So Rusty, we'll turn it over to you guys.
Burles Johnson III: Okay. Thank you, Dr. Chang. As Dr. Chang noted, this is a paper we published, Roy and myself with colleagues from Johns Hopkins, earlier this year on looking at a composite B cell/CD8 T cell biomarker and looking at circulating tumor DNA status to see whether using this composite biomarker can actually potentially augment the ability to see prognosis and prediction value of adjuvant immunotherapy in high-risk muscle-invasive urothelial carcinoma.
So it's a busy slide, but there's a lot going on in circulating tumor DNA that I wanted to just outline very quickly. A lot of it was presented at GU ASCO just recently this year. So we all know that circulating tumor DNA positive status is a prognosis for worse survival in both neoadjuvant and adjuvant muscle-invasive urothelial carcinoma spaces. It also defines two benefits from adjuvant immune checkpoint inhibitor therapy. And also that if ctDNA is cleared, whether it's pre-cystectomy or receipt of adjuvant immune checkpoint inhibitor therapy, that associates with improved overall survival. And also, that timing of the initial circulating tumor DNA positive status and the actual value initially positive also associates with outcomes.
CtDNA, and actually urine tumor DNA, are being shown to be complementary in recent studies such as NIAGARA.
Important to note that a small minority of patients who have repeat circulating tumor DNA (-) values actually have recurrence on CT scan. And also recently, as most of the audience is aware, perioperative EVP, whether a patient is cisplatin-eligible or not, has OS benefit in patients with muscle invasive urothelial. So the big questions are, can circulating tumor DNA be used to de-escalate therapy or intensify therapy? There's multiple trials that are studying this.
So looking at immune biomarkers, again, most of the audience knows this is just to review that high tumor B cell and T cell infiltration associates with treatment response and multiple cancers to immune checkpoint inhibitor therapy, that high tumor expression of a CD8 effector T cell gene signature associates with greater overall survival in patients with metastatic urothelial carcinomas, Tom Powell's work.
Our group looked at the IMvigor210 study, which is patients who had metastatic urothelial carcinoma and received the anti-PD-L1 immune checkpoint inhibitor, atezolizumab. They were stratified into four quadrants by B cell gene signature and a CD8 T cell gene signature into four groups based on high and low expression of both signatures. And we found that patients that had both high tumor B cell and CD8 T cell gene expression had the highest overall survival.
So despite advances in circulating tumor DNA, now we're using adjuvant immune checkpoint inhibitor therapy and now actually perioperative EVP in post-cystectomy space and adjuvant ICI in the high-risk post-cystectomy space. Not all patients who are circulating tumor DNA positive benefit from immune checkpoint inhibitor. And actually up to 35% of patients who are circulating tumor DNA (-) at cycle one day one post-cystectomy will actually have recurrence within three years.
So we wanted to look at the IMvigor010 study, which looked at patients with high-risk urothelial carcinoma after cystectomy who are randomized to atezolizumab versus observation and to see whether the tumor B cell gene signature and high CD8 T cell gene signature expression, we call that B8T, associated with overall survival in patients with high-risk urothelial carcinoma after cystectomy. So the overall goal of the study was to determine whether the B8T could provide additional prognostic or predictive information when used with circulating tumor DNA status.
So as you can see on the left, these are all patients who are circulating tumor DNA positive. We have the data on circulating tumor DNA (-) patients as well. But if you look at the left, you can see in that Kaplan-Meier curve the yellow line. Those are patients who have B cell high and CD8 T cell high gene signature expression on RNA sequencing. And when you look at those patients, those patients on the left-hand side received atezolizumab. The patients who are B8T high-high appeared to have the highest overall survival.
But that's also true in the middle. And in the middle, these are patients who had no atezolizumab after cystectomy in the high-risk space. No atezolizumab, no adjuvant atezolizumab, and they're circulating tumor DNA positive, and they were observed. And you can see that the yellow line again appears to be the highest. So the patients that had tumor B cell high, CD8 T cell high had the highest overall survival in both the atezolizumab on the left and the observation group in the middle.
