LE: It’s present in some cancers, but it’s not as frequent. I’m not sure the data is as clear in heme cancers, but it is seen in some.

DR LOVE: Any solid tumors where you never see it?

DR LE: It’s pretty uncommon in sarcoma, but we’ve seen it. So sometimes we’re looking at, like, 0.1%. And sometimes we’re looking at 4%, depending on the disease.

DR LOVE: Just to pick out a couple of common ones, breast and prostate?

DR LE: Yes. So again, I think that’s going to be in the very low single digits.

DR LE: So the NCI-MATCH, there was a nivolumab arm. And they looked at their patients by MLH1 and MSH2 IHC. So this is an economical way to do it, because MLH1 pairs with PMS2. So you should be able to catch those patients. Not always.

And MSH2 pairs with MSH6, so this should be able to detect patients with deficiencies. And so in their study, they had 35 patients and tested them with nivolumab.

Most of them were MLH1, so I’m guessing many of these were actually sporadic.

And you can see the diseases represented. Now you see some prostate cancer. You see some breast cancer.

And if you look at their waterfall, this is the way I like this. The more studies we see, even the common tumor types — and again, low frequency, but in a common tumor type, that’s going to be more patients.

So again, I think it’s hard. I mean, should we test everyone? In an ideal world, I wish we could, because, obviously, again, it’s the same idea about the durable benefit, the real benefit that patients get.

This is just again looking at the question of who should get tested. It is in the NCCN Guidelines for colon and rectal cancer that all patients should be tested, (1), to look for Lynch syndrome, (1), to look at predictive markers in Stage II disease, avoiding chemotherapy in those patients. And the final, for the potential for immunotherapy. Also we have now studies in the adjuvant Stage III patients, as well as first-line studies. So it’d be nice to know this data early, so we can enroll on those studies.

DR LOVE: Just out of curiosity, what is the stage distribution of MSI in the adjuvant setting? You said Stage III.

Do — first of all, do you think these patients should get chemo or adjuvant treatment? And second of all, how does the staging break down and — this trial’s looking at Stage III.

DR LE: Yes. So again, the more advanced stage you get, the less frequent it is. But it’s much higher in Stage III and Stage IV. So I think we can certainly find these patients.

DR LOVE: And is it higher in Stage II, because, kind of, that’s where you hear about maybe not treating them?

DR LE: Yes. So we definitely do not offer chemotherapy in that stage. The current stages for immunotherapy don’t include Stage II patients. I think that’s going to be really hard to do in Stage II patients, because they do so well. So that’d be a very difficult study to do, I think.

DR LOVE: How do Stage III MSI patients do with and without chemo?

DR LE: Yes. So there, the data suggests that they do benefit from chemotherapy. And there, that’s where the concern is not giving them chemotherapy. Again, we usually use a doublet with them, 5-FU and oxaliplatin. So the current studies are actually chemotherapy plus or minus checkpoint inhibitors. This particular study is atezolizumab. We’re not participating in it, but it is a study that is open. And there is no single-agent arm.

But I think that will be a very interesting question. But I think people are concerned that we know that chemotherapy does have a survival benefit. And so at this time, there’s not a single-agent arm, but we don’t really know in earlier-stage disease if checkpoint therapy is going to work in this disease. I guess now we have some data in melanoma, but that’s relatively new.

DR LOVE: But I don't know if I’m remembering correctly. There’s some kind of resistance to 5-FU in MSI patients. Is that right?

DR LE: There is some thought that there is some resistance. And usually when we’re giving therapy, we’re giving cocktails, not single-agent therapy.

Future directions with anti-PD-1/PD-L1 antibodies

DR LOVE: So maybe we can kind of finish out with your last slide there.

DR LE: So the last slide is just looking at future directions. And we already kind of talked about this, looking at do we really have to treat only chemotherapy-refractory patients?

Could some patients avoid chemotherapy altogether? Again, we don’t have the data beyond the 3-, 4- and 5-year mark on those patients, but hopefully we’ll be developing that data.

We’re looking at whether we should be looking at earlier stages of diseases, such as adjuvant therapy. We talked about, should we be testing patients? We aren’t going to be able to give them the therapy if we don’t find them. And so there’s IHC and PCR. And as a byproduct of more NGS testing, we’re finding more patients that way as well.

We need to continue to dissect mechanisms of primary and secondary resistance so we can pick the patients for the therapy and pick patients for, potentially, combination therapies as well. And then there is now a push to look at immunotherapy across tumor histologies, just looking at tumor mutation burden itself. We and others have cohorts that are tissue agnostic as well, looking at just the tumor mutation burden as a predictive marker.

