DSYB catalyses the key step of dimethylsulfoniopropionate biosynthesis in many phytoplankton

Dimethylsulfoniopropionate (DMSP) is a globally important organosulfur molecule and the major precursor for dimethyl sulfide. These compounds are important info-chemicals, key nutrients for marine microorganisms, and are involved in global sulfur cycling, atmospheric chemistry and cloud formation1–3. DMSP production was thought to be confined to eukaryotes, but heterotrophic bacteria can also produce DMSP through the pathway used by most phytoplankton4, and the DsyB enzyme catalysing the key step of this pathway in bacteria was recently identified5. However, eukaryotic phytoplankton probably produce most of Earth’s DMSP, yet no DMSP biosynthesis genes have been identified in any such organisms. Here we identify functional dsyB homologues, termed DSYB, in many phytoplankton and corals. DSYB is a methylthiohydroxybutryate methyltransferase enzyme localized in the chloroplasts and mitochondria of the haptophyte Prymnesium parvum, and stable isotope tracking experiments support these organelles as sites of DMSP synthesis. DSYB transcription levels increased with DMSP concentrations in different phytoplankton and were indicative of intracellular DMSP. Identification of the eukaryotic DSYB sequences, along with bacterial dsyB, provides the first molecular tools to predict the relative contributions of eukaryotes and prokaryotes to global DMSP production. Furthermore, evolutionary analysis suggests that eukaryotic DSYB originated in bacteria and was passed to eukaryotes early in their evolution. Identification of functional dsyB gene homologues for dimethylsulfoniopropionate production in eukaryotic phytoplankton allows estimation of the relative contributions of eukaryotes and prokaryotes to the global pool, and indicates that this enzyme originated in bacteria.


8
conditions, intracellular DMSP levels and DSYB transcription were relatively low, when 172 compared to (e.g.) P. parvum ( Supplementary Fig. 2). However, consistent with work in 173 other diatoms 18 , both F. cylindrus DMSP production and DSYB transcription increased with 174 nitrogen limitation and increased salinity ( Supplementary Fig. 2). The latter supports a role 175 for DMSP in osmoregulation and salinity-induced oxidative stress protection in F. cylindrus, 176 as suggested by Lyon et al. 7 . DSYB was not detected as one of the salinity-induced proteins 177 in Lyon et al. 7 , despite using the same salinity conditions for our experiments, reflecting the 178 nature of 2D gel electrophoresis studies, whereby not all proteins are identified. 179 Given the trend of intracellular DMSP concentration increasing with DSYB transcription, we 180 studied Symbiodinium microadriaticum CCMP2467, a dinoflagellate from a genus producing 181 high DMSP concentrations 6 . S. microadriaticum gave the highest intracellular DMSP (282 182 mM) and cumulative DSYB transcription of the tested phytoplankton ( Supplementary Fig. 2). 183 Similarly, available transcriptomic data showed that high DMSP-producing dinoflagellate 184 and haptophyte phytoplankton (see above) had the highest average DSYB transcription, which 185 was ~3 and 8-fold higher, respectively, than that in diatoms (Supplementary Table 2). 186 Transcriptomic data was also congruent with high variability in intracellular DMSP levels 187 within dinoflagellates and haptophytes 6,9 . While additional factors, such as DSYB protein 188 levels, DMSP excretion, DMSP catabolism and cell volume, will affect an organism's 189 intracellular DMSP concentration, the data presented here on a small number of 190 phytoplankton supports the hypothesis that DSYB transcription is a reasonably good indicator 191 of DMSP concentration. Some DSYB-containing phytoplankton may also contain MTHB 192 methyltransferase isoform enzymes or utilise other DMSP synthesis pathways, in which case 193 such predictions may be inaccurate. Further work is required to substantiate this hypothesis. 194 The prominence of environmental DMSP-producing bacteria and eukaryotes was examined 195 in the ocean microbial reference gene catalogue (OM-RGC) metagenomic dataset, generated 196 from samples fractionated to < 3 µm 26 (Supplementary Table 6 and Supplementary Fig. 6). 197 The dsyB gene was predicted to be present in 0.35% of total bacteria in these samples. For 198 comparison, DMSP lyase genes (dddD, dddL, dddK, dddP, dddQ, dddW, dddY and Alma1) 27 , 199 were also used. The dsyB gene was more abundant than dddL, dddW, dddY, and the algal 200 DMSP lyase gene Alma1, but was less abundant than dddD, dddK, dddP and dddQ in the 201 OM-RGC dataset. Despite only 3% of the OM-RGC microorganisms likely being 202 eukaryotes 26 , DSYB genes were detected and were ~25-fold less abundant than bacterial dsyB. 203 Since no DSYB sequences have been identified in bacteria, we conclude that picoeukaryotes 204 in these samples contain DSYB and thus, the genetic potential to make DMSP. The 205 production of DMSP by DSYB-containing picoeukaryotes could contribute, along with 206 DMSP-producing bacteria, to the DMSP measured from particles <2 µm in size in seawater 207 samples 28 . 208 We also investigated the occurrence of dsyB and DSYB in marine metatranscriptomes 209 (Supplementary Table 7). dsyB transcripts were detected in all tested Tara oceans 210 metatranscriptomic datasets apportioned to marine bacteria (Supplementary Table 8 and 211 Supplementary Fig. 6). dsyB transcript abundance (normalised to total sequence numbers) 212 was similar to dddD and greater than dddL, dddW, dddY and Alma1, but was far less than 213 dddK, dddP and dddQ. Although these datasets do not consider phytoplankton >3 µm, DSYB 214 transcripts, likely from picoeukaryotes, were detected at levels only 3-fold lower than the We also analysed the North Pacific Ocean metatranscriptomes (GeoMICS) which used 218 appropriate fractionation methods for bacteria and larger phytoplankton 29 . As expected, 219 eukaryotic DSYB transcript numbers were higher than those of bacterial dsyB in all of the 2-220 53 µm fractions, which should contain relatively more phytoplankton than bacteria, and the 221 opposite was true in most of the 0.2-2 µm fractions, which should have relatively more 222 bacteria but not contain the larger phytoplankton (Supplementary Table 9). Analysing data 223 from both the large and small size fractions at different sites allowed us to gauge the relative 224 total transcript numbers of DSYB and dsyB in these samples, as well as those of the DMSP 225 lyase genes. Prokaryotic dsyB transcripts (normalised to the recovery of an internal standard) 226 were more abundant than those for the bacterial DMSP lyase genes dddK, dddL, dddQ, dddY 227 and dddW, 3-fold less than dddP and Alma1 and 27-fold less than dddD (Supplementary 228 The DSYB gene from P. parvum CCAP946/6 was PCR-amplified from cDNA and cloned 291 into the IPTG-inducible wide host range expression plasmid pRK415 41 . All other DSYB 292 genes were synthesised by Eurofins Genomics, from sequences codon-optimised (using 293 Invitrogen GeneArt) for expression in E. coli, in the vector pEX-K4 (Eurofins Genomics).

