Search

Significance Single-cell high-throughput technologies enable the ability to identify combination cancer therapies that account for intratumoral heterogeneity, a phenomenon that has been shown to influence the effectiveness of cancer treatment. We developed and applied an approach that identifies top-ranking drug combinations based on the single-cell perturbation response when an individual tumor sample is screened against a panel of single drugs. This approach optimizes drug combinations by choosing the minimum number of drugs that produce the maximal intracellular desired effects for an individual sample., An individual malignant tumor is composed of a heterogeneous collection of single cells with distinct molecular and phenotypic features, a phenomenon termed intratumoral heterogeneity. Intratumoral heterogeneity poses challenges for cancer treatment, motivating the need for combination therapies. Single-cell technologies are now available to guide effective drug combinations by accounting for intratumoral heterogeneity through the analysis of the signaling perturbations of an individual tumor sample screened by a drug panel. In particular, Mass Cytometry Time-of-Flight (CyTOF) is a high-throughput single-cell technology that enables the simultaneous measurements of multiple (>40) intracellular and surface markers at the level of single cells for hundreds of thousands of cells in a sample. We developed a computational framework, entitled Drug Nested Effects Models (DRUG-NEM), to analyze CyTOF single-drug perturbation data for the purpose of individualizing drug combinations. DRUG-NEM optimizes drug combinations by choosing the minimum number of drugs that produce the maximal desired intracellular effects based on nested effects modeling. We demonstrate the performance of DRUG-NEM using single-cell drug perturbation data from tumor cell lines and primary leukemia samples.

Purpose: To identify an optimal drug combination for addressing intratumor heterogeneity by measuring single drug responses at the level of single cells. Background: Drug combination strategies have been derived from drug-drug and drug-target interaction inferences made based on genomic network analysis, such as synthetic lethality network analysis and protein network analysis. A key limitation with these approaches is that they do not account for intratumor heterogeneity. A next generation high throughput single cell technology such as Cytometry Time-of-Flight (CyTOF) has the potential to circumvent this limitation by providing the key information for optimally targeting intratumor heterogeneity within the context of personalized medicine. Methods: We propose a systems level modeling framework for analyzing heterogeneous drug response data across a population of single cells. Our approach, called Mixture Nested Effects Models (MNEM), identifies multi-target drug combinations from single drug effects measured at the level of single cells. MNEM analyzes drug responses among subpopulations (aka clusters) of similar cells, where the subpopulations are defined by cellular surface markers. MNEM infers the nested drug target hierarchy (directed network) from a heterogeneous population of drug effects by combining subpopulation identification, subpopulation matching to infer the drug response pre- and post-treatment, and nested effects modeling for each subpopulation individually or through a mixture of subpopulations. An MNEM infers that Drug A targets are a superset of Drug B targets, if the heterogeneous downstream target effects resulting from Drug B are a noisy subset of those resulting from Drug A. This relationship is represented as a directed edge from Drug A to Drug B. From the graphical model of nested drug effects, a scoring function is superimposed to identify the optimal drug combinations. Results: We apply MNEM to CyTOF drug screening data derived from treating three well-established cancer cell lines: HeLa (ovarian), NCIH460 (lung) and HCT116 (colon) with Mek, mTOR, pP38MAPK, JnK I and PI3K inhibitors at different optimal dose levels in addition to the base line treatment TRAIL, an apoptotic stimulator. MNEM indicates that pP38MAPK and Mek inhibitors are important candidates for targeting the ovarian and lung cancer cell lines. The MNEM analysis from the colon cancer cell line data indicates that several optimal drug regimens may have equivalent efficacy. Survival data from drug combination invitro experiments using clonogenic assays support the results from MNEM. Overall, these analyses suggest that MNEM could be used to optimize drug combinations based on responses from single drugs measured at the single cell level. Conclusion: Intratumor heterogeneity in tumor response, measured at the level of single cells, upon exposure to single drugs may be combined through Mixture Nested Effects Modeling (MNEM) to infer optimal drug combinations. Citation Format: Benedict Anchang, Harris Fienberg, Sean Bendall, Robert Tibshirani, Sylvia K. Plevritis. Multi-target drug combinations from single drug responses measured at the level of single cells using Mixture Nested Effects Models (MNEMs) applied to cancer. [abstract]. In: Proceedings of the AACR Special Conference on Computational and Systems Biology of Cancer; Feb 8-11 2015; San Francisco, CA. Philadelphia (PA): AACR; Cancer Res 2015;75(22 Suppl 2):Abstract nr B1-39.