DR LOVE: Do you think common multiplex assays, FoundationOne, Caris, et cetera — they report something related to mutation burden, correct?

DR LE: They do.

DR LOVE: Is that something a general oncologist in clinical practice should look at and think about?

DR LE: I think so, because why do we send the test? We’re sending the test because we have a patient in front of us and we’re trying to think outside the box. Are there options that we didn’t think of for this patient? And so if you have a tumor that has — the one thing about these tests is, sometimes it’s really just not cellular enough for you to get all the data. So sometimes they can’t actually report the tumor mutation burden or the microsatellite instability. So that’s the one caveat, just because you do have to have adequate tissue for the actual testing.

But if you had someone — our first patient, actually, on our tumor mutation burden cohort has over 60 mutations per megabase. And so we’re very excited — she just started actually yesterday on her treatment — to see if she’s going to respond. So if you see that, I think it’s kind of hard not to try to get your patient on a trial or expose them to immunotherapy. It may or may not work. We’re hoping that it will. So I do think you can’t ignore the very high ones. The ones that are intermediate, I’m not sure what to do with that information.

Case: A woman with recurrent small cell lung cancer receives pembrolizumab monotherapy

DR RIZVI: So small cell, I think, is a really tough cancer. We’ve made really no progress with anything beyond platinum doublet chemotherapy. And second-line therapies are pretty disappointing.

I recall a patient I had, in terms of immunotherapy discussion, who had a platinum-based doublet therapy with etoposide and carboplatin. She responded, as most of these patients do. Unfortunately, she progressed relatively quickly, within a couple of months of finishing her chemotherapy. And those patients are highly resistant to second-line therapy if they don’t have any durable benefit to first-line chemotherapy.

And at that point, she also was becoming increasingly debilitated from her disease, with a lot of osseous metastases, pelvic bone pain, basically wheelchair bound from pain that was difficult to manage. And so at that point, we really didn’t have any great hope that second-line chemotherapy would be working for her. So we didn’t feel that she could tolerate immunotherapy combinations or a trial. And she went on single-agent pembrolizumab.

DR LOVE: Can I just ask, did she have a PD-L1 level?

DR RIZVI: I don’t do it in small cell lung cancer.

DR LOVE: Do you think it would have been informative if it had been, for example, high? Would you have been more comfortable about using single-agent PD-1? If it had been low, would you have thought about not using it?

DR RIZVI: No. I mean, 80% of small cell lung cancer is PD-L1-negative. It’s one of the reasons PD-L1 is a challenging biomarker. It works for some tumor types and doesn’t work for other tumor types. And so we never do PD-L1 testing on small cell.

So what could be really helpful in these patients with small cell lung cancer is tumor mutation burden, right? And the data with TMB in small cell lung cancer shows that it’s a much better biomarker. And what’s also potentially very cool for the future is that small cell lung cancer, it sheds a lot of tumor DNA. So I think that the potential for blood-based TMB in small cell lung cancer is going to be very exciting.

DR LOVE: But would that inform in any way this issue of treatment benefit of single-agent versus combination?

DR RIZVI: I mean, I think that if you have high TMB, right, your chance of responding to either single-agent or combination therapy is very good. I mean, not very good, but it’s much better. If you have high TMB, you’re going to respond better to the combination. But at least you’re going to know that there’s a reasonable chance that PD-1 is going to work for you.

DR LOVE: Interesting.

Mechanism-driven biomarkers to guide immune checkpoint blockade in cancer therapy

DR RIZVI: This is a slide that gets shown a lot. And it really is a slide to show how tumors are rejected by cytotoxic T cells.

It’s really the CD8-positive effector T cells that lead to activation of macrophage-induced tumor cell death. But it’s important to note that what happens is that there is an interaction of both CD4-positive and CD8-positive T cells with antigen-presenting cells. And these tumors shed, basically, pieces of tumor protein. And those are processed by APCs and presented by MAC to the T-cell receptor.

And so it’s really that antigen that leads to activation of CD4 to release cytokines and activate CD8. And then CD8 also releases cytokines. That allows CD8 T cells, through interferon gamma, to activate macrophage-induced tumor death.

What happens, though, remember, is that the human immune system is all about homeostasis and balance. And things get turned on, things have to get turned off, otherwise you’ll have autoimmunity and excessive reactions to foreign stimuli. And so when interferon gamma activates CD8 T cells, it also, in a feedback loop, will activate PD-L1. And that’s what we measure as a biomarker.

And also, that’s why PD-L1 is a variable biomarker, because it’s subject to this feedback loop. It may be activated; it may not be. So I think when we think about tumor biomarkers, we can think about PD-L1 expression as a measurement of possible benefit from immune checkpoint blockade. Now, we know that the T cells are turned off by PD-L1, so we can turn them back on by blocking PD-L1. We can use RNA signatures, such as interferon gamma-type signatures. And those also portend response to immune checkpoint.

