TEAL-Seq: targeted expression analysis sequencing

Abstract

Metagenome sequencing enables the genetic characterization of complex microbial communities. However, determining the activity of isolates within a community presents several challenges, including the wide range of organismal and gene expression abundances, the presence of host RNA, and low microbial biomass at many sites. To address these limitations, we developed "targeted expression analysis sequencing" or TEAL-seq, enabling sensitive species-specific analyses of gene expression using highly multiplexed custom probe pools. For proof of concept, we targeted about 1,700 core and accessory genes of Staphylococcus aureus and S. epidermidis, two key species of the skin microbiome. Two targeting methods were applied to laboratory cultures and human nasal swab specimens. Both methods showed a high degree of specificity, with >90% reads on target, even in the presence of complex microbial or human background DNA/RNA. Targeting using molecular inversion probes demonstrated excellent correlation in inferred expression levels with bulk RNA-seq. Furthermore, we show that a linear pre-amplification step to increase the number of nucleic acids for analysis yielded consistent and predictable results when applied to complex samples and enabled profiling of expression from as little as 1 ng of total RNA. TEAL-seq is much less expensive than bulk metatranscriptomic profiling, enables detection across a greater dynamic range, and uses a strategy that is readily configurable for determining the transcriptional status of organisms in any microbial community.IMPORTANCEThe gene expression patterns of bacteria in microbial communities reflect their activity and interactions with other community members. Measuring gene expression in complex microbiome contexts is challenging, however, due to the large dynamic range of microbial abundances and transcript levels. Here we describe an approach to assessing gene expression for specific species of interest using highly multiplexed pools of targeting probes. We show that an isothermal amplification step enables the profiling of low biomass samples. TEAL-seq should be widely adaptable to the study of microbial activity in natural environments.

Keywords: metagenomics; skin microbiome; targeted sequencing.

Fig 1

Layout of the experimental design. The experimental variables used for evaluating the targeted RNA-seq approaches are shown, illustrating how RNA samples were processed for analysis to determine the optimal TEAL-seq method, outlined in red.

Fig 2

Distribution of the number of probes per gene in the targeted strains. (A) Probes target >50% Sa and Se CDS in SPE with at least one unique probe. (B) Probes target >50% Sa and Se CDS in MIP with at least one unique probe. (C) A high proportion of reads mapped on target for MIP, SPE, and Bulk RNA-seq without significant differences for RNA input amounts ranging from 10 ng to 100 ng as well as 1 ng with SPIA treatment. (D) A high proportion of reads mapped on target when tested on microbial gDNA controls and gDNA with obscuring mouse stool DNA or human gDNA.

Fig 3

Evaluation of reproducibility using a 50:50 mixture of Sa and Se RNA. (A) Technical replicates (n = 2) are highly correlated for bulk RNA-seq, MIP, and SPE. (B) Samples with and without SPIA have a good correlation for bulk RNA-seq and MIP, but less so for SPE. (C) Samples with low (1 ng) and standard (10 ng) input total RNA amounts correlate well in MIP and SPE. Correlation coefficients (R) and P-values from Pearson tests.

Fig 4

Probe performance varies and correlates with melting temperature in SPE data more so than MIP data. (A) Sequencing of genomic DNA with SPE shows a range of CPMs for Sa (black) and Se (gray) probes. (B) Probe CPM correlates with melting temperature in SPE data for both Sa (black) and Se (gray). (C) CPM has minimal correlation with melting temperature in MIP data for both Sa (black) and Se (gray).

Fig 5

Sources of variation in targeted RNA-seq. (A) Comparison of CDS-level expression in bulk RNA-seq and MIP (targeted) without SPIA (black) and with SPIA (grey) for Sa and Se. (B and C) MIP with and without RNA depletion without SPIA (black) and with SPIA (gray) for Sa and Se. (D and E) MIP with and without SPIA amplification with ribo-depletion (black) and with ribo-depletion (gray) for Sa and Se. (F and G) Quantification of the sum of squared error (SSE) across comparisons for Sa and Se. (H) Correlation coefficients (R) and P values from Pearson tests. ****P < 0.0001, Wilcoxon test.

Fig 6

TEAL-seq detects acid stress responses consistent with previous studies using bulk RNA-seq. (A) Differential expression between TSB pH 7 and pH 4.8, as determined by |median probe log₂FC| > 2 and median adjusted P value < 0.05 per gene in TEAL-seq, treatment shows upregulated (red) and downregulated (blue) regulated genes consistent with published bulk RNA-seq studies for Sa and Se (B) in similar conditions. In significant overrepresentation by hypergeometric test, 53 Sa genes were downregulated, and 39 Sa genes were upregulated by both bulk RNA-seq and TEAL-seq. Also in significant overlap, 59 Se genes were up-regulated and 26 Se genes were downregulated by both bulk RNA-seq and TEAL-seq. (C) Urease and acid response probes across conditions are consistent across CDSs for Sa and Se. Each column represents biological replicates (n = 3). (D) Sets of upregulated and downregulated acid response genes from TEAL-seq are over-represented by sets from published bulk RNA-seq studies of similar conditions. * adj. P < 0.05,** adj. P < 0.001 by hypergeometric test with Bonferroni correction.

Fig 7

Genes previously implicated in colonization are heterogeneously upregulated in Sa and Se when grown on RHE and as sampled in nasal swabs. (A) PCAs separate expression from growth on RHE and expression from human nasal swabs from in vitro (TSB) in Sa and Se (B) disparately. (C) Sample correlations are driven by source Sa and Se (D) with different RHE samples forming subgroups. Values are pairwise distances in Pearson correlation matrices. (E) Genes important for colonization and urease genes are induced on RHE and in nasal swabs with substantial heterogeneity in Sa and Se (F).