Mehdi Bouhaddou

Engineered Mammalian RNAi Can Elicit Antiviral Protection that Negates the Requirement for the Interferon Response

Although the intrinsic antiviral cell defenses of many kingdoms utilize pathogen-specific small RNAs, the antiviral response of chordates is primarily protein based and not uniquely tailored to the incoming microbe. In an effort to explain this evolutionary bifurcation, we determined whether antiviral RNAi was sufficient to replace the protein-based type I interferon (IFN-I) system of mammals. To this end, we recreated an RNAi-like response in mammals and determined its effectiveness to combat influenza A virus in vivo in the presence and absence of the canonical IFN-I system. Mammalian antiviral RNAi, elicited by either host- or virus-derived small RNAs, effectively attenuated virus and prevented disease independently of the innate immune response. These data find that chordates could have utilized RNAi as their primary antiviral cell defense and suggest that the IFN-I system emerged as a result of natural selection imposed by ancient pathogens.

Drug response consistency in CCLE and CGP

The Cancer Cell Line Encyclopedia1 (CCLE) and Cancer Genome Project2 (CGP) are two independent large-scale efforts to characterize genomes, mRNA expression, and anti-cancer drug dose–responses across cell lines, providing a public resource relating cellular biochemical context to drug sensitivity. A recent study3 analysed correlations between reported dose–response metrics and found inconsistency between CCLE and CGP, thus questioning the validity of not only these, but also other current and future costly large-scale studies. Here, we examine two metrics of drug responsiveness (slope and area under the curve) that we derive from the original CCLE and CGP data, and find reasonable and statistically significant consistency. Our results revive confidence that the CCLE and CGP drug dose–response data are of sufficient quality for meaningful analyses. There is a Reply to this Comment by Safikhani, Z. et al. Nature 540, http://dx.doi.org/10.1038/nature20581 (2016). CCLE and CGP share 2,520 dose–responses across 285 cell lines and 15 drugs, but cells were treated with different dose ranges. To compare

An Integrated Mechanistic Model of Pan-Cancer Driver Pathways Predicts Stochastic Proliferation and Death

Most cancer cells harbor multiple drivers whose epistasis and interactions with expression context clouds drug sensitivity prediction. We constructed a mechanistic computational model that is context-tailored by omics data to capture regulation of stochastic proliferation and death by pan-cancer driver pathways. Simulations and experiments explore how the coordinated dynamics of RAF/MEK/ERK and PI-3K/AKT kinase activities in response to synergistic mitogen or drug combinations control cell fate in a specific cellular context. In this context, synergistic ERK and AKT inhibitor-induced death is likely mediated by BIM rather than BAD. AKT dynamics explain S-phase entry synergy between EGF and insulin, but stochastic ERK dynamics seem to drive cell-to-cell proliferation variability, which in simulations are predictable from pre-stimulus fluctuations in C-Raf/B-Raf levels. Simulations predict MEK alteration negligibly influences transformation, consistent with clinical data. Our model mechanistically interprets context-specific landscapes between driver pathways and cell fates, moving towards more rational cancer combination therapy.

A Comparison of mRNA Sequencing with Random Primed and 3′-Directed Libraries

Creating a cDNA library for deep mRNA sequencing (mRNAseq) is generally done by random priming, creating multiple sequencing fragments along each transcript. A 3′-end-focused library approach cannot detect differential splicing, but has potentially higher throughput at a lower cost, along with the ability to improve quantification by using transcript molecule counting with unique molecular identifiers (UMI) that correct PCR bias. Here, we compare an implementation of such a 3′-digital gene expression (3′-DGE) approach with “conventional” random primed mRNAseq. Given our particular datasets on cultured human cardiomyocyte cell lines, we find that, while conventional mRNAseq detects ~15% more genes and needs ~500,000 fewer reads per sample for equivalent statistical power, the resulting differentially expressed genes, biological conclusions, and gene signatures are highly concordant between two techniques. We also find good quantitative agreement at the level of individual genes between two techniques for both read counts and fold changes between given conditions. We conclude that, for high-throughput applications, the potential cost savings associated with 3′-DGE approach are likely a reasonable tradeoff for modest reduction in sensitivity and inability to observe alternative splicing, and should enable many larger scale studies focusing on not only differential expression analysis, but also quantitative transcriptome profiling.

