Accelerating Bayesian inference of dependency between mixed-type biological traits.

Inferring dependencies between mixed-type biological traits while accounting for evolutionary relationships between specimens is of great scientific interest yet remains infeasible when trait and specimen counts grow large. The state-of-the-art approach uses a phylogenetic multivariate probit model to accommodate binary and continuous traits via a latent variable framework, and utilizes an efficient bouncy particle sampler (BPS) to tackle the computational bottleneck-integrating many latent variables from a high-dimensional truncated normal distribution. This approach breaks down as the number of specimens grows and fails to reliably characterize conditional dependencies between traits. Here, we propose an inference pipeline for phylogenetic probit models that greatly outperforms BPS. The novelty lies in 1) a combination of the recent Zigzag Hamiltonian Monte Carlo (Zigzag-HMC) with linear-time gradient evaluations and 2) a joint sampling scheme for highly correlated latent variables and correlation matrix elements. In an application exploring HIV-1 evolution from 535 viruses, the inference requires joint sampling from an 11,235-dimensional truncated normal and a 24-dimensional covariance matrix. Our method yields a 5-fold speedup compared to BPS and makes it possible to learn partial correlations between candidate viral mutations and virulence. Computational speedup now enables us to tackle even larger problems: we study the evolution of influenza H1N1 glycosylations on around 900 viruses. For broader applicability, we extend the phylogenetic probit model to incorporate categorical traits, and demonstrate its use to study Aquilegia flower and pollinator co-evolution.

Restriction Endonuclease-Based Modification-Dependent Enrichment (REMoDE) of DNA for Metagenomic Sequencing.

Metagenomic sequencing is a swift and powerful tool to ascertain the presence of an organism of interest in a sample. However, sequencing coverage of the organism of interest can be insufficient due to an inundation of reads from irrelevant organisms in the sample. Here, we report a nuclease-based approach to rapidly enrich for DNA from certain organisms, including enterobacteria, based on their differential endogenous modification patterns. We exploit the ability of taxon-specific methylated motifs to resist the action of cognate methylation-sensitive restriction endonucleases that thereby digest unwanted, unmethylated DNA. Subsequently, we use a distributive exonuclease or electrophoretic separation to deplete or exclude the digested fragments, thus enriching for undigested DNA from the organism of interest. As a proof of concept, we apply this method to enrich for the enterobacteria Escherichia coli and Salmonella enterica by 11- to 142-fold from mock metagenomic samples and validate this approach as a versatile means to enrich for genomes of interest in metagenomic samples. IMPORTANCE Pathogens that contaminate the food supply or spread through other means can cause outbreaks that bring devastating repercussions to the health of a populace. Investigations to trace the source of these outbreaks are initiated rapidly but can be drawn out due to the labored methods of pathogen isolation. Metagenomic sequencing can alleviate this hurdle but is often insufficiently sensitive. The approach and implementations detailed here provide a rapid means to enrich for many pathogens involved in foodborne outbreaks, thereby improving the utility of metagenomic sequencing as a tool in outbreak investigations. Additionally, this approach provides a means to broadly enrich for otherwise minute levels of modified DNA, which may escape unnoticed in metagenomic samples.

Methylation-Induced Hypermutation in Natural Populations of Bacteria.

Methylation of DNA at the C-5 position of cytosine occurs in diverse organisms. This modification can increase the rate of C→T transitions at the methylated position. In Escherichia coli and related enteric bacteria, the inner C residues of the sequence CCWGG (W is A or T) are methylated by the Dcm enzyme. These sites are hot spots of mutation during rapid growth in the laboratory but not in nondividing cells, in which repair by the Vsr protein is effective. It has been suggested that hypermutation at these sites is a laboratory artifact and does not occur in nature. Many other methyltransferases, with a variety of specificities, can be found in bacteria, usually associated with restriction enzymes and confined to a subset of the population. Their methylation targets are also possible sites of hypermutation. Here, I show using whole-genome sequence data for thousands of isolates that there is indeed considerable hypermutation at Dcm sites in natural populations: their transition rate is approximately eight times the average. I also demonstrate hypermutability of targets of restriction-associated methyltransferases in several distantly related bacteria: methylation increases the transition rate by a factor ranging from 12 to 58. In addition, I demonstrate how patterns of hypermutability inferred from massive sequence data can be used to determine previously unknown methylation patterns and methyltransferase specificities.IMPORTANCE A common type of DNA modification, addition of a methyl group to cytosine (C) at carbon atom C-5, can greatly increase the rate of mutation of the C to a T. In mammals, methylation of CG sequences increases the rate of CG→TG mutations. It is unknown whether cytosine C-5 methylation increases the mutation rate in bacteria under natural conditions. I show that sites methylated by the Dcm enzyme exhibit an 8-fold increase in mutation rate in natural bacterial populations. I also show that modifications at other sites in various bacteria also increase the mutation rate, in some cases by a factor of forty or more. Finally, I demonstrate how this phenomenon can be used to infer sequence specificities of methylation enzymes.

