BaM-seq and TBaM-seq, highly multiplexed and targeted RNA-seq protocols for rapid, low-cost library generation from bacterial samples

Abstract

The ability to profile transcriptomes and characterize global gene expression changes has been greatly enabled by the development of RNA sequencing technologies (RNA-seq). However, the process of generating sequencing-compatible cDNA libraries from RNA samples can be time-consuming and expensive, especially for bacterial mRNAs which lack poly(A)-tails that are often used to streamline this process for eukaryotic samples. Compared to the increasing throughput and decreasing cost of sequencing, library preparation has had limited advances. Here, we describe bacterial-multiplexed-seq (BaM-seq), an approach that enables simple barcoding of many bacterial RNA samples that decreases the time and cost of library preparation. We also present targeted-bacterial-multiplexed-seq (TBaM-seq) that allows for differential expression analysis of specific gene panels with over 100-fold enrichment in read coverage. In addition, we introduce the concept of transcriptome redistribution based on TBaM-seq that dramatically reduces the required sequencing depth while still allowing for quantification of both highly and lowly abundant transcripts. These methods accurately measure gene expression changes with high technical reproducibility and agreement with gold standard, lower throughput approaches. Together, use of these library preparation protocols allows for fast, affordable generation of sequencing libraries.

Figure 1.

BaM-seq library preparation workflow. (A) Fragmented RNA can be rapidly converted into barcoded cDNA libraries via a single-tube RT reaction with no intervening clean-up steps. RT products can be subsequently pooled, and downstream processing steps are performed as a single sample. Red arrows indicate clean-up steps. (B) The single-tube RT reaction, steps 1–3, involves dephosphorylation of RNA by T4 polynucleotide kinase, polyadenylation of 3’ ends by E. coli Poly(A) Polymerase, and reverse transcription by SSIII using barcoded oligo(dT) primers. Following pooling, an adapter is ligated to the 3’ end of cDNA molecule (step 4), and libraries PCR amplified (step 5).

Figure 2.

Validation of BaM-seq protocol. (A) Pearson correlation of log₁₀-transformed rpm values for genes with at least 100 mapped reads (212 genes) between 14 technical replicates. Minimum R-value = 0.983. The inset shows the cumulative distribution of median-normalized fold-changes for all pairs of genes between all pairwise combinations of replicate samples. (B) Representative example of rpm correlation between two replicates as in (A). Genes with more than 100 mapped reads in both samples are plotted. (C) 5’ mapped reads across fis and ahpC genomic loci in WT (153 and 1246 mapped reads, respectively), and Δfis (three mapped reads) and ΔahpC (10 mapped reads) E. coli strains. (D) Relative expression of genes as measured from three split RNA samples processed with BaM-seq or Rend-seq. Rpm was plotted for all genes with >100 reads in both samples.

Figure 3.

Overview of TBaM-seq protocol. (A) Primers for target-enrichment, each containing a 30-nt common adapter and 20-nt gene specific homology region. (B) Following single-tube RT, cDNA libraries can be enriched for transcripts of interest via a second-strand synthesis reaction with gene-specific primers as shown in (A) and subsequently PCR-amplified from the common adapter. (C) The percentage of reads mapping to 82 targeted genes in libraries prepared with either BaM-seq or TBaM-seq.

Figure 4.

Validation of TBaM-seq protocol. (A) Pearson correlation between replicates of log₁₀-transformed reads per second-strand primer. (B) Representative example of correlation across second-strand primers between replicates. A total of 162 second-strand primers targeting 82 genes were included in this pool. The second-strand primers that show >2-fold difference in rpm between these two samples target the genes polA (0.10× between sample 4 and 5), spo0E (0.20), lysC (0.28), and glnR (0.46). polA, spo0E, and lysC are the three lowest expressed genes targeted by this second-strand primer pool (as measured by non-targeted BaM-seq). (C) Reads per million primer-mapping reads for each second-strand primer between WT and Δrho strains. Primers targeting rho CDS and UTR, as well as those targeting sigB are highlighted. (D) lacZ expression relative to sample with highest lacZ induction. For BaM-seq samples, lacZ expression is calculated as rpkm. For TBaM-seq samples, lacZ expression is calculated from median normalized reads across 12 lacZ-targeting primers.