UCSF

Purpose: The purpose of this study is to measure microshear bond strength (µSBS) of glass ionomer cement (GIC) to carious dentin with and without silver diamine fluoride (SDF) treatment.

Methods: Permanent molars were sectioned and demineralized to create artificial carious lesions. Variables tested included the demineralization of the dentin, application of SDF, use of conditioner, and time between SDF and restoration. µSBS was measured after 24 hours using an UltraTester machine.

Results: The strongest bond strength was found when GIC was placed on conditioned and demineralized dentin treated with SDF one week prior. There was no statistical difference to µSBS with and without SDF. Statistically significant increases in bond strength were found when the dentin was demineralized, when conditioner was applied before SDF, and when one week elapsed between SDF application and GIC placement. The lowest bond strength was found with immediate GIC application after SDF.

Conclusions: Results suggest that optimal retention is obtained by conditioning with polyacrylic acid and allowing SDF treatment to set for one week prior to GIC placement.

The development of therapeutic agent against cancer is based on targeting key signaling proteins that cancer highjacks and uses to survive. Although progress has been made to define cancer’s vulnerabilities, a subset of cancer drivers remain undruggable. To address this problem the field has attempted to identify drug targets that would selectively kill cancer cells and spare wild type tissue, a concept known as synthetic lethality. The work outlined here seeks to address major challenges in identifying synthetic lethal targets. First, I provide an overview of the platforms for synthetic lethal screening and highlight the advantages and caveats of each approach. Chapter three is focused on a case study where we developed a network-based integration method for published KRAS synthetic lethal studies and derived principles for synthetic lethal screening. The major findings of this study highlight principles of synthetic lethal screening and identify a subset of genes, which may offer new therapeutic targets in the context of oncogenic KRAS. Chapter four explores the use of PARP inhibitors in non-small cell lung cancer cell lines and derive molecular signatures associated with response and resistance to PARP inhibitor. Chapter 5 reports the results of a KRAS 4a/4b drug screen which highlight isoform specific vulnerabilities that may inform therapeutic strategies for KRAS mutant cancers.

I present computational and experimental methods relating to the design of binding interactions involving proteins, including interactions of protein/small molecule, dimeric protein/protein, and tertiary protein/small molecule/protein systems. In chapter 2, I describe a benchmark comparison of flexible backbone design methods in Rosetta. Three methods, (1) BackrubEnsemble, (2) CoupledMoves, and (3) FastDesign, were tested for their ability to recapitulate observed protein sequence profiles assumed to represent the fitness landscapes of protein/protein and protein/small molecule binding interactions. We found that CoupledMoves, which combines backbone flexibility and sequence exploration into a single acceptance step during the sampling trajectory, better recapitulates sequence profiles than BackrubEnsemble and FastDesign, which separate backbone flexibility and sequence design into separate acceptance steps during the sampling trajectory. In chapter 3, I describe the screening and characterization of a chemically induced dimer (CID) that detects and responds to the presence of ibuprofen. The protein tool is composed of a sensor module and a reporter module, which are modular and can be interchanged. The sensor module is a heterodimer whose interface contains an ibuprofen binding site transplanted by computational design from a monomeric protein, such that ibuprofen binding induces heterodimerization. The reporter module is a protein complementation system whose complementation is induced by dimerization of the sensor domain. I present two methods to individually screen hundreds of designed CIDs targeting various proteins, (1) using a growth-based reporter module in E coli, and (2) using a luminescent reporter in a cell-free protein expression system. The work presented here represents methodological advances for both the computational and experimental design of protein binding interactions.

Infectious diseases are a leading cause of morbidity and mortality worldwide. Despite significant advancement in our understanding of infectious disease biology, existing microbiologic diagnostic tests often fail to identify etiologic pathogens in cases of suspected infection. Metagenomic next-generation sequencing (mNGS) offers the potential for a universal pathogen detection method, but analysis and interpretation of findings are challenging. This is especially true for lower respiratory tract infections (LRTIs) where mNGS data interpretation is complicated by the existence of a respiratory microbiome composed of pathobionts present in both health and disease.

