Determinants of microbiome composition: Insights from free-ranging hybrid zebras (Equus quagga × grevyi).

The composition of mammalian gut microbiomes is highly conserved within species, yet the mechanisms by which microbiome composition is transmitted and maintained within lineages of wild animals remain unclear. Mutually compatible hypotheses exist, including that microbiome fidelity results from inherited dietary habits, shared environmental exposure, morphophysiological filtering and/or maternal effects. Interspecific hybrids are a promising system in which to interrogate the determinants of microbiome composition because hybrids can decouple traits and processes that are otherwise co-inherited in their parent species. We used a population of free-living hybrid zebras (Equus quagga × grevyi) in Kenya to evaluate the roles of these four mechanisms in regulating microbiome composition. We analysed faecal DNA for both the trnL-P6 and the 16S rRNA V4 region to characterize the diets and microbiomes of the hybrid zebra and of their parent species, plains zebra (E. quagga) and Grevy's zebra (E. grevyi). We found that both diet and microbiome composition clustered by species, and that hybrid diets and microbiomes were largely nested within those of the maternal species, plains zebra. Hybrid microbiomes were less variable than those of either parent species where they co-occurred. Diet and microbiome composition were strongly correlated, although the strength of this correlation varied between species. These patterns are most consistent with the maternal-effects hypothesis, somewhat consistent with the diet hypothesis, and largely inconsistent with the environmental-sourcing and morphophysiological-filtering hypotheses. Maternal transmittance likely operates in conjunction with inherited feeding habits to conserve microbiome composition within species.

Reassessing the modularity of gene co-expression networks using the Stochastic Block Model.

Finding communities in gene co-expression networks is a common first step toward extracting biological insight from these complex datasets. Most community detection algorithms expect genes to be organized into assortative modules, that is, groups of genes that are more associated with each other than with genes in other groups. While it is reasonable to expect that these modules exist, using methods that assume they exist a priori is risky, as it guarantees that alternative organizations of gene interactions will be ignored. Here, we ask: can we find meaningful communities without imposing a modular organization on gene co-expression networks, and how modular are these communities? For this, we use a recently developed community detection method, the weighted degree corrected stochastic block model (SBM), that does not assume that assortative modules exist. Instead, the SBM attempts to efficiently use all information contained in the co-expression network to separate the genes into hierarchically organized blocks of genes. Using RNA-seq gene expression data measured in two tissues derived from an outbred population of Drosophila melanogaster , we show that (a) the SBM is able to find ten times as many groups as competing methods, that (b) several of those gene groups are not modular, and that (c) the functional enrichment for non-modular groups is as strong as for modular communities. These results show that the transcriptome is structured in more complex ways than traditionally thought and that we should revisit the long-standing assumption that modularity is the main driver of the structuring of gene co-expression networks.

Characterizing the landscape of gene expression variance in humans.

Gene expression variance has been linked to organismal function and fitness but remains a commonly neglected aspect of molecular research. As a result, we lack a comprehensive understanding of the patterns of transcriptional variance across genes, and how this variance is linked to context-specific gene regulation and gene function. Here, we use 57 large publicly available RNA-seq data sets to investigate the landscape of gene expression variance. These studies cover a wide range of tissues and allowed us to assess if there are consistently more or less variable genes across tissues and data sets and what mechanisms drive these patterns. We show that gene expression variance is broadly similar across tissues and studies, indicating that the pattern of transcriptional variance is consistent. We use this similarity to create both global and within-tissue rankings of variation, which we use to show that function, sequence variation, and gene regulatory signatures contribute to gene expression variance. Low-variance genes are associated with fundamental cell processes and have lower levels of genetic polymorphisms, have higher gene-gene connectivity, and tend to be associated with chromatin states associated with transcription. In contrast, high-variance genes are enriched for genes involved in immune response, environmentally responsive genes, immediate early genes, and are associated with higher levels of polymorphisms. These results show that the pattern of transcriptional variance is not noise. Instead, it is a consistent gene trait that seems to be functionally constrained in human populations. Furthermore, this commonly neglected aspect of molecular phenotypic variation harbors important information to understand complex traits and disease.

From GWAS to signal validation: An approach for estimating genetic effects while preserving genomic context.

