International Medical Conference Endometriosis 2025:
Endometriosis 2025: Your Mother Should Know, Your Doctor Should Know Better!
Decoding Menstrual Transcription: Insights into Endometriosis - Ridhi Tariyal, MBA, BS
Our next speaker is R Taral. She has an MS and a BS and is an innovator, a woman, and has the ability to advance women's health through technology and biotechnology. She is the co-founder and CEO of NextGen Jane. She is pioneering non-invasive diagnostics for reproductive health conditions, especially endometriosis.
Hi there. Good afternoon. Thank you, Dr. Seskin and Dr. Martin for the invitation. Okay, great. As mentioned, I run a small biotech out of Oakland, California where we are developing a menstrual database and molecular profiling of uterine shed uterine lining. I'm going to be talking about transcriptomics specifically of menstrual effluence in the context of endometriosis. The last decade has seen many useful molecular approaches taken to endometriosis. We have as a community done, RNA sequencing small RNA sequencing methylation arrays. Looked at microbiome mass spec, and we've explored so many different sample types. We have looked at serum, of course, endometrial biopsies, saliva, and there's this burgeoning field of people looking at menstrual effluence, and that's all very exciting. It has advanced our understanding of endometriosis, and yet we don't have a robust reproducible, non-invasive diagnostic to shorten that diagnostic odyssey that everyone has talked about today and is deeply familiar with. And so why do we have such promising advances that have failed to produce a set of biomarkers that have recapitulate in clinical validation or in any sort of broader dataset? There's a few potential culprits that have been brought up. They include small end studies. They include poorly defined control groups, hormone cycle variability, and a case population that has high comorbidities and more.
I put these two papers up here because I want to double click on one of those factors that are potential confounders and that is hormonal milieu and cycle phase. These two studies bracket the decade and have explored this factor deeply. TEI ET all in 2014 looked at hormone milieu and took it into consideration when evaluating and establishing cycle dependent markers for endometriosis. And in 2023, TE et al provided a more granular and molecular way of staging endometrial biopsies. They looked at over 200 samples of endometrial biopsies from women who were coming in with pelvic pain being evaluated by surgery, the majority of whom had endometriosis. And what the author suggested is that there is a remarkably coordinated and significant gene expression profile changes throughout menstrual progression. So potentially if you don't account for this variable, it's going to be a significant confounder in your data.
I am going to show you a series of slides how when we took that framing into account in our own data, we did find it to be a confounder. Over the last eight years, my team has explored many different genomic markers and menstrual blood for insight into endometriosis, and we have found that the hormonal milieu is relevant. We've looked at small RNA, we've looked at RNA, we've done methylation arrays. We've done 16 as sequencing. This is unpublished data from our first study where we did the classic pre-surgery teca menstrual sample from patients that went on to have surgical confirmation of endometriosis and compared it to tampons from individuals who were healthy controls. And we found 49 dysregulated, significantly dysregulated small RNA we're really excited because most of them overlapped with small RNA that were already known to be impacted in endometriosis patients. So great. We were on the right path.
In follow-up studies, we examined both RNA and small NA signatures to provide further biological support for a forming hypothesis. So to do this, we looked at target genes that were anti-correlated to their respective small RNA regulators. While this is not conclusive, this method narrows down the possible biological targets that these small RNAs are affecting. So the graph on the left shows you a single illustrative example of the negative correlation of target genes that we're interested in. And the similarity matrix on the right shows all 45 small RNAs that were dysregulated in our endometriosis patients. All the genes that show this negative regulatory correlation are genes involved in erythropoietic processes suggesting a tightly regulated biological process. So again, so far, this is looking all good to us. Now, we decided to add the framework of timing and disease to this analysis, and I should put timing in quotes because this is a self-reported subjective assessment that the patient herself provides as to when she collected the tampon.
Was it day one? Was it day two, day three? This will become more relevant in a second, but just know that this is a subjective assessment at first glance, which you'll see is that the small RNA expression levels by disease by day three, day one samples, they seem sort of equivalent, but the smaller expression on day three seems to diverge and that's interesting and potentially clinically useful. You might've already guessed where I'm going with this. When I put these two charts together and I graph the small RNA levels for both cases and controls on the same chart, the opportunity for confounding becomes more clear. Remember, this is a self-reported characterization of day of collection. So if it's incorrect or off by 12 hours, you could see where there would be a chance to mis call the sample. So the daily genomic changes that are inherent to menstrual dynamics, even within the three day window of menstruation, it's enough to erode the clinical utility of any perceived molecular difference between cases and controls if it's not accounted for.