A practical exact maximum compatibility algorithm for reconstruction of recent evolutionary history.

BACKGROUND: Maximum compatibility is a method of phylogenetic reconstruction that is seldom applied to molecular sequences. It may be ideal for certain applications, such as reconstructing phylogenies of closely-related bacteria on the basis of whole-genome sequencing. RESULTS: Here I present an algorithm that rapidly computes phylogenies according to a compatibility criterion. Although based on solutions to the maximum clique problem, this algorithm deals properly with ambiguities in the data. The algorithm is applied to bacterial data sets containing up to nearly 2000 genomes with several thousand variable nucleotide sites. Run times are several seconds or less. Computational experiments show that maximum compatibility is less sensitive than maximum parsimony to the inclusion of nucleotide data that, though derived from actual sequence reads, has been identified as likely to be misleading. CONCLUSIONS: Maximum compatibility is a useful tool for certain phylogenetic problems, such as inferring the relationships among closely-related bacteria from whole-genome sequence data. The algorithm presented here rapidly solves fairly large problems of this type, and provides robustness against misleading characters than can pollute large-scale sequencing data.

Sequential seasonal H1N1 influenza virus infections protect ferrets against novel 2009 H1N1 influenza virus.

Individuals <60 years of age had the lowest incidence of infection, with ~25% of these people having preexisting, cross-reactive antibodies to novel 2009 H1N1 influenza. Many people >60 years old also had preexisting antibodies to novel H1N1. These observations are puzzling because the seasonal H1N1 viruses circulating during the last 60 years were not antigenically similar to novel H1N1. We therefore hypothesized that a sequence of exposures to antigenically different seasonal H1N1 viruses can elicit an antibody response that protects against novel 2009 H1N1. Ferrets were preinfected with seasonal H1N1 viruses and assessed for cross-reactive antibodies to novel H1N1. Serum from infected ferrets was assayed for cross-reactivity to both seasonal and novel 2009 H1N1 strains. These results were compared to those of ferrets that were sequentially infected with H1N1 viruses isolated prior to 1957 or more-recently isolated viruses. Following seroconversion, ferrets were challenged with novel H1N1 influenza virus and assessed for viral titers in the nasal wash, morbidity, and mortality. There was no hemagglutination inhibition (HAI) cross-reactivity in ferrets infected with any single seasonal H1N1 influenza viruses, with limited protection to challenge. However, sequential H1N1 influenza infections reduced the incidence of disease and elicited cross-reactive antibodies to novel H1N1 isolates. The amount and duration of virus shedding and the frequency of transmission following novel H1N1 challenge were reduced. Exposure to multiple seasonal H1N1 influenza viruses, and not to any single H1N1 influenza virus, elicits a breadth of antibodies that neutralize novel H1N1 even though the host was never exposed to the novel H1N1 influenza viruses.

Comparative evolution of influenza A virus H1 and H3 head and stalk domains across host species.

IMPORTANCE: For decades, researchers have studied the rapid evolution of influenza A viruses for vaccine design and as a useful model system for the study of host/parasite evolution. By performing an exhaustive analysis of hemagglutinin protein (HA) sequences from 49 lineages independently evolving in birds, swine, canines, equines, and humans over the last century, our work uncovers surprising features of HA evolution. In particular, the canine H3 stalk, unlike human H3 and H1 stalk domains, is not evolving slowly, suggesting that evolution in the stalk domain is not universally constrained across all host species. Therefore, a broader multi-host perspective on HA evolution may be useful during the evaluation and design of stalk-targeted vaccine candidates.

