Characterization of sterol synthesis in bacteria

Sterols are essential components of eukaryotic cells whose biosynthesis and function in eukaryotes has been studied extensively. Sterols are also recognized as the diagenetic precursors of steranes preserved in sedimentary rocks where they can function as geological proxies for eukaryotic organisms and/or aerobic metabolisms and environments. However, production of these lipids is not restricted to the eukaryotic domain as a few bacterial species also synthesize sterols. Phylogenomic studies have identified genes encoding homologs of sterol biosynthesis proteins in the genomes of several additional species, indicating that sterol production may be more widespread in the bacterial domain than previously thought. Although the occurrence of sterol synthesis genes in a genome indicates the potential for sterol production, it provides neither conclusive evidence of sterol synthesis nor information about the composition and abundance of basic and modified sterols that are actually being produced. Here, we coupled bioinformatics with lipid analyses to investigate the scope of bacterial sterol production. We identified oxidosqualene cyclase (Osc), which catalyzes the initial cyclization of oxidosqualene to the basic sterol structure, in 34 bacterial genomes from 5 phyla (Bacteroidetes, Cyanobacteria, Planctomycetes, Proteobacteria and Verrucomicrobia) and in 176 metagenomes. Our data indicate that bacterial sterol synthesis likely occurs in diverse organisms and environments and also provides evidence that there are as yet uncultured groups of bacterial sterol producers. Phylogenetic analysis of bacterial and eukaryotic Osc sequences revealed two potential lineages of the sterol pathway in bacteria indicating a complex evolutionary history of sterol synthesis in this domain. We characterized the lipids produced by Osc-containing bacteria and found that we could generally predict the ability to synthesize sterols. However, predicting the final modified sterol based on our current knowledge of bacterial sterol synthesis was difficult. Some bacteria produced demethylated and saturated sterol products even though they lacked homologs of the eukaryotic proteins required for these modifications emphasizing that several aspects of bacterial sterol synthesis are still completely unknown. It is possible that bacteria have evolved distinct proteins for catalyzing sterol modifications and this could have significant implications for our understanding of the evolutionary history of this ancient biosynthetic pathway.

of steranes preserved in sedimentary rocks where they can function as geological proxies for 23 eukaryotic organisms and/or aerobic metabolisms and environments. However, production of 24 these lipids is not restricted to the eukaryotic domain as a few bacterial species also synthesize 25 sterols. Phylogenomic studies have identified genes encoding homologs of sterol biosynthesis 26 proteins in the genomes of several additional species, indicating that sterol production may be 27 more widespread in the bacterial domain than previously thought. Although the occurrence of 28 sterol synthesis genes in a genome indicates the potential for sterol production, it provides 29 neither conclusive evidence of sterol synthesis nor information about the composition and 30 abundance of basic and modified sterols that are actually being produced. Here, we coupled 31 bioinformatics with lipid analyses to investigate the scope of bacterial sterol production. We 32 identified oxidosqualene cyclase (Osc), which catalyzes the initial cyclization of oxidosqualene 33 to the basic sterol structure, in 34 bacterial genomes from 5 phyla (Bacteroidetes, Cyanobacteria,

34
Planctomycetes, Proteobacteria and Verrucomicrobia) and in 176 metagenomes. Our data indicate that 35 bacterial sterol synthesis likely occurs in diverse organisms and environments and also provides 36 evidence that there are as yet uncultured groups of bacterial sterol producers. Phylogenetic 37 analysis of bacterial and eukaryotic Osc sequences revealed two potential lineages of the sterol 38 pathway in bacteria indicating a complex evolutionary history of sterol synthesis in this domain. We 39 characterized the lipids produced by Osc-containing bacteria and found that we could generally 40 predict the ability to synthesize sterols. However, predicting the final modified sterol based on our current 41 knowledge of bacterial sterol synthesis was difficult. Some bacteria produced demethylated and 42 saturated sterol products even though they lacked homologs of the eukaryotic proteins required 43 Introduction of the organisms we tested were capable of sterol production under laboratory conditions. 111 Through these studies, it is evident that sterol production is more widespread in the bacterial 112 domain than previously thought and that the bacterial sterol biosynthetic pathway has a complex 113 evolutionary history.