To address the need for improved LRTI diagnostics, we first compared two fluid types commonly used for diagnosis of LRTI, showing that despite moderate microbiome differences, both mini-bronchioalveolar lavage (mBAL) and tracheal aspirate (TA) samples are suitable for identification of pathogens in the context of an infection. Then, we evaluated the utility of mNGS as a diagnostic for LRTI in a cohort of 92 TA samples from adults with acute respiratory failure. We developed methods for sifting putative pathogens from commensal microbiota as well as pathogen, microbiome diversity, and host gene expression metrics to identify LRTI-positive patients and differentiate them from critically ill controls with noninfectious acute respiratory illnesses. We applied the models developed for evaluation of LRTI status to several other cohorts and disease contexts to show their broad applicability.

The low sensitivity of existing clinical diagnostics results in an imperfect gold standard, complicating the development of mNGS-based biomarkers. We explored the impact of label noise on host gene expression classifiers and methods for circumventing the issue. First, we tested whether label-noise robust logistic regression approaches could improve classifier performance by enabling the use of a larger training set. Then, we tested whether variational autoencoders, an unsupervised dimensionality reduction approach, could generate novel insight from combined host and microbial mNGS data. Altogether, this work suggests that a single streamlined protocol offering an integrated genomic portrait of pathogen, microbiome, and host transcriptome may hold promise as a tool for diagnosis of infections and contextualization of patient response.

Mitochondria carry their own genome (mtDNA), which is present in multiple copies in all eukaryotic cells. Copy number of mtDNA in each cell is tightly maintained, yet surprisingly the cellular mechanisms that regulate mtDNA copy number remain poorly understood. To address this question, we carried out a forward genetic screen in the budding yeast S. cerevisiae and identified mutants exhibiting altered mtDNA levels. This screen revealed a previously uncharacterized mitochondrial gene, Mrx6, whose deletion results in a marked increase of mtDNA without affecting mitochondrial structure or cell size. We found that Mrx6 forms a complex with a sequence-related protein, Pet20, with Mam33, and with the conserved Lon protease Pim1, which is important for mitochondrial protein quality control. Furthermore, the Mrx6 complex colocalizes with mtDNA. Because human and bacterial Lon proteases have been proposed to regulate DNA replication by degrading replication initiation factors, our results suggest that the Mrx6 complex may similarly control mtDNA levels through degradation of key proteins regulating mtDNA replication.

The basal ganglia are a series of interconnected subcortical nuclei involved in movement, action selection, and decision making. These processes are dependent on the coordinated output of the striatal direct and indirect pathways, which is dysregulated in neurological conditions such as Parkinson’s disease, Huntington’s Disease, and forms of dystonia. However, the specific cells, circuits, and patterns of activity within the brain that contribute to disease manifestations, such as involuntary movements, or dyskinesias, are not well understood. To address this gap, we have used two mouse models of human dyskinesias, paroxysmal nonkinesigenic dyskinesia (PNKD) and levodopa-induced dyskinesia (LID), to investigate whether aberrant striatal activity is a root cause of dyskinesia. Using a variety of in vivo and ex vivo techniques to both record and manipulate neural activity, we found that abnormal patterns of activity in the striatum give rise to dyskinesia. Interestingly, in PNKD we see profound reductions in striatal indirect pathway activity during dyskinesia, whereas in LID there are decreases in indirect pathway activity, but also abnormally high direct pathway activity during dyskinesia. Further, we found that the decrease in indirect pathway activity is necessary and sufficient for drug-induced dyskinesia in PNKD, while an increase in activity in a subset of direct pathway neurons is necessary and sufficient for drug-induced dyskinesia in LID. In both models, we found evidence of aberrant striatal synaptic plasticity as a cellular correlate of dyskinesia: depressed excitatory input onto indirect pathway neurons in the case of PNKD, and enhanced excitatory input onto direct pathway neurons in LID. While these results are largely in support of the classical model of basal ganglia function, we also found heterogeneity within the canonical direct pathway in LID, and identified a subclass of direct pathway neurons with exceptionally high levodopa-evoked firing that correlates strongly with dyskinesia, Importantly, these results may guide the development of new therapeutics for hyperkinetic disorders based on targeting specific subclasses of striatal neurons or their connections.