Genomic perspectives on foodborne illness.

Whole-genome sequencing of bacterial pathogens is used by public health agencies to link cases of food poisoning caused by the same source of contamination. The vast majority of these appear to be sporadic cases associated with small contamination episodes and do not trigger investigations. We analyzed clusters of sequenced clinical isolates of Salmonella, Escherichia coli, Campylobacter, and Listeria that differ by only a small number of mutations to provide a new understanding of the underlying contamination episodes. These analyses provide new evidence that the youngest age groups have greater susceptibility to infection from Salmonella, Escherichia coli, and Campylobacter than older age groups. This age bias is weaker for the common Salmonella serovar Enteritidis than Salmonella in general. Analysis of these clusters reveals significant regional variations in relative frequencies of Salmonella serovars across the United States. A large fraction of the contamination episodes causing sickness appear to have long duration. For example, 50% of the Salmonella cases are in clusters that persist for almost three years. For all four pathogen species, the majority of the cases were part of genetic clusters with illnesses in multiple states and likely to be caused by contaminated commercially distributed foods. The vast majority of Salmonella cases among infants < 6 months of age appear to be caused by cross-contamination from foods consumed by older age groups or by environmental bacteria rather than infant formula contaminated at production sites.

T Residues Preceded by Runs of G Are Hotspots of T→G Mutation in Bacteria.

The rate of mutation varies among positions in a genome. Local sequence context can affect the rate and has different effects on different types of mutation. Here, I report an effect of local context that operates to some extent in all bacteria examined: the rate of T→G mutation is greatly increased by preceding runs of three or more G residues. The strength of the effect increases with the length of the run. In Salmonella, in which the effect is strongest, a G run of length three 3 increases the rate by a factor of ∼26, a run of length 4 increases it by almost a factor of 100, and runs of length 5 or more increase it by a factor of more than 400 on average. The effect is much stronger when the T is on the leading rather than the lagging strand of DNA replication. Several observations eliminate the possibility that this effect is an artifact of sequencing error.

Lineage-specific biology revealed by a finished genome assembly of the mouse.

The mouse (Mus musculus) is the premier animal model for understanding human disease and development. Here we show that a comprehensive understanding of mouse biology is only possible with the availability of a finished, high-quality genome assembly. The finished clone-based assembly of the mouse strain C57BL/6J reported here has over 175,000 fewer gaps and over 139 Mb more of novel sequence, compared with the earlier MGSCv3 draft genome assembly. In a comprehensive analysis of this revised genome sequence, we are now able to define 20,210 protein-coding genes, over a thousand more than predicted in the human genome (19,042 genes). In addition, we identified 439 long, non-protein-coding RNAs with evidence for transcribed orthologs in human. We analyzed the complex and repetitive landscape of 267 Mb of sequence that was missing or misassembled in the previously published assembly, and we provide insights into the reasons for its resistance to sequencing and assembly by whole-genome shotgun approaches. Duplicated regions within newly assembled sequence tend to be of more recent ancestry than duplicates in the published draft, correcting our initial understanding of recent evolution on the mouse lineage. These duplicates appear to be largely composed of sequence regions containing transposable elements and duplicated protein-coding genes; of these, some may be fixed in the mouse population, but at least 40% of segmentally duplicated sequences are copy number variable even among laboratory mouse strains. Mouse lineage-specific regions contain 3,767 genes drawn mainly from rapidly-changing gene families associated with reproductive functions. The finished mouse genome assembly, therefore, greatly improves our understanding of rodent-specific biology and allows the delineation of ancestral biological functions that are shared with human from derived functions that are not.

Recent Genetic Changes Affecting Enterohemorrhagic Escherichia coli Causing Recurrent Outbreaks.