And finally, we can use a measurement of tumor antigens. And the way we do that is with tumor mutational burden. So the more mutations you have, the more foreign or immunogenic a tumor is, the more pieces of antigen which will be presented by MAC.

So if you have a lot of them, then chances are you can turn on T-cell response to those antigens. If you have very few of them, you’re not going to have that many to turn back on.

DR LOVE: If you look at this model, how would you describe what happens to it when a cancer occurs?

DR RIZVI: Right. So I think that when a cancer evolves, it’ll evolve from 1 mutation, leading to 2, to 4 and then, at a certain point, it becomes clinically evident. I think that the reason that a cancer occurs is because it’s part of that balance that the human immune system is trying to achieve. It sees these tumor antigens turning on the immune system. And it’s programmed to say, “Okay. It’s turned on. It’s time to turn it off.”

And so the immune system turns off this response to the antigens, just because that’s what it’s supposed to do. It doesn’t know any better at a certain point. It gets overwhelmed. And so what these immune checkpoint inhibitors do is, they restore that memory of reacting against these foreign antigens.

DR LOVE: Do you think that it’s a fundamental issue with this system that leads to most cancers? Do you think — I think there’s been this thought out there from — I mean, my macro view as I look at cancer research, almost like of cancer as a biotic disease, that kind of model, infectious disease model, which is kind of like what targeted therapy is to me, just looking at it at a very simple level. How much of cancer do you think is that model, the altered bad seed type thing? And how much is it really an issue with the immune system and, particularly, maybe some genetic alterations in the immune system, very subtle changes that — maybe not that easy to appreciate?

DR RIZVI: Clearly, it’s multiple processes. I mean, if it were just 1 process, then immune checkpoint inhibitors would work in everyone. I think that what we all know now is that the tumors that respond best to these immune checkpoint inhibitors are those with a lot of mutations. And those are ones from too much UV light exposure or smoking carcinogen or impairment of a DNA pathway repair.

But you have other cancers which are driven by processes that are not completely foreign. Like, if you have a tumor mutation that’s completely foreign, that’s very clear. But if you have a breast cancer that’s driven by HER2 amplification, HER2 is a normal gene as well, so the body can’t recognize that as completely different. So that’s perhaps why breast cancer may not respond as well as melanoma, for example.

PD-L1 as a biomarker of response to immune checkpoint inhibitors in lung cancer

DR RIZVI: When we think about lung cancer, probably the most basic biomarker is smoking history, right? And we saw that early on in 2009, when we were first doing our Phase I trials with immune checkpoint inhibitor. And it was very clear that our smokers with lung cancer were responding and our never smokers were not, as well.

And I think that a never smoker with lung cancer is far less likely to respond than a patient with a prior smoking history. So it’s worth factoring that into one’s thinking in terms of a biomarker.

Beyond that, I think we’ve evolved into using PD-L1 as a biomarker. It has been very useful, particularly in nonsquamous histology. If you look at the squamous and the nonsquamous trials that were conducted with nivolumab versus docetaxel as second-line treatment, in the nonsquamous patients we see a clear improvement in response and survival in PD-L1 tumors that are high versus PD-L1 tumors that were negative, whereas in squamous histology, you don’t really see that much of a correlation, suggesting some different biology. And we see that clinically as well, where we often have squamous tumors that are PD-L1-negative that clearly respond to immune checkpoint inhibitors.

The first-line setting is where I think the PD-L1 biomarker is more robust. I think in the second-line setting for non-small cell lung cancer, it doesn’t really matter that much, right? I mean, pembrolizumab, nivolumab and atezolizumab have all beat second-line chemotherapy in survival. The first-line setting is a little bit more challenging, because chemotherapy works better in the first-line setting. So if you have a 30-percent-plus response in the first-line setting to chemotherapy, versus an 8% response in second line, I think you’ve got to be much better.

So the approval for pembrolizumab in the first-line setting is appropriately using a biomarker, PD-L1 expression, which is essentially now the standard of care for both squamous and nonsquamous lung cancer. So you do your molecular testing for adenocarcinomas and, for both adeno and squamous, you have to have a PD-L1 test done.

The prevalence of PD-L1 positivity at a 50% threshold is about a third of patients. And if you have a PD-L1-positive disease, it’s clear that pembrolizumab is better than platinum doublet chemotherapy. The response rates are 45% versus 28%. I think the most impressive data, from KEYNOTE-024, the first-line pembrolizumab trial, is that there’s nearly 10% complete response rates. And we almost never see complete responses with standard therapy. So to achieve a complete response rate of 9% is really remarkable. And at every parameter, progression-free survival as well as overall survival, pembrolizumab beat chemotherapy in PD-L1-positive disease.
 
