Synthetic Lethal and Resistance Interactions with BET Bromodomain Inhibitors in Triple-Negative Breast Cancer.

BET bromodomain inhibitors (BBDIs) are candidate therapeutic agents for triple-negative breast cancer (TNBC) and other cancer types, but inherent and acquired resistance to BBDIs limits their potential clinical use. Using CRISPR and small-molecule inhibitor screens combined with comprehensive molecular profiling of BBDI response and resistance, we identified synthetic lethal interactions with BBDIs and genes that, when deleted, confer resistance. We observed synergy with regulators of cell cycle progression, YAP, AXL, and SRC signaling, and chemotherapeutic agents. We also uncovered functional similarities and differences among BRD2, BRD4, and BRD7. Although deletion of BRD2 enhances sensitivity to BBDIs, BRD7 loss leads to gain of TEAD-YAP chromatin binding and luminal features associated with BBDI resistance. Single-cell RNA-seq, ATAC-seq, and cellular barcoding analysis of BBDI responses in sensitive and resistant cell lines highlight significant heterogeneity among samples and demonstrate that BBDI resistance can be pre-existing or acquired.

Breast tumours maintain a reservoir of subclonal diversity during expansion.

Our knowledge of copy number evolution during the expansion of primary breast tumours is limited1,2. Here, to investigate this process, we developed a single-cell, single-molecule DNA-sequencing method and performed copy number analysis of 16,178 single cells from 8 human triple-negative breast cancers and 4 cell lines. The results show that breast tumours and cell lines comprise a large milieu of subclones (7-22) that are organized into a few (3-5) major superclones. Evolutionary analysis suggests that after clonal TP53 mutations, multiple loss-of-heterozygosity events and genome doubling, there was a period of transient genomic instability followed by ongoing copy number evolution during the primary tumour expansion. By subcloning single daughter cells in culture, we show that tumour cells rediversify their genomes and do not retain isogenic properties. These data show that triple-negative breast cancers continue to evolve chromosome aberrations and maintain a reservoir of subclonal diversity during primary tumour growth.

The neuroimmune guidance cue netrin-1 promotes atherosclerosis by inhibiting the emigration of macrophages from plaques.

Atherosclerotic plaque formation is fueled by the persistence of lipid-laden macrophages in the artery wall. The mechanisms by which these cells become trapped, thereby establishing chronic inflammation, remain unknown. Here we found that netrin-1, a neuroimmune guidance cue, was secreted by macrophages in human and mouse atheroma, where it inactivated the migration of macrophages toward chemokines linked to their egress from plaques. Acting via its receptor, UNC5b, netrin-1 inhibited the migration of macrophages directed by the chemokines CCL2 and CCL19, activation of the actin-remodeling GTPase Rac1 and actin polymerization. Targeted deletion of netrin-1 in macrophages resulted in much less atherosclerosis in mice deficient in the receptor for low-density lipoprotein and promoted the emigration of macrophages from plaques. Thus, netrin-1 promoted atherosclerosis by retaining macrophages in the artery wall. Our results establish a causative role for negative regulators of leukocyte migration in chronic inflammation.

ESTIpop: a computational tool to simulate and estimate parameters for continuous-time Markov branching processes.

SUMMARY: ESTIpop is an R package designed to simulate and estimate parameters for continuous-time Markov branching processes with constant or time-dependent rates, a common model for asexually reproducing cell populations. Analytical approaches to parameter estimation quickly become intractable in complex branching processes. In ESTIpop, parameter estimation is based on a likelihood function with respect to a time series of cell counts, approximated by the Central Limit Theorem for multitype branching processes. Additionally, simulation in ESTIpop via approximation can be performed many times faster than exact simulation methods with similar results. AVAILABILITY AND IMPLEMENTATION: ESTIpop is available as an R package on Github (https://github.com/michorlab/estipop). SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

DIFFpop: a stochastic computational approach to simulate differentiation hierarchies with single cell barcoding.

SUMMARY: DIFFpop is an R package designed to simulate cellular differentiation hierarchies using either exponentially-expanding or fixed population sizes. The software includes functionalities to simulate clonal evolution due to the emergence of driver mutations under the infinite-allele assumption as well as options for simulation and analysis of single cell barcoding and labeling data. The software uses the Gillespie Stochastic Simulation Algorithm and a modification of expanding or fixed-size stochastic process models expanded to a large number of cell types and scenarios. AVAILABILITY AND IMPLEMENTATION: DIFFpop is available as an R-package along with vignettes on Github (https://github.com/ferlicjl/diffpop). SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

SIApopr: a computational method to simulate evolutionary branching trees for analysis of tumor clonal evolution.