Validating associations between genotypic and phenotypic variation remains a challenge, despite advancements in association studies. Common approaches for signal validation rely on gene-level perturbations, such as loss-of-function mutations or RNAi, which test the effect of genetic modifications usually not observed in nature. CRISPR-based methods can validate associations at the SNP level, but have significant drawbacks, including resulting off-target effects and being both time-consuming and expensive. Both approaches usually modify the genome of a single genetic background, limiting the generalizability of experiments. To address these challenges, we present a simple, low-cost experimental scheme for validating genetic associations at the SNP level in outbred populations. The approach involves genotyping live outbred individuals at a focal SNP, crossing homozygous individuals with the same genotype at that locus, and contrasting phenotypes across resulting synthetic outbred populations. We tested this method in Drosophila melanogaster, measuring the longevity effects of a polymorphism at a naturally-segregating cis-eQTL for the midway gene. Our results demonstrate the utility of this method in SNP-level validation of naturally occurring genetic variation regulating complex traits. This method provides a bridge between the statistical discovery of genotype-phenotype associations and their validation in the natural context of heterogeneous genomic contexts.

Dietary stress remodels the genetic architecture of lifespan variation in outbred Drosophila.

Evolutionary theory suggests that lifespan-reducing alleles should be purged from the gene pool, and yet decades of genome-wide association and model organism studies have shown that they persist. One potential explanation is that alleles that regulate lifespan do so only in certain environmental contexts. We exposed outbred Drosophila to control and high-sugar diets and genotyped more than 10,000 adult flies to track allele frequency changes over the course of a single adult lifespan. We identified thousands of lifespan-associated alleles associated with early versus late-life trade-offs, late-onset effects and genotype-by-environment interactions. Remarkably, a third of lifespan-associated genetic variation had environmentally dependent effects on lifespan. We find that lifespan-reducing alleles are often recently derived, have stronger effects on a high-sugar diet and show signatures of selection in wild Drosophila populations, consistent with the evolutionary mismatch hypothesis. Our results provide insight into the highly polygenic and context-dependent genetic architecture of lifespan variation and the evolutionary processes that shape this key trait.

The structure of behavioral variation within a genotype.

Individual animals vary in their behaviors. This is true even when they share the same genotype and were reared in the same environment. Clusters of covarying behaviors constitute behavioral syndromes, and an individual's position along such axes of covariation is a representation of their personality. Despite these conceptual frameworks, the structure of behavioral covariation within a genotype is essentially uncharacterized and its mechanistic origins unknown. Passing hundreds of inbred Drosophila individuals through an experimental pipeline that captured hundreds of behavioral measures, we found sparse but significant correlations among small sets of behaviors. Thus, the space of behavioral variation has many independent dimensions. Manipulating the physiology of the brain, and specific neural populations, altered specific correlations. We also observed that variation in gene expression can predict an individual's position on some behavioral axes. This work represents the first steps in understanding the biological mechanisms determining the structure of behavioral variation within a genotype.

TM3'seq: A Tagmentation-Mediated 3' Sequencing Approach for Improving Scalability of RNAseq Experiments.

RNA-seq has become the standard tool for collecting genome-wide expression data in diverse fields, from quantitative genetics and medical genomics to ecology and developmental biology. However, RNA-seq library preparation is still prohibitive for many laboratories. Recently, the field of single-cell transcriptomics has reduced costs and increased throughput by adopting early barcoding and pooling of individual samples -producing a single final library containing all samples. In contrast, RNA-seq protocols where each sample is processed individually are significantly more expensive and lower throughput than single-cell approaches. Yet, many projects depend on individual library generation to preserve important samples or for follow-up re-sequencing experiments. Improving on currently available RNA-seq methods we have developed TM3'seq, a 3'-enriched library preparation protocol that uses Tn5 transposase and preserves sample identity at each step. TM3'seq is designed for high-throughput processing of individual samples (96 samples in 6h, with only 3h hands-on time) at a fraction of the cost of commercial kits ($1.5 per sample). The protocol was tested in a range of human and Drosophila melanogaster RNA samples, recovering transcriptomes of the same quality and reliability than the commercial NEBNext kit. We expect that the cost- and time-efficient features of TM3'seq make large-scale RNA-seq experiments more permissive for the entire scientific community.

Searching for solutions to the missing heritability problem.

Rare genetic variants in yeast explain a large amount of phenotypic variation in a complex trait like growth.

Mouse Skull Mean Shape and Shape Robustness Rely on Different Genetic Architectures and Different Loci.

