Research Center

Quantitative and functional characterization of therapeutic resistance in cancer


Center Title

Quantitative and functional characterization of therapeutic resistance in cancer

Center Website

In progress

Center Summary

Fig. 1. Overview of Center and integration of Projects and Cores.

Despite tremendous advances in understanding of cancer pathogenesis, the treatment of individual patients with either conventional chemotherapy or targeted agents remains highly empiric. Current efforts to predict drug efficacy are generally focused on genetic and transcriptional markers of pathway activation or drug binding, such as resistance mutations that sterically hinder small molecule binding or activate parallel or orthogonal signaling pathways. These markers exist in a very small fraction of all cancers, such that most patients are treated with little or no understanding of whether they will respond to an individual therapy. This results in many patients receiving ineffective and/or unnecessarily toxic therapies. There is a desperate need to change this paradigm. The ideal for characterizing therapeutic sensitivity would allow for: real-time decision making, identification of rare subpopulations with therapeutic resistance, analysis of very small samples (e.g. MRD), and maintenance of viable individual cells for downstream assays to characterize phenotypic, genotypic, transcriptional and other determinants of sensitivity.

As shown in Fig. 1, the overall goal of our center is to address this need using new strategies for predicting therapeutic response in which paired phenotypic and genomic properties are measured at the single-cell level. Phenotypic properties will include both physical parameters (e.g. mass, mass accumulation rate) and molecular markers (e.g. protein secretion, surface immunophenotype) that are rapidly affected by effective therapeutics and precede longer-term phenotypes (e.g. loss of viability). Because these properties are measured for each single cell, clonal architectures based on therapeutic response will be established across each tumor sample by incorporating molecular and physical parameter data from large numbers of cells. In settings of deep treatment response, pre-treatment and MRD samples will be compared to define the effects of therapy on clonal architecture. The cells that exhibit particular functional properties (e.g. phenotypic non-responders) will be isolated and analyzed for genomic determinants of these properties. These data will then be incorporated into mathematical models to design and optimize therapeutic approaches that overcome the heterogeneity within individual tumors responsible for treatment failure. By pursuing this approach, our center will establish a framework that enables an iterative cycle between novel single-cell measurements from clinically-relevant specimens and computational approaches that result in testable predictions.


Principal Investigators

Dr. Scott Manalis

Dr. Scott Manalis has an undergraduate and doctorate degree in physics and applied physics, respectively, and his faculty appointment is in the departments of biological and mechanical engineering and an intramural member of MIT’s cancer center (Koch Institute for Integrative Cancer Research). He was the PI of MIT’s PSOC for Single Cell Dynamics in Cancer from 2012-2016 and a leader of a project and core within the center since it started in 2009. His lab developed suspended microchannel resonators for measuring the mass and mass accumulation rate of single cells with unprecedented precision – a capability that is used extensively in this center for measuring ex vivo drug sensitivity of tumor cells.

Dr. Douglas Lauffenburger

Dr. Douglas Lauffenburger is a biological engineer, formally educated in chemical engineering but with research program focused on quantitative, multi-variate studies of cell biology since beginning his academic faculty career in 1979. He is an affiliate member of the Koch Institute for Integrative Cancer Research, and has served as PI of the NCI-funded MIT Integrative Cancer Biology Program for the period 2007-2014. His expertise is in integration of computational analysis and modeling with quantitative cell biology and biochemistry experiments, toward development and testing of mathematical models for how phenotypic cell functions depend on cellular and extracellular molecular properties. He has extensive experience in combined experimental/computational studies of relating molecular properties and phenotypic behaviors on a single-cell basis.

Dr. William Hahn

Dr. William Hahn is a medical oncologist and professor of medicine at Harvard Medical School who has extensive experience in the genomic characterization and functional genomic analysis of cancers in vitro and in vivo. He is currently the Chief of the Division of Molecular and Cellular Oncology and Chair of the Executive Committee for Research at DFCI. Dr. Hahn is also an Institute Member of the Cancer Program at the Broad Institute. His lab has developed new experimental models of human cancer of defined genetic composition, created methods to perform systematic interrogation of gene function in mammalian cells and tissues and helped optimize new approaches to integrate genome scale data. Using these approaches, they have identified and credentialed new oncogenes and tumor suppressor genes and have performed preclinical studies that will form the foundation necessary for translational studies in patients.

