The antenna-like domain of the cyanobacterial ferrochelatase can bind chlorophyll and carotenoids in an energy-dissipative configuration

Ferrochelatase (FeCh) is an essential enzyme catalyzing the synthesis of heme. Interestingly, in cyanobacteria, algae, and plants, FeCh possesses a conserved transmembrane chlorophyll a/b binding (CAB) domain that resembles the first and the third helix of light-harvesting complexes, including a chlorophyll-binding motif. Whether the FeCh CAB domain also binds chlorophyll is unknown. Here, using biochemical and radiolabeled precursor experiments, we found that partially inhibited activity of FeCh in the cyanobacterium Synechocystis PCC 6803 leads to overproduction of chlorophyll molecules that accumulate in the thylakoid membrane and, together with carotenoids, bind to FeCh. We observed that pigments bound to purified FeCh are organized in an energy-dissipative conformation and further show that FeCh can exist in vivo as a monomer or a dimer depending on its own activity. However, pigmented FeCh was purified exclusively as a dimer. Separately expressed and purified FeCH CAB domain contained a pigment composition similar to that of full-length FeCh and retained its quenching properties. Phylogenetic analysis suggested that the CAB domain was acquired by a fusion between FeCh and a single-helix, high light-inducible protein early in the evolution of cyanobacteria. Following this fusion, the FeCh CAB domain with a functional chlorophyll-binding motif was retained in all currently known cyanobacterial genomes except for a single lineage of endosymbiotic cyanobacteria. Our findings indicate that FeCh from Synechocystis exists mostly as a pigment-free monomer in cells but can dimerize, in which case its CAB domain creates a functional pigment-binding segment organized in an energy-dissipating configuration.

Heme is an essential cofactor for virtually all forms of life, and it is produced by a conserved tetrapyrrole pathway. The last step of heme biosynthesis is catalyzed by ferrochelatase (FeCh; protoheme ferro-lyase, EC 4.99.1.1), 2 which inserts iron into protoporphyrin IX (PPIX), a metal-free tetrapyrrole ring. Apart from serving as a cofactor for a myriad of redox and sensor proteins and as a precursor for the synthesis of bilins, heme is also a crucial regulatory molecule (1). Among many other recognized regulatory functions, heme availability controls total metabolite flow through the tetrapyrrole pathway. This heme feedback loop, although not fully understood, seems to operate universally from unicellular bacteria to animals (mitochondria) and plants (plastids) (2)(3)(4).
In photosynthetic organisms, regulation of the tetrapyrrole pathway is more complicated, because PPIX is also a substrate for magnesium chelatase, which diverts PPIX toward the chlorophyll (Chl) branch ( Fig. S1A) (4). Considering the general phototoxicity of tetrapyrroles, there must be regulatory mechanisms to prevent an overproduction of one of the tetrapyrrole products at the expense of the others or a harmful accumulation of intermediates. Thus, in Chl producing organisms, the activity of FeCh not only just controls a regulatory pool of heme but also competes with magnesium chelatase for PPIX, the last common substrate of both the heme and Chl branches. Indeed, the activity of FeCh has been shown to be crucial for the regulation of the tetrapyrrole pathway in both plants and cyanobacteria; an impairment of FeCh causes a massive accumulation of PPIX (5,6). In contrast, low activity of magnesium chelatase resulted in a total down-regulation of the pathway but not in the accumulation of PPIX (7)(8)(9).
Interestingly, a unique feature of the FeCh enzymes of Chlproducing organisms is a conserved C-terminal transmembrane helix that resembles the first and third helix of lightharvesting complex (LHC) proteins including a Chl-binding motif (the so-called LHC or CAB domain; Fig. 1 and Fig. S1B). The FeCh in oxygenic phototrophs is thus the only known enzyme belonging to the LHC protein superfamily, a large and diverse group of proteins characterized by the presence of at least one LHC/CAB-domain. Apart from light-harvesting antennas, the superfamily includes LHC-like proteins that are apparently not involved in light harvesting but rather have regulatory or photoprotective functions (10 -15). Cyanobacterial LHC-like proteins are limited to the single-helix high lightinducible proteins (Hlips) and FeCh, but in algae and plants, the spectrum of these proteins is much broader. Apart from FeCh and the homologs of Hlips called one-helix proteins (OHPs), eukaryotes possess various two-and three-helix LHC-like proteins such as Lil3 proteins, early light-inducible proteins (ELIPs), and Psb33 (16 -21). ELIPs, Lil3s, and Hlips were already reported to bind Chl and carotenoids in vivo (22)(23)(24)(25). Based on the known structure of various LHC proteins (26,27), it is expected that two helices, each containing an LHC/CAB motif, must dimerize to create a functional Chl-binding site (25).
The function of the FeCh CAB domain has been previously addressed by genetic and biochemical approaches (28 -30); however, its role remains unclear. The CAB domain is connected to the catalytic part of FeCh via a variable linker (Fig. 1), and in the cyanobacterium Synechocystis PCC 6803 (hereafter Synechocystis), the FeCh enzyme lacking both the LHC domain and the linker still appeared to be fairly active but was unstable, and its accumulation in the cell was impaired (6). A truncation preserving the linker but eliminating the CAB domain yields an active and stable enzyme (29,30). Such a truncated enzyme is, however, strictly monomeric, whereas the full-length enzyme can dimerize, at least in vitro (29,30). In addition, a Synechocystis strain expressing FeCh without the CAB domain was sensitive to high light conditions (29).
In this work, we demonstrate that Synechocystis FeCh is present in the cell mostly as a pigment-less monomer but can conditionally dimerize, and in such cases, its CAB domain creates a functional pigment-binding segment. Interestingly, the Chl and carotenoids bound to the CAB domain are organized in an energy-dissipative configuration, as previously reported for Hlips (24). This is consistent with our phylogenetic analysis, suggesting that the FeCh CAB domain originates from a fusion between an ancient Hlip and FeCh enzyme early in cyanobacterial evolution.