SUMMARY: SIApopr (Simulating Infinite-Allele populations) is an R package to simulate time-homogeneous and inhomogeneous stochastic branching processes under a very flexible set of assumptions using the speed of C ++. The software simulates clonal evolution with the emergence of driver and passenger mutations under the infinite-allele assumption. The software is an application of the Gillespie Stochastic Simulation Algorithm expanded to a large number of cell types and scenarios, with the intention of allowing users to easily modify existing models or create their own. AVAILABILITY AND IMPLEMENTATION: SIApopr is available as an R library on Github ( https://github.com/olliemcdonald/siapopr ). SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online. CONTACT: michor@jimmy.harvard.edu.

Mathematical Modeling Identifies Optimum Palbociclib-fulvestrant Dose Administration Schedules for the Treatment of Patients with Estrogen Receptor-positive Breast Cancer.

Cyclin-dependent kinases 4/6 (CDK4/6) inhibitors such as palbociclib are approved for the treatment of metastatic estrogen receptor-positive (ER+) breast cancer in combination with endocrine therapies and significantly improve outcomes in patients with this disease. However, given the large number of possible pairwise drug combinations and administration schedules, it remains unclear which clinical strategy would lead to best survival. Here, we developed a computational, cell cycle-explicit model to characterize the pharmacodynamic response to palbociclib-fulvestrant combination therapy. This pharmacodynamic model was parameterized, in a Bayesian statistical inference approach, using in vitro data from cells with wild-type estrogen receptor (WT-ER) and cells expressing the activating missense ER mutation, Y537S, which confers resistance to fulvestrant. We then incorporated pharmacokinetic models derived from clinical data into our computational modeling platform. To systematically compare dose administration schedules, we performed in silico clinical trials based on integrating our pharmacodynamic and pharmacokinetic models as well as considering clinical toxicity constraints. We found that continuous dosing of palbociclib is more effective for lowering overall tumor burden than the standard, pulsed-dose palbociclib treatment. Importantly, our mathematical modeling and statistical analysis platform provides a rational method for comparing treatment strategies in search of optimal combination dosing strategies of other cell-cycle inhibitors in ER+ breast cancer. SIGNIFICANCE: We created a computational modeling platform to predict the effects of fulvestrant/palbocilib treatment on WT-ER and Y537S-mutant breast cancer cells, and found that continuous treatment schedules are more effective than the standard, pulsed-dose palbociclib treatment schedule.

Single-cell heterogeneity of EGFR and CDK4 co-amplification is linked to immune infiltration in glioblastoma.

Glioblastoma (GBM) is the most aggressive brain tumor, with a median survival of ∼15 months. Targeted approaches have not been successful in this tumor type due to the large extent of intratumor heterogeneity. Mosaic amplification of oncogenes suggests that multiple genetically distinct clones are present in each tumor. To uncover the relationships between genetically diverse subpopulations of GBM cells and their native tumor microenvironment, we employ highly multiplexed spatial protein profiling coupled with single-cell spatial mapping of fluorescence in situ hybridization (FISH) for EGFR, CDK4, and PDGFRA. Single-cell FISH analysis of a total of 35,843 single nuclei reveals that tumors in which amplifications of EGFR and CDK4 more frequently co-occur in the same cell exhibit higher infiltration of CD163+ immunosuppressive macrophages. Our results suggest that high-throughput assessment of genomic alterations at the single-cell level could provide a measure for predicting the immune state of GBM.

CREB downregulation in vascular disease: a common response to cardiovascular risk.

OBJECTIVE: To examine the impact of low-density lipoprotein (LDL), an established mediator of atherosclerosis, on the transcription factor cAMP-response element-binding protein (CREB), which is a regulator of vascular smooth muscle cell (VSMC) quiescence. METHODS AND RESULTS: VSMC CREB content is diminished in rodent models of diabetes and pulmonary hypertension. We examined aortic CREB content in rodent models of aging, hypertension, and insulin resistance, and we determined nuclear CREB protein in the medial VSMC of high-fat-fed LDL receptor-null mice. There was significant loss of CREB protein in all models. In vitro, primary culture rat aortic VSMC exposed to LDL and oxidized LDL exhibited a rapid, transient increase in CREB phosphorylation and transient phosphorylation/activation of Akt, ERK, JNK, ans p38 MAPK. Exposure to oxidized LDL, but not to LDL, for 24 to 48 hours decreased CREB protein in a dose-dependent fashion and led to nuclear exclusion of CREB. Pharmacological reactive oxygen species scavengers and inhibition of ERK activation blocked oxidized LDL-mediated CREB downregulation. CONCLUSIONS: These data support a model wherein loss of VSMC CREB protein, which renders these cells more susceptible to activation and apoptosis, is a common pathological response to vascular injury and potentially contributes to plaque progression.

