Cyclic peptides in the beyond rule of 5 space are increasingly gaining traction as a therapeutic platform, and their astonishing passive permeability and ability to bind difficult or “undruggable” targets is continually studied, developed, and exploited as potent lead drug compounds. RNA, as a cellular information carrier and gene regulator, has been implicated in many human diseases. The therapeutic potential of targeting RNA with drug-like molecules is being increasingly recognized and explored, but the physicochemical properties required to selectively target RNA from a membrane-permeable, therapeutically relevant paradigm remain unprobed, especially important considering existing efforts have yielded compounds that are largely antithetical to membrane permeability. We investigated the possibility of lipophilic, cell-permeable cyclic peptides binding RNA with high affinity and specificity. Chapter 1 describes the design and deployment of two choice cyclic peptides against a library of randomized RNAs to yield an RNA aptamer to each cyclic peptide that binds with high affinity and specificity. Sequence level apo and holo chemical probing efforts reveal crucial differences between the two complexes.

In a collaborative effort with the Theodore Holman lab (UC Santa Cruz), chapter 2 describes attempts towards the medicinal chemistry optimization of SLUG01, a derivative of a ML355, a previously reported human 12-Lipoxygenase inhibitor.

Investigating the mechanisms of action (MOAs) of bioactive compounds and the deconvolution of their cellular targets is an important and challenging undertaking. Drug resistance in model organisms such as S. cerevisiae has long been a means for discovering drug targets and MOAs. Strains are selected for resistance to a drug of interest, and the resistance mutations can often be mapped to the drug’s molecular target using classical genetic techniques. Here we demonstrate the use of next generation sequencing (NGS) to identify mutations that confer resistance to two well-characterized drugs, benomyl and rapamycin. Applying NGS to pools of drug-resistant mutants, we develop a simple system for ranking single nucleotide polymorphisms (SNPs) based on their prevalence in the pool, and for ranking genes based on the number of SNPs that they contain. We clearly identified the known targets of benomyl (TUB2) and rapamycin (FPR1) as the highest-ranking genes under this system. The highest-ranking SNPs corresponded to specific amino acid changes that are known to confer resistance to these drugs. We also found that by screening in a pdr1 null background strain that lacks a transcription factor regulating the expression of drug efflux pumps, and by pre-screening mutants in a panel of unrelated anti-fungal agents, we were able to mitigate against the selection of multi-drug resistance (MDR) mutants. We call our approach “Mutagenesis to Uncover Targets by deep Sequencing, or “MUTseq”, and show through this proof-of-concept study its potential utility in characterizing MOAs and targets of novel compounds.

Cationic amphiphilic drugs (CADs) are a broad class of chemicals that can cause adverse effects in humans, and particularly liver toxicity; the origin of this toxicity is not well understood. The CAD haloperidol, for example, is known to cause drug induced liver cholestasis and phospholipidosis (DIPL), severe side effects similar to those seen in familial intrahepatic cholestasis (FIC). In a high throughput screen in yeast, we discovered that several known CADs, as well as a number of novel related compounds, exhibit synthetic lethality with RCY1, a gene encoding an F-box protein involved in endocytic recycling of membrane proteins. To investigate the origin of this synthetic lethality, we isolated S. cerevisiae mutants resistant to the CADs haloperidol and a novel CAD identified in this study, 3346-2086. Whole genome sequencing revealed that resistance is often associated with mutations in the yeast phospholipid transporters, DRS2, DNF1 and DNF2. Further investigations showed that in the absence of these mutations, Rcy1- cells treated with 3346-2086 rapidly lose the ability to transport fluorescent analogues of phosphatidylcholine (PC) and phosphatidylethanolamine (PE), an effect that is largely blocked in the mutants. Our findings provide insight into the functional relationship between Rcy1p and phospholipid transporters and the conditions under which administration of CADs can give rise to drug-induced liver cholestasis and phospholipidosis.

