Coenzyme B₁₂-dependent and independent photoregulation of carotenogenesis across Myxococcales

Summary

Light-induced carotenogenesis in Myxococcus xanthus is controlled by the B₁₂-based CarH repressor and photoreceptor, and by a separate intricate pathway involving singlet oxygen, the B₁₂-independent CarH paralogue CarA and various other proteins, some eukaryotic-like. Whether other myxobacteria conserve these pathways and undergo photoregulated carotenogenesis is unknown. Here, comparative analyses across 27 Myxococcales genomes identified carotenogenic genes, albeit arranged differently, with carH often in their genomic vicinity, in all three Myxococcales suborders. However, CarA and its associated factors were found exclusively in suborder Cystobacterineae, with carA-carH invariably in tandem in a syntenic carotenogenic operon, except for Cystobacter/Melittangium, which lack CarA but retain all other factors. We experimentally show B₁₂-mediated photoregulated carotenogenesis in representative myxobacteria, and a remarkably plastic CarH operator design and DNA binding across Myxococcales. Unlike the two characterized CarH from other phyla, which are tetrameric, Cystobacter CarH (the first myxobacterial homologue amenable to analysis in vitro) is a dimer that combines direct CarH-like B₁₂-based photoregulation with CarA-like DNA binding and inhibition by an antirepressor. This study provides new molecular insights into B₁₂-dependent photoreceptors. It further establishes the B₁₂-dependent pathway for photoregulated carotenogenesis as broadly prevalent across myxobacteria and its evolution, exclusively in one suborder, into a parallel complex B₁₂-independent circuit.

Introduction

Light directs many key biological processes but can also cause photooxidative stress by generating highly reactive oxygen species (ROS), like singlet oxygen (¹O₂), which attack proteins, DNA and lipids (Ziegelhoffer and Donohue, 2009; Glaeser et al., 2011). Many organisms, including bacteria, combat photooxidative stress by inducing the synthesis of carotenoids, pigments that quench ROS (Armstrong, 1997; Ziegelhoffer and Donohue, 2009; Rodriguez-Concepción et al., 2018; Sandmann, 2019). How light is sensed and signalled to trigger carotenogenesis has been studied in depth in Myxococcus xanthus (Elías-Arnanz et al., 2011; Padmanabhan et al., 2021), a Gram-negative soil bacterium of the order Myxococcales, currently included within the class Deltaproteobacteria but it was recently proposed that this order be moved to a new and independent phylum with the name Myxococcata (Waite et al., 2020). Members of this order, the myxobacteria, form an important group of ubiquitous, predominantly aerobic, soil bacteria that share complex lifestyles and several traits typical of eukaryotes, such as multicellular development, biosynthesis of specialized lipids and steroids, social behaviour, kin recognition (the ability of an individual cell or organism to identify others as self-like), predation and motility (Muñoz-Dorado et al., 2016; Cao and Wall, 2019; Gallego-García et al., 2019; Hoshino and Gaucher, 2021). Given these eukaryotic-like traits in myxobacteria, the Syntrophy hypothesis envisages an ancient myxobacterium at the origin of eukaryotes (López-García and Moreira, 2020; Hoshino and Gaucher, 2021), and some phylogenetic signals supporting the theory can be traced to factors identified in M. xanthus light-induced carotenogenesis. Hence, exploring how this response and its unique factors are conserved can provide valuable molecular as well as evolutionary insights, and here we examined this across Myxococcales.

Two light sensing and signalling mechanisms control carotenogenesis in M. xanthus (Fig. 1A). One directly senses UV, blue or green light using a single protein, CarH_Mx (Mx for M. xanthus), which is the defining member of a large photoreceptor family whose chromophore is coenzyme B₁₂ or 5′-deoxyadenosylcobalamin (AdoCbl), a biological form of vitamin B₁₂ (Fig. S1, Supporting Information) (Ortiz-Guerrero et al., 2011; Jost et al., 2015b; Padmanabhan et al., 2017; Padmanabhan et al., 2019). The other more complex mechanism involves several factors including CarA_Mx, an AdoCbl-independent CarH_Mx paralogue and the extensive genetic rewiring that usually shapes the evolution of new regulatory networks (Perez and Groisman, 2009; Romero et al., 2012). It relies on ¹O₂ generated by blue light excitation of protoporphyrin IX, a photosensitizer and heme precursor (Fig. 1A) (Burchard and Dworkin, 1966; Galbis-Martínez et al., 2012). The protein known to act earliest in this pathway is CarF (Fontes et al., 2003), which we recently unmasked as the long-sought desaturase, widespread in metazoa but very rare in bacteria and required for the biosynthesis of plasmalogens, a special type of glycerophospholipids (Gallego-García et al., 2019). These lipids and ¹O₂, through a still unknown mechanism, provoke the release of the extracytoplasmic function σ (ECF-σ) factor CarQ from its membrane-bound anti-σ CarR (Galbis-Martínez et al., 2012; Gallego-García et al., 2019). CarQ then associates with RNA polymerase (RNAP) to turn on transcription of the carQRS operon (encoding CarQ, CarR and CarS, an antirepressor of CarA_Mx) and of the carotenogenic gene crtIb (Martínez-Argudo et al., 1998; Browning et al., 2003). This requires the CarD–CarG global regulatory complex, with some features reminiscent of multiprotein eukaryotic transcription complexes called enhanceosomes (Peñalver-Mellado et al., 2006; García-Heras et al., 2009; Elías-Arnanz et al., 2010).

Fig. 1

The two pathways for regulating M. xanthus light-induced carotenogenesis.

A. Left panel: B₁₂-dependent pathway. In the dark, CarH binds to its operator at P_crt (recognized by RNAP-σ^A) to repress expression of the carotenoid synthesis genes (crtE–orf9) of the carB operon. Light (UV/blue/green) disrupts CarH-operator binding to allow transcription from P_crt. Right panel: B₁₂-independent pathway. CarA represses P_crt in the dark and its action is counteracted by CarS, expressed when protoporphyrin IX (PPIX) exposed to blue light generates photoexcited PPIX (PPIX*) and ¹O₂. Plasmalogens (VEPE), whose biosynthesis requires CarF, are essential for inactivation of CarR by a mechanism that may involve VEPE cleavage by ¹O₂ to its lyso-PE form and a fatty aldehyde (Gallego-García et al., 2019) to free CarQ, which associates with RNAP to activate P_I and, together with the CarD–CarG complex, P_QRS.

B. Molecular mechanism of action of CarH and CarA. Top: Binding of AdoCbl to apoCarH (monomer or molten globule tetramer) forms active CarH repressor oligomers (tetramers: CarH_Tt, CarH_Bm; undefined oligomer: CarH_Mx). Photolysis of the AdoCbl Co-C bond liberates the Ado group as 4′, 5′-anhydroadenosine (unfilled red circles) (Jost et al., 2015a) and disrupts CarH oligomers to monomers (CarH_Tt, CarH_Mx) or dimers (CarH_Bm) with bound photolyzed AdoCbl (unfilled red triangles). Bottom: CarA dimers bind cooperatively to its operator, which overlaps with promoter elements at P_crt, to repress expression in the dark. Blue light induces expression of CarS, which sequesters the CarA DNA binding domain leading to antirepression.

Both light signalling pathways converge at P_crt (or P_B) (Fig. 1A), a promoter recognized by the primary σ^A-RNAP holoenzyme that drives expression of the carB operon, which encodes all the carotenoid synthesis enzymes except CrtIb, as well as CarH_Mx and CarA_Mx (López-Rubio et al., 2004; Pérez-Marín et al., 2008; Ortiz-Guerrero et al., 2011). CarH_Mx requires AdoCbl for oligomerization (its precise oligomeric state is unknown) and DNA binding to repress transcription from P_crt, and to be directly inactivated by light through the disassembly of the oligomer to monomers (Fig. 1B) (Pérez-Marín et al., 2008; Ortiz-Guerrero et al., 2011). In contrast to CarH_Mx, homologues from species in other bacterial phyla, Thermus thermophilus (CarH_Tt) and Bacillus megaterium (CarH_Bm), have been purified and characterized in vitro, and crystal structures of the dark (free or DNA-bound) and light-exposed forms have been resolved for CarH_Tt (Ortiz-Guerrero et al., 2011; Díez et al., 2013; Kutta et al., 2015; Jost et al., 2015a; Jost et al., 2015b; Fernández-Zapata et al., 2018; Miller et al., 2020). In the dark, both proteins form AdoCbl-bound tetramers that bind to operators comprising tandem 11-bp direct repeats, three for CarH_Tt or four for CarH_Bm. However, while apo and light-exposed holoCarH_Tt are both monomers, like CarH_Mx, apoCarH_Bm is a molten globule tetramer and light-exposed holoCarH_Bm is a dimer (Fig. 1B). By contrast, CarA_Mx is a dimer in the dark or light, with or without B₁₂, binds cooperatively to operator DNA as two dimers, and requires expression of CarS in the light for derepression (Fig. 1B) (López-Rubio et al., 2004; Navarro-Avilés et al., 2007; León et al., 2010; Ortiz-Guerrero et al., 2011). CarH_Mx and CarA_Mx thus differ in AdoCbl-dependence, oligomerization, and DNA binding modes to control the same promoter and target genes.

