The importance of input sequence set to consensus-derived proteins and their relationship to reconstructed ancestral proteins.

A protein sequence encodes its energy landscape-all the accessible conformations, energetics, and dynamics. The evolutionary relationship between sequence and landscape can be probed phylogenetically by compiling a multiple sequence alignment of homologous sequences and generating common ancestors via Ancestral Sequence Reconstruction or a consensus protein containing the most common amino acid at each position. Both ancestral and consensus proteins are often more stable than their extant homologs-questioning the differences between them and suggesting that both approaches serve as general methods to engineer thermostability. We used the Ribonuclease H family to compare these approaches and evaluate how the evolutionary relationship of the input sequences affects the properties of the resulting consensus protein. While the consensus protein derived from our full Ribonuclease H sequence alignment is structured and active, it neither shows properties of a well-folded protein nor has enhanced stability. In contrast, the consensus protein derived from a phylogenetically-restricted set of sequences is significantly more stable and cooperatively folded, suggesting that cooperativity may be encoded by different mechanisms in separate clades and lost when too many diverse clades are combined to generate a consensus protein. To explore this, we compared pairwise covariance scores using a Potts formalism as well as higher-order sequence correlations using singular value decomposition (SVD). We find the SVD coordinates of a stable consensus sequence are close to coordinates of the analogous ancestor sequence and its descendants, whereas the unstable consensus sequences are outliers in SVD space.

Ancestral sequence reconstruction dissects structural and functional differences among eosinophil ribonucleases.

Evolutionarily conserved structural folds can give rise to diverse biological functions, yet predicting atomic-scale interactions that contribute to the emergence of novel activities within such folds remains challenging. Pancreatic-type ribonucleases illustrate this complexity, sharing a core structure that has evolved to accommodate varied functions. In this study, we used ancestral sequence reconstruction to probe evolutionary and molecular determinants that distinguish biological activities within eosinophil members of the RNase 2/3 subfamily. Our investigation unveils functional, structural, and dynamical behaviors that differentiate the evolved ancestral ribonuclease (AncRNase) from its contemporary eosinophil RNase orthologs. Leveraging the potential of ancestral reconstruction for protein engineering, we used AncRNase predictions to design a minimal 4-residue variant that transforms human RNase 2 into a chimeric enzyme endowed with the antimicrobial and cytotoxic activities of RNase 3 members. This work provides unique insights into mutational and evolutionary pathways governing structure, function, and conformational states within the eosinophil RNase subfamily, offering potential for targeted modulation of RNase-associated functions.

Navigating contradictions: Salmonella Typhimurium chemotactic responses to conflicting chemoeffector signals show parity with bacterial growth benefits.

Many bacteria that colonize the guts of animals use chemotaxis to direct swimming motility and select sites for colonization based on sources of effectors derived from the host, diet, and microbial competitors of the local environ. The complex ecosystem of the gastrointestinal tract contains mixtures of chemoattractants and chemorepellents, but it remains poorly understood how swimming bacteria navigate conflicting signals. The enteric pathogen Salmonella Typhimurium possesses Tsr, a chemoreceptor protein that directs both chemoattraction and chemorepulsion responses, which we employed as a model to study chemotaxis in the presence of conflicting effector stimuli. We investigated how S. Typhimurium responds to human fecal matter, an effector source in the enteric lumen that contains high concentrations of indole, a bacteriostatic chemorepellent produced by the native commensals of the microbiota, and also nutrients such as l-serine, a chemoattractant. The indole concentration in human feces is more than 12-fold the concentration required for half-maximal chemorepulsion, however, we find S. Typhimurium, and various clinical isolates of non-typhoidal S. enterica serovars, are strongly attracted to liquid fecal matter. We further investigated the chemotactic responses of S. Typhimurium to titrations of indole and l-serine and revealed that chemorepulsion to indole is overridden in the presence of excess l-serine. We capture the inversion of these two opposing taxis behaviors in a phenomenon we define as "chemohalation" in which the bacteria organize into a halo around the treatment source with an interior zone of avoidance, which represents a compromise between chemoattraction and chemorepulsion. Growth analyses reveal that the chemotactic responses to these opposing effectors align chemoattraction and chemorepulsion with the relative growth of the bacteria in culture. Hence, our study supports the view that evolution has finely tuned chemotaxis to assess environmental habitability by evaluating the tradeoffs in bacterial growth based on the local combination of effectors.

