Read the Full Video Transcript

Andrea Miyahira: Hi, everyone. I'm Andrea Miyahira here at the Prostate Cancer Foundation. Today, I'm with Dr. Brad Gallent of Cedars-Sinai Medical Center, who will share his paper, "The Homeodomain Regulates Stable DNA-binding of Prostate Cancer Target ONECUT2" published recently in Nature Communications. Thanks so much for joining us.

Brad Gallent: A pleasure to be here. I appreciate the opportunity to share highlights from our work published last year in Nature Communications. Title of the work was "Homeodomain Regulates Stable DNA-binding of Prostate Cancer Target ONECUT2." The whole gist of this paper is looking at the structural biology of ONECUT2, which is the transcription factor and how a transcription factor binds to DNA, trying to gain some insights into how to target ONECUT2 for prostate cancer treatment. A lot of people on this paper did a lot of work, including Avradip Chatterjee, Madhu Katiki, who did a lot of structural work with me, as well as some of the biophysics.

So a little bit of background about ONECUT2. ONECUT2 is a highly expressed transcription factor in advanced prostate cancer. Some of this work was published by us several years ago. On the figure on the left, you see ONECUT2 gene activity as disease prostate cancer advances.

So in low Gleason score, low histological grade, we see less ONECUT2. But in metastatic disease, we see a lot of ONECUT2. ONECUT2 has been also co-discovered by other researchers, where they see a similar trend. Neuroendocrine, very advanced prostate cancer has very high expression of ONECUT2 versus CRPC. Prostate cancer has high amounts of ONECUT2 expression compared to benign prostate.

And then highlighting another paper that we published last year about how ONECUT2 activates lineage plasticity in prostate cancer. How we see in patients who take ADT or androgen receptor inhibitors, we see a large increase in ONECUT2 expression. And then looking at rapid autopsy data out of University of Washington, we see a lot of ONECUT2 expression in these patients. About 90% of biopsies from these patients showed ONECUT2 expression.

Essentially, a synopsis of that paper, as we see, essentially, early stage AR-dependent prostate cancer and then has some increase in ONECUT2 activity. And then with treatment, that ONECUT2 and expression increases, leading to lineage plasticity in this prostate cancer. So that's essentially the starting point for some of the structural work in wanting to target ONECUT2. And there's a little bit in that paper as well, some biology of the ONECUT2 inhibitors that we've been working on.

And so these ONECUT2 inhibitors, we've published previously, they decrease growth of CRPC in mouse models as a monotherapy. So these tumors, whether subcutaneous or metastatic models, we see a large decrease in these tumor growths when we inhibit ONECUT2. And the initial inhibitor was called CSRM-617. Which leads us to the starting point for this paper. We really didn't know much about ONECUT2 going in. We knew it was a transcription factor. Based on its sequence, its amino acid sequence, we could predict what it would look like. But we didn't have an actual structure of it.

So in this paper, we published the first crystal structure of ONECUT2. And what we see is, essentially, these two subdomains of the DNA-binding domain of ONECUT2, a CUT subdomain and a HOX subdomain. And each of these essentially bind on opposite sides of the DNA to make a very high affinity binding.

And then from this crystal structure, we can also see exact amino acids that interact with DNA, giving us an indication of the specificity of DNA-binding seen on that ladder on the right-hand side. Those are all the amino acids and the exact parts of DNA that they're interacting with, the phosphate backbone, the nucleotides themselves. And then you'll see some letters and some mutations that we do to really understand these DNA-binding amino acids, which I'll get to in a second.

And this work was just corroborated by mass spectrometry. So doing deuterium exchange mass spec, we can see the blue parts are the parts of the protein that don't have a lot of hydrogen exchange. Meaning there's not a lot of exposed solvent area in that area of the protein. So we see, as you'd expect, areas that are buried within DNA-binding portions don't have a lot of deuterium exchange, not a lot of solvent exposure, essentially corroborating what we're seeing in the crystal structure.

