Cancer cells survive by co-opting intracellular growth pathways regulated through kinase signaling. Many kinases in these pathways are validated drug targets and kinase inhibitors are first-line treatment for several advanced cancers. The first generation of kinase inhibitors were intended to target single proteins. However, the highly conserved active sites of protein kinases prohibited the design of perfectly selective inhibitors. Unexpectedly, this specificity problem led to the development of clinically useful, multi-targeted kinase inhibitors. Clinically, multi-targeted inhibitors were found to have high efficacy and low toxicity. At a molecular level, this translated into simultaneous inhibition of critical nodes in a signaling network regulated by multiple inputs and levels of feedback.

Two kinase families, phosphoinositide-3-kinases (PI3Ks) and protein tyrosine kinases (PTKs), are activated by mutations and amplifications in a disproportionate number of human cancers. This suggests multi-targeted therapy inhibiting PI3Ks and PTKs may be extremely effective at blocking the signaling networks that are essential for cancer cell survival. The design of dual PI3K/PTK inhibitors is chemically challenging because PI3Ks and PTKs are members of significantly divergent kinase families. However, druggable similarities exist between these two families due to structural features essential for phosphoryl transfer with ATP as the phosphate donor.

We report the first drug-like small molecules that potently and selectively inhibit PTKs and PI3Ks. These agents were developed by exploiting structural similarities in kinase active sites. Through iterative medicinal chemistry, kinome profiling and x-ray crystallography, the chemical and structural features that control the in vitro selectivity and potency of these compounds were defined. A subset of these compounds potently, and sometimes selectively, inhibit the mammalian target of rapamycin (mTOR), a Ser/Thr kinase in the PI3K family, placing mTOR in kinase space at the intersection of PI3Ks and PTKs. Subsequent cell-based experiments demonstrated these compounds inhibit their molecular targets in cells, block cancer cell proliferation and angiogenesis, and maintain activity against cells harboring mutations rendering them resistant to singly-targeted agents.

In response to external cues, kinases function in signaling networks to orchestrate a transcriptional response that results in a precisely altered cellular state. Chapter 1 introduces how kinases regulate histones and histone modifiers, the basal transcriptional machinery, and transcription factors and cofactors to control inducible gene expression. Due to an overwhelming number of cellular phosphorylated substrates and limited methods for studying kinase function, it is oftentimes difficult to link kinases to their transcriptional outputs. Chapter 2 describes how a new generation of ATP-competitive mammalian target of rapamycin (mTOR) inhibitors has revealed a new role for mTOR in regulating cholesterol biosynthetic gene expression and a global gene expression profile that is resistant to inhibition using the natural product rapamycin. Although pharmacological inhibitors have their advantages, they do not allow for discrimination between primary and secondary consequences of kinase inhibition. I have developed a method to identify the primary transcriptional targets of the direct substrates of an individual kinase. Chapter 3 describes a chemical genetic affinity labeling scheme for tracing the direct substrates of an individual kinase and Chapter 4 illustrates how this method may be combined with chromatin immunoprecipitation to elucidate the transcriptional regulatory networks controlled by a single kinase.

PI3Kdelta and PI3Kgamma regulate immune cell signaling, while the related PI3Kalpha; and PI3Kbeta; regulate cell survival and metabolism. Selective dual inhibitors of PI3Kdelta and PI3Kgamma; represent a potential class of anti-inflammatory agents lacking the anti-proliferative effects associated with PI3Kalpha and PI3Kbeta inhibition. Here we outline the discovery and synthesis of PI3Kdelta; and PI3Kdelta/gamma dual inhibitors that display up to 1,000-fold selectivity over PI3Kalpha and PI3Kbeta. We explain the criteria for selectivity for PI3Kdelta and PI3Kgamma and evaluate the selective compounds in a high-content inflammation assay using mixtures of primary human cells. We find selective inhibition of only PI3Kdelta is weakly anti-inflammatory, but PI3Kdelta/gamma inhibitors show superior inflammatory marker suppression through suppression of LPS-induced TNFalpha production and T cell activation. Moreover, PI3Kdelta/gamma inhibition yields an anti-inflammatory signature distinct from pan-PI3K inhibition and known anti-inflammatory drugs, yet bears striking similarities to glucocorticoid receptor agonists. These results highlight the potential of selectively designing drugs that target kinases with shared biological function. We also outline the discovery and synthesis of a new set of PI3K inhibitors in which a small change in structure yields an extremely promiscuous molecule in contrast to the much more selective members of the set. We explain the principles behind the extreme selectivity achieved with the PI3Kdelta/gamma and PI3Kdelta inhibitors and hypothesize what led to the extreme promiscuity displayed by one molecule.