Bacterial vampirism mediated through taxis to serum.

Bacteria of the family Enterobacteriaceae are associated with gastrointestinal (GI) bleeding and bacteremia and are a leading cause of death, from sepsis, for individuals with inflammatory bowel diseases. The bacterial behaviors and mechanisms underlying why these bacteria are prone to bloodstream entry remains poorly understood. Herein, we report that clinical isolates of non-typhoidal Salmonella enterica serovars, Escherichia coli, and Citrobacter koseri are rapidly attracted toward sources of human serum. To simulate GI bleeding, we utilized a custom injection-based microfluidics device and found that femtoliter volumes of human serum are sufficient to induce the bacterial population to swim toward and aggregate at the serum source. This response is orchestrated through chemotaxis, and a major chemical cue driving chemoattraction is L-serine, an amino acid abundant in serum that is recognized through direct binding by the chemoreceptor Tsr. We report the first crystal structures of Salmonella Typhimurium Tsr in complex with L-serine and identify a conserved amino acid recognition motif for L-serine shared among Tsr orthologues. By mapping the phylogenetic distribution of this chemoreceptor we found Tsr to be widely conserved among Enterobacteriaceae and numerous World Health Organization priority pathogens associated with bloodstream infections. Lastly, we find that Enterobacteriaceae use human serum as a source of nutrients for growth and that chemotaxis and the chemoreceptor Tsr provides a competitive advantage for migration into enterohaemorrhagic lesions. We term this bacterial behavior of taxis toward serum, colonization of hemorrhagic lesions, and the consumption of serum nutrients, as 'bacterial vampirism' which may relate to the proclivity of Enterobacteriaceae for bloodstream infections.

Ancestral Reconstruction and the Evolution of Protein Energy Landscapes.

A protein's sequence determines its conformational energy landscape. This, in turn, determines the protein's function. Understanding the evolution of new protein functions therefore requires understanding how mutations alter the protein energy landscape. Ancestral sequence reconstruction (ASR) has proven a valuable tool for tackling this problem. In ASR, one phylogenetically infers the sequences of ancient proteins, allowing characterization of their properties. When coupled to biophysical, biochemical, and functional characterization, ASR can reveal how historical mutations altered the energy landscape of ancient proteins, allowing the evolution of enzyme activity, altered conformations, binding specificity, oligomerization, and many other protein features. In this article, we review how ASR studies have been used to dissect the evolution of energy landscapes. We also discuss ASR studies that reveal how energy landscapes have shaped protein evolution. Finally, we propose that thinking about evolution from the perspective of an energy landscape can improve how we approach and interpret ASR studies. Expected final online publication date for the Annual Review of Biophysics, Volume 53 is May 2024. Please see http://www.annualreviews.org/page/journal/pubdates for revised estimates.

Simultaneous Mapping of DNA Binding and Nucleosome Positioning with SpLiT-ChEC.