And then on the right, you can see we took the B8T, or Roy took, he's the bioinformatician for this, we took the B8T high-high in A and the B8T high-high in B, the atezolizumab in the observation group, and graphed them together, and you can see they overlap. So that suggests that the B8T high-high group has favorable overall survival that's independent of receipt of atezolizumab.
Then we looked at the B8T high-low group. We looked at all the other groups, but we're just showing the groups where we see key differences here in this short presentation. You can see on the left when you group patients who are ctDNA positive and negative together, those patients who received adjuvant atezolizumab had an overall survival benefit versus those who were observed and did not receive adjuvant immune checkpoint inhibitor therapy.
And that also extended to whether the patients were circulating tumor DNA positive, which is in the middle, or whether patients were circulating tumor DNA (-) on the right. So the conclusion from this slide is that patients who are B8T high-low, they had tumor high B cell, low CD8 T cell gene signature expression, they had overall survival benefit from atezolizumab that was independent of circulating tumor DNA status.
So I invite you to look at the large black box towards the bottom here. This is multivariable analysis by, again, B8T gene signature expression. And you can see that independent of the B8T group, patients who had positive circulating tumor DNA status had worse overall survival.
However, in the smaller box, if you go up just a little bit, you can see that smaller box. That's in the B8T high-low group. You can see that patients that received atezolizumab in the adjuvant setting actually had better overall survival, or superior overall survival, compared with the observation arm. And that was statistically significant even after controlling for multiple variables such as tumor stage neural status, all the variables that you see there on the left-hand side.
And just to quickly look at the, and again this is ongoing, the actual B cells that are inside the tumor, Roy did a CIBERSORT analysis. You can see on the left that memory B cells are enriched in patients with both B8T high-high and tumor B8T high-low status.
However, on the right, you can see that naive B cells, and there's some reports of naive B cells expressing high levels of PD-L1, they're enriched primarily in patients with the B8T high-low status. So again, this is preliminary data, but this may start to explain why there's differences that we're seeing. And those differences are likely going to be teased out in the phenotype and the activity in the function of the B cells.
So to conclude, patients who had tumor B8T high-high status, had high overall survival. Adjuvant atezolizumab did not provide additional benefit, even if the patient was circulating tumor DNA positive. Atezolizumab receipt associated with longer overall survival in patients who were B8T high-low, despite whether they were circulating tumor DNA positive or circulating tumor DNA (-). So the study identified patients who were circulating tumor DNA positive who did not have benefit, but also patients with circulating tumor DNA (-) status who benefited from immune checkpoint inhibitor therapy.
A couple of limitations, again, retrospective nature of analyses. In particular, the B8T high-low group was small, but we still saw significant differences that we felt were worth reporting. And as molecular data for nivolumab and pembrolizumab were not available, we focused on atezolizumab here.
So hopefully, and as Sam, Dr. Chang, alluded to, the idea would be hopefully these immune biomarkers such as the B8T could be used in conjunction with circulating tumor DNA to stratify outcomes and help identify patients who'd benefit or maybe not benefit from adjuvant immune checkpoint inhibitor therapy.
And as we all know, it takes a village. So I just wanted to acknowledge everyone in our group and just say thank you for our financial support and also for the patients which inspire us to do the work that we do. Thank you.
Sam Chang: Well, fantastic work. Really, again, continues a pathway to attempt to personalize care for our patients. I'm just going to ask some general questions that urologic surgeons that think along the level of my intellectual capability can get just so that we have a rough idea of what we're talking about. So maybe Roy can answer this question. Can you tell me roughly just the percentage of the breakdown of B8T high-high, high-low in the different groups? It would give us an idea of, oh, they're equally split. This can really be helpful if it is only a certain small number. And I know in the multivariant there was clearly a relationship in this retrospective evaluation, but still really, really intriguing. So can you give us some general ideas regarding percentage breakdown?
Roy Elias: Yeah, absolutely. That's a great question, and it's really important when thinking about biomarkers is how prevalent are these groups amongst the population. So as a background, the way we determined our cutoffs was by first visualizing the distribution of the enrichment of these genes in the entire population. And actually, I think it made it in the supplement, Rusty. I don't know if it made it to the main figures, but in the paper you can see a very nice bimodal distribution in both the B cell signature and the T cell signature. And so it lended itself very well to using just a arbitrary cutoff of zero.