And we took that a step further to look at how these two, the CUT and the HOX domains, how their DNA-binding interplay between each other. So we see, using isothermal titration calorimetry, it's a biophysical technique to learn thermodynamics of interactions. We see a very high binding affinity for ONECUT2 for DNA. It's about 7 nanomolar, pretty typical for a lot of transcription factors

But if we piece out just the CUT subdomain, we see a much higher-- we see lower affinity binding to DNA. The CUT domain binds at about 2,000 nanomolar. High being lower affinity. The HOX domain alone actually doesn't bind DNA at all. But when you actually combine these two polypeptides not physically linked, we see a lot of the binding returns. We see a KD of about 1,000 nanomolar.

For a KD, we see a lot of the same interactions that delta H tells us the number of interactions are happening. So we see, essentially, the same interactions with DNA are happening. But the overall affinity is not the same with the two polypeptides compared to them together as a single peptide.

This essentially tells us the connection between the two, that loop, is very important. It keeps it, essentially, as we see, like a clamp. So one side is clamped onto DNA and keeps the other side clamped. So neither can really move, and gives us that really high affinity DNA-binding that we see for the whole protein.

So then we did a lot of mutations on those DNA-interacting amino acids. So SQ is one of them, essentially mutating two residues to see how that affects DNA-binding, as well as an RR mutation to arginines and an N mutation, asparagine. And the green being the CUT domain, where the purple on that ladder being the HOX domain.

So the important part of this is, if we essentially mutate amino acids in either portion of the protein, we lose DNA binding. So ONECUT2, as I mentioned, using ITC and similar with BLI, bilayer interferometry, which is like a kinetic analysis versus a thermodynamic analysis, we see a loss of binding affinity with both the SQ part of the CUT domain, or mutations in the HOX domain with that arginine mutation, or the RR mutation to arginines. And this, we wanted to see, does this actually translate into cell lines?

So if we mutate these amino acids, does ONECUT2 still have its normal function in cells. And in this case, we used LNCaP and we knocked in these mutant or wild type proteins. So you can see on the top part, on the right-hand panel, we look at cell growth. Wild-type ONECUT2 when knocked into these LNCaP cell lines, a castration sensitive cell line, we see large amount of growth in that cell line. However, if we knock in a mutant form of the protein, we don't actually see any increased growth. Telling us that the ability of ONECUT2 to actually bind to DNA with high affinity is very important for its function in the cell, as well as what we see in the test tube, looking at, essentially, thermodynamics or kinetic analyzes.

And then taking this to the next level, we want to look at neuroendocrine markers. ONECUT2 is very well known to increase neuroendocrine and lineage plasticity in prostate cancer cell lines, as well as clinically. So when we look at ONECUT2 wild-type overexpression, again, in LNCaP cells, we see the large expression of NSE, PEG10, and SYP, which are all neuroendocrine markers. However, if we knock in a mutant form of these ONECUT2 proteins, we actually don't see that. Telling us neuroendocrine growth, everything like that is very important for ONECUT2 to have its full functionality, bind DNA with high affinity to carry out its normal function.

So this brings us, essentially, to the next level. This is, essentially, a platform for us to do more drug development. So next level is for us to take the ONECUT2 inhibitors and learn where on ONECUT2 do they bind with atomic resolution, and what parts of the proteins are exactly interacting with, and how can we tailor the ONECUT2 inhibitors to bind to ONECUT2 better?

And another part of this, we want to learn the mechanism of these ONECUT2 inhibitors. We have that CSRM-617, which has shown in vivo efficacy as monotherapy. But we want to learn how it's working. How is it actually blocking ONECUT2 from binding to DNA? ONECUT2 is a very high affinity 7 nanomolar KD for binding DNA. How are we using a small molecule to actually do that? That's a big question we have that we're working on answering right now.

