Evolution of Ycf54-independent chlorophyll biosynthesis in cyanobacteria

Significance

Photosynthesis uses chlorophylls to utilize solar energy. In oxygenic phototrophs, formation of the isocyclic fifth ring of chlorophyll, responsible for its green color, is catalyzed by AcsF/CycI and the auxiliary protein Ycf54. Removal of Ycf54 causes severe chlorophyll deficiency and impaired photoautotrophic growth. We analyzed laboratory-evolved suppressor mutants of a Ycf54-less strain of the cyanobacterium Synechocystis where chlorophyll biosynthesis and phototrophy were restored. A single point mutation in CycI significantly weakens its dependence on Ycf54, mimicking natural evolution of the enzyme in marine cyanobacteria that lack Ycf54. A second mutation resulting in overaccumulation of chlorophyll inactivates an enzyme with in vitro chlorophyll dephytylase activity. Our results provide insights into the important regulatory role of Ycf54 in chlorophyll biosynthesis.

Abstract

Chlorophylls (Chls) are essential cofactors for photosynthesis. One of the least understood steps of Chl biosynthesis is formation of the fifth (E) ring, where the red substrate, magnesium protoporphyrin IX monomethyl ester, is converted to the green product, 3,8-divinyl protochlorophyllide a. In oxygenic phototrophs, this reaction is catalyzed by an oxygen-dependent cyclase, consisting of a catalytic subunit (AcsF/CycI) and an auxiliary protein, Ycf54. Deletion of Ycf54 impairs cyclase activity and results in severe Chl deficiency, but its exact role is not clear. Here, we used a Δycf54 mutant of the model cyanobacterium Synechocystis sp. PCC 6803 to generate suppressor mutations that restore normal levels of Chl. Sequencing Δycf54 revertants identified a single D219G amino acid substitution in CycI and frameshifts in slr1916, which encodes a putative esterase. Introduction of these mutations to the original Δycf54 mutant validated the suppressor effect, especially in combination. However, comprehensive analysis of the Δycf54 suppressor strains revealed that the D219G-substituted CycI is only partially active and its accumulation is misregulated, suggesting that Ycf54 controls both the level and activity of CycI. We also show that Slr1916 has Chl dephytylase activity in vitro and its inactivation up-regulates the entire Chl biosynthetic pathway, resulting in improved cyclase activity. Finally, large-scale bioinformatic analysis indicates that our laboratory evolution of Ycf54-independent CycI mimics natural evolution of AcsF in low-light–adapted ecotypes of the oceanic cyanobacteria Prochlorococcus, which lack Ycf54, providing insight into the evolutionary history of the cyclase enzyme.

All oxygenic phototrophs rely on the unique chemical properties of chlorophyll (Chl) molecules, the cofactors that enable cyanobacteria, algae, and plants to carry out the light-harvesting and photochemical reactions of photosynthesis. Together with hemes, bilins, and vitamin B₁₂, Chls are produced by a branched tetrapyrrole biosynthetic pathway (1). The heme/bilin and Chl pathways bifurcate at the point of metal insertion into protoporphyrin IX (PPIX), where iron is chelated by ferrochelatase to generate heme or magnesium is inserted by magnesium chelatase (MgCH) to form Mg-protoporphyrin IX (MgP), the first dedicated intermediate of the Chl branch. Along with those for MgCH, the enzymes for Chl biosynthesis have been identified and together produce Chl when assembled in the heterologous host Escherichia coli (2). MgP is first methylated by MgP methyltransferase to produce MgP monomethyl ester (MgPME) and the pathway continues with MgPME cyclase catalyzing the formation of the isocyclic E ring to generate 3,8-divinyl protochlorophyllide a (DV PChlide a). In the next step, the D pyrrole ring is reduced by PChlide oxidoreductase (POR) to produce 3,8-divinyl chlorophyllide a (DV Chlide a), which is reduced by 8-vinyl reductase to produce 3-vinyl chlorophyllide a (MV Chlide a). Chl a biosynthesis is completed by attachment of hydrophobic phytol chain to MV Chlide a by Chl synthase. The core pathway from PPIX to DV Chlide a is shared by all phototrophs, and further reactions modify DV Chlide a to produce the variety of bacteriochlorophylls (BChls) and Chls that occur in nature (1).

Two forms of the MgPME cyclase are found in BChl/Chl biosynthesis (Fig. 1). In contrast to the O₂-sensitive radical-SAM enzyme (BchE) found in most anoxygenic phototrophs, a mechanistically unrelated O₂-dependent enzyme catalyzes the formation of DV PChlide a in cyanobacteria, algae, plants, and some purple bacteria (3). A genetic study with the purple betaproteobacterium Rubrivivax (Rvi.) gelatinosus identified a putative diiron monooxygenase named AcsF (aerobic cyclization system Fe-containing subunit) as the catalytic subunit of the O₂-dependent enzyme (4). AcsF homologs are now known to be widespread in photosynthetic organisms (5, 6). Intriguingly, most cyanobacteria contain two distinct AcsF isoforms, AcsFI and AcsFII, named CycI and CycII, respectively, in Synechocystis sp. PCC 6803 (hereafter Synechocystis). CycI is constitutively expressed, whereas CycII is additionally required under microoxic conditions (7, 8).

Fig. 1.

The O₂-dependent MgPME cyclase reaction. In most anoxygenic phototrophs, AcsF is the only subunit required for formation of the isocyclic E ring (highlighted). Alphaproteobacterial AcsF requires an auxiliary subunit (BciE) for activity, and another auxiliary subunit (Ycf54) is required for cyclase activity in oxygenic phototrophs. e⁻ represents the electron donor to the diiron center of AcsF. The relevant macrocycle carbons are numbered according to the International Union of Pure and Applied Chemistry.

In most anoxygenic phototrophs, AcsF is active without any extra subunit, whereas in photosynthetic alphaproteobacteria an additional small protein (BciE) is required for cyclase activity (6) (Fig. 1). Conversely, in oxygenic phototrophs, a small (∼15 kDa) protein, Ycf54, is required for the cyclase reaction; in Synechocystis Ycf54 copurifies with both CycI and CycII (9) and the single AcsF homolog in higher plants also interacts with Ycf54 (10, 11). Inactivation of the ycf54 gene strongly impairs cyclase activity (12) and Ycf54-less mutants exhibit severe phenotypes, including lower levels of the AcsF subunit, a buildup of the cyclase substrate MgPME, and lower synthesis of DV PChlide a and Chls (9–11, 13). The strict requirement of Ycf54 for the cyclase activity in vivo has also been demonstrated by heterologous coexpression of Synechocystis, algal, and plant CycI/AscF enzymes with their cognate Ycf54 in E. coli and in Rvi. gelatinosus (14). However, the exact role of Ycf54 remains enigmatic, and a possible catalytic function has not been tested due to the absence of an in vitro cyclase assay. It is noteworthy that, in contrast to all other cyanobacteria, low-light (LL) ecotypes of Prochlorococcus do not contain Ycf54 (15); thus, a Ycf54-independent cyclase evolved naturally in these abundant marine microorganisms.