The organization of chromatin - including the positions of nucleosomes and the binding of other proteins to DNA - helps define transcriptional profiles in eukaryotic organisms. While techniques like ChIP-Seq and MNase-Seq can map protein-DNA and nucleosome localization separately, assays designed to simultaneously capture nucleosome positions and protein-DNA interactions can produce a detailed picture of the chromatin landscape. Most assays that monitor chromatin organization and protein binding rely on antibodies, which often exhibit nonspecific binding, and/or the addition of bulky adducts to the DNA-binding protein being studied, which can affect their expression and activity. Here, we describe SpyCatcher Linked Targeting of Chromatin Endogenous Cleavage (SpLiT-ChEC), where a 13-amino acid SpyTag peptide, appended to a protein of interest, serves as a highly-specific targeting moiety for in situ enzymatic digestion. The SpyTag/SpyCatcher system forms a covalent bond, linking the target protein and a co-expressed MNase-SpyCatcher fusion construct. SpyTagged proteins are expressed from endogenous loci, whereas MNase-SpyCatcher expression is induced immediately before harvesting cultures. MNase is activated with high concentrations of calcium, which primarily digests DNA near target protein binding sites. By sequencing the DNA fragments released by targeted MNase digestion, we found that this method recovers information on protein binding and proximal nucleosome positioning. SpLiT-ChEC provides precise temporal control that we anticipate can be used to monitor chromatin under various conditions and at distinct points in the cell cycle.

Disentangling contact and ensemble epistasis in a riboswitch.

Mutations introduced into macromolecules often exhibit epistasis, where the effect of one mutation alters the effect of another. Knowing the mechanisms that lead to epistasis is important for understanding how macromolecules work and evolve, as well as for effective macromolecular engineering. Here, we investigate the interplay between "contact epistasis" (epistasis arising from physical interactions between mutated residues) and "ensemble epistasis" (epistasis that occurs when a mutation redistributes the conformational ensemble of a macromolecule, thus changing the effect of the second mutation). We argue that the two mechanisms can be distinguished in allosteric macromolecules by measuring epistasis at differing allosteric effector concentrations. Contact epistasis manifests as nonadditivity in the microscopic equilibrium constants describing the conformational ensemble. This epistatic effect is independent of allosteric effector concentration. Ensemble epistasis manifests as nonadditivity in thermodynamic observables-such as ligand binding-that are determined by the distribution of ensemble conformations. This epistatic effect strongly depends on allosteric effector concentration. Using this framework, we experimentally investigated the origins of epistasis in three pairwise mutant cycles introduced into the adenine riboswitch aptamer domain by measuring ligand binding as a function of allosteric effector concentration. We found evidence for both contact and ensemble epistasis in all cycles. Furthermore, we found that the two mechanisms of epistasis could interact with each other. For example, in one mutant cycle we observed 6 kcal/mol of contact epistasis in a microscopic equilibrium constant. In that same cycle, the maximum epistasis in ligand binding was only 1.5 kcal/mol: shifts in the ensemble masked the contribution of contact epistasis. Finally, our work yields simple heuristics for identifying contact and ensemble epistasis based on measurements of a biochemical observable as a function of allosteric effector concentration.

Topiary: Pruning the manual labor from ancestral sequence reconstruction.

Ancestral sequence reconstruction (ASR) is a powerful tool to study the evolution of proteins and thus gain deep insight into the relationships among protein sequence, structure, and function. A major barrier to its broad use is the complexity of the task: it requires multiple software packages, complex file manipulations, and expert phylogenetic knowledge. Here we introduce topiary, a software pipeline that aims to overcome this barrier. To use topiary, users prepare a spreadsheet with a handful of sequences. Topiary then: (1) Infers the taxonomic scope for the ASR study and finds relevant sequences by BLAST; (2) Does taxonomically informed sequence quality control and redundancy reduction; (3) Constructs a multiple sequence alignment; (4) Generates a maximum-likelihood gene tree; (5) Reconciles the gene tree to the species tree; (6) Reconstructs ancestral amino acid sequences; and (7) Determines branch supports. The pipeline returns annotated evolutionary trees, spreadsheets with sequences, and graphical summaries of ancestor quality. This is achieved by integrating modern phylogenetics software (Muscle5, RAxML-NG, GeneRax, and PastML) with online databases (NCBI and the Open Tree of Life). In this paper, we introduce non-expert readers to the steps required for ASR, describe the specific design choices made in topiary, provide a detailed protocol for users, and then validate the pipeline using datasets from a broad collection of protein families. Topiary is freely available for download: https://github.com/harmslab/topiary.