Integrating Transcriptomic Data with Mechanistic Systems Pharmacology Models for Virtual Drug Combination Trials

Monotherapy clinical trials with mutation-targeted kinase inhibitors, despite some success in other cancers, have yet to impact glioblastoma (GBM). Besides insufficient blood-brain barrier penetration, combinations are key to overcoming obstacles such as intratumoral heterogeneity, adaptive resistance, and the epistatic nature of tumor genomics that cause mutation-targeted therapies to fail. With now hundreds of potential drugs, exploring the combination space clinically and preclinically is daunting. We are building a simulation-based approach that integrates patient-specific data with a mechanistic computational model of pan-cancer driver pathways (receptor tyrosine kinases, RAS/RAF/ERK, PI3K/AKT/mTOR, cell cycle, apoptosis, and DNA damage) to prioritize drug combinations by their simulated effects on tumor cell proliferation and death. Here we illustrate a first step, tailoring the model to 14 GBM patients from The Cancer Genome Atlas defined by an mRNA-seq transcriptome, and then simulating responses to three promiscuous FDA-approved kinase inhibitors (bosutinib, ibrutinib, and cabozantinib) with evidence for blood-brain barrier penetration. The model captures binding of the drug to primary targets and off-targets based on published affinity data and simulates responses of 100 heterogeneous tumor cells within a patient. Single drugs are marginally effective or even counterproductive. Common copy number alterations (PTEN loss, EGFR amplification, and NF1 loss) have a negligible correlation with single-drug or combination efficacy, reinforcing the importance of postgenetic approaches that account for kinase inhibitor promiscuity to match drugs to patients. Drug combinations tend to be either cytostatic or cytotoxic, but seldom both, highlighting the need for considering targeted and nontargeted therapy. Although we focus on GBM, the approach is generally applicable.

Validating Antibodies for Quantitative Western Blot Measurements with Microwestern Array

Fluorescence-based western blots are quantitative in principal, but require determining linear range for each antibody. Here, we use microwestern array to rapidly evaluate suitable conditions for quantitative western blotting, with up to 192 antibody/dilution/replicate combinations on a single standard size gel with a seven-point, two-fold lysate dilution series (~100-fold range). Pilot experiments demonstrate a high proportion of investigated antibodies (17/24) are suitable for quantitative use; however this sample of antibodies is not yet comprehensive across companies, molecular weights, and other important antibody properties, so the ubiquity of this property cannot yet be determined. In some cases microwestern struggled with higher molecular weight membrane proteins, so the technique may not be uniformly applicable to all validation tasks. Linear range for all validated antibodies is at least 8-fold, and up to two orders of magnitude. Phospho-specific and total antibodies do not have discernable trend differences in linear range or limit of detection. Total antibodies generally required higher working concentrations, but more comprehensive antibody panels are required to better establish whether this trend is general or not. Importantly, we demonstrate that results from microwestern analyses scale to normal “macro” western for a subset of antibodies.

Analysis of copy number loss of the ErbB4 receptor tyrosine kinase in glioblastoma

Current treatments for glioblastoma multiforme (GBM)—an aggressive form of brain cancer—are minimally effective and yield a median survival of 14.6 months and a two-year survival rate of 30%. Given the severity of GBM and the limitations of its treatment, there is a need for the discovery of novel drug targets for GBM and more personalized treatment approaches based on the characteristics of an individual’s tumor. Most receptor tyrosine kinases—such as EGFR—act as oncogenes, but publicly available data from the Cancer Cell Line Encyclopedia (CCLE) indicates copy number loss in the ERBB4 RTK gene across dozens of GBM cell lines, suggesting a potential tumor suppressor role. This loss is mutually exclusive with loss of its cognate ligand NRG1 in CCLE as well, more strongly suggesting a functional role. The availability of higher resolution copy number data from clinical GBM patients in The Cancer Genome Atlas (TCGA) revealed that a region in Intron 1 of the ERBB4 gene was deleted in 69.1% of tumor samples harboring ERBB4 copy number loss; however, it was also found to be deleted in the matched normal tissue samples from these GBM patients (n = 81). Using the DECIPHER Genome Browser, we also discovered that this mutation occurs at approximately the same frequency in the general population as it does in the disease population. We conclude from these results that this loss in Intron 1 of the ERBB4 gene is neither a de novo driver mutation nor a predisposing factor to GBM, despite the indications from CCLE. A biological role of this significantly occurring genetic alteration is still unknown. While this is a negative result, the broader conclusion is that while copy number data from large cell line-based data repositories may yield compelling hypotheses, careful follow up with higher resolution copy number assays, patient data, and general population analyses are essential to codify initial hypotheses prior to investing experimental resources.

