Read the Full Video Transcript

Andrea Miyahira: Hi everyone. I'm Andrea Miyahira at the Prostate Cancer Foundation. Joining me is Dr. Estefania Labanca, an instructor at MD Anderson Cancer Center. She will discuss her team's recent paper, "Integrative Molecular Analyses of the MD Anderson Prostate Cancer Patient-Derived Xenograft Series," published in Clinical Cancer Research. Dr. Labanca, thank you for joining me and sharing your work with us today.

Estefania Labanca: Thank you so much for having me. I'm very excited to share our work. I would like to start by highlighting some of the contributors to this work. In particular, Dr. Anselmino as the co-first author of this work, Dr. Yu Chen from MSPCC, who had substantial insights into this project and is co-corresponding author, Peter Shepherd, who is the program manager of the MDA PCa PDX Program, and importantly Dr. Navone, who initiated these efforts and has led this work all the way.

This work builds upon the establishment of the MD Anderson Prostate Cancer Patient-Derived Xenograft, or MDA PCa PDX Program, that Dr. Navone started over 30 years ago in efforts to address the challenge in modeling prostate cancer, which remains lethal, particularly when the disease metastasizes to its dominant site, the bone.

In this program, we take samples from patients and inoculate them subcutaneously in mice. Then we serially propagate these models until they're established. To date, the collection includes over 150 PDXs encompassing the clinical spectrum of potentially lethal prostate cancer. From those 150 PDXs, for this published paper, we selected 44 derived from 38 patients and performed a typical characterization by whole genome, targeted, and RNA sequencing. Importantly, these analyses were done in representative samples from the same tumor, which allowed us to integrate the analysis.

This is just to show you how we account for the heterogeneity of the disease that is typically seen at the morphologic level with different morphologic subtypes. We also clinically annotate these models based on the treatment statuses that the donor was subjected to, and we collect and establish samples from primary as well as different metastatic sites, including a good number, 20%, derived from bone metastases. The collection also has samples established from circulating tumor cells and paired samples from different areas of the same tumor, allowing us to study heterogeneity, intra-tumor heterogeneity, and samples from the same patients before and after therapy, allowing us to perform longitudinal studies of the progression of the disease.

We also observed heterogeneity at the genomic level when we analyzed the sequencing data. But what we see is that the MDA PCa PDXs reflect the genomic alterations or the genes and pathways that are typically altered in the clinic. In particular, the main drivers are seen highly represented: TP53, PTEN, RB1, and AR. We also detect the fusions that are seen in prostate cancers such as TMPRSS2 and ERG in a good number of models.

When we analyzed these samples transcriptomically, we observed that they cluster very well based on their morphology, so at the histologic level. So overall, the aim of this platform is to provide a resource to complement publicly available datasets in order to make robust hypothesis generation and testing by functionalizing clinical observations using these models. This all together allows us to interrogate from different perspectives such as clinical, basic, and translational.

I will give you an illustration of how we can apply this resource focusing on the fibroblast growth factor or the FGF axis, in particular FGFR1, that our group implicated in prostate cancer progression to bone metastasis, actually using one of these models and further validating in the different PDXs.

Furthermore, our group, by clinical studies, determined that targeting FGF receptors has clinical activity in a subset of men with castration-resistant prostate cancer bone metastases. Therefore, highlighting the need to identify patients that will benefit from targeted therapy of this pathway. However, pathway activation in particular in prostate cancer still remains poorly defined. Therefore, we applied our scientific background in the lab with our experimental data, the PDX platform that we have deeply characterized, and human datasets to better understand FGFR1 regulation.

So this shows particularly how we did that. We used PC3 cells that overexpress FGFR1, which we identified as increasing bone metastasis in our experimental models, and found by RNA sequencing and principal component analysis that those cells expressing FGFR1, shown in red and yellow, separate very well from the controls. So we used the genes differentially expressed in this dimension and went to our PDX cohort where we have a subset of models that express FGFR1 as expected and many that are low expressors or negative for FGFR1.

From the top thousand genes that were differentially expressed in our cell lines, after several rounds of iteration, we identified three genes that were highly expressed when FGFR1 expression was high. And with principal component analysis using those three genes, we were able to separate very well the populations that had high expression of FGFR1 or low. We then went into human datasets and further confirmed this association by finding positive correlation between the expression of these three genes and FGFR1 in the TCGA-PRAD dataset as well as the Stand Up to Cancer PCF dataset.

