Novel Computational Approach Identifies NME2-MYC Axis as Biomarker and Therapeutic Target in Enzalutamide-Resistant CRPC - Antonina Mitrofanova

March 11, 2024

Antonina Mitrofanova delves into her team's research published in Nature Communications. The study uncovers the NME2 and MYC programs as pivotal markers of enzalutamide resistance in castration-resistant prostate cancer (CRPC). Dr. Mitrofanova's team embarked on this journey following observations that patients with elevated MYC levels showed increased resistance to androgen receptor signal inhibitors (ARSIs). By employing a novel computational algorithm, they explored upstream regulators of MYC in enzalutamide resistance, unveiling potential therapeutic targets. Through rigorous analysis and experimental validation, the study indicates that CRPC patients with heightened NME2 and MYC activity could see benefits from combined therapeutic targeting, offering new avenues for managing enzalutamide resistance and proposing these markers for future clinical trials.


Antonina Mitrofanova, PhD, Department of Health Informatics, Rutgers Health, Newark, NJ

Andrea K. Miyahira, PhD, Director of Global Research & Scientific Communications, The Prostate Cancer Foundation

Read the Full Video Transcript

Andrea Miyahira: Hi, everyone. I'm Andrea Miyahira at the Prostate Cancer Foundation. Today I'm joined by Dr. Antonina Mitrofanova, an associate professor and Associate Dean for Research at Rutgers University. We'll be discussing Dr. Mitrofanova's recent paper "Mechanism-centric regulatory network identifies NME2 and MYC programs as markers of Enzalutamide resistance in CRPC." This was published in Nature Communications. Dr. Mitrofanova, thank you for joining me today to discuss this study.

Antonina Mitrofanova: Thank you so much for your kind invitation, Andrea. Thank you very much, Andrea, for your kind introduction and for the opportunity to present details of this work. This work was done in collaboration with Northwestern University Vishal Kothari, Ted Schaeffer, and Sarki Abdulkadir. My postdoc Sukanya Panja and Sarki's postdoc Mihai Truica were co-first authors.

The story motivations started with our publication in Nature Cancer. This was a collaboration between my lab and the lab of Cory Abate-Shen. One of the stories that we investigated in that manuscript was that patients with increased levels of MYC showed increased resistance to androgen receptor signal inhibitors, ARSIs. So my lab dug deeper into this story and took Abida et al.'s cohort and divided it into patients that received enzalutamide and those that received abiraterone. What we observed was that patients who received enzalutamide were at high risk of developing resistance to this treatment if they had high MYC activity. We did not observe the same association in abiraterone-treated patients.

Inspired by this observation, we asked ourselves a couple of questions. Could we, in an unbiased way, elucidate upstream regulators of MYC in enzalutamide resistance and could we use this information to define potential additional axes for salvage therapeutics for patients that develop resistance to enzalutamide? To address these questions, my lab developed a novel computational algorithm. This algorithm combined two ideas that we have been working on for a long time: transcription regulatory modeling and pathway-based modeling. We combined these two modeling ideas into one to reconstruct the mechanism-centric regulatory network.

In this network, each node is not a gene or mutation or specific entity. It's the whole mechanism. So, let's say here the green node is the whole pathway mechanism, and orange nodes that surround green nodes are upstream transcription regulatory programs. We reconstructed this network in a way so that upstream transcription regulatory programs surround green nodes, there are arrows that point to it, so that it addresses our question: Can we identify upstream transcription programs from the MYC pathway?

In order to reconstruct this network, we utilized the carefully curated Stand Up To Cancer East Coast patient cohort. This is a CRPC cohort. For each patient, we estimated activity levels for each pathway and activity levels for each transcription regulator so that we were able to define transcriptional activity vectors and pathway activity vectors. So for each transcriptional regulator, the vector would correspond to activity levels of that particular regulator in each patient. It can be down-regulated, here shown in blue, or upregulated, shown in red. The same applies to pathways.

We then compared all pairs of transcription regulators and pathways for potential association using linear regression analysis. This association can be positive, or it can be negative; we then corrected it for multiple hypothesis testing and also performed bootstrap analysis. Those associations that were selected by these analyses were then considered as edges in our network. Such a network is CRPC-specific, and you could utilize the network to address multiple CRPC-specific problems. The next question was, can we make it enzalutamide specific?

For that, we developed a network mining query step, with the general idea of this step as the following: If we take this network through phenotypes of interest, and it can be two phenotypes, for example, responders, non-responders; it can be three phenotypes like we utilized here, untreated, enzalutamide-sensitive, and enzalutamide-resistant; or it can be an array of phenotypes. It really depends on the phenotypes you're interested in. And so, if we take the whole network through these phenotypes, we can then identify parts of the network or subnetworks with differential behavior, and these subnetworks would then be potential biomarkers and therapeutic targets for the conditions of interest.

We hypothesized that subnetworks that are active in the untreated stage, then are brought down, they are repressed by enzalutamide treatment, so in the enzalutamide-sensitive condition, but then they bounce back and become activated again in the enzalutamide-resistant condition. Those subnetworks would be utilized as our biomarkers and potential therapeutic targets.