A mechanistic pan-cancer pathway model informed by multi-omics data interprets stochastic cell fate responses to drugs and mitogens

Most cancer cells harbor multiple drivers whose epistasis and interactions with expression context clouds drug and drug combination sensitivity prediction. We constructed a mechanistic computational model that is context-tailored by omics data to capture regulation of stochastic proliferation and death by pan-cancer driver pathways. Simulations and experiments explore how the coordinated dynamics of RAF/MEK/ERK and PI-3K/AKT kinase activities in response to synergistic mitogen or drug combinations control cell fate in a specific cellular context. In this MCF10A cell context, simulations suggest that synergistic ERK and AKT inhibitor-induced death is likely mediated by BIM rather than BAD, which is supported by prior experimental studies. AKT dynamics explain S-phase entry synergy between EGF and insulin, but simulations suggest that stochastic ERK, and not AKT, dynamics seem to drive cell-to-cell proliferation variability, which in simulations is predictable from pre-stimulus fluctuations in C-Raf/B-Raf levels. Simulations suggest MEK alteration negligibly influences transformation, consistent with clinical data. Tailoring the model to an alternate cell expression and mutation context, a glioma cell line, allows prediction of increased sensitivity of cell death to AKT inhibition. Our model mechanistically interprets context-specific landscapes between driver pathways and cell fates, providing a framework for designing more rational cancer combination therapy.

Fluorescence Multiplexing with Spectral Imaging and Combinatorics

Ultraviolet-to-infrared fluorescence is a versatile and accessible assay modality but is notoriously hard to multiplex due to overlap of wide emission spectra. We present an approach for fluorescence called multiplexing using spectral imaging and combinatorics (MuSIC). MuSIC consists of creating new independent probes from covalently linked combinations of individual fluorophores, leveraging the wide palette of currently available probes with the mathematical power of combinatorics. Probe levels in a mixture can be inferred from spectral emission scanning data. Theory and simulations suggest MuSIC can increase fluorescence multiplexing ∼4-5 fold using currently available dyes and measurement tools. Experimental proof-of-principle demonstrates robust demultiplexing of nine solution-based probes using ∼25% of the available excitation wavelength window (380-480 nm), consistent with theory. The increasing prevalence of white lasers, angle filter-based wavelength scanning, and large, sensitive multianode photomultiplier tubes make acquisition of such MuSIC-compatible data sets increasingly attainable.

Abstract 1568: Predicting stochastic proliferation and death in response to drugs with mechanistic models tailored to genomic, transcriptomic, and proteomic data

Over the past decade we have seen a shift in cancer therapy from broadly cytotoxic drugs to molecular therapies targeting “driver” mutations. Although targeted therapy has seen great success for some cancers (e.g. imatinib for leukemia), it has struggled with poor efficacy in treating other cancers that can sometimes possess multiple “driver” mutations. This highlights the complex, and at times non-intuitive, interplay between multiple players in a signaling cascade, which can be highly dependent on the biological context – that is, gene expression levels and mutational architecture – of a tumor or cell line. A quantitative, mechanistic, biologically-tailored understanding of how these signaling dynamics drive proliferation and death could improve precision pharmacology approaches to treat cancer. Here, we constructed the first highly detailed, large-scale ordinary differential equation (ODE) mechanistic mathematical model depicting the most commonly mutated cancer signaling pathways across human cancers, as indicated by a pan-cancer analysis by The Cancer Genome Atlas (TCGA). The model includes the RTK/Ras/MAPK, PI3K/AKT/mTOR, CDK/RB cell cycle, p53/MDM2 DNA damage response, and BCL/Caspases apoptosis pathways. The adjustable parameters of the model can be informed by measurements from patients or cell lines, including copy number alterations, mutations, and gene expression levels. This single-cell model links stochastic gene expression processes to quantitative signaling dynamics, and once tailored to a biological context can be used to simulate the effect of various anti-cancer therapies on cell fate behavior such as proliferation and death for a population of cells. The first instance of the model integrated genomic, transcriptomic, and proteomic data from the MCF10A cell line, a non-transformed cell line with predictable phenotypic behaviors. We trained the model using western blot and flow cytometry experiments to refine various biochemical parameters and phenotypic outcomes. Many fundamental questions in signal transduction arose during this process, such as how EGF and insulin synergize to drive S-phase entry or how a specific biological context confers sensitivity or resistance to inhibitors of the ERK and AKT pathways. Simultaneously, we are tailoring the model to patient-derived genetic information from primary glioblastoma tumors and screening brain-penetrable compounds in a patient-specific manner. In conclusion, a quantitative, mechanistic, biologically-tailored mathematical model depicting the major cancer pathways allows us to probe the mechanisms that underlie how signaling dynamics drive proliferation and death in response to various perturbations, and gain insight into their dependence on the biological context of cell lines and patient tumors. Note: This abstract was not presented at the meeting. Citation Format: Mehdi Bouhaddou, Anne Marie Barrette, Rick J. Koch, Marc R. Birtwistle. Predicting stochastic proliferation and death in response to drugs with mechanistic models tailored to genomic, transcriptomic, and proteomic data [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 1568. doi:10.1158/1538-7445.AM2017-1568