CarH homologues are widely distributed in several bacterial phyla, usually as a single copy (Padmanabhan et al., 2019). Hence, it is intriguing that CarH_Mx coexists with CarA_Mx, whose function has required coevolution of a cohort of other unique factors. Such duplications and divergence, and their retention or loss are of great interest in understanding transcription factor evolution (Hittinger and Carroll, 2007; Chapal et al., 2019; Kuzmin et al., 2020). We therefore examined conservation of carotenogenesis and of the two M. xanthus photosensory regulatory pathways in all three Myxococcales suborders: Cystobacterineae (includes M. xanthus), Sorangiineae and Nannocystineae. This established that: (i) genes for carotenoid biosynthesis enzymes and for CarH are present in all myxobacterial genomes examined except in Anaeromyxobacter dehalogenans and Vulgatibacter incomptus, both with reduced genomes; (ii) that carH in Cystobacterineae always occurs in a large syntenic cluster of carotenogenic genes that in some species also contains, interestingly, a gene for 3-hydroxy-3-methyl-glutaryl-CoA reductase or HMGR (a separate syntenic operon encodes the primary HMGR and other enzymes of the mevalonate or MVA pathway for the synthesis of isoprenoids, the precursors of carotenoids); (iii) even though carotenogenic genes are rearranged in Sorangiineae and Nannocystineae, carH frequently occurs in the genomic vicinity. We provide evidence supporting B₁₂-dependent photoregulation of carotenogenesis in representative myxobacteria from all three suborders and its likely mediation by CarH, whose operator design and DNA binding are, we show, remarkably plastic. The parallel CarA pathway and all its associated players appear strictly confined to suborder Cystobacterineae, among which the Cystobacter/Melittangium genera have CarH and all factors of the CarA pathway yet, surprisingly, not CarA. We purified and characterized Cystobacter CarH (the one myxobacterial homologue thus far amenable to analysis in vitro) and show that it differs from the tetrameric CarH_Tt and CarH_Bm, the only others that have been characterized in vitro and which belong to different phyla. Thus, Cystobacter CarH acts as a dimer with CarH-like B₁₂-dependent photosensory mode of action as well as CarA-like DNA binding and its antirepression by CarS. Altogether, this work reveals that B₁₂-based photoregulation of carotenogenesis pervades across Myxococcales and provides molecular-evolutionary insights into this photoprotective response in these bacteria.

Results

CarH is present in all three myxobacterial suborders and CarA only in Cystobacterineae

Myxococcales currently comprise three suborders and 10 families, and we examined 27 complete, publicly accessible myxobacterial genomes for genes encoding CarA/CarH homologues and other elements of photoinduced carotenogenesis identified in M. xanthus (Table S1). These myxobacteria are soil-dwelling, except for the aquatic Enhygromyxa salina, Haliangium ochraceum and Plesiocystis pacifica, and obligate aerobes except for A. dehalogenans, a facultative anaerobe. Their GC-rich (66%–75%) genomes are among the largest in bacteria (~9–16 Mb) but are reduced, presumably by extensive gene loss, in V. incomptus (~4.4 Mb) and A. dehalogenans (~5 Mb) (Huntley et al., 2011; Yamamoto et al., 2014). BLAST searches of these genomes using CarA_Mx or CarH_Mx (35% identical) as query yielded one hit in Sorangiineae and Nannocystineae, and two hits in Cystobacterineae except for a single hit in Cystobacter ferrugineus, Cystobacter fuscus and Melittangium boletus, and none in A. dehalogenans or V. incomptus. In Cystobacterineae with two hits, one was most similar to CarA_Mx (70%–98% identity, versus 33%–38% to CarH_Mx) and the other to CarH_Mx (62%–97% identity, versus 32%–34% to CarA_Mx), suggesting the first and second sets to be CarA_Mx and CarH_Mx orthologs respectively (Fig. 2A and Table S2). Interestingly, the single hits in C. ferrugineus, C. fuscus and M. boletus were most similar to CarH_Mx (57%–60% identity) than to CarA_Mx (33%–36%), while sequence identities to CarH_Mx (32%–39%) or to CarA_Mx (29%–34%) were comparable for hits in Sorangiineae and Nannocystineae (Fig. 2A and Table S2).

Fig. 2

CarA and CarH homologues in Myxococcales.

A. Heat map of the sequence identities of myxobacterial CarH/A homologues relative to CarA_Mx, CarH_Mx and their DBDs or CBDs (see Table S1 for species abbreviations).

B. Left: Structure of the CarH_Tt protomer (PDB code: 5C8D) with its DBD (DNA recognition helix and wing in dark blue) and its CBD, with bound AdoCbl (sticks in magenta with upper axial Ado group in cyan). Right: Weblogos of myxobacterial CarH/A homologues alongside the corresponding subdomains in CarH_Tt. Dots indicate DNA contacts (blue), Wx₉EH motif (red), lower axial B₁₂-binding motif (magenta), dimer contacts (green) and dimer-dimer interface (black), based on the CarH_Tt and DBD-CarA_Mx structures, with filled dots indicating residues tested by mutational analysis (Navarro-Avilés et al., 2007; Jost et al., 2015b).

C. Maximum likelihood phylogeny (blue dots >80% bootstrap values) for myxobacterial CarH and CarA homologues.

Whether the observed hits are bona fide CarA_Mx or CarH_Mx orthologs was assessed further by sequence analysis and experimentally. CarA_Mx and CarH_Mx have similar architectures, with an N-terminal MerR-type winged-helix DNA binding domain (DBD) and a C-terminal cobalamin-binding domain (CBD) (Fig. 2B). Sequence comparisons revealed patterns in each of these domains closely parallel to those for whole CarA_Mx or CarH_Mx, with the highest sequence identities between CarA_Mx domains and likely CarA homologues, and CarH_Mx domains and likely CarH homologues within Cystobacterineae (Fig. 2A). Our previous studies of CarH_Tt, CarH_Bm, CarA_Mx and CarH_Mx established key motifs in both the DBD and CBD (Navarro-Avilés et al., 2007; Jost et al., 2015b; Fernández-Zapata et al., 2018). In the CarH_Tt structure, the CBD comprises: (i) an N-terminal four-helix bundle subdomain with a Wx₉EH motif (x = any residue), conserved in CarH_Mx and CarH_Bm, which contacts the upper axial Ado group of CarH-bound AdoCbl and is crucial for CarH photoreceptor activity; (ii) a C-terminal five-stranded α/β Rossmann module housing the classic D/ExHx₂G/Px₄₁SxT/Vx_22–27GG B₁₂-binding motif, whose His supplies the lower axial ligand of bound AdoCbl (Fig. S1B and Fig. S2) (Jost et al., 2015b; Fernández-Zapata et al., 2018). The CarA_Mx CBD retains the B₁₂-binding motif and EH of the Wx₉EH motif but lacks the Trp and four flanking residues (Fig. S2). Accordingly, whether an intact Wx₉EH motif co-occurs or not with the classic B₁₂-binding motif may help distinguish CarH and CarA homologues. Classification of the Cystobacterineae homologues applying this criterion matched that based on overall sequence homology to CarA_Mx and CarH_Mx; moreover, all the single Sorangiineae and Nannocystineae homologues could be assigned as CarH. Consistent with these assignments, CarA homologues form a monophyletic clade within Cystobacterineae, separate from the coexisting CarH or the single Sorangiineae or Nannocystineae homologues (Fig. 2C). Sequence alignment of myxobacterial homologues thus classified as CarA or CarH yielded logos for the signature motifs in each domain shown in Fig. 2B. Besides the presence of a classic B₁₂-binding motif in all myxobacterial homologues, and an intact Wx₉EH motif only in CarH homologues, the logos revealed that the RxWERRY motif in the DNA recognition α-helix and several other residues implicated in DNA binding are conserved in the predicted DBDs of both CarH and CarA homologues, suggesting a common mode of DNA recognition. Also, many CarH homologues have: Trp of the Wx₉EH motif followed by a His, which in light-exposed CarH_Tt is the upper ligand in a bis-His-cobalt linkage; a Glu-Arg or Asp-Arg pair equivalent to the crucial Arg-Asp salt-bridge pair of the CarH_Tt dimer interface; and one of the two Gly at the dimer–dimer interface important in CarH_Tt tetramer assembly (the other Gly is often Pro) conserved in CarH as well as CarA homologues (Fig. 2B).

To experimentally test our assignments, we first analysed several CarA and CarH homologues for self-interactions using the bacterial adenylate cyclase two-hybrid (BACTH) assay in E. coli (which cannot synthesize B₁₂ de novo but can take up and convert exogenous B₁₂ to AdoCbl for use in vivo). This assay has demonstrated that CarH_Mx self-interacts via its CBD only in the dark and when B₁₂ is present, whereas CarA_Mx self-interacts in the dark or light, irrespective of the presence or not of B₁₂ (Ortiz-Guerrero et al., 2011). In the BACTH assay, the homologues we assigned as CarA and CarH behaved like CarA_Mx and CarH_Mx, respectively, except for the sole C. ferrugineus, C. fuscus and M. boletus homologues, whose behaviour mirrored CarA_Mx rather than CarH_Mx (Fig. 3A). Negative controls with only one fusion protein ruled out autoactivation in the behaviour of the C. fuscus, C. ferrugineus and M. boletus homologues (Fig. S3).

Fig. 3

Experimental tests of myxobacterial CarH/CarA assignments.

A. BACTH analysis of the effect of B₁₂ and light on self-interaction of various myxobacterial CarH/CarA homologues. Cells expressing T25 and T18 fusions of CarH or CarA were spotted on X-Gal-LB plates with or without vitamin B₁₂ and incubated for 48 h in the dark or exposed to light.

B. Complementation analysis in M. xanthus. Reporter P_{crt Mx}::lacZ expression (specific β-galactosidase activity shown as the mean and standard error of three independent measurements) in the dark and in the light with AdoCbl present or absent in M. xanthus strain MR2649 bearing the indicated myxobacterial homologue, expressed from a vanillate-inducible promoter using vanillate at 100 μM (CarH_Mx, CarH_Sa, CarH_Cfu, CarH_Cfe), 2 μM (CarH_Sc) or 500 μM (CarH_Ho).

C. Size-exclusion chromatography elution profiles tracked using absorbance at 280 or 522 nm for CarH_Cfu and CarH_Cfe in the apo form in the dark, and with AdoCbl present in the dark (red) or after 5-min exposure to green light (green). Apparent M_r (kDa) for each eluted peak is on top.

D. UV–visible absorbance spectra for the photoconversion of AdoCbl-bound CarH_Cfu and CarH_Cfe upon stepwise illumination with green light for the times (in seconds) indicated.

