Join MySLAS Social

Identifying Cancer Cell Subsets and Therapeutic Responses with High-Content Single-Cell Screening

Increasing drug resistance and recent cancer stem cell research have revealed a critical set of potent cancer cells that continue to wreak havoc on patients even when the majority of proliferating cells have been eliminated by effective therapies. But finding ways to better understand these cellular subsets and how they respond to treatment in individual patients remains a challenge, according to Tiffany (TJ) Chen, Ph.D., director of Informatics at Cytobank and a researcher at Stanford University.


One answer to the challenge is single-cell mass cytometry, a combination of flow cytometry and mass spectrometry that enables the measurement of up to 40 parameters simultaneously in a single cell. This and fluorescence flow technology enabled Chen, working in the Stanford University (Palo Alto, CA) laboratories of Garry Nolan, Ph.D. and Serafim Batzoglou, Ph.D., with colleagues Matthew Clutter, Ph.D., now at Northwestern University (Evanston, IL); Nikesh Kotecha, Ph.D, cofounder of Cytobank (Mountain View, CA); Karen Sachs, Ph.D.; and Wendy Fantl, Ph.D. (Stanford, CA) to help create a system that can target identification in drug discovery efforts as well as identify the right combination of therapies for individual patients in the clinic. The system uses novel tagging technology, equipment, imaging and software to detect and identify cancer cell subsets based on their responses to therapies.

Getting Started

As a graduate student, Chen discovered the importance of ensuring that experimental designs and computational models work together. "It really clicked when I was taking my basic coursework—generating my own microarrays in one course but at the same time doing microarray analysis and learning dynamic programming and statistical methods in my other classes. Taking those courses at the same time made me see how limiting it was to consider them in isolation. Experimental design has to work synergistically with computational models. Otherwise, you can have an elegant algorithm that doesn't achieve any real results."

"Anyone in a Ph.D. program who truly believes in what he or she is doing dreams big," Chen continues. "I joke that every graduate student who joins the Stanford Biomedical Informatics program dreams of a Nobel prize and a Turing award. Of course, that's not going to happen. But we do end up really and truly believing in the value of both biological data and computational/statistical methods."

As a post-doctoral research fellow at Stanford and now also at Cytobank, Chen has been able to use her understanding and drive to help make a difference in cancer research. It began with the Nolan Lab's work with the CyTOF ("cytometry by time-of-flight") machine, invented by University of Toronto professor Scott Tanner and initially distributed by DVS Sciences (acquired in 2014 by Fluidigm). "Flow cytometry and fluorescent cytometry, the traditional methods used to analyze and sort cells to help diagnose disease, both have limitations," Chen says. "The main problem is that the fluorescent dyes used to tag antibodies limit the number of proteins that can be measured simultaneously to 15-18 at most. In addition, certain therapeutics cause autofluorescence, which confounds results."

By contrast, with mass cytometry, antibodies are tagged with rare earth metal isotopes instead of fluorophores, thereby eliminating light-related problems. Cells of interest, labeled with the metal-tagged antibodies, move through the machine and then are analyzed by a mass spectrometer.

"The fact that we can measure so many parameters at once is really huge, because so much of the work in academia and industry involves measuring all the cell types in a sample. Until now, to do that, I would have had to build, say, five different experiments with five different groups of biomarkers for each cell type. That's time consuming and it still doesn't give you a picture of all the cell types together. Also, it's especially difficult to find rare cell types that might take a lot of different biomarkers to identify."

Homing in on Cells and Cell Cycles

At Stanford, Peter Krutzik, Ph.D. (co-founder of Cytobank) and Clutter built an automation system with a custom robotic arm and several plate readers and dispensers, with the goal of speeding up the more than 100 preparation steps required to get samples ready for testing in the CyTOF, Chen explains. "The components for the first iteration of the automation platform were bought from online auctions. Today, the system has been refined and newer members of the lab such as Zach Bjornson have extended this to work with the FDA; industry and research cores are using similar systems to provide contract research services. What I brought to the table then and what we do at Cytobank now is help develop computational methods to deal with all the additional data that results from these experiments."

Some of the earliest work on single-cell mass cytometry coming out of the Nolan Lab was published in Science in 2011. The team used the technology to examine healthy human bone marrow, measuring 34 parameters simultaneously (as opposed to the 10-15 parameters measured by conventional methods). In addition, the signaling behavior of cell subsets was examined in response to various stimuli, including small-molecule therapeutic agents. Nolan and his colleagues wrote:

"In this study of the immune system, coupling classical phenotypic organization to cellular functional responses was unrestricted by the inherent limitations imposed by fluorescence. This merging provided a systems-level view of human hematopoiesis and immunology from the perspective of immunophenotype and coupled it to underlying events as measured through receptor engagement and small-molecule drug actions."

Simply put, that study established a reference, or map, against which diseases of the immune system and certain types of cancers could be compared, as well as a process by which the effects of candidate drugs might be evaluated, Chen explains. A subsequent study expanded the approach to include measurements of cell cycle phases in immune cell subsets in healthy human bone marrow, and in various human and mouse tumor cell lines.