Partially inhibited FeCh binds Chl and carotenoids
To address the longstanding question of whether the FeCh enzyme in oxygenic phototrophs can bind Chl, we affinity-purified a 3ϫFLAG-tagged version of FeCh (f.FeCh; Fig. 1) ectopically expressed in a Synechocystis strain with the original hemH gene deleted (29). The obtained eluate was further separated by size-exclusion chromatography (SEC), and the Chl absorbance was measured by a diode-array detector. However, the only detectable Chl in the eluate comigrated with traces of photosystem I complexes (PSI) and with detergent micelles (DDM), but not with the purified f.FeCh ( Fig. 2A). The FeCh, migrating close to DDM micelles, was the only protein detectable on the stained gel after the separation of all fractions collected from the SEC column (Fig. S2). It should be noted that PSI complexes are extremely Chl-rich (90 Chl molecules per monomeric PSI), and thus their Chl absorbance is typically detectable even if the level of PSI protein subunits is below the detection limit of our protein stain (compare Fig. 1A and Fig. S2). We hypothesized that the FeCh could only bind (sense) Chl molecules if there was an abnormally large pool of unbound "free" Chl in the thylakoid membranes. Low FeCh activity in Synechocystis was previously shown to up-regulate the levels of Chl precursors (6,31); however, whether such up-regulation leads to an accumulation of free Chl molecules in the membrane was not known.
To clarify this point we treated the Synechocystis cells with a specific FeCh inhibitor, N-methyl mesoporphyrin IX (MPP) (32). After 4 h of the MPP treatment, the cells were fed with 14 C-labeled glutamate to follow the rate of de novo Chl formation. The concentration of MPP (275 nM) was not significantly toxic to Synechocystis under our standard conditions, because the proliferation rate of the cells treated with this concentration of MPP for 24 h was only 6% slower than that of untreated cells (Fig. S3). However, the partial FeCh inhibition significantly accelerated Chl biosynthesis compared with untreated cells (Fig. 3, A and B). Further, to assess the amount of free (or loosely bound) Chl in the cell, the membrane fraction was isolated after 4 h of MPP-treatment and separated by clear-native electrophoresis. In the case of MPP-treated cells, the Chl fluorescence detected in the zone of free pigments (DDM micelles) was more than doubled than in untreated control (Fig. 4, A and B). We eluted and quantified pigments from this part of the gel, and the content of Chl correlated well with the fluorescence measurement (Fig. 4B). According to these results, ϳ0.7% of the total Chl loaded on the clear-native gel comigrated with DDM micelles, and this amount is roughly doubled after treatment with MPP (Fig. 4B). Given these results, we concluded that decreased FeCh activity results in excess production of Chl in the cell.
Notably, when we purified f.FeCh from the cells treated with MPP, it eluted from the SEC column as a higher-mass protein complex than when it was purified from cells without the inhibitor. Moreover, a peak of Chl absorbance (675 nm) clearly comigrated with the f.FeCh (Fig. 2B), and the absorbance spectrum of the f.FeCh peak (450 -500 nm region; Fig. 2C) also signaled the presence of carotenoids. We extracted pigments from the collected f.FeCh peak and separated the extract by HPLC, which revealed a complex mixture of carotenoids in addition to Chl molecules. ␤-Carotene (␤-car) was the most abundant

Binding of pigments on the ferrochelatase CAB domain
carotenoid but zeaxanthin, echinenone, and an unknown carotenoid species were present as well (Fig. 2D). We also found MPP bound to f.FeCh, which explains the high absorbance of f.FeCh in the blue spectral region ( Fig. 2C and Fig. S4). By recording Chl fluorescence during SEC, we discovered that the fluorescence of Chl molecules bound to f.FeCh is quenched: the fluorescence yield of Chl was only ϳ14% when compared with the fluorescence of Chl molecules in detergent micelles (Fig. 5). From these observations, we conclude that the Syn-echocystis FeCh possesses a functional Chl and carotenoidbinding motif, which coordinates pigments in an energy-dissipative configuration.

The pigment-binding form of FeCh is a homodimer
Based on the elution time from the SEC column, the apparent mass of f.FeCh was greater after incubation with MPP (Fig. 2, A  and B). However, on the stained gel, we detected no other proteins except f.FeCh (Fig. S2), which implies the formation of a Chl, carotenoids, and UV absorptions were recorded by a diode-array detector; minute fractions were collected and separated by SDS-electrophoresis, and f.FeCh protein was stained by SYPRO Orange (below SEC chromatogram, see Fig. S2 for full-sized gel pictures). The peak of f.FeCh is highlighted by pink shading. The asterisk marks the absorption of the 3ϫFLAG peptide used to elute the f.FeCh protein from the anti-FLAG gel; DDM marks detergent micelles containing Chl molecules. PSI [3] and PSI [1] indicate Chl bound to traces of trimeric and monomeric photosystem I, respectively. B, SEC separation of f.FeCh purified from Synechocystis cells treated for 4 h with a specific FeCh inhibitor (MPP). The f.FeCh peak, highlighted by cyan shading, was collected for pigment analysis. C, absorbance spectrum of the pigmented form of f.FeCh; the unusually high absorbance in the region of 415-435 nm is due to the MPP bound to f.FeCh (see Fig.  S4 for an HPLC analysis). D, pigments were extracted by methanol from the collected f.FeCh peak (B) and separated on a C18 column. Zea, ech, and ␤-car denote zeaxanthin, echinenone and, ␤-carotene, respectively. An asterisk indicates an unknown carotenoid species.

