ResMiCo: Increasing the quality of metagenome-assembled genomes with deep learning.

The number of published metagenome assemblies is rapidly growing due to advances in sequencing technologies. However, sequencing errors, variable coverage, repetitive genomic regions, and other factors can produce misassemblies, which are challenging to detect for taxonomically novel genomic data. Assembly errors can affect all downstream analyses of the assemblies. Accuracy for the state of the art in reference-free misassembly prediction does not exceed an AUPRC of 0.57, and it is not clear how well these models generalize to real-world data. Here, we present the Residual neural network for Misassembled Contig identification (ResMiCo), a deep learning approach for reference-free identification of misassembled contigs. To develop ResMiCo, we first generated a training dataset of unprecedented size and complexity that can be used for further benchmarking and developments in the field. Through rigorous validation, we show that ResMiCo is substantially more accurate than the state of the art, and the model is robust to novel taxonomic diversity and varying assembly methods. ResMiCo estimated 7% misassembled contigs per metagenome across multiple real-world datasets. We demonstrate how ResMiCo can be used to optimize metagenome assembly hyperparameters to improve accuracy, instead of optimizing solely for contiguity. The accuracy, robustness, and ease-of-use of ResMiCo make the tool suitable for general quality control of metagenome assemblies and assembly methodology optimization.

Multisubstrate DNA stable isotope probing reveals guild structure of bacteria that mediate soil carbon cycling.

Soil microorganisms determine the fate of soil organic matter (SOM), and their activities compose a major component of the global carbon (C) cycle. We employed a multisubstrate, DNA-stable isotope probing experiment to track bacterial assimilation of C derived from distinct sources that varied in bioavailability. This approach allowed us to measure microbial contributions to SOM processing by measuring the C assimilation dynamics of diverse microorganisms as they interacted within soil. We identified and tracked 1,286 bacterial taxa that assimilated 13C in an agricultural soil over a period of 48 d. Overall 13C-assimilation dynamics of bacterial taxa, defined by the source and timing of the 13C they assimilated, exhibited low phylogenetic conservation. We identified bacterial guilds composed of taxa that had similar 13C assimilation dynamics. We show that C-source bioavailability explained significant variation in both C mineralization dynamics and guild structure, and that the growth dynamics of bacterial guilds differed significantly in response to C addition. We also demonstrate that the guild structure explains significant variation in the biogeographical distribution of bacteria at continental and global scales. These results suggest that an understanding of in situ growth dynamics is essential for understanding microbial contributions to soil C cycling. We interpret these findings in the context of bacterial life history strategies and their relationship to terrestrial C cycling.

Interpreting tree ensemble machine learning models with endoR.

Tree ensemble machine learning models are increasingly used in microbiome science as they are compatible with the compositional, high-dimensional, and sparse structure of sequence-based microbiome data. While such models are often good at predicting phenotypes based on microbiome data, they only yield limited insights into how microbial taxa may be associated. We developed endoR, a method to interpret tree ensemble models. First, endoR simplifies the fitted model into a decision ensemble. Then, it extracts information on the importance of individual features and their pairwise interactions, displaying them as an interpretable network. Both the endoR network and importance scores provide insights into how features, and interactions between them, contribute to the predictive performance of the fitted model. Adjustable regularization and bootstrapping help reduce the complexity and ensure that only essential parts of the model are retained. We assessed endoR on both simulated and real metagenomic data. We found endoR to have comparable accuracy to other common approaches while easing and enhancing model interpretation. Using endoR, we also confirmed published results on gut microbiome differences between cirrhotic and healthy individuals. Finally, we utilized endoR to explore associations between human gut methanogens and microbiome components. Indeed, these hydrogen consumers are expected to interact with fermenting bacteria in a complex syntrophic network. Specifically, we analyzed a global metagenome dataset of 2203 individuals and confirmed the previously reported association between Methanobacteriaceae and Christensenellales. Additionally, we observed that Methanobacteriaceae are associated with a network of hydrogen-producing bacteria. Our method accurately captures how tree ensembles use features and interactions between them to predict a response. As demonstrated by our applications, the resultant visualizations and summary outputs facilitate model interpretation and enable the generation of novel hypotheses about complex systems.