We next checked our assignments by testing if CarH_Mx could be functionally replaced by representative CarH homologues: CarH_Sa, CarH_Cfu and CarH_Cfe from S. aurantiaca, C. fuscus and C. ferrugineus, respectively, outside the Myxococcus–Corallococcus branch in Cystobacterineae; CarH_Sc from S. cellulosum in Sorangiineae; and CarH_Ho from H. ochraceum in Nannocystineae. Each homologue was expressed from a vanillate-inducible promoter in an M. xanthus strain (MR2649) with a P_{crt Mx}::lacZ transcriptional fusion and which lacks endogenous CarH, CarA, CarS and PduO, the ATP:corrinoid adenosyltransferase essential for AdoCbl generation, to enable controlled exogenous AdoCbl supply. All myxobacterial homologues tested, including CarH_Cfu and CarH_Cfe, downregulated reporter P_{crt Mx}::lacZ expression in the dark when AdoCbl was present but not when absent or when the cells were exposed to light, and hence are functionally equivalent to CarH_Mx (Fig. 3B).

We could purify CarH_Cfu and CarH_Cfe in the apo form as soluble proteins, unlike other myxobacterial CarH tested, and analyse them in vitro. Consistent with the BACTH assay, in size exclusion chromatography apoCarH_Cfu and apoCarH_Cfe eluted predominantly as dimers (M_r  ~59 kDa, compared to ~35 kDa determined from the sequence or by mass spectrometry) (Fig. 3C). In the presence of AdoCbl, both continued to elute as dimers in the dark (M_r ~62 and ~68 kDa respectively), but as monomers when exposed to light (M_r ~31 and ~33 kDa respectively) (Fig. 3C), in contrast to their self-interaction in BACTH analysis. In the latter assay, photolysis of intracellular AdoCbl and consequent poor binding to the two homologues probably causes them to persist as apoproteins and self-interact. The absorbance of the holoproteins at 522 nm (besides at 280 nm) indicated that both the dark and light forms bind to cobalamin (Fig. 3C) at ~1:1 stoichiometry. Moreover, the dark to light photoconversion of their AdoCbl-bound form upon stepwise illumination with green light (Fig. 3D) was similar to CarH_Tt or CarH_Bm (Kutta et al., 2015; Fernández-Zapata et al., 2018; Miller et al., 2020). CarH_Cfu and CarH_Cfe are therefore AdoCbl-based CarH photoreceptors but transition from an apodimer to a holodimer to a photolyzed monomer, in a new twist to the oligomeric plasticity of these photoreceptors. Altogether, our analyses indicate that CarH occurs in all three myxobacterial suborders, but CarA only in Cystobacterineae and always with CarH.

Genes for CarA, CarH and carotenoid synthesis usually occur in genomic proximity and occasionally linked to that for HMGR

To determine whether the link between carotenogenesis and CarH/A is maintained in other myxobacteria, we analysed the genomic context of carA and carH, and performed a search for carotenoid synthesis genes. We found the latter in all myxobacteria examined except A. dehalogenans and V. incomptus, which also lack carA and carH (Fig. 4 and Table S2). Within Cystobacterineae, one phylogenetic branch that includes all Myxococcus, Corallococcus, Stigmatella, Hyalangium and Archangium species presents a genomic arrangement of carotenoid genes (with crtIb also at a different genomic location), carA and carH matching that in M. xanthus. Synteny is conserved in the branch containing C. ferrugineus, C. fuscus and M. boletus, except for the absence of carA (Fig. 4). In M. xanthus, the primary carotenoid end product in light-grown cells is myxobacton ester (a monocyclic carotenoid), although significant amounts of phytoene (the first C₄₀ compound in the biosynthetic pathway) are also found (Fig. S4) (Iniesta et al., 2007). This biosynthetic pathway is likely conserved in the other Cystobacterineae with the same set of carotenogenic genes as M. xanthus.

Fig. 4

CarH, CarA and other factors in the B₁₂-independent pathway identified from myxobacterial genome analysis. The 16S rRNA-based phylogenetic tree for myxobacteria highlighting the three suborders and species, and the genes for carA (A), carH (H), and carotenoid synthesis. B: crtB (phytoene synthase); C: crtC (neurosporene hydroxylase); cruA: CruA-type lycopene cyclase; cyc: possible lycopene cyclase; D: crtD (hydroxyneurosporene dehydrogenase); E: crtE (geranyl geranyl pyrophosphate synthase); I/Ia/Ib: crtI (phytoene desaturase); L: crtL (CrtL-type lycopene cyclase); Y: lycopene cyclase fused to CrtU; Yc: crtYc (CrtYc subunit of heterodimeric lycopene cyclase); Yd: crtYd (CrtYd subunit of heterodimeric lycopene cyclase); Ycd: crtYcd (heterodimeric fusion type CrtYcd lycopene cyclase); Z: crtZ (β-carotene hydroxylase); 6: orf6 (possible glycosyltransferase); 9: orf9 (possible acyltransferase). Genes carF, carQ, carR, carS, carD and carG are indicated by the red circles if present and by unfilled circles if absent.

Interestingly, five species within Cystobacterineae (Stigmatella erecta, Hyalangium minutum, Archangium gephyra, Archangium violaceum and M. boletus) have a gene encoding HMGR between crtB and crtD in the syntenic carotenoid cluster (Fig. 4 and Table S2). HMGR is an enzyme that acts early in the MVA pathway for producing isopentenyl pyrophosphate (IPP) and its dimethylallyl pyrophosphate (DMAPP) isomer, the C₅ building blocks for the biosynthesis of isoprenoids, including carotenoids (Fig. S4). The MVA pathway is one of the two known for isoprenoid biosynthesis and is used exclusively by archaea, fungi and animals; within bacteria, it is used only by a few species, since the majority employ the alternative methylerythritol 4-phosphate (MEP) pathway (Moise et al., 2014; Rodriguez-Concepción et al., 2018; Hoshino and Gaucher, 2021). Our observation that a gene for HMGR lies within the carotenogenic cluster in some myxobacteria prompted us to examine isoprenoid biosynthesis pathways in the species studied here. Interestingly, while A. dehalogenans and V. incomptus have the complete gene set for the MEP pathway, all the remaining myxobacteria have MVA pathway genes and are thus among the minority of bacteria employing this typically eukaryotic route for isoprenoid biosynthesis (Hoshino and Gaucher, 2021). The gene for the primary HMGR, a class II type conserved in all myxobacteria (except the two mentioned earlier) tends to occur in a syntenic cluster with genes for enzymes catalysing the subsequent four steps of the MVA pathway (Figs. S4 and S5A, and Table S2). The primary HMGRs in S. erecta, H. minutum, A. gephyra, A. violaceum and M. boletus share 46%–51% sequence identity (all catalytic residues are conserved; Fig. S5B) with the second HMGRs encoded by the gene within the carotenogenic cluster.

Compared to Cystobacterineae, carotenoid synthesis genes in Sorangiineae and Nannocystineae are often rearranged and/or redistributed as various clusters, with crtI existing as a single copy (Fig. 4). Even so, syntenic blocks of carotenogenic genes with carH are observed in the Sorangiineae examined (except for Sandaracinus amylolyticus and Labilithrix luteola), which are split into smaller blocks in Nannocystineae (Fig. 4). Some species appear to conserve only genes acting in the early biosynthetic steps, suggesting that their carotenoid end products may differ. Also, while M. xanthus and other Cystobacterineae (based on protein conservation relative to M. xanthus) use a heterodimeric CrtYc–CrtYd cyclase (Iniesta et al., 2008) to catalyse the formation of cyclic carotenoids from acyclic lycopene (Fig. S4), lycopene cyclases of the CruA family were found in Sorangiineae and of the CrtL family in Nannocystineae (Fig. 4). CruA-type cyclases typically occur in green sulfur bacteria and cyanobacteria, and the CrtL-type in cyanobacteria and plants (Maresca et al., 2007). Despite these differences, carH was generally found adjacent to carotenoid synthesis genes in Sorangiineae and Nannocystineae. Neither of these two suborders has a gene encoding HMGR within a carotenoid synthesis gene cluster. Their only gene for HMGR occurs with MVA pathway genes, and a second copy is found only in H. ochraceum and S. amylolyticus (Fig. S5A). Interestingly, in the latter and in Nannocystineae, the MVA cluster includes genes for the synthesis of hopanoids (polycyclic triterpenes, like sterols, with six IPP units), which have been functionally and evolutionarily linked to carotenoids (Santana-Molina et al., 2020). All the myxobacterial genomes analysed, excluding the downsized ones in A. dehalogenans and V. incomptus, thus have carotenoid synthesis genes, with proximal carH and/or carA in the majority, hinting at light and CarH/CarA-mediated regulation of carotenogenesis in these myxobacteria, as in M. xanthus.

The B₁₂-independent CarA pathway evolved solely in Cystobacterineae

We next surveyed myxobacterial genomes for the other essential factors of the M. xanthus B₁₂-independent, CarA-dependent regulatory pathway (CarF, CarQ, CarR, CarS, CarD and CarG), which have singular features. Our search for myxobacterial homologues of CarF_Mx, the desaturase required for the biosynthesis of plasmalogen lipids typically found in metazoa (Gallego-García et al., 2019), yielded single hits (each encoded by an isolated gene transcribed divergently relative to its flanking genes) in all Cystobacterineae except A. dehalogenans and V. incomptus, and in the other two suborders only in N. exedens (Fig. 4, Fig. S6 and Table S2). All these hits share sequence identities to CarF_Mx of ≥ 70% among Cystobacterineae (Fig. S6A) and 54% for that in N. exedens, and they conserve the nine histidines shown to be essential for function in CarF_Mx and its metazoan homologues (Gallego-García et al., 2019).

Of the three proteins encoded by the carQRS operon, CarQ belongs to the large ECF-σ factor family, whose members are widespread, because use of such alternative σ factors constitutes a fundamental mechanism of signal transduction across eubacteria. Moreover, the number of ECF-σ factors in myxobacteria tends to be high (∼45 in M. xanthus, ∼87 in S. cellulosum, ∼118 in P. pacifica) (Abellón-Ruiz et al., 2014). Consequently, unequivocal assignment of genuine CarQ orthologs relied on the simultaneous presence of the far more distinct CarR and CarS. CarR has the unusual six transmembrane-helix topology of DUF1109 family anti-σ factors largely restricted to Proteobacteria, some of which have been implicated in heavy metal or oxidative stress responses (Kohler et al., 2012; Masloboeva et al., 2012). However, CarR and its myxobacterial orthologs share low sequence similarity to other DUF1109 factors. Interestingly, CarR is a dual response anti-σ that can be inactivated by the joint action of light-induced ¹O₂ and plasmalogens (Galbis-Martínez et al., 2012; Gallego-García et al., 2019) as well as by copper (Moraleda-Muñoz et al., 2005). On the other hand, the CarS antirepressor is a DNA mimic (León et al., 2010) with no sequence homologues beyond Myxococcales. We identified CarR and CarS homologues only in Cystobacterineae and, except for A. dehalogenans and V. incomptus, in all species analysed in this suborder, with carR and carS always with carQ as a carQRS operon, as in M. xanthus (Fig. 4, Fig. S6 and Table S2).