Binding of pigments on the ferrochelatase CAB domain
f.FeCh oligomer. According to the known structure of LHCII (33) and a structural model of Hlips (24,25), two LHC/CAB domains should dimerize to create a canonical 4-Chls and 2-carotenoid binding motif. The observed shift in f.FeCh mobility suggests that the capability of the Synechocystis FeCh to bind Chl is related to its dimerization. To exclude the possibility that the f.FeCh oligomerization is an artifact of protein purification or its tagging, we separated Synechocystis membrane complexes by SEC and immunodetected FeCh in the collected minute fractions. A similar shift in the mass of FeCh was detected in membranes from WT cells: MPP treatment resulted in a higher-mass FeCh complex (Fig. 6A). According to the column calibration (Fig. S5), this shift corresponds to 50 kDa, which is very close to the shift expected for the presence of a FeCh dimer. Therefore, the discrepancy (ϳ2 min) in the elution times of FeCh and f.FeCh from the SEC column is most likely because of the presence of the highly charged 3ϫFLAG tag.
To verify the inducible FeCh dimerization by an independent approach, we separated WT membrane proteins by 2D bluenative/SDS electrophoresis (BN/SDS-PAGE), blotted, and immunodetected FeCh. Although the FeCh pattern was a bit smeared, after treatment with MPP, it was evident that a new spot corresponding to a size of a FeCh dimer appeared (Fig. 6B). We repeated the same experiment using membranes isolated from the Synechocystis hemH-⌬347 strain containing FeCh without the CAB domain ( Fig. 1) (29). As expected, the FeCh-⌬347 enzyme was exclusively monomeric, and its mobility was not affected by MPP (Fig. 6B). These data demonstrate that the Synechocystis FeCh can exist in vivo as a monomer or as a homodimer, and the formation of FeCh dimer is stimulated by MPP.
To prove that the FeCh CAB domain is the pigment-binding site, the FeCh C terminus, including the linker, was expressed in Synechocystis as a His 6 -tagged protein (His-C-tn, 9 kDa; Fig. 1). We have previously shown that the His-C-tn protein interacts in vivo with the FeCh CAB domain and can be copurified with the full-length FeCh but not with a truncated, CAB-less derivative (29). To isolate the His-C-tn domain not "contaminated" by His-C-tn-FeCh heterodimers, we expressed this protein in the Synechocystis hemH-⌬324 strain containing FeCh lacking both the linker and the LHC domain ( Fig. 1) (6). The FeCh-⌬324 truncation, apart from its inability to bind His-C-tn (29), lowers the cellular activity of FeCh because of impaired stability of the enzyme, thus mimicking the free Chl accumulation effect of MPP (Fig. 4C). We purified His-C-tn using a nickel affinity column, and after removing contaminating proteins and pigments by anion-exchange chromatography, we obtained pure His-C-tn protein (Fig. 7A). Fig. 7B shows the absorption spectrum of the isolated His-C-tn protein; as revealed by HPLC analysis of the extracted pigments, the protein binds Chl, ␤-car, and zeaxanthin at a molar ratio 6: 0.9: 1. A small amount of echinenone and the unknown carotenoid copurified with fulllength FeCh was also detectable (Fig. 7C). Furthermore, the fluorescence of Chls bound to His-C-tn was also quenched (Fig.  7A) as in the case of purified FeCh. Therefore, we can conclude that the FeCh CAB domain is solely responsible for the binding of pigments; the catalytic domain is not required for its folding into a Chl-binding protein structure.

Occurrence and phylogeny of FeCh in cyanobacterial genomes
The FeCh enzyme found in distinct groups of oxygenic phototrophs, such as plants, green algae, diatoms, haptophytes, and  [1] and PSI [3], monomeric and trimeric photosystem I, respectively. PSII [1] and PSII [2], monomeric and dimeric photosystem II, respectively. Pigments migrating in gel with DDM micelles are indicated by dotted boxes. B, Chl fluorescence in the gel region marked by dotted boxes in A was quantified. This part of gel was further cut, the pigments were eluted by repeated incubation in an elution buffer containing DDM (see "Experimental procedures"), and the Chl content was quantified by HPLC. Error bars show S.D., n ϭ 3 independent experiments. C, membrane complexes prepared from WT and hemH-⌬324 mutant were separated by clear native electrophoresis, and Chl fluorescence was detected as described for A.