Imaging dynamic mTORC1 pathway activity in vivo reveals marked shifts that support time-specific inhibitor therapy in AML.

Acute myeloid leukemia (AML) is a high remission, high relapse fatal blood cancer. Although mTORC1 is a master regulator of cell proliferation and survival, its inhibitors have not performed well as AML treatments. To uncover the dynamics of mTORC1 activity in vivo, fluorescent probes are developed to track single cell proliferation, apoptosis and mTORC1 activity of AML cells in the bone marrow of live animals and to quantify these activities in the context of microanatomical localization and intra-tumoral heterogeneity. When chemotherapy drugs commonly used clinically are given to mice with AML, apoptosis is rapid, diffuse and not preferentially restricted to anatomic sites. Dynamic measurement of mTORC1 activity indicated a decline in mTORC1 activity with AML progression. However, at the time of maximal chemotherapy response, mTORC1 signaling is high and positively correlated with a leukemia stemness transcriptional profile. Cell barcoding reveals the induction of mTORC1 activity rather than selection of mTORC1 high cells and timed inhibition of mTORC1 improved the killing of AML cells. These data define the real-time dynamics of AML and the mTORC1 pathway in association with AML growth, response to and relapse after chemotherapy. They provide guidance for timed intervention with pathway-specific inhibitors.

Identification of optimal dosing schedules of dacomitinib and osimertinib for a phase I/II trial in advanced EGFR-mutant non-small cell lung cancer.

Despite the clinical success of the third-generation EGFR inhibitor osimertinib as a first-line treatment of EGFR-mutant non-small cell lung cancer (NSCLC), resistance arises due to the acquisition of EGFR second-site mutations and other mechanisms, which necessitates alternative therapies. Dacomitinib, a pan-HER inhibitor, is approved for first-line treatment and results in different acquired EGFR mutations than osimertinib that mediate on-target resistance. A combination of osimertinib and dacomitinib could therefore induce more durable responses by preventing the emergence of resistance. Here we present an integrated computational modeling and experimental approach to identify an optimal dosing schedule for osimertinib and dacomitinib combination therapy. We developed a predictive model that encompasses tumor heterogeneity and inter-subject pharmacokinetic variability to predict tumor evolution under different dosing schedules, parameterized using in vitro dose-response data. This model was validated using cell line data and used to identify an optimal combination dosing schedule. Our schedule was subsequently confirmed tolerable in an ongoing dose-escalation phase I clinical trial (NCT03810807), with some dose modifications, demonstrating that our rational modeling approach can be used to identify appropriate dosing for combination therapy in the clinical setting.

Beyond Deterministic Models in Drug Discovery and Development.

The model-informed drug discovery and development paradigm is now well established among the pharmaceutical industry and regulatory agencies. This success has been mainly due to the ability of pharmacometrics to bring together different modeling strategies, such as population pharmacokinetics/pharmacodynamics (PK/PD) and systems biology/pharmacology. However, there are promising quantitative approaches that are still seldom used by pharmacometricians and that deserve consideration. One such case is the stochastic modeling approach, which can be important when modeling small populations because random events can have a huge impact on these systems. In this review, we aim to raise awareness of stochastic models and how to combine them with existing modeling techniques, with the ultimate goal of making future drug-disease models more versatile and realistic.

Pharmacokinetic Profiles Determine Optimal Combination Treatment Schedules in Computational Models of Drug Resistance.