Our earlier studies established the unique two-domain architecture of CarD, comprising an N-terminal RNAP-interacting domain, widely distributed among bacteria, which in CarD also interacts with CarG; a C-terminal K/A/P-rich DNA-binding domain akin to eukaryotic high mobility group A (HMGA) proteins, with four or three repeats of their characteristic AT-hook motifs in CarD_Mx and CarD_Sa respectively, but which lacks the AT-hooks and resembles the histone H1 C-terminal region in CarD_Ad (Padmanabhan et al., 2001; Cayuela et al., 2003; García-Heras et al., 2009; García-Moreno et al., 2010). On the other hand, an inactive version of the metzincin-type zinc-binding metalloprotease motif typically found in Archaea and vertebrates (Marino-Puertas et al., 2017) occurs in CarG, which regulates gene expression by interacting with another transcription factor (CarD) rather than directly with DNA (Peñalver-Mellado et al., 2006). Our search identified CarD and CarG orthologs (encoded in an operon, as in M. xanthus) only among Myxococcales and exclusively in Cystobacterineae, and in every species analysed in this suborder (Fig. 4, Fig. S6 and Table S2). Among CarD orthologs, only CarD_Vi has an H1-like C-terminal domain-like CarD_Ad, the rest having four AT-hooks or three (in CarD_Sa, CarD_Se and CarD_Hm).

In sum, given the absence of all six factors among Sorangiineae and Nannocystineae (except for CarF in N. exedens), the entire B₁₂-independent pathway appears to have evolved uniquely in Cystobacterineae, in parallel with the advent of CarA. It is therefore intriguing that all six factors occur in C. ferrugineus, C. fuscus and M. boletus, even though none of the three species have CarA. Despite lacking CarH as well as CarA and four of its associated factors, A. dehalogenans and V. incomptus retain CarD and CarG, possibly because the two proteins act as a global regulatory complex in processes other than light-induced carotenogenesis (García-Heras et al., 2009; Abellón-Ruiz et al., 2014; Bernal-Bernal et al., 2018).

P_crt operator design in Cystobacterineae

Most Cystobacterineae conserve the M. xanthus structural and regulatory genes for light-induced carotenogenesis, as well as their genomic arrangement (Fig. 4). This suggests a common strategy to control carotenoid synthesis, with both CarA and CarH-dependent pathways in the Myxococcus, Corallococcus, Stigmatella, Hyalangium and Archangium species, but only the latter in C. ferrugineus, C. fuscus and M. boletus. A common mode of regulation would likely be reflected by DNA sequence conservation around P_crt, which we analysed next.

DNA binding assays with purified CarA_Mx revealed its binding as two dimers in a stepwise cooperative manner to an ~55 bp operator at P_{crt Mx}, with the proposed CarA_Mx binding site containing two inverted repeats (Fig. 5A): a perfect one upstream of the −35 promoter region and a second imperfect one overlapping with it (López-Rubio et al., 2004). However, among Cystobacterineae with CarA, only Myxococcus species and C. coralloides conserve a perfect or near-perfect match to the M. xanthus inverted repeats, which steadily degenerate in the rest. CarH_Tt and CarH_Bm, on the other hand, have been shown to bind as AdoCbl-bound tetramers to tandem 11-bp direct repeats (DRs) with the sequence 5´-nnnnTnnACAn-3′ (n is any base). While CarH_Tt uses three of its DBDs to bind three tandem DRs, one of which includes the −35 promoter element, CarH_Bm binds to operators comprising four such tandem DRs, or three DRs and one pseudorepeat (hereafter dr, where the TnnACA motif is not fully conserved), which overlap with the − 35 or − 10 promoter elements (Jost et al., 2015b; Fernández-Zapata et al., 2018). Soluble native CarH_Mx has resisted purification for analysis in vitro, but a functionally equivalent chimeric protein (CTt2) with CarH_Mx CBD replaced by that of CarH_Tt recognized the same ~55-bp operator at P_{crt Mx} as CarA_Mx (Ortiz-Guerrero et al., 2011; Padmanabhan et al., 2017). Inspection of the ~55-bp DNA segment at P_{crt Mx} revealed three 11-bp DRs (DR1, DR3 and DR4) conforming to those identified for CarH_Tt and CarH_Bm, and two drs (dr2 and dr5), the −35 TTGACA promoter element being within DR4 (Fig. 5A and Fig. S7). Sequence comparisons revealed a similar DR1-dr2-DR3-DR4-dr5 arrangement in all other Myxococcus species, C. coralloides, S. aurantiaca and S. erecta, which is dr1-DR2-DR3-DR4-dr5 in H. minutum, C. fuscus, C. ferrugineus and M. boletus, and dr1-dr2-DR3-DR4-dr5 in A. gephyra and A. violaceum (Fig. 5A and Fig. S7). DNA sequence logos from aligning the repeats at each promoter highlight the TnnACA motif, but also reveal variations between closely related species, particularly in the less conserved bases that can contribute additional sequence information to fine–tune DNA binding affinity and specificity (Fig. 5A and Fig. S7) (Jost et al., 2015b; Fernández-Zapata et al., 2018). Yet, despite these variations, all Cystobacterineae CarH homologues tested could replace CarH_Mx to regulate P_{crt Mx} in M. xanthus (Fig. 3B), demonstrating their functional equivalence as well as DNA-binding plasticity.

Fig. 5

B₁₂-dependent light-induced carotenogenesis and CarH DNA binding in Cystobacter.

A. Alignment of the CarA_Mx/CarH_Mx operator DNA sequence (sense strand) at P_{crt Mx} and the corresponding segments for Cystobacterineae species with CarH. Red dots align with T and ACA of the consensus DR motif, and the span of the CarA_Mx/CTt2 DNase I footprint at P_{crt Mx} is indicated; below the logo is a perfect inverted repeat (red arrows) and a less perfect one (purple arrows) found in some Myxococcus species. The −35 promoter element within DR4 and the remaining four 11-bp repeats are indicated at the bottom.

B. Left: Colour of cell culture spots on plates with or without vitamin B₁₂ incubated for 48 h in the dark or in the light for M. xanthus (wild-type and ΔcarA strains), C. fuscus and C. ferrugineus with the corresponding carotenogenic gene cluster and sequence logo (TnnACA segment or equivalent shaded) of the five direct repeats at P_crt.

C. Expression of crtE from C. fuscus and C. ferrugineus measured by qRT-PCR with rpoD as the reference gene reported relative to the level in the dark with vitamin B₁₂ present (mean and standard error from three biological replicates).

D. Representative EMSA for the binding of probes P_{crt Cfu} and P_{crt Cfe} to increasing concentrations of AdoCbl-bound CarH_Cfu or CarH_Cfe (0.1, 0.25, 0.5, 1, 5, 10, 100 nM), respectively, in the dark or after 5-min exposure to green light.

B₁₂-dependent photoregulation of carotenogenesis, CarH DNA binding and its abrogation by light or CarS in Cystobacter

Among Cystobacterineae, C. fuscus, C. ferrugineus and M. boletus are particularly interesting because, besides CarH, they have all the factors in the CarA pathway except CarA (Fig. 4). Moreover, the Cystobacter CarH AdoCbl-bound form is a dimer (Fig. 3C) whereas CarH_Tt or CarH_Bm are tetramers, which can have consequences on operator size and/or mode of DNA recognition. We therefore examined how B₁₂ affects carotenogenesis in C. fuscus and C. ferrugineus and how purified Cystobacter CarH recognizes operator DNA in vitro.

First, we checked if cell culture colour depends on vitamin B₁₂ and/or light. Wild-type M. xanthus cell cultures are yellow (due to non-carotenoid pigments called DKxanthenes) (Meiser et al., 2006) in the dark with or without B₁₂ in the growth medium because P_{crt Mx} is repressed by CarA_Mx under both conditions, and also by CarH_Mx when B₁₂ is available; but a carA-deleted (ΔcarA) strain is orange in the dark without B₁₂, and yellow when B₁₂ is present since only CarH_Mx can repress P_{crt Mx} (Fig. 5B). Both strains turn red in the light, irrespective of the presence or not of B₁₂, due to the expression of the carB operon as well as of gene crtIb (Pérez-Marín et al., 2008; Ortiz-Guerrero et al., 2011). Like the ΔcarA M. xanthus strain, C. fuscus and C. ferrugineus were orange without B₁₂, yellow or whitish with B₁₂ in the dark, and more intensely pigmented under light (Fig. 5B). Consistent with these colour changes, expression of the first gene (crtE) in the carotenogenic cluster [estimated by Quantitative Reverse Transcription Polymerase Chain Reaction (qRT-PCR)] was lower in the dark with B₁₂ present, and higher under light (Fig. 5C). Altogether, these data indicate that B₁₂ controls light-induced carotenogenesis in C. fuscus and C. ferrugineus.

We next tested in vitro if and how purified CarH_Cfu and CarH_Cfe bind to 170-bp DNA probes containing the 55-bp segment mentioned earlier. In gel-shift assays (EMSA), apoCarH_Cfu or apoCarH_Cfe yielded only a diffuse smear even at high concentrations, suggesting weak non-specific binding (Fig. S8A). However, in the dark with AdoCbl present, defined retarded bands were observed, which were suggestive of stepwise DNA binding, from a smaller to a larger stable DNA-protein complex as the protein concentration increased (Fig. 5D and Fig. S8A). Binding was cooperative with an affinity (apparent K_D ≈ 35 ± 1 nM; Fig. S8B) comparable to the estimates reported for CarH_Tt and CarH_Bm to their respective operators (Jost et al., 2015b; Fernández-Zapata et al., 2018). Exposure to light eliminated the retarded bands, indicating poor DNA binding by the photolyzed CarH_Cfu or CarH_Cfe monomer (Fig. 5D).