DR LOVE: Do you think that it should be such a black-and-white — I mean, I can understand, maybe, from a regulatory, or whatever, perspective. But on and off, 50%. Is it a continuous variable? How does, globally, treatment benefit vary as you start with zero and move up to 50?

DR RIZVI: I don’t think it’s linear. If you look at the response rates in the middle level, they’re definitely lower. And then if you have no expression, it’s much lower. Also, the expression of PD-L1 is not linear. You do see a lot of patients at either end, those that are negative or low level and those that are 50% or higher. So the middle ground is less frequent. So it’s not like the expression is linear, either, so…

DR LOVE: What do we know about the heterogeneity in PD-L1 expression in the same patient at the same time and the same patient over time?

DR RIZVI: In terms of PD-L1 expression, it’s very clear that it’s not a perfect biomarker. You can biopsy different tumor locations or even the same tumor in 2 separate parts of that tumor and get different PD-L1 expression. So it’s definitely very heterogeneous.

DR LOVE: How often does that happen?

DR RIZVI: I mean, I think the concordance is probably around 70%, would be my estimate. So it’s not bad, I mean, but it’s not perfect. There are a number of companies that are trying to develop a PD-L1 PET radiotracer. So they take the PD-L1 antibody, and they link it to zirconium. And they’re trying to evaluate that as a radiotracer. And that may actually be a very cool way to more globally assess PD-L1 expression.

DR LOVE: What about PD-L1 expression over time in the same patient, Point A on their first diagnosis, Point B after, say, first-line therapy?

DR RIZVI: Again, I think that there’s not a lot of data. But the data that we have would suggest — and we know that from the very first pembrolizumab trials, where everyone that we treated had new biopsies on treatment. And then we were able to compare it to archival tissue. And the concordance, again, was about 70%. So we don’t fully know whether there is a negative effect of getting chemotherapy first, though. If you look at the KEYNOTE-024 trial where there was crossover allowed — so people were PD-1-positive. They got pembro alone or chemotherapy. And at progression on chemotherapy, they were allowed to cross over to pembrolizumab.

And if you look at just the patients that crossed over, the survival, versus those that got pembro first line, they still didn’t do as well. So maybe there is some deleterious effects of getting chemotherapy first. We do know that chemotherapy is associated with corticosteroids. And maybe that could be a negative effect. We also know that chemotherapy can induce subclonal mutations. So if you have a tumor that you induce a large number of subclonal mutations, that could be associated with a poorer response to immunotherapy. We’ve seen data, and we have data around that.

First-line therapeutic options combining immune checkpoint inhibitors with chemotherapy for metastatic non-small cell lung cancer

DR LOVE: What about the issue of combining chemotherapy with a checkpoint inhibitor? Now we have 2 trials suggesting positivity. Any predictors of whether or not patients could receive just PD-1 versus a combination?

DR RIZVI: I think the first-line space in lung cancer is becoming more complicated. We have KEYNOTE-021G, which was a trial combining pembrolizumab with pemetrexed and carboplatin versus chemotherapy alone. It was a Phase II randomized trial that showed a significant response, as well as progression-free survival and survival benefit. It actually led to accelerated approval of pembrolizumab with chemotherapy for nonsquamous lung cancer first line.

A recent press release showed that the confirmatory Phase III trial with a similar study design was positive for progression-free survival and overall survival.

We also have a recent Phase III trial called IMpower150 with carbo/paclitaxel and bevacizumab with or without atezolizumab. And that actually showed a progression-free survival benefit. It was a positive trial as well.

Interestingly, there was also a comparison between the standard regimen of carboplatin/paclitaxel and bevacizumab versus carboplatin/paclitaxel and atezolizumab. And that actually did not beat in terms of PFS. And so that is also — it’s going to be followed for overall survival, so we’ll see how the data matures.

But it was interesting that the chemotherapy plus bevacizumab plus atezolizumab was positive, but chemotherapy plus atezolizumab was not positive.

DR LOVE: What about the issue of anti-angiogenics in combination with checkpoint inhibitors? You hear about that in trials, renal cell, for example. This is an example, I guess, they talk about with bevacizumab. What’s the thinking, theoretically, about why there would be synergy and, again, if there’s any way to predict who would benefit by that strategy?

DR RIZVI: Yes. I mean, I think that’s an important question. In renal cancer, for example, there are certain RNA signatures that have been found to predict better response to combination of anti-angiogenics with atezolizumab. I think it’d be useful or interesting to see if we could try to explore those sort of RNA signatures for combination with anti-angiogenesis in other tumor types as well.

