Read the Full Video Transcript

Andrea Miyahira: Hi, everyone. I'm Andrea Miyahira at the Prostate Cancer Foundation, and thank you for joining me today. With me is Dr. Abhijit Parolia, an assistant professor at the University of Michigan. He will present his group's latest paper, "NSD2 is a Requisite Subunit of the AR/FOXA1 neo-Enhanceosome in Promoting Prostate Tumorigenesis." This has been released as a pre-print in Bioarchives and is under review. Thanks so much for taking the time to be with us today, Dr. Parolia.

Abhijit Parolia: Great. Thank you, Andrea, for that kind introduction as well as the invitation to come and share this new work from our group. It really centers on studying the mechanistic details of AR's functional duality in normal prostate epithelium as opposed to the cancerous or malignant variant of prostate cancer cells.

Let me begin by introducing you to the normal prostate. It is a walnut-shaped organ that sits at the base of the bladder and microscopically comprises glandular structures, predominantly made up of two types of epithelial cells: the outer basal epithelium and the inward lumen-facing epithelial cells called luminal cells, which distinctively express the androgen receptor protein.

Now, androgen receptor protein activity in these luminal cells is absolutely critical. That terminal differentiation, moving from these luminal progenitors, shown in red, to this terminal state, shown in blue, is where AR really drives their proper maturation into a non-dividing state and rather promotes their secretory luminal functions in producing that prostatic fluid. Now, the androgen receptor primarily does so by binding to its ligand, testosterone or the primary male hormone. Once internalized, it gets converted into DHT and binds the AR, which triggers AR dimerization. Now, this homodimer translocates into the nucleus and binds to the DNA in a sequence-specific manner. These are primarily called androgen response elements and comprise palindromic sequences containing two half AR sites oriented in an inverted conformation. Now, like most transcription factors, the androgen receptor doesn't work in isolation. It rather works in concert with a host of epigenetic and chromatin remodeling proteins that regulate its activating or repressive transcriptional duties. So in the normal epithelium, AR really drives a non-dividing secretory luminal cell state.

In contrast, and actually very intriguingly, in cancer that originates from the prostate epithelium, AR is predominantly expressed in almost 90-95% of that primary disease. These two seminal studies back in the 1940s realized that if you remove androgen or outcompete it by estrogen administration, you trigger dramatic atrophy of the malignant prostate epithelium. And ever since, androgen deprivation therapies following radiation or prostatectomy of the primary disease have been the mainstay of therapy in the clinical management of the disease. You see these profound cyclical patterns of disease remission with either ADT in the form of castration or second-line therapies like enzalutamide or abiraterone that directly bind to AR in the ligand pocket and outcompete DHT's activity and, therefore, repress AR function.
So at least in the malignant context, in contrast to the normal epithelium, it appears that oncogenic AR really drives the hyperproliferative, invasive, and survival properties of the malignant prostate epithelium. This brings me to the primary thesis of this project: how do we explain, mechanistically, the profound antitumor effect that we see in prostate cancer cells upon inhibiting this terminal differentiation factor?

One of the primary implications, therefore, that the wider field has also started believing and interrogating, is that AR activity gets dramatically reprogrammed likely from its terminal differentiation axes in the normal prostate epithelium to more of an oncogenic activity once the prostate epithelium is transformed. Expanding on these ideas, the seminal papers from the Pomerantz, Freedman, and Zwart laboratories suggest that somatically altered or cancer-specific epigenetic or chromatin remodeling factors could likely be involved or implicated in reprogramming AR once this transformation occurs.

In this state, we hypothesize that there are two modes or forms of the AR protein: one in the normal epithelium that drives a nonproliferative state, and the second, an oncogenic variant that sustains survival and metastatic and invasive properties of the malignant prostate epithelium.

To identify these subunits or factors that could be implicated, we have followed various orthogonal approaches. One unique approach we have taken is synthesizing these reporter cell lines, which we refer to as the endogenous AR reporters. Endogenous is critical because we hope to capture epigenetic regulation, in this case, of the androgen receptor protein. So we went into the KLK locus in LNCaP epithelial cells or prostate cancer cells, and we engineered the KLK3 locus, KLK3 being a cognate AR target, to incorporate right downstream of the KLK promoter an mCherry construct fused by a P2A endopeptidase sequence to the KLK3 gene.