174
Oxidosqualene cyclases catalyze the conversion of oxidosqualene to lanosterol in vertebrates and 175 fungi and to cycloartenol in land plants (Desmond and Gribaldo, 2009). Deletion of this protein 176 in yeast completely blocks sterol production and therefore its occurrence in a genome is a good 177 indicator of sterol production by an organism (Lees et al., 1995). BLASTP analysis recovered 34 178 bacterial Osc homologs in five different bacterial phyla with an e-value equal to or lower than e -179 100 and greater than 30% similarity (Table 2). As expected, Osc homologs are found in the 180 genomes of five organisms that have been previously shown to produce sterols: M. capsulatus 181 Bath, N. exedens, Cystobacter fuscus, Stigmatella aurantiaca and G. obscuriglobus (Table 2).

182
The myxobacterium Corallococcus coralloides also contains an Osc homolog, however, a 183 previous study of myxobacterial species did not detect any sterols in this bacterium (Bode et al.,  The 34 bacterial species with Osc homologs in their genomes were isolated from a 210 variety of environments indicating that bacterial sterol producers are not restricted to a specific 211 ecological niche ( Table 2). The majority of the myxobacterial sterol producers were acquired  Osc metagenome sequences were identified with an e-value of e -50 or lower. The majority of the 224 metagenomic sequences were from soil, marine or freshwater environments similar to the 225 distribution of isolate environments described above (Figure 4). In addition, Osc sequences were 226 found in metagenomes from estuarine microbial mats, hydrothermal vent fluids and two 227 sequences from sponge symbionts.

228
To distinguish bacterial from eukaryotic Osc sequences in the metagenomes, we 229 generated a maximum likelihood phylogenetic tree of metagenomic and genomic Osc homologs.  Thirty-seven of the Osc metagenomics sequences retrieved clustered within the two 237 bacterial Osc clades ( Figure 5). Some of these sequences grouped with known sterol producers 238 like the Methylococcales and Myxococcales. However, some of these sequences formed their 239 own clades within the bacterial groups or clustered with organisms that have yet to be shown to 240 produce sterols. Thus, identification of these bacterial Osc homologs in metagenome datasets 241 indicates that there are novel sterol-producing bacteria yet to be discovered and that the sterol 242 producing bacteria inhabit diverse environments. Our bioinformatics analysis of metagenomic 243 databases did identify 18 eukaryotic Osc sequences ( Figure 6) which were related to algal, plant 244 or fungal Osc homologs. Given how widespread sterol synthesis is in eukaryotes, we had 245 expected to detect more eukaryotic Osc sequences than bacterial sequences in metagenomic 246 databases. However, it has been documented that metagenomic sequencing tends to recover few 247 eukaryotic sequences in general (Lindahl and Kuske, 2013). Therefore, the low number of eukaryotic metagenomic sequences more likely reflects a limited number of eukaryotic 249 sequences in metagenome databases rather than the true prevalence of eukaryotic sterol 250 producers in the environment.

251
Lipid analysis of potential sterol producers 252 Our identification of Osc homologs in bacterial genomes demonstrates that the potential 253 for sterol synthesis exist in a variety of bacteria. However, the majority of these potential sterol-254 producing bacterial strains have not been tested for sterol production. In addition, the occurrence 255 of Osc in a genome only suggests the production of the most basic sterols, lanosterol or 256 cycloartenol. Thus, lipid analysis is needed not just to verify sterol production but also to 257 determine if and how sterols are modified in bacteria. We performed lipid analysis on 11 Osc-258 containing bacteria that included five myxobacteria, four Methylococcales, one Bacteriodetes 259 and one α-Proteobacterium (Table 1). In addition, we searched the genomes of these 11 260 organisms for other sterol biosynthesis protein homologs. Our goal was to link the occurrence of 261 these downstream biosynthesis genes with any sterol modifications, such as saturations and 262 demethylations, these bacteria may be carrying out.