Enterohemorrhagic E. coli (EHEC) is responsible for significant human illness, death, and economic loss. The main reservoir for EHEC is cattle, but plant-based foods are common vectors for human infection. Several outbreaks have been attributed to lettuce and leafy green vegetables grown in the Salinas and Santa Maria regions of California. Bacteria causing different outbreaks are mostly not close relatives, but one group of closely-related O157:H7 has caused several of them. This unusual pattern of recurrence may have some genetic basis. Here I use whole-genome sequences to reconstruct the genetic changes that occurred in the recent ancestry of this EHEC. In a short period of time corresponding to little genetic change, there were several changes to adhesion-related sequences, mainly adhesins. These changes may have greatly altered the adhesive properties of the bacteria. Possible consequences include increased persistence of cattle infections, more bacteria shed in cattle feces, and greater virulence in humans. Similar constellations of genetic change, which are detectable by current sequencing-based surveillance, may identify other bacteria that are particular threats to human health. In addition, the Santa Maria subclade carries a nonsense mutation affecting ArsR, a repressor of genes that confer resistance to arsenic and antimony. This suggests that the persistent source of Santa Maria contamination is located in an area with arsenic-contaminated groundwater, a problem in many parts of California. This inference may aid identification of the reservoir of EHEC, which would greatly aid mitigation efforts. IMPORTANCE Food-borne bacterial infections cause substantial illness and death. Understanding how bacteria contaminate food and cause disease is important for combating the problem. Closely-related E. coli, likely originating in cattle, have repeatedly caused outbreaks spread by vegetables grown in California. Such recurrence is atypical, and might have a genetic basis. The genetic changes that occurred in the recent ancestry of these E. coli can be reconstructed from their DNA sequences. Several mutations affect genes involved in bacterial adhesion. These might affect persistence of infection in cattle, quantity of bacteria in their feces, and human disease. They also suggest a way of detecting dangerous bacteria from their genome sequences. Furthermore, a subgroup carries a mutation affecting the regulation of genes conferring arsenic resistance. This suggests that the reservoir for contamination utilizes groundwater contaminated with arsenic, a problem in parts of California. This observation may be an aid to locating the persistent reservoir of contamination.

Origins and Evolution of Seasonal Human Coronaviruses.

Four seasonal human coronaviruses (sHCoVs) are endemic globally (229E, NL63, OC43, and HKU1), accounting for 5-30% of human respiratory infections. However, the epidemiology and evolution of these CoVs remain understudied due to their association with mild symptomatology. Using a multigene and complete genome analysis approach, we find the evolutionary histories of sHCoVs to be highly complex, owing to frequent recombination of CoVs including within and between sHCoVs, and uncertain, due to the under sampling of non-human viruses. The recombination rate was highest for 229E and OC43 whereas substitutions per recombination event were highest in NL63 and HKU1. Depending on the gene studied, OC43 may have ungulate, canine, or rabbit CoV ancestors. 229E may have origins in a bat, camel, or an unsampled intermediate host. HKU1 had the earliest common ancestor (1809-1899) but fell into two distinct clades (genotypes A and B), possibly representing two independent transmission events from murine-origin CoVs that appear to be a single introduction due to large gaps in the sampling of CoVs in animals. In fact, genotype B was genetically more diverse than all the other sHCoVs. Finally, we found shared amino acid substitutions in multiple proteins along the non-human to sHCoV host-jump branches. The complex evolution of CoVs and their frequent host switches could benefit from continued surveillance of CoVs across non-human hosts.

Highly expressed and slowly evolving proteins share compositional properties with thermophilic proteins.

The sequences of proteins encoded by a genome evolve at different rates. A correlate of a protein's evolutionary rate is its expression level: highly expressed proteins tend to evolve slowly. Some explanations of rate variation and the correlation between rate and expression predict that more slowly evolving and more highly expressed proteins have more favorable equilibrium constants for folding. Proteins from thermophiles generally have more stable folds than proteins from mesophiles, and it is known that there are systematic differences in amino acid content between thermophilic and mesophilic proteins. I examined whether there are analogous correlations of amino acid frequencies with evolutionary rate and expression level within genomes. In most of the organisms analyzed, there is a striking tendency for more slowly evolving proteins to be more thermophile-like in their amino acid compositions when adjustments are made for variation in GC content. More highly expressed proteins also tend to be more thermophile-like by the same criteria. These results suggest that part of the evolutionary rate variation among proteins is due to variation in the strength of selection for stability of the folded state. They also suggest that increasing strength of this selective force with expression level plays a role in the correlation between evolutionary rate and expression level.