Now, this allows us to use mCherry as a cognate AR target gene. We treat these cells with a focused epigenetic guide RNA library, incubate them for about 5 to 7 days, treat them with DHT, and then FACS sort them into mCherry high and low populations. We then look at the guide RNA sequences, and if you rank these enrichment profiles from the low mCherry cells to the high mCherry cells, you find that, amidst other known co-activators of AR like BRD4 and TRIM24, NSD2 guide RNAs are enriched in the mCherry-low population. This suggests that NSD2 is actually a co-activator of AR activity in prostate cancer cells.

Now, what is NSD2? It's actually a K36 dimethyltransferase. It is primarily involved in the activation of gene expression, and this is thought to happen by antagonism of the PRC2 or EZH2 repressive duties that deposit the K36 trimethyl mark on the histones of the nuclear zones. It belongs to this NSD2 family or NSD gene family, where NSD2 actually notably templates two splice isoforms. So it's the same gene producing two splice isoforms, with the long form having the SET domain, which is the catalytically active variant.

Now, intriguingly, and this, again, builds on some of the primary work from the primary reports from Cory Abate-Shen group, where they've actually gone in and suggested that NSD2 shows this increase upon transformation. But what we've done in this paper is we've connected a unique cohort of prostatectomy specimens, and here in a patient-matched setting, you can actually appreciate that the adjacent or the normal tissue proximal to the cancerous lesion has very little expression of NSD2, but in the same patient, the malignant lesions acquire NSD2 expression. So, really, NSD2 is expressed upon malignant transformation. And here you can see a summary from about four independent patients. We've done this in a TMA cohort as well using IF, which is a very sensitive approach where you can see AR is expressed in the benign epithelium as well as the malignant forms of prostate cancer, but the NSD2 protein is almost undetectable in the normal epithelial cells and is only expressed when these cells are formed either in the primary prostate cancer or the castration-resistant variant of the metastatic form of this disease.

In terms of functional validation, we went ahead and took these parental lines, and if you knock down both the long or the short isoform, you can actually see a downregulation of KLK3, which validates a CRISPR screen. But more importantly, on the right, I actually show you some models that we have CRISPR-engineered, and you can appreciate either loss of both the long and the short isoform or the long isoform alone, which is, again, the catalytically active variant. You see a marked loss in the activity of AR target genes or the expression of AR target genes.

Now, notably, what's really impressive to note is that the AR protein levels themselves do not change, but its ability to activate or drive the expression of its cognate targets is dramatically attenuated in the absence of the catalytic activity of NSD2. And even moving away from these select targets in a gene expression transcriptomic assay in RNA-seq, you can see a wider AR-activated gene set to be dramatically blunted in its expression upon the loss of the NSD2 protein, which parallels a loss in the proliferative signature and some of the other oncogenic hallmarks that I will share in subsequent slides.

But more impressively, given that we actually see no change in the androgen receptor protein itself, we wanted to see if there are any changes in the binding of AR to the chromatin. And here we were in for a surprise. We actually found that in the wild-type NSD2 cells, AR bound a sizable cistrome, but when you remove NSD2, despite no change in the levels of the AR protein, we saw this dramatic downsizing of the androgen receptor cistrome, with almost 65% of the sites being absolutely wiped clean in cells that lack the activity of the NSD2 protein. Now, this is actually true even when you look at the heat map. There's absolutely no binding of these elements while AR continues to bind just fine at the independent sites. And this is what really led us to now divide AR binding elements into NSD2-dependent and independent compartments.

When you now interrogate FOXA1 binding, it remains just intact. NSD2 doesn't regulate the binding of FOXA1, but in the absence of AR loading, you actually see a dramatic attenuation of the activation mark in the form of K27 acetylation, really suggesting that in the absence of NSD2, the dependent elements are decommissioned as enhancer sites in these NSD2-null prostate cancer cells.

Now, if you look at the motifs that really underlie these elements, we really wanted to distinguish what is the distinctive feature of dependent versus independent sites. So when you interrogate just the independent elements, you can actually see the recurrence of these full palindromic AREs that I introduced to you earlier, where you see the palindromic sequence. But this sequence wasn't evident when we looked at the dependent cistrome. The top-ranking motif was, in fact, the FOXA1 AR half-motif, where the FOXA1 DNA sequence is juxtaposed to only one half of the AR sequence or the AR motif. Now, this is really intriguing and really suggests that NSD2 is perhaps essential in AR binding at these chimeric AR half-sites, and this is perhaps more evident in these read-density plots where you can appreciate that the NSD2-independent site remains unaltered while in cis-proximity and chromosome 10, you see a complete removal of AR binding at this site. And if you interrogate the motifs, you actually see the NSD2-dependent site has the FOXA1 AR half-element while the NSD2-independent site comprises these full palindromic AREs.