In the present study, we used adaptive laboratory evolution to generate a Ycf54-independent cyclase in the model cyanobacterium Synechocystis. By placing a Synechocystis ∆ycf54 mutant under selective pressure, we isolated two strains where cyclase activity and Chl biosynthesis were restored. Genome sequencing revealed the changes necessary to compensate for the lack of Ycf54 in these suppressor mutants were a D219G substitution in CycI and inactivation of a putative esterase, Slr1916. We present evidence that Ycf54 is required for both normal accumulation of CycI and full cyclase activity. The Slr1916 protein also affects the CycI level and activity, but the mechanism seems to be indirect through up-regulation of the whole Chl biosynthetic pathway. Synechocystis was also used as a host to test the activity of cyclase enzymes from the marine picocyanobacterium Prochlorococcus in the presence and absence of Ycf54. The role of Ycf54 and the evolution of the O₂-dependent cyclase reaction are discussed.

Results

Identification of Mutations Suppressing the Deletion of the ycf54 Gene.

The ∆ycf54 mutant of Synechocystis has severely impaired Chl biosynthesis (∼13% of wild-type [WT] Chl levels) and is incapable of photoautotrophic growth (12). We reported previously that the purple bacterial cyclase gene from Rvi. gelatinosus, acsF^Rg, complemented the loss of cycI in Synechocystis, irrespective of the presence of Ycf54 (6). The photoautotrophic growth rate of complemented strains was comparable with the WT under 30 µmol photons·m⁻²·s⁻¹ (Fig. 2A) (see SI Appendix, Table S1 for list of strains and plasmids described in this study). The presence of the foreign AcsF^Rg does not affect the distribution of Ycf54 between membrane and soluble fractions, nor is the level of AcsF^Rg protein or its association with membranes affected by the absence of Ycf54 (Fig. 2B). However, complemented strains did suffer from growth retardation when the light intensity increased to 400 µmol photons·m⁻²·s⁻¹ (hereafter referred to as HL for high light) (Fig. 2A), indicating the advantage of the native CycI-Ycf54 couple under less favorable growth conditions.

Fig. 2.

Analysis of Synechocystis mutants complemented with the Rvi. gelatinosus cyclase gene (acsF^Rg). (A) Drop growth assays of the described strains grown on BG11 agar under the indicated light intensity. The cycI⁺ ΔcycI strain, which expresses the native cycI gene under the control of the psbAII promoter, is shown as a control for the expression method. Photographs were taken after incubation for 6 d. (B) Immunodetection of selected Chl biosynthetic enzymes in the indicated strains. Membrane (M) and soluble (S) protein fractions were isolated from an equal number of cells from each strain grown under 30 µmol photons·m⁻²·s⁻¹. CycI and AcsF^Rg were detected by an antibody raised against the Arabidopsis AcsF homolog, which cross-reacts significantly less strongly with AcsF^Rg than with CycI (SI Appendix, Fig. S11). Part of the SDS-PAGE gel was stained with SYPRO Orange as a loading control.

Such observations led us to explore whether photoautotrophy could be restored to the ∆ycf54 mutant through adaptive evolution. We incubated the mutant on BG11 agar without glucose under 15 µmol photons·m⁻²·s⁻¹; after 3 wk, a few tiny colonies arose and were restreaked onto a new plate and incubated under 30 µmol photons·m⁻²·s⁻¹. The restreak procedure was repeated every fortnight, and after 12 wk two photoautotrophic strains were isolated with Chl levels (monitored at ∼680 nm) similar to the WT (see SI Appendix, Fig. S1 for whole-cell spectra), designated suppressor mutant 1 (SM1) and SM2.

Next-generation sequencing was used to analyze the genomes of SM1 and SM2, together with the “parent” ∆ycf54 mutant and the isogenic WT strain (GT-W) (16). Variants were identified by mapping the obtained sequences to a reference strain GT-S (17). Those found in SM1 or SM2 but not in the ∆ycf54 strain were identified as putative suppressor mutations and are listed in SI Appendix, Table S2. Intriguingly, both SM1 and SM2 contain mutations in the cycI gene and in an open reading frame, slr1916. A D219G substitution in CycI is shared by SM1 and SM2, while slr1916, provisionally annotated to encode a 283-aa esterase, is truncated due to frameshifts that leave 129 and 104 aa intact in SM1 and SM2, respectively.

D219G Substitution in cycI or Inactivation of slr1916 Individually Restore Photoautotrophy to the Δycf54 Mutant.

To determine the contribution made by the D219G substituted CycI (hereafter CycI^SM) to the observed suppressor effects, we constructed a Δycf54 cycI^SM+ strain in which the redundant psbAII gene was replaced with the cycI^SM mutant gene (see SI Appendix, Fig. S2 for colony PCR screening of Synechocystis strains). A strain expressing cycI^SM in the WT background and another strain, Δycf54 cycI⁺ expressing an extra copy of the native cycI gene, served as controls. Remarkably, the Δycf54 mutant complemented with the cycI^SM gene was able to grow photoautotrophically even under HL (Fig. 3A), and its Chl level, although still not matching the WT level, increased dramatically when compared with Δycf54 (Fig. 3B). Complementation of the ycf54 mutant phenotypes was not due to increased dosage as the control Δycf54 cycI⁺ strain still has very low levels of Chl and was unable to grow in the absence of glucose (SI Appendix, Fig. S3A). The growth of Δycf54 cycI^SM+ was close to the WT under 100 µmol photons·m⁻²·s⁻¹ (Fig. 3A), and this was used as standard light (SL) intensity for the remainder of the study.

Fig. 3.

