High-throughput phenogenotyping clinical strains reveals bacterial determinants of treatment outcomes.

BACKGROUND

Combatting the tuberculosis (TB) epidemic caused by ( ) necessitates a better understanding of the factors contributing to patient clinical outcomes and transmission. While host and environmental factors have been evaluated, the impact of genetic background and phenotypic diversity is underexplored. Previous work has made associations between genetic lineages and some clinical and epidemiological features, but the bacterial traits underlying these connections are largely unknown.

METHODS

We developed a high-throughput functional genomics platform for defining genotype-phenotype relationships across a panel of clinical isolates. These phenotypic fitness profiles function as intermediate traits which can be linked to genetic variants and associated with clinical and epidemiological outcomes. We applied this approach to a collection of 158 strains from a study of transmission in Ho Chi Minh City, Vietnam. strains were genetically tagged in multiplicate, which allowed us to pool the strains and assess competitive fitness using deep sequencing across a set of 14 host-relevant antibiotic and metabolic conditions. Phylogenetic and monogenic associations with these intermediate traits were identified and then associated with clinical outcomes.

FINDINGS

clinical strains have a broad range of growth and drug response dynamics that can be clustered by their phylogenetic relationships. We identified novel monogenic associations with fitness in various metabolic and antibiotic conditions. Among these, we find that mutations in , a phosphodiesterase, which were identified through their association with slow growth in glycerol, are further associated with treatment failure. We also identify a previously uncharacterized subclade of Lineage 1 strains (L1.1.1.1) that is phenotypically distinguished by slow growth under most antibiotic and metabolic stress conditions . This clade is associated with cavitary disease, treatment failure, and demonstrates increased transmission potential.

INTERPRETATION

High-throughput phenogenotyping of Mtb clinical strains enabled bacterial intermediate trait identification that can provide a mechanistic link between genetic variation and patient clinical outcomes. strains associated with cavitary disease, treatment failure, and transmission potential display intermediate phenotypes distinguished by slow growth under various antibiotic and metabolic conditions. These data suggest that Mtb growth regulation is an adaptive advantage for host bacterial success in human populations, in at least some circumstances. These data further suggest markers for the underlying bacterial processes that govern these clinical outcomes.

FUNDING

National Institutes of Allergy and Infectious Diseases: P01 AI132130 (SS, SMF); P01 AI143575 (XW, SMF); U19 AI142793 (QL, SMF); 5T32AI132120-03 (SS); 5T32AI132120-04 (SS); 5T32AI049928-17 (SS) Wellcome Trust Fellowship in Public Health and Tropical Medicine: 097124/Z/11/Z (NTTT) National Health and Medical Research Council (NHMRC)/A*STAR joint call: APP1056689 (SJD) The funding sources had no involvement in study methodology, data collection, analysis, and interpretation nor in the writing or submission of the manuscript.

RESEARCH IN CONTEXT

We used different combinations of the words mycobacterium tuberculosis, tuberculosis, clinical strains, intermediate phenotypes, genetic barcoding, phenogenomics, cavitary disease, treatment failure, and transmission to search the PubMed database for all studies published up until January 20 , 2022. We only considered English language publications, which biases our search. Previous work linking determinants to clinical or epidemiological data has made associations between bacterial lineage, or less frequently, genetic polymorphisms to or models of pathogenesis, transmission, and clinical outcomes such as cavitary disease, treatment failure, delayed culture conversion, and severity. Many of these studies focus on the global pandemic Lineage 2 and Lineage 4 strains due in part to a deletion in a polyketide synthase implicated in host-pathogen interactions. There are a number of GWAS studies that have led to novel genetic determinants of drug resistance and tolerance. Previous GWAS analyses with clinical outcomes did not experimentally test any predicted phenotypes of the clinical strains. Published laboratory-based studies of clinical strains involve relatively small numbers of strains, do not identify the genetic basis of relevant phenotypes, or link findings to the corresponding clinical outcomes. There are two recent studies of other pathogens that describe phenogenomic analyses. One study of 331 clinical strains performed one-by-one phenotyping to identify bacterial features associated with clearance of infection and another details a competition experiment utilizing three barcoded clinical isolates to assay antimalarial fitness and resistance. We developed a functional genomics platform to perform high-throughput phenotyping of clinical strains. We then used these phenotypes as intermediate traits to identify novel bacterial genetic features associated with clinical outcomes. We leveraged this platform with a sample of 158 clinical strains from a cross sectional study of transmission in Ho Chi Minh City, Vietnam. To enable high-throughput phenotyping of large numbers of clinical isolates, we applied a DNA barcoding approach that has not been previously utilized for the high-throughput analysis of clinical strains. This approach allowed us to perform pooled competitive fitness assays, tracking strain fitness using deep sequencing. We measured the replicative fitness of the clinical strains in multiplicate under 14 metabolic and antibiotic stress condition. To our knowledge, this is the largest phenotypic screen of clinical isolates to date. We performed bacterial GWAS to delineate the genetic variants associated with each fitness phenotype, identifying monogenic associations with several conditions. We then defined phenotypic and genetic features associated with clinical outcomes. We find that a subclade of strains, defined by variants largely involved in fatty acid metabolic pathways, share a universal slow growth phenotype that is associated with cavitary disease, treatment failure and increased transmission potential in Vietnam. We also find that mutations in , a poorly characterized phosphodiesterase, also associate with slow growth and with treatment failure in patients. Phenogenomic profiling demonstrates that strains exhibit distinct growth characteristics under metabolic and antibiotic stress conditions. These fitness profiles can serve as intermediate traits for GWAS and association with clinical outcomes. Intermediate phenotyping allows us to examine potential processes by which bacterial strain differences contribute to clinical outcomes. Our study identifies clinical strains with slow growth phenotypes under models of antibiotic and host-like metabolic conditions that are associated with adverse clinical outcomes. It is possible that the bacterial intermediate phenotypes we identified are directly related to the mechanisms of these outcomes, or they may serve as markers for the causal yet unidentified bacterial determinants. Via the intermediate phenotyping, we also discovered a surprising diversity in responses to the new anti-mycobacterial drugs that target central metabolic processes, which will be important in considering roll-out of these new agents. Our study and others that have identified determinants of TB clinical and epidemiological phenotypes should inform efforts to improve diagnostics and drug regimen design.