Binding of pigments on the ferrochelatase CAB domain
glaucophytes, is of cyanobacterial origin and possesses a conserved CAB domain (Fig. S1) (34). However, the origin of the FeCh CAB domain, as well as its distribution in currently sequenced cyanobacterial species, remains unclear. We searched for hemH genes in all of the currently available cyanobacterial genomes (a total of 462 genomes) and found that the vast majority (453 genomes) code for FeCh with the CAB domain. The only exceptions are the most basal clades of the cyanobacterial evolutionary tree (primordial Gloeobacter, Synechococcus, and Pseudanabaena strains) and two members of a derived clade of unicellular endosymbionts of diatoms ( Fig. 8 and Table 1). A complete list of the analyzed genomes is provided in Table S1. In all analyzed CAB domains (453 genomes), the EXXNXR Chl-binding motif is conserved.
Three Pseudoanabena strains, forming a sister group to the basal, CAB domain-less clade, (Fig. 8) possess two similar hemH genes (identity ϳ60%); one hemH copy contains the CAB domain, but this domain is absent in the second gene. This situation is unique, because no other cyanobacterial genomes contain a combination of hemH genes with and without the CAB domain (Table 1). Intriguingly, 90 cyanobacterial genomes code for a "FeCh-like" protein with very limited similarity to the cyanobacterial FeCh (20 -30%; Table 1). This gene also seems to be cyanobacteria-specific, because we found no homologs in the NCBI database (by BLAST search) after excluding cyanobacteria from the search. Whereas the FeCh-like protein substantially differs from archeal, bacterial, or eukaryotic FeChs on a sequence level (not shown), its tertiary structure predicted by I-TASSER (35) aligned very tightly with the structure of canonical human FeCh (root-mean-square deviation between 256 pruned atom pairs ϭ 0.37 Å; Fig. S7A). Moreover, His-263, Asp-340, and Glu-343 residues (if counted according to the human FeCh), which are essential for the extraction of protons from PPIX nitrogens (36), are conserved in all 90 sequences of the FeCh-like proteins. The same is true for the absolutely conserved Phe-337 residue placed onto the PPIX-binding pocked (Fig. S7B) (37). Strains containing the putative second FeCh are listed in Table S1.
Single-helix Hlips have already been proposed as ancestors of the whole LHC superfamily (16). Indeed, genes coding for Hlips are present in all of the analyzed genomes, including the strains lacking thylakoids and/or the FeCh CAB domain (Table 1). We therefore conclude that the FeCh CAB domain most likely originates from an ancient fusion between an Hlip protein and the FeCh enzyme in the ancestor of all modern cyanobacteria except for the three deepest-branching lineages (Fig. 8). Chl absorption and fluorescence were recorded by diode array and fluorescent detectors, respectively. PSI [3] and PSI [1] indicate Chl bound to traces of trimeric and monomeric PSI, respectively. Numbers in parentheses indicate Chl fluorescence yield. The efficiency of quenching was calculated from the ratio of fluorescence/absorbance at their respective maxima for Chls bound to FeCh and for Chl molecules in detergent micelles. The same extinction coefficient was assumed for both Chl populations. B, fluorescence maxima of the f.FeCh-pigment complex and free Chl molecules in DDM micelles recorded during SEC chromatography by a fluorescence detector after excitation at 440 nm.  Fig. S5 for details). PSI [3] and PSI [1] show the position of trimeric and monomeric PSI complexes. B, solubilized membrane fractions were separated by 2D BN-PAGE and blotted, and the FeCh was immunodetected. Only slices of the blot with the signal of FeCh are shown (see Fig. S6 for full-sized gels). The position of monomeric and dimeric FeCh is indicated.

Discussion
The FeCh enzyme equipped with the pigment-binding CAB domain is apparently an old invention of cyanobacteria (Fig. 8). Comparing our phylogenetic tree to the recent calibrated molecular clock analyses, the evolution of this fused protein can be dated back to the early Proterozoic, roughly coinciding with the Great Oxygenation Event (38). Our data support a scenario in which the CAB domain originates from a fusion between an ancient cyanobacterial FeCh and an Hlip. Although we cannot completely exclude the possibility that the FeCh CAB domain is the progenitor of the whole LHC superfamily including Hlips, it is indeed very unlikely that the CAB segment of FeCh, in such a case already present in the last common cyanobacterial ancestor, would be selectively lost in basal cyanobacterial lineages (Fig. 8). Furthermore, in the literature there is an example of another fusion between an Hlip and a conserved cyanobacterial protein, a homolog of the Synechocystis Ssl2148 protein (39), supporting the plausibility of FeCh-Hlip fusion.
The function of Hlips and their plant homologs (OHPs) has been connected to the biogenesis of Chl-binding proteins, particularly the subunits of PSII (21,40). All four Hlips present in Synechocystis (HliA-D) interact with either the D1 subunit (HliC/D) or the CP47 subunit of PSII (HliA/B/C) soon after these Chl proteins are synthesized (21). Hlips then remain associated with D1/CP47 to support certain stages of PSII assembly or, possibly, PSII repair (21). In addition, the HliC/D proteins associate with the Chl-synthase enzyme (41). The exact roles that Hlips play are not clear yet, but they most likely fulfill several different tasks: (i) photoprotection of PSII biogenesis (dissipation of absorbed energy), (ii) delivery of Chl molecules, and (iii) regulation of Chl metabolism (HliC/D) (10). We suggest that once the FeCh enzyme linked physically with an Hlip, the FeCh became targeted to the machinery responsible for the biogenesis and assembly of Chl-binding proteins. There is general agreement about the existence of specific membrane compartments in which the photosystems are assembled, in both cyanobacteria (42) and plastids (43). As a crucial regulatory enzyme, FeCh probably got close to a site where Chl molecules are produced (44) and distributed into Chl-binding apoproteins (45).
The ability of Hlips to bind Chl and carotenoids in an energy dissipative configuration (24) has been preserved in the FeCh CAB domain. The absorbance spectrum of the His-C-tn-pigment complex is similar to two Hlips that have been previously characterized (HliC and HliD) (24,46). However, unlike HliC and HliD proteins that exclusively contain ␤-car, the CAB domain also binds zeaxanthin and echinenone (Fig. 7C). The calculated molar ratio of Chl:␤-car:zeaxanthin (6:0.9:1) in the CAB domain is very similar to the ratio of Chl:␤-car identified in HliC and HliD (6 or 4 Chl per 2 ␤-car, respectively) (24,25). Two different ␤-carbinding sites have been unambiguously identified in Hlips (24,47). However, in CAB domain, one carotenoid-binding site could be less specific and more polar than that in Hlips, preferentially accommodating xanthophylls or keto-carotenoids. A similar shift to polar xanthophylls (lutein, violaxanthin) also happened during the evolution of the eukaryotic major and minor antenna proteins (26,48).
The original cyanobacterial FeCh was probably monomeric, because it is strictly monomeric if the CAB domain is eliminated ( Fig. 6B) (29), and the recombinant full-length Synechocystis FeCh can be prepared in an active monomeric form (30). Although there is a small fraction of dimeric FeCh in Synechocystis grown under our standard laboratory conditions (29), the vast majority of FeCh appears to be monomeric in vivo (Fig. 6B). Because the pigment-binding motif in Hlips is, however, formed by two parallel helices (24), the monomeric CAB domain cannot serve a photoprotective or Chl carrier role. To be capable of pigment binding the FeCh needs to dimerize via CAB domain (29); most likely by mimicking the LHC-like pair of helices (26) clipped together by salt bridges between Glu and Arg residues on the opposite helices (see also Fig. S1B).
A mechanistic explanation or a function of the observed conditional FeCh dimerization is beyond the scope of this report and must be addressed by future studies; however, based on the present data, we speculate that the formation of FeCh dimers is triggered in response to disturbed tetrapyrrole metabolism. The rather mild dose of MPP used in our study allows treated cells to produce enough heme for close to normal proliferation (Fig. S3); yet the regulatory function of heme was apparently