So is there any relevance when it comes to these motifs being detected in primary prostate cancer patients? To really address this question, we went back to Mark Pomerantz's studies from Matthew Freedman's lab where we downloaded the primary AR cistromes generated from patient tumor biopsies. We defined normal CRPC and primary specific AR binding sites, and when you now interrogate what motifs are specifically enriched in the primary cancer versus the normal epithelium or mCRPC versus the normal epithelium, you realize that the FOXA1 AR half-element or the AR FOX, so this is the inverted orientation, are specifically and markedly enriched in the cancer epithelium.

Another notable feature is that the full palindromic AREs, if anything, showed slight depletion upon progression to a malignant state. So that's really exciting and suggests that the cancerous form of AR binding really happens at these chimeric sequences as opposed to the full palindromic motifs.

With the loss of NSD2, therefore, you see this marked attenuation in proliferation, you see a loss in the invasion or migration ability, as well as a loss in the ability to form colonies as well as graft in mice, all of this being hallmark cancerous phenotypes. So what we really believe is happening is that NSD2-deficient cells are resembling more a normal epithelium as opposed to a cancerous state.

Now, all this was well and good, but we've always had a translational focus in our group, so our obvious question was, can it be therapeutically targeted in advanced castration-resistant prostate cancers? And in the interest of time, while I do not share the entirety of how we worked on the compensation and so on and so forth, we realized that NSD1 plus 2-targeting seemed to be the most optimal way of inducing cytotoxicity. And to really be able to make pharmacological agents that could allow us to trigger dual degradation of both of these isoforms, we actually leveraged this warhead that was published in 2021 by a group and tried to, with our chemist collaborators, convert them into these powerful PROTACs that would bind to these proteins of interest and induce their proteasomal degradation, thereby kind of degrading these and affecting a therapeutic effect.

We synthesized almost upwards of 150 odd derivatives and arrived at this particular dual compound, which we call LLC0150. It primarily engages both NSD2 and 1, we believe, through the WWP1 domain, which is found in the extreme N-terminal half, and you can appreciate in these Westerns, actually triggering a considerable loss of both NSD1 as well as 2 without altering NSD3 expression. And in the cancerous cells, when you treat the VCaP cells, in this case, with LLC0150, you induce a marked loss in the expression of AR MYC as well as AR target genes while inducing c-PARP accumulation, which suggests that these cells are experiencing some sort of apoptotic cell death or stress.

We've now gone on to profile this dual compound more comprehensively in a larger panel of about 122 normal and cancerous human-derived cell lines. Here we find that while the NSD2-mutant forms of cancer—mostly heme malignancies where NSD2 is rearranged or mutated in an activating fashion in almost 15-20% of the cases—are, as expected, the most sensitive. But this is immediately followed by AR-positive forms of the prostate epithelium, shown in red, and, to our surprise, even AR-positive breast tumors, which is really reassuring and distinguishes quite dramatically from the side of toxicity seen in normal cells, shown in gray, or even AR-negative forms of the tumor, shown in blue.

So with this, we are quite excited about reaching the conclusion that there are two modes in which AR interacts with the chromatin. The predominant mode occurs at the palindromic full AREs, perhaps in the normal epithelium, where AR is able to occupy these sites even independently of NSD2 function. But upon transformation, with the gain or the ectopic gain, rather, of NSD2, AR gains access to a larger cistrome comprising these FOX AR half-motifs that we believe constitute the AR neo-enhanceosome, or rather its cancer-specific complex, which amplifies or wires its oncogenic activity, driving proliferation, metastases, and other cancer hallmark phenotypes.

With that, I'm happy to introduce you to the team. Arul, my primary mentor, is a critical player, as well as Irfan, with whom we collaborated at UPenn, who has done a lot of the interactome work that I did not present due to time constraints. In my lab, primarily two individuals, both of them graduate students, Sanjana and Yihan, have significantly contributed to this work. So I'm happy to take any questions you may have, Andrea.

Andrea Miyahira: Thank you so much for sharing this really exciting study. What causes NSD2 upregulation in prostate cancer, and do you see any genomic alterations?