The genetic architecture of skull shape has been extensively studied in mice and the results suggest a highly polygenic and additive basis. In contrast few studies have explored the genetic basis of the skull variability. Canalization and developmental stability are the two components of phenotypic robustness. They have been proposed to be emergent properties of the genetic networks underlying the development of the trait itself, but this hypothesis has been rarely tested empirically. Here we use outbred mice to investigate the genetic architecture of canalization of the skull shape by implementing a genome-wide marginal epistatic test on 3D geometric morphometric data. The same data set had been used previously to explore the genetic architecture of the skull mean shape and its developmental stability. Here, we address two questions: (1) Are changes in mean shape and changes in shape variance associated with the same genomic regions? and (2) Do canalization and developmental stability rely on the same loci and genetic architecture and do they involve the same patterns of shape variation? We found that unlike skull mean shape, among-individual shape variance and fluctuating asymmetry (FA) show a total lack of additive effects. They are both associated with complex networks of epistatic interactions involving many genes (protein-coding and regulatory elements). Remarkably, none of the genomic loci affecting mean shape contribute these networks despite their enrichment for genes involved in craniofacial variation and diseases. We also found that the patterns of shape FA and individual variation are largely similar and rely on similar multilocus epistatic genetic networks, suggesting that the processes channeling variation within and among individuals are largely common. However, the loci involved in these two networks are completely different. This in turn underlines the difference in the origin of the variation at these two levels, and points at buffering processes that may be specific to each level.

Using the Mus musculus hybrid zone to assess covariation and genetic architecture of limb bone lengths.

Two subspecies of the house mouse, Mus musculus domesticus and Mus musculus musculus, meet in a narrow contact zone across Europe. Mice in the hybrid zone are highly admixed, representing the full range of mixed ancestry from the two subspecies. Given the distinct morphologies of these subspecies, these natural hybrids can be used for genomewide association mapping at sufficiently high resolution to directly infer candidate genes. We focus here on limb bone length differences, which is of special interest for understanding the evolution of developmentally correlated traits. We used 172 first-generation descendants of wild-caught mice from the hybrid zone to measure the length of stylopod (humerus/femur), zeugopod (ulna/tibia) and autopod (metacarpal/metatarsal) elements in skeletal CT scans. We find phenotypic covariation between limb elements in the hybrids similar to patterns previously described in Mus musculus domesticus inbred strains, suggesting that the hybrid genotypes do not influence the covariation pattern in a major way. Mapping was performed using 143,592 SNPs and identified several genomic regions associated with length differences in each bone. Bone length was found to be highly polygenic. None of the candidate regions include the canonical genes known to control embryonic limb development. Instead, we are able to identify candidate genes with known roles in osteoblast differentiation and bone structure determination, as well as recently evolved genes of, as yet, unknown function.

Genomic regions controlling shape variation in the first upper molar of the house mouse.

Numerous loci of large effect have been shown to underlie phenotypic variation between species. However, loci with subtle effects are presumably more frequently involved in microevolutionary processes but have rarely been discovered. We explore the genetic basis of shape variation in the first upper molar of hybrid mice between Mus musculus musculus and M. m. domesticus. We performed the first genome-wide association study for molar shape and used 3D surface morphometrics to quantify subtle variation between individuals. We show that many loci of small effect underlie phenotypic variation, and identify five genomic regions associated with tooth shape; one region contained the gene microphthalmia-associated transcription factor Mitf that has previously been associated with tooth malformations. Using a panel of five mutant laboratory strains, we show the effect of the Mitf gene on tooth shape. This is the first report of a gene causing subtle but consistent variation in tooth shape resembling variation in nature.

Craniofacial shape transition across the house mouse hybrid zone: implications for the genetic architecture and evolution of between-species differences.

Craniofacial shape differences between taxa have often been linked to environmental adaptation, e.g., new food sources, or have been studied in the context of domestication. Evidence for the genetic basis of such phenotypic differences to date suggests that between-species as well as between-population variation has an oligogenic basis, i.e., few loci of large effect explain most of the variation. In mice, it has been shown that within-population craniofacial variation has a highly polygenic basis, but there are no data regarding the genetic basis of between-species differences in natural populations. Here, we address this question using a phenotype-focused approach. Using 3D geometric morphometrics, we phenotyped a panel of mice derived from a natural hybrid zone between Mus musculus domesticus and Mus mus musculus and quantify the transition of craniofacial shape along the hybridization gradient. We find a continuous shape transition along the hybridization gradient and unaltered developmental stability associated with hybridization. This suggests that the morphospace between the two subspecies is continuous despite reproductive isolation and strong barriers to gene flow. We show that quantitative changes in overall genome composition generate quantitative changes in craniofacial shape; this supports a highly polygenic basis for between-species craniofacial differences in the house mouse. We discuss our findings in the context of oligogenic versus polygenic models of the genetic architecture of morphological traits.

Mapping of Craniofacial Traits in Outbred Mice Identifies Major Developmental Genes Involved in Shape Determination.