I think one of the uncertainties around the chemotherapy trials is whether they’re positive, because those patients that were going to benefit from immunotherapy benefited from immunotherapy anyways. And if you look at these trials so far, it’s clear that the PD-L1-positive subset did really well. And it’s not clear how much additional benefit the PD-L1-negative subset benefited by adding the chemotherapy — by adding the immunotherapy to the chemotherapy. And I think we have to wait for other more mature Phase III data for us to try to know that.

DR LOVE: I think that’s reflected, too, in the surveys we do of investigators when we ask, “If you have a patient with metastatic nonsquamous that has a high PD-L1 level, over 50%, what do you do?”

DR RIZVI: Yes.

DR LOVE: And, of course, they all give single-agent pembro. But then you say, “Okay. Suppose the patient’s super symptomatic?” I don’t know what you do. Sometimes people will add in chemo, but a lot of people don’t, because they really don’t think there’s extra benefit. What are your thoughts about that?

DR RIZVI: I think that’s a really important question. And I don't know the answer to that. I think that my default is, again, is to give single-agent pembro for the positives, but I think that for those patients that are really symptomatic, I think it may be worth adding some chemotherapy.

I had a patient recently who was in his forties. And he had lung to bone. He was actually PD-L1-negative. And so I started him on chemotherapy plus pembrolizumab. And, interestingly, his tumor mutational profile then came back showing that he was superhigh with tumor mutational burden. He had over 30 mutations per megabase, which is a really big number. And so, I mean, we, at that point — I mean, the question is, is this analogous to getting an EGFR mutation result back, right, after you start chemotherapy? And you just switch them to a more specific therapy.

And so for him, he finished his 4 cycles of chemotherapy with pembro, and then I continued the pembro alone. I didn’t continue any pemetrexed maintenance at all in him.

DR LOVE: Just out of curiosity, if you have gotten that TMB back before treatment, would you still have given him chemo?

DR RIZVI: I personally think that those patients who are PD-L1-negative, high TMB, I think that they will benefit more from combination immunotherapy. I think that we’ll see that — and we saw that. I think the data with small cell is proof of concept of that where if you have PD-L1-negative tumor, like small cell, and you have high TMB, yes, you’ll respond to PD-1 therapy. But adding CTLA-4 to drive the immune system further in terms of priming the T cells is probably going to be more effective.

Blood-based biomarkers for cancer immunotherapy

DR LOVE: So you want to comment about liquid biopsies or liquid assessment?

DR RIZVI: Yes. So I think that there’s a lot of excitement around liquid biopsies when you are able to detect circulating tumor DNA. There’s a recent paper around that showing that even you can pick up — make early diagnoses of cancers. And I think that everything is going to be moving towards blood-based assays.

There was data presented at ESMO last year with atezolizumab from some of the randomized trials in lung cancer showing that you actually can detect tumor DNA in probably two thirds of the patients that they collected plasma on. And what they did was essentially run the same Foundation-type panel on the blood and were — found that you could identify patients that had high TMB that would benefit from immunotherapy using a blood-based assay.

And I think it gets around sufficient tumor that’s required, timelines of getting tumor back. You see a newly diagnosed patient; you send a plasma test. You can get a result back very quickly.

DR LOVE: How about PD-L1 levels?

DR RIZVI: I’m not aware that we have a blood assay for PD-L1.

DR LOVE: Is there some reason to think we couldn’t?

DR RIZVI: It’s really specifically looking at membranous PD-L1 expression. So I think that people have tried to look at soluble PD-L1 in the bloodstream. And I don’t think they’ve seen good correlations.

DR LOVE: Interesting. So one other, quote, marker of response, at least in terms of the pathophysiology, what’s going on, and clinically, are tumor mutations, EGFR, ALK, et cetera. We’ve been hearing now from investigators that there’s a lower mutational burden in these patients. They didn’t respond as well.

In a sense, are these markers of response or nonresponse?

DR RIZVI: I mean, we don’t have a specific mutation marker of response to immunotherapy. We know that a subset of lung cancer that coexpress KRAS and LKB1 or 2 mutations, which actually would suggest less likely to have PD-L1 expression, less likely to respond to immunotherapy. But there really isn’t 1 specific marker that I think will tell us that this patient will respond. Some lung cancer tumors can have DNA pathway repair mutations, such as POLE and POLD. And those are seen in colorectal and other tumor types as well.

And those patients, if they have one of these mutations in the proofreading domain, can’t repair DNA. And so they’re, like, superhighly mutated. So that could be a marker. But, I mean, TMB is not that difficult a test, really, and I think it’s probably the one that’s going to gain the most traction as a single mutation test.