Dr. Alex K. Shalek

Dr. Alex K. Shalek received training in physics, mathematics and chemistry, and his faculty appointment is in the Institute for Medical Engineering and Science and the department of Chemistry. He is also an Associate Member of the Broad and Ragon Institutes, where he has additional labs and access to an array of cutting-edge equipment, platforms and approaches. His expertise relevant to this proposal is on developing and utilizing nanoscale manipulation and measurement technologies to understand how small components (molecules, cells) drive systems of vast complexity (cellular responses, population behaviors). As a postdoctoral fellow, he developed a strategy that uses single-cell RNA-Seq to identify distinct cell states and circuits from the natural variation that exists between seemingly identical cells. His lab is determining how cell-to-cell variability arises from intra- and inter-cellular regulatory circuits in healthy and diseased states, as well as to explore the causes and consequences of cellular heterogeneity. These approaches are used extensively in this center.

Dr. David Weinstock

Dr. David Weinstock is an oncologist at DFCI, Associate Professor at Harvard Medical School and Associate Member of the Broad Institute. His laboratory has made a series of high-impact discoveries in leukemias, including the identification of CRLF2 as a targetable oncoprotein in poor-risk B-cell acute lymphoblastic leukemia (B-ALL), the therapeutic targeting of CRLF2-dependent leukemias with inhibitors of JAK2, HSP90, MDM2 and BRD4, the role of the nucleosome remodeling protein HMGN1 in cases with Down Syndrome, and the identification of alterations in G protein subunits as a mechanism of targeted therapy resistance. The Weinstock lab leads the effort to establish leukemia and lymphoma PDX models at the DFCI, Brigham and Women’s Hospital, Boston Children’s Hospital and Massachusetts General Hospital. He has already established a repository of >300 models serially passaged through immunodeficient mice and characterized by exome and transcriptome sequencing, and made these available open-source ( as well as through the Jackson Laboratories. He currently leads a multi-institutional Specialized Center of Research that is supported by the Leukemia and Lymphoma Society and focuses on lymphoid cancer biology and targeting.

Dr. Christopher Love

Dr. Christopher Love has an undergraduate and doctoral degree in physical chemistry, and his faculty appointment is in the department of chemical engineering. He is an intramural member of the Koch Institute for Integrative Cancer Research and an associate member at both the Broad Institute and Ragon Institute. He has been part of a working group on systems biology and engineering solutions for sparse samples in mucosal immunology (DAIDS/HVTN MIG). He has served as the primary PhD thesis advisor of graduate students from the departments of chemical engineering and biological engineering. His lab developed the nanowell technology that is used in our center. It has been optimized in all aspects of the technology, including imaging of cells in nanowells, time-resolved, multiplexed analysis of cytokines, and software tools for data extraction. New advances in plate formats allow for 24-array processing in parallel, and up to 16-channel imaging cytometry on a microscope. These capabilities will be employed in both projects.

Ms. Kristen Stevenson

Ms. Kristen Stevenson is a biostatistician and has worked for the past 10 years at DFCI in research of hematologic malignancies primarily including myelodysplastic syndrome, chronic lymphocytic leukemia, and acute lymphoblastic leukemia. She also has undergraduate degrees in chemical engineering, computer science, and industrial mathematics and statistics. Her expertise is in statistical modeling and analysis, power and samples size calculation, and experimental and clinical trial design.


There are three novel single-cell analysis approaches that extensively utilized in our center:

[1] Suspended microchannel resonator. Our approach for measuring cell growth, or mass accumulation rate, is based on a sensing technology known as the suspended microchannel resonator (SMR). When an object that is denser than water passes through the SMR, the net increase in mass (i.e. the buoyant mass of the object) lowers the resonant frequency. By continually measuring buoyant mass when the cell travels back and forth through the sensor, growth of individual bacteria, yeast, and mammalian cells has been quantified with a precision that has not been achievable by conventional microscopy. In our center, we will use a novel SMR array-based technique for high-throughput mass accumulation rate measurements that increases throughput by 100-fold without sacrificing precision. This is accomplished by an array of SMRs microfluidically connected in series, with “delay” channels in between each cantilever (Fig. 2). These delay channels give the cell time to continue changing mass as it flows between cantilevers. After a cell exits a cantilever, other cells are free to enter it and be weighed. We are not limited to flowing only one cell through the array at a time, but can have many cells flowing through the array in a queue, enabling precision growth measurements with a throughput approaching 100 cells per hour per device. In a collaboration between MIT and DFCI researchers, the serial SMR device has been validated for ex vivo drug susceptibility testing in several cell lines, patient-derived cell lines and primary cells from mouse models.

