Predicting Treatment Response in Prostate Cancer: The Power of Spatial Proteomic Approaches - Christina Curtis

July 10, 2023

Christina Curtis shares insights from her work on breast and gastrointestinal cancers, discussing their potential application to prostate cancer. The focus of her discussion revolves around innovative strategies for defining biomarkers of therapy response using spatial proteomic approaches. Dr. Curtis describes how this new technology lets researchers examine the composition of tumor cells and the surrounding microenvironment, potentially predicting which patients will achieve a pathologic complete response to targeted therapy. Furthermore, Dr. Curtis explains that spatial proteomics allow for a multi-marker analysis which provides a more in-depth understanding of cell changes during therapy, and has been instrumental in revealing the role of immune cell influx in predicting patient responses. The conversation also explores the potential for integrating AI techniques with spatial proteomics and highlights the need for composite biomarkers to enhance risk stratification. The conversation concludes with optimism for the future application of these techniques in clinical trials and their potential to move effective therapies earlier into biomarker-stratified populations.


Christina Curtis, PhD, MSc, Endowed Professor of Medicine and Genetics, Stanford University, Cancer Computational and Systems Biology Group, Director of Breast Cancer Translational Research, Co-Director, Molecular Tumor Board, Stanford Cancer Institute

Alicia Morgans, MD, MPH, Genitourinary Medical Oncologist, Medical Director of Survivorship Program at Dana-Farber Cancer Institute, Boston, Massachusetts

Read the Full Video Transcript

Alicia Morgans: Hi, I'm so excited to be here with Professor Christina Curtis, who is a professor at Stanford and is here at the Prostate Cancer Foundation Annual Scientific Retreat to talk with all of us about some of her experience that can be applied from her work in breast cancer and GI cancers and how we can think about some of these innovative strategies in prostate cancer. So, thank you so much for your innovative speech. Can you tell us a little bit about what you discussed and teach all of the rest of us who were not at the retreat this year?

Christina Curtis: Absolutely. So really what I was sharing is how we've been approaching new strategies to define biomarkers of therapy response. I used the HER2 positive breast cancer setting and the neoadjuvant treatment paradigm as a model for this in part because we of course have many FDA approved therapeutics here. But it's also a place where we have an opportunity to deescalate therapies in some instances by omitting chemo and escalate therapies in the context for patients who may not respond to those agents.

So what I described is really using new technologies that allow us to peer into tissue in situ. So taking whole tissue sections and using spatial proteomic approaches to now ask about the composition of not only the tumor cells but also the surrounding microenvironment. And really excitingly in some of the first applications of these new technologies, we were able to show that comparing pre-treatment samples to an on-treatment biopsy taken during the course of a targeted therapy, that we could predict which patients would achieve a pathologic complete response to that agent. And so this is really a powerful way to be thinking about these new tools and how we can stratify patients.

Alicia Morgans: It is extremely exciting. And I'm just wondering, can you dive into spatial proteomics a little bit? Tell us exactly what you're doing, how you're able to really understand on a tissue level. I assume you're somehow understanding which patients are achieving drug near target, is that part of this?

Christina Curtis: That's part of it, absolutely. And so again, the HER2 positive setting was a nice place because we of course know what our target is, the HER2 or HER2 oncogene in this patient population. And so what these spatial approaches allow us to do is to multiplex really. And so we can go beyond the sort of classical immunohistochemistry and now look at let's say 40 markers and those markers can be predefined, we can use antibody based approaches and really now look not just at one marker, say HER2, but at a whole composite. And in doing that, we can ask which cells within the neighborhood are expressing those markers, what their relative abundance is of particular cell types and how that changes through therapy. So did we in fact hit the target? Did the downstream signaling axes also go down?

What we found was really striking that in the patients that did achieve a pathologic complete response, not only did you see down regulation of HER2 and all of its downstream targets, but there was a concomitant influx of immune cells and it was actually those immune cells, predominantly lymphocytes, that were most predictive of which patients responded. And so that's telling us about a whole nother facet of the biology that if we were just to look at the genome, which I'm a big proponent of as well, or even the transcriptome we might have missed. And so it really allows us to ask about the signal and hitting the target or not. And I think a really great extension of this kind of approach is in early trials where we're trying to ask, do we know what this biologic is doing? And can we start to learn that biology from these kinds of correlative studies, especially when we compare pre and on, because of course we need to know the baseline status of that tumor.

Alicia Morgans: And I think further applications too, at least in prostate cancer, we have multiple drugs that are supposedly hitting the same target. But I say supposedly because we actually don't know. But I wonder if spatial proteomics could help us understand on a much deeper level, of course, if we're hitting the target, if we're hitting the downstream. And then what other effects may be going on, whether it's immune cell infiltration or other effects. HER2 targeted by multiple drugs at this time, Herceptin's one of the early ones but there are multiple others. Are you able to look at different drugs and see if they may be hitting near target a little bit differently?