Higher-order interactions shape microbial interactions as microbial community complexity increases.

Non-pairwise interactions, or higher-order interactions (HOIs), in microbial communities have been described as significant drivers of emergent features in microbiomes. Yet, the re-organization of microbial interactions between pairwise cultures and larger communities remains largely unexplored from a molecular perspective but is central to our understanding and further manipulation of microbial communities. Here, we used a bottom-up approach to investigate microbial interaction mechanisms from pairwise cultures up to 4-species communities from a simple microbiome (Hafnia alvei, Geotrichum candidum, Pencillium camemberti and Escherichia coli). Specifically, we characterized the interaction landscape for each species combination involving E. coli by identifying E. coli's interaction-associated mutants using an RB-TnSeq-based interaction assay. We observed a deep reorganization of the interaction-associated mutants, with very few 2-species interactions conserved all the way up to a 4-species community and the emergence of multiple HOIs. We further used a quantitative genetics strategy to decipher how 2-species interactions were quantitatively conserved in higher community compositions. Epistasis-based analysis revealed that, of the interactions that are conserved at all levels of complexity, 82% follow an additive pattern. Altogether, we demonstrate the complex architecture of microbial interactions even within a simple microbiome, and provide a mechanistic and molecular explanation of HOIs.

Evolution avoids a pathological stabilizing interaction in the immune protein S100A9.

Stability constrains evolution. While much is known about constraints on destabilizing mutations, less is known about the constraints on stabilizing mutations. We recently identified a mutation in the innate immune protein S100A9 that provides insight into such constraints. When introduced into human S100A9, M63F simultaneously increases the stability of the protein and disrupts its natural ability to activate Toll-like receptor 4. Using chemical denaturation, we found that M63F stabilizes a calcium-bound conformation of hS100A9. We then used NMR to solve the structure of the mutant protein, revealing that the mutation distorts the hydrophobic binding surface of hS100A9, explaining its deleterious effect on function. Hydrogen-deuterium exchange (HDX) experiments revealed stabilization of the region around M63F in the structure, notably Phe37. In the structure of the M63F mutant, the Phe37 and Phe63 sidechains are in contact, plausibly forming an edge-face π-stack. Mutating Phe37 to Leu abolished the stabilizing effect of M63F as probed by both chemical denaturation and HDX. It also restored the biological activity of S100A9 disrupted by M63F. These findings reveal that Phe63 creates a molecular staple with Phe37 that stabilizes a nonfunctional conformation of the protein, thus disrupting function. Using a bioinformatic analysis, we found that S100A9 proteins from different organisms rarely have Phe at both positions 37 and 63, suggesting that avoiding a pathological stabilizing interaction indeed constrains S100A9 evolution. This work highlights an important evolutionary constraint on stabilizing mutations, namely, that they must avoid inappropriately stabilizing nonfunctional protein conformations.

Probing the origin of prion protein misfolding via reconstruction of ancestral proteins.

Prion diseases are fatal neurodegenerative diseases caused by pathogenic misfolding of the prion protein, PrP. They are transmissible between hosts, and sometimes between different species, as with transmission of bovine spongiform encephalopathy to humans. Although PrP is found in a wide range of vertebrates, prion diseases are seen only in certain mammals, suggesting that infectious misfolding was a recent evolutionary development. To explore when PrP acquired the ability to misfold infectiously, we reconstructed the sequences of ancestral versions of PrP from the last common primate, primate-rodent, artiodactyl, placental, bird, and amniote. Recombinant ancestral PrPs were then tested for their ability to form ß-sheet aggregates, either spontaneously or when seeded with infectious prion strains from human, cervid, or rodent species. The ability to aggregate developed after the oldest ancestor (last common amniote), and aggregation capabilities diverged along evolutionary pathways consistent with modern-day susceptibilities. Ancestral bird PrP could not be seeded with modern-day prions, just as modern-day birds are resistant to prion disease. Computational modeling of structures suggested that differences in helix 2 could account for the resistance of ancestral bird PrP to seeding. Interestingly, ancestral primate PrP could be converted by all prion seeds, including both human and cervid prions, raising the possibility that species descended from an ancestral primate have retained the susceptibility to conversion by cervid prions. More generally, the results suggest that susceptibility to prion disease emerged prior to ~100 million years ago, with placental mammals possibly being generally susceptible to disease.