Determining proper sensitivity to incoming signals is fundamental to all regulated biological responses, but how thresholds are established within eukaryotic signal transduction networks is not understood. Here, we approach this question using chemical biology and single-cell western blotting, and then verify our conclusions with mathematical modeling. Focusing on the bistable kinase network that governs progesterone-induced meiotic entry in Xenopus oocytes, we characterize glycogen-synthase kinase-3&beta (GSK-3&beta) as a dampener of progesterone sensitivity. GSK-3&beta engages the meiotic kinase network through a double-negative feedback loop; this specific feedback architecture raises oocytes' progesterone threshold in proportion to the strength of double-negative signaling. These data demonstrate that intracellular signaling nodes can tune a system's response threshold away from the basal EC_{50 established by ligand-receptor interactions. Because oocytes actively depress their progesterone sensitivity, they can integrate additional signals into their decision to enter meiosis, an impossibility if they were maximally sensitive to one stimulus. Accordingly, we identify the branched-chain amino acid L-Leucine as a natural, sensitizing co-stimulus and show that an oocyte's native progesterone threshold can be lowered under specific nutrient conditions. In a chemical biology experiment that is conceptually similar to genetic epistasis, we show that L-Leucine adjusts oocytes' progesterone thresholds in a GSK-3&beta-dependent manner.}

The G protein-coupled receptor cascade leading to production of the second messenger cAMP is replete with pharmacologically targetable proteins with the exception of the Gα subunit, Gαs. GTPases remain largely undruggable given the difficulty of displacing high-affinity guanine nucleotides and the lack of other drug binding sites. We explored a chemical library of 10^12 cyclic peptides to expand the chemical search for inhibitors of this enzyme class. We identified two macrocyclic peptides, GN13 and GD20, that antagonize the active and inactive states of Gαs, respectively. Both macrocyclic peptides fine-tune Gαs activity with high nucleotide-binding-state selectivity and G protein class-specificity. Co-crystal structures reveal that GN13 and GD20 distinguish the conformational differences within the switch II/α3 pocket. Cell-permeable analogs of GN13 and GD20 modulate Gαs/Gβγ signaling in cells through binding to crystallographically defined pockets. The discovery of cyclic peptide inhibitors targeting Gαs provides a path for further development of state-dependent GTPase inhibitors.

Biologically active compounds exert their effects on living systems by binding to and modulating particular nodes among a vast network of macromolecules within the cell. The complexity of cellular machinery renders understanding which nodes are affected (i.e. direct targets or related pathways) a major challenge in the development of novel therapeutics. Furthermore, compounds of therapeutic interest might interact with cellular processes that were previously unknown, providing an opportunity for biological discovery in addition to informing drug chemical optimization. Here, we combined CRISPR (clustered regularly interspaced short palindromic repeats)-based functional genomics technologies with an atomistic approach to drug design across two important therapeutic targets, KRASG12C and MTOR. Using covalent switch-II pocket (S-IIP) inhibitors of KRASG12C, we nominated multiple genetic targets that could be co-inhibited to increase KRASG12C target engagement or block residual cancer survival pathways. We then applied a similar approach to uncover chemical-genetic interactions with a prototype bitopic MTOR inhibitor (molecular weight: 1784 g/mol), leading to the identification of an endogenous chemical uptake pathway involving interferon-induced transmembrane (IFITM) proteins. These proteins facilitate the cellular entry of diverse linked chemotypes, expanding cell permeable chemical space beyond traditional guidelines for drug-like molecules (e.g. Lipinski’s rule of five). These findings demonstrate the mutual utility of a combined functional genomics and chemical approach toward the development of therapeutics with novel mechanisms of action.

As mRNA translation is an energetically costly and crucial operation, cells have developed a plethora of mechanisms to regulate protein production at the post-transcriptional level, controlling initiation and elongation as well as mRNA stability to finely tune translation. In yeast, the DEAD-box protein Dhh1 has been implicated in coordinating translational repression and mRNA decay as a sensor of codon optimality but the exact molecular mechanism and whether this is recapitulated in mammalian cells through the homologue DDX6 still remain unclear. We are developing a chemical genetic inhibition strategy to elucidate the functions of DDX6 as well as other DEAD-box proteins using small molecules that covalently bind an engineered cysteine in the enzyme active site. This approach provides improved temporal and dosing control compared to traditional genetic methods. The latest efforts on this project aim to improve upon the first generation of probes, to discover new small molecule scaffolds, and to develop tools to monitor the probe’s efficacy in cells.Dysregulated protein synthesis is also a hallmark of many diseases, including cancer. The natural product Rocaglamide A (RocA) targets the DEAD-box protein eIF4A to selectively inhibit translation initiation on mRNAs with long, structured 5’-UTRs, including many oncogenes. We hypothesize that dimerizing RocA to add a layer of cell specificity through cellular uptake by IFITM proteins could “tune” the effects of such a potent molecule and broaden its therapeutic index by decreasing on target, off tumor side effects.

The proteomic mapping of enzyme-substrate interactions is challenged by their transient nature. A method to capture interacting protein kinases in complexes with a single substrate of interest would provide a new tool for mapping kinase signaling networks. Here we describe a nucleotide based substrate analog capable of reprogramming the wild-type phosphoryl-transfer reaction to produce a kinase-acrylamide based thioether crosslink to mutant substrates with a cysteine nucleophile substituted at the native phosphorylation site. A previously reported ATP-based methacrylate crosslinker (ATP-MA) was capable of mediating kinase crosslinking to short peptides but not protein substrates. Exploration of structural variants of ATP-MA to enable crosslinking of protein substrates to kinases led to the discovery that an ADP-based methacrylate (ADP-MA) crosslinker was superior to the ATP scaffold at crosslinking in vitro. The improved efficiency of ADP-MA over ATP-MA is due to reduced inhibition of the second step of the kinase-substrate crosslinking reaction by the product of the first step of the reaction. The new probe, ADP-MA, demonstrated enhanced in vitro crosslinking between the Src tyrosine kinase and its substrate Cortactin in a phosphorylation site specific manner. The kinase-substrate crosslinking reaction can be carried out in a complex mammalian cell lysate setting, although the low abundance of endogenous kinases remains a significant challenge for efficient capture.