Analysis of tumor mutation burden on the Phase III CheckMate 026 trial

DR RIZVI: So around the same time as we saw that first-line pembrolizumab in PD-L1-positive lung cancer trial was a positive trial in the first-line setting. There’s also data from CheckMate 026, which used a lower threshold of PD-L1 expression. Instead of 50%, they used 5% PD-L1 expression. And that actually was a negative trial, did not meet its primary endpoint of PFS.

And so the investigators went back and actually looked at tumor mutation burden in these patients that were treated on CheckMate 026. They had matched tumor and germline DNA for performing the sequencing on close to 60% of patients and actually found that when you used the TMB estimate, it was very clear that the high TMB patients did far better than the low or the medium TMB. And we saw a PFS of 10 months versus 4 months in the patients that had high versus low, nivolumab.

So comparing nivolumab to chemotherapy, the high TMB patients were 10 months versus 6 months’ PFS. And they didn’t specially look at a threshold; rather, they just divided patients by tertile, so the upper third of mutations, middle third and lower third. And just by that simple division, they were able to show that the high TMB patients had a significant benefit.

What was interesting is that the low and the medium TMB didn’t really perform that differently, suggesting there really is some threshold number of mutations that’s required.

What was also interesting is that if you have both high TMB and PD-L1 more than 50%, the response and progression-free survival was really off the wall. I mean, it was really pretty dramatic. And so you’ve got a subgroup that’s as good a biomarker as EGFR mutation testing, for example. I mean, you can really identify those patients with incredible granularity.

So one of the issues around mutation burden that was done in CheckMate 026 is, they did whole exome sequencing on patients’ tumors. So to do whole exome sequencing, you need to sequence the tumor of the whole exome. And they also needed to get blood for germline sequencing. It’s a much more complicated, laborious analysis.

So the question is, is there an easier way to get a TMB estimate? And in this analysis around the CheckMate 026 patients, as well as other analyses, looking at a targeted panel of 300+ genes, like the Foundation Medicine panel, basically if you just count all the mutations within that more limited coverage, you are able to show a very good correlation.

And so a TMB report at this point will give you a readout of what the number of mutations per megabase your patient that you have has. So I think it’s another important component of these targeted panel reports.

DR LOVE: How close do you think we are to having TMB as a marker that’s used in clinical practice?

DR RIZVI: I think it’s close.

DR LOVE: Seems like it. I mean, are any investigators — or do you actually use it in your decisions yet?

DR RIZVI: I guess the challenges would be, in the first-line setting, is getting a TMB report back in a timely enough a manner, right? So you have a newly diagnosed patient. You can get EGFR, ALK and PD-L1 back within 7 days. To get a TMB report through Foundation can take 2 weeks or more by the time they get the material and so forth. So I think that — and then second line for lung cancer, they’re all going to get immune checkpoint therapy anyways. So I think that it’s going to be helpful, if there is — and also, it’s not approved for giving it first line based on TMB.

I think if you have an FDA approval for first-line therapy based on TMB, you can be sure that everyone’s going to do it. The timelines are going to get compressed, and you’re going to get a result back in a way that you’ll be able to make a first-line decision. But I think it’s going to be hugely important not just for lung cancer but really for all tumor types, right? I mean, there’s patients who with any tumor could have a high TMB. And I think it’s going to be just like MSI testing, right? MSI is basically just a surrogate for TMB, right? And the approval for MSI is pan cancer. So I think in the near future, you’re going to see a TMB approval pan cancer.

DR LOVE: Hmm. Fascinating.

Neoantigen clonality and PD-1 benefit

DR LOVE: Why don’t we just continue?

DR RIZVI: Ah. No. So I think there some of the work that we and others are going on is to try to define the mutation landscape better.

We know that a certain number of mutations correlates with response to immune checkpoint therapy, but there’s data now that the clonality of a tumor also affects response to PD-L1 inhibitors.

Work from Charlie Swanton’s group, published in Science in 2016, they actually looked at our data from our Science paper with pembrolizumab-treated patients, showing a correlation between TMB and benefit with immunotherapy. And they looked at it a little bit further and looked at the neoantigen clonality of these tumors and found those tumors that were highly clonal responded, whereas those that had a large subclone fraction did not respond as well. So I think this is more of a research tool at this point, but I think that as we are starting to use TMB more and more, we’ll be able to refine it using some of these additional pieces of information.

DR LOVE: Yes. Can you explain a little bit more about what neoantigen clonality is?