This observation is not limited to reproductive health. It's not limited to endometriosis, and it's not limited to endometrial biopsies or menstruation. The effect of hormone milieu or where a person is in their cycle has been known to affect biomarker readouts as broad as cholesterol readings, depression, small RNA and serum saliva biopsies. It affects mental health assessment tools. When you send a woman in, at what point in her monthly cycle, you evaluate her for any biomarker, it's relevant and we know it's relevant because these fluctuations in hormone cycles and the subsequent noise they add to experiments, we know it so well that it's part of the reason that motivates most animal models to be male models, right? You don't want to include hormone variability in your animal models in the beginning, and so you reduce that by looking exclusively at male models. So it's only natural to assume that this factor of noise in clinical work is also going to be relevant.
Okay, so then we wanted to address this challenge. How are we going to address this challenge? We continue to collect thousands of tampons from hundreds of participants and performed unbiased transcriptomics to establish a large RNA-Seq dataset. These samples included 50 patients that provided us serial samples on day one, day two, and day three. They included patients that provided us pre-surgery and post-surgery tampons, so pre-surgery, and then immediate post six months post, 12 months post. We now have people who were providing us samples 18 and 24 months past surgery, multiple cycles from individuals. So we could compare a cycle in January to a cycle in June as well as non menstruating cycles. So for a subset of patients, we'll send them a tampon, ask them to wear it for 15 minutes, 30 minutes, mail it back in, so we have that data point as well. And all of this so that we could try to define the key variables that add noise to any attempt to find reliable differences between endometriosis cases and controls.
The two big ones I want to walk you through are bleeding phenotype and tissue as well as timing, and when in the cycle you do your sample collection. Okay, so you're looking at this heat map. I'm showing you select genes that we've identified that help define uterine tissue from vaginal background. And so as part of our tampon collection, we're getting not just menstrual effluence, but we're also getting cervical vaginal background. So we need clear ways to identify when we have sufficient uterine tissue to look at the sample. As you can imagine, if someone is giving you a sample on day one, let's just say they have a lot of vaginal vaginal background because they're not providing much blood versus someone who's giving you a tampon on heavy flow day, there's a lot of uterine tissue there. If you compare those two, one of the biggest differences between those tampons will just be tissue, necessarily any sort of signal related to disease.
And so you have to account for that. So this was where we looked at a set of tampons that we knew were blood, a set of tampons that we knew had no blood. So just vaginal background. We established these gene modules. And you can see on the left hand side, the vaginal gene modules are highly upregulated when you're looking at vaginal background. And at the bottom is the uterine module, which is highly upregulated when you're looking at menstrual samples. So great. Then we tested it out. We looked at over 1800 samples and we said, how well does this work? So I just want to 0.1 thing out is, so you'll see that this again is the vaginal gene module here. This is the menstrual slash uterine gene module here and here, this bar on the top is the patient self-assessment of whether or not she thinks it's a light tampon, a medium tampon, or a heavy tampon.
And so there's sort of two observations that should jump out. One is that in general, there is a trend, right? People are directionally correct. You can see that these are all tampons in the beginning that are highly expressed in this vaginal module where the individuals themselves have said, this is a light tampon and the maroon are heavy tampons. And again, directionally, they are highly expressed towards this area. So people generally know how heavy the tampon is that they have provided. The second observation is, and yet there's room for improvement. There are oftentimes where people are imprecise. And so having a way to molecularly identify that is useful.
So we did the same thing for timing. Now, this was a little bit trickier. We used the dataset, and again, we have people self-report to say, this is my day one, this is my day two, this is my day three. And so we broadly use that as a way to identify cycle start gene modules. So these are going to be the genes that are upregulated when you're closer to the beginning of menstrual initiation and then cycle end gene modules to do the same exact thing to say, is there a way that we could actually using molecular signals, identify when you're closer to the beginning of a cycle and then we validated it or looked at it in 1800 tampons. I don't know if it would be surprising to you, but it's a little bit more messy. It doesn't have quite the clean architecture that the tissue did.
There's still, so again, right here at the bars at the top are these self-described assessments. The lighter samples are day one, the darker samples are day three. There still is a trend. People who say that this is day one, it is closer to the start, and this is darker and it is closer to the end. And there's two observations to make there. One is that people are generally more reliable narrators as to how heavy the tampon was when they took it out. As you can imagine, they look at it, they say, oh, this has a lot of blood. It's saturated. And so being able to identify subjectively the tissue in the tampon is just much easier. But for day, someone could think when they start spotting, that's day one, or is it when you actually start bleeding? And so that's one thing to note that you're just going to see less tightness in this particular graph.
But the second thing is that you're more likely to get consensus markers at the beginning of the cycle. Most people have a cycle start, whether you're off by T 12 hours in either direction. However, day three is not necessarily the end of everyone's cycle. In fact, probably for anyone who's a heavy bleeder, that's midway in their cycle and they're going to go on for day seven, day 10. And so you would see less consensus markers. So this is also something that we would expect. Okay, so we, as I've mentioned, do tampons only, and there's a lot of benefit to that as I've, hopefully you've gleaned the fact that you can collect pre and postsurgery tampons. The fact that you can do serial collections every four hours, every six hours across multiple cycles, that's all great. But the shortcoming is that you don't get to look at secretory or proliferative samples because we don't do endometrial biopsies.