Ensemble epistasis: thermodynamic origins of nonadditivity between mutations.

Epistasis-when mutations combine nonadditively-is a profoundly important aspect of biology. It is often difficult to understand its mechanistic origins. Here, we show that epistasis can arise from the thermodynamic ensemble, or the set of interchanging conformations a protein adopts. Ensemble epistasis occurs because mutations can have different effects on different conformations of the same protein, leading to nonadditive effects on its average, observable properties. Using a simple analytical model, we found that ensemble epistasis arises when two conditions are met: (1) a protein populates at least three conformations and (2) mutations have differential effects on at least two conformations. To explore the relative magnitude of ensemble epistasis, we performed a virtual deep-mutational scan of the allosteric Ca2+ signaling protein S100A4. We found that 47% of mutation pairs exhibited ensemble epistasis with a magnitude on the order of thermal fluctuations. We observed many forms of epistasis: magnitude, sign, and reciprocal sign epistasis. The same mutation pair could even exhibit different forms of epistasis under different environmental conditions. The ubiquity of thermodynamic ensembles in biology and the pervasiveness of ensemble epistasis in our dataset suggests that it may be a common mechanism of epistasis in proteins and other macromolecules.

A conserved folding nucleus sculpts the free energy landscape of bacterial and archaeal orthologs from a divergent TIM barrel family.

The amino acid sequences of proteins have evolved over billions of years, preserving their structures and functions while responding to evolutionary forces. Are there conserved sequence and structural elements that preserve the protein folding mechanisms? The functionally diverse and ancient (ßα)1-8 TIM barrel motif may answer this question. We mapped the complex six-state folding free energy surface of a ∼3.6 billion y old, bacterial indole-3-glycerol phosphate synthase (IGPS) TIM barrel enzyme by equilibrium and kinetic hydrogen-deuterium exchange mass spectrometry (HDX-MS). HDX-MS on the intact protein reported exchange in the native basin and the presence of two thermodynamically distinct on- and off-pathway intermediates in slow but dynamic equilibrium with each other. Proteolysis revealed protection in a small (α1ß2) and a large cluster (ß5α5ß6α6ß7) and that these clusters form cores of stability in Ia and Ibp The strongest protection in both states resides in ß4α4 with the highest density of branched aliphatic side chain contacts in the folded structure. Similar correlations were observed previously for an evolutionarily distinct archaeal IGPS, emphasizing a key role for hydrophobicity in stabilizing common high-energy folding intermediates. A bioinformatics analysis of IGPS sequences from the three superkingdoms revealed an exceedingly high hydrophobicity and surprising α-helix propensity for ß4, preceded by a highly conserved ßα-hairpin clamp that links ß3 and ß4. The conservation of the folding mechanisms for archaeal and bacterial IGPS proteins reflects the conservation of key elements of sequence and structure that first appeared in the last universal common ancestor of these ancient proteins.

Were Ancestral Proteins Less Specific?