DR RIZVI: So why do patients who have high TMB respond better to immunotherapy? I think that one level of explanation is that the more mutations you have, the more likely the cancer can see those tumors as foreign, right, because they have a different genetic code than your own genetic code. But remember, not all mutations change your amino acid. So you could have a silent mutation, which won’t actually change the coding, and so your amino acid remains the same. So it’s not seen any different by the immune system.

You have nonsynonymous mutations, or missense mutations, which actually change the amino acid. And when we look at all these TMB estimates, largely we’re looking at nonsynonymous mutations. There actually can be mutations which are even more immunogenic. And these are frame-shift insertions or — they’re deletions, these indels, that we call them. And if you have a deletion or insertion and you change the reading frame of the amino acid, then you will have potentially many new amino acids and new foreign amino acids that are generated from 1 mutation.

So beyond that, mutations that are immunogenic are what we call new antigens. And what that means is that they must be able to bind to MAC Class 1 tightly. And they also need to bind to the T-cell receptor tightly. And so if you have tumor DNA and it’s basically broken down into amino acid — or — of peptide fragments. And there are usually about 9 or 10 amino acids long. And those 9 or 10 amino acids are presented by MAC1 to the T-cell receptor. So if you have a mutation, it could change it to one that doesn’t bind to them well, and it’s not really immunogenic.

So the ones that change the peptide fragment in a way where you have high affinity binding to MAC and T-cell receptor, those are the ones which, when we give immune checkpoint inhibitor therapy, we’re probably turning on the T-cell response to those specific immunogenic mutations or new antigens. And so when we try to count neoantigens, we make estimates of what a neoantigen is based on some peptide prediction algorithms.

And so you need mutations, you need immunogenic mutations, you need ones that are neoantigens, and then, ideally, you need tumors that have very clonal neoantigens. So if you have a tumor and it has many different populations of cells which are very different, you probably aren’t going to have a lot of neoantigens. If you have 1 tumor and you have 1 or 2 neoantigens that your T cells can target against the neoantigen, you can eradicate the whole tumor, those are better neoantigens.

Effect of tumor mutation burden on efficacy of nivolumab alone or nivolumab and ipilimumab for recurrent small cell lung cancer on the CheckMate 032 trial

DR RIZVI: So small cell lung cancer, 80% of them are PD-L1-negative. I don’t even test for PD-L1. We have immune checkpoint inhibitors not approved for small cell, but they’re on NCCN Guidelines, so they’re often used.

So data that was presented last year from World Lung looked at small cell lung cancer patients that were treated with either nivolumab or ipilimumab and nivolumab. We know that PD-L1 is basically negative in most patients with small cell lung cancer. We know that there’s about 20%, 25% response rates with immunotherapy. We know that small cell is almost entirely associated with smoking, so tumors should respond to immune therapy.

And in this analysis, patients were — their tumors underwent whole exome sequencing to look at whether tumor mutational burden played a role in response to immune checkpoint therapy.

And they did the same thing as in 026, where they broke the patients down by tertiles — so low TMB, intermediate and high TMB — and found that both for nivolumab and ipilimumab and nivolumab, there’s a correlation between TMB and response. So if you have high TMB, the response rate was 46% with ipi/nivo versus 21% with the nivo alone.

And I think even more impressively, if you look at the overall survival, if you had high TMB and you received ipi/nivo, the 1-year survival was 62%, which is really remarkable data for small cell lung cancer. I mean, there’s nothing that will give you a 1-year survival in pretreated small cell of 60%.

And even with nivolumab alone, if you have high TMB, there seemed to be a differential benefit. But if you look — again, I think the most important data is the OS. If you have high TMB and you got ipi/nivo, the median OS was 22 months. If you had low or intermediate TMB, the OS was 3 to 4 months.

So I do think the ipilimumab does add to the nivolumab. And in general, if I’m using immunotherapy for small cell lung cancer, I will use the combination.

DR LOVE: Did they do PD-L1 testing on these tumors?

DR RIZVI: They did. And there’s no correlation between response and PD-L1 expression.

Tumor mutation burden as a potential predictor of benefit of immune checkpoint inhibitors in squamous cell carcinoma of the lung

DR LOVE: I was going to ask you in terms of the issue of combining chemo and checkpoint inhibitors where we are with that strategy in squamous cell, when we might be seeing some data on it and any predictions about what we’re going to see.

DR RIZVI: So there are trials that are ongoing that do include squamous histology, and there are trials specifically for squamous cell lung cancer as well. I think that they’re going to be reading out in the near future. I don’t know exactly when. I haven’t heard a timeline on when to expect that.

DR LOVE: Any prediction about what we’ll see?

DR RIZVI: I think the squamous patients are potentially going to benefit from giving chemotherapy with PD-1. I think the correlation between PD-L1 expression and response is not as good in squamous cell lung cancer. So I think there’s a lot of patients who test negative for PD-L1 with squamous cell lung cancer that are not getting first-line therapy but could benefit. And I think that some of these first-line trials with combination immunotherapy with chemotherapy will help squamous patients.