Analysis of the Synechocystis ∆ycf54 mutant complemented with single and double suppressor mutations. (A) Drop growth assays of the described strains grown on BG11 agar under different light intensities. Photographs were taken after incubation for 6 d. (B) Whole-cell absorption spectra of the described strains grown autotrophically under SL, except the ∆ycf54 strain, which was grown mixotrophically under LL. (C) CN-PAGE separation of membrane proteins isolated from the described strains. The growth conditions were as described in B. The loading corresponds to the same number of cells from each strain, except the ∆ycf54 strain, for which four times the number of cells were loaded to detect traces of PSII in this strain. Pigmented complexes were detected by their color (Scan) and Chl fluorescence with excitation by blue light (Chl FL). PSI[1] and PSI[3] indicate monomeric and trimeric PSI, respectively; PSII[1] and PSII[2] indicate monomeric and dimeric PSII, respectively. See SI Appendix, Fig. S4 for the second-dimension separation of selected CN-gel strips by SDS-PAGE. (D) Immunodetection of selected Chl biosynthetic enzymes in the indicated strains. Membrane fractions were isolated and analyzed by SDS-PAGE with loading on an equal cell number basis, followed by immunodetection. The WT sample was also loaded at 25%, 50%, and 200% levels for ease of comparison. Part of the SDS-PAGE gel was stained with SYPRO Orange as a loading control.

We also constructed Δycf54 slr1916^SM and Δycf54 Δslr1916 strains in which slr1916 was either truncated, resembling the SM1 mutation, or deleted, respectively. The two strains appeared identical, with Chl contents (SI Appendix, Fig. S3B) notably higher than the Δycf54 strain but lower than the Δycf54 cycI^SM+ strain (Fig. 3B). The improvement achieved by inactivation of slr1916 was thus less prominent than with the cycI^SM mutation and the Δycf54 slr1916^SM strain grew more slowly under SL and HL than the Δycf54 cycI^SM+ strain.

In an attempt to reproduce the phenotypes of the suppressor mutants, we combined the two suppressor mutations to make a Δycf54 cycI^SM+ slr1916^SM strain. These mutant cells grew better than the Δycf54 cycI^SM+ strain under 30 µmol photons·m⁻²·s⁻¹, the light intensity used for generating the suppressor mutants, but less well under SL and HL (Fig. 3A). Whole-cell absorption spectra show that the truncation of slr1916 in the Δycf54 cycI^SM+ strain further increased accumulation of Chl, as well as that of phycobilisomes, to a level significantly higher than in the WT (Fig. 3B). Despite the presence of other putative suppressor mutations in SM1 and SM2 (see SI Appendix, Table S2 for list of identified suppressor mutations), these data suggest that the combination of the D219G substitution in cycI and the truncation of slr1916 principally account for the suppressor effects observed in SM1 and SM2.

Accumulation of Chl-binding proteins in the strains described above was analyzed by clear-native polyacrylamide gel electrophoresis (CN-PAGE). Visibly green bands and detection of Chl fluorescence showed that both photosystem I (PSI) and photosystem II (PSII) levels were partially restored in the Δycf54 cycI^SM+ and Δycf54 Δslr1916 strains (Fig. 3C). In the Δycf54 cycI^SM+ slr1916^SM strain, the PSII level was similar to WT and there was a noticeably higher level of PSI (Fig. 3C), which was further supported by two-dimensional CN/SDS-PAGE analysis of membrane complexes (SI Appendix, Fig. S4). There was no apparent effect of expressing the cycI^SM gene in the WT background (see SI Appendix, Fig. S5 A–D for the comparison between the WT and cycI^SM+ strains).

The accumulation of Chl biosynthetic enzymes is severely hindered in the Δycf54 mutant, which contains only ∼15% of WT CycI levels and ∼50% the level of POR (12). Our immunoblot analysis revealed that expression of the cycI^SM gene increased the levels of CycI and POR (Fig. 3D). On the other hand, inactivation of slr1916 only had marginal effects on the CycI level, but resulted in increased accumulation of POR, with the level in the Δycf54 Δslr1916 and Δycf54 cycI^SM+ slr1916^SM strains being several times higher than in the WT (Fig. 3D).

Synechocystis CycI^SM has Cyclase Activity when Heterologously Expressed in Rvi. gelatinosus.

We have shown that the cycI^SM mutation restored the WT level of CycI in the absence of Ycf54 (Fig. 3D), but it is unclear whether the mutated CycI^SM has the same catalytic activity as the WT enzyme. As an in vitro assay with purified AcsF has not been reported yet, we assayed the heterologous activity of Synechocystis cyclase in a Rvi. gelatinosus mutant that lacks both the O₂-sensitive and O₂-dependent cyclase enzymes (6). We grew the Rvi. gelatinosus strain expressing cycI^SM together with a control strain expressing the native Rvi. gelatinosus acsF gene in liquid culture and monitored the content of BChl a. In agreement with the previous report (6), the coexpression of Synechocystis cycI and ycf54 is strictly required for the synthesis of BChl a (Fig. 4). Intriguingly, there was some residual activity of CycI^SM in Rvi. gelatinosus in the absence of Ycf54, allowing the synthesis of ∼1% of BChl measured for the CycI–Ycf54 pair, which was boosted to ∼50% by the inclusion of Ycf54 (Fig. 4). These results show that CycI^SM can work as a stand-alone cyclase, but still relies on Ycf54 for optimal activity when heterologously expressed in Rvi. gelatinosus.

Fig. 4.

Heterologous activity of Synechocystis cyclase in Rvi. gelatinosus. Plasmid-borne genes encoding Synechocystis cyclase components were tested in the Rvi. gelatinosus ΔbchE ΔacsF Rif^R mutant. Pigment extracts from Rvi. gelatinosus strains were analyzed by HPLC. (A) No plasmid negative control. (B) pBB[cycI]. (C) pBB[cycI^SM]. (D) pBB[cycI-ycf54]. (E) pBB[cycI^SM-ycf54]. (F) pBB[acsF^Rg]. For A–C, pigments were extracted from 10 times as many cells as in D–F. Immunodetection of cyclase proteins in whole-cell lysates prepared from the same number of cells of each of the Rvi. gelatinosus strains using specific antibodies is also shown (Inset).

Ycf54 and Slr1916 Affect the Cyclase Level during Nitrogen Deficiency.

To investigate the role of Ycf54 and the effects of the suppressor mutations, we monitored Chl biosynthesis in the WT and complemented strains grown in a nitrogen-fluctuating regime. Nitrogen deficiency is known to diminish the whole tetrapyrrole pathway, and the metabolic flow can be restored quickly (<2 h) upon nitrogen repletion (18). Such regulation requires tightly synchronized levels/activities of all enzymes involved in tetrapyrrole metabolism and any defect in the accumulation/activity of CycI should be much more pronounced than under conditions with sufficient levels of nutrients.

