Xanthophyll carotenoids stabilise the association of cyanobacterial chlorophyll synthase with the LHC-like protein HliD

Carotenoids are isoprenoid pigments that are categorised into two main classes, xanthophylls, which contain oxygen, and carotenes, which do not. In chlorophototrophs carotenoids are important for light harvesting, photoprotection and structural stabilisation of proteins and membranes [1,2]. For example, the trimeric photosystem I (PSI) from Synechocystis sp. PCC 6803 (hereafter Synechocystis) contains 72 carotenoids and is dependent upon β-carotene for trimerisation [3–6]. β-carotene is also a cofactor in photosystem II (PSII) and is required for its assembly [7,8], is present in the cytochrome b₆f complex [9,10] and photosynthetic complex I [11], and plays a role in the assembly of the phycobilisome antenna complex [5]. In plants, xanthophylls play important roles in structural stabilisation of light-harvesting complexes (LHCs), light harvesting and photoprotection, and act as lipid-soluble antioxidants [12–14]. However, the role of xanthophylls in cyanobacteria, where they do not participate in light harvesting, is less obvious. There is a pool of free xanthophylls in cyanobacterial membranes that modulates membrane rigidity/fluidity [15,16] and it is widely accepted that these pigments provide protection against photooxidative stress and reactive oxygen and nitrogen species, especially under high light conditions [17,18]. Although not integral components of either photosystem, xanthophylls stabilise oligomers of PSI and PSII [5,19], while keto-xanthophylls are required for non-photochemical dissipation of excess energy by the orange carotenoid protein [20].

Chlorophyll (Chl) synthase (ChlG) is an integral thylakoid membrane protein that catalyses the addition of a geranylgeranyl or phytyl tail to the chlorophyllide (Chlide) macrocycle as one of the terminal steps of Chl biosynthesis (Figure 1A). In Synechocystis, tagged ChlG co-purifies with the carotenoids β-carotene, zeaxanthin and myxoxanthophyll in a pigment-protein complex also containing Chl, high light-inducible proteins (Hlips), the membrane insertase YidC and the PSII assembly factor Ycf39 [21,22]. ChlG binds tightly to HliD in a ChlG–HliD ‘core’, with YidC and Ycf39 present in sub-stoichiometric amounts [21]. Hlips are single helix transmembrane pigment-binding proteins that are thought to be the ancestors of LHCs in eukaryotic phototrophs [23]. Synechocystis contains four Hlips (HliA-D), which appear to be primarily involved in Chl biosynthesis/recycling and biogenesis and photoprotection of Chl-binding proteins [24], although their exact roles remain enigmatic. An Hlip domain is also found fused to the C-terminus of cyanobacterial and plant ferrochelatases [25]. YidC assists in the integration of translated proteins into the membrane bilayer [26] and its association with ChlG is presumed to facilitate insertion of Chl molecules into newly synthesised Chl-binding proteins [21]. The function of Ycf39 in the ChlG–HliD complex is unclear but it could play a regulatory role in re-modelling ChlG–Hlip assemblies in response to stress [27]; Ycf39 and HliD interact in a separate complex that promotes the synthesis and assembly of the core PSII subunits D1 and D2 during exposure to high light [28], conditions under which Ycf39 dissociates from the ChlG complex [27] and HliD appears to be partially replaced by HliC [22].

The approximate molar ratio of pigments in the ChlG complex is Chl (6): zeaxanthin (2.1-2.7): β-carotene (1): myxoxanthophyll (0.6–1) [21,22]. HliD likely binds to ChlG as a dimer as the predicted structure of Hlips shows that they cannot bind pigments as monomers [29]; purified HliD dimers bind 6 Chl molecules and 2 β-carotenes [22]. Ultrafast transient absorption spectroscopy indicates that one of the β-carotenes in the HliD adopts a ‘twisted’ configuration that can quench excited states of Chl, resulting in the safe dissipation of excitation energy as heat [22,30,31], leading to the suggestion that HliD photoprotects Chl-binding proteins and Chl-biosynthesis enzymes [24,28,30,31]. Given that neither isolated HliD nor the Ycf39–HliD complex binds xanthophylls [22,28,30], and specific removal of Ycf39 does not alter the pigment composition of the larger ChlG complex [27], the presence of zeaxanthin and myxoxanthophyll appears to be strictly dependent on the interaction of ChlG and HliD, but their functional roles are unknown.