Incorporating genome-based phylogeny and functional similarity into diversity assessments helps to resolve a global collection of human gut metagenomes.

Tree-based diversity measures incorporate phylogenetic or functional relatedness into comparisons of microbial communities. This can improve the identification of explanatory factors compared to tree-agnostic diversity measures. However, applying tree-based diversity measures to metagenome data is more challenging than for single-locus sequencing (e.g. 16S rRNA gene). Utilizing the Genome Taxonomy Database for species-level metagenome profiling allows for functional diversity measures based on genomic content or traits inferred from it. Still, it is unclear how metagenome-based assessments of microbiome diversity benefit from incorporating phylogeny or function into measures of diversity. We assessed this by measuring phylogeny-based, function-based and tree-agnostic diversity measures from a large, global collection of human gut metagenomes composed of 30 studies and 2943 samples. We found tree-based measures to explain phenotypic variation (e.g. westernization, disease status and gender) better or equivalent to tree-agnostic measures. Ecophylogenetic and functional diversity measures provided unique insight into how microbiome diversity was partitioned by phenotype. Tree-based measures greatly improved machine learning model performance for predicting westernization, disease status and gender, relative to models trained solely on tree-agnostic measures. Our findings illustrate the usefulness of tree- and function-based measures for metagenomic assessments of microbial diversity, which is a fundamental component of microbiome science.

Bacterial community dynamics explain carbon mineralization and assimilation in soils of different land-use history.

Soil dwelling microorganisms are key players in the terrestrial carbon cycle, driving both the degradation and stabilization of soil organic matter. Bacterial community structure and function vary with respect to land use; yet the ecological drivers of this variation remain poorly described and difficult to predict. We conducted a multi-substrate DNA-stable isotope probing experiment across cropland, old-field, and forest habitats to link carbon mineralization dynamics with the dynamics of bacterial growth and carbon assimilation. We tracked the movement of 13 C derived from five distinct carbon sources as it was assimilated into bacterial DNA over time. We show that carbon mineralization, community composition, and carbon assimilation dynamics all differed with respect to land use. We also show that microbial community dynamics affect carbon assimilation dynamics and are associated with soil DNA content. Soil DNA yield is easy to measure and may be useful in predicting microbial community dynamics linked to soil carbon cycling. Soil dwelling microorganisms are key players in the terrestrial carbon cycle, driving both the degradation and stabilization of soil organic matter. Microbial communities vary with respect to land use, but we still have an incomplete understanding of how variation in community structure links to variation in community function. DNA stable isotope probing (DNA-SIP) is a high-resolution method that can identify specific microbial taxa that assimilate carbon in situ. We conducted a large-scale multi-substrate DNA-SIP experiment to explore differences in bacterial activity across land-use regimes. We show that microbial community dynamics vary with land use, that these dynamics are linked to soil carbon cycling, and that they are associated with easily measured soil properties.

Tracing Carbon Metabolism with Stable Isotope Metabolomics Reveals the Legacy of Diverse Carbon Sources in Soil.