We found that CycI was unstable in the WT during nitrogen deficiency and decreased to ∼25% of the predepletion level after 6-h nitrogen deprivation (Fig. 5A), becoming virtually undetectable after 18 h (Fig. 5B). Conversely, ChlM, POR, and Ycf54 were much more stable during nitrogen depletion (Fig. 5B). CycI was still barely detectable after 2-h nitrogen repletion but was restored to the predepletion level after 6 h (Fig. 5B). A similar pattern was observed in the cycI^SM+ strain, confirming that expression of the cycI^SM gene from the psbAII promoter in the WT background does not alter the level or regulation of the protein (SI Appendix, Fig. S5E). We repeated the same experiments with the Δycf54 cycI^SM+ and Δycf54 cycI^SM+ slr1916^SM strains. Remarkably, CycI was still present in the Δycf54 cycI^SM+ slr1916^SM strain even after 18-h nitrogen depletion (Fig. 5B). In addition, both complemented strains exhibited a faster recovery of CycI levels upon nitrogen restoration, with a significant CycI signal detectable after only 2 h (Fig. 5B). These results indicate misregulation of the cellular level of CycI in the absence of Ycf54, and that the mutated CycI is stabilized, particularly in combination with the slr1916^SM mutation.

Fig. 5.

Immunodetection of selected Synechocystis Chl biosynthetic enzymes in the indicated strains during nitrogen depletion and restoration. (A) Time course analysis of the level of CycI in the WT during nitrogen depletion. Cells were collected before (+N) and after 1, 3, and 6 h of nitrogen starvation. Membrane (M) and soluble (S) protein fractions were isolated from the collected cells and loaded on an equal cell number basis for SDS-PAGE, followed by immunodetection. The +N sample was also loaded at 25% and 50% levels for comparison. (B) Immunodetection of indicated Chl biosynthetic enzymes in the described strains upon nitrogen depletion and subsequent restoration with 10 mM NaNO₃. Cells were collected before (+N) and after 18-h nitrogen starvation, and after 2- and 6-h nitrogen restoration. SDS-PAGE analysis and immunodetection were conducted as in A.

We also measured the Chl precursor pool to analyze the overall consequence of disruption of the ycf54 and slr1916 genes on Chl biosynthesis. Before nitrogen deprivation, the Δycf54 cycI^SM+ strain contained approximately four times the amount of MgPME and only half the amount of DV PChlide a as the WT, indicating a deficiency in cyclase activity (Fig. 6). This pigment profile is shared by the Δycf54 Δslr1916 strain but with an even higher MgPME level, ∼70 times greater than in the WT (Fig. 6), in line with its low CycI level (Fig. 3D). The overaccumulation of MgPME is sustained in the Δycf54 cycI^SM+ slr1916^SM strain despite its near WT level of DV PChlide a (Fig. 6). After 18-h nitrogen deprivation, the entire Chl biosynthetic pathway was shut down with only traces of precursors detected in the WT, except for MV Chlide a (Fig. 6), which mostly originates from the dephytylation of Chl in the Chl recycling process (19, 20).

Fig. 6.

Analysis of Chl precursors in Synechocystis strains recovering from nitrogen depletion. Strains were grown autotrophically under SL and subjected to nitrogen starvation for 18 h, followed by nitrogen repletion by addition of 10 mM NaNO₃. Pigments were extracted from cells harvested before (+N 0 h) and after (−N 18 h) nitrogen starvation, and after 2, 4, 12, and 24 h of nitrogen repletion. Pigments were analyzed by HPLC to allow detection of protoporphyrin IX (PPIX), Mg-PPIX (MgP), MgP monomethyl ester (MgPME), 3,8-divinyl protochlorophyllide a (DV PChlide), and 3-vinyl chlorophyllide a (MV Chlide). The level of precursors is shown as ratio to the WT level before nitrogen starvation, and the error bars indicate the SD from the mean of biological triplicates.

Following nitrogen repletion, the WT gradually built up the precursor pool without anomalous accumulation of intermediates, and predepletion precursor levels were restored within 24 h (Fig. 6). In sharp contrast, MgP and MgPME were rapidly restored in the Δycf54 cycI^SM+ strain after only 2-h nitrogen repletion and their levels continued to increase up to 12 h (Fig. 6). The fast recovery of CycI in this strain (Fig. 5B) did not result in an abrupt recovery of DV PChlide a, which instead built up more gradually (Fig. 6). It seems that the “regreening” process was stalled at the cyclase step, as further evidenced by the drastic increase of MgPME upon nitrogen repletion in Δycf54 Δslr1916, a strain clearly deficient in cyclase activity. A fast recovery of MgP and MgPME was also observed in the Δycf54 cycI^SM+ slr1916^SM strain but in a less dramatic manner, with the finishing precursor levels even greatly surpassing the predepletion ones (Fig. 6). Surprisingly, this strain showed levels and recovery of DV PChlide a similar to the WT (Fig. 6), indicating the large pool of MgP and MgPME in this strain was caused by up-regulated metabolic flow in Chl biosynthesis, rather than very low cyclase activity.

It is noticeable that the depletion of MgPME after 18-h nitrogen starvation was less severe in the mutant strains, with Δycf54 cycI^SM+ still containing ∼30% of the WT predepletion level of MgPME and Δycf54 cycI^SM+ slr1916^SM ∼47%, indicating that MgPME may stabilize CycI, and/or its mutated form, CycI^SM. To check this possibility, we added gabaculine to inhibit the tetrapyrrole pathway (21) and monitored the stability of CycI. The gabaculine-treated cells continued to proliferate for 12 h, “diluting” Chl-containing complexes and phycobilisomes (SI Appendix, Fig. S6). However, even after 24-h gabaculine treatment, neither the WT nor the complemented strains lost CycI (Fig. 7), implying that the level of CycI/CycI^SM does not simply match the availability of MgPME.

Fig. 7.

Immunodetection of Chl biosynthetic enzymes in Synechocystis strains before and after treatment with gabaculine. Strains were grown autotrophically under SL. Cells were collected before (0 h) and after 6-, 12-, and 24-h treatment with 5 µM gabaculine. SDS-PAGE analysis and immunodetection were conducted as in Fig. 3D. The 0-h sample was also loaded at 25% and 50% levels for ease of comparison.