Extreme C-to-A Hypermutation at a Site of Cytosine-N4 Methylation.

Methylation of cytosine in DNA at position C5 increases the rate of C→T mutations in bacteria and eukaryotes. Methylation at the N4 position, employed by some restriction-modification systems, is not known to increase the mutation rate. Here, I report that a Salmonella enterica Type III restriction-modification system that includes a cytosine-N4 methyltransferase causes an enormous increase in the rate of mutation of the methylated cytosines, which occur at the overlined C in the motif CACC̅GT Mutations consist mainly of C→A transversions, the rate of which is increased ∼500-fold by the restriction-modification system. The rate of C→T transitions is also increased and somewhat exceeds that at C5-methylated cytosines in Dcm sites. Two other Salmonella N4 methyltransferases investigated do not have such dramatic effects, although in one case there is a modest increase in C→A mutations along with an increase in C→T mutations. The sensitivity of the C→A rate to orientation with respect to both DNA replication and transcription is higher at hypermutable sites than at other cytosines, suggesting a fundamental mechanistic difference between hypermutation and ordinary mutation.IMPORTANCE Mutation produces the raw material for adaptive evolution but also imposes a burden because most mutations are deleterious. The rate of mutation at a particular site is affected by a variety of factors. In both prokaryotes and eukaryotes, methylation of C at the C5 position, a naturally occurring DNA modification, greatly increases the rate of C→T mutation. A distinct C modification that occurs in prokaryotes, methylation at N4, is not known to increase mutation rate. Here, I report that a bacterial restriction-modification system, found in some Salmonella bacteria, increases the rate of C→A mutation by a factor of 500 at sites that it methylates at N4. This rate increase is much greater than that caused by C5 methylation. Although fewer than 1 in 1,600 positions analyzed are methylation sites, over 10% of all mutations occur at these sites. Like other examples of extremely high mutation rate, whether naturally occurring or the result of laboratory mutation, this phenomenon may shed light on the mechanism of mutation in general.

Selection-Driven Gene Inactivation in Salmonella.

Bacterial genes are sometimes found to be inactivated by mutation. This inactivation may be observable simply because selection for function is intermittent or too weak to eliminate inactive alleles quickly. Here, I investigate cases in Salmonella enterica where inactivation is instead positively selected. These are identified by a rate of introduction of premature stop codons to a gene that is higher than expected under selective neutrality, as assessed by comparison to the rate of synonymous changes. I identify 84 genes that meet this criterion at a 10% false discovery rate. Many of these genes are involved in virulence, motility and chemotaxis, biofilm formation, and resistance to antibiotics or other toxic substances. It is hypothesized that most of these genes are subject to an ongoing process in which inactivation is favored under rare conditions, but the inactivated allele is deleterious under most other conditions and is subsequently driven to extinction by purifying selection.

Selection, subdivision and extinction and recolonization.

In a subdivided population, the interaction between natural selection and stochastic change in allele frequency is affected by the occurrence of local extinction and subsequent recolonization. The relative importance of selection can be diminished by this additional source of stochastic change in allele frequency. Results are presented for subdivided populations with extinction and recolonization where there is more than one founding allele after extinction, where these may tend to come from the same source deme, where the number of founding alleles is variable or the founders make unequal contributions, and where there is dominance for fitness or local frequency dependence. The behavior of a selected allele in a subdivided population is in all these situations approximately the same as that of an allele with different selection parameters in an unstructured population with a different size. The magnitude of the quantity N(e)s(e), which determines fixation probability in the case of genic selection, is always decreased by extinction and recolonization, so that deleterious alleles are more likely to fix and advantageous alleles less likely to do so. The importance of dominance or frequency dependence is also altered by extinction and recolonization. Computer simulations confirm that the theoretical predictions of both fixation probabilities and mean times to fixation are good approximations.

Selection in a subdivided population with dominance or local frequency dependence.