Life often diversifies through changes in gene expression patterns. These patterns evolve via changes in transcriptional regulatory circuits that are determined by transcriptional regulatory proteins and the cis-regulatory sequences they bind in the genome. While it has long been known that changes in cis-regulatory sequences can affect the evolution of gene expression patterns and that transcriptional regulatory proteins can themselves evolve, we know little of how these two types of regulatory changes occur together to generate new circuits. I discerned a stepwise order of evolutionary events in which both regulator protein-coding and cis-regulatory changes were necessary to evolve a new transcriptional regulatory circuit (repression of the a-specific genes by Mat⍺2 in yeast). The two changes evolved at separate points in time, millions of years apart. First to evolve were coding-sequence changes in the regulator that formed new protein-protein interaction regions. In one lineage, these new protein-protein interactions became necessary for Mat⍺2’s ancestral gene regulatory function (repression of the haploid-specific genes with Mata1). In another lineage, millions of years after the coding-sequence changes to Mat⍺2, cis-regulatory changes occurred in the a-specific genes, thereby co-opting Mat⍺2 for regulation of this new set of target genes. We propose that this evolutionary trajectory is an example of constructive neutral evolution in that Mat⍺2’s new protein-protein interactions initially had no consequence to the logic of cell-type specific gene regulation, but eventually allowed for the creation of a novel circuit (Chapter 2). In the course of these investigations, I also observed additional coding-sequence changes in the DNA-binding domain of Mat⍺2 (Chapter 3), and evolutionary changes in the identities of some of the yeast cell-type specific genes (Chapter 4). The results presented here add to our understanding of the ways in which transcriptional regulatory circuits diversify.

Human disease risk has been linked to hundreds of variants in our DNA. Understanding of these unbiased genetic associations has long held the promise of revealing insights into disease mechanisms. However, disease-risk variants overwhelmingly reside in non-coding sequences – long stretches of our chromosomes that we still know relatively little about. Here I develop new tools and methodologies to understand genetic risk for disease for a critical autoimmunity locus, IL2RA. I first demonstrate that CRISPR activation can be adapted for high throughput enhancer screens. By tiling CRISPR-activation across the super-enhancer within the IL2RA locus, I systematically map functional IL2RA enhancers in the disease-associated non-coding sequences. Undertaking a genetic perturbation approach, I dissect how distinct IL2RA enhancers regulate immune cell function as well as shape risk of autoimmunity in vivo. Using CRISPR-engineered enhancer deletion mice and human immune cells I identified two novel IL2RA enhancers; a maintenance enhancer that controls IL2RA expression in anti-inflammatory regulatory T cells and a disease-associated IL2RA enhancer that controls the timing of IL2RA induction in pro-inflammatory immune cells. Having discovered enhancers that regulate IL2RA in different contexts I interrogated their effects in an in vivo model of autoimmune disease. Deletion of the conserved stimulation-responsive enhancer that harbors a human variant protective against T1D completely protected non-obese diabetic (NOD) mice from diabetes. This work decodes a critical autoimmunity association, develops a cis-regulatory framework at the IL2RA locus, and causally links IL2RA gene regulation to autoimmunity. The tools and strategies developed in these studies can be used to decode disease-associated loci in the human genome.

