Andrea Miyahira: Hi, I am Andrea Miyahira here at The Prostate Cancer Foundation. I'm excited to be joined by Dr. Jindan Yu of Emory University. She will discuss her team's recent paper, NKX2-1 drives neuroendocrine transdifferentiation of prostate cancer via epigenetic and 3D chromatin remodeling. This was published in Nature Genetics.

Dr. Yu, thanks for joining us.

Jindan Yu: Thank you, Andrea, for the opportunity to discuss our work. So today, I'm going to talk about our study of neuroendocrine prostate cancer transformation. Here is a quick summary of what we found.

So, we found out that FOXA2, which is a protein that's very highly and specifically expressed in many neuroendocrine prostate cancers, it can function as a pioneer factor to bind to the enhancer elements of neuroendocrine genes. It will modulate DNA demethylation to open the chromatin.

Then FOXA2 will interact with another neuron transcription factor, NKX2-1. So, this protein was not expressed in the adenocarcinoma, but FOXA2 initiating effect was able to turn on NKX2-1. Then these two proteins interact. These in turn stabilize FOXA2 binding at the neuroendocrine enhancers.

So you can see this interaction was through a DNA looping, a promoter enhancer looping. And then eventually, these two proteins recruit P300, which is a histone acetyltransferase. These catalyze histone 3 lysine 27 acetylation. The promoters of neuroendocrine genes turn on a whole new set of neuroendocrine gene expression. And we found that this can be blocked by P300 inhibitors.

So, we set out to address this mechanism of neuroendocrine differentiation of prostate cancer, thinking that epigenetic mechanisms are usually fundamental to cell identity. So, in this linear transformation, or linear plasticity from prostate adenocarcinoma to neuroendocrine prostate cancer, there's this huge shift of cell identity. So we thought it would be very important to look at the fundamental epigenetics. So, we use this Hi-C technology to map the chromatin architecture in neuroendocrine prostate cancer models, and also in the adenocarcinoma carcinogenic system prostate cancer models. We identify there are many neuroendocrine, specific chromatin interactions, between promoters and enhancers. If you look at the genes that are involved, a lot of these are associated with neuroendocrine or neuron functions.

Then we expand these to many models of CRPC and NEPC PDX models, and also our cell line model, to map the chromatin architecture in these cells. And you can see clearly there's two clusters. One is CRPC, and there's a separate cluster of neuroendocrine indicating that there's major differences in the 3D chromatin architecture between these two kinds of subtypes.

Similarly, we mapped a whole genome methylation profile of these PDX models. And again, there's a 2D level. We found that there's a huge difference in the epigenetic landscape of DNA methylation. All these data are available in the GEO database. We put it there and hope it'll be helpful for people who are interested in this question.

In order to understand the mechanism how these huge shifts of epigenetic landscapes happened, we built a model by over-expressing FOXA2 in the LNCaP cells, which is a typical androgen-dependent prostate carcinoma cell line. You can see over 28 days these LNCaP cells get transformed to demonstrate a more neuron-like, with some neuron type of features in these cells.

Looking at the gene expression, we see a complete transcriptional reprogramming from these luminal genes to the upregulation of many neuron genes, indicating neuroendocrine transformation. If we compare these cells with the known neuroendocrine prostate cancer or adenocarcinoma, so these are different CRPC, PDX models with different subtypes and AR-positive, the different subtypes. We can see that our early stage of LNCaP cells, they are luminal, being clustered with typical AR-positive and AR-negative PDX tumors. And our day 28 cells are closely clustered with a neuroendocrine subtype of CRPC PDX models. We actually also made many additional models using different genetic background as shown here. If anyone is interested, we're happy to share these models with the field.

And we also want to show the importance of the histology. When we put these cells into the mice, you can see first of all these LNCaP cells with FOXA2 overexpression. So these are the neuroendocrine model. It grows much faster than LNCaP cells itself in mice. And then the cells exhibit a small cell carcinoma phenotype compared to adenocarcinoma. This is a small population of the cells, showed a small cell carcinoma phenotype. So, it is really a mixed histology, which is very similar to what we see in patients. But I want to point out in the FOXA2, the NE model, although a lot of these, morphological-wise, they still look like adenocarcinoma, they are mostly SYP positive.

