Summary
I'm interested in how cancer cells respond to therapy. My PhD work examines the cell- and molecular-level adaptations that rapidly allow cancer cells to survive targeted therapies, with a focus on EGFR, BRAF, MET, and ALK inhibitors in model systems of lung cancer and melanoma. The goal of this work is to better understand oncogenic signaling networks and the pharmacology of targeted therapies; with this understanding we can begin to nominate promising new combination therapies for preclinical follow-up. I make heavy use of phosphoproteomics by mass spectrometry to characterize the kinetics of signaling networks in patient-derived cell lines, and I use unsupervised machine learning and statistical methods to interpret these large datasets. I enjoy thinking of clever ways to integrate experiment and computation to learn about cancer biology.
Research Objectives
Most of my experiments are powered by phosphoproteomics and drug sensitivity measurements. The major thrusts of my work are aimed at the following questions:
Why do some cancer cells survive therapy while others experience growth arrest or cell death?
What happens to cancer cell signaling networks in the early stages of targeted therapy exposure (the so-called "adaptive response" or "adaptive signaling")?
What can we learn from looking at the kinetics of adaptive signaling?
Can we use knowledge of tumor cell-intrinsic adaptive signaling to propose rational drug combinations that may improve treatment efficacy or reduce toxicity?
How can targeted MS assays help us find new drug targets in human tumors to inform precision oncology?
Motivation
The remarkable clinical success of imatinib (Gleevec) for patients with chronic myeloid leukemia is irrefutable evidence that oncogenic processes can be understood and translated into highly effective, non-toxic therapies. Since its FDA approval in 2001, dozens of other targeted therapies have been approved for cancers with defined oncogenic drug targets; these drugs have allowed many patients to enjoy longer lives and reduced disease burden. However, most patients don't qualify for targeted therapy because they lack an established biomarker for sensitivity, and patients who do qualify often fail to show a durable response – this unfortunate status reflects our limited ability to reliably select drug targets, assess drug sensitivity, and predict therapeutic indices on a patient-specific level.
The recent explosion of genomic and systems-level approaches to query cell biology, paired with computational advances to sift through the massive datasets they produce, have confirmed that cancer is an enormously complex problem. While we still have much to learn from these and future data, we now understand that the cancer phenotype is a function of more than just mutations in DNA – cancer cells emerge, persist, metastasize, and evolve at the mercy of dysregulated molecular events on many levels:
In the White Lab at MIT, my research has focused on the role of signaling in cancer. Proteins signal to one another by physically interacting and covalently modifying each other with post-translational modifications, the most common of which is phosphorylation. Protein phosphorylation plays a ubiquitous role in maintaining cellular homeostasis; many essential processes in human cells depend on phosphorylation as a dynamic and reversible mechanism of protein functional regulation. Kinases are the proteins responsible for catalyzing phosphorylation, and together form one of the largest families of protein-coding genes in the human genome.
Since regulation of signaling is crucial for proper cellular function, it's no surprise that kinases are among the most commonly mutated genes in cancer. (In their seminal 2000 paper, Doug Hanahan and Bob Weinberg named sustained growth signaling one of the "hallmarks of cancer", characteristics shared by all known cancer types.) Many of these mutations occur in tyrosine kinases – a subclass of kinases that tend to transmit growth signals to the cell – and oncogenic mutations often lead to kinase overexpression or structural changes that increase kinase activity. As a result, those growth signals are amplified and the cell divides uncontrollably.
These properties have flagged tyrosine kinases as extremely attractive drug targets, and indeed they have been successfully exploited for a number of different cancers. Early success stories included: imatinib (Gleevec) for inhibition of BCR-ABL (a fusion tyrosine kinase) in chronic myeloid leukemia; trastuzumab (Herceptin) for HER2 (a receptor tyrosine kinase) in breast cancer; and gefitinib (Iressa) for EGFR (another receptor tyrosine kinase) in lung cancer. We now have over 70 FDA-approved kinase inhibitors for cancer treatment.
While oncogenic signaling can originate from changes in DNA sequence and/or gene regulation, it is very difficult to infer signaling activity from genomic and gene expression data alone. This inconvenience is borne out of the underlying complexity of cell signaling networks, and explains why some patients matched to a targeted therapy on the basis of tumor genotype don't exhibit a strong initial or long-term response.
The human "kinome": a phylogenetic depiction of all 538 kinase genes in the human genome. (Adapted from Cell Signaling TechnologyⓇ.)
Fortunately, rather than trying to infer kinase activity from mutation or gene expression data, we can measure kinase signaling networks directly using phosphoproteomics. In this approach, proteins from whole cell lysate or tissue homogenate are digested into short, sequenceable peptides. Those peptides are then subjected to phosphate enrichment, allowing only the phosphorylated peptides through. The phosphopeptide mixture is then ionized and analyzed by mass spectrometry (MS). The process of tandem MS involves breaking apart each peptide into fragments, and using a high-resolution mass analyzer, we can measure the masses of fragment ions to within 0.01 Da (one percent of a hydrogen atom!), letting us unambiguously assign the peptide sequence and discern which of its amino acids was endogenously phosphorylated. Peptide sequences are then mapped back to the human proteome for interpretation. Modern phosphoproteomic workflows can quantify thousands of protein phosphorylation sites with ease, enabling statistical and computational modeling of entire cell signaling networks on an "omics" scale. I've been fortunate to learn this approach from Forest White, my PhD advisor who led the first large-scale phosphoproteome analysis with Donald Hunt and colleagues in 2002.
A mass spectrum of a peptide derived from the human cortactin (CTTN) protein, harboring a phosphorylation site at tyrosine 421. Y421 is known to be phosphorylated by the Src-family kinases (SFKs) and is important for cortactin's role in cell migration. The TMT reporter ions (light blue) are used for multiplexed quantitation of the peptide.
Together with genomics, transcriptomics, and other modalities, phosphoproteomics can help us understand cancer biology and prioritize targets for cancer therapy.