263
Sterol production in the myxobacteria 264 Four of the five myxobacterial strains tested were found to produce sterols (Table 3). C. 265 fuscus strains were previously reported to produce either lanosterol or cycloartenol (Bode et al.,266 2003) and the C. fuscus strain we analyzed produced cycloartenol. We identified homologs for 267 C-14 demethylation and C-24 reduction in the C. fuscus genome but did not observe any sterols 268 with these modifications (Table 4). The other three myxobacteria, E. salina, P. pacifica and S.  (Table 4). Thus, it is unclear how these myxobacterial strains are fully 280 demethylating at the C-4 position.

281
In agreement with a previous study, the myxobacterium C. coralloides produces 282 significant amounts of squalene but no sterol-like molecules despite having a copy of both 283 squalene epoxidase (SE), required for the conversion of squalene to oxidosqualene prior to 284 cyclization, and oxidosqualene cyclase in its genome (Table 4 and 295 The lipid profiles of the four Methylococcales species tested were similar to what was 296 previously observed in M. capsulatus Bath (Volkman, 2005), with some exceptions (Figure 9).

297
M. lacus did not saturate the sterol side chain at C-24 as would be predicted because it lacks a 298 homolog of the C-24(28) sterol reductase (ERG4 in yeast or DHCR24 in humans) (Table 4). M. 299 luteus, on the other hand, only produced sterols that were saturated at the C-24 position ( Figure   300 9). Interestingly, while all of the Methylococcales tested produced sterols that were partially 301 demethylated at the C-4 position, none had homologs of any of the eukaryotic C-4 demethylase 302 genes (Table 4). These methanotrophs also had sterols in which the unsaturation generated 303 during C-14 demethylation was subsequently removed even though they lack a homolog of the 304 C-14 reductase (ERG24) ( Table 4). This is in contrast to the previously tested M. alcaliphilum there may be more than one mechanism for this reaction even within the Methylococcales.

307
Sterol production in other bacterial species 308 We also observed production of cycloartenol in one Bacteriodetes species, F. taffensis, 309 and one α-Protebacterium, M. caenitepidi ( Figure 10 and Table 3). Neither of these strains had 310 homologs of sterol biosynthesis genes downstream of osc in their genomes and this was in 311 agreement with our observations of only cycloartenol production (Table 4).

312
Cycloartenol versus lanosterol synthesis is likely correlated with a single residue 313 The production of cycloartenol by some strains in our survey and lanosterol by others 314 prompted us to investigate if specific residues were indicative of whether a cyclase was a 315 lanosterol or cycloartenol synthase. Site-directed mutagenesis studies have previously identified 316 three amino acids changes that seem to control the product profile of oxidosqualene cyclases 317 (Meyer et al., 2000;Meyer et al., 2002;Lodeiro et al., 2004). Specifically, the amino acid residues 318 T381/C,Q449/V453 (numbering based on human Osc) were indicative of a lanosterol synthase 319 while Y381/H449/I453 suggested a cycloartenol synthase (Summons et al., 2006). Comparative 320 genomics of three bacterial cyclases with eukaryotic cyclases revealed that only one of these 321 residues was conserved and suggested that a valine (V) or isoleucine (I) at residue 453 suggested 322 lanosterol or cycloartenol production, respectively (Summons et al., 2006). Our lipid analyses 323 and alignments (Figure 8) verify that the bacterial oxidosqualene cyclases in the organisms we 324 tested completely correlated with the observation that a V453 was indicative of lanosterol 325 production while I453 signified cycloartenol production.