Millions of people world-wide are affected by diseases and conditions related to microbial pathogens and the immune system. The demand for new therapies against antibiotic resistant bacteria and anti-inflammatories is urgent. My dissertation encompasses two distinct avenues of modern drug discovery: 1) A chemical biology approach to improve the physical properties and potency of novel cyclic peptide inhibitors to combat the virulence of aggressive gram negative bacteria by inhibiting the Type 3 Secretion System (T3SS) and 2) The development of a high-content image-based screening assay for the detection of anti-inflammatories. Chapter 1 describes the structural activity relationship studies of a novel cyclic peptide T3SS inhibitor and its potency against various gram negative bacteria T3SS. Chapter 2 describes the development of a high content image-based screen using macrophages and pro-inflammatory endotoxin lipopolysaccharide (LPS). We compare this screening platform to classic HeLa cytological profiling (CP) and observe the limitations of image based feature extraction techniques to cluster by MOA based on similarity profiles. Furthermore, we use supervised deep learning approaches to identify compounds that reverse the LPS-stimulated phenotype and confirm the selected deep learning hits exhibit an immunomodulatory response to LPS. Lastly we briefly describe our efforts in developing a costume-built online database for viewing the images generated from the HCS screen. This database brings side-by-side comparisons of macrophages treated with compounds +/- LPS stimuli and both stain sets. In addition, it's searchable and includes compound metadata associated with every image, allowing for quick inspection of the images.

Much of modern pharmaceutical development has occurred within or nearby the chemical space delineated by Lipinski’s rules of five (< 500 molecular weight, < 5 hydrogen bond donors, < 5 octanol/water partition coefficient, < 10 hydrogen bond acceptors) due to the practical advantages in pharmacokinetic properties evinced by such small molecules, especially cell permeation. This leaves intracellular interactions which occur over a larger surface area (i.e. not involving native small molecules) “undruggable”. Cyclic peptides of up to 1200 molecular weight have demonstrated the ability to inhibit a wide variety of such undruggable intracellular interactions. However, engineering cell permeability into cyclic peptides remains a major barrier to their therapeutic utility. Here I present my work using tandem mass spectrometry and the PAMPA artificial membrane permeability assay to Much of modern pharmaceutical development has occurred within or nearby the chemical space delineated by Lipinski’s rules of five (< 500 molecular weight, < 5 hydrogen bond donors, < 5 octanol/water partition coefficient, < 10 hydrogen bond acceptors) due to the practical advantages in pharmacokinetic properties evinced by such small molecules, especially cell permeation. Their small size has made them ideal for occupying small binding pockets instead of a protein’s intended substrate and achieving sufficient oral bioavailability for oral delivery is generally facile. More recently, therapeutic enzymes and antibodies have joined Insulin in a class of injectable macromolecule therapeutics to great success – multiple new therapeutic antibodies are approved each year. Their large size gives them superior selectivity, specificity, and potency compared to small molecules, but also precludes them from entering cells to modify intracellular interactions (and from oral delivery). These two therapeutic modalities leave a great many intracellular interactions which occur over a larger surface area (i.e., not involving native small molecules) “undruggable”, and the continual discovery of actionable disease-relevant interactions in this category has prompted a search for new therapeutic modalities. Cyclic peptides of up to 1200 molecular weight have demonstrated the ability to inhibit a wide variety of such “undruggable” intracellular interactions. Their ease of synthesis combined with advances in DNA/mRNA encoded library construction and screening technologies have ensured that obtaining potent cyclic peptide leads against virtually any protein target is possible. However, these powerful encoding technologies cannot screen for permeability and engineering cell permeability into cyclic peptides remains a major barrier to their therapeutic utility. The conformational nature of cyclic peptide passive cell permeability has thus far defied computational prediction over a broad set of compounds and empirical evaluations of cyclic peptide passive permeability have been of limited size (tens to low hundreds). The overall goals of this dissertation are to empirically evaluate the prevalence of passively cell permeable backbone geometries across thousands of geometrically diverse cyclic hexa- and heptapeptides, to derive general insights into the design of passively cell permeable cyclic peptides, and to publish a database of known permeable backbone geometries with which screening libraries can be biased towards passive cell permeability. Chapter one covers the details of CycLS, a non-de novo cyclic peptide sequencing program using a database-matching approach to allow sequencing of entire libraries of cyclic peptides in a timely manner. I validated CycLS against a unique-mass library of 400 cyclic hexapeptomers, achieving 95% sequencing accuracy despite a ratio greater than 500:1 of decoy sequences to true sequences, and against a mass-redundant 1800-member library of cyclic hexapeptides and hexapeptomers (also found in chapter two) by resynthesis of twenty-two individual compounds