So then, one major mechanism for, as I mentioned earlier for FOXA2 to drive a neuroendocrine transformation is really by functioning as a pioneer factor to initiate the neuroendocrine gene enhancer activation, and some critical new neural transcription factors. One of these we identified is NKX2-1. So you can see over the time goes from two days all the way to 28 days, FOXA2 gradually binds to the enhancer elements of NKX2-1 gene to turn on this expression of this NKX2-1 gene. And when we knock down NKX2-1 in this model, as you can see here, this NKX2-1 expression over the time course is gradually turned on. If we knock down NKX2-1 expression, we no longer see this neuroendocrine gene expression. So, suggesting that it is required for the FOXA2 to drive a neuroendocrine transformation.

Looking at the epigenetic level, so in the controlled cells you see the shift of monomethylation, which indicates enhancer priming from the adenocarcinoma to the neuroendocrine enhancers. This is totally blocked when we knock down NKX2-1, indicating this mechanism involving a neural transcription factor that is bound to the promoter forming loop with FOXA2 that is bound to enhancer altering these 2D epigenetic landscape, and the 3D chromatin architecture that is important for neuroendocrine transformation of prostate cancer.

Then I also want to show another really important action point in this neuroendocrine transformation process. Eventually, we need the neuroendocrine genes to get expressed for the neuroendocrine prostate cancer cells. Here we show that the P300 CBP are really important, because these two proteins directly interact with P300 to recruit them to the neuroendocrine enhancers. Shown here are the data. Basically, when we knock down P300 or CBP, we have a similar regulation with gene expression as FOXA2 and NKX2-1 knocked down. They are critical to maintain this neuroendocrine gene expression.

So these give us an opportunity to target P300 CBP in these neuroendocrine cells as a way to block a neuroendocrine gene expression. So we use the CCS1477, which is a P300 inhibitor that has been in clinical trial, as you can see either reduces the expression of many of neuroendocrine genes. And consequently, we see this is done by the blocking the H3K27 acetylation, and these neuroendocrine enhancers. So mechanistically, [inaudible 00:08:57] this epigenetic mechanism.

And then looking at tumor growth in vivo, you can see this P300 inhibitor really reduced a neuroendocrine prostate cancer cell tumor growth in our model, and also in a PDX model.

Andrea Miyahira: Thank you so much, Dr. Yu, for sharing this study with us. So what is the role of FOXA2 in normal prostate cells?

Jindan Yu: Thank you, Andrea, for the great question. So in the normal prostate cells, the normal prostate FOXA2 is usually not expressed. So it's not expressed in the majority of prostate, it only expressed in the neuroendocrine cells, which is very small population of the entire prostate.

Andrea Miyahira: Okay, thanks. And there have been several NEP subtypes reported such as NeuroD1, NeuroD2, and ASCL1 driven. Was this FOXA2 pathway found to be specifically driving any of these subtypes?

Jindan Yu: That's a great question. So, we did a look at NeuroD1 ASCL1, a number of these gene expressions. So, what we found is NeuroD1 was not upregulated in our FOXA2 induced neuroendocrine transformation like LNCaP cells. For NeuroD1, the cell was not upregulated.

Another the same for ASCL1. So we think NeuroD1, ASCL1, they might have a similar function as FOXA2. This is a hypothesis we have not yet tested, but we think NKX2-1 as a neuro changer factor that binds the promoter is able to cooperate with a number of different factors, or enhance the binding of transcription factors. So we did look deeper into the NKX2-1 cooperation with ASCL1. There was another study about ASCL1 from Matthew Freedman's lab.

So we look at our supplementary data, we show that NKX2-1 can also cooperate with ASCL1 to drive this neuroendocrine transformation subtype. We put NKX2-1 into some of these not fully transformed models, and it was able to drive this process.

So that the short answer is NKX2-1 as a promoter-bound transcription factor might be able to cooperate with different enhancer-bound lineage factors to mediate this process.

