Dissecting cell-to-cell regulatory heterogeneity by stochastic profiling

Uncovering single-cell regulatory states in microdissected cells by stochastic profiling
Uncovering single-cell regulatory states in microdissected cells by stochastic profiling
(Nat Methods 7:311-7 [2010])

There is tremendous enthusiasm for using the power of transcriptomics to unravel the state of single cells in tissues and tumors. If a single-cell profiling method is to be meaningful, however, it must be i) technically reproducible, ii) sensitive to low-abundance molecules, and iii) compatible with cells isolated in situ. There are many exciting techniques under development to profile the transcriptomes of single cells; unfortunately, none of them meet any of these three criteria (1).

To address the challenge, we have developed an alternative approach that provides global information about single-cell regulatory states without the need to profile single cells. The technique, called stochastic profiling (2, 3), analyzes the transcriptomic fluctuations of multiple 10-cell pools collected randomly from a tissue or tumor context. The purpose of pooling is to increase the input material, which dramatically improves the reproducibility and sensitivity of the resulting profiles, enabling pools to be collected by laser capture microdissection. The subsequent statistical analysis deconvolves the 10-cell distributions to identify heterogeneous regulatory states and infer their underlying single-cell distribution in the sampled population (4). We have trained labs across the country in this method and anticipate many applications in the coming years. Ongoing work has adapted stochastic profiling to RNA sequencing, fluorescence-guided applications in genetically engineered mouse models, and primary patient material.

1. Janes KA. (2016) Single-cell states versus single-cell atlases–two classes of heterogeneity that differ in meaning and method. Curr Opin Biotechnol, 39, 120-5.

2. Janes KA, Wang CC, Holmberg KH, Cabral K, Brugge JS. (2010) Identifying single-cell molecular programs by stochastic profiling. Nat Methods, 7, 311-7.

3. Wang L, Janes KA. (2013) Stochastic profiling of transcriptional regulatory heterogeneities in tissues, tumors and cultured cells. Nat Protoc, 8, 282-301.

4. Bajikar SS, Fuchs C, Roller A, Theis FJ*, Janes KA*. (2014) Parameterizing cell-to-cell regulatory heterogeneities via stochastic transcriptional profiles. Proc Natl Acad Sci, 111, E626-35.

 

 

Oxidative stress caused by compound disruption of RUNX1 and FOXOs
Oxidative stress caused by compound disruption of RUNX1 and FOXOs (Proc Natl Acad Sci, 108:E803-12 [2011])

Functional regulatory heterogeneities in breast cancer

In cancer, single-cell differences could be critically important for tumor progression, metastases, and therapeutic response. Or, they could simply be consequences of unselected genome instability and transcriptional noise. Distinguishing these possibilities requires formal hypothesis-testing experiments with contemporary approaches from cancer biology, which we adapt in the spirit of biomedical engineering.

Informed by stochastic profiling, we were among the first to suggest a role for the myeloid transcription factor RUNX1 in breast cancer (1, 2). This work was later verified by TCGA and mouse models, and it ultimately led to a collaborative effort with the Bushweller lab (UVA Department of Chemistry) toward testing a novel RUNX-family inhibitor in the setting of triple-negative breast cancer (3). The most-recent work focuses on another stress-related transcription factor that may collaborate with the tumor suppressor p53.

A second research subtheme emerged from a dynamic regulatory circuit identified by stochastic profiling and comprised of the JUND transcription factor together with the TGFβ-family co-receptor TGFBR3 (4). This work led to the TGFβ-family ligand GDF11, where the in vitro, in vivo, and clinical evidence points to it acting as a novel tumor suppressor in triple-negative breast cancer (5). We are interested in pursuing the unique mechanism of GDF11 deactivation in triple-negative disease.

1. Wang L*, Brugge JS, Janes KA*. (2011) Intersection of FOXO and RUNX1 gene-expression programs in single breast epithelial cells during morphogenesis and tumor progression. Proc Natl Acad Sci, 108, E803-12.