Alzheimer’s disease (AD) is the leading cause of dementia worldwide, and no effective therapies are available. The multifactorial etiology and pathophysiological complexity of AD cause patient heterogeneity and pose challenges for drug development, with almost all efforts to target AD-related pathways having failed in human trials. Although apolipoprotein (apo) E4 is the major genetic risk factor for AD—60–80% of patients have at least one APOE4 allele and ~70% of homozygotes develop AD by age 85—it has not been actively considered in drug target stratification and development for AD. Here, we used an apoE-genotype-specific drug repositioning approach to screen for drugs to treat apoE4-related AD. From a meta-analysis of 610 human temporal lobar samples from public databases, we established apoE-genotype-specific transcriptomic signatures of AD and applied them to a validated Connectivity Map (CMap) database containing transcriptomic perturbation signatures of 1300 existing drugs to identify those capable of perturbing an entire gene-expression network away from the apoE-genotype-driven disease state towards a normal state. The loop-diuretic bumetanide was the top predicted drug candidate for apoE4/4 AD. Treating aged apoE4 knock-in (apoE4-KI) mice with bumetanide rescued cognitive and neuronal plasticity deficits, warranting further efficacy tests in AD clinical trials. This study highlights the power of combining precision medicine, computational drug repositioning, and targeting network alterations in developing new therapies for AD and other neurodegenerative disorders.

The molecular chaperone Hsp90 interacts with around 10% of the proteome, facilitating folding and regulating the biological function of its clients. Decades of work have led to a general understanding of the ATP-driven conformational cycle of Hsp90 and its role in client regulation, however the molecular details of this mechanism remain unknown. This is due in part to the intrinsic instability of Hsp90 clients, which has until recently hampered in-vitro and structural work, as well as to the difficulty in breaking down Hsp90s ATP-driven cycle into discrete steps that can be observed and measured together with changes in client structure and function. The work described in this thesis addresses two of the main questions that the Hsp90 field faces: 1) How is Hsp90’s conformational cycle and ATP hydrolysis used to remodel clients and 2) is there a conserved mechanism of client recognition and engagement that allows Hsp90 to functionally interact with dozens of clients of different size, sequence and structure.

The first chapter of this thesis addresses the first question and builds on work from Laura Lavery and James Partridge which revealed that the closed state of the mitochondrial Hsp90, TRAP1, is structurally asymmetric with one protomer in a ‘straight’ conformation while the other one adopts a novel ‘buckled’ conformation. Starting from this catalytic closed state, we investigate the effect of each ATP hydrolysis event on Hsp90’s conformation and how Hsp90 uses these hydrolysis-driven changes to remodel and regulate clients. We show that this asymmetry sets up sequential and deterministic hydrolysis, with the buckled protomer hydrolyzing first followed by a flip of the asymmetry that positions the unhydrolyzed protomer in the hydrolysis-favorable buckled conformation, promoting the second hydrolysis leading to reopening of the chaperone. While this asymmetry has not been observed in eukaryotic Hsp90 homologs, we propose that, despite being homodimers, all Hsp90s work as asymmetric machines, either through a structural asymmetry like in the case of TRAP1 or through formation of asymmetric complexes with clients and co-chaperones in the case of eukaryotic Hsp90s.

The third chapter of this thesis addresses the second question through the determination of the cryoEM structure of the bacterial Hsp90, HtpG, in complex with its client ribosomal protein L2. Though this work is still ongoing, the current cryoEM reconstruction of the apo-HtpG:L2 complex shows that, surprisingly, HtpG is in a GRP94-like conformation. Also striking is the fact that L2 interacts with HtpG in a Cdk4-like fashion, with density observed on both sides of HtpG and connected by a thread of density going through the lumen of HtpG along the same region previously identified for L2 and the model client Δ131Δ binding to HtpG. Additionally, L2 makes further interactions with HtpG different from those observed with Cdk4, notably with the N-terminal domain of HtpG around the lid area.

Overall, this work provides further evidence supporting that the different conformational states of Hsp90 are conserved across homologs but differentially regulated, and supports a conserved mechanism of interaction between Hsp90 and different classes of clients where Hsp90 clamps between domains stabilizing structural and likely short-lived transitions in the client.