To investigate the requirement of xanthophylls in the ChlG–HliD complex, we generated a series of Synechocystis strains lacking either combinations of xanthophylls or HliD and performed immunoprecipitations using tagged ChlG as bait. Our results demonstrate that zeaxanthin is required for stable formation of the ChlG–HliD complex, both in vitro and in vivo. ChlG alone does not co-purify with carotenoids and its function appears to be perturbed in the absence of the xanthophyll-mediated association with HliD. Possible roles of HliD in Chl trafficking and photosystem assembly/repair, and candidates which may perform analogous functions in higher phototrophs, are discussed.

Synechocystis was grown at 30°C with moderate light (30–50 µmol photons m⁻² s⁻¹) in BG11 medium [32] supplemented with 10 mM TES (Sigma–Aldrich)-KOH pH 8.2 (BG11-TES). Liquid cultures were shaken at ∼150 rpm. For growth on plates, BG11-TES was supplemented with 1.5% (w/v) agar and 0.3% (w/v) sodium thiosulphate. Antibiotics were included where appropriate (as detailed below). Cultures for purification of protein complexes were grown photoautotrophically with ∼100 µmol photons m⁻² s⁻¹ illumination in 8 L vessels which were mixed by bubbling with sterile air and maintained at 30°C using a temperature coil connected to a thermostat-controlled circulating water bath.

All reported mutant strains were prepared in the Synechocystis WT-P (WT) substrain [33]; the FLAG-chlG ΔchlG (FG/ΔG) strain generated in this background has been reported previously [27]. To generate a strain producing 10xhistidine-tagged ChlG, the NdeI-BglII fragment encoding the 3xFLAG-tagged chlG in pPD-NFLAG::chlG [21] was replaced with a sequence encoding 10xhistidine-tagged chlG (synthesised as a gBLOCK by Integrated DNA Technologies; see Supplementary Table S1 for sequence) and the resulting allele exchange construct (pAH97) was introduced into WT Synechocystis with selection and segregation on kanamycin, as detailed below. The zeocin resistance mutagenesis construct described by Chidgey et al. [21] was used to delete the native chlG gene (slr0056) from the His-chlG strain, generating the strain His-chlG ΔchlG (HG/ΔG).

To generate a crtR null mutant, a linear mutagenesis construct was generated to replace the central 555 bp of the 939 bp gene (sll1468) with an erythromycin resistance cassette by allele exchange. Similar constructs were generated to replace 568 bp of the 912 bp cruF gene (sll0814) with the aadA gene (streptomycin resistance) from pCDFDuet-1 (Novagen), 710 bp of the 1629 bp crtO gene (slr0088) with the chloramphenicol acetyl transferase (cat) from pACYC184 (NEB) or 722 bp of the 1185 bp cruG gene (sll1004) with aadA. The erythromycin resistance mutagenesis construct described by Xu et al. [34] was used to delete hliD (ssr1789). Linear DNA constructs were introduced to Synechocystis by natural transformation and transformants were selected on BG11 agar with 7.5 µg ml⁻¹ kanamycin, 7.5 µg ml⁻¹ erythromycin, 5 µg ml⁻¹ streptomycin, 12.5 µg ml⁻¹ chloramphenicol or 2.5 µg ml⁻¹ zeocin, as appropriate. Segregation of genome copies was achieved by sequential plating with increasing antibiotic concentration up to 20 µg ml⁻¹ (for zeocin), 30 µg ml⁻¹ (for erythromycin/streptomycin/kanamycin) and 50 µg ml⁻¹ (for chloramphenicol) and was confirmed by colony PCR. All primers used to generate constructs and screen Synechocystis strains/mutants are provided in Supplementary Table S1.

FLAG-immunoprecipitations were performed as reported in our previous work [21,27]. Synechocystis cultures were grown to an OD₇₅₀ of ∼0.7–1.0, harvested by centrifugation (17 700×g, 4°C, 20 min), resuspended in binding buffer (25 mM sodium phosphate pH 7.4, 10 mM MgCl₂, 50 mM NaCl, 10% (w/v) glycerol and EDTA-free Protease Inhibitor [Roche]) and lysed in a Mini-Beadbeater-16. The lysed cells were collected from atop the glass beads and thylakoid membranes were pelleted by centrifugation (48 400×g, 4°C, 30 min) and solubilised by incubation with 1.5% (w/v) n-dodecyl-β-D-maltoside (β-DDM; Anatrace) at 4°C for 1 h. Following centrifugation (48 400×g, 4°C, 30 min) to pellet insoluble debris, the solubilised thylakoid fraction (the supernatant) was diluted 2-fold and applied to a 300 µl anti-FLAG-M2 agarose (Sigma-Aldrich) column equilibrated in wash buffer (binding buffer with 0.04% (w/v) β-DDM). The resin was washed with 20 resin volumes of wash buffer to remove contaminating proteins and the FLAG-tagged bait protein and associated interaction partners were eluted in 400 µl of the same buffer containing 187.5 µg ml⁻¹ 3xFLAG peptide (Sigma-Aldrich). The mixture was filtered through a 0.22 µm spin column (Sigma–Aldrich) to separate the resin from the eluted protein. Eluates were analysed immediately or stored at −80°C.