Tracking the metabolic activity of whole soil communities can improve our understanding of the transformation and fate of carbon in soils. We used stable isotope metabolomics to trace 13C from nine labeled carbon sources into the water-soluble metabolite pool of an agricultural soil over time. Soil was amended with a mixture of all nine sources, with one source isotopically labeled in each treatment. We compared changes in the 13C enrichment of metabolites with respect to carbon source and time over a 48-day incubation and contrasted differences between soluble sources (glucose, xylose, amino acids, etc.) and insoluble sources (cellulose and palmitic acid). Whole soil metabolite profiles varied singularly by time, while the composition of 13C-labeled metabolites differed primarily by carbon source (R2 = 0.68) rather than time (R2 = 0.07), with source-specific differences persisting throughout incubations. The 13C labeling of metabolites from insoluble carbon sources occurred slower than that from soluble sources but yielded a higher average atom percent (atom%) 13C in metabolite markers of biomass (amino acids and nucleic acids). The 13C enrichment of metabolite markers of biomass stabilized between 5 and 15 atom% 13C by the end of incubations. Temporal patterns in the 13C enrichment of tricarboxylic acid cycle intermediates, nucleobases (uracil and thymine), and by-products of DNA salvage (allantoin) closely tracked microbial activity. Our results demonstrate that metabolite production in soils is driven by the carbon source supplied to the community and that the fate of carbon in metabolites do not generally converge over time as a result of ongoing microbial processing and recycling. IMPORTANCE Carbon metabolism in soil remains poorly described due to the inherent difficulty of obtaining information on the microbial metabolites produced by complex soil communities. Our study demonstrates the use of stable isotope probing (SIP) to study carbon metabolism in soil by tracking 13C from supplied carbon sources into metabolite pools and biomass. We show that differences in the metabolism of sources influence the fate of carbon in soils. Heterogeneity in 13C-labeled metabolite profiles corresponded with compositional differences in the metabolically active populations, providing a basis for how microbial community composition correlates with the quality of soil carbon. Our study demonstrates the application of SIP-metabolomics in studying soils and identifies several metabolite markers of growth, activity, and other aspects of microbial function.

Struo: a pipeline for building custom databases for common metagenome profilers.

SUMMARY: Taxonomic and functional information from microbial communities can be efficiently obtained by metagenome profiling, which requires databases of genes and genomes to which sequence reads are mapped. However, the databases that accompany metagenome profilers are not updated at a pace that matches the increase in available microbial genomes, and unifying database content across metagenome profiling tools can be cumbersome. To address this, we developed Struo, a modular pipeline that automatizes the acquisition of genomes from public repositories and the construction of custom databases for multiple metagenome profilers. The use of custom databases that broadly represent the known microbial diversity by incorporating novel genomes results in a substantial increase in mappability of reads in synthetic and real metagenome datasets. AVAILABILITY AND IMPLEMENTATION: Source code available for download at https://github.com/leylabmpi/Struo. Custom genome taxonomy database databases available at http://ftp.tue.mpg.de/ebio/projects/struo/. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

DeepMAsED: evaluating the quality of metagenomic assemblies.

MOTIVATION: Methodological advances in metagenome assembly are rapidly increasing in the number of published metagenome assemblies. However, identifying misassemblies is challenging due to a lack of closely related reference genomes that can act as pseudo ground truth. Existing reference-free methods are no longer maintained, can make strong assumptions that may not hold across a diversity of research projects, and have not been validated on large-scale metagenome assemblies. RESULTS: We present DeepMAsED, a deep learning approach for identifying misassembled contigs without the need for reference genomes. Moreover, we provide an in silico pipeline for generating large-scale, realistic metagenome assemblies for comprehensive model training and testing. DeepMAsED accuracy substantially exceeds the state-of-the-art when applied to large and complex metagenome assemblies. Our model estimates a 1% contig misassembly rate in two recent large-scale metagenome assembly publications. CONCLUSIONS: DeepMAsED accurately identifies misassemblies in metagenome-assembled contigs from a broad diversity of bacteria and archaea without the need for reference genomes or strong modeling assumptions. Running DeepMAsED is straight-forward, as well as is model re-training with our dataset generation pipeline. Therefore, DeepMAsED is a flexible misassembly classifier that can be applied to a wide range of metagenome assembly projects. AVAILABILITY AND IMPLEMENTATION: DeepMAsED is available from GitHub at https://github.com/leylabmpi/DeepMAsED. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Reclassification of Catabacter hongkongensis as Christensenella hongkongensis comb. nov. based on whole genome analysis.