We next mapped the site recognized by CarH_Cfu at P_{crt Cfu} using footprinting analyses. CarH_Cfu binding yielded a large DNase I footprint, which coincides well with the segment based on alignment with the CarA_Mx/CarH_Mx operator (Fig. 5A) and which includes part of dr1 and DR2-dr5. Within the DNase I footprint, four tracts (2–3 nt each) spaced 10–11 nt apart (~1 pitch of double-stranded B-DNA) were protected from hydroxyl radical attack on each strand (Fig. 6A and C). Moreover, CarH_Cfu binding arrested Exo III at ~4 nt downstream of dr5 (strongly) and DR4 (weakly) in the sense strand, and at ~2–4 nt upstream of DR2 (strongly) and dr1 (weakly) in the antisense strand (Fig. 6B and C). Taken together, the sizes of the DNase I/Exo III footprints and the four evenly spaced hydroxyl radical footprints are consistent with cooperative binding to one face of a DNA site composed of the three tandem DRs plus one dr (dr1-DR4 or DR2-dr5, both including the −35 promoter region) by two CarH_Cfu dimers, each using its two DBDs. Accordingly, CarH_Cfu could bind to 52-bp probes spanning dr1-DR4 or DR2-dr5, yielding the two retarded bands observed with the 170-bp P_{crt Cfu} DNA fragment or with a 63-bp probe spanning all five repeats (Figs. 5D and 6D). By contrast, the upper band normally observed at higher CarH_Cfu concentrations was weak for probes with three DRs (2–4) or two DRs and one dr (1–3, 3–5), and absent for probes with two DRs (2*–3–4, 2–3–4*, generated by mutating one of the two outlying DRs of the three DR-probe), with which a stable retarded band nonetheless continued to be observed (Fig. 6D). Moreover, the CarH_Cfu dimer requires the two repeats to be in tandem, since no binding was detected for probe 2–3*–4, with the central DR of the three-DR probe mutated (Fig. 6D). Stable formation of the larger complex therefore requires at least four repeats and two CarH_Cfu dimers, since two DBDs from each dimer would bind to two tandem repeats.

Fig. 6

Mapping of the binding site at P_{crt Cfu} and mode of action of CarH_Cfu.

A. Representative DNase I and hydroxyl radical footprints upon binding of 100 nM CarH_Cfu to the sense and antisense strands of P_{crt Cfu}. Lines to the left: DNase I footprints; lines to the right: hydroxyl radical footprints; arrowheads: DNase I hypersensitive sites.

B. Representative Exo III footprints for CarH_Cfu binding to P_{crt Cfu}. Positions of Exo III arrest: horizontal lines on the side (thicker for stronger arrest).

C. Summary of footprint data generated by CarH_Cfu binding at P_{crt Cfu}. DNase I footprints: horizontal lines; DNase I hypersensitive sites: arrowheads; sites protected from hydroxyl radical attack: dots; positions of Exo III arrest: vertical lines (thicker for stronger arrest); −35 element: lettered red; −10 element (part): blue. DR: tandem 11-bp direct repeat with the sequence 5 ́-nnnnTnnACAn-3 ́(n is any base); dr: 11-bp direct repeat where the TnnACA motif is not fully conserved.

D. Top: Sequences (sense strand) of oligonucleotide probes used, with mutations in blue lowercase. Bottom: Binding to the indicated probes (number of repeats shown below) by CarH_Cfu (50, 100 and 200 nM; 100 and 200 nM for mutant DR2-4 probes).

E. Representative EMSA with CarH_Cfu (100 nM) and increasing concentrations of its cognate CarS_Cfu (100, 200, 400, 800 nM, added after prior incubation of CarH_Cfu with P_{crt Cfu}). DNA binding assays in A–E were performed in the dark with AdoCbl present at five-fold excess relative to protein.

F. Model summarizing CarH_Cfu mode of action. ApoCarH_Cfu dimers bind to AdoCbl to form active CarH_Cfu dimers, which bind cooperatively to an operator at P_{crt Cfu} with two overlapping sites, each comprising three tandem DRs (filled grey arrows) and one dr (unfilled arrows). Binding of two CarH_Cfu dimers to dr1–DR4 (as shown) or to DR2–dr5 would block the −35 promoter region and repress transcription. Light (UV/blue/green) cleaves the AdoCbl Co–C bond to liberate the upper axial ligand (unfilled red circles) provoking disassembly of CarH_Cfu dimers to monomers that retain photolyzed AdoCbl (unfilled red triangles) and do not bind DNA. Alternatively, P_{crt Cfu} can be derepressed by CarS_Cfu, which sequesters CarH_Cfu to block DNA binding.

Since CarH_Cfu can regulate P_{crt Mx} in M. xanthus, we probed how it binds to P_{crt Mx} in vitro. We also tested whether and how CarA_Mx, a dimer like CarH_Cfu but B₁₂-independent, binds to P_{crt Cfu}, which lacks the palindromes (Fig. 5A) that were suggested as recognized by CarA_Mx (López-Rubio et al., 2004). In EMSA, DNase I and Exo III analyses (Fig. S9) both proteins bound in a similar cooperative, stepwise manner to target the same DNA stretch at both P_{crt Cfu} or P_{crt Mx}, despite variations in the DR sequences and their arrangement and the absence of the inverted repeats in P_{crt Cfu}, suggesting that CarA_Mx also recognizes DRs rather than inverted repeats. Differences in some features, like DNase I hypersensitive sites (usually indicative of DNA distortions from protein binding), possibly reflect intrinsic differences between the two proteins, such as AdoCbl dependence only for CarH_Cfu. Thus, even though CarH_Cfu requires AdoCbl for DNA binding and CarA_Mx does not, both are dimers with similar modes of DNA recognition. In addition, purified CarS_Cfu abolished DNA binding by AdoCbl-bound CarH_Cfu (Fig. 6E), just as CarS_Mx abolishes DNA binding by CarA_Mx. CarH_Cfu therefore combines specific AdoCbl-dependent DNA binding in the dark and direct disruption by light characteristic of its tetrameric homologues CarH_Tt and CarH_Bm, with the cooperative stepwise binding as a dimer to operator DNA and its inhibition by the cognate CarS_Cfu that characterizes the B₁₂-independent CarA_Mx dimer (Fig. 6F).

B₁₂-dependent photoregulation of carotenogenesis, CarH operator design and DNA binding in Sorangiineae and Nannocystineae

Sorangiineae and Nannocystineae differ from Cystobacterineae, where the majority of carotenogenic genes and carH occur as a large syntenic operon with a single promoter region that generally contains three DRs, with at least two in tandem, and two drs. Among Sorangiineae, carotenogenic genes are located at several genomic loci and unlinked to carH in S. amylolyticus and L. luteola, and our search in their putative promoter regions for tandem 11-bp direct repeats as possible CarH binding sites revealed none in S. amylolyticus and only one upstream of carH in L. luteola, with two tandem DRs and one dr (Fig. S10A). In the remaining Sorangiineae, we could identify one DR flanked by two or three drs in the intergenic segments of the divergently expressed crtI-carH and crtE-crtD gene pairs, the latter with three DRs in tandem in S. cellulosum (P_{crt Sc2}; Fig. S10A). Similar analyses in Nannocystineae (where carH is isolated in N. exedens but amidst carotenogenic genes in the others) identified at least one putative CarH binding site with two tandem DRs and two adjacent drs in all except E. salina, which has only one DR at both possible sites flanked by 2–3 drs (Fig. S10B). To experimentally test CarH binding to these diverse possible sites, we chose S. cellulosum and H. ochraceum, which we could grow in liquid cultures and/or plates with or without B₁₂, as representative of Sorangiineae and Nannocystineae respectively.

We found that S. cellulosum undergoes B₁₂- and light-dependent cell culture colour changes, being orange in the dark without B₁₂ or when exposed to light but yellow in the dark with B₁₂ present (Fig. 7A), analogous to the ΔcarA M. xanthus and the two Cystobacter strains (Fig. 5B). Moreover, expression of the divergent crtI–carH and crtE–crtD genes was repressed in the dark in the presence of B₁₂ and derepressed upon exposure to light (Fig. 7B), confirming that B₁₂ and light modulate expression of carotenoid genes in S. cellulosum. We then examined AdoCbl-dependent binding at the crtI–carH (P_{crt Sc1}) and crtE–crtD (P_{crt Sc2}) intergenic regions in vitro. Because native soluble CarH_Sc could not be purified, we resorted to the strategy used previously to test CarH_Mx binding to P_{crt Mx} in vitro (Ortiz-Guerrero et al., 2011). For this, the CarH_Sc DBD (and its specific DNA-binding determinants) and linker were fused to the CarH_Tt CBD to generate chimera CSc, which was purifiable. In EMSA, binding of CSc to P_{crt Sc1} and P_{crt Sc2} was clearly observed only in the dark and with AdoCbl present (Fig. 7C). Consistent with three tandem DRs identified in P_{crt Sc2} and only one DR and two drs in P_{crt Sc1}, CSc exhibited higher affinity for P_{crt Sc2}, since an eight-fold lower CSc concentration was sufficient for maximal binding to P_{crt Sc2} compared to P_{crt Sc1}. Moreover, CSc yielded clearer DNase I, hydroxyl and Exo III footprints at P_{crt Sc2} than at P_{crt Sc1} (Fig. 7D–F). Nonetheless, the three tandem repeats in either probe mapped to the ~34–38 bp DNase I footprints and were delimited by positions where Exo III advance was blocked by CSc. Thus, CSc can bind to a site with three tandem DRs (P_{crt Sc2}), like CarH_Tt, but also to one with just one DR and two drs (P_{crt Sc1}). Putative −35 and/or −10 promoter elements that could be identified within the footprinted regions of both probes hint at how CarH_Sc binding at a single site in either probe can simultaneously regulate expression from divergent promoters (Fig. 7F).

Fig. 7

B₁₂-dependent light-induced carotenogenesis and CarH DNA binding in S. cellulosum.