This work summarizes cellular apoptotic and signaling response to TRAIL across 10 cell lines and for the first time links signaling diversity to nongenetic resistance. This high-dimensional, single-cell approach toward TRAIL resistance sets a standard in studying nongenetic resistance., TNFα-related apoptosis-inducing ligand (TRAIL), specifically initiates programmed cell death, but often fails to eradicate all cells, making it an ineffective therapy for cancer. This fractional killing is linked to cellular variation that bulk assays cannot capture. Here, we quantify the diversity in cellular signaling responses to TRAIL, linking it to apoptotic frequency across numerous cell systems with single-cell mass cytometry (CyTOF). Although all cells respond to TRAIL, a variable fraction persists without apoptotic progression. This cell-specific behavior is nonheritable where both the TRAIL-induced signaling responses and frequency of apoptotic resistance remain unaffected by prior exposure. The diversity of signaling states upon exposure is correlated to TRAIL resistance. Concomitantly, constricting the variation in signaling response with kinase inhibitors proportionally decreases TRAIL resistance. Simultaneously, TRAIL-induced de novo translation in resistant cells, when blocked by cycloheximide, abrogated all TRAIL resistance. This work highlights how cell signaling diversity, and subsequent translation response, relates to nonheritable fractional escape from TRAIL-induced apoptosis. This refined view of TRAIL resistance provides new avenues to study death ligands in general.

In fluorescence-based flow cytometry, cellular viability is determined with membrane-impermeable fluorescent reagents that specifically enter and label plasma membrane-compromised non-viable cells. A recent technological advance in flow cytometry uses antibodies conjugated to elemental metal isotopes, rather than to fluorophores, to allow signal detection by atomic mass spectrometry. Unhampered by the limitations of overlapping emission fluorescence, mass cytometry increases the number of parameters that can be measured in single cells. However, mass cytometry is unable to take advantage of current fluorescent viability dyes. An alternative methodology was therefore developed here in which the platinum-containing chemotherapy drug cisplatin was used to label cells for mass cytometry determinations of live/dead ratios. In a one-minute incubation step, cisplatin preferentially labeled non-viable cells, from both adherent and suspension cultures, resulting in a platinum signal quantifiable by mass cytometry. This protocol was compatible with established sample processing steps for cytometry. Furthermore, the live/dead ratios were comparable between mass and fluorescence based cytometry. Importantly, although cisplatin is a known DNA-damaging agent, a one-minute “pulse” of cisplatin did not induce observable DNA damage or apoptotic responses even within 6 hours post-exposure. Cisplatin can therefore be used as a viability reagent for a wide range of mass cytometry protocols.

The influence of host and parasite genetic background on infection outcome is a topic of great interest because of its pertinence to theoretical issues in evolutionary biology. In the present study, we use a classical genetics approach to examine the mode of inheritance of infection outcome in the crustacean Daphnia magna when exposed to the bacterial parasite Pasteuria ramosa. In contrast to previous studies in this system, we use a clone of P. ramosa, not field isolates, which allows for a more definitive interpretation of results. We test parental, F1, F2, backcross and selfed parental clones (total 284 genotypes) for susceptibility against a clone of P. ramosa using two different methods, infection trials and the recently developed attachment test. We find that D. magna clones reliably exhibit either complete resistance or complete susceptibility to P. ramosa clone C1 and that resistance is dominant, and inherited in a pattern consistent with Mendelian segregation of a single-locus with two alleles. The finding of a single host locus controlling susceptibility to P. ramosa suggests that the previously observed genotype–genotype interactions in this system have a simple genetic basis. This has important implications for the outcome of host–parasite co-evolution. Our results add to the growing body of evidence that resistance to parasites in invertebrates is mostly coded by one or few loci with dominance.