The genera Catabacter (family 'Catabacteraceae') and Christensenella (family Christensenellaceae) are close relatives within the phylum Firmicutes. Members of these genera are strictly anaerobic, non-spore-forming and short straight rods with diverse phenotypes. Phylogenetic analysis of 16S rRNA genes suggest that Catabacter splits Christensenella into a polyphyletic clade. In an effort to ensure that family/genus names represent monophyletic clades, we performed a whole-genome based analysis of the genomes available for the cultured representatives of these genera: four species of Christensenella and two strains of Catabacter hongkongensis. A concatenated alignment of 135 shared protein sequences of single-copy core genes present in the included strains indicates that C. hongkongensis is indeed nested within the Christensenella clade. Based on their evolutionary relationship, we propose the transfer of Catabacter hongkongensis to the genus Christensenella as Christensenella hongkongensis comb. nov.

Phytoplankton succession affects the composition of Polynucleobacter subtypes in humic lakes.

Phytoplankton influence the composition of bacterial communities, but the taxonomic specificity of algal-bacterial interactions is unclear due to the aggregation of ecologically distinct bacterial populations by community characterization methods. Here we examine whether phytoplankton seasonal succession affects the composition of subtypes within the cosmopolitan freshwater bacterial genus Polynucleobacter. Changes in the composition of Polynucleobacter subtypes were characterized in samples collected weekly from May to August in 2003 and 2008 from three humic lakes using terminal restriction fragment length polymorphism fingerprinting of the protein-encoding cytochrome c oxidase ccoN gene. Changes in phytoplankton population abundances explained, on average, 30% of temporal variation in the composition of Polynucleobacter subtypes and the interaction between phytoplankton and the environment explained an additional 18% of temporal variation. The effect of phytoplankton on specific Polynucleobacter subtypes was experimentally confirmed by changes in Polynucleobacter subtype composition following incubation with different phytoplankton assemblages or a no-phytoplankton control. Phytoplankton-associated subtypes and differentiation in substrate use among subtypes likely contribute to the effects of phytoplankton on Polynucleobacter subtype composition. Interactions between unique Polynucleobacter populations and phytoplankton highlight the ecological significance and specificity of species interactions in freshwater communities.

Differentiation between sediment and hypolimnion methanogen communities in humic lakes.

The traditional view of carbon cycling within the pelagic zone of freshwater lakes has consisted of methane production within the anoxic sediment, followed by diffusive flux and ebullition through the water column. Methanogenic archaea have been shown to be present within the water columns of freshwater lakes; however, little is known about whether these methanogenic communities are distinct from those in the sediment or how these communities change over space and time. We used the methanogen-specific phylogenetic marker mcrA to perform a 3-year study focusing on the community structure of methanogens within the sediment and anoxic hypolimnion water layer of five humic lakes in WI, USA. The hypolimnion and sediment communities were distinct in composition, richness and phylogenetic diversity. Hypolimnion communities displayed a temporally stable biogeographical pattern among lakes, which was driven by both lake-specific environmental variables and barriers to dispersal. We conclude that the hypolimnion comprised communities of methanogens that are distinct from those in the sediment, differentiated among lakes, and likely have unique ecological roles and evolutionary trajectories in these anaerobic environments.

Lineage-specific responses of microbial communities to environmental change.

A great challenge facing microbial ecology is how to define ecologically relevant taxonomic units. To address this challenge, we investigated how changing the definition of operational taxonomic units (OTUs) influences the perception of ecological patterns in microbial communities as they respond to a dramatic environmental change. We used pyrosequenced tags of the bacterial V2 16S rRNA region, as well as clone libraries constructed from the cytochrome oxidase C gene ccoN, to provide additional taxonomic resolution for the common freshwater genus Polynucleobacter. At the most highly resolved taxonomic scale, we show that distinct genotypes associated with the abundant Polynucleobacter lineages exhibit divergent spatial patterns and dramatic changes over time, while the also abundant Actinobacteria OTUs are highly coherent. This clearly demonstrates that different bacterial lineages demand different taxonomic definitions to capture ecological patterns. Based on the temporal distribution of highly resolved taxa in the hypolimnion, we demonstrate that change in the population structure of a single genotype can provide additional insight into the mechanisms of community-level responses. These results highlight the importance and feasibility of examining ecological change in microbial communities across taxonomic scales while also providing valuable insight into the ecological characteristics of ecologically coherent groups in this system.