For purification of His-ChlG (H.ChlG), solubilised thylakoids were applied three times to a Ni²⁺ NTA Agarose (Qiagen) immobilised metal affinity chromatography (IMAC) column that had been pre-equilibrated in binding buffer (as above) supplemented with 5 mM imidazole. Approximately 250 µl of resin was used per 8 L of cell culture from which the membrane fraction was isolated. The column was washed with 20 column volumes of binding buffer followed by washes with binding buffer containing progressively higher (20, 50 and 80 mM) imidazole concentrations, each containing 0.04% (w/v) β-DDM. H.ChlG was eluted by incubation of the resin with 400 µl of elution buffer (binding buffer containing 400 mM imidazole and 0.04% (w/v) β-DDM) for 1 h with gentle agitation at 4°C before the solution was filtered through a 0.22 µm spin column, separating the resin from the eluted protein. Eluates were analysed immediately or stored at −80°C.

Pigments were extracted from cell pellets or FLAG-/His-eluates in 100% methanol and separated by RP-HPLC on an Agilent 1200 HPLC system using a Discovery HS C₁₈ column (5 µm, 250 × 4.6 mm) according to the slightly modified method of that of Lagarde and Vermaas [35] described in Proctor et al. [27]. Absorbance was monitored at 450 nm and 665 nm and carotenoid species and Chl a were identified by their known absorption spectra (Supplementary Figure S1) and retention time (Supplementary Figure S2 and reference [27]).

Proteins were extracted from the FLAG eluates by precipitation using a 2-D clean-up kit (GE Healthcare) and processed according to Hitchcock et al. [36] to generate tryptic peptides. Analysis was performed by nano-flow reversed-phase chromatography coupled to a mass spectrometer using system parameters described by MacGregor-Chatwin et al. [37] with the exception that peptides were resolved with a 75 min gradient in this study. Proteins were identified and quantified using MaxQuant v.1.5.3.30 [38] to search a Synechocystis proteome database (http://genome.microbedb.jp/cyanobase/).

Protein electrophoresis was performed as reported in our previous work [21,27]. Proteins in the FLAG- and His-eluates were separated by SDS–PAGE on Invitrogen 12% Bis–Tris NuPage gels (Thermo Fisher Scientific) and visualised by staining with Coomassie Brilliant Blue (Bio-Rad). For blue-native (BN)-PAGE, solubilised thylakoid membrane protein complexes, prepared as outlined above, were separated on 8–16% BN-gels [25] to resolve ChlG–HliD and Ycf39–HliD complexes. Protein complexes were further resolved by incubating the BN-gel strip in 2% (w/v) SDS and 1% (w/v) dithiothreitol for 30 min at room temperature followed by separation of individual protein components in the second dimension by SDS–PAGE in a denaturing 12 to 20% (w/v) polyacrylamide gel containing 7 M urea. For immunoblotting, proteins were transferred onto polyvinylidene fluoride (PVDF) membranes (Thermo Fisher Scientific) and incubated with specific primary antibodies against the 3xFLAG tag (Sigma-Aldrich), His₆-tag (Merck), HliD (Agrisera AS10-1615), ChlG (described in reference [21]) or Ycf39 [21] followed by an appropriate secondary antibody (anti-rat for 3xFLAG, anti-mouse for His₆ and anti-rabbit for HliD, ChlG and Ycf39) conjugated with horseradish peroxidase (Sigma-Aldrich) to allow detection using the WESTAR ETA C 2.0 chemiluminescent substrate (Cyanagen) with an Amersham Imager 600 (GE Healthcare).

Chl content was determined spectrophotometrically following extraction from cell pellets (from 1 ml of culture at OD₇₅₀ ≈ 0.4) with 100% methanol according to Porra et al. [39]. Chl precursors were extracted from cell pellets (from 2 ml of culture at OD_750nm ≈ 0.4) of WT and mutant Synechocystis strains (five biological replicates per strain) and analysed by RP-HPLC with two fluorescence detectors, as described previously [40]. Equivalent peaks were integrated, summed, and calculated as a percentage of the WT values, which were set as 100%.

