Read the Full Video Transcript

Andrea Miyahira: Hello, I'm Andrea Miyahira here at the Prostate Cancer Foundation. Today I'm joined by Dr. Max Loda, a professor at Weill Cornell Medicine. Dr. Loda's group recently published the paper "Distinct Mesenchymal Cell States Mediate Prostate Cancer Progression" in Nature Communications. Dr. Loda, thank you for joining me to present and discuss this paper today.

Massimo Loda: Thank you, Andrea. Thank you very much for the opportunity. We started this work several years ago on the role of the microenvironment in prostate cancer initiation, progression, and metastasis. This is work done by three co-first authors, Hubert Pakula in the first place, Mohamed Omar, and Ryan Carelli.

It's well known, it's been well known for a while, that the stroma is very important in determining both the genesis of the prostate and in carcinogenesis. In very old experiments by Gerry Cunha's group, it has been shown the urogenital sinus is required to generate tumors when recombination occurs in the mouse, and as you can see here, the tumors with the urogenital sinus are formed, whereas those with immortalized cells alone do not. We undertook a study several years ago in human tumors where we microdissected the epithelial and stromal component, and to make a long story short, what we showed is that the stromal component had significant changes from benign to prostate intraepithelial neoplasia to invasive cancers, and that these were actually suggestive of a bone-like environment, which we'll get back to later. Shown here, there are changes that occur only in invasive tumors versus the others.

When we distilled the signature of genes, the majority of which were stromal and only one was epithelial and a target of many of the stromal genes, in a separate dataset, this signature predicted metastasis and lethality in patients. So this suggested to us that we should look at this more carefully. What we did is we took four mouse models representing the disease progression from no epithelial phenotype, all the way to neuroendocrine differentiation, and did single-cell RNA sequencing on roughly 9,000 stromal cells in over 40 mice and compared those to human tumors, in which we did the same thing, single-cell RNA sequencing.

What were our findings? So first of all, there is an increase in the stromal component as we go from the less aggressive to the very aggressive neuroendocrine tumors, as seen here. When we look at single-cell RNA sequencing using gene expression proof, single-cell RNA sequencing, and several algorithms, we identified eight clusters that were separate in terms of their gene expression profiles. Three of these were common to any genetic alterations, and the rest were specific for genetic alterations that occur. Seen here is how these clusters are placed compared to wild type and mutants, wild type being on the right, of these mice, and you can see that there's an enrichment and changes in these stromal cells as we introduced genetic mutations in the epithelium. We also did interactions, ligand-receptor interactions in a virtual manner using single-cell RNA-seq data and came up with significant interactions between stromal cells and epithelial cells and between immune cells, stromal cells, and epithelial cells, and these are very interesting, and I'll touch upon very briefly on a couple of them.

So these clusters are common to all genetic alterations, shown here in this UMAP. If you look at these clusters, the expression of genes is highest. There's many genes that are high in this expression in terms of expression level. You can see in cluster zero there's mostly myoepithelial cells, but in cluster one, there are many genes that are associated with an innate immune response. If you look at these by expression, immunofluorescence, you can see that these genes are located predominantly in the stroma of, in this case, neuroendocrine tumors versus virtually none in the stroma of the corresponding wild type.

Then we look at signaling networks. These are complicated circle plots, but you can see that there are communications between the epithelial luminal cells, some of the normal basal cells, and transformed luminal cells as well as towards these clusters that I pointed out. So thrombospondin and macrophage inhibiting factors are two examples that are very prominent in these clusters. Very interestingly, the composition of the immune set is different in the different models. So for example, in the TMPRSS2-ERG and Hi-MYC, which are on the earlier side of the spectrum, the majority of cells are cytotoxic T lymphocytes and NK cells. In contrast, the PTEN loss and the neuroendocrine tumors are mostly composed of monocyte macrophages.

There are clusters, as I mentioned, specific to genotypes, and these are them. We don't have to go over this dot plot, but just to show you that there are genes that are preferentially expressed in these clusters. These are mapped here. On the top, in light blue, are the two AR-positive clusters, and in dark blue, in the center, are those that are associated with neuroendocrine tumors. Importantly, you can see here that the top clusters are AR-positive. These are stromal cells, I'll remind you, whereas in neuroendocrine tumors, the clusters are for the most part either AR or low-AR-negative, and these AR expressions are completely opposite of the expression of another gene called periostin, which we feel is associated with neuroendocrine differentiation.

This is shown here in situ, the areas of neuroendocrine differentiation in the mouse. So in the human, they are completely surrounded by tumor cells that express periostin, so a potential target. This is corroborated by other GEM models that give rise to neuroendocrine differentiation. So there's a strong anti-correlation of androgen receptor in periostin associated with neuroendocrine differentiation.

This is probably the most important functional experiment. If we take epithelial cells with genetic alterations and co-culture them with fibroblasts, these fibroblasts then get transformed in a sense to express a gene set that is similar to the one found in mouse. So you can see here the corresponding gene sets in these cultured fibroblasts with two genetic alterations, TMPRSS2-ERG and MYC amplification, that really overlap with the clusters found in the mouse themselves. So the genetic alterations determine how the fibroblasts behave and create an environment that is important for their survival and progression. So seeing here are the clusters one and three, and seen here are the clusters three, six, and seven that are associated with neuroendocrine tumors or PRN.