Currently, cancer drug combinations primarily focus on mutational heterogeneity of the primary tumor and do not account for single-cell variations that can give rise to drug resistance. Moreover, even with the increasing number of potential FDA-approved targeted drugs including immunotherapies, methods are needed to identify better combination therapy that leverages intratumor heterogeneity, thereby potentially mitigating the need for trials with large numbers of patients. Despite advances in single-cell technologies that capture intratumor heterogeneity, there are no drug combination strategies that utilize single-cell platforms. One idea that has been proposed is to use Mass Cytometry Time-of-Flight (CyTOF) for drug screening by producing drug perturbation effects at the level of the single cell, but the required analytics for the resulting complex data were not addressed. To address this unmet need, we have developed a novel algorithm to optimize combination therapy for an individual patient by analyzing distinct single-cell drug perturbation responses on a tumor sample. This model framework, called “DRUGNEM,” can be applied to CyTOF data, single-cell RNA-seq, or any single-cell imaging data currently available. DRUGNEM is composed of three steps: (1) identify the subpopulations that make up the tumor and may respond differently to treatment; (2) reconstruct a drug-nested-effects model that integrates the drug effects across all subpopulations to capture sub-setting relationships among individual drug effects; and (3) systematically score potential drug combinations to identify or prioritize strategies that will be clinically (and economically) sustainable. Currently, DRUGNEM is optimized to select the minimum number of drugs that produces the maximal desired intracellular effect, predicated on the premise that fewer drugs lower treatment-related toxicities and costs, but the final selection criterion can be easily modified. As proof of concept, we applied the DRUGNEM framework to individualize drug combinations based on CyTOF data generated on de-identified malignant research samples from 30 ALL pediatric patients before and after exposure to 3 targeted FDA -pproved single drugs (dasatinib, tofacitinib and BEZ235). We found that the most common combination treatment strategy (dasatinib and BEZ235) might not be optimal for all 30 ALL patients, with 2 of the 30 likely responding best to tofacitinib alone. Using in vitro survival assays, we validated the DRUGNEM prediction of BEZ235 and dasatinib as a potential synergistic combination on ALL cell lines. In summary, DRUGNEM is a novel framework using single-cell technologies to guide drug-combination strategies and can be adapted to incorporate complementary molecular data and computational methods to ultimately achieve more effective therapy for the individual cancer patient. Citation Format: Benedict Anchang, Kara Davis, Harris Fienberg, Sean Bendall, Loukia Karacosta, Garry Nolan, Sylvia K. Plevritis. Individualized drug combination based on single-cell drug perturbations [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 2275.

Millions of seahorses a year are traded internationally, with many eventually sold at retail shops as curios, as aquarium pets, and especially for use in traditional medicine. The entire genus is now protected by CITES, but conservation measures have been limited in part by difficulties with species identification and incomplete understanding of trade patterns. In this study, we use molecular techniques to make species- and population-level identifications of 56 seahorses sampled from both traditional medicine and curio shops in San Francisco and central California. Seahorses from medicinal sources included unexpectedly large numbers of the eastern Pacific Hippocampus ingens, a species poorly protected by current CITES export recommendations. Curio shops were dominated by H. barbouri, a spiny species that has been reportedly confused in trading records as H. histrix. Specific populations of origin could be inferred for several species through comparison with publicly available phylogeographic data. Our results underscore the strengths and weaknesses of current recommended export regulations, and we suggest that molecular forensics can help in verifying trade documentation and developing more effective conservation measures.