over a broad range of sequencing scores. I then devised a normalized sequencing confidence metric that was able to divide the seventeen successfully sequenced resynthesized compounds from the five unsuccessfully sequenced resynthesized compounds. Direct sequencing of cyclic peptides rather than by linearization or encoding is critical for passive cell permeability assays, which are sensitive to any encoding tag or the structural elements pre-requisite to many linearization strategies. CycLS is freely available online and improves on previous work in this area by inclusion of extensive spectral pre-processing to remove noise and boost signal, which is critical to ensure sequencing quality. CycLS is the first step of a computational workflow to associate individual compound identities with corresponding mass spectrometry quantified assay results by matching individual peak retention times given an identical LCMS method. The remainder of that workflow can be found in appendix A and is composed of CycLS and two additional steps. Processing of the assay data is done via AutoPAMPA, which performs automated peak-finding, integration, and calculation of permeation rates for the PAMPA artificial membrane permeability assay. AutoPAMPA removes data analysis as the limiting factor to throughput for mass spectrometry quantified assays by batch-processing of hundreds to thousands of peaks per job. Finally, RTMerge performs peak alignment and matching by retention time between CycLS and AutoPAMPA outputs and generates statistics describing the physicochemical properties and structure of each compound. In addition to the work included here, this computational workflow has already been successfully used to explore the permeabilities of hundreds of lariat peptides and enables further such projects in the future. Chapter two investigates the passive permeability of 1800 cyclic hexamers and 3600 structurally related cyclic heptamers with highly variant backbone geometries by PAMPA. Of especial interest were the effects of N-methyl residues, peptoid residues, and beta residues on passive permeability. I identified 823 hexamers and 1330 heptamers with permeation rates greater than 1 * 10-6 cm/s, a threshold at which compounds are considered passively cell permeable. I confirmed the utility of these library-derived permeabilities by correlation with the pure permeabilities of 9 resynthesized hexamers and 10 resynthesized heptamers. A matched-pair analysis revealed that peptoid and beta residues have a negative structural contribution to passive permeability that I hypothesize originates from their increased flexibility. As expected, a matched-pair analysis of stereochemistry showed little effect averaged over such diverse backbone geometries. Library generation technologies with complete synthetic control stand to benefit greatly from the permeable hexamer and heptamer backbone geometries discovered in chapter two by selecting only the most permeable of them, but combinatorically generated libraries doing the same would be limited to small library sizes or few backbone geometries. To enable combinatorically generated libraries to better utilize known permeable backbone geometries I defined and investigated passive permeability “motifs” of length three, holding three residues of the library design static while allowing the rest to vary over a number of known permeable backbone geometries. The best motifs had median permeabilities four-fold greater than the median permeability of all other compounds with the same number of hydrogen bond donors. Bundled into motifs, these sets of permeable backbone geometries allow combinatorically generated libraries increased size and some degree of geometric diversity. This dissertation has thoroughly explored the impact of stereochemistry, N-methylation, and peptoid residues on the passive permeability of cyclic hexa- and heptamers and gained some insights into the effect of beta residues. In addition to these insights into the average effects on permeability, this dissertation emphasizes that backbone geometries with high intrinsic permeability may be designed into DNA/mRNA encoded screening libraries to improve the likelihood of hits with favorable pharmacokinetic properties. Lastly, the computational workflow necessary to gain these insights can be used to obtain a similar register of “privileged” backbone geometries for larger ring sizes.

The conformational hypothesis of membrane permeability hypothesizes that certain macrocyclic scaffolds can exhibit an abnormally high degree of permeability producing a larger logP value than expected due to intramolecular hydrogen bonding which locks the macrocycle into a specific conformation where the hydrophilic functional groups are buried within the molecule causing it to become lipophilic enough to cross the membrane. Exploring this concept in previous work resulted in the synthesis of a 32 membered library of cyclic peptides with the same sequence varying in stereochemistry and N-methylation pattern. From this library two cyclic peptides, compounds 1 and 2 were found to exhibit an abnormally large logP value when assayed using the PAMPA permeability assay. This result lead us to synthesize more of these compounds and subject them to a series of NMR and other analytical experiments to computationally determine these two molecules 3D structure as they exist in chloroform solution, used to mimic the environment of the inside of membrane bilayer. These NMR structures were produced using NOESY derived distance restraints and the molecular modeling program CYANA. These structures provide valuable information about what types of conformational motifs can be used when designing novel therapeutic compounds.