Photosynthetic carotenoids are C₄₀ molecules synthesised from 8 C₅ isoprene units [41]. Synechocystis accumulates four major carotenoid species, β-carotene, zeaxanthin, myxoxanthophyll and echinenone, and produces others in lesser amounts, including synechoxanthin, 3′-hydroxy-echinenone and β-cryptoxanthin [35,42]; the molecular structures of the major carotenoids are presented in Figure 1B. An overview of carotenoid biosynthesis from all-trans-lycopene, the last common precursor of all the mature carotenoids synthesised by Synechocystis, is given in Supplementary Figure S3. Briefly, lycopene is cyclised at one or both of its ψ-ends producing γ-carotene (one β-ionone ring) or β-carotene (two β-rings); the myxoxanthophyll biosynthesis pathway branches from γ-carotene, whereas the other carotenoids are produced by modification of β-carotene. The carotenoid contents of WT Synechocystis and a strain containing an N-terminally 3xFLAG-tagged ChlG but lacking the native chlG (FG/ΔG) were analysed by RP-HPLC and the four expected major carotenoid species were identified (Supplementary Figure S2).

Mutants unable to synthesise β-carotene display severe growth phenotypes because it is required for assembly of PSII [4,7] but xanthophylls are dispensable for photoautotrophic growth under our low-stress laboratory conditions [35,42,43]. Mutants lacking xanthophyll biosynthesis genes were generated in the WT and FG/ΔG strains (Table 1; Supplementary Figure S4), resulting in identical carotenoid deficiencies in both backgrounds (Supplementary Figure S2; summarised in Table 2).

Deletion of the crtR gene, which encodes the β-carotene hydroxylase acting at the 3/3′ positions of the ionone rings of β-carotene, prevented biosynthesis of both zeaxanthin and myxoxanthophyll and resulted in the appearance of a new carotenoid species, previously identified as 3-dehydroxy-myxoxanthophyll (myxoxanthophyll missing the hydroxyl group on the β-ring [35,44]). It is not possible to generate a knockout strain that produces myxoxanthophyll but not zeaxanthin owing to the shared requirement of CrtR for synthesis of both carotenoids, however, myxoxanthophyll biosynthesis is specifically halted at the first dedicated step, 1′,2′-hydroxylation of lycopene/γ-carotene, in the absence of the C-1′-hydroxylase CruF (Supplementary Figure S3) [45,46]. Deletion of cruF generated a mutant that did not contain any myxoxanthophyll or myxoxanthophyll-specific precursors but otherwise had a normal carotenoid quota; cruF deletion in the ΔcrtR strain resulted in a strain that lacks 3-dehydroxy-myxoxanthophyll as well as zeaxanthin and myxoxanthophyll, accumulating only β-carotene and echinenone in significant amounts (Supplementary Figure S2). The only other confirmed enzyme in myxoxanthophyll biosynthesis in cyanobacteria is the 2′-O-glycosyltransferase CruG, which adds a sugar moiety to the carotenoid backbone (Supplementary Figure S3; [45]). Deletion of the Synechocystis cruG homologue from the WT and ΔcrtR strains resulted in accumulation of myxol or 3-dehydroxy-myxol, respectively (Supplementary Figure S2). Finally, we constructed a ΔcrtO mutant lacking the FAD-dependent β-ionone ring ketolase, which produces myxoxanthophyll and zeaxanthin but is unable to synthesise the keto-carotenoids echinenone and 3′-hydroxy-echinenone [43].

Parallel immunoprecipitations of FLAG-ChlG (F.ChlG) from the FG/ΔG strain and the carotenoid mutants were performed to investigate the effect of xanthophyll deficiency on the ChlG–HliD interaction in vitro (Figure 2). Retrieval of F.ChlG from each strain was confirmed by SDS–PAGE (Figure 2A) and immunoblotting with FLAG and ChlG specific primary antibodies (Figure 2B). F.ChlG from the FG/ΔG strain co-eluted with both HliD and Ycf39 as observed previously [21,27]; note that F.ChlG from strains lacking crtR migrated slightly further than expected on SDS–PAGE gels, which is discussed below.

Only a residual immunoblot signal for HliD was observed in the FG/ΔG/ΔcrtR and FG/ΔG/ΔcrtR/ΔcruF eluates, accompanied by the loss of signal for Ycf39; the level of HliD was comparable in thylakoids isolated from each of the strains (Figure 2C), ruling out the possibility of any pleiotropic effect on the production of HliD associated with deletion of crtR. Consistent with the absence of HliD, direct measurement of the absorbance spectra of the eluates from the FG/ΔG/ΔcrtR and FG/ΔG/ΔcrtR/ΔcruF strains revealed a drastic reduction in pigmentation compared with the visibly orange eluate from the FG/ΔG strain (Figure 2D). Qualitative analysis of the pigments in the eluates by RP-HPLC confirmed Chl, β-carotene myxoxanthophyll and zeaxanthin were present in the complex isolated from the FG/ΔG strain (Figure 2E). Small amounts of dehydroxy-myxoxanthophyll, Chl and β-carotene were present in the FG/ΔG/ΔcrtR elution; the Chl and β-carotene likely originate from trimeric PSI which contaminates FLAG-tag pulldowns [21].