Fig. 2: A) Micrograph of serial SMR device. Cantilever lengths range from 380 to 470 μm to provide unique resonant frequencies, which are readout with integrated piezoresistors. The MEMS foundry CEA-LETI recently fabricated several 8” wafers, each of which includes ~200 hundred devices. B) Buoyant mass data from mouse lymphoblast cells extracted from frequency shifts of cantilevers, color-coded by cantilever. 7 and 9 μm polystyrene particles are added as calibration and negative control, respectively. Each cell requires ~20 min to pass through the serial array, which is sufficient for measuring its mass accumulation rate with high precision. Over the course of incubation with a selected drug (2-24 hrs), the serial SMR can detect changes in the growth of single cells to predict therapeutic response without the need for extended culture. In some cases, measuring single cell growth can result in drug sensitivity test that is sufficient more rapid than viability. This is important since the properties of primary cells change quickly when moved to an ex vivo environment.

Through a seed grant from the MIT/DFCI Bridge program (co-Principal Investigators: Manalis, Weinstock), two serial SMR platforms are now being operated in CLIA-approved space at the DFCI/Brigham and Women’s Cancer Center by a dedicated technician in the Weinstock lab. This was an essential step toward translating the SMR into a clinical assay for predicting therapeutic response in heterogeneous tumor populations directly from primary specimens.

[2] Single-cell genomic profiling. scRNA-Seq, an approach co-developed by the Shalek Lab in both basic and clinical settings, is utilized throughout Projects 1 and 2. In previous work, we have established and validated robust experimental and computational pipelines for scRNA-Seq. This work has demonstrated that scRNA-Seq enables unbiased identification of the cell types, states, and circuits that drive complex biological systems, such as tumors, at the single-cell level. More specifically, in the context of human cancer biopsies, we have optimized strategies for the capture and lysis of individual cancer, stromal, and immune cells, and developed computational approaches to identify the cell states of each and their molecular underpinnings. Here, scRNA-Seq will enable not only the precise identification of malignant and non-malignant cells from the bone marrow microenvironment, but also the nomination of putative extracellular factors that influence malignant cell drug responses through the identification of co-expressed gene programs downstream of secreted molecules or receptor-ligand pairs. Recent work in the Shalek Lab, performed in collaboration with the Regev (Broad, MIT) and Garraway (DFCI) Labs, identified intratumor and intertumor heterogeneity in melanoma biopsies and discovered patient-specific immune and stromal cell populations that correlate with prognosis and potentially explain some instances of resistance (Fig. 3). Here we propose to specifically examine the phenotype of cells with unaffected MAR in the presence of drug using scRNA-Seq in order to define resistant cell states and intuit treatment strategies that target them.

Fig. 3: scRNA-Seq in human cancer. scRNA-Seq of primary human melanoma samples demonstrates (A) inter- and intratumor heterogeneity among malignant (left) and non-malignant (right) cells. Further, in the context of malignant cells, (B) substantial variability is observed in the fractional abundance of a drug-resistant cell state (AXL-high) across patient samples (labeled MelXX). Adapted from Tirosh et al, Science, 2016.


[3] Nanowells. We will utilize a nascent technology for single-cell analysis that relies on dense arrays of subnanoliter wells to profile the phenotypic and molecular properties of tumor cells and for precise control of a cell’s microenvironment (Project 2). The majority of examples, and the development of this technology to date, has emphasized the characterization of the diversity and functions of antigen-specific B- and T-cells of the immune system. The Love Lab has developed a strategy for bioanalytics that uses these arrays in defined processes comprising one or more modular unit operations—each one reporting on different characteristics of the cells analyzed, such as phenotypic lineages or functions (Fig. 4). Examples include on-chip cytometry (51), quantitative measures of secreted proteins by a printing method called microengraving, single-cell cytolysis assays and gene expression. To date, this nanowell-based platform has been used to characterize mouse hybridomas, mouse B and T cells, yeast cells, and human PBMCs.