Obesity is the main driver of altered gut microbiome functions in the metabolically unhealthy.

Obesity (OB) and cardiometabolic disease are major public health issues linked to changes in the gut microbiome. OB and poor cardiometabolic health status (CHS) are often comorbid, which hinders efforts to identify components of the microbiome uniquely linked to either one. Here, we used a deeply phenotyped cohort of 408 adults from Colombia, including subjects with OB, unhealthy CHS, or both, to validate previously reported features of gut microbiome function and diversity independently correlated with OB or CHS using fecal metagenomes. OB was defined by body mass index, waist circumference, and body fat; CHS as healthy or unhealthy according to blood biochemistry and anthropometric data. We found that OB, more so than metabolic status, drove associations with gut microbiome structure and functions. The microbiome of obese individuals with and without co-existing unhealthy CHS was characterized by reduced metagenomic diversity, reduced fermentative potential and elevated capacity to respond to oxidative stress and produce bacterial antigens. Disease-linked features were correlated with increased host blood pressure and inflammatory markers, and were mainly contributed by members of the family Enterobacteriaceae. Our results link OB with a microbiome able to tolerate an inflammatory and oxygenated gut state, and suggest that OB is the main driver of microbiome functional differences when poor CHS is a comorbidity.

An integrated systems biology approach reveals differences in formate metabolism in the genus Methanothermobacter.

Methanogenesis allows methanogenic archaea to generate cellular energy for their growth while producing methane. Thermophilic hydrogenotrophic species of the genus Methanothermobacter have been recognized as robust biocatalysts for a circular carbon economy and are already applied in power-to-gas technology with biomethanation, which is a platform to store renewable energy and utilize captured carbon dioxide. Here, we generated curated genome-scale metabolic reconstructions for three Methanothermobacter strains and investigated differences in the growth performance of these same strains in chemostat bioreactor experiments with hydrogen and carbon dioxide or formate as substrates. Using an integrated systems biology approach, we identified differences in formate anabolism between the strains and revealed that formate anabolism influences the diversion of carbon between biomass and methane. This finding, together with the omics datasets and the metabolic models we generated, can be implemented for biotechnological applications of Methanothermobacter in power-to-gas technology, and as a perspective, for value-added chemical production.

Silent recognition of flagellins from human gut commensal bacteria by Toll-like receptor 5.

Flagellin, the protein subunit of the bacterial flagellum, stimulates the innate immune receptor Toll-like receptor 5 (TLR5) after pattern recognition or evades TLR5 through lack of recognition. This binary response fails to explain the weak agonism of flagellins from commensal bacteria, raising the question of how TLR5 response is tuned. Here, we screened abundant flagellins present in metagenomes from human gut for both TLR5 recognition and activation and uncovered a class of flagellin-TLR5 interaction termed silent recognition. Silent flagellins were weak TLR5 agonists despite pattern recognition. Receptor activity was tuned by a TLR5-flagellin interaction distal to the site of pattern recognition that was present in Salmonella flagellin but absent in silent flagellins. This interaction enabled flagellin binding to preformed TLR5 dimers and increased TLR5 signaling by several orders of magnitude. Silent recognition by TLR5 occurred in human organoids and mice, and silent flagellin proteins were present in human stool. These flagellins were produced primarily by the abundant gut bacteria Lachnospiraceae and were enriched in nonindustrialized populations. Our findings provide a mechanism for the innate immune system to tolerate commensal-derived flagellins while remaining vigilant to the presence of flagellins produced by pathogens.