Identification of optimal schedules for combination drug administration relies on accurately estimating the correct pharmacokinetics, pharmacodynamics, and drug interaction effects. Misspecification of pharmacokinetics can lead to wrongly predicted timing or order of treatments, leading to schedules recommended on the basis of incorrect assumptions about absorption and elimination of a drug and its effect on tumor growth. Here, we developed a computational modeling platform and software package for combination treatment strategies with flexible pharmacokinetic profiles and multidrug interaction curves that are estimated from data. The software can be used to compare prespecified schedules on the basis of the number of resistant cells where drug interactions and pharmacokinetic curves can be estimated from user-provided data or models. We applied our approach to publicly available in vitro data of treatment with different tyrosine kinase inhibitors of BT-20 triple-negative breast cancer cells and of treatment with erlotinib of PC-9 non-small cell lung cancer cells. Our approach is publicly available in the form of an R package called ACESO (https://github.com/Michorlab/aceso) and can be used to investigate optimum dosing for any combination treatment. SIGNIFICANCE: These findings introduce a computational modeling platform and software package for combination treatment strategies with flexible pharmacokinetic profiles and multidrug interaction curves that are estimated from data.

Acquired resistance to combined BET and CDK4/6 inhibition in triple-negative breast cancer.

BET inhibitors are promising therapeutic agents for the treatment of triple-negative breast cancer (TNBC), but the rapid emergence of resistance necessitates investigation of combination therapies and their effects on tumor evolution. Here, we show that palbociclib, a CDK4/6 inhibitor, and paclitaxel, a microtubule inhibitor, synergize with the BET inhibitor JQ1 in TNBC lines. High-complexity DNA barcoding and mathematical modeling indicate a high rate of de novo acquired resistance to these drugs relative to pre-existing resistance. We demonstrate that the combination of JQ1 and palbociclib induces cell division errors, which can increase the chance of developing aneuploidy. Characterizing acquired resistance to combination treatment at a single cell level shows heterogeneous mechanisms including activation of G1-S and senescence pathways. Our results establish a rationale for further investigation of combined BET and CDK4/6 inhibition in TNBC and suggest novel mechanisms of action for these drugs and new vulnerabilities in cells after emergence of resistance.

Serum amyloid A3 does not contribute to circulating SAA levels.

Adipose tissue secretes proteins like serum amyloid A (SAA), which plays important roles in local and systemic inflammation. Circulating SAA levels increase in obese humans, but the roles of adipose-derived SAA and hyperlipidemia in this process are unclear. We took advantage of the difference in the inducible isoforms of SAA secreted by adipose tissue (SAA3) and liver (SAA1 and 2) of mice to evaluate whether adipose tissue contributes to the circulating pool of SAA in obesity and hyperlipidemia. Genetically obese (ob/ob) mice, but not hyperlipidemic mice deficient in apolipoprotein E (Apoe(-/-)), had significantly higher circulating levels of SAA than their littermate controls. SAA1/2 mRNA expression in the liver and SAA3 mRNA expression in intra-abdominal fat were significantly higher in obese than thin mice, but they were not affected by hyperlipidemia in Apoe(-/-) mice. However, only SAA1/2 and the constitutive form of SAA (SAA4) could be detected in the circulation by mass spectrometric analysis of HDL, the major carrier of circulating SAA. In contrast, SAA3 could be detected in medium from cultured adipocytes. Our findings indicate that the expression of SAA3 in adipose tissue is upregulated by obesity, but it does not contribute to the circulating pool of SAA in mice.

KDM5 Histone Demethylase Activity Links Cellular Transcriptomic Heterogeneity to Therapeutic Resistance.

Members of the KDM5 histone H3 lysine 4 demethylase family are associated with therapeutic resistance, including endocrine resistance in breast cancer, but the underlying mechanism is poorly defined. Here we show that genetic deletion of KDM5A/B or inhibition of KDM5 activity increases sensitivity to anti-estrogens by modulating estrogen receptor (ER) signaling and by decreasing cellular transcriptomic heterogeneity. Higher KDM5B expression levels are associated with higher transcriptomic heterogeneity and poor prognosis in ER+ breast tumors. Single-cell RNA sequencing, cellular barcoding, and mathematical modeling demonstrate that endocrine resistance is due to selection for pre-existing genetically distinct cells, while KDM5 inhibitor resistance is acquired. Our findings highlight the importance of cellular phenotypic heterogeneity in therapeutic resistance and identify KDM5A/B as key regulators of this process.

Diabetes and arterial extracellular matrix changes in a porcine model of atherosclerosis.