Andrea Miyahira: Okay, thanks. And in prior trials testing P300 CBP inhibitors such as CCS1477, were there any patients with NEPC that were treated, and are there clinical responses compared to other patients known?

Jindan Yu: Again, that's a very important question. So that's also something we are very interested. So, in the field, there's couple of P300 behind inhibitor that's in clinical trial. So CCS1477, which is more for metastatic NEPC. And there's another FT-7051 and which is also in phase 1B 2A trial. It did pretty well in first phase 1A trial. So, they are specifically looking at a CRPC without neuroendocrine feature.

So for the CCS1477, they did not really select or exclude neuroendocrine prostate cancer. So I don't know how many of their patients might have neuroendocrine features in their CRPC patients. It could be. So they might have treated on some CRPC patients with NE features. But most of their patients should be the CRPC. So, it seems right now these P300 behind the inhibitor have not been specifically tested in any PC patients.

Andrea Miyahira: Okay, thanks. And based on these data, what do you think might be the optimal therapeutic targeting strategy for preventing or reversing NEPC?

Jindan Yu: That's a very important question. I think we are definitely looking into that. So what we found for the P300 inhibitors? It targets H3K27 acetylation. So it will also reduce H3K27 acetylation at some of the housekeeping gene promoters, for example. So, the therapeutic opportunity is really the P300 inhibitor are targeting a lot of CP enhancers. I have shown you neuroendocrine enhancers and the adenocarcinoma enhancers, or luminal enhancers. And so a lot of these kinds are addicted to these enhancers, like SEMA3C, AR, and those enhancers. So that's really the opportunity.

But then on the other hand, because it's targeting P300 that's expressed in almost every cell, so you have to also consider potential toxicity. So, although there's a preferential targeting of the enhancers, but there could be toxicities when targeted broadly. So, I think there's a very important couple lines of research to figure out, is really what's the therapeutic window, what doses you use so that you kill the tumor cells without causing major side effects. And also, maybe potentially using combination therapies to help using a lower dose to kill the cancer cells.

And also, as you probably alluded to, is like, what's the timing to do this? So we're using our model. So we have a very nice time course model. So we are right now looking at using P300 at a different time point of this progression. So, this neuroendocrine transformation process, a different time point. If we give the P300 inhibitor, how sensitive these cells are. Also, earlier you mentioned the different subtype of neuroendocrine prostate cancer or CRPC in general, there's a number of different subtypes. So, can we give them P300 inhibitor to see what kind of genes it's targeting in the different subtypes? Because P300 inhibitor is going to target it. Luminal enhancers as well, like AR, SEMA3C. In our study, which is also targeting the NE enhancers. So, in different cells it's probably inhibiting different genes. So, a lot to figure out.

Andrea Miyahira: Definitely. And what are your next steps in these studies?

Jindan Yu: Yeah, so I already mentioned in terms of P300 inhibitor, we really want figure out in different CRPC subtype, or different cell lines, more the PDX models. What are the target genes or pathways of the different P300 inhibitors?

Then the other is, can we actually develop a neuroendocrine signature to predict this CRPC patient is starting to develop neuroendocrine features, and it's become more a deadly disease. Maybe that's the opportunity or that's a time we should give a higher dose of P300 inhibitor. We want to be a little bit more aggressive in the treatment, or add a P300 inhibitor in addition to AR pathway inhibitor. In those patients, maybe the combination will do better. But for a regular CRPC patient, maybe the AR pathway inhibitor is sufficient to do a good job. So, adding P300 inhibitor might not be worthwhile considering potential side effects. So, we really want to identify this signature, and you can see in our model we find that there's a lot of epigenetic change that are happening before the gene expression happened. So we want to look at these epigenetic signatures to develop either based on tissue, or based on blood. So self-redeeming to predict a neuroendocrine transformation of CRPC.

Andrea Miyahira: Thank you so much, Dr. Yu, for coming on and sharing this study with us today.

Jindan Yu: Thank you, Andrea, for the opportunity.