And we want to because we want to see how relevant the markers that we've developed based on menstrual fluid is actually abstractable. And then can you actually figure out menstrual dynamics beyond menstruation using these markers? So this is actually the TAY dataset. They obviously do endometrial biopsies all throughout the entire monthly cycle. And we decided to use our biomarkers on their dataset. And so this is, again, I think over 200 endometrial biopsies that they did. We just picked a single marker that we were interested in and to see first to overlay it onto the TAY dataset. And you can see that there is a nice pattern and flow across the seven stages that TE uses. And then we added our menstrual data here on the left, and I think one of the points that we want to make is that you can actually get more resolution if you add more granularity.
You can see there's even further pattern as you look at day one, day two, day three, and how far does that go? Would you get further granularity if you looked at early in the morning versus evening? Would you get further granularity if you looked at every four hours? We're not sure, but it's something that we're exploring internally. Now, continuing to look at this dataset. As we mentioned, we have uterine and vaginal gene modules, and so we applied those modules to the TE dataset. And so in this particular Disney, oops, I was doing so well with not moving forward. This is TE data, and it is colored by the stage that the authors have identified. So whether it's proliferative, secretory, or menstrual, and it is our gene modules that have separated it out. So what that's showing you is that these menstrual markers that were developed at a completely different dataset actually do a really great job of getting the same disaggregation on these endometrial biopsies.
We keep applying our framework of identifying timing and tissue to the endometrial biopsy dataset because we're really interested in this opportunity of data agreement across sample types. So the differences that are associated with timing and tissue are so important to us that we want to validate on other people's datasets. On the left, this is just to figure directly from the tape paper. You can see that using their markers, they're able to actually get really nice disaggregation on this PCA plot by menstrual, proliferative and secretory phases. And then we took their dataset, we applied our modules, and we're not only able to get the same disaggregation. We actually would offer up that we get even more clear disaggregation in menstrual samples. And maybe that's obvious because our dataset was built on menstrual samples, and so it maybe would be more refined and offer more resolution. There.
We have more work to do to determine other variables that can shift genomic signals and interfere with our ability to really reliably phase a sample. What you're looking at here are 18 individuals who gave serial samples on day one, two, and three. They have been organized so that the plots on the left are the stable profiles. So these individuals, what you're looking at is relative abundance of lactobacillus good bacteria. And so this is a healthy sign. You're seeing that the good lactobacillus is stable across day one, day two, and day three. So all of these nine individuals, there's nothing really remarkable about what's happening in their microbiome profile. On the right hand side, you're seeing, well, these people have some sort of unstable microbiome profile. In fact, their lactobacillus is either decreasing or moving in wonky directions. And so we know that this might actually impact gene expression profiles that we've selected as indicative of timing.
Like if there are inflammatory markers in our modules, then they would respond not only to what's happening from a menstrual dynamic perspective, but potentially they would respond to what's happening in your microbiome. And so these are questions that we're currently investigating, and you might ask why this environment see seems noisy, right? How do you account for all of these variables, timing, tissue, bacterial content? The reason is, I think probably came up in discussion for many of the PIs. There's no non-invasive diagnostic. If you think about this, disease impacts children, that's when the disease process starts. Maybe it should be picked up in pediatrics. This is not an area where you want your child to have surgery to find out that they have a disease. And so there has to be a noninvasive diagnostic that is explored, and we think that menstrual effluence is just a key aperture by which to continue to evaluate what's happening with this patient population.
So this last slide I'm showing you is really exciting. It's part of why we want to continue to build the foundation models to understand menstrual effluence. This is about 1400 samples from our dataset, a mix of cases and controls. We are agnostic to any of the comorbidities that our patient population has. I mentioned that because we know that these patients have autoimmune and fibroids and adeno MiSiS, so we don't exclude them from the data at all. each.is a sample, and they have been plotted on an X axis that represents our uterine tissue. So the further you move right on the axis, the more enriched the sample is for uterine signatures. The dots are colored by whether they're on birth control of any type. So you can see most of 'em are on no birth control. They're yellow. If they're a red dot, they're on a combo, estrogen, progestin pill.
And if they're blue, they're just on a progestin only pill. And as you might expect, many patients that are in some type of oral contraceptive respond to the drug and have lighter, more manageable periods. So there's this concentration of red and blue dots here. You can see where they have lighter periods. But what's happening here, I mean, these are individuals who are on some type of progestin and they have extremely high uterine signatures. Could this possibly be an opportunity to evaluate potential non-responders to progestin to responders? And it's these type of questions that we think really should be explored and we're excited to explore and why we keep going, trying to understand and unpack the noise that's inherent in this fascinating environment. Thank you.