In contrast with those from strains lacking crtR, FLAG-immunoprecipitation eluates from the FG/ΔG/ΔcruF mutant were visibly orange and spectrally similar to those from the FG/ΔG parent strain (Figure 2D). Immunoblot analysis of the FG/ΔG/ΔcruF eluate gave clear signals for both HliD and Ycf39 and zeaxanthin was identified by RP-HPLC analysis of the extracted pigments (Figure 2B,E). Removing the sugar group from myxoxanthophyll (producing myxol) or 3-dehydroxy-myxoxanthophyll (producing 3-dehydroxy-myxol) by deletion of cruG from the FG/ΔG and FG/ΔG/ΔcrtR backgrounds, respectively, did not alter the results compared with those of the respective parent strain (Supplementary Figure S5). Finally, deletion of crtO did not affect the composition of the F.ChlG complex (Supplementary Figure S6), confirming that keto-carotenoids are not involved in the ChlG–HliD interaction.

Quantitative proteomic analysis by mass spectrometry was also used to compare the levels of HliD and Ycf39 co-isolated with F.ChlG from the different carotenoid mutants (Figure 2F). When normalised to the amount of bait protein, the absence of myxoxanthophyll (FG/ΔG/ΔcruF) did not significantly change the amount of HliD co-purified with F.ChlG compared with the FG/ΔG control (P = 0.35). However, the absence of zeaxanthin (in the FG/ΔG/ΔcrtR strain) or both zeaxanthin and myxoxanthophyll (in FG/ΔG/ΔcrtR/ΔcruF) drastically reduced the level of HliD to 2–3% of the control level (Supplementary Table S2). These effects on the HliD:F.ChlG stoichiometry were mirrored by that of Ycf39:F.ChlG, with no significant change after elimination of myxoxanthophyll (P = 0.67) and average decreases to 10–20% of the control level in the zeaxanthin-less strains.

As stated above, F.ChlG isolated from strains lacking crtR migrated slightly faster on SDS–PAGE gels (Figure 2A), suggesting the protein was somehow smaller. Sequencing of the psbAII locus in these FG/ΔG/ΔcrtR strains confirmed the loss of one of the 3×FLAG-epitopes, leaving a 2×FLAG-tag in frame with the chlG gene (Supplementary Figure S7). Using a freshly isolated FG/ΔG/ΔcrtR mutant confirmed to replace the 3×FLAG-tag and performing parallel co-immunoprecipitations alongside a strain with a 2×FLAG-tagged enzyme showed that the length of the tag did not affect the protein or pigment profiles of the eluates (Supplementary Figure S8).

The 3xFLAG-tag is relatively long (24 amino acids) and highly positively charged. To confirm that this extra amino-acid sequence does not generate artefacts regarding the interaction with xanthophylls, the experiments were repeated with an N-terminally 10xHis-tagged ChlG (H.ChlG; Supplementary Figures S4, S9). Purification of the His-tagged enzyme by immobilised nickel-affinity chromatography resulted in a visibly pigmented eluate with very similar absorbance properties to the FLAG-immunoprecipitation complex (Supplementary Figure S10C). Analysis of the eluate by SDS–PAGE and immunoblotting showed a prominent band corresponding to H.ChlG (Supplementary Figure S10A,B); immunoblots also confirmed the presence of HliD (Supplementary Figure S10B), and zeaxanthin and myxoxanthophyll were identified by RP-HPLC (Supplementary Figure S10D). Consistent with the results with the FLAG-tagged enzyme (Figure 2), HliD does not co-elute with H.ChlG in the absence of zeaxanthin and myxoxanthophyll (Supplementary Figure S10A,B) and the eluate contained very low levels of pigments (Supplementary Figure S10C,D).