Some have hypothesized that ancestral proteins were, on average, less specific than their descendants. If true, this would provide a universal axis along which to organize protein evolution and suggests that reconstructed ancestral proteins may be uniquely powerful tools for protein engineering. Ancestral sequence reconstruction studies are one line of evidence used to support this hypothesis. Previously, we performed such a study, investigating the evolution of peptide-binding specificity for the paralogs S100A5 and S100A6. The modern proteins appeared more specific than their last common ancestor (ancA5/A6), as each paralog bound a subset of the peptides bound by ancA5/A6. In this study, we revisit this transition, using quantitative phage display to measure the interactions of 30,533 random peptides with human S100A5, S100A6, and ancA5/A6. This unbiased screen reveals a different picture. While S100A5 and S100A6 do indeed bind to a subset of the peptides recognized by ancA5/A6, they also acquired new peptide partners outside of the set recognized by ancA5/A6. Our previous work showed that ancA5/A6 had lower specificity than its descendants when measured against biological targets; our new work shows that ancA5/A6 has similar specificity to the modern proteins when measured against a random set of peptide targets. This demonstrates that altered biological specificity does not necessarily indicate altered intrinsic specificity, and sounds a cautionary note for using ancestral reconstruction studies with biological targets as a means to infer global evolutionary trends in specificity.

Evolutionary Conservation of Structural and Functional Coupling between the BRM AT-Hook and Bromodomain.

The BAF chromatin remodeling complex is critical for genome regulation. The central ATPase of BAF is either BRM or BRG1, both of which contain a C-terminal bromodomain, known to associate with acetylated lysines. We have recently demonstrated that in addition to acetyl-lysine binding, the BRG1/BRM bromodomain can associate with DNA through a lysine/arginine rich patch that is adjacent to the acetyl-lysine binding pocket. Flanking the bromodomain is an AT-hook separated by a short, proline-rich linker. We previously found that the AT-hook and bromodomain can associate with DNA in a multivalent manner. Here, we investigate the conservation of this composite module and find that the AT-hook, linker, and lysine/arginine rich bromodomain patch are ancient, conserved over ~1 billion years. We utilize extensive mutagenesis, NMR spectroscopy, and fluorescence anisotropy to dissect the contribution of each of these conserved elements in association of this module with DNA. Our results reveal a structural and functional coupling of the AT-hook and bromodomain mediated by the linker. The lysine/arginine rich patch on the bromodomain and the conserved elements of the AT-hook are critical for robust affinity for DNA, while the conserved elements of the linker are dispensable for overall DNA affinity but critical for maintaining the relative conformation of the AT-hook and bromodomain in binding to DNA. This supports that the coupled action of the AT-hook and bromodomain are important for BAF activity.

Exploring the Evolutionary History of Kinetic Stability in the α-Lytic Protease Family.

In addition to encoding the tertiary fold and stability, the primary sequence of a protein encodes the folding trajectory and kinetic barriers that determine the speed of folding. How these kinetic barriers are encoded is not well understood. Here, we use evolutionary sequence variation in the α-lytic protease (αLP) protein family to probe the relationship between sequence and energy landscape. αLP has an unusual energy landscape: the native state of αLP is not the most thermodynamically favored conformation and, instead, remains folded due to a large kinetic barrier preventing unfolding. To fold, αLP utilizes an N-terminal pro region similar in size to the protease itself that functions as a folding catalyst. Once folded, the pro region is removed, and the native state does not unfold on a biologically relevant time scale. Without the pro region, αLP folds on the order of millennia. A phylogenetic search uncovers αLP homologs with a wide range of pro region sizes, including some with no pro region at all. In the resulting phylogenetic tree, these homologs cluster by pro region size. By studying homologs naturally lacking a pro region, we demonstrate they can be thermodynamically stable, fold much faster than αLP, yet retain the same fold as αLP. Key amino acids thought to contribute to αLP's extreme kinetic stability are lost in these homologs, supporting their role in kinetic stability. This study highlights how the entire energy landscape plays an important role in determining the evolutionary pressures on the protein sequence.

Identification and Characterization of Zebrafish Tlr4 Coreceptor Md-2.