Christina Curtis: Absolutely. Yeah. And so in this trial, which was an older study, it was a phase two neoadjuvant trial comparing actually Herceptin versus lapatinib. So small molecule inhibitor, and we know the outcome of that, but in fact it is quite interesting to then start to ask, are they eliciting the immune system in the same fashion? So it's not just the tumor cells and are we hitting the target in the same way and do we see the same degree of modulation of signaling, but also what's happening in the periphery. And I think that's really powerful is that we're keeping the tissue in context, we're not dissociating it. So we can really read out these patterns quite elegantly, I would say. And of course, it requires now new computational tools to parse these data. It also begs the question, what kind of biomarkers do we need? And in this case, we were able to distill this down to a single protein that we could then measure with a routinely available clinical assays such as CD45, reading out CD45.

So I think the biology will be different for different tumor types and for different therapeutic agents, and that's where having the ability to look in multiplex and then potentially distill it down to something far simpler is going to be very powerful. I think these kinds of approaches, the spatial technologies, whether it's proteomic or transcriptomic, I would say the transcriptomics, still catching up, can also be paired with really powerful approaches to use digital pathology. So right, digitizing these H&E images that are routinely collected and using them as yet another source to learn about the underlying biology. And really these kinds of data types can actually be integrated. So that's going into the frontier is how do we leverage the wealth of information that's collected in some cases in the context of routine care and in other cases, these are still just research assays. Right?

Alicia Morgans: Absolutely. And so take us a few steps further along that line. So in prostate cancer, we are newly seeing some data with this AI derived or reviewed pathology. And I think at least at this point in time, there are signatures that been, are being developed that help us predict which patients may actually respond to one treatment or another based solely on AI on these pathology specimens and really interesting work with these digitized slides. So how is it that we can combine that kind of an approach, or maybe I'm misunderstanding, but how could we combine some of these AI techniques with these spatial proteomics?

Christina Curtis: Yeah, it's a great question. It is the new frontier and I think there's a couple different ways to do that. So the beauty of the H&E is they sort of capture a lot of different tissue regions, because we're typically looking at whole slides. So one approach is to use the information in those H&E to focus in on areas that maybe variable grade that we want to have a better resolution on. And in prostate, this is of course hugely relevant because of the extensive sampling that often occurs. And so you can apply approaches such as attention learning, machine learning, AI based approaches to parse out what those regions are that are most informative and then perhaps go deeper with some of these spatial proteomic or transcriptomic approaches. That's one strategy.

The other would be to really try to co-register the H&E to the spatial proteomics or transcriptomics. And that would be done on adjacent sections. So certainly there are research groups thinking very hard about the appropriate algorithms to do that. And then the real question is in which context is that going to be most informative and where do we need all of this information? But certainly, the field's moving more towards composite biomarkers as a way to better risk stratify. And I think we're going to perhaps need different approaches in different contexts, but we really need to get to grips with these new methods.

Alicia Morgans: I could not agree more. And final question, and thank you for letting me pick your brain, since prostate is not your area and you are being very gracious to review these things. I think, and this may be the case in breast cancer, but in prostate cancer the sooner we get an effective treatment to a patient the better the patient does. And what's really interesting is that in clinical trials, even when there's crossover in a control arm, that patient population even in less advanced settings, often never catches up and there is a survival difference just from starting that effective treatment earlier despite crossover. So these kinds of techniques from my perspective, might be able to help us early on identify which patients are going to benefit, which aren't, even if we don't have something at baseline that may inform us. Is this something that you see coming down the pike? Is this something that might be possible perhaps even if we can distill it down, we can find something at baseline that can help us predict. What are your thoughts?

Christina Curtis: That's right, absolutely. And we're thinking very much about it the same way that we need to be moving therapies forward in biomarker stratified populations earlier. And of course the balance there is achieving the response while minimizing toxicity of any of these agents. So really having the risk stratification upfront. And I think this is where these kinds of biomarkers can help define those populations. And I'm a big proponent, I think there's a huge opportunity to be doing this and embedding these approaches in clinical trials. That's really where we're going to learn is to start to ask how are these agents working? Which patients are responding? What are the maybe even compensatory pathways that are causing a lack of response? So, I think it's a huge learning opportunity and we need to be moving the needle and now we have a whole new set of tools in our armamentarium to start to do that.

Alicia Morgans: Wonderful. Well, thank you again for your expertise, for your continued efforts in other cancers. At some point we'll have to pull you over into the prostate cancer and GU space-

Christina Curtis: So I'm told.

Alicia Morgans: You should be told many more times. But I really appreciate your time and your expertise today.

Christina Curtis: Thank you. Thank you for having me.