br Authors contributions br Acknowledgements br Introduction
Introduction Retinoic 8-Azido-ATP is a key signalling molecule in healthy development and in differentiation of stem cells, albeit uncontrolled levels of retinoic acid can lead to mutagenesis . In eukaryotes retinoic acid is made by oxidation of all-trans-retinal by retinal/aldehyde dehydrogenases (ALDH1A1, ALDH1A2, ALDH1A3, ALDH8) . This step is preceded by the production of retinal from either retinol or beta-carotene (Fig. 1a). As yet, this canonical signalling pathway has only been characterised in animals. Retinoic acid has more recently been identified in cyanobacteria both grown in culture and in blooms isolated in China, where it is thought to be responsible for the mutagenesis of frogs [, , , , , , ] Thus far, no mechanism for cyanobacterial retinoic acid production has been defined, beyond it being a potential degradation product of retinal . Orthologues of carotenoid dioxygenases, used by humans to produce retinal from beta-carotene, have previously been characterised from the cyanobacteria Synechocystis PCC 6803 and Nostoc PCC 7107 [, , , ]. These orthologues are able to convert beta-carotene and apo-8-carotenal into all-trans-retinal. An orthologue of a cytochrome P450 enzyme known as CYP26 in animals and CYP120 in cyanobacteria has also been characterised and found to oxidise retinoic acid [, , ]. All-trans-retinal and the precursor beta-carotene have been identified in 24 species of cyanobacteria . Production of all-trans-retinal, beta-carotene or retinoic acid in cyanobacteria occurs independently of whether the species is grown in culture or isolated from a bloom [4,18]. Therefore, given the conservation of genes involved in the synthetic pathway, one hypothesis is that the retinoic acid produced by cyanobacteria could be produced by the same biosynthetic pathway as in eukaryotes, potentially starting from beta-carotene (Fig. 1a). Attempts have previously been made to identify a cyanobacterial aldehyde dehydrogenase that is able to convert all-trans-retinal to retinoic acid, bridging the gap between the previous characterised carotenoid dioxygenase and CYP120 . This Synechocystis PCC 6803 slr0091 protein was able to convert long chain aldehydes into their respective acids but had no activity against retinal. We have previously postulated that retinoic acid is not of animal origin as previously thought, but may be a product of bacterial origin and present in humans through a lateral gene transfer event of aldehyde dehydrogenase and CYP120 from cyanobacteria to animals . Previously, lateral gene transfers between cyanobacteria and animals, in particular of cyclooxygenases , calpains  and WD40 domains have been suggested , indicating that these events are likely to have shaped processes across the animal kingdom. Our phylogenetic analysis identified orthologues of human ALDH1A1 in cyanobacteria that have a very high sequence conservation (>60%) and retain the catalytic GQCC motif. Herein we characterise in vitro a cyanobacterial ALDH and demonstrate that it can convert retinal to retinoic acid.
Results A good model for the study of retinoic acid production in cyanobacteria would have orthologues of all the proteins in the putative pathway. Therefore, we searched the NCBI database of non-redundant protein sequences for a cyanobacterium that contains orthologues of human ALDH1A2 (BAA34785.1), beta-carotene monooxygenase (AAI26211.1) and cytochrome P450 CYP26A1 (AAB88881.1). One of 65 cyanobacteria identified that possess all three orthologues (Table S1), was Chlorogloeopsis fritschii PCC 6912. To confirm that the C. fritschii ALDH orthologue also appears to have been involved in lateral gene transfer from cyanobacteria to animals we undertook phylogenetic analysis. Sequences were chosen from Refs. [20,24], alongside the sequence of the previously studied Synechocystis sp. PCC 6803 ALDH (Slr0091) and ALDH from Bacillus subtilis which is the first bacterial aldehyde dehydrogenase to be identified that can convert retinal to retinoic acid . Phylogenetic analysis indicated that the C. fritschii ALDH orthologue (henceforth cfALDH), which has 64% sequence identity with human ALDH1A1, groups with other previously identified cyanobacterial proteins (Fig. 1b and Supplementary Fig. 1). When the slr0091 protein from Synechocystis PCC 6803 is included in the phylogenetic analysis (Fig. 1b), this protein does not group with any of the other cyanobacterial proteins but rather groups with human ALDH8. This is consistent with the lack of function seen with the purified slr0091 . The Bacillus cereus ALDH in our phylogenetic analysis shows a closer relationship with yeast ALDH and mammalian ALDH8 then it does with the cyanobacterial ALDH .