Nongenetic resistance has recently been described as a major impediment to effective cancer therapy. Nongenetic resistance is challenging to study since it occurs nonuniformly, even in cell lines, and can involve the interplay of multiple survival pathways. Until recently, no technology allowed measurement of large-scale alterations in survival pathways with single-cell resolution. Mass cytometry, a flow-based technique in which the activation of up to 50 proteins can be measured simultaneously in single-cell, now provides the ability to examine nongenetic resistance on the functional level on a cell-by-cell basis. The application of mass cytometry, in combination with new bioinformatic techniques, will allow fundamental questions on nongenetic resistance to be addressed: Is resistance caused by selection of cells with a pre-existing survival phenotype or induction of a survival program? Which survival pathways are necessary for nongenetic resistance and how do they interact? Currently, mass cytometry is being used to investigate the mechanism of nongenetic resistance to TRAIL-induced apoptosis. The approaches being developed to understand resistance to TRAIL will likely be applied to elucidate the mechanisms of nongenetic resistance broadly and in the clinic.

Acute myeloid leukemia (AML) can present with multiple concurrent subclones at diagnosis. Subclone-specific mutations may confer resistance to molecular-targeted drugs through loss of antigen expression or rewiring of intracellular signaling pathways, leading to relapse. Deep sequencing approaches improve the detection of rare subclones, but the ability to prospectively identify subclones (i.e. without relapse material) is limited, and the effects of subclone-specific mutations on phenotype are poorly understood. In the course of a broader study of oligoclonal pediatric AML patients, we focused investigation on a diagnosis bone marrow sample from a patient harboring 3 presumed subclones (two with distinct NRAS-G12D, NRAS-G13D mutations) as determined by whole-genome sequencing (WGS). We employed a combination of 31-parameter mass cytometry, deep sequencing, FACS sorting, and computational modeling to produce a detailed profile of the subclonal genotypes and phenotypes in this patient. The 3 anticipated subclones did not correlate with a clear subset of surface markers in the mass cytometry analysis. Instead, we observed a single continuous ‘differentiation trajectory’ from progenitor-like to monocyte-like blasts. To dissect this trajectory, 7 distinct myeloid/progenitor subsets were FACS-sorted, plus T/B cells as a non-leukemic control. For each subset we performed capture-based deep sequencing of 307 tumor-specific variants (Tier 1: 13; Tier 2: 33; Tier 3: 261; median depth: 1420x). Shifts in allele frequencies among the FACS-sorted subsets provided critical information to a Bayesian mixture modeling algorithm, allowing identification of 5 subclones, as opposed to the 3 subclones anticipated by WGS. The inferred subclonal genotypes were validated and further refined by targeted single-cell Sanger sequencing of multiple Tier 3 loci. Although all 5 subclones were present throughout the differentiation trajectory, some were enriched in certain phenotypic states or ‘reservoirs’. For example, the NRAS-G12D subclone was enriched in a progenitor-like subset, but its daughter subclone, which harbored an additional mutation in RAC2 (implicated in HSC engraftment), was enriched in the more differentiated subsets, suggesting an opposing effect. Notably, the T/B cell population harbored 2 tumor-specific Tier 1 mutations at >15% allele frequency, suggesting its presence in a preleukemic, multilineage-competent HSC. Taken together, subclone-specific mutations appear to skew cells toward either progenitor or mature phenotypes, but the developmental trajectory enforced by parental mutations is resistant to change. Furthermore, characterization of subclones is possible at diagnosis, and may improve the selection of targeted therapies. *Contributed equally: EFS, SCB & ALG Citation Format: Erin F. Simonds, Sean C. Bendall, Amanda L. Gedman, Jacob H. Levine, Kara L. Davis, Harris G. Fienberg, Astraea Jager, El-ad D. Amir, Ina Radtke, Wendy J. Fantl, Dana Pe'er, James R. Downing, Garry P. Nolan. Deep sequencing of immunophenotypically distinct subsets in acute myeloid leukemia reveals reservoirs of genetically distinct subclones along a conserved differentiation trajectory. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr 3178. doi:10.1158/1538-7445.AM2013-3178