The interplay between population structure and natural selection is an area of great interest. It is known that certain types of population subdivision do not alter fixation probabilities of selected alleles under genic, frequency-independent selection. In the presence of dominance for fitness or frequency-dependent selection these same types of subdivision can have large effects on fixation probabilities. For example, the barrier to fixation of a fitter allele due to underdominance is reduced by subdivision. Analytic results presented here relate a subdivided population that conforms to a finite island model to an approximately equivalent panmictic population. The size of this equivalent population is different from (larger than) the actual size of the subdivided population. Selection parameters are also different in the hypothetical equivalent population. As expected, the degree of dominance is lower in the equivalent population. The results are not limited to dominance but cover any form of polynomial frequency dependence.

Selection in a subdivided population with local extinction and recolonization.

In a subdivided population, local extinction and subsequent recolonization affect the fate of alleles. Of particular interest is the interaction of this force with natural selection. The effect of selection can be weakened by this additional source of stochastic change in allele frequency. The behavior of a selected allele in such a population is shown to be equivalent to that of an allele with a different selection coefficient in an unstructured population with a different size. This equivalence allows use of established results for panmictic populations to predict such quantities as fixation probabilities and mean times to fixation. The magnitude of the quantity N(e)s(e), which determines fixation probability, is decreased by extinction and recolonization. Thus deleterious alleles are more likely to fix, and advantageous alleles less likely to do so, in the presence of extinction and recolonization. Computer simulations confirm that the theoretical predictions of both fixation probabilities and mean times to fixation are good approximations.

A diffusion approximation for selection and drift in a subdivided population.

The population-genetic consequences of population structure are of great interest and have been studied extensively. An area of particular interest is the interaction among population structure, natural selection, and genetic drift. At first glance, different results in this area give very different impressions of the effect of population subdivision on effective population size (N(e)), suggesting that no single value of N(e) can completely characterize a structured population. Results presented here show that a population conforming to Wright's island model of subdivision with genic selection can be related to an idealized panmictic population (a Wright-Fisher population). This equivalent panmictic population has a larger size than the actual population; i.e., N(e) is larger than the actual population size, as expected from many results for this type of population structure. The selection coefficient in the equivalent panmictic population, referred to here as the effective selection coefficient (s(e)), is smaller than the actual selection coefficient (s). This explains how the fixation probability of a selected allele can be unaffected by population subdivision despite the fact that subdivision increases N(e), for the product N(e)s(e) is not altered by subdivision.

Deleterious mutation and the evolution of eusociality.

Certain arguments concerning the evolution of eusociality form a classic example of the application of the principles of kin selection. These arguments center on the different degrees of relatedness of potential beneficiaries of an individual's efforts, for example a female's higher relatedness to her sisters than to her daughters in a haplodiploid system. This type of reasoning is insufficicnt to account for the evolution and maintainence of sexual reproduction, because parthenogenic females produce offspring that are more closely related to them than are offspring produced sexually. Among the forces invoked to explain sexual reproduction is deleterious mutation. This factor can be shown to favor eusociality as well, because siblings produced by helping carry fewer deleterious alleles on average than would offspring. The strength of this effect depends on the genomewide deleterious mutation rate, U, and on the selection coefficient, s, associated with deleterious alleles. For small s, the effect depends approximately on the product Us. This phenomenon illustrates that an assumption implicit in some analyses-that the relatedness of an individual to an actor is all that matters to its value to that actor-can fail for the evolution of eusociality as it does for the evolution of sex.

Why does a protein's evolutionary rate vary over time?

The sequences of different proteins evolve at different rates. The relative evolutionary rate (ER) of a single protein also changes over evolutionary time. The cause of this ER fluctuation remains uncertain, and study of this phenomenon may shed light on protein evolution more broadly. We have characterized ER fluctuation in mammals and Drosophila. We found little correlation between the amount of rate variation observed for a protein and such factors as its expression level or phylogenetic distribution. Perhaps more surprisingly, we found little correlation between our measure of rate variation and ER itself. We also investigated the extent to which the ERs of different domains of a protein vary independently. We found that rates of different domains do tend to vary together. In fact, rates at positions in different domains are coupled just as strongly as rates at equally distant positions in the same domain. These findings provide clues to the protein evolutionary process.