Patients with diabetes are at substantially increased risk for atherosclerosis and clinical cardiovascular events. Because arterial extracellular matrix contains several molecules, including biglycan, versican, hyaluronan, and elastin, that may affect plaque lipid retention and stability, we determined whether diabetes affects plaque content of these molecules in a porcine model of hyperlipidemia and diabetes. Coronary artery sections were studied from non-diabetic normolipidemic (n=11, N-NL), diabetic normolipidemic (n=10, DM-NL), non-diabetic hyperlipidemic (n=16, N-HL), and diabetic hyperlipidemic (n=15, DM-HL) animals. Hyaluronan, biglycan, versican, and apolipoprotein B (apoB) were detected with monospecific peptides or antisera, and elastin with Movat's pentachrome stain, and contents of each were quantified by computer-assisted morphometry. In the hyperlipidemic groups, diabetes was associated with a 4-fold increase in intimal area, with strong correlations between intimal area and immunostained areas for hyaluronan (R(2) = 0.83, p<0.0001), biglycan (R(2) = 0.72, p<0.0001), and apoB (R(2) = 0.23, p=0.0069). In contrast, median (interquartile range) intimal elastin content was significantly lower with diabetes [N-HL: 5.2% (2.4-8.2%) vs DM-HL: 1.5% (0.5-4.2%), p=0.01], and there was a strong negative correlation between intimal total and elastin areas (Spearman r = -0.62, p=0.001). In this porcine model, diabetes was associated with multiple extracellular matrix changes that have been associated with increased lesion instability, greater atherogenic lipoprotein retention, and accelerated atherogenesis.

Advanced glycation end product precursors impair ABCA1-dependent cholesterol removal from cells.

Abnormal HDL metabolism may contribute to the increased atherosclerosis associated with diabetes. The ATP-binding cassette transporter A1 (ABCA1) is an atheroprotective cell protein that mediates cholesterol transport from cells to apolipoprotein (apo) A-I, the major protein in HDL. Because formation of advanced glycation end products (AGEs) is associated with diabetic vascular complications, we examined the effects of carbonyls implicated in AGE formation on the ABCA1 pathway in cultured fibroblasts and macrophages. Treating cells with glycolaldehyde (GA) and glyoxal (GO) strongly inhibited ABCA1-dependent transport of cholesterol from cells to apoA-I, while methylglyoxal had little effect. This occurred under conditions where other lipoprotein receptors or lipid metabolic pathways were little affected, indicating that ABCA1 was uniquely sensitive to these carbonyls. GA and GO destabilized ABCA1 and nearly abolished its binding of apoA-I, indicating that these carbonyls directly modified ABCA1. Immunohistology of coronary arteries from hyperlipidemic swine revealed that inducing diabetes with streptozotocin increased atherosclerotic lesion area and dramatically reduced the fraction of macrophages that expressed detectable ABCA1. These results raise the possibility that reactive carbonyl-mediated damage to ABCA1 promotes accumulation of cholesterol in arterial macrophages and thus contribute to the increased cardiovascular disease associated with diabetes, insulin resistance, and other inflammatory conditions.

Serum amyloid A and lipoprotein retention in murine models of atherosclerosis.

OBJECTIVE: Elevated serum amyloid A (SAA) levels are associated with increased cardiovascular risk in humans. Because SAA associates primarily with lipoproteins in plasma and has proteoglycan binding domains, we postulated that SAA might mediate lipoprotein retention on atherosclerotic extracellular matrix. METHODS AND RESULTS: Immunohistochemistry was performed for SAA, apolipoprotein A-I (apoA-I), apolipoprotein B (apoB), and perlecan on proximal aortic lesions from chow-fed low-density lipoprotein receptor (LDLR)-/- and apoE-/- mice euthanized at 10, 50, and 70 weeks. SAA was detected on atherosclerotic lesion extracellular matrix at all time points in both strains. SAA area correlated highly with lesion areas (apoE-/-, r=0.76; LDLR-/-, r=0.86), apoA-I areas (apoE-/-, r=0.88; LDLR-/-, r=0.80), apoB areas (apoE-/-, r=0.74; LDLR-/-, r=0.89), and perlecan areas (apoE-/-, r=0.83; LDLR-/-, r=0.79) (all P<0.0001). In vitro, SAA enrichment increased high-density lipoprotein (HDL) binding to heparan sulfate proteoglycans, and immunoprecipitation experiments using plasma from apoE-/- and LDLR-/- mice demonstrated that SAA was present on both apoA-I-containing and apoB-containing lipoproteins. CONCLUSIONS: In chow-fed apoE-/- and LDLR-/- mice, SAA is deposited in murine atherosclerosis at all stages of lesion development, and SAA immunoreactive area correlates highly with lesion area, apoA-I area, apoB area, and perlecan area. These findings are consistent with a possible role for SAA-mediated lipoprotein retention in atherosclerosis.