Previous studies have shown that Synechocystis ΔhliD mutants do not display a growth phenotype or altered accumulation of photosystems under standard low-stress laboratory growth conditions [47–49]. However, deletion of hliD did decrease the level of ChlG, resulting in a concomitant six-fold increase in the level of its substrate Chlide a [21]. We generated an independent hliD deletion in the FG/ΔG background (Supplementary Figures S2, S4) in order to determine the inherent pigment binding properties of isolated F.ChlG in the absence of HliD. In agreement with previous reports, there was a marked reduction in the level of F.ChlG isolated from FG/ΔG/ΔhliD using the same amount of starting material (solubilised thylakoid membranes), although it was possible to isolate Coomassie-stainable quantities of the protein (Figure 3A). As expected, Ycf39 was not detectable by immunoblot (Figure 3B) and the eluate lacked Chl and carotenoids, evident from both the absorbance spectra (Figure 3C) and RP-HPLC analysis of extracted pigments (Figure 3D), confirming that ChlG alone does not bind carotenoids. Consistent with the result obtained with the FLAG-tagged enzyme, isolation of H.ChlG from a ΔhliD background also resulted in a decreased level of ChlG (Supplementary Figure S10A,B) and an immobilised nickel-affinity chromatography eluate that lacked pigments (Supplementary Figure S10C,D).

The results presented above indicate that xanthophylls are required to maintain the interaction between ChlG and HliD, and that ChlG does not co-purify with carotenoids in the absence of HliD. To determine whether the interaction could be restored in isolated membranes, solubilised membranes from the FG/ΔG/ΔhliD strain, which synthesises the normal complement of carotenoids but lacks HliD, were incubated with those from the FG/ΔG/ΔcrtR/ΔcruF strain, which lacks zeaxanthin and myxoxanthophyll but produces HliD (schematically illustrated in Figure 4A). F.ChlG was subsequently isolated from the individual or mixed membrane samples by immunoprecipitation (Figure 4B). Unlike eluates from either individual sample, immunoblotting detected both HliD and Ycf39 in the elution from the mixed membranes (Figure 4C). Although considerably less pigmented than that from the FG/ΔG strain, the eluate from the mixed sample was visibly coloured and the absorbance spectra confirmed a small but clear increase in pigmentation compared with the eluates from the two mutant strains (Figure 4D). Zeaxanthin and myxoxanthophyll were both present in the complex isolated from the combined membranes (Figure 4E), confirming that the ChlG–HliD interaction can reform upon provision of xanthophylls. The ChlG–HliD complex was also restored in an analogous experiment with membranes from the equivalent H.ChlG strains (Supplementary Figure S11).

Isolation of FLAG- or His-tagged ChlG complexes requires washing steps that might disrupt interactions with weakly binding proteins. To further ascertain the effects of the loss of xanthophylls on the ChlG–HliD complex we used an alternative approach, separating solubilised thylakoid membrane proteins by two-dimensional BN/SDS–PAGE followed by immunoblotting (Figure 5). Most of the detectable ChlG in WT membranes appears to co-migrate with HliD in a ∼100 KDa complex; this complex likely forms the major fraction of our co-immunoprecipitated F.ChlG and it is likely to be composed of ChlG associated with several copies of HliD [21]. A larger ChlG–HliD complex is also detected; we refer to the smaller ChlG–HliD assembly as complex 1 and the larger one as complex 2, as indicated above the figure. Small amounts of free ChlG and HliD, and the Ycf39–HliD complex [28], which migrates slightly faster than the ChlG–HliD complex 1, are also clearly observed.

Repeating the analysis with membranes of the ΔcrtR/ΔcruF double mutant lacking both zeaxanthin and myxoxanthophyll, the ratio between the ChlG–HliD complex 1 and unattached ChlG differed from the WT sample, with much less ChlG present in the complex with HliD and more in a free form. Along with the clear reduction in ChlG–HliD complex 1, the larger complex 2 is almost completely absent from the ΔcrtR/ΔcruF mutant. In contrast, the Ycf39–HliD interaction is unaffected by the removal of xanthophylls. This 2D-BN/SDS–PAGE analysis supports the proposed requirement for xanthophylls for the stability of the ChlG–HliD complexes in the thylakoid membranes of Synechocystis.

As discussed above, deletion of hliD results in a decreased cellular level of ChlG and accumulation of its substrate, Chlide [21]. To determine the effect of the absence of xanthophylls on Chl biosynthesis, the levels of Chl and its biosynthetic intermediates (see Supplementary Figure S12 for an overview of the Chl biosynthesis pathway) were compared in WT, ΔcrtR, ΔcruF and ΔcrtR/ΔcruF cells (Figure 6). Under standard growth conditions the whole-cell absorbance spectra were similar for all four strains except for the region where carotenoids absorb (450–500 nm; Figure 6A). The Chl level of the ΔcruF (4.8 ± 0.3 µg Chl per ml of 1 OD₇₅₀ unit cells) and ΔcrtR/ΔcruF (5.2 ± 0.1 µg ml⁻¹) strains were not significantly different from that of the WT (5.1 ± 0.2 µg ml⁻¹), but the ΔcrtR mutant did display a small but significant (P = 0.04) decrease in Chl (4.6 ± 0.2 µg ml⁻¹) (Figure 6B).