. The C-terminal segment of FeCh lacking the catalytic domain (an artificial His-C-tn protein) binds Chl, ␤-car, and zeaxanthin in vivo. A,
the His-C-tn protein was expressed in the Synechocystis hemH-⌬324 genetic background and purified on a nickel column. The obtained eluate was further separated by anion-exchange chromatography, and the Chl absorbance (675 nm) and Chl fluorescence were recorded (Chl FL; 440-nm excitation/675-nm emission). Note the strikingly different emission intensity of Chls in DDM micelles and Chls bound to His-C-tn. Minute fractions were collected and separated by SDS-electrophoresis, and the presence of His-C-tn protein was detected by antibodies raised against the FeCh C terminus. The core photosystem I and II subunits PsaA and D1 were also immunodetected as a control of purity; dashed line indicates blot splicing. Fractions highlighted by the gray box were pooled and used for the analysis of pigments associated with His-C-tn. B, the absorption spectrum of the His-C-tn protein purified by a combination of nickel-affinity and anion-exchange chromatography as described above. C, the extracted pigments from the purified His-C-tn protein were separated on a C18 column; the molar stoichiometry of the major constituents is indicated. zea, zeaxanthin; ech, echinenone. The unknown carotenoid (see Fig. 2D) is marked by an asterisk.

Binding of pigments on the ferrochelatase CAB domain
disturbed, resulting in a strongly up-regulated tetrapyrrole pathway. Under such conditions, a large pool of FeCh dimerizes. It should be noted that the His-C-tn protein (an artificial CAB domain) readily attaches to full-length FeCh via its CAB domain in vivo (29), suggesting that there is a control mechanism such that two FeChs form a stable dimer only under certain conditions. A possible explanation is that the monomeric FeCh is incorporated into a protein complex. A colocalization of cyanobacterial FeCh with protoporphyrinogen oxidase, the preceding enzyme in the pathway, has been reported (49), and it is worth noting that the mitochondrial FeCh is a component of large protein assembly dealing with heme metabolism (50). The proposed spatial separation of FeCh monomers in the cell is consistent with the fact that purified pigment-less Synechocystis FeCh spontaneously dimerizes (29). However, there is the possibility of in vitro artifacts caused by the high concentrations of the protein and/or handling and freezing the sample. In this work, the purified sample was immediately loaded into a SEC column. An alternative scenario is that the in vivo stability of the pigment-less FeCh dimer is weak, but after association with Chl and carotenoids, the dissociation constant of the pigmented dimer becomes much lower, resulting in the accumulation of FeCh dimer. This is indeed an attractive hypothesis, providing a mechanism for how the FeCh enzyme senses a risky accumulation of Chl molecules in the membrane, particularly if this occurs in the specialized compartment for Chl-protein biogenesis. The Synechocystis mutant lacking FeCh CAB domain was reported to be sensitive to high light and to accumulate chlorophyllide, the substrate of Chl synthase (29). We did not observe a large pool of FeCh dimer under high light conditions (not shown); however, it is possible that under physiological conditions, the level of FeCh dimer does not exceed a certain threshold and/or it is short-lived.
The amount of FeCh in Synechocystis appears to be much higher than that needed for synthesizing a sufficient amount of heme: strains with ϳ10% of normal FeCh activity do not exhibit any signs of phycobilisome deficiency, at least under low-stress conditions (6,31). This situation resembles the reserve flux capacity that enables an immediate (Ͻ1 min) rerouting of glucose-6-phosphate into the oxidative pentose phosphate path-