A. Colour of S. cellulosum cell culture spots on plates with or without vitamin B₁₂ incubated for 48 h in the dark or in the light.

B. Expression of the indicated gene measured by qRT-PCR using rpoD as the reference gene, and reported relative to the level in the dark with vitamin B₁₂ present (mean and standard error from three biological replicates).

C. Top: S. cellulosum carotenoid synthesis gene cluster with sequence logos (TnnACA segment or equivalent shaded) from the three putative 11-bp repeats at P_{crt Sc1} (dr1, dr2, DR3) and P_{crt Sc2} (DR1–DR3). Bottom: Representative EMSA for the binding (in the dark or after 5-min green light exposure in the presence or absence of AdoCbl) of CSc to P_{crt Sc1} (left gel: 60, 120, 240, 480, 960 nM; right gel: 1000 nM CSc) and to P_{crt Sc2} (left gel: 10, 20, 40, 80, 120 nM CSc; right gel: 120 nM CSc).

D. Representative DNase I and hydroxyl radical footprints on the sense and antisense strands (relative to the indicated gene) for binding of 1000 nM CSc to P_{crt Sc1} and 240 nM CSc to P_{crt Sc2} (in the dark and presence of AdoCbl).

E. Representative Exo III footprint on the sense and antisense strands of probes P_{crt Sc1} and P_{crt Sc2} upon CSc binding as in D.

F. Summary of footprint data at P_{crt Sc1} and at P_{crt Sc2} upon CSc binding. Proposed promoter elements (red, −35 region; blue, −10 region) are shown (note overlap of the two divergent promoters). Symbols in D–F are as in Fig. 6A–C respectively.

A similar analysis was carried out for H. ochraceum, which was also less pigmented in the dark with B₁₂ present than without, or in the light (Fig. 8A). Like CarH_Sc, CarH_Ho was not purifiable as a soluble native protein. We therefore again generated and purified a chimera, CHo, fusing the CarH_Ho DBD and linker to the CarH_Tt CBD and tested its binding to a 170-bp probe (P_{crt Ho}) corresponding to the upstream non-coding region of the carotenogenic gene cluster containing carH (Fig. 8B). CHo could bind to P_{crt Ho} only in the dark with AdoCbl present (Fig. 8B) and footprinting analysis mapped the binding site, which overlaps with putative −35 and −10 promoter regions, to four tandem 11-bp repeats comprising two DRs and two drs (Fig. 8C–E). Thus, light and B₁₂ also appear to regulate via CarH the expression of carotenogenic gene clusters in Sorangiineae and Nannocystineae, despite their distinct arrangements from that in Cystobacterineae. Moreover, the data reveal, for the first time, that even operators with only one or two DRs (with flanking drs) can be recognized by CarH, highlighting unusual plasticity in DNA recognition and operator design.

Fig. 8

B₁₂-dependent light-induced carotenogenesis and CarH DNA binding in H. ochraceum.

A. Colour of H. ochraceum cells streaked on plates with or without vitamin B₁₂ incubated for 48 h in the dark or in the light.

B. Top: H. ochraceum carotenoid synthesis gene cluster and sequence logo (TnnACA segment or equivalent shaded) derived from the four repeats at P_{crt Ho}. Bottom: representative EMSA for the binding of CHo to a 221-bp P_{crt Ho} probe in the presence or absence of AdoCbl in the dark (left and middle gels: 2.5, 5, 10, 15 nM CHo) or after 5-min green light exposure (right gel: 10 nM CHo).

C. Representative DNase I and hydroxyl radical footprinting on the sense and antisense strands of probe P_{crt Ho} with or without CHo in the dark and presence of AdoCbl.

D. Representative Exo III footprinting of CHo binding to probe P_{crt Ho} as in C.

E. Summary of footprint data on each strand at P_{crt Ho} upon CHo binding. Proposed promoter elements (red, −35 region; blue, −10 region) are shown. Symbols in C–E are as in Fig. 6A–C respectively.

Discussion

Studies on M. xanthus light-induced carotenogenesis have uncovered new paradigms in bacterial light sensing, signalling and response, and led to the discoveries of widely distributed protein families with novel functions. Central players in the genetic circuit underlying this M. xanthus response to light are two transcriptional factors with similar domain architectures and DNA target sites, CarH_Mx and CarA_Mx, which control the expression of the same genes by distinct mechanisms. While an AdoCbl chromophore directly senses light to allosterically modulate CarH repressor activity, regulation by CarA requires the antirepressor CarS, whose light-induced expression relies on a multifactorial pathway that additionally includes CarF, CarQ, CarR, CarD and CarG. Our present study on the conservation of these elements for photoinduced carotenogenesis across diverse myxobacteria provides molecular and evolutionary insights into these factors and establishes that B₁₂-dependent photoregulation of carotenogenesis pervades in this major class of social and predatory soil and marine bacteria.

Our sequence analysis with experimental validation assigned CarH homologues in all the myxobacteria examined, except in two species with reduced genomes, A. dehalogenans and V. incomptus. Both of these also lack genes for carotenoid synthesis and probably rely on ROS detoxifying enzymes and/or light-avoidance mechanisms to combat photooxidative stress. The majority have carH in the genomic neighbourhood of genes for carotenogenesis, which occur as a large syntenic block in Cystobacterineae and as smaller syntenic blocks in Sorangiineae or Nannocystineae. Genomic proximity of carH to carotenogenic genes points to functional, presumably evolutionary, links between them that may be related to CarH's ability to sense light and elicit the photoprotective response over a broader spectral range (UV to green) than most other known photoreceptors. In contrast to CarH, our analysis revealed CarA only in suborder Cystobacterineae, always with CarH, and invariably encoded by the gene immediately upstream of carH in the carotenogenesis gene cluster characteristic of this suborder. That carA and carH exist in tandem and their close phylogenetic relationship suggests that their co-occurrence is likely due to gene duplication. Gene duplication and divergence is a major mechanism for the evolution of new genes or paralogues with overlapping or non-overlapping functions and has been proposed as a key contributor to genomic expansion and sensory complexity in myxobacteria (Goldman et al., 2006; Schneiker et al., 2007). Tandemly duplicated genes, usually arising from recombination-based mechanisms, are often restricted to closely related species, tend to share similar functions and regulatory elements, and typically contribute to adaptation and response to environmental stimuli and species-specific traits (Simon et al., 2015; Chapal et al., 2019; Kuzmin et al., 2020). All of these apply to the CarH–CarA pair. Given the ancient prebiotic origin proposed for B₁₂ (Roth et al., 1996; Monteverde et al., 2017), the phylogeny, and far greater prevalence of CarH over CarA, it is likely that an ancestral AdoCbl-dependent CarH duplicated and diverged to create a B₁₂-independent CarA. The distribution of carH and carA genes in the myxobacterial genomes analysed in this study suggests that such a gene duplication could have occurred in the ancestor of the Myxococcus-Archangium branch or, alternatively, in an even earlier Cystobacterineae predecessor, with subsequent gene loss resulting in just one copy in C. ferrugineus, C. fuscus and M. boletus, and none in A. dehalogenans and V. incomptus.

Transition from CarH to CarA entailed loss of the Wx₉EH motif, crucial for AdoCbl-specific binding, oligomerization and DNA binding of CarH, together with changes to ensure that CarA forms stable dimers independently of B₁₂. Most remarkably, photosensory regulation by CarA has required coevolution of various singular proteins, some eukaryotic-like, and a signature metazoan lipid. Myxobacteria are the only strictly aerobic bacteria known to have the oxygen-dependent desaturase (CarF) for plasmalogen biosynthesis whose equivalent, despite the enormous evolutionary distance, is specifically conserved in metazoa (Gallego-García et al., 2019). The existence in myxobacteria of factors considered typically eukaryotic point to tangible evolutionary links to eukaryotes and possible horizontal gene transfer events that provide supporting evidence for the Syntrophy hypothesis, which posits an ancient myxobacterium as crucial in eukaryogenesis (López-García and Moreira, 2020; Hoshino and Gaucher, 2021). Thus, steroid biosynthesis has been proposed to have first evolved in an early myxobacterium in response to the rise of oxygen and later horizontally transferred to an early eukaryote, because myxobacteria form the select set of aerobic bacteria to have both the steroid and its isoprenoid precursor (MVA) biosynthesis pathways homologous to those in eukaryotes (Hoshino and Gaucher, 2021). It is tempting to speculate that this also extends to the oxygen-based biosynthesis of plasmalogens, given the existence of CarF in myxobacteria and in metazoans, and its absence in nearly all other bacteria. However, while eukaryotic-type steroid biosynthesis enzymes appear across all three myxobacterial suborders (Hoshino and Gaucher, 2021), we found CarF almost entirely in Cystobacterineae. Curiously, besides the gene for HMGR that is part of an operon with various other MVA pathway genes, the large syntenic block of carotenogenic genes in some Cystobacterineae includes a second copy of a gene for HMGR. Since HMGR is generally a rate-limiting enzyme in the MVA pathway, its photoinduced co-expression with carotenogenic enzymes possibly ensures ample supply of the earliest precursors (IPP/DMAPP) of carotenoids for light-induced production.

CarR orthologs also appear to have evolved solely in Cystobacterineae to act in concert with plasmalogens and ¹O₂ in a photoprotective stress response. They diverge significantly in sequence from other DUF1109 anti-σ but may nonetheless retain a role in responses to metals, as observed for CarR_Mx in the copper response. Moreover, CarR has had to co-evolve with CarQ to recognize it specifically from among the numerous ECF-σ factors typically found in myxobacteria. CarS homologues are yet another class of proteins found exclusively in Cystobacterineae. CarS is a small, taxonomically restricted protein acting in a stress response and thus has features typically ascribed to a de novo gene product (Schlotterer, 2015; Van Oss and Carvunis, 2019). Also crucial in the CarA pathway is the highly unique CarD–CarG complex, formed by two proteins encoded by adjacent genes that always co-exist in every (and only in) Cystobacterineae analysed here. In CarD, the fusion of an N-terminal RNAP-interacting module of exclusively bacterial origin with a C-terminal DNA-binding domain that resembles eukaryotic HMGA or histone H1 is intriguing at the evolutionary and functional levels. It has been proposed that the origins of the characteristic K/A/P-rich H1 C-terminal region can be traced to eubacteria, from where it was acquired by horizontal gene transfer and evolved in eukaryotes (Kasinsky et al., 2001). If so, an ancestral CarD could have arisen from fusion of two bacterial modules, one of which was a C-terminal K/A/P-rich H1-like region as in A. dehalogenans/V. incomptus, from which the remaining Cystobacterineae acquired the HMGA AT-hooks by convergent evolution. Two independent fusion events, one to an H1-like region and another to an HMGA-like region, would be less probable. Since CarD always functions as a single unit with CarG, the two proteins must have co-evolved. Horizontal gene transfer followed by functional divergence may have led to the emergence of CarG, as it lacks any protease activity yet has a zinc-binding metalloprotease motif of the type typically found only in archaeal and metazoan metzincins.