A lot of people to thank. Authors on this paper, Avradip, Madhu, PIs that I work with, Ram Murali, the structural biology PI, Michael Freeman, Chen, who worked on the other paper published in Nucleic Acids Research last year, and then a lot of collaborators, NIH for funding T32 funding for me. The DOD has funded a career development work for me, an ID award to help us work on this. Prostate Cancer Foundation has been immensely helpful with a Young Investigator Award for me and a new challenge award to help work on this.

Andrea Miyahira: So thank you so much, Dr. Gallent, for sharing this with us. So what insights critical to drug discovery did you specifically gain from the atomic structure of ONECUT2 that would not have been found by other strategies?

Brad Gallent: I think a couple of things. From this paper, I think the ability to understand amino acids interacting directly with DNA, and the fact that just mutating one or two of those amino acids and having a small effect on ONECUT2 function was really important for us to understand this is accomplishable. There's something-- we don't have to knock out the whole protein. Previously, we used shRNA, knocking out the whole protein, or large portions of the protein. That's not necessary. Some kind of small inhibition would be sufficient to actually stop ONECUT2 from its normal function.

Andrea Miyahira: Thank you. And based on this, what regions of ONECUT2 do you think should be targeted for an optimal on-target, low-toxicity drug development approach? And how does this compare with the already developed ONECUT2 inhibitors?

Brad Gallent: That's a great question. So actually, we were thinking-- part of this paper was to look at, was the CUT or the HOX remain the better part of the protein to go after. And our answer was both are fine. Essentially, knocking out amino acids on either [inaudible] function. But then taking that next further, part of your question was low toxicity. It's actually very important-- coming from a clinician standpoint, very important for new drug development.

So part of this paper, we don't go into it entirely, but looking at the homology of the CUT and the HOX domain compared to other proteins in the human genome. So actually, there's not a lot of homology similarity in either the CUT or the HOX. HOX is very prevalent in our genome. There's about 300 proteins that have a HOX domain within them.

But sequence similarity is very low. So 30% to 40% outside of the ONECUT2 family. So meaning that although the overall scaffold of the protein is very similar, the actual three-dimensional structure is not. Meaning that we have a lot of leeway to make a ONECUT2 inhibitor, as long as it's very specific for the HOX, same with the CUT, we should be fine with producing an inhibitor with low off-target toxicity.

Andrea Miyahira: OK, thank you. And based on this, how do you think these findings impact what's known about the role of ONECUT2 in prostate cancer, CRPC, and neuroendocrine prostate cancer?

Brad Gallent: Yeah, a really good question as well. I think what we learned is that any kind of small effect on ONECUT2 function blocks its ability to carry out its cancer progression activity. We block it and we just have a loss of 7 from 7 nanomolar KD to 40 nanomolar KD binding affinity for DNA, all of a sudden, it's not able to induce neuroendocrine features. It's not able to produce cancer cell growth like it normally does.

Andrea Miyahira: All right, thanks. And what are your next steps? Is your team utilizing this information to develop novel ONECUT2 targeting agents?

Brad Gallent: Yes, that's the exciting part. So we have our initial inhibitor CSRM-617. But we've made a host of analogs of that, about 80. Michael Jung, well known for inventing enzalutamide and apalutamide, our close collaborator, so he's made about 80 of these. And we've learned, one, what features of the inhibitor are very important for the inhibitor to bind to the protein?

But also we have solved the crystal structure of a bound. Now, we're learning a lot more about how some of these improved analogs actually interact with ONECUT2, taking it to the next level. And then now, we can rationally design these inhibitors to tailor interactions with the protein and also mess with pharmacodynamics and other things that we want to play with as well now that we have a visual of the inhibitor interacting with the ONECUT2 protein.

Andrea Miyahira: OK, well, I wish your team luck. And thanks for sharing this with us.

Brad Gallent: Absolutely, I appreciate the opportunity.