sv_association_nature

SV term

The spectrum of human genetic variation ranges from the single base pair to large chromosomal events, but it has become apparent that human genomes differ more as a consequence of structural variation than of single-base-pair differences

Structural variation was originally defined as insertions, deletions and inversions greater than 1kb in size. With the sequencing of human genomes now becoming routine , the operational spectrum of structural variants (SVs) and copy number variants (CNVs) has widened to include much smaller events (for example, those >50 bp in length).

The challenge now is to discover the full extent of structural variation and to be able to genotype it routinely in order to understand its effects on

• human disease,
• complex traits
• evolution

Two distinct models to associate the sv and disease

two distinct models have been proposed with respect to associations between disease and structural variation.

• The first involves large variants (typically gains and losses several hundred kilobase pairs in length) that are individually rare in the population (<1%) but collectively account for a significant fraction of disease

• The second includes multicopy gene families that are commonly copy number variable and contribute to disease susceptibility, as seen for traits related to immune gene functions

classes of SV

• Structural variant (SV). Genomic rearrangements that affect >50bp of sequence, including deletions, novel insertions, inversions, mobile-element transpositions, duplications and translocations

• Copy number variant (CNV). Also defined as unbalanced structural variants; variants that change the number of base pairs in the genome.

• Mobile elements DNA sequences that move location within the genome. Active mobile elements (transposons) in the human genome include Alu, L1 and SVA sequences

Scientific Questions

The discovery and genotyping of structural variation has been central to understanding these disease associations

• Systematic and comprehensive assessment of structural variation has been problematic owing to the complexity and multifaceted features of SVs

• SV discovery and genotyping requires accurate prediction of three features: copy, content and structure

• SVs tend to reside within repetitive DNA, which makes their characterization more difficult. SVs vary widely in size and there are numerous classes of structural variation: deletions, translocations, inversions, mobile elements, tandem duplications and novel insertions

• Once a variant has been detected, validated and characterized at the sequence level (discovery), a different suite of methods may be applied to infer genotypes with relaxed threshold

SV calling methods

• array-based

• sequencing-based There are four general types of strategy, all of which focus on mapping sequence reads to the reference genome and subsequently identifying discordant signatures or patterns that are diagnostic of different classes of SV

Importance of This work

To detect segments of varying lengths, potentially containing multiple genes, that are deleted from certain bacteria in some individuals or present in a variable number of copies in others

identify microbial genomic structural variants (SVs) and find them to be prevalent in the human gut microbiome across phyla and to replicate in different cohorts. SVs are enriched for CRISPRassociated and antibiotic-producing functions and depleted from housekeeping genes, suggesting that they have a role in microbial adaptation. We find multiple associations between SVs and host disease risk factors, many of which replicate in an independent cohort. Exploring genes that are clustered in the same SV, we uncover several possible mechanistic links between the microbiome and its host, including a region

our results uncover a nascent layer of variability in the microbiome that is associated with microbial adaptation and host health

Accurate metagenomic read assignment using ICRA

Problem: over 15% of the metagenomic reads were assigned ambiguously to multiple references upon mapping to a database of 3,953 bacterial genomes

Solution: To address this problem, we devised the ICRA algorithm which uses read assignments, read and mapping qualities, sequencing coverage depth along microbial entities (for example, bacterial genomes) and microbial relative abundances to reassign ambiguously mapped reads

ICRA introduces a demand for sufficient coverage over entities that are to be considered present in a sample, making it robust to genomic regions with extremely high or low coverage that may arise from misassemblies, homology to other microbes or phage activation

Such regions could otherwise bias the estimated relative abundances, potentially even assigning abundances to genomic entities that are absent from the sample.

To test the performance of ICRA, we validated the two key components of the algorithm: its ability to resolve ambiguous read assignments and the accuracy of the species relative abundances that it infers

SGV-Finder to seek to systematically characterize structural variation

we applied to ICRA-corrected read assignments of 887 metagenomic samples to a reference database of 3,953 representative microbial genomes

core of SGV:

SGV-Finder analyses coverage depth across all microbial genomes in all samples to characterize SVs with respect to the standardized coverage of a genome in a given sample

We differentiate between deletion SVs, that are deleted and not covered in 25–75% of samples, and variable SVs, that have highly variable coverage across samples. In both SV types, segments are united based on co-occurrence (deletion SVs) or correlation (variable SVs).

An online metagenome explorer for all SVs and the genes they encompass is available at SV Finder

SV summarized information about bacterial metagenome (SVs are highly prevalent in the microbiome)

• detected 2,423 variable SVs and 5,056 deletion SVs in 56 bacteria that passed our coverage thresholds

• SVs were detected in six bacterial phyla and one archaeal phylum, with 5–241 SVs per species in 1.4–18.6 kilobase pairs (kbp) average size per species

• Variable and deletion SVs make up 0.3–8.4% and 5.0–26.9% of the microbial genome, respectively

Information to infer:

• This apparent disparity in size may suggest inherent differences in the formation of the two types of SVs

• We detected SVs in every subject and strain analysed, demonstrating the ubiquity of such variations.

SV is prevalent across distinct populations

to To test the universality of these regions and reinforce their biological relevance, we applied ICRA and SGV-Finder independently in

• study cohorts (Iseri)

• validation cohorts (Dutch Lifelines)

SVs are potentially involved in microbial adaptation

characterize the function of SVs by searching for enriched or depleted genetic functions. We annotated gene functions across variable SVs, deletion SVs and ‘conserved’ regions

Using the KEGG BRITE hierarchy, we found that ‘housekeeping’ modules such as nucleotide and amino acid metabolism or carbohydrate and lipid metabolism were significantly depleted from SVs and significantly enriched in conserved regions

Conversely, modules classified as ABC-2 type- and other transport systems were significantly enriched in variable SVs

bacterial conjugation systems, to which the T4SS is related, strongly associate with variability, further implicating SVs as tools of adaptation and speciation

To further characterize the potential contribution of SVs to microbial adaptation, we searched for SVs associated with the fitness of their harbouring microbe. As a proxy for fitness, we calculated bacterial growth rates of 21 strains with sufficient coverage and complete reference genomes, using a method that estimates growth from DNA copy number differences created during DNA replication

suggesting that certain SVs may be important for bacterial adaptation and fitness.

To probe the mechanisms potentially underlying this adaptation, we systematically examined the genetic content of growth-associated deletion SVs and found similar functional profiles as in all SVs, with a

• depletion of housekeeping functions and
• enrichment for genes involved with CRISPR-, transposon- and HGT-associated genes

SVs associate with risk factors across cohorts (SVs associate with disease risk, replicated in another cohort.)

we associated the abundance of variable SVs and the presence of deletion SVs with metrics of health and risk factors

We found 81 (Spearman’s correlation) and 43 (Mann– Whitney U-test; Fig. 3b) significant associations FDR corrected at 0.1 for variable and deletion SVs, respectively, demonstrating the potential importance of microbial SVs to the human host.

the associations of specific SVs with risk factors allows us to pinpoint specific regions and mechanisms that may underlie the association.

These seemingly paradoxical associations between SVs and risk factors further suggest that SVs represent a different layer of information compared to the taxonomic level, one which may assist in obtaining mechanistic insights into host–microbe interactions

generality of this methodology

Our methodology is highly adaptable to any metagenomic scenario and could be used, for example, to detect SVs in the soil microbiome