Codiversification of gut microbiota with humans.

The gut microbiomes of human populations worldwide have many core microbial species in common. However, within a species, some strains can show remarkable population specificity. The question is whether such specificity arises from a shared evolutionary history (codiversification) between humans and their microbes. To test for codiversification of host and microbiota, we analyzed paired gut metagenomes and human genomes for 1225 individuals in Europe, Asia, and Africa, including mothers and their children. Between and within countries, a parallel evolutionary history was evident for humans and their gut microbes. Moreover, species displaying the strongest codiversification independently evolved traits characteristic of host dependency, including reduced genomes and oxygen and temperature sensitivity. These findings all point to the importance of understanding the potential role of population-specific microbial strains in microbiome-mediated disease phenotypes.

Struo2: efficient metagenome profiling database construction for ever-expanding microbial genome datasets.

Mapping metagenome reads to reference databases is the standard approach for assessing microbial taxonomic and functional diversity from metagenomic data. However, public reference databases often lack recently generated genomic data such as metagenome-assembled genomes (MAGs), which can limit the sensitivity of read-mapping approaches. We previously developed the Struo pipeline in order to provide a straight-forward method for constructing custom databases; however, the pipeline does not scale well enough to cope with the ever-increasing number of publicly available microbial genomes. Moreover, the pipeline does not allow for efficient database updating as new data are generated. To address these issues, we developed Struo2, which is >3.5 fold faster than Struo at database generation and can also efficiently update existing databases. We also provide custom Kraken2, Bracken, and HUMAnN3 databases that can be easily updated with new genomes and/or individual gene sequences. Efficient database updating, coupled with our pre-generated databases, enables "assembly-enhanced" profiling, which increases database comprehensiveness via inclusion of native genomic content. Inclusion of newly generated genomic content can greatly increase database comprehensiveness, especially for understudied biomes, which will enable more accurate assessments of microbiome diversity.

Genomic Insights into Adaptations of Trimethylamine-Utilizing Methanogens to Diverse Habitats, Including the Human Gut.

Archaea of the order Methanomassiliicoccales use methylated amines such as trimethylamine as the substrates for methanogenesis. They form two large phylogenetic clades and reside in diverse environments, from soil to the human gut. Two genera, one from each clade, inhabit the human gut: Methanomassiliicoccus, which has one cultured representative, and "Candidatus Methanomethylophilus," which has none. Questions remain regarding their distribution across biomes and human populations, their association with other taxa in the gut, and whether host genetics correlate with their abundance. To gain insight into the Methanomassiliicoccales clade, particularly its human-associated members, we performed a genomic comparison of 72 Methanomassiliicoccales genomes and assessed their presence in metagenomes derived from the human gut (n = 4,472, representing 22 populations), nonhuman animal gut (n = 145), and nonhost environments (n = 160). Our analyses showed that all taxa are generalists; they were detected in animal gut and environmental samples. We confirmed two large clades, one enriched in the gut and the other enriched in the environment, with notable exceptions. Genomic adaptations to the gut include genome reduction and genes involved in the shikimate pathway and bile resistance. Genomic adaptations differed by clade, not habitat preference, indicating convergent evolution between the clades. In the human gut, the relative abundance of Methanomassiliicoccales spp. correlated with trimethylamine-producing bacteria and was unrelated to host genotype. Our results shed light on the microbial ecology of this group and may help guide Methanomassiliicoccales-based strategies for trimethylamine mitigation in cardiovascular disease.IMPORTANCE Methanomassiliicoccales are less-known members of the human gut archaeome. Members of this order use methylated amines, including trimethylamine, in methane production. This group has only one cultured representative; how its members adapted to inhabit the mammalian gut and how they interact with other microbes is largely unknown. Using bioinformatics methods applied to DNA from a wide range of samples, we profiled the abundances of these Archaea spp. in environmental and host-associated microbial communities. We observed two groups of Methanomassiliicoccales, one largely host associated and one largely found in environmental samples, with some exceptions. When host associated, these Archaea have smaller genomes and possess genes related to bile resistance and aromatic amino acid precursors. We did not detect Methanomassiliicoccales in all human populations tested, but when present, they were correlated with bacteria known to produce trimethylamine. Due to their metabolism of trimethylamine, these intriguing Archaea may form the basis of novel therapies for cardiovascular disease.