The zebrafish (Danio rerio) is a powerful model organism for studies of the innate immune system. One apparent difference between human and zebrafish innate immunity is the cellular machinery for LPS sensing. In amniotes, the protein complex formed by TLR4 and myeloid differentiation factor 2 (Tlr4/Md-2) recognizes the bacterial molecule LPS and triggers an inflammatory response. It is believed that zebrafish have neither Md-2 nor Tlr4; Md-2 has not been identified outside of amniotes, whereas the zebrafish tlr4 genes appear to be paralogs, not orthologs, of amniote TLR4s We revisited these conclusions. We identified a zebrafish gene encoding Md-2, ly96 Using single-cell RNA sequencing, we found that ly96 is transcribed in cells that also transcribe genes diagnostic for innate immune cells, including the zebrafish tlr4-like genes. In larval zebrafish, ly96 is expressed in a small number of macrophage-like cells. In a functional assay, zebrafish Md-2 and Tlr4ba form a complex that activates NF-κB signaling in response to LPS. In larval zebrafish ly96 loss-of-function mutations perturbed LPS-induced cytokine production but gave little protection against LPS toxicity. Finally, by analyzing the genomic context of tlr4 genes in 11 jawed vertebrates, we found that tlr4 arose prior to the divergence of teleosts and tetrapods. Thus, an LPS-sensitive Tlr4/Md-2 complex is likely an ancestral feature shared by mammals and zebrafish, rather than a de novo invention on the tetrapod lineage. We hypothesize that zebrafish retain an ancestral, low-sensitivity Tlr4/Md-2 complex that confers LPS responsiveness to a specific subset of innate immune cells.

Inferring a complete genotype-phenotype map from a small number of measured phenotypes.

Understanding evolution requires detailed knowledge of genotype-phenotype maps; however, it can be a herculean task to measure every phenotype in a combinatorial map. We have developed a computational strategy to predict the missing phenotypes from an incomplete, combinatorial genotype-phenotype map. As a test case, we used an incomplete genotype-phenotype dataset previously generated for the malaria parasite's 'chloroquine resistance transporter' (PfCRT). Wild-type PfCRT (PfCRT3D7) lacks significant chloroquine (CQ) transport activity, but the introduction of the eight mutations present in the 'Dd2' isoform of PfCRT (PfCRTDd2) enables the protein to transport CQ away from its site of antimalarial action. This gain of a transport function imparts CQ resistance to the parasite. A combinatorial map between PfCRT3D7 and PfCRTDd2 consists of 256 genotypes, of which only 52 have had their CQ transport activities measured through expression in the Xenopus laevis oocyte. We trained a statistical model with these 52 measurements to infer the CQ transport activity for the remaining 204 combinatorial genotypes between PfCRT3D7 and PfCRTDd2. Our best-performing model incorporated a binary classifier, a nonlinear scale, and additive effects for each mutation. The addition of specific pairwise- and high-order-epistatic coefficients decreased the predictive power of the model. We evaluated our predictions by experimentally measuring the CQ transport activities of 24 additional PfCRT genotypes. The R2 value between our predicted and newly-measured phenotypes was 0.90. We then used the model to probe the accessibility of evolutionary trajectories through the map. Approximately 1% of the possible trajectories between PfCRT3D7 and PfCRTDd2 are accessible; however, none of the trajectories entailed eight successive increases in CQ transport activity. These results demonstrate that phenotypes can be inferred with known uncertainty from a partial genotype-phenotype dataset. We also validated our approach against a collection of previously published genotype-phenotype maps. The model therefore appears general and should be applicable to a large number of genotype-phenotype maps.

Learning peptide recognition rules for a low-specificity protein.