2. Janes KA. (2011) RUNX1 and its understudied role in breast cancer. Cell Cycle, 10, 3461-5.

3. Illendula A, Gilmour J, Grembecka J, Tirumala VSS, Boulton A, Kuntimaddi A, Schmidt C, Wang L, Pullikan JA, Zong H, Parlak M, Kuscu C, Pickin A, Zhou Y, Gao Y, Mishra L, Adli M, Castilla LH, Rajewski RA, Janes KA, Guzman ML, Bonifer C, Bushweller JH. (2016) Small molecule inhibitors of CBFβ-RUNX binding for RUNX transcription factor driven cancers. EBioMedicine, 8, 117-31.

4. Wang CC, Bajikar SS, Jamal L, Atkins KA, Janes KA. (2014) A time- and matrix-dependent TGFBR3–JUND–KRT5 regulatory circuit in single breast epithelial cells and basal-like premalignancies. Nat Cell Biol, 16, 345-56.

5. Bajikar SS, Wang CC, Borten MA, Pereira EJ, Atkins KA, Janes KA. (2017) GDF11 tumor suppression is sequestered in triple-negative breast cancer. In revision.

 

Calibration of subcellular phosphatase activities

Calibration of subcellular phosphatase activities (Mol Cell Proteomics 16:S244-62 [2017])

Methods engineering of biomolecular and cellular assays

Problem solving subject to constraints is the hallmark of engineering design (1). We adopt a design approach to invent new methods for interrogating cells and biomolecules. Our goal is to develop bioassays that are sensitive, quantitative, and as high-throughput and multiplex as possible. Most importantly, they should be reliable, generalizable, and shareable.

In the realm of signal transduction, we have developed and expanded high-throughput methods for profiling protein phosphatase activities (2, 3). These methods are now being applied to chronic viral infections along with a technique under development for a multiplex kinase activity assay. We also seek to extend or standardize established methods. For example, we have scaled quantitative real-time PCR to the entire human signaling receptome (4) and set standards for quantitative immunoblotting (5). Beyond empirical methods, we have completed a software platform for digital segmentation of brightfield images of 3D organoid cultures (to be submitted).

1. Janes KA, Chandran PL, Ford RM, Lazzara MJ, Papin JA, Peirce SM, Saucerman JJ, Lauffenburger DA. (2017) An engineering design approach to systems biology. Integr Biol, doi:10.1039/c7ib00014f.

2. Bose AK, Janes KA. (2013) A high-throughput assay for phosphoprotein-specific phosphatase activity in cellular extracts. Mol Cell Proteomics, 12, 797-806.

3. Shah M, Smolko CM, Kinicki S, Chapman ZD, Brautigan DL, Janes KA. (2017) Profiling Subcellular Protein Phosphatase Responses to Coxsackievirus B3 Infection of Cardiomyocytes. Mol Cell Proteomics, 16, S244-62.

4. Kang BH*, Jensen KJ*, Hatch JA, Janes KA. (2013) Simultaneous profiling of 194 distinct receptor transcripts in human cells. Sci Signal, 6, rs13.

5. Janes KA. (2015) An analysis of critical factors for quantitative immunoblotting. Sci Signal, 8, rs2.

 

 

CVB3-infected cardiomyocytes

CVB3-infected cardiomyocytes
(Cell Host Microbe, 13:67-76 [2013].)

Systems virology of coxsackievirus B3 pathogenesis

Coxsackievirus B3 (CVB3) is a cardiotropic positive-strand RNA virus that is a leading cause of viral myocarditis and heart failure in infants and young children. Nearly all work on CVB3 virology has focused on the docking interactions and gene products of the virus. Sparked by an early collaboration with the McManus laboratory (University of British Columbia), my group has taken a fundamentally different view of CVB3 pathogenesis as a systems-level perturbation of the host.

By exploiting the polypharmacology of host-cell signaling inhibitors, we have uncovered novel posttranslational and autocrine mechanisms during acute CVB3 infection (1). Moreover, we have shown that dynamic changes in the host-cell signaling network are sufficient to predict the outcome of acute infection to within 97% accuracy (2). Combinatorial perturbation of the most-informative host-cell pathways reduced CVB3 transmission by 1000-fold and revealed new mechanisms of cytotoxicity during infection. Ongoing work is focused on systems-level analysis of signaling dysfunction during chronic CVB3 infection in an in vitro model. Chronic CVB3 infection ultimately leads to dilated cardiomyopathy and the need for heart transplantation in infected patients. We seek to determine whether systems-biology approaches can have the same impact on virology as they have had on cancer biology (3).