Strikingly, Cystobacter/Melittangium species in Cystobacterineae lack CarA despite having all other known factors of this pathway. The single Cystobacter CarH/A homologue, our data show, is an AdoCbl-based CarH photoreceptor that transitions from an inactive dimeric apo form to active AdoCbl-bound dimers, which become inactive monomers upon photolysis. The contrast with the oligomeric transitions of the only other well-studied CarH homologues, CarH_Tt (monomer–tetramer–monomer) and CarH_Bm (tetrameric molten globule-tetramer–dimer) highlights the oligomerization plasticity of these photoreceptors. Oligomerization confers properties crucial for protein function, like cooperative ligand binding, allosteric regulation and stability and changes in oligomerization through evolution usually enable new regulatory modes. But how a given multimer and its functions arise is poorly understood. Extant oxygen-binding haemoglobin tetramers, with their striking cooperativity and regulatory modes, may have evolved by successive gene duplication-divergence events through surprisingly few mutations from an ancestral dimeric precursor, which itself arose from an ancient monomer (Pillai et al., 2020). Single mutations at the interface between the two head-to-tail dimers that constitute the CarH_Tt tetramer suffice to shift the tetramer to a native dimer, which retains AdoCbl-dependent DNA binding, but with altered affinity and cooperativity (Jost et al., 2015b); CarH_Tt tetramerization traces a kinetic pathway from monomer to dimer to tetramer (Camacho et al., 2021). Hence, few genetic changes may underlie the shift between the AdoCbl-dependent Cystobacter CarH dimer and its tetrameric homologues.

We found that despite the distinct oligomerization, the Cystobacter CarH dimer emulates its tetramer homologues in its AdoCbl-dependent binding to one face of a DNA region comprising 11-bp tandem direct repeats. It does so by cooperative binding as two dimers, effectively a tetramer, to two overlapping sites, each composed of three tandem DRs and one dr. Whereas each CarH_Cfu dimer requires the two repeats to be in tandem, in the case of the CarH_Tt tetramer one head-to-tail dimer interacts with two non-contiguous DRs (DR1 and DR3) and the other contacts the central DR (DR2) using only one of its two DBDs (Jost et al., 2015b). Furthermore, CarH_Cfu and CarA_Mx recognize each other's operators in vitro and employ a similar DNA binding mode, regardless of the variations between their natural sites and the requirement for AdoCbl by CarH_Cfu but not by CarA_Mx; CarH_Cfu can direct AdoCbl-based regulation of P_{crt Mx} in vivo, emphasizing the evolutionary relationship between the two proteins and their operators. Indeed, in every CarH-containing Cystobacterineae analysed we discerned counterparts to the M. xanthus and Cystobacter operators in size and sequence, generally with three DRs (and two drs) and an identical −35 promoter region embedded in one of them, such that binding of CarH would block access to RNAP. In Sorangiineae and Nannocystineae, CarH operators are less like those in Cystobacterineae. Most species appear to have two operators that differ among themselves, each expected to control two divergent promoters, given the overlap with putative promoter regions. Thus, two CarH operators were experimentally mapped in S. cellulosum, one comprising three tandem DRs and the other just one DR and two drs, while the single operator mapped in H. ochraceum has two DRs and two drs. Even though CarH operators across the three suborders are quite varied, CarH homologues from the three myxobacterial suborders could functionally replace CarH_Mx for AdoCbl-dependent photoregulation of P_{crt Mx} in M. xanthus. This highlights an unusual DNA-binding plasticity of these homologues, possibly providing cells a strategy to fine–tune affinities and operator occupancies for optimal regulation in vivo.

The B₁₂-independent CarA pathway, absent in both Sorangiineae and Nannocystineae, affords an alternative mechanism to the B₁₂-dependent CarH one in Cystobacterineae. Parallel mechanisms usually confer a fitness advantage by boosting response efficiency and control, and enable adaptive evolution through cooperation and functional divergence to bestow benefits not gained from just a single mechanism (Chapal et al., 2019; Kuzmin et al., 2020). Direct light-sensing and regulation by AdoCbl-CarH ensures a very rapid and efficient response. However, myxobacteria cannot perform the expensive de novo biosynthesis of the complex AdoCbl cofactor, which requires as many as 30 steps, because they do not have all the necessary enzymes (Shelton et al., 2019). They consequently must depend on their environmental niche to acquire B₁₂, a frequently transacted resource in microbial consortia (Sokolovskaya et al., 2020), by preying on microorganisms that can supply B₁₂ or a suitable precursor, or by importing these directly from the milieu. A need to cope with limited B₁₂ probably drove the emergence of the CarA pathway in myxobacteria, which allows them to still regulate the photoprotective response when B₁₂ is unavailable. An additional adaptative advantage conferred by this pathway is its possible implication in other independent or overlapping responses, as illustrated by its role in the response to copper. Even without CarA, presence of the other factors in the B₁₂-independent pathway would expand the range of responses beyond that to light. Moreover, given that CarQ is an alternative ECF-σ factor, its regulon may extend beyond genes for carotenoid synthesis or their specific regulators (our unpublished data point to various genes against oxidative stress). This may explain the occurrence of these other B₁₂-independent pathway factors even in the absence of CarA, as in Cystobacter, and opens the possibility that these other factors may have emerged prior to the appearance of CarA from an ancestral B₁₂-dependent CarH photoreceptor.

Experimental procedures

Bacterial strains, growth conditions and complementation analysis

Bacterial strains, their source and plasmids used in this study are listed in Table S3. E. coli strain DH5α used for plasmid constructs, BL21(DE3) for protein overexpression, and BTH101 (cya⁻) for BACTH analysis, were grown in LB (Luria–Bertani) liquid medium or LB-agar (1.5%) plates at 37°C or 25°C/18°C for inducing His₆-tagged protein overexpression. M. xanthus was grown at 33°C in CTT medium (10 g L⁻¹ Bacto-casitone, 10 mM Tris-HCl pH 7.6, 1 mM KH₂PO₄, 8 mM MgSO₄), as reported previously (Ortiz-Guerrero et al., 2011). S. cellulosum was grown at 33°C in M medium (Müller and Gerth, 2006). C. fuscus, C. ferrugineus and M. boletus were grown at 30°C in 3 g L⁻¹ Bacto-casitone, 1 g L⁻¹ glucose, pH 7.2, and H. ochraceum at 30°C on CY 1:5 plates (0.6 g L⁻¹ Bacto-casitone, 0.2 g L⁻¹ yeast extract, 0.27 g L⁻¹ CaCl₂·2H₂O, 21.1 g L⁻¹ NaCl, 0.6 g L⁻¹ KCl, 3.6 g L⁻¹ MgCl₂·6H₂O, 0.09 g L⁻¹ NaHCO₃, 2.6 g L⁻¹ MgSO₄.7H₂O, 15 g L⁻¹ Difco agar, pH 7.2). Media were supplemented with 1 μg ml⁻¹ of vitamin B₁₂ (which is converted intracellularly to AdoCbl) or AdoCbl and, as required, grown under light using three 18-W fluorescent lamps (10 Wm⁻²). Genes for CarA_Av, CarH_Av, CarH_Ho, CarH_Pp, CSc and Cho were synthesized (GenScript). The others examined were PCR-amplified from genomic DNA purchased from DSMZ, Germany (C. coralloides), kindly provided by Prof. Rolf Müller (M. fulvus), or isolated using Wizard genomic DNA purification kit (Promega) from cells (C. fuscus, C. ferrugineus and M. boletus) grown as described above. For complementation analysis, the gene of interest was cloned into plasmid pMR3679 (Km^R), which allows conditional expression from a vanillate-inducible promoter, tunable by inducer concentration (Iniesta et al., 2012). The construct, verified by sequencing, was electroporated into M. xanthus strain MR2649, which harbours a Tn5-lac-132 insertion in the carB operon to serve as a P_{crt Mx}::lacZ reporter probe and lacks endogenous CarH, CarA, CarS and to allow controlled supply of exogenous AdoCbl, PduO (Ortiz-Guerrero et al., 2011). Transformants with plasmid integration at a heterologous site by homologous recombination were selected using the Km^R marker. M. xanthus cells were grown to exponential phase (OD₆₀₀ ~ 0.8) in the dark in liquid CTT media with vanillate (2–500 μM depending on the CarH homologue), antibiotics and 1 μg ml⁻¹ of AdoCbl as required, diluted into fresh medium of the same composition and grown to OD₆₀₀ ~ 0.8, then split into two cultures, one for growth in the dark and the other under light for 8 h. Cells were spotted (5 μl) on plates of the same composition and grown under the same conditions, or collected for P_{crt Mx}::lacZ β-galactosidase specific activity measurements (in nmol o-nitrophenyl-β-d-galactoside hydrolyzed/min mg⁻¹ protein reported as the mean and standard error of three biological replicates) in a SpectraMax 340 microtitre plate reader (Molecular Devices), as described previously (Ortiz-Guerrero et al., 2011).

