Interrogating cancer, one cell at a time: Q&A with Hisham Mohammed

Hisham Mohammed, Ph.D., helped create a way to peer into individual cells and take multiple measures of gene activity and gene regulation at the same time. He’s leading the first research to apply the method to cancer – work that’s earned recognition including the Endocrine Society’s Early Investigators Award. Mohammed is an assistant professor of molecular and medical genetics in the OHSU School of Medicine and scientist in CEDAR, the Knight Cancer Institute’s Cancer Early Detection Advanced Research Center. Cancer Translated talked with him about his single-cell multi-omic research and its prospects for finding better ways to detect and treat cancer.

Cancer Translated: What can we learn about early cancer biology and detection using the single cell analysis techniques you’ve helped develop?

Hisham Mohammed: We’re working on understanding what’s going on in hormone driven cancers such as breast and prostate cancer. The exact same hormones and hormone receptors that regulate normal cell functions have different roles in cancer cells. We think that it could be because of differences in the underlying genetics of each cell, or it could be underlying epigenetics, which is how genes are controlled. Whilst we know that these differences exist, we haven’t been able to explain how or why. Our multi-omic assay allows us to look at RNA, DNA methylation and chromatin accessibility in the same cell at the same time, and we can now resolve questions into how fundamental biological processes go wrong and start regulating and driving cancer. Understanding this is important to detecting early cancer, and treating early cancer.

CT: How is this working in the case of breast cancer?

HM: More than 70% of all breast cancers are estrogen receptor positive. Estrogen signaling is thought to drive cancer in these patients, and you have both very aggressive and nonaggressive forms of these diseases. But we don’t know why tumors in some patients respond to anti-estrogen therapy and some don’t. From a histology perspective, they look almost exactly the same, they all have estrogen receptors, but some cells are really responsive to estrogen or to estrogen blocking, and some aren’t. Now the question is: what makes these nonresponsive cells different? We think, again, that it is probably a combination of the underlying genetics and epigenetics of the cell.

We are doing experiments where we get patient samples and we challenge them with different hormones and then do our multi-omic analysis to understand why some cells respond to anti-hormone therapies and others don’t. And because we have these multi-omic analyses we can understand what’s different about particular cells that have become nonresponsive. As a result, we have found several signaling pathways that we think not only regulate an alternative signaling route [in nonresponsive cells], but also may have a role in controlling tumor heterogeneity [the diversity of cancer cells arising from the same tumor]. This is something we are actively working on right now.

CT: How early in the course of tumor development are these things happening?

HM: Probably very early. We find that there exists some heterogeneity at the single cell level in normal breast tissue, and you have multiple different lineages of cells in normal breast tissue. There is also a lot of shared heterogeneity between normal and cancer cells, suggesting that in cancer you have a loss of lineage identity and increased plasticity perhaps even preceding cancer transformation. Understanding this is important to identifying which cancers progress to become lethal and which don’t.

“we find that not only are these cancers different, the surrounding cells such as the fibroblasts are also very different between a benign tumor and an aggressive cancer”

‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾

CT: Some altered cells and incipient tumors stay benign while others run rampant. Will it ever be possible to draw a line between altered cells that are never going to become dangerous cancers, and altered cells that are going to be a dangerous?

HM: I would say yes. We have some level of information, using traditional approaches such as histology, to identify which cancers are likely to become somewhat aggressive and which ones are not. And now we’re adding much more detailed information on top of that, which we can combine to give us much better predictors. We can use this information to develop very specific testing tools. For example, existing bulk genetic testing strategies, such as the PAM50 [a tumor profiling test that helps show whether some estrogen receptor-positive, HER2-negative breast cancers are likely to metastasize] do work to a certain extent. The granular detail that we will be providing will give much better resolution.

CT: Have you been able to study pre-cancerous patient samples?

HM: The experiments that I mentioned earlier were being performed in all types of early stage cancer including benign tissues. And we can see very clear differences at the single cell level. We find that not only are these cancers different, the surrounding cells such as the fibroblasts are also very different between a benign tumor and an aggressive cancer. We are looking at these fundamental differences in the early stage cancer and we think there are probably very stark differences between them.

CT: That’s the power of single cell analysis. Before this capability, what were people doing?

HM: They were correlating bulk averages, they were averaging out all of the signals.

CT: Are you selecting cells one by one from specific locations?

HM: Right now, we are dissociating tumor biopsies. We have a project where we are sectioning tissue, capturing individual cells based on their spatial location and then performing single cell omics on them, which will allow us to ask questions within the same patient about why some cells expand beyond the ducts and become invasive cancers while other cells remain confined within it. We know that genomic aberrations occur in cancers: DNA mutations, copy number alterations. But if you look at benign breast cancer, or DCIS [ductal carcinoma in situ], there is almost no difference in the genetic profile. It’s almost as messed up as invasive cancer. We hypothesize that it is probably because of changes in the microenvironment, but also because of epigenetic changes or lack of epigenetic changes, which haven’t allowed the cells to become invasive. That could be what makes the difference between DCIS and invasive cancer. We now have the technologies to go beyond bulk analysis and cut out individual DCIS and non-DCIS cells out and analyze those.

CT: You mentioned the role of stromal cells around tumors. How do you identify these non-cancerous cells for single cell analysis?

HM: We use the RNA results from our multi-omic assay to ground a lot of our questions, such as what cell types are we looking at, in addition to cell surface markers. We can identify cells that are slightly different from other cells, and see if that is because they have an underlying copy number aberration, or because they have an epigenetic difference. Without a multi-omic assay, you would not be able to do this because you lack a matched epigenome and transcriptome from the same cell.