We applied this analysis to the MYC pathway and identified an array of transcriptional regulatory programs. As you can see, there is quite some number of them, and the question we asked ourselves was, are they multicollinear and do they work in groups in the way they affect the MYC pathway and can they prioritize those groups for potential therapeutic targeting?

To address these questions, we developed a second network mining step that was inspired by our understanding of Partial Least Squares. This analysis allowed us to identify multicollinear groups of transcriptional regulators and further prioritize them based on their effect on the MYC pathway. This analysis identified seven groups or seven clusters, with cluster number three that just had one transcriptional regulator, NME2, having the biggest effect on the MYC pathway. NME2 is known to bind to the promoter of MYC and is better studied in lung adenocarcinoma.

First step was to confirm the NME2 and MYC association in CRPC conditions. For this, we analyzed three patient cohorts that were CRPC and were able to demonstrate that there was a strong association between MYC pathway activity levels and NME2 transcriptional activity levels. We took this discovery back to the Abida cohort and confirmed the same association in this cohort as well that I'm showing on the left. We have also confirmed that patients with increased levels of MYC and NME2 are at higher risk of developing resistance to enzalutamide, and in fact, they developed resistance to enzalutamide five times faster compared to the rest of the patients. We also validated this discovery in the Stand Up To Cancer West Coast cohort. This is a cohort with mixed treatment, but we know that 67% of this cohort received enzalutamide. We were able to show the NME2-MYC association there as well, and also demonstrated that patients with high MYC and NME2 levels had a high risk of resistance to enzalutamide.

For experimental validation, we collaborated with Sarki Abdulkadir's lab and the experimental data I'm showing in this slide was done in his lab by his postdoc Mihai Truica. They have shown that knocking down NME2 significantly reduces MYC levels, confirming that NME2 is an upstream regulator. And then they performed in vivo investigations of therapeutic targeting along the MYC axis. They were able to show that tumors that were resistant to enzalutamide, after the administration of MYC and NME2 inhibitors, were able to get resensitized to enzalutamide, basically prolonging enzalutamide action and prolonging survival.

Our results suggest that CRPC patients that fail enzalutamide and have increased levels of NME2 and MYC activity could benefit from combined therapeutic targeting along the NME2-MYC and AR axis. These markers are valuable candidates for future clinical trials. Even though developed to investigate MYC in CRPC, our network and our method could be broadly used to investigate many CRPC-related questions and beyond.

Andrea Miyahira: Thank you for sharing this really informative study. Might the NME2-MYC axis be a biomarker for patients who should receive abiraterone instead of enzalutamide or who may benefit from abiraterone after enzalutamide?

Antonina Mitrofanova: Andrea, this is a really great question. I would hypothesize that, yes, patients who cannot receive enzalutamide because they are at high risk should receive abiraterone, or those that failed enzalutamide should be subjected to abiraterone. We didn't investigate this specific question in the study, but given the data that we have, I would hypothesize that. What we were able to show, though, is that patients that are at risk of resistance to enzalutamide would benefit from inhibition along the MYC axis and patients that also failed enzalutamide treatment would benefit from MYC inhibition as well, coupled with enzalutamide.

Andrea Miyahira: Thank you. Did you investigate whether this resistance mechanism overlaps with other CRPC resistance mechanisms such as AR amplification, for instance?

Antonina Mitrofanova: It's a wonderful question. When we compared patients with high MYC and NME2 levels to the rest of the patients, we saw enrichment in AR expression and AR activity in patients with high MYC and NME2. We also saw that about 50% of patients who have high MYC had AR amplification. So there is definitely a signal.

Andrea Miyahira: Okay. What frequency of ARSI-naive patients have this pathway activated?

Antonina Mitrofanova: From the cohorts we investigated in this study, we observed that at least 55-60% of CRPC ARSI-naive patients have MYC and NME2 active.

Andrea Miyahira: When you observe changes in the MYC-NME2 axis during progression to enzalutamide resistance, are you able to differentiate between intrinsically changing cellular programs versus changing frequencies of tumor subclones?

Antonina Mitrofanova: This is a wonderful question. What we've seen so far is that we see the intrinsic signal, so the signal where MYC and NME2 are high and active and can predict patients ahead of the therapy before they're treated, we see that signal in CRPC ARSI-naive condition. So it is intrinsic; we see it at the very beginning. Then, when there is sensitivity, it kind of goes down and then it bounces back when the resistance develops.

When we look at the single-cell level, we saw that the majority of lung carcinoma cells showed increased activity of MYC and NME2 before treatment. And then after patients develop resistance, also the majority of adenocarcinoma cells. So it looks like it's a major clone that is present before the treatment is given, then it subsides, but then that major clone recovers back after the resistance develops.

Andrea Miyahira: Okay, thanks. What are your next steps for the studies?

Antonina Mitrofanova: We're working as a team now, and we are really hoping to take it to the clinical trial so that we can help this 60% of CRPC patients define the best treatment for them and improve their quality of life and prolong their survival.

Andrea Miyahira: Well, thank you so much for sharing this study with us.

Antonina Mitrofanova: Thank you so much for your kind invitation, Andrea.