Genome and phylogenetic analysis

Specific M. xanthus proteins were used as queries in BLASTP analysis of myxobacterial genomes at the JGI integrated microbial genomes (IMG) resource (https://jgi.doe.gov/data-and-tools/img) to identify myxobacterial orthologs in the light response and in the MVA pathway as well as their genome contexts. The default E-value cut-off (1E-5) was used in all the analyses (Table S2 lists percent identities, E-values and other details). Among those retrieved with CarA_Mx or CarH_Mx as queries against each genome, the best hits were assigned as true homologues, and all had E-values ≤ 1E-17, percent identities ≥ 29%, contained an N-terminal MerR family DBD and a C-terminal CBD, and were in the 286–330 amino acid residue size range. Myxobacterial carotenoid, isoprenoid and cobalamin biosynthetic pathways were additionally analysed using the KEGG pathways database (https://www.genome.jp/kegg), BLASTP (https://blast.ncbi.nlm.nih.gov/Blast.cgi), and Conserved Domains search tool (https://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi). Protein sequence identities were visualized with Heatmapper (http://www.heatmapper.ca/). Maximum likelihood phylogenetic trees were generated using MEGA7 (Molecular Evolutionary Genetics Analysis version 7.0; https://www.megasoftware.net), the JTT (Jones–Taylor–Thornton) substitution model and 200 bootstrap iterations after alignment of CarH/A homologues with MUSCLE (in MEGA7) or of 16S rRNA sequences for Myxococcales species with MAFFT (version 7.0; https://www.ebi.ac.uk/Tools/msa). Trees were visualized and displayed using FigTree v1.4.4 (http://tree.bio.ed.ac.uk/software/figtree) and iTOL (version 5; https://itol.embl.de). Putative CarH DNA-binding sites were identified by visual inspection of regions upstream of candidate target genes; and bioinformatic searches between positions −150 and + 25 relative to the annotated translation start codon for sites conforming to three 11-bp direct repeats with an N₄-T-N₃-C[C][G] consensus sequence (N is any base), and at least one N₄-T-N₂-RCR[G] repeat (R is A or G) using R packages: Biostrings (https://bioconductor.org/packages/Biostrings), IRanges (Lawrence et al., 2013), dyplr (https://CRAN.R-project.org/package=dplyr) and SeqinR (Charif and Lobry, 2007). Genomic data downloaded from Ensembl Bacteria database in FASTA and GFF3 formats were appropriately converted for data handling. WebLogo (https://weblogo.berkeley.edu/logo.cgi) was used to generate logos from sequences aligned with MUSCLE (gaps removed in MEGA7) or ClustalOmega (https://www.ebi.ac.uk/Tools/msa).

Quantitative RT-PCR

RNA was isolated from 5 ml of myxobacterial cultures grown in liquid media with or without 1 μg ml⁻¹ of vitamin B₁₂ in the dark or under light, reverse transcribed and analysed by qRT-PCR, as described in detail previously (Fernández-Zapata et al., 2018). Primers used (Table S4) were designed using Primer Express 3.0 software to amplify an ~ 50–150 bp region within each transcript. Target gene expression is reported relative to rpoD as reference gene and normalized to the level of expression of the gene in the dark with vitamin B₁₂ present, as the mean and standard error for three biological replicates.

BACTH analysis

Two-hybrid analysis of light and B₁₂-dependent oligomerization employed the BACTH system (based on protein fusion to the T25 and T18 fragments of Bordetella pertussis adenylate cyclase), as described previously (Ortiz-Guerrero et al., 2011). Self-interaction was inferred from the blue colour of cell spots on 40 μg ml⁻¹ of X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactoside)-LB plates with or without 1 μg ml⁻¹ of vitamin B₁₂ grown in the dark or under light and, as controls, CarH_Mx and CarA_Mx. Pairs with just one fusion protein were used as negative controls and no autoactivation was observed, as shown in Fig. S3 for CarH_Cfu, CarH_Cfe and CarH_Mb.

Protein purification and characterization

His₆-tagged proteins were overexpressed in E. coli BL21(DE3) using pET15b constructs, purified and stored as reported before (López-Rubio et al., 2002; Ortiz-Guerrero et al., 2011). Native apoproteins purified in a final size-exclusion chromatography step (in 150 mM NaCl, 50 mM phosphate buffer, pH 7.5, 2 mM β-mercaptoethanol) were concentrated and their identities checked by SDS-PAGE and electrospray ionization-time-of flight mass spectrometry (Agilent 1100 Series HPLC, Supelco Discovery BioWide Pore C5 2.1 × 10 cm, 5 μm HPLC column, Ion Trap XCT Plus mass spectrometer with an electrospray interface; mass spectra in the positive ion mode, m/z range 100 to 2200 for samples separated using a linear 4%–90% water/acetonitrile/formic acid gradient). AdoCbl was always handled under dim light and checked by UV–visible (UV–Vis) spectroscopy. Protein concentrations were determined from absorbance at 280 nm in a Cary 60 spectrophotometer, using molar extinction coefficients ε₂₈₀ (in M⁻¹ cm⁻¹) of 37,930, 36,440, 30,480, 32,430 and 30,940 for CarH_Cfu, CarH_Cfe, CarS_Cfu, CSc and Cho, respectively, and the Bio-Rad protein assay kit.

UV–Vis absorption spectroscopy and analytical size exclusion chromatography

UV–visible spectra for AdoCbl-bound CarH_Cfu and CarH_Cfe photoconversion were recorded for samples in the dark and with stepwise green light irradiation (522 nm, 10 Wm⁻², from a computer-controlled LED array) from 1 to 450 s. Analytical size-exclusion chromatography to estimate the apparent molecular mass (M_r) was performed in an AKTA HPLC unit using a Superdex200 HR 10/30 column and calibrated yielding log M_r = 8.1537–0.2316 V_e, where V_e is the elution volume. Pure protein (100 μl, 50–100 μM), alone or after 15-min incubation with five-fold molar excess of AdoCbl was injected into the column in the dark or after 5-min green light exposure. Elution (0.4 ml min⁻¹ flow) was tracked by absorbance at 280, 361 and 522 nm, the peaks collected and analysed for M_r, and by UV–Vis spectroscopy and SDS-PAGE.

DNA binding assays

DNA probes > 100 bp were obtained by PCR using appropriate primers, one of them ³²P-labelled at the 5′-end with T4 polynucleotide kinase (Takara). For DNA probes ≤ 63 bp, complementary HPLC-purified synthetic oligonucleotides (Merck), one ³²P-labelled at the 5′-end and the other unlabeled, were mixed at a 1:2 ratio, incubated at 100°C for 2 min and hybridized by slow cooling. EMSA was carried out using 20 μl of samples containing the DNA probe (1.2 nM, ~13,000 cpm) and protein at specific concentrations, with or without a 5-fold excess of AdoCbl relative to protein in 0.1 M KCl, 10 mM MgCl₂, 0.025 M Tris-HCl (pH 8), 1 mM dithiothreitol, 10% glycerol, 200 ng μl⁻¹ of bovine serum albumin and 1 μg sheared salmon sperm DNA as non-specific competitor. Samples were incubated at 37°C for 30 min and loaded onto a 6% non-denaturing polyacrylamide gel (37.5:1 acrylamide: bisacrylamide) pre-run for 30 min in 0.5× TBE buffer (45 mM Tris base, 45 mM boric acid, 1 mM EDTA). All steps were carried out in the dark (under dim light) except when, prior to electrophoresis, samples were exposed for 5 min to green light. Electrophoresis was carried out in the dark for 1.5 h at 200 V and 10°C. Gels were vacuum-dried and scanned using Personal Molecular Imager™ (PMI™) FX with Quantity One 4.4 software (BioRad). The fraction bound, estimated from band intensities (quantified using ImageJ, NIH) for the free DNA, was fit to a 3-parameter Hill equation using SigmaPlot (Systat Software) to estimate the apparent equilibrium dissociation constant (K_D, the protein concentration at half-maximal binding) and Hill coefficient. For DNase I footprinting, 20 μl of samples of a DNA probe (∼20,000 cpm with either the sense or the antisense strand 5′-end ³²P-labelled) and protein with AdoCbl at a 5-fold excess in EMSA buffer lacking glycerol were incubated for 30 min at 37°C, then treated with 0.07 units of DNase I (Promega) at 37°C for 2 min and quenched with 25 mM EDTA, all in the dark. DNA was ethanol-precipitated, washed twice with 70% ethanol and dried. For hydroxyl radical footprinting, samples were prepared as for DNase I footprinting but without MgCl₂, and then treated for 4 min at 25°C with 2 μl each of freshly prepared Fe(II)-EDTA solution (1 mM ammonium iron (II) sulfate, 2 mM EDTA), 0.01 M sodium ascorbate and 0.6% hydrogen peroxide (Merck). The reaction was stopped with 2 μl each of 0.1 M thiourea and 0.5 M EDTA (pH 8), and DNA precipitated and dried as before. For Exo III (3′ → 5′ exodeoxyribonuclease; ThermoFisher Scientific) footprinting, 20 μl samples of the DNA probe 5′-end ³²P-labelled on a specific strand were mixed with protein and a 5-fold excess of AdoCbl, incubated in DNase I buffer for 30 min in the dark, then treated with 2.5–5 U μl⁻¹ of Exo III for 15 min at 37°C, quenched with 20 mM EDTA, DNA precipitated and dried. Each dried DNA sample for footprinting assays was resuspended in 5 μl of formamide loading buffer, incubated at 95°C for 3 min, and loaded onto 6% polyacrylamide-8 M urea sequencing gels. After electrophoresis (2–3 h, 50 mA/60 W/1.90 kV), the gels were vacuum-dried and scanned using PMI™ (BioRad). A + G chemical sequencing ladders were used to map the footprints.

Acknowledgements

This work was supported by the Ministerio de Ciencia, Innovación y Universidades (MCIU)/Agencia Estatal de Investigación (AEI)/European Regional Development Fund (ERDF): grants PGC2018-094635-B-C21 to M.E.-A. and PGC2018-094635-B-C22 to S.P, both funded by MCIU/AEI 10.13039/501100011033 and by ‘ERDF A way of making Europe’; by grant 20992/PI/18 to M.E.-A. from Fundación Séneca (Murcia)-Spain; by Ph.D. fellowship contracts to R.P.-C (BES-2016-076883, MINECO/AEI), to J. F.-Z. (BES-2013-063376, MINECO), and to E.B.-M. (FPU17/05685, Ministerio de Educación, Cultura y Deporte). We thank J.A. Madrid and V. López-Egea for technical assistance.