Binding of pigments on the ferrochelatase CAB domain
way in Escherichia coli (51). Synechocystis could have a similar capacity to promptly reroute PPIX into the heme pathway, away from magnesium chelatase. Whether FeCh dimerization together with pigment binding on the CAB domain influences its activity is unknown at this point. Unfortunately, we cannot assess the activity of pigmented FeCh because of the presence of MPP in the active site. However, in an in vitro assay, FeCh lacking the CAB domain shows increased catalytic turnover number (30), suggesting that there is an effect of this segment on the enzymatic parameters in vivo. In the same report, the addition of a methanol extract from Synechocystis membranes containing pigments lowered the FeCh activity irrespectively of the CAB domain (30). This observation might indicate that free Chl molecules have a direct effect on the enzyme catalytic site. However, the conditions in in vitro assay are very artificial, and also the methanol extract is a complex mixture of compounds; in the work of Storm et al. (30), a measurement with chemically pure Chl a is not provided. Potential modulation of FeCh activity by Chl, either via CAB domain or by interfering with catalytic site, thus remains an open question. Alternatively, instead of changing the catalytic properties, dimerization and binding of pigments on the CAB domain may alter the availability of PPIX for chelation by changing the FeCh interactome (as discussed above). In this scenario, FeCh would be regulated by substrate availability rather than by enzymatic activity.
Although FeCh activity is undoubtedly very important for the regulation of Chl biosynthesis (31,52), the molecular mechanism of this regulation is not known. The classical model of labile free heme directly controlling the formation of ALA (53) appears to be oversimplified. In plants, it has been proposed that two spatially separated pools of PPIX are located in plastids, each used for the synthesis of heme, but only one pool of PPIX also serving as a substrate for magnesium chelatase (52,54). In addition, there seems to be a dedicated regulatory heme pool produced by plastid FeCh1 isozyme (lacking the CAB domain) that is involved in plastid to nucleus signaling (55). Plant cells are much more complex than cyanobacterial cells; yet the CAB domain (FeCh2 isozyme) is conserved in plants, and there is no reason to expect that it has a fundamentally different function than in cyanobacteria. Predictably, FeCh2 is the key regulator of plant tetrapyrrole metabolism; FeCh1 activity does not affect the synthesis of tetrapyrroles in plastids (56). Because many cyanobacterial stains contain a putative second FeCh (Table 1), we speculate that the FeCh-controlled ALA synthesis and flow of PPIX into the Chl branch are limited to a specific compartment. This is in line with our proposed targeting of FeCh with the CAB domain into a site of Chl-protein biogenesis, where enzymes of the Chl pathway should be also localized (44,57).
Different oligomeric states of the FeCh enzyme are most likely involved in the sophisticated regulatory machinery controlling Chl metabolism, which in cyanobacteria also includes Hlips (10) and other proteins (58,59). Intriguingly, Synechocystis lycopene cyclase (CruA), an enzyme in carotenoid biosynthesis, was recently reported to bind Chl and carotenoids (60). This enzyme does not possess a CAB domain or helical transmembrane segments; the mechanism through which the lycopene cyclase binds Chl and carotenoids in a ratio and spectral characteristics resembling a typical Chl-complexes such as photosystem I thus remains enigmatic. In plants and algae, the conserved CAB/LHC motif is present within a broader spectrum of proteins that are otherwise diverse but also generally connected to Chl biosynthesis. For instance, the Msf1 protein recently identified in the alga Chlamydomonas reinhardtii is required for the accumulation of photosystem I and interacts with the cyclase enzyme in Chl biosynthesis (61). In Arabidopsis thaliana, Lil3.1/2 proteins are crucial for the synthesis of the Chl phytol chain (12,13,(62)(63)(64)(65), whereas OHPs assist during the synthesis of photosystem II as described above for the cyanobacterial Hlips (14,15). Association with a biogenesis compartment and (regulatory) pigment-driven dimerization could be thus common themes for LHC-like proteins.

Synechocystis strains and growth conditions
The glucose-tolerant Synechocystis WT-P substrain is referred to here as WT (66). The Synechocystis strain expressing 3ϫFLAG-tagged FeCh and lacking the native hemH gene has been described previously (29), as well as the strains possessing truncated FeCh derivatives (hemH-⌬324, hemH-⌬347) (6,29) or expressing the FeCh C-terminal domain as a His 6 -tagged protein (His-C-tn) (29). To express the His-C-tn protein in the hemH-⌬324 genetic background, the His-C-tn cells were transformed with chromosomal DNA isolated from the hemH-⌬324 cells, and transformants were selected on a BG-11 plate with Zeocin (3 mg/liter). Complete segregation of the mutant was verified by PCR. All strains were grown in BG-11 medium; solid media were further supplemented with 10 mM TES/NaOH (pH 8.2), 1.5% agar, and 0.3% sodium thiosulfate. Liquid cultures were grown either grown in 50 ml of BG-11 in 250-ml conical flasks on a rotary shaker or in 800 ml of BG-11 in 1-liter cylin- Table 1 Numbers of cyanobacterial genomes coding for the Hlips, the FeCh with or without the CAB domain, and the FeCh-like proteins A detailed list of all strains and identified hemH genes are provided in Table S1.

Binding of pigments on the ferrochelatase CAB domain
ders bubbled with air at 29°C with a surface irradiance of 40 mol photons m Ϫ2 s Ϫ1 of white light. The cultures were harvested in the exponential phase at a cell density of ϳ10 8 /ml (A 750 in a range of 0.8 -1.0) by centrifugation (5000 RCF, 5 min). Where applicable the FeCh inhibitor MPP (32) (Frontier Scientific) was added from a 10 mM stock in DMSO to a final concentration of 275 nM.