Slr1916 Has Chl Dephytylase Activity.

Unlike the obvious link between the D219G substitution in CycI and the complementation of Δycf54, it was unclear why inactivation of slr1916, which encodes a putative esterase of unknown function, increased Chl content and restored photoautotrophic growth in the Δycf54 background (Fig. 3 A and B). We generated a Δslr1916 mutant in the WT background and found a higher Chl content (Fig. 8A) with extra Chl molecules allocated mostly to trimeric PSI complexes (Fig. 8B and SI Appendix, Fig. S7). The increased levels of POR (Fig. 3D) and PSI (Fig. 8B and SI Appendix, Fig. S7) upon slr1916 deletion, as well as the aberrant acceleration of Chl biosynthesis during the regreening process (Fig. 6), collectively indicate that Slr1916 negatively regulates Chl biosynthesis.

Fig. 8.

Analysis of the ∆slr1916 mutant and the Chl dephytylase activity of FLAG-tagged Slr1916 purified from Synechocystis. (A) Whole-cell absorption spectra and (B) CN-PAGE separation of membrane proteins isolated from the WT and Δslr1916 strains grown autotrophically under SL. For CN-PAGE analysis, the loading corresponds to the same number of cells from each strain. Pigmented complexes were detected and annotated as in Fig. 3C. See also SI Appendix, Fig. S7 for the second-dimension separation of CN-gel strips. (C) SDS-PAGE analysis of 15 µL of purified Slr1916-FLAG from detergent-solubilized Synechocystis membranes with protein staining with Coomassie Brilliant Blue staining. (D) HPLC-based in vitro Chl dephytylase assays with Slr1916-FLAG. A positive control using clarified E. coli lysate containing Arabidopsis CLH1 (SI Appendix, Fig. S8B) and negative controls with the FLAG-immunoprecipitation elution from WT Synechocystis or purified E. coli MenH (SI Appendix, Fig. S8C) were also performed. Retention times and absorption spectra of peaks were used to identify MV Chlide a and Chl a.

Slr1916 belongs to the functionally diverse alpha/beta hydrolase superfamily that contains proteases, lipases, peroxidases, esterases, epoxide hydrolases, and dehydrogenases (22, 23). In some databases, Slr1916 is annotated as MenH, an enzyme required for synthesis of vitamin K; however, the phenotype of the Δslr1916 strain does not suggest a role of Slr1916 in phylloquinone biosynthesis (see Discussion for more details). Chl dephytylating enzymes are not known in cyanobacteria, but like Slr1916, plant Chl dephytylases are alpha/beta hydrolases (24–26). Based on this classification, we hypothesized that the Slr1916 may be a Chl dephytylase that could act on free Chl molecules that accumulate in the membrane (see Discussion). Slr1916 was insoluble when overproduced in E. coli so we produced the enzyme with a 3×FLAG tag in Synechocystis (expressed from the psbAII locus as described previously) (9, 27). Although not predicted to be a membrane protein, both N- and C-terminally tagged Slr1916 were more abundant in the membrane fraction than the soluble lysate (SI Appendix, Fig. S8A). FLAG-tagged Slr1916, isolated from the solubilized membranes, was pure with no obvious partner proteins (Fig. 8C and SI Appendix, Fig. S8A). For the dephytylase assay, the Slr1916 protein was incubated with Chl a and stopped assays were analyzed by high-performance liquid chromatography (HPLC) to look for the formation of MV Chlide a. Assays containing FLAG-tagged Slr1916 eluates or Arabidopsis thaliana (hereafter Arabidopsis) chlorophyllase (CLH1, positive control) (SI Appendix, Fig. S8B) gave an MV Chlide a peak at 10.8 min, whereas FLAG pulldown eluates from WT Synechocystis membranes or purified E. coli MenH (negative controls; see Discussion for details) did not degrade Chl a (Fig. 8D).

LL-Adapted Prochlorococcus Ecotypes Lack the ycf54 Gene.

We have shown that the Synechocystis Δycf54 mutant was rescued by a spontaneous single point mutation in the cycI gene. We hypothesized that this type of event could happen naturally during evolution if there was less stringent need for a Ycf54 protein to modulate the levels and activity of the O₂-dependent cyclase, particularly in combination with severe constraints on the number of genes that a cell could maintain. Indeed, it has been reported that six LL-adapted Prochlorococcus ecotypes, known to have streamlined genomes, do not contain the ycf54 gene (15). We did a thorough BLAST search against all cyanobacteria with sequenced genomes (1,048 quality-checked genomes available in the Genome Taxonomy Database [GTDB]) (28) and found that all 89 Prochlorococcus genome assemblies clustering in the LL-adapted clades lack the ycf54 gene. LL-adapted Prochlorococcus ecotypes also lack a bchE ortholog and so appear to rely solely on an O₂-dependent cyclase for Chl biosynthesis.

To gain more detailed insight into the evolution of AcsF and Ycf54 in Prochlorococcus ecotypes, we constructed a phylogenetic tree inferred from AcsF proteins (394 aligned positions), which was compared to a species tree based on concatenated sequences (3,182 aligned positions) of 13 universally conserved proteins (29) (Fig. 9). Representatives of all phototrophic phyla and LL- and HL-adapted Prochlocococcus ecotypes were included. All Prochlorococcus ecotypes form a monophyletic lineage within the clade of other picocyanobacteria (marine Synechococcus and Cyanobium) (Fig. 9). The LL-adapted ecotypes are ancestral in the Prochlorococcus lineage and form three paraphyletic branches, whereas the HL-adapted ones form a single compact branch (Fig. 9). The HL-adapted ecotypes contain a typical AcsFI, in keeping with other picocyanobacteria. In contrast, the LL-adapted strains possess only an AcsFII that is phylogenetically distant from other AcsFII proteins (Fig. 9) (see SI Appendix, Fig. S9 for sequence alignments). Most cyanobacteria contain both AcsFI and AcsFII (SI Appendix, Fig. S10); however the latter protein is expressed only under microoxic conditions (30). Apart from Prochlorococcus species, most marine Synechococcus also contain only one AcsFI homolog (SI Appendix, Fig. S10).

Fig. 9.