"If you open up David Morgan's cell cycle textbook, you see diagrams of cyclins as they move through the cell cycle. Those diagrams represent decades of scientific knowledge," Chen says. "But now, with just a single experiment, we can recreate all of the diagrams by putting them through a computational pipeline to identify cell cycle trajectories. Not only does the map recapitulate what we know, but it also allows us to insert new markers that aren't traditionally thought of as cell cycle proteins, but that we know are associated with cancer."

Homing in on Specific Drug Responses

The team then took mass cytometry profiling to the next level with mass-tag cellular barcoding, a cell-based multiplexing technique that improves sample throughput, reduces antibody consumption and helps ensure uniformity of the antibody stain across samples.

"This type of high-content, high-throughput screening could play an important role in drug discovery as well as preclinical testing with existing drugs in advance of controlled trials," Chen says. Indeed, after Nolan received the Ovarian Cancer Research Program's Teal Innovator Award, Chen is continuing to contribute to the lab's efforts to use mass cytometry to identify relationships among cells in individual patients, with investigator Fantl as well as Veronica Gonzalez Muñoz, Ph.D. and Nikolay Samusik, Ph.D.

Leveraging their fluorescence flow cytometry work as well, the team found that the cell cycle and cell signaling seem to be more important than cell death as a means of stratifying cellular responses to therapy. "For the screening that led to this finding, we treated cells for 24 hours, and that's when we discovered that cell cycle signaling yielded much more information on what was happening to the cells compared to just looking at cell death," Chen explains. "Then we did another experiment where we looked at very early points—one hour, four hours and six hours after treatment. And we noticed the mechanisms of different therapeutics changing as early as one hour post-treatment, especially in response to DNA damage. This is important because most researchers who do a general screening look at 24 or 48 hours; they don't think that there might be some critical points early on. As it turned out, the changes in DNA damage we saw early in the cell cycle, before the S phase, really differentiated the therapies."

The finding has potential application in the clinic, Chen says. "I imagine that some day a patient undergoing cancer treatment will have their blood drawn and tissue collected at each doctor's visit, and that every week throughout therapy, we'll be profiling the effects of treatment, looking for rogue cells—hidden cells that are mutating and might become the next group of cancerous cells that must be treated. It's critical to understand not just the treatment, but how the patient has evolved during therapy. It makes sense that with immunotherapy, we need to monitor the immune system and cell signaling and cycling to understand how these things are dynamically changing—and what the implications might be for treatment."

Data Interpretation

Before this can happen, as indicated earlier, researchers need to meet the challenge of data analysis and interpretation, which is Chen's area of expertise. That work continues to evolve. For the Science study, the Nolan team developed an algorithm called SPADE to handle the bioinformatic demands of analyzing the grouping of thousands of cells by 34 different parameters (version 2.0 of the software is freely available online). Subsequently, in collaboration with Columbia University (New York, NY) researchers, the team also developed viSNE (visual interactive Stochastic Neighbor Embedding), which is based on a tool that translates high-dimensional data into two-dimensional visual representations. viSNE was able to identify previously unrecognized heterogeneity in bone marrow cells and to detect minimal residual disease (~20 cancer cells among tens of thousands of healthy cells), which is important for efforts to treat and cure cancer, Chen says. viSNE is also available on the Cytobank platform as well as from the Pe'er lab at Columbia.

Most recently, the Nolan Lab developed "scaffold maps"—algorithms that transform the single-cell mass cytometry data into intuitive maps that provide interpretations of the organization of immune cells in various organs in humans and mice. The raw data from the experiments and a package for viewing the maps are available on the Cytobank website.

"Different cancer drugs target different aspects of the cell," Chen explains. "We have folate inhibitors, kinase inhibitors and other newer therapies. The idea is to put all the mechanistic information into a universal context, so we can directly compare the systemic mechanisms of action in a single framework, qualitatively and quantitatively. Then we can look side by side at, for example, dasatinib and methotrexate and see that a drug that's been used for more than 60 years and a drug that's only been on the market for five years or so both target an early part of the cell cycle and cause signaling profiles to change. This information has potential not only for drug repurposing but also for combination therapies, which we'll be studying in the future."

Learn More at SLAS2016

At SLAS2016, Chen will describe the latest applications of her computational work in high-content fluorescence and mass cytometry analysis, including the creation of a "new landscape" for the classification of cancer therapies and the identification of rare and differential cell subsets. Her podium presentation, "Identifying Druggable Cells: Automated Methods for High-Content Single-Cell Screening," is scheduled on Wednesday, Jan. 27, from 1:30 – 2:00 p.m. at the San Diego Convention Center. Chen is one of nine presenters competing for the 2016 SLAS Innovation Award and $10,000 cash prize at SLAS2016.

"Chen and many other SLAS early career professionals are actively contributing their expertise and enthusiasm to innovative and potentially ground-breaking projects; others are looking for opportunities to realize their goals," commented Ellen Berg, SLAS2016 Informatics Track Chair. "SLAS2016 offers researchers at all levels many opportunities for networking, collaborating and career enhancement."

January 4, 2016