The forces and energetic principles that guide molecular structure and dynamics in cellular lipid bilayers are poorly understood, largely due to limited familiarity with physical chemistry outside of water. We strive to enhance our fundamental chemical intuition concerning membrane proteins because their biochemical behavior within lipid membranes governs critical cellular processes and dictates health or disease. Likewise, this intuition paired with the ability manipulate membrane protein structure and function using computational modeling, theoretical calculations, and molecular design are instrumental to the engineering of new therapeutics. The work described here focuses on using protein design and integrative computational approaches to dissect the fundamental biophysical principles underlying protein interactions and folding within the lipid environment, spanning studies of both natural and totally synthetic proteins. Chapter 1 describes the design and experimental characterization of a family of simple synthetic model membrane proteins that rely entirely on the steric packing of lipid-embedded apolar sidechains for their stabilization and folding. This work reveals new fundamental principles governing van der Waals forces for proteins in lipid which should dictate protein structure, function, and evolution. Chapter 2 reports discovery of an evolutionarily conserved interaction motif within the transmembrane (TM) domains of integrin family proteins regulating their biochemical function. This discovery was driven by molecular modeling and all-atom dynamics simulations paired with co-workers’ optical tweezer force spectroscopy experiments. Chapter 3 details the development and application of software to design lipid-soluble peptides that target the transmembrane domains of proteins. We report several molecules that activate integrin α5β1 in human endothelial cell culture, and one with binding specificity for the β1 subunit. Chapter 4 outlines a second campaign with similar software to engineer small synthetic transmembrane proteins that bind and inhibit signal-amplifying erythropoietin receptor (EpoR), a canonical cytokine family receptor. Here, we demonstrate computational design can be used to target membrane-soluble molecules to specifically adopt custom, non-native binding modes to engage targeted regions of protein embedded in lipid. Thus, the efforts described represent leaps in our capacity to compute designer tool molecules entirely from scratch useful for manipulating membrane-spanning regions. These works together define progress in our fundamental biophysical understanding of protein interactions and folding in lipid, as well as our ability to dissect membrane protein structure and function with in silico molecular modeling algorithms and novel chemical biology tools.

Amyloid fibrils are insoluble protein aggregates with a broad range of biophysical properties and biological functions. Potentially all proteins can form amyloids, each of which can adopt a multitude of distinct molecular conformations, or polymorphs. Characterizing the molecular structure of various polymorphs and linking these features to their function or pathology is a prevalent challenge across biology, medicine, and technology. In this dissertation, I explore multiple aspects of amyloid structure and its relationship to biology. In Chapter 1, I work towards expanding a fluorogenic dye toolkit that can report on structural differences among amyloids; determining the structural composition of fibrils generated from multiple distinct species of amyloid beta (Aβ); and interrogating the atomic connections governing a functional amyloid’s macrostructure. In Chapter 2, I investigate the prevalence of different species of Aβ along the trajectory of Alzheimer’s disease (AD) pathology in Down syndrome (DS). In the latter work, I find that different subsets of Aβ may predominate in the brain tissue of people with early compared to advanced AD pathology and that a more heterogeneous population of Aβ species are present at the late stages of disease in DS.