Phylogenetic analysis of AcsF proteins and their parent organisms. Phylogenetic tree (Left) inferred from AcsF proteins (394 aligned positions; Synechocystis CycI is highlighted) is compared to a species tree (Right) based on concatenated sequences of 13 conserved proteins (3,182 aligned positions). Trees containing 103 sequences were calculated using Bayesian inference employing the LG+G+I substitution model; posterior probabilities are displayed near the nodes, and the trees are rooted with Chloracidobacterium thermophilum B. AcsF is accompanied by BciE in Alphaproteobacteria, while Ycf54 is present in oxygenic phototrophs (cyanobacteria, algae, and plants). Among picocyanobacteria, the clade containing HL-adapted Prochlorococcus ecotypes has the canonical AcsF plus Ycf54 arrangement. Conversely, the LL-adapted lineages of Prochlorococcus lack Ycf54 and contain a distinct AcsFII. Note that the AcsFII sequences were retrieved by a BLAST search using the Synechocystis CycII sequence as the query, while other AcsF sequences including the AcsFI sequences were retrieved using the Synechocystis CycI sequence.

To test whether the acsF gene from Prochlorococcus can function in Synechocystis, we expressed the genes from a representative LL-adapted strain, Prochlorococcus marinus MIT 9313, and a representative HL-adapted strain, Prochlorococcus marinus MED4, in the WT background and subsequently attempted to delete the native cycl. We were not able to fully segregate the acsF⁹³¹³⁺ΔcycI strain, suggesting that AcsF⁹³¹³ cannot functionally replace CycI in Synechocystis under our standard laboratory conditions. This result contrasted with the successful complementation of Synechocystis ΔcycI strain by the acsF^MED4 gene; although the resulting acsF^MED4+ΔcycI strain still contained low levels of Chl (Fig. 10A), it was able to proliferate autotrophically under SL. On the other hand, the Δycf54 acsF^MED4+ strain showed no improvement in Chl content (Fig. 10B) or autotrophic growth, demonstrating the dependence of AcsF^MED4 on Ycf54.

Fig. 10.

Heterologous activity of Prochlorococcus cyclase enzymes in Synechocystis. Whole-cell absorption spectra of the described strains grown autotrophically under SL (A), mixotrophically under LL (B), and autotrophically under 30 µmol photons·m⁻²·s⁻¹ in a gas mixture of 2% O₂ and 0.5% CO₂ in N₂ (C). The Inset in C shows an expanded view of the absorption of Chl at 682 nm and Chl contents of the two strains (*P < 0.02, n = 4, Student’s t test). (D) Visual comparison of pigmentation of the Δycf54 cycII⁺ and Δycf54 acsF⁹³¹³⁺ strains grown under the same conditions as in C.

The inactivity of AcsF⁹³¹³ in Synechocystis is further demonstrated by the phenotype of the Δycf54 acsF⁹³¹³⁺ strain, which was unable to grow photoautotrophically and had a Chl level similar to the Δycf54 strain (Fig. 10B). The AcsF⁹³¹³ is, however, of the AcsFII type, and the Synechocystis CycII contributes to Chl biosynthesis only under microoxic conditions, likely due to the oxygen-sensitive nature of AcsFII/CycII-type enzymes (7). We therefore tested these strains under low oxygen conditions using a gas mixture containing 2% O₂ and 0.5% CO₂ in N₂ with a light intensity of 30 µmol photons·m⁻²·s⁻¹. The Δycf54 acsF⁹³¹³⁺ strain exhibited slow autotrophic growth with a doubling time of ∼95 h, whereas the control Δycf54 cycII⁺ strain showed only a negligible increase in turbidity after 5 d. The Δycf54 acsF⁹³¹³⁺ strain accumulated significantly more Chl and carotenoids than the Δycf54 cycII⁺ strain (Fig. 10C). The observed lack of CycII activity in the Δycf54 cycII⁺ strain grown under low oxygen conditions is consistent with the expected dependence of CycII on Ycf54 (9). In summary, our data support a model that the LL-adapted Prochlorococcus ecotypes have evolved a distinct Ycf54-independent AcsFII that does not require regulation by Ycf54, a mechanism that is otherwise conserved in cyanobacteria, algae, and plants.

Discussion

In order to investigate the role of Ycf54, we conducted laboratory evolution experiments with the ∆ycf54 mutant and identified suppressor mutations that restore photoautotrophic growth. Our results clearly demonstrate that a D219G substitution significantly weakens the dependence of CycI on Ycf54 to allow CycI to accumulate without Ycf54, but that Ycf54 is required for optimal cyclase activity. It has been demonstrated that AcsF and Ycf54 form a stable, membrane-bound complex in various model phototrophs (9–11, 31, 32); the docking of Ycf54 onto CycI/CycII requires a region of positive surface potential on Ycf54 (31). The following enzyme in the pathway, POR, is likely to be a component of the same complex (12, 33), perhaps along with several other Chl biosynthetic enzymes (34); consistently, the absence of Ycf54 in Synechocystis destabilizes CycI and POR (9, 12).

Our results support that Ycf54 is required for the stability/accumulation of CycI as well as for optimal cyclase activity in vivo. A recent report shows recombinant barley AcsF does not accumulate in E. coli unless coexpressed with Ycf54 (35), indicating a possible role of Ycf54 in the folding and/or maturation of plant AcsF. In addition, Synechocystis CycI requires Ycf54 for heterologous cyclase activity in E. coli (2) and Rvi. gelatinosus (6). On the other hand, Bollivar et al. (36) showed that recombinant Ycf54 stimulates in vitro cyclase activity with barley extracts. A direct role of Ycf54 in the cyclase reaction is further supported by PChlide (Fig. 6) and Chl (Fig. 3B) deficiency in the Δycf54 cycI^SM+ strain, despite restoration of WT-like CycI levels by the D219G substitution (Fig. 3D).

However, why is Ycf54 present in almost all oxygenic phototrophs despite the apparent relative ease for the gene encoding AcsF to mutate to form a Ycf54-independent enzyme? An analogy between the O₂-dependent cyclase and the first committed enzyme in (B)Chl biosynthesis, MgCH, can be drawn here. Although structurally and mechanistically conserved in all phototrophs, MgCH in Chl-producing organisms requires an auxiliary protein, Gun4, which is not found in anoxygenic phototrophs (37, 38), and like the Δycf54 mutant, Synechocystis Δgun4 mutants have severely lowered levels of Chl (39). Gun4 directly interacts with ChlH, the catalytic subunit of MgCH, and has been shown to control the accumulation of ChlH during the first few hours of recovery from nitrogen depletion (18), enhance enzyme activity in vitro (38, 40, 41), and control the metabolic flux within the tetrapyrrole biosynthesis pathway in cyanobacteria, green algae, and various plant species (37–39, 42, 43).