Innovative aspects of the nanowell technology relevant to this proposal include:

  • Modular operations that link complex phenotypic traits like viability to genotypic features (clonal lineages determined by single-cell sequencing)
  • Resolution and recovery of extremely rare cells (~1 in 10,000 or fewer), including circulating tumor cells, for clonal expansion or single-cell sequencing
  • Deep immunophenotyping by multi-spectral imaging cytometry (MuSIC) for up to 16 channels of data (comparable to state-of-the-art flow cytometers)
  • Quantitative measures of secretion from one or a few cells with sensitivities 10-100x greater than conventional clinical assays like ELISpot
  • Discrete co-cultures of multiple cells to measure intercellular communication among tumor cells, or mixed populations.




Fig. 5. Functional characterization of therapeutic resistance in cancer.Cells are treated in vivo before target populations are extracted, purified and measured immediately upon removal from the mouse or patient. Alternatively, the target population is treated ex vivo prior to the measurement. Upon measurement of mass accumulation rate (MAR), each cell is sorted into a well and lysed within ~1 min, thereby allowing scRNA-seq of a given cell to be linked to its MAR. Cells that maintain MAR similar to vehicle-treated cells after sufficient drug exposure are considered to be non-responders. The overall goal of this proposal is to determine mechanisms of resistance and make testable predictions to overcome that resistance through transcriptome analysis of phenotypically-defined single cells.

As illustrated in Fig. 5, scRNA-Seq will be acquired from cells that are determined by the SMR growth assay to be non-responders to a particular treatment (i.e. phenotypically drug resistant). These data will provide unprecedented insight into: i) pathways that distinguish response after target inhibition compared to untreated cells, ii) whether these same pathways distinguish response upon relapse and thus could be predicted at the time of minimal residual disease, and iii) pathways that associate with differential response to targeted therapeutics between different compartments (e.g. bone marrow and peripheral blood leukemia cells at the untreated and relaps ed time points). Based on these findings, we expect to nominate combination strategies that target the phenotypically-resistant cells as well as come up with drugs that might be appropriate targets currently not appreciated as being relevant for a particular cancer. Project 1 will focus on cell-intrinsic properties of tumor cells while Project 2 will focus on cell-extrinsic effects through the use of nanowells to control paracrine and juxtacrine interactions that mimic the microenvironment.


Project 1. Systematic discovery of cell-intrinsic mechanisms of cancer drug resistance

PIs: Manalis (lead) and Lauffenburger (co-lead), Shalek, Weinstock and Hahn
We aim to utilize high precision single-cell growth measurements together with single-cell RNA-Seq (scRNA-Seq) to profile the intrinsic factors that inform the responses of individual cancer cells to therapeutic interventions. We will ask to what extent paired phenotypic and transcriptomic measurements can identify pathways that mediate cell autonomous resistance and highlight therapeutic approaches to overcome that resistance. Cancer cells will be isolated from primary tumors or from patient-derived cell lines/xenografts of both leukemias (as a liquid tumor model) and colon/pancreatic cancers (as a solid tumor model). In contrast to Project 2, cells will be measured in isolation without mimicking aspects of the microenvironment. Over a period of many hours, we will examine distinct phenotypic attributes of the cells (mass and mass accumulation rate) and link these attributes to the transcriptome at the single-cell level. We will then determine cell intrinsic mechanisms for resistance by analyzing transcriptomic features of responding and non-responding tumor cells.

Project 2. Systematic discovery of cell-extrinsic mechanisms of cancer drug resistance

PIs: Shalek (lead), Hahn and Love (co-leads), Weinstock, Lauffenburger and Manalis
We aim to utilize high precision single-cell growth measurements and single-cell RNA-Seq (scRNA-Seq) together with nanowells to systematically examine how the extracellular factors present in leukemias and colon and pancreatic cancers influence drug responses. First, we will identify the signals and non-malignant cells present in each tumor type by performing scRNA-Seq on primary biopsies. We will then generate an atlas of implicated cell types/states and putative signaling molecules that may influence cancer cell drug responses in vivo. Second, we will use high precision single-cell growth measurement and nanowells to systematically uncover how these soluble factors and tumor cells inform on cancer cell drug responses; we will similarly examine the impact of previously implicated environmental factors as well as other cancer cells (of the same and different intrinsic states; from Project 1). By analyzing and modeling our results, we will uncover previously unknown microenvironmental synergies (e.g., cytokines, receptor-ligand pairing) that may modulate cancer cell drug responses in vivo. Collectively, these aims will afford an unprecedented view of the tumor microenvironment and shed light on current therapeutic bottlenecks, while suggesting potential new and more effect therapeutic inroads for treating cancer.