Furthermore, because our interest is in bone metastasis, we found that FGFR1, as well as the three genes implicated and associated in this work, were highly expressed in bone metastasis compared to other metastatic sites. Therefore, we have a candidate signature of pathway activation comprised of NRP2, LRP4, and TGFBI that we now have the right models in the PDX collection to further study and understand the clinical relevance of these genes in progression and in targeted therapy.

With these examples, I hope I have shown you how we can do an informed selection of clinical annotated models in the PDX platform suitable for experimentation. To further facilitate this, the data has been deposited into cBioPortal, so users can start by interrogating the platform and selecting the models that are right for their projects.

I present to you the work that we recently published, and in the picture, you can see once again some of the main contributors to this work. In particular, this publication and this whole effort is dedicated to the memory of Dr. Navone, who founded this and was fundamental for all of this to happen. She recently passed away, and we keep working to honor her legacy and all her work. Thank you.

Andrea Miyahira: Thank you so much Dr. Labanca and congratulations to you and your team for establishing such a tremendous resource. So what is the racial distribution of your models?

Estefania Labanca: That's a great question and something that in the past year has raised a lot of interest and we have a lot of collaborations and have devoted a lot of efforts in the past year to establishing models from different races. So in the cohort and represented in the 44 models, we have a racial distribution that reflects what is seen in the clinic in our institution with about 80% being white, 13% black, and 5% Hispanic. In particular, African Americans, we have a lot of collaborators reach out for studying the black race. So we have five models, all of which are included in the characterization through sequencing. So five models derive from black. Furthermore, when Dr. Navone started this work, she established two cell lines, MDAPCA 2A and 2B, and those two are also derived from African-American.

Andrea Miyahira: Okay, thank you. Your team has established significantly more PDX models than cell lines. What do you think the barrier is to creating cell lines and for people who are looking to establish patient-derived tumor models, what is your biggest piece of advice?

Estefania Labanca: So another great question. So as I mentioned, when Dr. Navone started this, she devoted a lot of efforts to establishing cell lines actually derived from patient samples. However, this was really challenging as you can see, and it resulted in these very good cell lines that we can use nowadays. However, yeah, I think hypothesize that the challenges are first of all, as we know, prostate cancer is a very indolent disease, typically slow growing. So that could be one of the challenges to establish these cell lines as with the PDX and that's part of the advice it takes time. And most likely, and given the evidence that we have with the PDXs, the other thing is definitely there's some stromal component and in particular a contribution of the microenvironment that is provided when we grow the tumors in the mice as PDXs.

For the advice, I would say patience is crucial. So again, as we know, they might take a very long time, so we cannot expect results that fast and there is a big percentage of failure. So that needs to be accounted for. I guess I would say also a very good interaction with the clinic to make sure that the samples arrive as fresh as possible. There is a time window that we need to account for, so we need to be close to the clinic. And I didn't mention this, but obviously there is a big interaction of a lot of players that needs to take place for this to happen. And then the quality of the sample, ideally that is tied to the time also, so I didn't present it, but we also devoted efforts to establishing organoids derived from these PDXs. And what we learned for instance, is that if the tissue is necrotic and that happens a lot or there is more prevalence in neuroendocrine samples, it makes it more challenging to establish the organoids. So therefore good quality and timing.

Andrea Miyahira: Okay, thank you. Are interested researchers able to gain access to your models and how would they go about doing this?

Estefania Labanca: Absolutely, that is the whole point of this resource. I place it in the presentation, we have an email that you can reach out to or you can reach out to me, Peter Shepherd, our program manager at . So I show that we have cBioPortal for accessing the sequencing data and if researchers are interested in the models itself, they can be shared to our material transfer agreement. So yes, they're more than welcome to reach out and we can discuss and happy to collaborate.

Andrea Miyahira: Thank you. And what are your plans for next steps with these models?

Estefania Labanca: Next steps, a lot. But some of them we want to continue characterizing these models. One of the efforts includes epigenomic. We are currently analyzing some of this data in this same for PDXs so we can link it to expression. We also want to make a proteomic-metabolomic characterization that would add further to their value. And as part of our plans, as I mentioned, we have deep interest in bone metastases in particular. So we are planning to further understand the impact of the microenvironment in the growth of these PDXs and whether that has an impact and also treatment. So at least knowing what happens under their standard of care treatment with the growth of these PDXs and how they respond to that, those are only some of the plans in the near future.

Andrea Miyahira: Okay, well I look forward to seeing your next studies. Thank you again for sharing this with us and congratulations again on producing this incredible resource.

Estefania Labanca: Thank you. Thank you so much. I appreciate the invitation and I look forward to discussing very soon.