Interestingly, the level of the ChlG substrate monovinyl (MV)-Chlide was 6–7 times higher in the ΔcrtR and ΔcrtR/ΔcruF mutants compared with the WT (set as 100%) (Figure 6C), indicating that ChlG activity is affected by the loss of zeaxanthin. Similarly, divinyl (DV)-Chlide, the substrate of the preceding enzyme in the pathway, 8-vinyl reductase (8VR), also accumulated ∼4-fold in both strains compared with the WT, suggesting the 8VR reaction is also affected, either directly or by the build-up of MV-Chlide. In contrast, the ΔcruF strain contained both Chlide species at levels comparable to the WT. The only other statistically significant differences in precursor levels were the decrease in magnesium-protoporphyrin IX in the ΔcrtR mutant and the increase in magnesium protoporphyrin monomethyl ester in ΔcruF. The reason(s) for these small variations in earlier pathway intermediates are likely due to pleiotropic effects resulting from the loss of xanthophylls in these strains.

We have shown that zeaxanthin stabilises the ChlG–HliD interaction in Synechocystis membranes. The alternating single–double carbon–carbon bonds in the polyene chain of carotenoids makes them much more rigid than other hydrophobic molecules, such as comparatively flexible lipids, and a scaffolding role for carotenoids in LHC proteins in plants has been proposed previously (reviewed by [13]). Although HliD and ChlG still associate to a minor extent in the absence of both zeaxanthin and myxoxanthophyll, this complex appears to be considerably less stable than in WT cells since the HliD protein is almost completely absent from F.ChlG immunoprecipitation eluates from a ΔcrtR mutant. Therefore, neither β-carotene, echinenone nor 3-dehydroxy-myxoxanthophyll (myxoxanthophyll with an unhydroxylated β-ring which accumulates in the ΔcrtR mutant [35,44]) can substitute for zeaxanthin/myxoxanthophyll in stabilising the ChlG–HliD interaction. These results suggest that the hydroxyl groups on the β-ring(s) of zeaxanthin and myxoxanthophyll, which allow xanthophylls to be held perpendicular to the membrane through strong hydrogen bonds, enhancing their structural function [13], are essential for their interaction with ChlG and HliD. Another Hlip (HliA/B) pigment-protein complex has also been reported to contain zeaxanthin and myxoxanthophyll [50], while zeaxanthin binds to the conserved C-terminal transmembrane Chl a/b binding (CAB) domains of dimeric ferrochelatase [25]; thus, xanthophylls may play similar roles in other Hlip-/CAB-domain protein assemblies in Synechocystis.

The ChlG–HliD interaction was maintained in a ΔcruF mutant that produces zeaxanthin but not myxoxanthophyll, indicating that zeaxanthin alone can mediate association of the two proteins. However, reconstitution of the ChlG–HliD interaction in isolated membranes showed both myxoxanthophyll and zeaxanthin were incorporated into the re-formed complex; thus, promiscuity versus specificity of zeaxanthin and myxoxanthophyll binding sites in the ChlG–HliD complex requires further study. Insight into any specific role of myxoxanthophyll requires a strain that produces myxoxanthophyll in the absence of zeaxanthin, but this is not possible by deletion of native genes because of the shared requirement of CrtR for biosynthesis of both carotenoids. We attempted to generate such a strain using CrtR from the filamentous cyanobacteria Nostoc sp. PCC 7120, which produces keto-myxoxanthophyll species but does not synthesise zeaxanthin [51–53]. However, our preliminary results indicate that neither myxoxanthophyll nor zeaxanthin biosynthesis was restored when alr4009 was expressed in place of the native crtR or at the psbAII locus in the ΔcrtR mutant (data not shown).

Isolation of native ChlG complexes from other chlorophototrophic organisms has not yet been reported but given the high-level of sequence conservation of cyanobacterial ChlG enzymes (e.g. 84% identity in Synechocystis and Synechococcus sp. PCC 7002) we predict that ChlG–xanthophyll–HliD complexes will be conserved in cyanobacteria. In support of this, we previously reported that ChlG from Synechococcus sp. PCC 7002 produced in Synechocystis interacts with HliD, zeaxanthin and myxoxanthophyll [27]. The myxoxanthophyll species produced in Synechocystis is myxol-2′ dimethylfucoside, whereas Synechococcus sp. PCC 7002 produces myxol-2′ fucoside, which lacks two methyl groups on the sugar moiety [45,54], indicating that the fucose group is not important for the interaction of myxoxanthophyll with the complex, as found with the ΔcruG mutant here.