The vertebrate cranium is a prime example of the high evolvability of complex traits. While evidence of genes and developmental pathways underlying craniofacial shape determination is accumulating, we are still far from understanding how such variation at the genetic level is translated into craniofacial shape variation. Here we used 3D geometric morphometrics to map genes involved in shape determination in a population of outbred mice (Carworth Farms White, or CFW). We defined shape traits via principal component analysis of 3D skull and mandible measurements. We mapped genetic loci associated with shape traits at ~80,000 candidate single nucleotide polymorphisms in ~700 male mice. We found that craniofacial shape and size are highly heritable, polygenic traits. Despite the polygenic nature of the traits, we identified 17 loci that explain variation in skull shape, and 8 loci associated with variation in mandible shape. Together, the associated variants account for 11.4% of skull and 4.4% of mandible shape variation, however, the total additive genetic variance associated with phenotypic variation was estimated in ~45%. Candidate genes within the associated loci have known roles in craniofacial development; this includes 6 transcription factors and several regulators of bone developmental pathways. One gene, Mn1, has an unusually large effect on shape variation in our study. A knockout of this gene was previously shown to affect negatively the development of membranous bones of the cranial skeleton, and evolutionary analysis shows that the gene has arisen at the base of the bony vertebrates (Eutelostomi), where the ossified head first appeared. Therefore, Mn1 emerges as a key gene for both skull formation and within-population shape variation. Our study shows that it is possible to identify important developmental genes through genome-wide mapping of high-dimensional shape features in an outbred population.

Use of a natural hybrid zone for genomewide association mapping of craniofacial traits in the house mouse.

The identification of the genes involved in morphological variation in nature is still a major challenge. Here, we explore a new approach: we combine 178 samples from a natural hybrid zone between two subspecies of the house mouse (Mus musculus domesticus and Mus musculus musculus), and high coverage of the genome (~ 145K SNPs) to identify loci underlying craniofacial shape variation. Due to the long history of recombination in the hybrid zone, high mapping resolution is anticipated. The combination of genomes from subspecies allows the mapping of both, variation within subspecies and inter-subspecific differences, thereby increasing the overall amount of causal genetic variation that can be detected. Skull and mandible shape were measured using 3D landmarks and geometric morphometrics. Using principal component axes as phenotypes, and a linear mixed model accounting for genetic relatedness in the mapping populations, we identified nine genomic regions associated with skull shape and 10 with mandible shape. High mapping resolution (median size of significant regions = 148 kb) enabled identification of single or few candidate genes in most cases. Some of the genes act as regulators or modifiers of signalling pathways relevant for morphological development and bone formation, including several with known craniofacial phenotypes in mice and humans. The significant associations combined explain 13% and 7% of the skull and mandible shape variation, respectively. In addition, a positive correlation was found between chromosomal length and proportion of variation explained. Our results suggest a complex genetic architecture for shape traits and support a polygenic model.

Exploring the effects of gene dosage on mandible shape in mice as a model for studying the genetic basis of natural variation.

Mandible shape in the mouse is a complex trait that is influenced by many genetic factors. However, little is known about the action of single genes on adult mandible shape so far, since most developmentally relevant genes are already required during embryogenesis, i.e., knockouts lead to embryonic death or severe deformations, before the mandible is fully formed. We employ here a geometric morphometric approach to identify subtle phenotypic differences caused by dosage effects of candidate genes. We use mouse strains with specific gene modifications (knockouts and knockins) to compare heterozygous animals with controls from the same stock, which is expected to be equivalent to a change of gene expression of the respective locus. Such differences in expression level are also likely to occur as part of the natural variation. We focus on Bmp pathway genes (Bmp4, its antagonist Noggin, and combinations of Bmp5-7 genotypes), but include also two other developmental control genes suspected to affect mandible development in some way (Egfr and Irf6). In addition, we study the effects of Hoxd13, as well as an extracellular matrix constituent (Col2a1). We find that subtle but significant shape differences are caused by differences in gene dosage of several of these genes. The changes seen for Bmp4 and Noggin are partially compatible with the action of these genes known from birds and fish. We find significant shape changes also for Hoxd13, although this gene has so far only been implicated in skeletal patterning processes of the limbs. Comparing the effect sizes of gene dosage changes to the variation found in natural populations of mice as well as quantitative trait loci (QTL) effects on mandible shape, we find that the effect sizes caused by gene dosage changes are at the lower end of the spectrum of natural variation, but larger than the average additive effects found in QTL studies. We conclude that studying gene dosage effects have the potential to provide new insights into aspects of craniofacial development, variation, and evolution.