CT: This research seems very promising for breast cancer. What about prostate cancer?

HM: We have a very similar project. We know that androgen receptor is very important in normal prostate development. It’s also highly present in benign, early, and aggressive, and also metastatic forms of prostate cancer, suggesting that androgen receptor is a master regulator that assumes different roles in different settings.

We are taking prostate samples of normal, benign and early invasive cancer and challenging them with androgens or antiandrogens to see how the cells respond. Understanding the response to a given stimulant or inhibitor allows us to unmask the true identities, in a sense, at a single cell level. The question we are asking is: can we understand signaling changes in somewhat early cancers that may be more aggressive?

In prostate cancer, you have patients who are on the fence on whether they need to be treated or not. We know that a subset of these patients will have aggressive cancers. And we think we will be able to better detect them using these single cell methods. We think there are a lot of existing drugs which can easily treat these patients. We are hoping to be able to identify which drug matches these early aggressive cancers better so that we can identify and treat them.

CT: Do you see a future in which single cell, multi-omic analysis become a part of the routine workup for cancer patients?

HM: Right now we are answering fundamental research questions and once we validate key findings, we’re confident they can be converted into blood- or biopsy-based assays such as DNA methylation panels, or RNA signature panels or a combination of these. We use the single cell analysis to tease apart in fine resolution what is happening, and then we will have to translate that into a more clinic-friendly approach.

“we’ve uncovered signaling networks that we haven’t seen before in cancer, and many of these networks change in response to hormones”

‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾‾

CT: What avenue of research are you most excited about now?

HM: We’ve developed a new computational approach to look at single cells. Traditional computational approaches cluster cells into different groups. Our approach uses a machine learning algorithm that allows us to better understand the plasticity of cells, especially hormone responsive cells. And we’ve uncovered signaling networks that we haven’t seen before in cancer, and many of these networks change in response to hormones. What’s even more exciting is that we’ve applied these networks in patient single cell data and we can see that these underlying, not very obvious networks are very strong regulators and differentiators of why, within a patient, some cells are estrogen receptor positive and some cells aren’t.

CT: Are these networks avenues for intervention?

HM: Absolutely. These networks change with different interventions; when we give different hormones and hormone agonists and antagonists, some networks change and some don’t. We can manipulate these networks in a lab setting, and if we can translate that into a patient setting we can better personalize and identify therapies, you know, perhaps patient A should get anti-hormone therapy and patient B shouldn’t based on their underlying genetic networks. That would allow us to move away from more simplistic classifications of tumors into classifications that appreciate the heterogeneity of tumor cells in one patient.

CT: There’s quite a few cells in a tumor, and you are only sampling some of them, how do you know that you’re capturing enough?

HM: You don’t. You try and capture as much as possible. You don’t know if you’ve lost things especially from a patient biopsy, in which the amount of tissue is a limiting factor. From a nonpatient sample it’s not a big issue because you can capture a lot of cells. But in patient samples, especially in prostate cancer, we have to appreciate the fact that it’s more multi-focal than other diseases; one biopsy might be very different from another from the same patient. We don’t get all of the tissue from a given patient, it has to go to multiple different needs, pathology and other research needs. We try to do the best we can and factor this into our analyses.

It is also one of the challenges in working with early cancer, because the tumor is small. But that’s also an advantage of single cell because if you have a biopsy with 90 cells and only 10 cells of tumor you can at least confidently conclude that observed signals are from tumor cells, whereas if you did a bulk assay they would probably be drowned or averaged out.

CT: For people who don’t work in single cell biology, what are the biggest misconceptions about it?

HM: The issue with single-cell, like any new technology, is to ask whether your question warrants its use. The early days of single-cell sequencing provided limited information beyond creating biological atlases. However, ourselves and others are now moving towards using these technologies to ask focused questions.

CT: The technology has just been roaring ahead. Where do you see the field of single cell analysis in 5 years?

HM: I think it will be in stages. Stage one, will be primarily assaying and  explaining the existence of heterogeneity. The next stage is explaining why cells are heterogeneous. The next thing could be drugs or interventions to minimize heterogeneity, then you can have diseases that escape therapy less. The other huge and exciting avenue for this is exploring how heterogenous cancer cells interact with the  tumor microenvironment, including the immune system and stromal architecture. Can we better perform non-tumor interventions to the patient based on these data? That’s probably the hardest bit, but also one of the most promising avenues.

CT: Have you done any work with groups that are bioprinting model systems that have vasculature and 3-D architecture?

HM: We have performed some early experiments on bioprinted materials with the Sears and Gray labs. This could be a potentially exciting avenue to ask tumor vs microenvironment questions in a 3D setting.

CT: It’s been a busy time for  you.

HM: Busy but exciting. We’re applying single cell multi-omic analysis in cancer for the first time. And we’re pretty confident nobody else will anytime soon because it is extremely labor and cost intensive to establish the capabilities. The labs that have set it up are not cancer labs, they are developmental biology labs, as far as we know. I’m one of the first ones that developed the method and it still took me more than a year to get it set up here. It’s very difficult to implement. We’ve had lots of failures along the way. We needed to have the vision and the confidence that this is something worth investing in and persevering. And we’ve done that and we’re starting to get results.

CT: CEDAR, the Knight Cancer Institute’s Cancer Early Detection Advanced Research Center, does not require its scientists to support themselves through external grants, hire new employees, conduct basic lab operations, or manage equipment ordering and maintenance. Has this structure made a difference?

HM: Being at CEDAR has allowed me tremendous latitude on the risk-taking front. And a lot of our successes are because we have been able to take risks. That’s the environment CEDAR is trying to create.

Leave a Reply