Many proteins interact with short linear regions of target proteins. For some proteins, however, it is difficult to identify a well-defined sequence motif that defines its target peptides. To overcome this difficulty, we used supervised machine learning to train a model that treats each peptide as a collection of easily-calculated biochemical features rather than as an amino acid sequence. As a test case, we dissected the peptide-recognition rules for human S100A5 (hA5), a low-specificity calcium binding protein. We trained a Random Forest model against a recently released, high-throughput phage display dataset collected for hA5. The model identifies hydrophobicity and shape complementarity, rather than polar contacts, as the primary determinants of peptide binding specificity in hA5. We tested this hypothesis by solving a crystal structure of hA5 and through computational docking studies of diverse peptides onto hA5. These structural studies revealed that peptides exhibit multiple binding modes at the hA5 peptide interface-all of which have few polar contacts with hA5. Finally, we used our trained model to predict new, plausible binding targets in the human proteome. This revealed a fragment of the protein α-1-syntrophin that binds to hA5. Our work helps better understand the biochemistry and biology of hA5, as well as demonstrating how high-throughput experiments coupled with machine learning of biochemical features can reveal the determinants of binding specificity in low-specificity proteins.

Evolution of multifunctionality through a pleiotropic substitution in the innate immune protein S100A9.

Multifunctional proteins are evolutionary puzzles: how do proteins evolve to satisfy multiple functional constraints? S100A9 is one such multifunctional protein. It potently amplifies inflammation via Toll-like receptor four and is antimicrobial as part of a heterocomplex with S100A8. These two functions are seemingly regulated by proteolysis: S100A9 is readily degraded, while S100A8/S100A9 is resistant. We take an evolutionary biochemical approach to show that S100A9 evolved both functions and lost proteolytic resistance from a weakly proinflammatory, proteolytically resistant amniote ancestor. We identify a historical substitution that has pleiotropic effects on S100A9 proinflammatory activity and proteolytic resistance but has little effect on S100A8/S100A9 antimicrobial activity. We thus propose that mammals evolved S100A8/S100A9 antimicrobial and S100A9 proinflammatory activities concomitantly with a proteolytic 'timer' to selectively regulate S100A9. This highlights how the same mutation can have pleiotropic effects on one functional state of a protein but not another, thus facilitating the evolution of multifunctionality.

A single protein sometimes does multiple jobs. For instance, our immune system uses a small number of multipurpose proteins to respond quickly to a large number of threats. One example is the protein S100A9. It acts as an antimicrobial by preventing microbes from getting the nutrients they need, while also stimulating inflammation by inducing the release of molecules that recruit white blood cells. S100A9, like all proteins, is made up of a chain of small building blocks. These building blocks interact with each other and with other molecules in the environment. The sequence of the building blocks thus determines what jobs the protein can do. Therefore, a single change to the sequence of building blocks can have a dramatic effect: one change might render the protein faulty, while another change might allow it to do a new job. Proteins face similar challenges humans do when trying to do several things at once. A person driving a car while using their phone will not do either task well. Likewise, a protein that does two jobs faces challenges a single-purpose protein does not. Harman et al. were interested in how S100A9 was able to evolve and maintain its dual functionality, despite this potential problem. They started by asking when S100A9 acquired its two purposes. They measured the antimicrobial and inflammatory activity of S100A9 proteins from humans, mice and opossums. The activities of S100A9 in these species was similar, suggesting that S100A9 acquired its different jobs in the ancestor of mammals, some 160 million years ago. Next, Harman et al. computationally reconstructed ancestral forms of S100A9 by comparing hundreds of similar proteins and building an evolutionary tree. They then measured the antimicrobial and inflammatory activity of these ancestral proteins. By comparing the last ancestor that did not have these activities to the first ancestor that did, they identified the sequence changes that gave S100A9 its dual activity. Importantly, these changes are located in separate regions of the protein, meaning they could occur independently, without affecting each other. Further, the same sequence change that converted S100A9 into an inflammatory signal also introduced a mechanism to regulate this activity. The results suggest that a small number of sequence changes ­ or even a single change ­ can make a protein more versatile. This means that evolving multipurpose proteins may not be as difficult as is often thought.