The process by which various upstream sensor and signal-transduction domains of bacterial histidine kinases (HKs) modulate the activity of the conserved autokinase domain remains poorly understood. Specifically, why do most HKs contain modularly inserted signal transduction domains? How do HKs robustly evolve and finetune the coupling between stimulus sensor domains and the conserved autokinase domain, which are often separated by 10s of nanometers? What is the role of these intervening domains in fine-tuning signaling parameters such as the minimum/maximum responsiveness, mid-point, and steepness of signal transition of an HK? In this work, we examine signal transduction through model E. coli HKs, PhoQ and CpxA, which contain one of the most abundant signal transduction domains in HKs, the HAMP domain. We first generate a large set of single-point mutants of PhoQ, and simultaneously measure the signaling state of the ligand-binding sensor and the kinase activity of the autokinase in vitro, at several inducing ligand concentrations to assess the coupling between these two domains. We demonstrate that point mutants in the HAMP signal transduction domain significantly modulate the coupled behavior of the sensor and autokinase, producing markedly varied ligand-dependent responses. We further use the insertion of poly-glycine motifs (Gly7) to decouple domains from one another and qualitatively show that, intrinsically, the sensor domain has a drastically poor ligand-dependent state transition propensity, and similarly, the autokinase domain has a drastically high basal kinase activity. The HAMP domain strongly couples to both domains and is sufficient to adjust these propensities to what is observed in the full length PhoQ. We suggest that signal transduction in PhoQ occurs by an allosteric coupling mechanism, in which the HAMP domain strongly couples to and acts in opposition the underlying signaling state equilibria of PhoQ such that it is maximally responsive to physiologically relevant ranges of stimuli. We demonstrate the same phenomenon in two other E. coli HAMP containing HKs, CpxA and BaeS, and suggest this may be a common theme in the evolution of signal transduction domains in HKs. In order to quantitatively examine the feasibility of modulating various ligand-dependent properties that inform HK function through evolution, we next establish and experimentally fit a three-domain, two-state equilibrium allosteric signaling model. We demonstrate that small changes to the HAMP domain sequence allow for robust modulation of the signaling ensemble and provide quantitative measures for the strong modulation of both sensor and autokinase domains by the HAMP, as well as the effects of point-mutations and Gly7 insertions. We more fully examine the ability of the HAMP to couple strongly and influence the sensor and autokinase domains of PhoQ by introducing a large library of variants in the HAMP four-helix bundle hydrophobic core, as well as the junction between the HAMP and autokinase domains (the S-Helix) and selecting for variants with high PhoQ activity. We find that destabilizing the HAMP four-helix bundle hydrophobic core does indeed lead to higher kinase activity. Furthermore, we find that the wild-type S-Helix sequence is enriched in the high-activity population, along with sequences with comparable polarity or poor helical propensity. Taken together, these observations lend credence to the hypothesis that the thermodynamically preferred signaling state of the HAMP behaves as a negative allosteric regulator of the autokinase, and that this regulation is alleviated by destabilizing the core helical bundle structure as well as the alpha-helical motif that connects it to the autokinase. We investigate this relationship further using a deep learning method to establish sequence-activity predictive relationships and extract structural features that are essential for this behavior. Finally, we examine the question of whether the HAMP domain exists in two distinct structural states, or rather conformational ensembles that can be classified into one of two functional states. We examine signaling through the HAMP domain of an E. coli histidine kinase, CpxA, by constructing a small library of structurally diverse inputs into the HAMP domain and evaluate the resulting autokinase activity as a function of several S-helix point mutations. This analysis allows us to discern the relationship between different signal inputs into the HAMP domain as the linkage to the output domain (autokinase) is varied. We find that the HAMP seems to have a multiconformational landscape that is not explained by 2 unique structural conformations. In this thesis, we show that the insertion of signal transduction domains in HKs can significantly alter both the intrinsic behaviors of sensor and autokinase domains, as well as the coupling between them. These properties can be well-described through a coupled two-state allosteric mechanism, and easily finetuned through simple mutations to the signal transduction domain and its linkage to adjacent domains to achieve the desired physiologically relevant activity profile.

The M2 proton channel is a drug target of the influenza A virus. It is also a model system for the study of the selective, unidirectional transport of protons across the membrane bilayer. The work presented here focuses on the structural study of the M2 proton channel using X-ray crystallography. High resolution (1.10 Å) X-ray crystal structures of the Inward(open) conformation of M2 reveal layers of ordered, carbonyl-associated waters in the M2 pore. These waters form hydrogen-bonded networks of "water wires" that potentially play a role in the transport of protons down to the channel's gating His37 residues. This crystal form was examined using three data collection techniques: cryogenic diffraction at a synchrotron source, room temperature diffraction at a synchrotron source, and room temperature diffraction at an XFEL source. The degree of solvent ordering observed within the M2 pore is dependent on pH, and is also sensitive to temperature and the effects of radiation damage. New structures of M2 bound to amantadine and rimantadine reveal that the adamantane drugs bind by forming hydrogen bonds with ordered waters in the M2 pore, interrupting the water wires that lead down to the His37 gate. These drugs take advantage of the channel's essential ability to stabilize hydronium; the drug ammonium group acts as a hydronium mimic, which positions the bulky adamantyl group to sterically block the diffusion of hydronium into the channel. Additionally, structural characterization of drug-resistant S31N and V27A mutants of M2 characterize the mechanisms through which mutation of pore-lining residues at the N-terminus of the channel disrupt the binding of the adamantane drugs. These structures provide information for the design of new drugs to target these adamantane-resistant M2 mutants that have become prevalent among currently circulating strains of influenza.