And the way that played out in this cohort was that roughly half of the patients were B cell... It cut them down the middle. Half were B cell high, half were low, and then half are T cell high, T cell low. So there was approximately 25%. The patients fell into quadrants, pretty close.
Burles Johnson III: Yeah. And Roy's exactly right. That's in the supplemental data. And then if you look at the four quadrants themselves, and then you look at the high and the low-low, and then the high-low and the low-high, and I had to look at it right here, I always think of the high-high as the highest. I think the high-low and low-high, Roy, are probably, high-low and low-high are probably 15% a piece. And then the high-high and low-low probably make up about a third and a third.
So it's a good point. And Sam, I know you were so smart. You asked the perfect questions, how much of a impact is this going to have? Because if the biomarker for high-low is not that high, then why do this in high-high as well? So high-high, I'd say 30-ish percent. And then high-low 15%. That's just roughly guesstimating based on figure 1A and looking at the Kaplan-Meier curves that Roy so kindly generated for me. I would've been completely lost without him in this research.
Sam Chang: So then obviously this is, everything we're doing is trying to get that next step to both, just as your introductory slide intimated, who do we really need to escalate versus who can we de-escalate? Or that might've been the second slide. If you look at that, then tell me what you all are planning next steps. Are you going to go back and look at other trials and see if this is corroborated perhaps with other forms of immunotherapy in perhaps the metastatic versus the edge? Tell me what you all are thinking next step-wise, because I don't know the availability of the gene signature data that you've pulled with this atezolizumab trial. So tell me your next steps.
Burles Johnson III: I asked Joaquin when the IMvigor011 data is going to be available when we were at a talk in Savannah, Georgia a couple weeks ago. So that's, obviously, these publicly available datasets are important. So there's a couple of them that Roy and I have discussed, and that's all well and good. I think we all know it's harder to get the nivolumab data, and so that's just one challenge.
And then the other ideas are to try and generate our own data. So that's another future direction because in order to be able to validate this, retrospective's not going to be enough. We're going to need to do some prospective work.
Roy Elias: So I thought it was very provocative. This model, Rusty's being very modest. Rusty developed this model and this composite score. I thought it was so provocative that the state of the tumor microenvironment could lend itself to responses or predict responses. And so I think that's something else that we're both working on. And we've actually got some experiments and grants planned where we're hoping to start dissecting the compositions and the phenotype of the tumor microenvironment to really understand what it is that enables one subtype to be sensitive versus not.
Sam Chang: I think that being able to look beyond just the tumor itself versus the drug itself, clearly interplay of everything really is going to be really, really important. But then to be able to tease out, just as you said, Rusty's being all modest, but point of fact, being able to actually consider that group.
As you, Dr. Elias, looked at the composite data, was there anything else as you looked at the expression signals, anything else that was intriguing or is that for us to figure out or maybe hear from you guys in the future? I don't want the cat to get out of the bag, but anything else that seems at least intriguing?
Roy Elias: Well, I will say molecular subtyping is something that we all care quite a bit about. And also in the supplement, we looked at the distribution of these B8T classes across the TCGA subtypes and the consensus bladder subtypes, which have been previously defined. And in running this analysis, I had expected some subtypes to be dominated by a certain tumor microenvironment. But what we actually saw is even though the distributions weren't the same across all the subtypes, you can clearly find all four groups amongst all the different subtypes. So it shows that even though the tumor microenvironment is related to the underlying molecular properties of the tumor itself, there's clearly other factors at play that influence this. And I think understanding that will be very, very fruitful for our patients one day.
Sam Chang: Oh, no question. And no question we're scratching the surface. But we've got to, I think, continue, and it's with individuals like yourself that will only give us more information to help us take those next steps. So I look forward to having both Dr. Johnson and Dr. Elias again to help elucidate the next findings that will help better identify patients appropriate for care. And then also, as you all have both mentioned, de-escalate the care as well. So thanks to you both and look forward to talking to both of you again soon.
Roy Elias: Thank you for having us.
Burles Johnson III: Yeah, thank you so much.