Managing soil health is crucial for the long-term fertility and ecological integrity of agricultural lands. Soil health encompasses various soil properties contributing to agroecosystem function, including nutrient and water cycling, biodiversity, and pathogen suppression. Biological, physical, and chemical indicators are used to monitor soil health, but ideally, these indicators should be directly linked to soil function, interpretable, and dynamically responsive to management practices. The soil microbiome, with its sensitivity to management practices, offers significant potential as a source of such indicators. Microbial communities are highly sensitive to management practices and shape properties determining soil health. The diversity of soil bacteria provides valuable information about soil conditions, recently used to predict soil health status. However, limited understanding of the ecology and function of most soil bacteria hinders the interpretation of bacterial bioindicators. This research bridges the gap between soil microbial ecology and soil health by leveraging genomic data and amplicon sequencing data to understand the associations of bacterial communities with soil health. Genomic data, particularly traits like genome size and rrn copy number, can offer ecological insights based on the evolutionary trade-offs influencing these traits. While promising, genomic inference has limitations as many active soil microorganisms lack representative genomes. Environment-wide association studies (EWAS) offer a complementary approach, profiling phylogenetic gene markers across numerous amplicon sequencing projects to infer ecological information without prior knowledge. An EWAS approach is primarily limited by metadata quality and standardization in sequencing workflows, though the large volume of available sequencing projects partially compensates for these issues. This study aims to identify and characterize bacterial bioindicators of soil properties relevant to soil health assessment using a large amplicon sequencing survey and genomic inference.

The introduction extensively reviews existing literature on soil health assessment, emphasizing the importance of integrating microbial ecology with soil health monitoring. It cites several studies demonstrating the sensitivity of soil microbiomes to various agricultural practices and the potential of microbial communities as indicators of soil health. The review also highlights the challenges in interpreting bioindicator responses due to limited understanding of microbial ecology and function. Furthermore, it discusses existing methods for ecological inference from genomic and amplicon sequencing data, including the strengths and weaknesses of each approach. The studies cited cover various aspects of soil health assessment, including the development of comprehensive assessment frameworks (e.g., CASH framework), the use of specific microbial indicators, and the application of genomic and metagenomic approaches to understand microbial contributions to soil function.

This study used a two-pronged approach. First, it employed 16S rRNA gene sequencing data from 778 soil samples across the USA, representing diverse cropping systems and management practices, as part of a larger soil health initiative. The soil properties were assessed using the Comprehensive Assessment of Soil Health (CASH) framework, which includes biological, chemical, and physical indicators. Bacterial community composition was determined using Illumina MiSeq sequencing of the V4 region of the 16S rRNA gene. Data processing involved demultiplexing, filtering, trimming, chimera removal, and operational taxonomic unit (OTU) assignment. Bacterial OTUs correlated with soil health ratings were identified using Spearman rank correlations, with weak correlations removed. Indicator species analysis identified bioindicators for tillage intensity. Second, the study used inferred genomic traits (genome size, coding density, rrn copy number, CRISPR arrays, and biosynthetic gene clusters) from available databases to characterize the bioindicators. Community-weighted average trait values were calculated for whole communities. Finally, an environment-wide association study (EWAS) was conducted using data from 89 studies totaling 14,780 16S rRNA gene amplicon libraries. OTUs identified as significant EWAS indicators of study factors were used to calculate community-weighted averages for broad categories (management practice, disturbance, plant association, biome). Statistical analyses included PERMANOVA, and assessment of relative importance of community-weighted traits and EWAS in explaining community composition and variation in health scores. Co-occurrence networks were also constructed to visualize relationships among bacterial taxa.

The study identified numerous bacterial OTUs (8.7%) as bioindicators of soil health, with most positively correlated with biological ratings and negatively correlated with physical and chemical ratings. Many bioindicators belonged to candidate groups or unclassified genera. Higher soil health ratings were associated with smaller genome size, higher coding density, and lower rrn copy number. Community-weighted genome size was the best predictor of total health score. The EWAS analysis revealed that key bioindicators were associated with environmental disturbances, particularly tillage. Several genomic traits (genome size, CRISPR array frequency, and number of BGCs) were negatively correlated with total health scores. The relationships between these traits in the community-weighted data partly reflected those observed in the genomic database, but with some contrasting trends. Community-weighted CRISPR array frequency showed strong correlations with various health ratings, while coding density was positively correlated with total health score and organic matter quality. Community-weighted rrn copy number was significantly higher in tilled soils and correlated with surface and subsurface hardness. The EWAS revealed that community-weighted EWAS explained more variation in bacterial community composition than community-weighted genomic traits, but genome size explained more variation in total health score. Analysis of the taxa driving the relationship between community-weighted genome size and active carbon showed that taxa with larger genomes had higher relative abundance in soils with low active carbon ratings and were associated with tilled soils and bulk soil rather than rhizosphere soil. Some apparent contradictions emerged, such as “Chthoniobacter” strains having larger genomes and being more prevalent in tilled soils yet indicative of high biological ratings, possibly indicating niche differentiation or the effects of other soil characteristics. Bioindicators of physical and chemical health ratings increased in relative abundance at lower soil health ratings, potentially indicating stress-tolerant taxa. Bioindicators of biological health ratings increased in relative abundance at higher soil health ratings, possibly suggesting enrichment of organoheterotrophs in soils with higher organic matter.

The findings highlight the utility of integrating genomic traits and EWAS approaches for understanding the ecological basis of soil microbiome-soil health relationships. The strong association of community-weighted genome size with soil health ratings suggests that genome size is a key trait reflecting microbial adaptation to soil conditions. The contrasting responses of various bacterial taxa to tillage and other disturbances reveal complex interactions within the soil microbiome. The study's findings support the hypothesis that soil health is closely linked to microbial community composition and function, and that genomic traits can provide valuable insights into these relationships. The discrepancies observed between expected and observed correlations, such as those for coding density and CRISPR array abundance, underscore the need for further research to understand the complex interplay of factors influencing soil microbial communities. The results highlight the potential for using microbial community composition as a diagnostic tool to assess soil health and guide management practices aimed at improving soil health and function.

This study demonstrates the value of combining genomic traits and EWAS to understand the ecological roles of bacteria in soil health. Community-weighted genome size emerged as the strongest predictor of overall soil health, and the results reveal complex relationships between microbial community composition, genome traits, and various soil properties. Future research should focus on validating these findings using shotgun metagenomics and investigating the functional mechanisms linking genome size and carbon cycling to soil health. The study provides a valuable framework for assessing the ecological attributes of soil bacteria, including unclassified taxa, contributing to a more comprehensive understanding of soil microbiome-soil health dynamics and informing sustainable agricultural management practices.

The study's reliance on inferred genomic traits from reference genomes and low phylogenetic resolution for many taxa could limit the accuracy of trait assignments. The EWAS analysis is dependent on the quality and availability of metadata from various studies, and biases inherent in the datasets used could influence the results. Furthermore, the study primarily focuses on correlational relationships; it doesn't directly establish causality between microbial community composition and soil health. The relatively large number of unclassified taxa, while highlighting the gap in our knowledge, may also limit the inferences we can draw about specific functional groups.