Given the central importance of Chl for the function of photosynthetic complexes, and the photolability and phototoxicity of its biosynthetic intermediates, multiple layers of regulation are required to adjust production of Chl in response to fluctuating levels of nutrients and light. Apart from MgCH directing PPIX into Chl biosynthesis, the cyclase step is also expected to be tightly regulated. The following enzyme in the pathway, POR, is light activated, and thus its activity is difficult to modulate under fluctuating light. It may therefore be important to control the availability of the POR substrate, i.e., the cyclase product, either by direct intervention in catalysis or controlling the stability of AcsF/CycI. We showed that CycI stability is not impaired by the lack of substrate (Fig. 7), suggesting that the CycI level is controlled by a more sophisticated mechanism that presumably involves Ycf54. Our analysis of the Δycf54 cycI^SM+ strain showed fast recovery of CycI shortly after nitrogen was restored following depletion, which contrasts with the WT-like slower restoration of CycI in the presence of Ycf54 (Fig. 5B and SI Appendix, Fig. S5E). We speculate that Ycf54 and Gun4 evolved in cyanobacteria to stabilize and/or regulate the catalytic subunits of Chl biosynthetic enzymes, and later started to modulate the activity of these enzymes providing an additional (strict) measure of control to avoid aberrant accumulation of phototoxic Chl precursors. As photosynthetic bacteria perform anoxygenic photosynthesis and production of BChl is largely controlled by environmental oxygen tension (44), they may not need regulators equivalent to Ycf54 or Gun4.

Intriguingly, unlike other cyanobacteria, the LL-adapted Prochlorococcus ecotypes lack the ycf54 and acsFI genes. Given the dependence of CycII on Ycf54 (9) (Fig. 10 C and D), we propose that the acsFI gene was initially lost in a subpopulation of ancient LL-ecotype Prochlorococcus. The loss of acsFI can be rationalized, given that the habitat of LL-adapted strains has a low oxygen saturation level as well as low light intensities (45), which can only support a low rate of oxygen evolution. The remaining AcsFII may then have mutated to become less dependent on Ycf54, allowing a subsequent loss of the ycf54 gene due to genome streamlining, a well-documented phenomenon in Prochlorococcus (46); the likelihood of this mutation event is high, given that we were able to generate a Ycf54-independent mutant of AcsF by laboratory microevolution. The scenario of the HL-adapted ecotypes is completely different as it is unlikely that they could rely on AcsFII for Chl biosynthesis due to the higher levels of oxygen in the upper layers of the oceans. Thus, the HL-adapted ecotypes most likely evolved from a population still possessing acsFI/ycf54 genes by a subsequent loss of acsFII (Fig. 9 and SI Appendix, Fig. S10). Nonetheless, it is notable that, despite extreme genome reduction during the evolution of Prochlorococcus species (47), HL ecotypes retain Ycf54. Thus, under high light intensities, the role of Ycf54 appears essential (Fig. 3A).

In contrast to the cycI^SM mutation, inactivation of slr1916 seems to stimulate the cyclase activity indirectly, rather than by restoring the CycI level (Fig. 3D). After inactivation of the Slr1916 in the Δycf54 cycI^SM+ strain, the synthesis of PChlide doubled (Fig. 6) and the cellular Chl level increased significantly (Fig. 3B). However, the resulting strain is more photosensitive (Fig. 3A) and the regulation of the CycI level is disrupted (Fig. 5B). The KEGG database annotates Slr1916 as the MenH enzyme required for phylloquinone biosynthesis, which is supported by BLAST searches revealing that Slr1916 is the sole homolog of the E. coli enzyme in Synechocystis (93% coverage, 42% similarity, 29% identity, E value 3e-10). However, it is worth noting that MenH sequences are highly variable, and only 15 residues were strictly conserved across 47 homologs analyzed by Jiang et al. (48). In plants, mutation of the menH locus causes a pale green phenotype due to phylloquinone deficiency, which results in reduced Chl content and stability of PSI (49, 50), and in other Synechocystis men mutants the absence of phylloquinone results in a lowered level of PSI (51–53). This contrasts with the significantly increased PSI level in Δslr1916 mutants (54) (Fig. 8B and SI Appendix, Fig. S7), indicating Slr1916 is not a MenH enzyme. We noticed that Slr1916 contains a GHSLG motif, similar to the PPH motif (GNS[L/I/V]G) identified in plant pheophytinases and Chl dephytylases (25, 26) and the lipase motif (GHSRG) in chlorophyllase (24). Consistently, we found that the purified Slr1916 has Chl dephytylase activity in vitro (Fig. 8D).

We do not expect Slr1916 to be a major Chl dephytylase in Synechocystis as during nitrogen starvation the Δycf54 cycI^SM+ slr1916^SM strain contained high level of MV Chlide a that can only originate from Chl dephytylation (Fig. 6). Screening of an inducible CRISPRi gene repression library in Synechocystis identified slr1916 as one of the few genes that, when down-regulated, leads to significantly increased growth rates in a turbidostat (55). Slr1916 is proposed to play a global regulatory role, and its activity somehow limits the cellular level of PSI, the main sink for the Chl molecules in Synechocystis (56). We hypothesize that Slr1916 catabolizes unbound Chl accumulating in the membrane if Chl biosynthesis exceeds production of Chl-binding apoproteins (57). The activity of Slr1916 may therefore provide a feedback mechanism to synchronize the biosynthesis of Chl and Chl-binding proteins. As CycI and POR probably form an enzymatic complex (33), the greatly elevated level of POR in slr1916 mutants (Fig. 3D) might account for the increased activity of cyclase in the absence of Ycf54 and the higher stability of CycI^(SM) during nitrogen depletion. However, addressing the exact function of Slr1916 requires further study.

Conclusion

We have applied adaptive laboratory evolution to the Chl-deficient Δycf54 mutant of Synechocystis, in combination with genomic sequencing, molecular genetics, phenotypic analyses, biochemical assays, and bioinformatic approaches. We have 1) shown that a point mutation allows the CycI cyclase to accumulate in the absence of Ycf54; 2) presented evidence that Ycf54 regulates Chl biosynthesis by controlling the activity and level of CycI in response to fluctuating environmental factors; 3) demonstrated that the putative esterase Slr1916 has Chl dephytylase activity in vitro; and 4) investigated the evolution of O₂-dependent cyclase in Prochlorococcus, the most abundant photosynthetic organism on Earth.

