Benefit from decline: the primary transcriptome of Alteromonas macleodii str. Te101 during Trichodesmium demise

Interactions between co-existing microorganisms deeply affect the physiology of the involved organisms and, ultimately, the function of the ecosystem as a whole. Copiotrophic Alteromonas are marine gammaproteobacteria that thrive during the late stages of phytoplankton blooms in the marine environment and in laboratory co-cultures with cyanobacteria such as Trichodesmium. The response of this heterotroph to the sometimes rapid and transient changes in nutrient supply when the phototroph crashes is not well understood. Here, we isolated and sequenced the strain Alteromonas macleodii str. Te101 from a laboratory culture of Trichodesmium erythraeum IMS101, yielding a chromosome of 4.63 Mb and a single plasmid of 237 kb. Increasing salinities to ≥43 ppt inhibited the growth of Trichodesmium but stimulated growth of the associated Alteromonas. We characterized the transcriptomic responses of both microorganisms and identified the complement of active transcriptional start sites in Alteromonas at single-nucleotide resolution. In replicate cultures, a similar set of genes became activated in Alteromonas when growth rates of Trichodesmium declined and mortality was high. The parallel activation of fliA, rpoS and of flagellar assembly and growth-related genes indicated that Alteromonas might have increased cell motility, growth, and multiple biosynthetic activities. Genes with the highest expression in the data set were three small RNAs (Aln1a-c) that were identified as analogs of the small RNAs CsrB-C in E. coli or RsmX-Z in pathogenic bacteria. Together with the carbon storage protein A (CsrA) homolog Te101_05290, these RNAs likely control the expression of numerous genes in responding to changes in the environment.


Library preparation, read cleaning and taxonomic classification
Details of the RNA isolation and sampling were described previously (Pfreundt et al., 2014;Pade et al., 2016). Primary transcriptomes were inferred by the genome-wide mapping of transcription start sites (TSSs). For this aim, differential RNA-Seq (dRNA-Seq) (Sharma et al., 2010) was used, in which the primary transcripts resulting from the initiation of transcription are selectively sequenced. This approach relies on the 5'P-dependent terminator exonuclease (TEX) activity, which specifically degrades processed transcripts while primary transcripts with their 5'triphosphates are kept intact. Recently, we have extended this approach to the microbial community sampled from the Red Sea and identified the suite of active TSSs from five different organisms representing all three domains of life, showing the potential of this approach in a complex microbial community context (Hou et al., 2016).

Prediction of TSSs
For the bioinformatic analysis of dRNA-Seq data and TSS prediction we applied a replicateassisted background subtraction algorithm. Reads from all libraries were aligned to the Alteromonas Te101 genome at 99% identity using segemehl v0.2.0 (Hoffmann et al., 2009), alignments were then converted to Artemis (Rutherford et al., 2000) compatible tabular files (GRP format) with genome coordinates, number of reads starting at each nucleotide position ("number of reads starting", NRS), and per-nucleotide-coverage using GRPutils were called iTSS and aTSS, respectively. TSS located in intergenic regions or upstream of an ncRNA including rRNA and tRNA genes were designated as non-coding TSS (nTSS). Finally, the original alignments were mapped and aggregated to these bona fide TSS positions to get the raw read counts initiated from these positions. The implementation of this algorithm can be found at https://github.com/housw/GRPutils/blob/master/tss_analysis_pipline.sh.

Differential expression analysis
For differential expression analysis, the TSS count table obtained from the TSS prediction was filtered for lowly expressed TSS (<10 reads from the 4 libraries combined) and TSS associated with rRNAs and tRNAs and then normalized using the Trimmed Mean of M-values (TMM) method in edgeR v3.14.0 (Robinson et al., 2010). Dispersions were estimated by treating samples from 37 ppt and 43 ppt as replicates using the quantile-adjusted conditional maximum likelihood method (qCML) and differentially expressed TSS between 43 ppt and 37 ppt were called using the exactTest function in edgeR with an adjusted p-value cutoff of 0.05.
To characterize the functions of genes up-regulated at 43 ppt, GO terms were annotated for the whole proteome using Blast2GO v4.0.7 (Conesa et al., 2005). All the determined differentially expressed TSSs were selected, then the genes driven by gTSS, the genes downstream of nTSS, iTSS and aTSS within 1 kb were extracted as the query gene set. Enriched GO terms were determined using hypergeometric tests implemented in GOstats v2.38.1 (Falcon and Gentleman, 2007) with the GO terms of the whole proteome as background. The multiple test corrections were performed using qvalue v2.4.2 (available at http://github.com/jdstorey/qvalue) in R.
Enriched GO terms were semantically clustered using the REVIGO (Supek et al., 2011) online server. to the one used in a previous study (Naghdi et al., 2017), which looks for a GGA motif 4 to 70 nt upstream of the Shine-Dalgarno (SD) sequence within 5'UTRs. Genes without SD sequences 5 5 to 15 nt upstream of translation start sites were considered as targets when ≥3 "GGA" were found in their 5'UTR and the normalized frequency of "GGA" per 100 nt was ≥3. To include genes without gTSSs, we also applied the CSRA_TARGET (Kulkarni et al., 2014) algorithm to genome-wide scan intergenic regions 300 nt upstream of and 50 nt following the ORF start.

CsrA target prediction and motif finding
To detect regulatory motifs in promoter regions, sequences 200 nt upstream of predicted TSSs were extracted and submitted to the XXmotif web server (Luehr et al., 2012) with default parameters except no masking, the E-value cutoff for trusted identified motifs was set to 0.001. Figure S1. Spectral scan of pigment absorbance of the Trichodesmium cultures on day 9.

Supplementary Figures
Cells were collected on GFF and pigments were extracted using 90% boiling methanol (de Marsac and Houmard, 1988). Pigment absorbance scans were analyzed with Cary 300 spectrophotometer (Agilent Technologies) between 520-760 nm in intervals of 1 nm.     TSSs were identified. The sequences within the dark blue square were used to generate the motif profile, which was further used to scan the A. macleodii Te101 genome to detect the other occurrences using FIMO (Grant et al., 2011) from MEME Suite v4.12.0 (Bailey et al., 2015).
The probabilities and adjusted q-values of each occurrence were shown in the right columns.