Although we cannot rule out that altered thylakoid membrane carotenoid content in the xanthophyll mutants may affect the detergent-sensitivity of protein complexes, our data still supports a role for xanthophylls in stabilising the interaction of ChlG and HliD. Destabilising this association by removal of xanthophylls or deletion of HliD results in a significant build-up of Chlide and, in the latter case, the accumulation of ChlG is also reduced. However, consistent with previous results, there are no major phenotypic consequences relating to growth or Chl biosynthesis upon the loss of HliD, at least under our standard, low-stress laboratory growth conditions. It is possible that the ChlG–xanthophylls–HliD complex is particularly important under specific stress-conditions. For example, the xanthophylls could quench singlet oxygen produced during photo-oxidative stress; work is underway to elucidate how xanthophyll-mediated recruitment of HliD affects Chl metabolism in cyanobacteria in vivo.

In contrast with cyanobacterial ChlG, the enzymes from Arabidopsis thaliana (Arabidopsis) and the green alga Chlamydomonas reinhardtii do not associate with HliD or xanthophylls when heterologously produced in Synechocystis [27]. These phototrophs lack Hlips, and decreased sequence identity between plant and cyanobacterial ChlG (e.g. 63% for Arabidopsis and Synechocystis) may explain why the eukaryotic enzymes do not interact with HliD. Plants do, however, contain single helix LHC-like proteins that are homologous to Hlips called One-Helix Proteins (OHPs) (Supplementary Figure S13) [55]. Like HliD and HliC, which are predicted to form heterodimers in cyanobacteria [22,28], Arabidopsis OHP1 and OHP2 have been shown to dimerise in vivo [56]. Furthermore, OHPs bind the plant homologue of Ycf39 (HCF244) and are suggested to function in pigment delivery to newly synthesised PSII subunits [56–58], as proposed for the cyanobacterial Ycf39–Hlip complex [28,59]. It is currently unclear whether Ycf39–Hlip-dependent synthesis of PSII subunits relies on the additional interaction of Hlips with ChlG. If so, then OHPs (and HCF244) may associate with ChlG in plants as well, although evidence of such a complex is yet to be reported. Plants also synthesise LIL3 [55], a two transmembrane helix protein with a proposed role in Chl biosynthesis; LIL3 interacts with the Chl biosynthesis enzymes protochlorophyllide oxidoreductase (POR) and geranylgeranyl diphosphate reductase (ChlP), although whether it forms a complex with ChlG is not clear and may be species-specific [60–62]. Further work is required to clarify the interactions between these LHC-like proteins and Chl biosynthesis enzymes in plants, but it is feasible that zeaxanthin and/or other abundant plant xanthophylls, such as lutein, may play stabilisation roles in such complexes.

The authors declare that there are no competing interests associated with the manuscript.

M.S.P. was supported by a University of Sheffield Faculty of Science PhD Studentship. M.P., J.P. and R.S. were supported by grant 19-29225X from the Czech Science Foundation. M.J.D. acknowledges award BB/M012166/1 from the Biotechnology and Biological Sciences Research Council (BBSRC) UK. M.P.J. acknowledges award RPG-2019-045 from the Leverhulme Trust. C.N.H. acknowledges funding from the European Research Council (Synergy Award 854126) and the BBSRC UK (award BB/M000265/1). A.H. acknowledges support from a Royal Society University Research Fellowship (award number URF\R1\191548).

M.S.P., D.P.C., C.N.H., R.S. and A.H. conceived the study and designed the experiments. M.P.J., C.N.H., R.S. and A.H. supervised the project. M.S.P., M.P., P.J.J., J.P., E.C.M., M.J.D., R.S. and A.H. performed the experiments and/or analysed the data. M.S.P., R.S. and A.H. wrote the manuscript.

 
• BN-PAGE

  blue native polyacrylamide gel electrophoresis

 
• Chl(s)

  chlorophyll(s)

 
• ChlG

  chlorophyll synthase

 
• Chlide

  chlorophyllide

 
• F.ChlG

  FLAG-chlorophyll synthase

 
• H.ChlG

  His-chlorophyll synthase

 
• HliC

  high light-inducible protein C

 
• HliD

  high light-inducible protein D

 
• Hlips

  high light-inducible proteins

 
• LHCs

  light-harvesting complexes

 
• PSI

  photosystem I

 
• PSII

  photosystem II

 
• RP-HPLC

  reverse-phase high-performance liquid chromatography

 
• SDS–PAGE

  sodium dodecyl sulfate polyacrylamide gel electrophoresis

 
• WT

  wild-type

 
• β-DDM

  n-dodecyl-β-maltoside