Materials and Methods

Bacterial Strains and Growth Conditions.

Bacterial strains described in this study are listed in SI Appendix, Table S1. Synechocystis strains were grown at 28 °C under constant illumination in BG11 medium buffered with 10 mM Tes, pH 8.2 (adjusted with KOH). Unless otherwise specified, light conditions were as follows: 10, 100, and 400 µmol photons·m⁻²·s⁻¹, referred as LL, SL, and HL, respectively. Photoautotrophic liquid cultures were grown in air-bubbled 100-mL cylinders in a water-tempered growth chamber under SL. Synechocystis strains that are not photoautotrophic were grown in Erlenmeyer flasks on a rotary shaker in BG11 medium with 5 mM glucose under LL. For low-oxygen cultivation, cells were grown in Erlenmeyer flasks on a rotary shaker in a laboratory incubator (Memmert) with adjustable CO₂ and O₂ levels and equipped with a light source. For FLAG-immunoprecipitation experiments, photoautotrophic liquid cultures were grown in 8-L vessels under SL bubbled with sterile air and mixed using a magnetic stirrer. For plate-based drop growth assays, Synechocystis cell cultures were adjusted to OD_750nm of 0.4 and diluted to 0.04 and 0.004. All three concentrations were spotted (5 µL) on solid medium, left to dry, and incubated under conditions as indicated in the text. E. coli strains were grown at 37 °C in LB medium, and if required antibiotics were added at 30, 34, and 100 μg·mL⁻¹ for kanamycin, chloramphenicol, and ampicillin, respectively. Rvi. gelatinosus strains were grown at 30 °C in PYS medium (58), and where required, kanamycin and rifampicin were added at 50 and 40 μg·mL⁻¹, respectively.

Construction of Plasmids and Bacterial Strains.

Plasmids described in this study are listed in SI Appendix, Table S1. Sequences of synthesized genes and primers described in this study are shown in SI Appendix, Tables S3 and S4, respectively. The procedures for constructing plasmids and bacterial strains are described in SI Appendix.

Genome Sequencing and Variant Calling.

High-integrity Synechocystis genomic DNA was isolated, fragmented by nebulization with N₂ gas, and used for construction of a DNA library for paired-end sequencing using the Nextera DNA Library Preparation Kit (Illumina) with a median insert size of ∼300 bp. The constructed library was subjected to 100-bp paired-end sequencing on an Illumina HiSeq 2000 platform according to the manufacturer’s instructions. Variants were called using the mapping-based method, and those found in the suppressor mutants but not in the Δycf54 strain were identified as putative suppressor mutations and listed in SI Appendix, Table S2. Details of genomic DNA isolation, genome sequencing, and variant calling are described in SI Appendix.

Protein Electrophoresis and Immunodetection.

For native electrophoresis, solubilized membrane proteins were separated on 4 to 12% gels (59). Individual components of protein complexes were resolved by incubating the gel strip from the first dimension in 2% (wt/vol) SDS and 1% (wt/vol) dithiothreitol for 30 min at room temperature, and proteins were separated in the second dimension by SDS-PAGE in a denaturing 12 to 20% (wt/vol) polyacrylamide gel containing 7 M urea (60). The procedures for standard single-dimension SDS-PAGE, immunodetection, and assessment of antibody reactivity are detailed in SI Appendix.

Whole-Cell Absorption Spectroscopy.

Synechocystis whole-cell spectra were measured using a Shimadzu UV-3000 spectrophotometer and normalized to light scattering at 750 nm.

Pigment Extraction and Analysis by HPLC.

Pigments were extracted from Rvi. gelatinosus cells with an excess of 0.2% (wt/vol) ammonia in methanol by vigorous shaking using a Mini-Beadbeater (BioSpec). Clarified pigment extracts were vacuum dried, reconstituted in 0.2% (wt/vol) ammonia in methanol, and analyzed by HPLC as previously described (61). Synechocystis Chl intermediates were analyzed by a previously described method (62).

Chl Dephytylase Assays.

Anti-FLAG immunoprecipitaion experiments were performed with the Synechocystis WT, FLAG-slr1916⁺, and slr1916-FLAG⁺ strains as described previously (9, 27). Recombinant Arabidopsis CLH1 protein was produced in E. coli BL21(DE3) as described previously (63), and clarified cell lysates were used as a positive control for the assay. E. coli MenH was overproduced with a His₆-tag in E. coli BL21(DE3) and purified by Ni-affinity and size exclusion chromatography to determine whether it has nonspecific Chl dephytylase activity. Chl dephytylase assays were performed by adding 5 µL of 500 µM Chl a in acetone to 45 µL of sample (FLAG-elution, CLH1 lysate, or purified MenH) so at a final concentration of 50 µM pigment and 10% (vol/vol) acetone. The assay mixture was incubated at 35 °C for 30 min in darkness before stopping by adding 200 µL of acetone, followed by vortexing and centrifugation. One hundred microliters of the resulting supernatant was diluted four times in methanol, and 100 µL was loaded onto a Discovery HS C18 column (5 μm; 250 × 4.6 mm) and analyzed on an Agilent 1200 HPLC system as described previously (27).

Phylogenetic Analyses.

A representative set of 1,048 publicly available cyanobacterial genome assemblies, quality-checked by the GTDB toolkit, version 1.0.2 (28, 64), was downloaded from the National Center for Biotechnology Information and utilized to create a custom BLAST database (65). Ycf54 from Synechocystis was used as a query for a tBLASTn search with a cutoff E value of 1e-10 against this database. All hits were automatically harvested and aligned using MAFFT, version 7 (66), to check their overall homology. The presence/absence of Ycf54 homologs was mapped to the current GTDB phylogenomic species tree of Cyanobacteria (based on 120 conserved proteins) to investigate their phylogenetic distribution among the LL- and HL-adapted Prochlorococcus clades. To further compare the evolutionary scenarios between the AcsF protein and its parental organisms, we employed two phylogenetic analyses using an identical representative set of 103 organisms ranging from Acidobacteria, photosynthetic Proteobacteria, Chloroflexi, and Cyanobacteria to plant and algal plastids. The first tree was based on alignments of AcsF proteins, while the second tree was inferred from 13 universally conserved proteins selected from those used previously for studying plastid evolution (29). See details in SI Appendix.

Data Availability

All study data are included in the article and/or SI Appendix.