Abstract 2565 Acute myeloid leukemia (AML) encompasses a broad spectrum of cellular morphologies. This spectrum exists both within each patient in the form of a malignant developmental hierarchy, and across patients in the form of distinct disease classes with distinguishing mutational or pathological hallmarks. Flow cytometry has long been an essential tool for classifying AML surface markers, and has enabled improved diagnosis, minimal residual disease monitoring, and cell sorting for functional studies. Previous work by our laboratory and others (Irish et al., Cell 2004; Kornblau et al., Clin. Cancer Res. 2010) has shown that clinically relevant heterogeneity also exists at the level of single-cell intracellular signaling capabilities, particularly in response to ex vivo perturbations. Due to the constraints of fluorescent flow cytometry, the limited number of simultaneous parameters measured per cell in these studies was limited (e.g., 3–6). We hypothesized that there may exist additional layers of heterogeneity in both surface marker phenotypes and signaling responses which were invisible to the low-parameter systems used in these studies. To address this challenge, we used a next-generation 31-parameter ‘mass cytometry’ platform to survey the diversity of signaling responses and their relationship to subsets defined by surface marker patterns. We measured the simultaneous co-expression of 16 surface markers and 15 dynamic intracellular signaling epitopes (e.g. phosphorylated kinase substrates) in a cohort of pediatric AML diagnosis bone marrow samples (n=18) and healthy adult bone marrow controls (n=3). Signaling dynamics were measured under 18 stimulation conditions, including a battery of cytokines, chemical stimuli, and small molecule kinase inhibitors. The resulting single-cell data was overlaid and compared using a new, unsupervised flow cytometry analysis and visualization program–SPADE (Spanning-tree Progression Analysis of Density-Normalized Events)–which links rare, transitional, and abundant cell types alike along a branching tree organized by phenotypic progression. Thus, we distilled the diversity of cellular phenotypes, their relative frequencies, and the corresponding signaling dynamics of each cell population onto a single graph structure representing the distinct and overlapping expression patterns in the 16-dimensional “immunophenotypic space” explored by the 21 individuals. Striking signaling and immunophenotypic diversity was observed between the AML samples, particularly in response to pervanadate stimulation. Pervanadate liberates intracellular kinase cascades by broadly inhibiting tyrosine phosphatases, and thus reveals the cell's potential for kinase activity and the interplay of signaling circuits in different cell populations. Interestingly, two transcription factors implicated in AML pathogenesis – CREB and c-Cbl–often exhibited mutually exclusive phosphorylation patterns in different subsets of primitive (CD34+) cells. Specifically, CREB was readily phosphorylated at Ser133 upon pervanadate exposure in the CD34+CD123- cells, while c-Cbl phosphorylation at Tyr700 was restricted to CD34+CD123+ cells. As CD123 (IL-3Ra) has been implicated as a marker of AML stem cells, the finding that these cells are poised for c-Cbl signaling may suggest a potential therapeutic avenue. This work demonstrates the utility of 31-parameter mass cytometry for the simultaneous characterization of multiple cancer stem cell markers and comparison of their expression patterns to the normal tissue hierarchy. These results will inform future studies of AML signaling biology, prognostic biomarkers, and indicators of minimum residual disease. EFS and SCB contributed equally to this work. Disclosures: No relevant conflicts of interest to declare.