Affinity, size-exclusion, and anion-exchange chromatography
Pelleted cells were resuspended in isolation buffer A (20 mM MES, pH 6.5, 150 mM NaCl, complete protease inhibitor mixture; Roche) and mixed with glass beads of 100 -200 M diameter at a 1:1 ratio (1 volume of dense cell solution to 1 volume of glass beads). This suspension was vortexed six times for 10 s with 1 min on ice in between each cycle. After vortexing, the glass beads were washed four times with buffer A, and the obtained suspension of broken cells was pooled and centrifuged (40,000 RCF, 20 min). The pelleted membranes were resuspended in buffer A and frozen (Ϫ80°C) for further use. Isolated membranes containing 500 g Chl/ml were solubilized using 1% (w/v) DDM for 5 min on ice. His-and FLAG-tagged proteins were purified in buffer A at 10°C according to the manufacturer's instructions using Protino nickel-nitrilotriacetic acidagarose (Macherey-Nagel) and an anti-FLAG M2 affinity gel (Sigma-Aldrich), respectively. Prior to ion-exchange/size-exclusion chromatography, samples were concentrated approximately four times using 10-kDa cutoff microconcentrators (AmiconR Ultra 0.5 ml; Merck). The eluted and concentrated f.FeCh protein was immediately injected onto an Agilent-1200 HPLC system and separated on a Yarra SEC-3000 column (Phenomenex) using mobile phase (20 mM MES, pH 6.5, 100 mM NaCl, pH 6.5, containing 0.06% (w/v) DDM) at a flow rate of 0.25 ml min Ϫ1 at 15°C. Eluted proteins and complexes were detected using a diode-array detector and a fluorescence detector set to 440/680 nm (excitation/emission wavelengths). The membrane fraction solubilized by 1% (w/v) DDM was separated with an identical procedure as that of the f.FeCh eluate.
For anion-exchange separation of the His-C-tn protein, the concentrated eluate from the nickel column was injected onto an Agilent-1200 HPLC system and separated on a MonoQ 4.6/ 100 PE column (GE Healthcare) using a gradient of NaCl in 20 mM HEPES, pH 8.1, 5 mM MgCl 2 , 0.06% (w/v) DDM. The highsalt (HS) buffer was identical to the low-salt buffer but contained 1 M NaCl. The gradient was run as follows: 10 min, 0% HS; 15 min, 20% HS; 45 min, 55% HS; and 47 min, 100% HS at a flow rate of 0.8 ml min Ϫ1 at 15°C. Eluted proteins and complexes were detected using a diode-array detector and a fluorescence detector set to 440/675 nm (excitation/emission wavelengths).

Native and denaturing protein electrophoresis
Harvested cells were dissolved in isolation buffer B (25 mM MES, pH 6.5, 4 mM MgCl 2 , 4 mM CaCl 2 , 25% (w/v) glycerol, complete protease inhibitor mixture; Roche), mixed with glass beads of 100 -200 M diameter at a 1:1 ratio (1 volume of dense cell solution to 1 volume of glass beads), and broken (three times for 20 s) using a Mini-Beadbeater-16 (Biospec). The cells were kept on ice for 2 min between each cycle. The membranes were isolated as described above.
For the separation of membrane complexes on BN-PAGE, the isolated membranes containing 400 g Chl/ml were solubilized using 1% (w/v) DDM for 5 min on ice. Nonsolubilized material was removed by centrifugation (20,000 RFC, 10 min). Extracted membrane proteins corresponding to 4 g of Chl were mixed with 1/10 volume of sample buffer (750 mM aminocaproic acid, 5% (w/v) CBB-G) and separated at 4°C on an 8 -16% gradient polyacrylamide gel containing 500 mM aminocaproic acid and 50 mM Bis-Tris/HCl, pH 7.6, at room temperature according to Ref. 67. The starting blue cathode buffer (50 mM Tricine, 50 mM Bis-Tris/HCl, pH 7.0, 0.2% (w/v) CBB-G) was exchanged after one-third of the run for the clear cathode buffer (50 mM Tricine, 50 mM Bis-Tris/HCl, pH 7.0). As anode buffer, 50 mM Bis-Tris/HCl, pH 7.0, was used. In clear-native electrophoresis the cathode buffer contained 0.05% (w/v) sodium deoxycholate and 0.02% (w/v) DDM, and a 4 -12% gradient polyacrylamide gel was used. After separation, Chl fluorescence in the gel was detected using LAS-4000 (Fuji) after illumination with blue light.
To separate proteins using denaturing SDS-PAGE, a 16 -20% linear gradient polyacrylamide gel containing 7 M urea, 650 mM Tris/HCl, pH 8.6, was prepared according to (68). For the second dimension, gel strips from the native gel were excised and incubated for 30 min in 25 mM Tris/HCl, pH 7.5, containing 1% (w/v) SDS and 2% (w/v) DTT, and placed on top of the denaturing gel; two strips were analyzed on a single denaturing gel. Cathode buffer (25 mM Tris, 192 mM glycine, 0.2% (w/v) SDS) and anode buffer (25 mM Tris/HCl, pH 8.3) were used. A ratio of 1:60 for acrylamide to bis-acrylamide was used for all gels.

Protein immunodetection
Proteins separated by SDS-PAGE were stained for 50 min with SYPRO Orange (Thermo Fisher) 5000ϫ diluted in 10 mM Tris/HCl, pH 7.6, 150 mM NaCl, 10% (v/v) methanol and subsequently transferred (3 mA/cm 2 , 3h) onto an Immobilon-P membrane (0.45 m; Millipore). After electrotransfer, the membrane was blocked by 0.05% (w/v) Tween 20 in 10 mM Tris/HCl, pH 7.6, 150 mM NaCl for 30 min and subsequently incubated with a specific primary antibody and then with a secondary antibodyhorseradish peroxidase conjugate (Sigma). The peroxidase activity was visualized by incubating the membrane for 20 s in Luminata Crescendo Western HRP substrate (Sigma) and detected using LAS-4000 (Fuji).