Importantly, this is conserved in the human. So in the human, we only had one genetic alteration, TMPRSS2-ERG, but you can see that the cluster in red is conserved in the mouse and in the human. This is very important because now this has translational implications for patients.

Finally, what you can see here is that the same clusters that we find in tumors that are advanced are found in advanced metastatic sites in humans. And you can see here the cells that are expressed in the mouse and in the human are similar.

If you use these biomarkers, you can very strongly, as we showed five years ago in the other paper, predict metastasis, and these are purely stromal genes, and it's independent of [inaudible 00:10:53]. So again, we have a prognostic signature that is quite important.

So this is a summary slide. I don't have to go over it, but we went from mouse, to single-cell RNA sequencing, and then a direct comparison between human and mouse, and came up with stromal markers that can be predictive of lethal disease.

Andrea Miyahira: Thank you so much for sharing that with us.

So my first question is, did you identify any targetable epithelium-stroma or immune interactions in prostate cancer?

Massimo Loda: Yes. So I glossed over very quickly in the presentation, but the fact that some of the genetic alterations attract a certain type of immune cells, like macrophages and monocytes, is a significant finding, whereas other alterations attract different types of cells. Within these cells, these circle plots identify potential markers that are targetable. I mentioned thrombospondin and macrophage-inhibiting factors as just two examples of many. In the paper, there are extensive tables of ligand-receptor interactors that are predicted bioinformatically that need to be followed up by functional analysis.

Andrea Miyahira: Okay, thank you. And did you perform any temporal analyses to see how these stromal or immune populations and their activities changed during prostate cancer development and progression in these models?
Massimo Loda: So these are not in the paper right now, but we were doing this concomitantly. So we have, for instance, at the present time, the high-MYC mice were followed and analyzed by single-cell RNA sequencing at three, six, and nine months. We did the same for another model that is not in the paper, the TRAMP model, that evolves towards neuroendocrine differentiation, and we are finding that things do change over time. Even in the wild type, the stroma changes over time, which is a finding that has been found by others before. So it's a dynamic environment that needs to be analyzed over time, but there is enrichment, if you will, as you go from early lesions to late lesions of certain clusters and certain ligand-receptor pairs that predict lethal disease. But these are ongoing studies, and we will talk about these soon again, hopefully, as the studies evolve.

Andrea Miyahira: I look forward to that.

Did you look at any single-cell profiles of stromal or immune cells from metastatic lesions and from metastatic lesions from different sites?

Massimo Loda: Yes. So we are looking right now in metastatic lesions in the lung and the liver of mouse models that actually have metastasis. As you know, metastasis to the bone is not frequent in mouse models, and particularly in these models, they do not occur almost at all. We have one or two.

So those studies are ongoing, but in the paper, we compared very deeply the stromal metastasis of human cases that were published in a Cell paper three or four years ago now. That paper addressed only immune cells at that time, but the comparison between the stroma of metastatic cases in human and the stroma that is found in the more aggressive mouse models was significant, and that is an area that we're exploring actively because finding, of course, markers in the stroma that are novel targets. I'll emphasize that most of the targets, if not all of the targets we have these days, are epithelial, and some, of course, are immune, but the fibroblasts, myofibroblasts stromal component has not been targeted as of yet. So we are identifying novel targets, but what we do want to do is make sure that what we found in the mouse is conserved in the human and both in expression and in site. So the advantage of the human case is that these are bone mets, which is, of course, the most frequent site of metastases in prostate cancer.

Andrea Miyahira: Thank you. And I think that leads nicely to the next question, is that what mechanisms do you think govern the possibility that the mesenchymal bone signatures in primary tumors that you're seeing may promote bone metastases?

Massimo Loda: That's probably, as you've mentioned, the most interesting question that is as yet unanswered. So either there are cells in the fibroblast, myofibroblast population that are there and are recruited and expanded as a result of tumor formation or transformation of the epithelial cells, or the transformation of epithelial cells and genetic alterations drives transcriptional changes in the stromal cells that then form an environment that is commensurate to their state. So in tumors that do not metastasize, they have a certain microenvironment there. In tumors that metastasize, what we think is happening is that the epithelial cells create an environment that is similar to that of bone, and we've shown that in the paper by many genes that are expressed in the fibroblast in the primary setting that in general are expressed in osteoblast or stromal cells in the bone.

So either it's a clonal expansion of preexisting cells or an induction, as some of our data suggests, of run-of-the-mill fibroblasts that then become specialized in a certain way. We don't know that, but we're following and trying to answer that question with mouse models that are clonal and with promoters that will help us follow their fate over time. Again, we don't have the answers to that yet, but these experiments are ongoing.

Andrea Miyahira: Well, thank you so much for joining us and sharing this work with us today.

Massimo Loda: Thank you, Andrea, very much for the opportunity.