DR LOVE: What do we know about TMB as a predictor of treatment benefit with IOs in squamous cell?

DR RIZVI: There’s no specific data around TMB and squamous, but I don’t expect that it should perform any differently. I think it’ll be a good predictor of benefit in squamous lung cancer.

DR LOVE: Do you think in general there are higher TMBs in general with squamous cell? I always think more about smokers with squamous cell, or not really?

DR RIZVI: I mean, I think that the medians for squamous and nonsquamous of TMB are about the same. You see more compression of squamous cell histology around the median, whereas in nonsquamous, you see more of a spread. But the median’s around the same, in the 200-mutations range.

I think that we can use PD-L1 expression and tumor mutation burden together.

If you look at the CheckMate 026 data, those patients who have both high TMB and PD-L1, I think I can envision somewhere in the future, as we get TMB and PD-L1 data back in a real-time way and make decisions on first-line therapy, someone like that probably could or should get PD-L1 monotherapy. If you have high TMB and PD-L1 of anything less than 50%, frankly, they should get a combination therapy. And that could be PD-1/PD-L1 plus CTLA-4, or it could be combination of PD-1 with chemotherapy.

Maybe that’s a subset where chemotherapy really adds something. You’re immunosuppressed for reasons other than PD-L1. You have a high TMB. Maybe you release antigens with a little chemotherapy, and then the immunotherapy can work better. And then I think for patients who have a low TMB, I think I’m questioning how much these immune checkpoint inhibitors are helping patients. And I think we may need a more personalized approach, looking at some of the T-cell bispecifics or other ways of using immunotherapy.

And whether immune checkpoint inhibitors can help patients with lung cancer who have low TMB I think I’m less certain about.

DR LOVE: Any specific other immune combination strategies that you think might lend themselves to that subset with low PD-L1 and low TMB? I don't know that much about IOs. I know there’s epacadostat. Or any of those alternative agents in combination? Would they make sense to look at in those patients?

DR RIZVI: It’s possible. I mean, we are seeing combinations with IDO inhibitors that are — the data with pembrolizumab and epacadostat does show about a 30% response rate in PD-L1-negative tumors. Whereas the expected response rate in PD-L1-negative tumors with pembro alone would be 10%.

I think it’d be really useful to go back and get genetic data on those patients. What’s their tumor mutational burden status? Do they have high TMB or low TMB? Is there an opportunity there?

DR LOVE: Maybe you can just go through the last 2 slides and kind of summarize this.

DR RIZVI: I think that the biomarkers for lung cancer is an evolving story. We also had a paper in Science a few weeks ago looking at patients’ own genetics, what their HLA composition was and how it affected outcome with immunotherapy. We know that the HLA zygosity within the infectious disease literature can affect HIV outcomes and response to HIV therapies. So we explored that in cancer patients. And so —

DR LOVE: What did you see?

DR RIZVI: So what we saw was that — so basically for HLA Class 1 presentation, which — you have 3 classes, A, B and C. And you inherit 1 set from each parent. So you have potentially 6 classes. And if you have full heterozygosity, if you have all 6 of them, you respond better to immune checkpoint therapy. If you lose any of your HLA zygosity, then you respond less well to immune checkpoint therapy. So I think it’s interesting that beyond the tumor genetics we’re seeing, the patient’s own genetics can impact response to immune checkpoint therapy.

Then there’s a recent data from Charlie Swanton that close to half of patients with lung cancer have loss of heterozygosity at HLA. So if you lose, you’re basically going to go automatically from 6 to 3 alleles. And so if you’re already homozygous at 1 or 2, you’re going to reduce your number quite a bit.

It makes sense that if you have twice as many MAC molecules able to present antigen, you’re going to have twice as many chances of activating T cells. So it’s potentially just a simple math effect. So what we’re going to see, that the biomarker program for cancer immunotherapy is going to get more and more complicated, but I think also more and more granular. The more we understand, the more we’ll know, why did that patient with high TMB not respond to immunotherapy? And why did that one?

Because it’s more than just the number of mutations. It’s the type of mutation, whether they’re nonsynonymous, whether they’re indels, what the HLA composition is, how many alleles does that patient have? It’s all going to come together in a very granular way for us to predict benefit.

DR LOVE: Subsequent to this recording, it was announced that the Phase III IMpower131 study demonstrated that the combination of atezolizumab and carboplatin/nab paclitaxel improved progression-free survival compared to chemotherapy alone in the first-line treatment of advanced squamous non-small cell lung cancer.