1. Garmaroudi FS, Marchant D, Si X, Khalili A, Bashashati A, Wong BW, Tabet A, Ng RT, Murphy K, Luo H, Janes KA*, McManus BM*. (2010) Pairwise network mechanisms in the host signaling response to coxsackievirus B3 infection. Proc Natl Acad Sci, 107, 17053-8.

2. Jensen KJ, Garmaroudi FS, Zhang J, Lin J, Boroomand S, Zhang M, Luo Z, Yang D, Luo H, McManus BM*, and Janes KA*. (2013) An ERK–p38 subnetwork coordinates host-cell apoptosis and necrosis during coxsackievirus B3 infection. Cell Host Microbe, 13, 67-76.

3. Shah M, Smolko CM, Kinicki S, Chapman ZD, Brautigan DL, Janes KA. (2017) Profiling subcellular phosphatase responses to coxsackievirus B3 infection of cardiomyocytes. Mol Cell Proteomics, 16, S244-62.

 

 

Predictive modeling of biomolecular networks

Exhaustive modeling of pharmacologic and genetic perturbations of a three-tiered enzyme cascade
Exhaustive modeling of pharmacologic and genetic perturbations of a
three-tiered enzyme cascade (Cell Syst 2:112-21 [2016])

Thousands of biomolecular measurements mean little without a way to interpret them. As engineers, we believe that predictive modeling of biomolecular data is an important tool for interpretation and understanding. Using partial least squares regression (PLSR), we built a predictive model of host-cell responses to virus infection, which uncovered new connections between MAP kinase signaling pathways (1) We have also structured such models as mathematical tensors that retain how the data were collected (2). Application of tensor PLSR to a joint signaling–transcriptomic dataset revealed a function for a novel phosphorylation site on an understudied transcription factor (3). These powerful approaches now inform ongoing and future data-collection efforts in the lab.

For other applications, we have also built predictive models based on physicochemical mechanisms or logical rules. By modeling simple signaling circuits as ordinary differential equations, we defined network motifs that were especially susceptible to pharmacologic or genetic perturbations (4). Mechanistic as well as agent-based models were applied to a dynamic and context-specific regulatory circuit in basal-like breast premalignancies (5). Currently, we are reconstructing a physicochemical model of coxsackievirus infection to investigate the competition between viral pathogenesis and the antiviral interferon response.

1. Jensen KJ, Garmaroudi FS, Zhang J, Lin J, Boroomand S, Zhang M, Luo Z, Yang D, Luo H, McManus BM*, and Janes KA*. (2013) An ERK–p38 subnetwork coordinates host-cell apoptosis and necrosis during coxsackievirus B3 infection. Cell Host Microbe, 13, 67-76.

2. Shah M, Chitforoushzadeh Z, Janes KA. (2016) Statistical data analysis and moceling. In: Uncertainty in Biology – A Computational Modeling Approach (eds. L Geris, D Gomez-Cabrero), pp 478. Springer International.

3. Chitforoushzadeh Z, Ye Z, Sheng Z, LaRue S, Fry RC, Lauffenburger DA, Janes KA. (2016) TNF-insulin crosstalk at the transcription factor GATA6 is revealed by a model that links signaling and transcriptomic data tensors. Sci Signal, 9, ra59.

4. Jensen KJ*, Moyer CB*, Janes KA. (2016) Network architecture predisposes an enzyme to pharmacologic or genetic targeting. Cell Syst, 2, 112-21.

5. Wang CC, Bajikar SS, Jamal L, Atkins KA, Janes KA. (2014) A time- and matrix-dependent TGFBR3–JUND–KRT5 regulatory circuit in single breast epithelial cells and basal-like premalignancies. Nat Cell Biol, 16, 345-56.