Abhijit Parolia: Right, that's a very interesting question. Actually, that's the first thing we examined in our genomic datasets. We see no amplifications of the NSD2 gene. In heme malignancies, there are actually activating hotspot mutations in the SET domain. We don't see those in prostate cancer. There is one group—there's a JCI paper, I forget the lead investigator—they projected that AKT phosphorylates NSD2, causing stabilization, but that is still, I think, an open question and very much of interest to our group in terms of what really leads to aberrant or this foreign expression of NSD2 in the cells. I can tell you it seems to be more through protein modifications rather than mRNA at this point.

Andrea Miyahira: Okay, thank you. And have you evaluated what fraction of prostate cancers, including across different stages, are driven by NSD2?

Abhijit Parolia: This is very intriguing. There are two reasons for this. I think, prior to this, there have been a couple of publications focused on NSD2. So we are not the first to highlight this protein, but they focused on its expression in the neuroendocrine or the AR-negative form of the tumor. Objectively, if you rank its expression in all forms of normal or malignant prostate tissue, neuroendocrine prostate cancer has the highest expression of NSD2. It almost mirrors what happens to EZH2 expression in the prostate epithelium. So there is certainly evidence that NSD2 plays a role in the neuroendocrine tumor, which is not something we are interrogating actively, but I know Michael Shen's group is working on it.

In terms of our understanding, almost every form of the normal or primary tumors that express AR seem to also express NSD2. So I would speculate that all forms, AR-positive as well as AR-negative, express NSD2. However, its function may evolve from supporting and unlocking the oncogenic activity of AR in the primary transformation to antagonizing, perhaps, EZH2 in the neuroendocrine form of the tumor, where EZH2 also takes on a larger role in terms of its hyper expression.

Andrea Miyahira: Okay, thank you. What are NSD1 and NSD2's functions in normal tissues? Do they interact with AR in normal contexts?

Abhijit Parolia: NSD2 is very easy; it's not even expressed in the epithelium in the normal tissue that we have looked at so far. We are still trying to use more antibodies to validate that with IF, which is seemingly the most sensitive methodology. We are also trying to do it at the RNA level just to be sure in terms of the production of the active transcript. So in terms of NSD2, I don't speculate that to be a normal function. For NSD1, we are in the process of validating some of these antibodies for IHC. At this point, I don't know if it's expressed in the normal epithelium. But I can tell you that it's thought to be the primary NSD2 paralog that competes with EZH2 and its repressive functions. So I wouldn't be surprised if it has some role in the normal epithelium in that capacity.

Andrea Miyahira: Okay, thank you. And what cancer features are activated through the NSD2-dependent chimeric AR-binding sites?

Abhijit Parolia: Actually, the first thing we thought when we arrived at this protein was, "Oh, if we now inhibit it, we are going to see acute cytotoxicity." And we don't. We have never actually seen cytotoxic effects from NSD2 inhibition alone, and that's what led us to really trying to understand if that was paralog compensation, perhaps. And that's what really made us arrive at this dual PROTAC of 1+2.

But one thing is very, very certain. NSD2 loss alone certainly strips prostate cancer cells of the primary malignant phenotype. So they grow dramatically slower, they do not invade at all, there's absolutely no ability to migrate and invade, and this might be something that NSD2 wires across epithelial tumors, it appears, and they certainly do not graft in mice. So all of that makes us believe that perhaps cardinal hallmark features of a malignant cell are being wired through NSD2's redistribution of AR, but survival is not one of them. So you have to hit NSD1 plus NSD2 to induce cytotoxicity.

Andrea Miyahira: Okay, thanks. What are your translational plans for the NSD1/2 PROTAC? And do you foresee any negative consequences to NSD1/2 blockade?

Abhijit Parolia: Right. Honestly, at this point, we only have a dual compound. LLC0150 has very little in vivo bioavailability, which we are trying to improve, and solubility as well. So we've been a bit hampered in our ability to test its tolerability in mice in terms of safety. Another issue is this variant doesn't bind mouse orthologs, so being able to perform these studies in mice has been a bit of a challenge. That said, we are actively working with our medicinal chemist collaborators. We are doing a lot of SAR studies to improve on this compound and perhaps arrive at a more drug-like variant that we can inhibit in the in vivo context to address those questions.

Andrea Miyahira: Okay. Well, thank you so much for sharing this, and good luck with the rest of your studies.

Abhijit Parolia: Thank you so much, Andrea. I really appreciate the invitation. Thank you.