Binding of pigments on the ferrochelatase CAB domain Chl radiolabeling and measurement of Chl precursors and pigments
To assess the rate of de novo Chl synthesis, cells corresponding to 50 mg of Chl in an exponential growth phase at density ϳ5 ϫ 10 7 /ml (OD 750 ϭ 0.5) were harvested by centrifugation (2700 RCF, 3 min), washed, and resuspended in fresh BG-11 medium containing 20 mM TES to a final volume of 540 l. The cell suspension was shaken in 10-ml glass tubes for 30 min at 29°C at 40 mol photons m Ϫ2 s Ϫ1 . Then 60 l of [ 14 C]Glu with specific activity 248 mCi/mmol (Moravek Biochemicals) was added. After 60 min, the cells were harvested (10,000 RCF, 2 min) and frozen in liquid nitrogen. MPP (275 nM) was added to the cells 3.5 h before the addition of [ 14 C]Glu and also kept at the 275 nM concentration during radiolabeling. Pelleted cells were washed and resuspended in 50 l of water; Chl was extracted with 1 ml of methanol, and insoluble material was discarded by centrifugation (10,000 RCF, 4 min). The extract was injected into an Agilent-1260 HPLC system equipped with a Radiomatic 150 TR scintillation detector (PerkinElmer Life Sciences) and a diode-array detector. Pigment separation was carried out on a Zorbax Eclipse Plus C18 column (4.6-m particle size, 3.9 ϫ 100 mm; Agilent) with 35% methanol, 15% acetonitrile in 0.25 M pyridine as solvent A and 20% methanol, 20% acetone in acetonitrile as solvent B. Pigments were eluted with a linear gradient of solvent B (30 -95% in 25 min) followed by 95% solvent B at a flow rate of 0.8 ml min Ϫ1 at 40°C. The relative rate of Chl synthesis was estimated from the ratio of Chl radio emission to Chl absorbance between MPP-treated and untreated samples.
To quantify free Chl in the native gel, a piece of gel (ϳ4 ϫ 5 mm) corresponding to the region of DDM micelles was cut and further chopped up into pieces (1 ϫ 1 mm). The obtained gel pieces were incubated for 2 h in 125 l of buffer A containing 0.04% DDM, and then the buffer was collected after centrifugation (10,000 RCF, 2 min). The incubation of gel pieces in the buffer A was repeated, the buffer fractions were pooled, and the eluted pigments were separated by HPLC as described above. The absolute Chl concentration was calculated from peak area using a Chl-a standard.
Chl precursors were quantified as described in Ref. 69. To identify pigments bound to the isolated f.FeCh and His-C-tn proteins, fractions collected during SEC and anion-exchange chromatography were concentrated on 10-kDa cutoff microconcentrators (AmiconR Ultra 0.5 ml; Merck). This concentrated solution was extracted with 90% methanol, and the extract was analyzed using the Agilent-1260 HPLC system, essentially as described above for the radiolabeled samples. Pigment stoichiometry was calculated from calibration curves prepared using authentic standards.

Nonphotochemical quenching of Chl bound to f.FeCh
Quenching of Chl fluorescence was estimated from the ratio of fluorescence/absorbance of Chl molecules bound to f.FeCh compared with unquenched Chl in DDM micelles. Identical fluorescence constants and extinction coefficients were assumed for both Chl populations.

Phylogenetic and bioinformatic analysis
A representative set of 550 publicly available cyanobacterial genome assemblies was downloaded from NCBI and utilized to create a custom BLAST database (70). FeCh protein from Synechocystis (slr0839 gene) was used as the query for a tBLASTn search with a cutoff value of 1.e Ϫ10 against this database, allowing multiple hits for each genome. All hits were automatically harvested and aligned using MAFFT v. 7 (71). Sequence analysis in Geneious Prime v. 2019 allowed us to remove short hits, reducing the number of genomes in the final data set to 462 and to assess the presence of a terminal CAB domain, specific conserved residues (Chl-binding motif), and pairwise amino acid sequence identity between individual deduced proteins in the matrix.
The multilocus phylogenetic tree was generated using a set of 22 conserved single-copy orthologous protein-coding genes selected from loci previously tested for congruent cyanobacterial phylogenies (72,73). The selection was made based on the presence of a full unambiguous sequence of the genes in a representative subset of genomes, and comprised these loci: dnaG, frr, infC, nusA, pgk, pyrG, rplB, rplC, rplD, rplE, rplF, rplM, rplP, rplT, rpoB, rpsC, rpsE, rpsI, rpsK, rpsM, rpsS, and smpB. First, BLAST queries were derived from respective proteins mined from Synechocystis (BA000022). Each of the 22 queries was then used in tBLASTn (cutoff value 1.e Ϫ15 ) searches against the custom database. Hits for each protein were aligned using MAFFT v. 7, and the alignments were manually reviewed to remove ambiguous sites and gap regions. A maximum likelihood tree was calculated using each of the 22 alignments with 1000 bootstrap pseudoreplications in RaxML v. 8 (74). The resulting phylogenies were inspected manually for topological incongruence. Because individual gene trees were generally consistent, all alignments were concatenated into a 5117-amino acid-long matrix containing 220 sequence lines. A Bayesian inference (BI) tree was obtained using MrBayes v. 3.2.6 (75). Two independent runs of eight Markov chains were performed for 1 million generations, sampling every 100th tree, until the likelihood values were stable, and the divergence criterion was lower than 0.01. The BI calculation employed a common LG ϩ I ϩ G evolutionary model because of the excessive computational demands of a calculation using partitioned models. Posterior probabilities were estimated from branch frequencies in the sampled trees, discarding the first 25% of the harvested data as burn-in. The maximum likelihood and BI calculations were run via the CIPRES supercomputing facility (76).