294
The synthesised genes were then subcloned into pLMB509 42 , a taurine-inducible plasmid for 295 the expression of genes in Rhizobium and Labrenzia, using NdeI and BamHI or EcoRI 296 restriction enzymes. All plasmid clones are described in Supplementary Table 10.      showed that all standard was recovered following our extraction and measurement procedure.    Table 11) were designed, using 434 Primer3Plus 43 , to amplify ~130 bp region, with an optimum melting temperature of 60 °C.

435
Melting temperature difference between primers in a pair was 2 °C and GC content was kept 436 between 40-60%.

437
Quantitative PCR was performed with a C1000 Thermal cycler equipped with a CFX96 Real-        Statistical methods for RT-qPCR are described in the relevant section above. All 619 measurements for DMSP production or DSYB/DsyB enzyme activity (in algal strains or 620 enzyme assays) are based on the mean of at least three biological replicates per 621 strain/condition tested, with all experiments performed at least twice. To identify statistically 622 significant differences between standard and experimental conditions in Supplementary Fig.   623 2, a two-tailed independent Student's t-test (P<0.05) was applied to the data, using R 51 .

637
Each potential DsyB/DSYB sequence was manually curated by BLASTP analysis against the 638 RefSeq database and discounted as a true DSYB sequence if the top hits were not to ratified 639 DSYB sequences detailed in Fig. 1b. DSYB sequences identified from iMicrobe 640 transcriptomes were aligned to ratified DsyB and DSYB sequences and included in the 641 evolutionary analysis (Fig. 1b). All DsyB and DSYB protein sequences identified from   Table 7). Sequences were trimmed using TrimGalore (default parameters, 702 paired-end mode, https://www.bioinformatics.babraham.ac.uk/projects/trim_galore/) and 703 overlapping paired-end reads were joined using PandaSeq 80 . To create peptide databases, the 704 joined reads were translated using the translate function in Sean Eddy's squid package 705 (http://selab.janelia.org/software.html) to generate all ORFs above 20 amino acids in length.

706
The resulting peptide sequences were used to retrieve dsyB and DSYB sequences using HMM  Table 9) and ratios of dsyB/DSYB calculated.