Core 1: Biospecimens and Patient-derived xenografts

PIs: Hahn (lead) and Weinstock (co-lead)
The overarching goal of Core 1 is to provide the infrastructure and professional expertise needed to bank primary leukemia, colon, and pancreatic cancer specimens, establish patient-derived xenograft (PDX), organoid, and cell models, and make these patient-derived resources available to all Program investigators, thereby facilitating the translational and laboratory-based research performed by CSBC investigators in Projects 1 and 2. Both projects will require access to patient samples and cells derived from PDX and organoid models. The core supports 3 functions: (1) cryopreservation and distribution of primary cells from patients with leukemia, colon, and pancreatic cancer, (2) establishment, characterization, and distribution of xenograft, organoid, and cell culture models of these diseases, and (3) isolation of cell populations from primary tumor samples and tissue models for single cell analysis. The functions of the core include documentation and tracking of informed consent, isolation and cryopreservation of viable tissue, plasma or serum, and genomic DNA from patient samples and distribution of samples to CSBC investigators. A large number of primary leukemia, colon and pancreatic tumor samples are currently available in our existing repositories for use by Project Investigators. Additional samples will be acquired from patients routinely evaluated in our gastrointestinal oncologic and hematologic malignancies clinics. In addition to distributing primary cells to program investigators, Core 1 will also maintain and distribute a bank of well characterized PDXs selected for passage in immunocompromised mice. Methods for engrafting leukemia cells and primary epithelial tumor tissues in immunodeficient mice are now well established, and we and others have successfully used PDX models for drug testing. Organoid culture approaches will be used to complement the PDX effort, and will provide an additional way of propagating and segregating patient-derived epithelial tumor cells and stromal cells. Core 1 will maintain a bank of PDX models and organoids and distribute to program investigators as needed. Overall, Core 1 has extensive experience in the processing, cryopreservation, flow cytometric analysis, purification, propagation, banking, and genomic characterization of primary specimens and tumor models, as well as their use in preclinical drug evaluation.

Core 2: Computational Analysis Core

PIs: Lauffenburger (lead), Shalek and Stevenson
The Computational Analysis Core (Core 2) will support Projects 1 & 2 by bringing necessary bioinformatics and computational methods to bear on the questions addressed by each project. More specifically, we will apply established computational pipelines to perform quality control on, and extract maximal information (cell types, states, circuits, and drivers) from, our rich data sets. We will help our CSBC team identify the cell types, states, and circuits active in the Leukemia and Colon and Pancreatic tumor models of Core 1 from the single-cell RNA-Seq data collected in Projects 1 & 2, as well as the molecular drivers of interesting behaviors. We will deploy numerical algorithms to uncover the predictive power of phenotypic measurements, alone or in combination, made via our Nanowell (NW) and Suspended Microchannel Resonator (SMR) platforms to explain cancer cell drug responses. When necessary, we will identify and/or develop new algorithms, enhancing our CSBC’s quantitative and multivariate capabilities. Finally, we will work closely with our CSBC team members to help with experimental design, ensuring that an appropriate number of cells/samples are processed to test hypotheses of interest and that appropriate statistical testing is performed. By providing comprehensive analysis (QC, power, gene set enrichment) and modeling capabilities, we will dramatically enhance our ability to draw crucial insights from our data. This will lead to new hypotheses that can be tested by designing new experiments.

Education and Outreach Core

PI: Lauffenburger
The mentorship of current and future trainees who can tackle cancer-related problems with computational systems biology approaches is an integral part of fulfilling our commitment to catalyze and generate new bodies of knowledge and fields of cancer study. To achieve this goal, we will: 1) Establish graduate student fellowships for students jointly mentored in computational systems biology, precision measurement or oncology. 2) Provide undergraduate research opportunities for MIT students to work in laboratories at DFCI. 3) Provide outreach to the biotech/pharmaceutical industry. 4) Establish an NCI CSBC Junior Investigators program. 5) Facilitate monthly meetings and annual retreats that will be open to the MIT/DFCI community. 6) Offer mini-courses and training in experimental and computational methods. 8) Provide outreach to the community. and 7) Establish and maintain a website for disseminating research activities of our center as well as relevant techniques and applications.

Administrative Core

The overall mission of the Administrative Unit is to provide oversight and coordination on administrative and fiscal aspects of the MIT/DFCI CSBC.