Vertebrate host phylogeny influences gut archaeal diversity.

Commonly used 16S rRNA gene primers do not detect the full range of archaeal diversity present in the vertebrate gut. As a result, several questions regarding the archaeal component of the gut microbiota remain, including which Archaea are host-associated, the specificities of such associations and the major factors influencing archaeal diversity. Using 16S rRNA gene amplicon sequencing with primers that specifically target Archaea, we obtained sufficient sequence data from 185 gastrointestinal samples collected from 110 vertebrate species that span five taxonomic classes (Mammalia, Aves, Reptilia, Amphibia and Actinopterygii), of which the majority were wild. We provide evidence for previously undescribed Archaea-host associations, including Bathyarchaeia and Methanothermobacter, the latter of which was prevalent among Aves and relatively abundant in species with higher body temperatures, although this association could not be decoupled from host phylogeny. Host phylogeny explained archaeal diversity more strongly than diet, while specific taxa were associated with both factors, and cophylogeny was significant and strongest for mammalian herbivores. Methanobacteria was the only class predicted to be present in the last common ancestors of mammals and all host species. Further analysis indicated that Archaea-Bacteria interactions have a limited effect on archaeal diversity. These findings expand our current understanding of Archaea-vertebrate associations.

Free-Living, Psychrotrophic Bacteria of the Genus Psychrobacter Are Descendants of Pathobionts.

Host-adapted microorganisms are generally assumed to have evolved from free-living, environmental microorganisms, as examples of the reverse process are rare. In the phylum Gammaproteobacteria, family Moraxellaceae, the genus Psychrobacter includes strains from a broad ecological distribution including animal bodies as well as sea ice and other nonhost environments. To elucidate the relationship between these ecological niches and Psychrobacter's evolutionary history, we performed tandem genomic analyses with phenotyping of 85 Psychrobacter accessions. Phylogenomic analysis of the family Moraxellaceae reveals that basal members of the Psychrobacter clade are Moraxella spp., a group of often-pathogenic organisms. Psychrobacter exhibited two broad growth patterns in our phenotypic screen: one group that we called the "flexible ecotype" (FE) had the ability to grow between 4 and 37°C, and the other, which we called the "restricted ecotype" (RE), could grow between 4 and 25°C. The FE group includes phylogenetically basal strains, and FE strains exhibit increased transposon copy numbers, smaller genomes, and a higher likelihood to be bile salt resistant. The RE group contains only phylogenetically derived strains and has increased proportions of lipid metabolism and biofilm formation genes, functions that are adaptive to cold stress. In a 16S rRNA gene survey of polar bear fecal samples, we detect both FE and RE strains, but in in vivo colonizations of gnotobiotic mice, only FE strains persist. Our results indicate the ability to grow at 37°C, seemingly necessary for mammalian gut colonization, is an ancestral trait for Psychrobacter, which likely evolved from a pathobiont.IMPORTANCE Host-associated microbes are generally assumed to have evolved from free-living ones. The evolutionary transition of microbes in the opposite direction, from host associated toward free living, has been predicted based on phylogenetic data but not studied in depth. Here, we provide evidence that the genus Psychrobacter, particularly well known for inhabiting low-temperature, high-salt environments such as sea ice, permafrost soils, and frozen foodstuffs, has evolved from a mammalian-associated ancestor. We show that some Psychrobacter strains retain seemingly ancestral genomic and phenotypic traits that correspond with host association while others have diverged to psychrotrophic or psychrophilic lifestyles.