Abstract 3160 While oncogene addiction is a well-documented phenomenon, the molecular mechanisms by which oncogene withdrawal triggers cell death are poorly understood. Interrogation of this phenomenon in a manipulatable murine model, coupled to concomitant analysis of human AML samples, could elucidate this phenomenon for therapeutic applications. In order to decipher these molecular mechanisms, we employ a murine model harboring a tetracycline repressible, activated NRAS (NRASG12V) transgene along with an MLL/AF9 transgene to induce AML development. Primary leukemia cells are then transplanted into SCID mice and, upon development of full-blown leukemia, NRASG12V transgene expression is repressed with doxycycline. Previous work has shown that repression of NRASG12V in this model leads to widespread apoptosis of the leukemia cells (Kim et al. Blood 2009). To analyze the kinetics of this response, we found that the tumor burden, as assayed by the white blood cell (WBC) count, declines by 60 hours of doxycycline treatment. NRASG12V message levels are undetectable by 12 hours while protein expression begins to decline after 48 hours. To dissect the signaling network directing the apoptotic response, the phosphorylation status of critical signaling intermediates was analyzed by flow cytometry at time points from 48–96 hours post-doxycycline treatment. This analysis revealed numerous modifications in known NRAS effectors including loss of phosphoErk1/2, phosphoSTAT3, and phosphop38. These alterations correlate with immunophenotypic cell surfaces markers. Whereas leukemia cells expressing mature myeloid markers (Mac1+) do not exhibit alterations in any of the phosphoproteins tested, only Mac1- leukemia cells show meaningful changes in RAS-activated signaling molecules. Whether mouse leukemic stem cells reside in the Mac1+ fraction or in the less differentiated subpopulation remains unclear. These findings suggest that subpopulations of leukemia cells exhibit differential vulnerabilities to oncogene addiction. Furthermore, these studies also reveal that oncogene withdrawal leads to a reduction of the Mac1+Gr1- population and an enrichment of the Mac1-Gr1- and Mac1+Gr1+ populations. Therefore, in addition to effecting apoptosis, oncogene withdrawal leads to alterations in the differentiation status of the leukemia, which could alter the self-renewal capacity of these cells. Using these results, we have designed an extensive antibody panel and are currently using a CyTOF mass spectrometer, a new technology that allows us to perform 30 dimensional measurements to profile immunophenotypic markers and phosphoprotein states with single cell resolution. Simultaneous, high-dimensional single cell profiling of signaling states enables integrative network analysis of these data and as such will allow us to discern pathway dependencies and regulatory relationships that traditional low dimensional flow cytometry cannot (Sachs et al. Science 2005, Sachs et al. IEEE Eng Med Biol Soc 2009). This approach will provide a novel and powerful method to elucidate the critical pathways and leukemic subpopulations that define the response to oncogene withdrawal. Furthermore, these findings will be compared to a concomitant study of 30-dimensional measurements of human AML samples which could facilitate harnessing the oncogene addiction phenomenon in therapeutic applications. Disclosures: No relevant conflicts of interest to declare.

The maintenance of genetic variation [1 3] and sex [4 8] despite its costs [9] has long puzzled biologists. A popular idea the Red Queen Theory [4 8] is that under rapid antagonistic coevolution between hosts and their parasites the formation of new rare host genotypes through sex can be advantageous as it creates host genotypes to which the prevailing parasite is not adapted. For host parasite coevolution to lead to an ongoing advantage for rare genotypes parasites should infect specific host genotypes and hosts should resist specific parasite genotypes. The most prominent genetics capturing such specificity are matching allele models (MAMs) which have the key feature that resistance for two parasite genotypes can reverse by switching one allele at one host locus. Despite the lack of empirical support MAMs have played a central role in the theoretical development of antagonistic coevolution [4 7 8 10] local adaptation [11 12] speciation [13] and sexual selection [14]. Using genetic crosses we show that resistance of the crustacean Daphnia magna against the parasitic bacterium Pasteuria ramosa follows a MAM. Simulation results show that the observed genetics can explain the maintenance of genetic variation and contribute to the maintenance of sex in the facultatively sexual host as predicted by the Red Queen Theory. © 2013 Elsevier Ltd. All rights reserved.